Originally published as Kalthoff N Adler B Wieser A Kohler M Traumlumner K Handwerker J Corsmeier U Khodayar S Lambert D Kopmann A Kunka N Dick G Ramatschi M Wickert J Kottmeier C (2013) KITcube ndash a mobile observation platform for convection studies deployed during HyMeX ‐ Meteorologische Zeitschrift 22 6 633‐647 DOI 1011270941‐294820130542

eschweizerbart_xxx

KITcube ndash a mobile observation platform for convectionstudies deployed during HyMeX

Norbert Kalthoff1

Bianca Adler1 Andreas Wieser

1 Martin Kohler

1 Katja Traumner

1

Jan Handwerker1 Ulrich Corsmeier

1 Samiro Khodayar

1 Dominique Lambert

2

Andreas Kopmann3 Norbert Kunka

3 Galina Dick

4 Markus Ramatschi

4 Jens Wickert

4

and Christoph Kottmeier1

1Institut fur Meteorologie und Klimaforschung (IMK-TRO) Karlsruher Institut fur Technologie (KIT) KarlsruheGermany

2Laboratoire drsquoAerologie Universite de Toulouse Toulouse France3Institut fur Prozessverarbeitung und Elektronik (IPE) Karlsruher Institut fur Technologie (KIT) KarlsruheGermany

4Department of Geodesy and Remote Sensing German Research Centre for Geosciences (GFZ) PotsdamGermany

(Manuscript received August 9 2013 in revised form October 2 2013 accepted October 3 2013)

AbstractWith the increase of spatial resolution of weather forecast models to order O(1 km) the need for adequateobservations for model validation becomes evident Therefore we designed and constructed the lsquolsquoKITcubersquorsquoa mobile observation platform for convection studies of processes on the meso-c scale The KITcube consistsof in-situ and remote sensing systems which allow measuring the energy balance components of the Earthrsquossurface at different sites the mean atmospheric conditions by radiosondes GPS station and a microwaveradiometer the turbulent characteristics by a sodar and wind lidars and cloud and precipitation properties byuse of a cloud radar a micro rain radar disdrometers rain gauges and an X-band rain radar The KITcubewas deployed fully for the first time on the French island of Corsica during the HyMeX (Hydrological cyclein the Mediterranean eXperiment) field campaign in 2012 In this article the components of KITcube and itsimplementation on the island are described Moreover results from one of the HyMeX intensive observationperiods are presented to show the capabilities of KITcube

Keywords HyMeX convection observation platform

1 Introduction

During the last years spatial resolution of weather fore-cast models increased to order O (1 km) allowing toresolve larger convective systems like thunderstormsConvection however encompasses multiple scales frommicro- to mesoscale Therefore the transition from con-vection initiation via shallow to deep convection is notsimulated properly (BRYAN et al 2003) Besides under-standing of processes and mechanisms governing theconvection development has been limited up to now

To improve the knowledge about convection scale-adapted measurements were performed (WECKWERTH

et al 2004 WILSON and ROBERTS 2006 BROWNING

et al 2007 KOTTMEIER et al 2008 WULFMEYER

et al 2011) So-called supersites were installed wherecoordinated measurements of different in-situ and remotesystems were exploited synergetically so that convectiveprocesses could be analysed (eg MILLER and SLINGO2007 BEHRENDT et al 2011 CORSMEIER et al 2011

KALTHOFF et al 2013) Often smaller-scale meteorolog-ical networks were embedded in larger-scale investiga-tion domains to capture spatial inhomogeneitiesAircraft missions including dropsonde releases wereperformed to flexibly respond to convection evolution

From these and previous observations it was foundthat convection in many cases was initiated by low-levelconvergence zones (KHODAYAR et al 2010 2013)and favoured by small-scale moisture variability(WECKWERTH 2000) Convergence zones are often gen-erated by mesoscale surface heterogeneities eg givensoil moisture (TAYLOR et al 2011) and land-sea contrastor orography (KALTHOFF et al 2009) Besides trigger-ing convergence zones also support continued growthof convection because of the permanent upward motionand moisture transport there For mountainous terrainadditional types of thunderstorm initiation mechanismswere distinguished (eg BANTA 1990) direct orographiclifting and obstacle or aerodynamic effects (blockingflow deflection gravity wave effects) The influence ofislands on the triggering of thunderstorms was also doc-umented well (eg WILSON et al 2001 QIAN 2007)Mountainous islands turned out to be most effectivebecause sea breezes and valley winds that are roughly

Corresponding author Norbert Kalthoff Institut fur Meteorologie undKlimaforschung (IMK-TRO) Karlsruher Institut fur Technologie (KIT)POB 3640 76021 Karlsruhe Germany e-mail norbertkalthoffkitedu

Meteorologische Zeitschrift Vol 22 No 6 633ndash647 (February 2014) Open Access Article by Gebruder Borntraeger 2013

DOI 1011270941-2948201305420941-294820130542 $ 675

Gebruder Borntraeger Stuttgart 2013

eschweizerbart_xxx

in phase with each other may combine to strengthen thediurnal cycle of winds and form convergence zones dur-ing daytime (KOTTMEIER et al 2000)

The lsquolsquoKITcubersquorsquo was developed for state-of-the-artobservations the objective being to contribute to theunderstanding of the process chain from convection ini-tiation via cloud formation to precipitation It is amobile observation platform consisting of various in-situand remote sensing systems Most of the measurementdevices are preferably concentrated at one supersiteAdditional instruments are distributed within an areaof meso-c scale extension to exploit the synergy ofcomplementary measurements and to cover the spatialextension of convection-related phenomena KITcubewas implemented first during the international projectHyMeX (Hydrological cycle in the MediterraneaneXperiment DROBINSKI et al 2014 DUCROCQ et al2013) on the French island of Corsica in theMediterranean

During the late summer and autumn the westernMediterranean is often affected by heavy precipitation(DUCROCQ et al 2008) mainly convective in nature(DOSWELL et al 1998) Both the orography and thesea surface influence the formation and evolution of theconvective systems in the region (HOMAR et al 1999BUZZI and FOSCHINI 2000 ROTUNNO and FERRETTI2001) To improve the understanding of the initiationand development of heavy precipitation events in theMediterranean region the HyMeX project was initiatedUnder this project a special observation period (SOP)was conducted in autumn 2012 to provide comprehen-sive data sets for process studies and model evaluationalike Due to the processes involved in convective precip-itation observations on different scales were carried outDuring the SOP the KITcube was positioned at one ofHyMeXrsquos supersites on the Corsican Island (Fig 1)

Two main scientific goals were pursued with thedeployment of KITcube at this position (i) Corsica unitesboth features necessary to cause orographically inducedconvection or to modify it ie a landmass of sufficientsize (about 80 km by 180 km) and a high mountain ridgeof up to 2700 m above mean sea level (MSL) The KIT-cube investigated the influence of the island on initiationand modification of isolated convection (ii) For most ofthe synoptic conditions under which severe precipitationevents occur in the Ligurian Sea the Corsican Island ison the upwind side (LAMBERT et al 2011) ThusKITcube was well-positioned to provide upstream condi-tions for the intense precipitation events affecting conti-nental south-eastern France and northern and centralItaly

The aim of this paper is to introduce the KITcubeobservation platform present its deployment on Corsicaduring the HyMeX field campaign in 2012 and to dem-onstrate its observation capabilities to investigate convec-tive processes For the latter a time section from theintense observation period (IOP) 12a (11 October2012) was selected

2 The mobile observation platformKITcube

KITcube consists of measurement systems which provideinformation about the energy exchange at the surface tur-bulence and mean conditions in the convective boundarylayer (CBL) and whole troposphere respectively andabout cloud and precipitation properties The differentmeasurement systems of KITcube are listed in Table 1including eg temporal resolutions and measurementranges of the main derived parameters Details are givenbelow

21 Surface energy exchange andnear-surface observations

The energy balance stations measure meteorological vari-ables (temperature humidity wind speed and directionair pressure precipitation) radiation temperature of thesurface solar and reflected irradiance long-wave incom-ing and outgoing radiation soil heat sensible heat latentheat and momentum fluxes (KALTHOFF et al 2006) aswell as the moisture and temperature in the soil The tur-bulent fluxes are measured with an ultrasonic anemome-terthermometer and a fast infrared hygrometer Forpost-processing of turbulent fluxes the eddy-correlationsoftware package TK31 described by MAUDER andFOKEN (2011) is used

Seven additional flux stations provide temperaturehumidity (both at three levels) wind speed and wind

Figure 1 Deployment of KITcube devices on the Corsican Islandmain deployment sites (yellow) mobile towers (white) GPSstations operated by IGN and ACTIPLAN (magenta) and additionalGPS stations deployed by GFZ (red) For details see section 3

634 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

direction momentum and sensible heat fluxes The sta-tions deliver information on the spatial variability ofthe sensible heat fluxes Additionally a scintillometer(SLS20 Scintec) measures the sensible heat flux for spa-tial averages along the laser beam with an averaging timeof 1 minute The transmitting laser beam and receiverare typically separated by a distance of about 100 m to200 m

A 20 m high meteorological tower provides windspeed temperature and humidity profiles as well as winddirection and sensible heat and momentum fluxes at20 m above ground level (AGL) Soil temperatures aremeasured at six different depths

Battery-powered mobile towers that are deployedflexibly in the investigation area measure temperaturehumidity wind speed and wind direction air pressureprecipitation and sensible heat and momentum fluxes

22 Mean and turbulent atmosphericconditions

To obtain vertical profiles of temperature humidity windspeed and wind direction in the troposphere two radio-sonde systems (DFM-09 Graw) are available Dependingon height the 1 s raw data are averaged over 2 s (betweenground and 3 km AGL) 4 s (3 km and 10 km) and 8 s(gt 10 km) During IOPs radiosondes are typicallylaunched at 2- to 3-hour intervals The sounding data arealso used to derive significant levels and convection-related parameters such as the convective condensationlevel (CCL) lifting condensation layer (LCL) level of freeconvection (LFC) convective inhibition (CIN) and con-vective available potential energy (CAPE)

A ground-based scanning microwave radiometer(humidity and temperature profiler HATPRO) designedby Radiometer Physics measures the sky brightnesstemperature at 12 frequencies distributed within the22-30 GHz and 51-59 GHz bands With statistical retri-evals provided by the University of Cologne (LOHNERTand CREWELL 2003 CREWELL and LOHNERT 2003)these sky brightness temperatures are inverted to obtaintemperature and humidity profiles up to 10 km heightUsing additional boundary layer scans vertical tempera-ture profiles are obtained in 50 m steps close to the sur-face and in steps of 200 m to 400 m in the freetroposphere The system also provides integrated watervapour (IWV) and liquid water path (LWP) data alongthe line of sight 360 azimuth scans at fixed elevationangles carried out on a regular basis provide informationabout spatial humidity distribution The temperature ofcloud base is determined with an upwards directed infra-red thermometer

Additionally a JAVAD GPS receiver combined withsurface observations of temperature and pressure is usedto determine IWV with a temporal resolution of 15 minThe calculations are performed by GFZ Potsdam based

on algorithms described by DICK et al (2001) andGENDT et al (2004)

To determine the mean horizontal and vertical veloc-ity and vertical velocity variance a phased array sodar(MFAS Scintec) is used Operation frequencies in the2 kHz band and multi-frequency operation allow for aspatial resolution of 10 m and a maximum measurementrange of up to 1000 m AGL Backscatter data can beused for mixing height detection

One 2 lm and another 16 lm heterodyne wind lidarwith beam width of 75 mm at laser exit (WindTracerLockheed Martin) are part of KITcube to measure aerosolbackscatter and calculate radial velocity (TRAUMNER

et al 2011) Applying the vertical stare mode immedi-ately yields vertical velocity with a time resolution of1 s from about 375 m AGL to the top of the boundarylayer and partly above (depending on the aerosol concen-tration) The measurements are mainly restricted to thecloud-free atmosphere because wind lidars only partlypenetrate clouds The uncorrelated noise of velocitycould be estimated to be less than 02 m s1 for a highsignal to noise ratio (SNR) range Both lidars havetwo-axis scanners and yield profiles of the horizontalwind velocity via the velocity-azimuth display (VAD)technique (eg BROWNING and WEXLER 1968)Synchronised monitoring of the two lidars allow fordual-Doppler measurements (STAWIARSKI et al 2013)The CBL depth can be estimated using the profiles ofthe vertical velocity variance

In order to cover the range between the surface andthe lowest WindTracer measurement heights a third windlidar (Windcube 8 Leosphere) with a wavelength of154 lm is used which measures the wind profile from40 m AGL up to about 600 m AGL with averaging timesof 16 s to 10 min and a vertical range resolution of 20 mIn normal operation mode this small wind lidar has afixed elevation angle of 752 and an azimuth scannerBy means of a 4-point stop and stare plane position indi-cator (PPI) scan applying the VAD algorithm the windprofile is calculated Running a special operation modeby using a motorised tilted table allows for continuousvertical stare measurements for the direct detection ofvertical velocity In combination with a WindTracer thisleads to a full vertical coverage of vertical velocity fromthe ground up into the entrainment zone

23 Measurements of clouds andprecipitation properties

Clouds and precipitation properties are measured byremote sensing techniques as well as by in-situ observa-tions Every two minutes two cameras with fish eyelenses take photos of the upper hemisphere from whichthe cloud type and cloud cover can be estimated

A 355 GHz scanning cloud radar (MIRA36-SMetek) is implemented in the KITcube to measure cloud

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 635

eschweizerbart_xxx

properties It is dopplerised and dual-polarised but trans-mits horizontally polarised radiation only and detects hor-izontally and vertically polarised reflections providing thelinear depolarisation ratio (LDR) of the hydrometeorsWith a 200 ns pulse duration and a 07 beam width it

reaches a very fine spatial resolution of 30 m for radialvelocity Additional data measured in the absence ofclouds are available to determine radial velocity whenclear-air returns are sufficient (eg caused by insects orother floating materialdebris dust and Bragg scattering

Table 1a Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement heights(m AGL)

Surface energy exchange and near-surface observations

Energy balance stations(Different

manufacturers)

2

Temperature 10 min (1 s) 3Humidity 10 min (1 s) 33-d wind 10 min (005 s) 4Humidity 10 min (005 s) 4

Temperature 10 min (005 s) 4Air pressure 10 min (1 s) 01Precipitation 10 min 1

Radiation temperatureof surface

10 min (1 s) 0

Solar and reflected 10 min (1 s) 3irradiance

Long wave incomingand outgoing radiation

10 min (1 s) 3

Sensible and latent 30 min (005 s) 4heat and momentum

fluxesSoil heat flux 10 min (1 s) 005Soil moisture 10 min 4 variable depths

Soil temperature 10 min 4 variable depths

Flux stations(Different

manufacturers)

7

Temperature 10 min (1 s) 1 2 4Humidity 10 min (1 s) 1 2 43-d wind 10 mm (005 s) 5

Temperature 10 min (005 s) 5Air pressure 10 min (1 s) 01

Sensible heat and momentum fluxes 30 min (005 s) 5

Scintillometer (Scintec) 1 Sensible heat flux 1 min 2ndash3

20-m tower(Different

manufacturers)

1

Temperature 10 min (1 s) 05 1 2 4 8 16 19Humidity 10 min (1 s) 05 1 2 4 8 16 19Wind speed 10 min (1 s) 05 1 2 4 8 16 19

Wind direction 10 min (1 s) 20Temperature 10 min (005 s) 20

Soil temperature 10 min (1 s) 002 004 008 016032 064

3-d wind 10 min (005 s) 20Sensible heat andmomentum fluxes

30 min (005 s) 20

Mobile towers (Young) 3

Temperature 10 min 2Humidity 10 min 23-d wind 003 s 4

Temperature 003 s 4Air pressure 10 min 1Precipitation 10 min 1

Sensible heat andmomentum fluxes

30 min (003 s) 4

636 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

from refractive-index fluctuations BANTA et al 2013)The cloud radar is able to scan at all azimuths fromvertical direction down to the 45 zenith angle Typicallya time resolution of 10 s is used in the vertical staremode Additional instrument specifications are given byTRAUMNER et al (2011)

To measure the cloud base height (CBH) the cloudradar is complemented by a ceilometer (CHM15k Jenop-tik) The ceilometer is a lidar providing uncalibratedbackscatter at a spatial resolution of 15 m From these

raw data CBHs intrusion depths and other propertiesof up to three different cloud layers can be determined

Volume filling information on precipitation is gath-ered by a mobile X-band radar (Meteor 50DX Gematro-nik) This radar is dopplerised and able to measurepolarisation properties of the hydrometeors It transmitsradiation polarised at 45 whereas the received signalsare split up into the horizontal and vertical polarisationthus providing polarisation properties such as differentialreflectivity (ZDR) and specific differential phase (KDP)

Table 1b Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments The measurement range column includes the minimal distance spatial resolution and maximum distance x stands for avariable resolution

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement range (m)

Mean and turbulent atmospheric conditionsRadiosonde system

(Graw)2 Temperature 2ndash8 (1 s) 0x15000

Humidity 2ndash8 (1 s) 0x15000Wind velocity and direction 2-8 (1 s) 0xl5000

Microwave radiometer(Radiometer Physics)

1 Temperature 10 s 050120012002005000500040010000

Humidity 10 s 0200200020004005000500080010000

IWV 10 s ndashLWP 10 s ndash

Infrared temperature 10 s ndash

GPS receiver (JAVAD) 1 IWV 15 min ndash

Sodar (Scintec) 1 Wind velocity and direction 30 min 30101000Vertical velocity variance 30 min 30101000

Wind lidar WindTracer(Lockheed Martin)

2 Radial velocity 01ndash1 s 375x12000Aerosol backscatter 01ndash1 s 375x12000

Wind velocity and direction scan dependent 75508475

Wind lidar Windcube(Leosphere)

1 Radial velocity 16 s 4020600Wind speed and direction 10 min (7 s) 4020600

Measurements of clouds and precipitation propertiesCloud camera (Mobotix) 2 Hemispheric photo 2 min ndash

Cloud radar (Metek) 1 Radial velocity 1ndash10 s 1503014460

Ceilometer (Jenoptik) 1 Cloud base heights 1 min 1501515000

X-band radar (Gematronik) 1 Precipitation 5 min 125250100000

Micro rain radar (Metek) 1 Precipitation Drop size distribution 1 min 1 min 10010032001001003200

Disdrometer (KIT) 1 Precipitation Drop size distribution 1 min 1 min 1 1

Disdrometer (Distromet) 1 Precipitation Drop size distribution 1 min 1 min 01 01

Rain gauge (EML) 2 Precipitation 1 min 1

height dependent

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 637

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

KITcube ndash a mobile observation platform for convectionstudies deployed during HyMeX

Norbert Kalthoff1

Bianca Adler1 Andreas Wieser

1 Martin Kohler

1 Katja Traumner

1

Jan Handwerker1 Ulrich Corsmeier

1 Samiro Khodayar

1 Dominique Lambert

2

Andreas Kopmann3 Norbert Kunka

3 Galina Dick

4 Markus Ramatschi

4 Jens Wickert

4

and Christoph Kottmeier1

1Institut fur Meteorologie und Klimaforschung (IMK-TRO) Karlsruher Institut fur Technologie (KIT) KarlsruheGermany

2Laboratoire drsquoAerologie Universite de Toulouse Toulouse France3Institut fur Prozessverarbeitung und Elektronik (IPE) Karlsruher Institut fur Technologie (KIT) KarlsruheGermany

4Department of Geodesy and Remote Sensing German Research Centre for Geosciences (GFZ) PotsdamGermany

(Manuscript received August 9 2013 in revised form October 2 2013 accepted October 3 2013)

AbstractWith the increase of spatial resolution of weather forecast models to order O(1 km) the need for adequateobservations for model validation becomes evident Therefore we designed and constructed the lsquolsquoKITcubersquorsquoa mobile observation platform for convection studies of processes on the meso-c scale The KITcube consistsof in-situ and remote sensing systems which allow measuring the energy balance components of the Earthrsquossurface at different sites the mean atmospheric conditions by radiosondes GPS station and a microwaveradiometer the turbulent characteristics by a sodar and wind lidars and cloud and precipitation properties byuse of a cloud radar a micro rain radar disdrometers rain gauges and an X-band rain radar The KITcubewas deployed fully for the first time on the French island of Corsica during the HyMeX (Hydrological cyclein the Mediterranean eXperiment) field campaign in 2012 In this article the components of KITcube and itsimplementation on the island are described Moreover results from one of the HyMeX intensive observationperiods are presented to show the capabilities of KITcube

Keywords HyMeX convection observation platform

1 Introduction

During the last years spatial resolution of weather fore-cast models increased to order O (1 km) allowing toresolve larger convective systems like thunderstormsConvection however encompasses multiple scales frommicro- to mesoscale Therefore the transition from con-vection initiation via shallow to deep convection is notsimulated properly (BRYAN et al 2003) Besides under-standing of processes and mechanisms governing theconvection development has been limited up to now

To improve the knowledge about convection scale-adapted measurements were performed (WECKWERTH

et al 2004 WILSON and ROBERTS 2006 BROWNING

et al 2007 KOTTMEIER et al 2008 WULFMEYER

et al 2011) So-called supersites were installed wherecoordinated measurements of different in-situ and remotesystems were exploited synergetically so that convectiveprocesses could be analysed (eg MILLER and SLINGO2007 BEHRENDT et al 2011 CORSMEIER et al 2011

KALTHOFF et al 2013) Often smaller-scale meteorolog-ical networks were embedded in larger-scale investiga-tion domains to capture spatial inhomogeneitiesAircraft missions including dropsonde releases wereperformed to flexibly respond to convection evolution

From these and previous observations it was foundthat convection in many cases was initiated by low-levelconvergence zones (KHODAYAR et al 2010 2013)and favoured by small-scale moisture variability(WECKWERTH 2000) Convergence zones are often gen-erated by mesoscale surface heterogeneities eg givensoil moisture (TAYLOR et al 2011) and land-sea contrastor orography (KALTHOFF et al 2009) Besides trigger-ing convergence zones also support continued growthof convection because of the permanent upward motionand moisture transport there For mountainous terrainadditional types of thunderstorm initiation mechanismswere distinguished (eg BANTA 1990) direct orographiclifting and obstacle or aerodynamic effects (blockingflow deflection gravity wave effects) The influence ofislands on the triggering of thunderstorms was also doc-umented well (eg WILSON et al 2001 QIAN 2007)Mountainous islands turned out to be most effectivebecause sea breezes and valley winds that are roughly

Corresponding author Norbert Kalthoff Institut fur Meteorologie undKlimaforschung (IMK-TRO) Karlsruher Institut fur Technologie (KIT)POB 3640 76021 Karlsruhe Germany e-mail norbertkalthoffkitedu

Meteorologische Zeitschrift Vol 22 No 6 633ndash647 (February 2014) Open Access Article by Gebruder Borntraeger 2013

DOI 1011270941-2948201305420941-294820130542 $ 675

Gebruder Borntraeger Stuttgart 2013

eschweizerbart_xxx

in phase with each other may combine to strengthen thediurnal cycle of winds and form convergence zones dur-ing daytime (KOTTMEIER et al 2000)

The lsquolsquoKITcubersquorsquo was developed for state-of-the-artobservations the objective being to contribute to theunderstanding of the process chain from convection ini-tiation via cloud formation to precipitation It is amobile observation platform consisting of various in-situand remote sensing systems Most of the measurementdevices are preferably concentrated at one supersiteAdditional instruments are distributed within an areaof meso-c scale extension to exploit the synergy ofcomplementary measurements and to cover the spatialextension of convection-related phenomena KITcubewas implemented first during the international projectHyMeX (Hydrological cycle in the MediterraneaneXperiment DROBINSKI et al 2014 DUCROCQ et al2013) on the French island of Corsica in theMediterranean

During the late summer and autumn the westernMediterranean is often affected by heavy precipitation(DUCROCQ et al 2008) mainly convective in nature(DOSWELL et al 1998) Both the orography and thesea surface influence the formation and evolution of theconvective systems in the region (HOMAR et al 1999BUZZI and FOSCHINI 2000 ROTUNNO and FERRETTI2001) To improve the understanding of the initiationand development of heavy precipitation events in theMediterranean region the HyMeX project was initiatedUnder this project a special observation period (SOP)was conducted in autumn 2012 to provide comprehen-sive data sets for process studies and model evaluationalike Due to the processes involved in convective precip-itation observations on different scales were carried outDuring the SOP the KITcube was positioned at one ofHyMeXrsquos supersites on the Corsican Island (Fig 1)

Two main scientific goals were pursued with thedeployment of KITcube at this position (i) Corsica unitesboth features necessary to cause orographically inducedconvection or to modify it ie a landmass of sufficientsize (about 80 km by 180 km) and a high mountain ridgeof up to 2700 m above mean sea level (MSL) The KIT-cube investigated the influence of the island on initiationand modification of isolated convection (ii) For most ofthe synoptic conditions under which severe precipitationevents occur in the Ligurian Sea the Corsican Island ison the upwind side (LAMBERT et al 2011) ThusKITcube was well-positioned to provide upstream condi-tions for the intense precipitation events affecting conti-nental south-eastern France and northern and centralItaly

The aim of this paper is to introduce the KITcubeobservation platform present its deployment on Corsicaduring the HyMeX field campaign in 2012 and to dem-onstrate its observation capabilities to investigate convec-tive processes For the latter a time section from theintense observation period (IOP) 12a (11 October2012) was selected

2 The mobile observation platformKITcube

KITcube consists of measurement systems which provideinformation about the energy exchange at the surface tur-bulence and mean conditions in the convective boundarylayer (CBL) and whole troposphere respectively andabout cloud and precipitation properties The differentmeasurement systems of KITcube are listed in Table 1including eg temporal resolutions and measurementranges of the main derived parameters Details are givenbelow

21 Surface energy exchange andnear-surface observations

The energy balance stations measure meteorological vari-ables (temperature humidity wind speed and directionair pressure precipitation) radiation temperature of thesurface solar and reflected irradiance long-wave incom-ing and outgoing radiation soil heat sensible heat latentheat and momentum fluxes (KALTHOFF et al 2006) aswell as the moisture and temperature in the soil The tur-bulent fluxes are measured with an ultrasonic anemome-terthermometer and a fast infrared hygrometer Forpost-processing of turbulent fluxes the eddy-correlationsoftware package TK31 described by MAUDER andFOKEN (2011) is used

Seven additional flux stations provide temperaturehumidity (both at three levels) wind speed and wind

Figure 1 Deployment of KITcube devices on the Corsican Islandmain deployment sites (yellow) mobile towers (white) GPSstations operated by IGN and ACTIPLAN (magenta) and additionalGPS stations deployed by GFZ (red) For details see section 3

634 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

direction momentum and sensible heat fluxes The sta-tions deliver information on the spatial variability ofthe sensible heat fluxes Additionally a scintillometer(SLS20 Scintec) measures the sensible heat flux for spa-tial averages along the laser beam with an averaging timeof 1 minute The transmitting laser beam and receiverare typically separated by a distance of about 100 m to200 m

A 20 m high meteorological tower provides windspeed temperature and humidity profiles as well as winddirection and sensible heat and momentum fluxes at20 m above ground level (AGL) Soil temperatures aremeasured at six different depths

Battery-powered mobile towers that are deployedflexibly in the investigation area measure temperaturehumidity wind speed and wind direction air pressureprecipitation and sensible heat and momentum fluxes

22 Mean and turbulent atmosphericconditions

To obtain vertical profiles of temperature humidity windspeed and wind direction in the troposphere two radio-sonde systems (DFM-09 Graw) are available Dependingon height the 1 s raw data are averaged over 2 s (betweenground and 3 km AGL) 4 s (3 km and 10 km) and 8 s(gt 10 km) During IOPs radiosondes are typicallylaunched at 2- to 3-hour intervals The sounding data arealso used to derive significant levels and convection-related parameters such as the convective condensationlevel (CCL) lifting condensation layer (LCL) level of freeconvection (LFC) convective inhibition (CIN) and con-vective available potential energy (CAPE)

A ground-based scanning microwave radiometer(humidity and temperature profiler HATPRO) designedby Radiometer Physics measures the sky brightnesstemperature at 12 frequencies distributed within the22-30 GHz and 51-59 GHz bands With statistical retri-evals provided by the University of Cologne (LOHNERTand CREWELL 2003 CREWELL and LOHNERT 2003)these sky brightness temperatures are inverted to obtaintemperature and humidity profiles up to 10 km heightUsing additional boundary layer scans vertical tempera-ture profiles are obtained in 50 m steps close to the sur-face and in steps of 200 m to 400 m in the freetroposphere The system also provides integrated watervapour (IWV) and liquid water path (LWP) data alongthe line of sight 360 azimuth scans at fixed elevationangles carried out on a regular basis provide informationabout spatial humidity distribution The temperature ofcloud base is determined with an upwards directed infra-red thermometer

Additionally a JAVAD GPS receiver combined withsurface observations of temperature and pressure is usedto determine IWV with a temporal resolution of 15 minThe calculations are performed by GFZ Potsdam based

on algorithms described by DICK et al (2001) andGENDT et al (2004)

To determine the mean horizontal and vertical veloc-ity and vertical velocity variance a phased array sodar(MFAS Scintec) is used Operation frequencies in the2 kHz band and multi-frequency operation allow for aspatial resolution of 10 m and a maximum measurementrange of up to 1000 m AGL Backscatter data can beused for mixing height detection

One 2 lm and another 16 lm heterodyne wind lidarwith beam width of 75 mm at laser exit (WindTracerLockheed Martin) are part of KITcube to measure aerosolbackscatter and calculate radial velocity (TRAUMNER

et al 2011) Applying the vertical stare mode immedi-ately yields vertical velocity with a time resolution of1 s from about 375 m AGL to the top of the boundarylayer and partly above (depending on the aerosol concen-tration) The measurements are mainly restricted to thecloud-free atmosphere because wind lidars only partlypenetrate clouds The uncorrelated noise of velocitycould be estimated to be less than 02 m s1 for a highsignal to noise ratio (SNR) range Both lidars havetwo-axis scanners and yield profiles of the horizontalwind velocity via the velocity-azimuth display (VAD)technique (eg BROWNING and WEXLER 1968)Synchronised monitoring of the two lidars allow fordual-Doppler measurements (STAWIARSKI et al 2013)The CBL depth can be estimated using the profiles ofthe vertical velocity variance

In order to cover the range between the surface andthe lowest WindTracer measurement heights a third windlidar (Windcube 8 Leosphere) with a wavelength of154 lm is used which measures the wind profile from40 m AGL up to about 600 m AGL with averaging timesof 16 s to 10 min and a vertical range resolution of 20 mIn normal operation mode this small wind lidar has afixed elevation angle of 752 and an azimuth scannerBy means of a 4-point stop and stare plane position indi-cator (PPI) scan applying the VAD algorithm the windprofile is calculated Running a special operation modeby using a motorised tilted table allows for continuousvertical stare measurements for the direct detection ofvertical velocity In combination with a WindTracer thisleads to a full vertical coverage of vertical velocity fromthe ground up into the entrainment zone

23 Measurements of clouds andprecipitation properties

Clouds and precipitation properties are measured byremote sensing techniques as well as by in-situ observa-tions Every two minutes two cameras with fish eyelenses take photos of the upper hemisphere from whichthe cloud type and cloud cover can be estimated

A 355 GHz scanning cloud radar (MIRA36-SMetek) is implemented in the KITcube to measure cloud

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 635

eschweizerbart_xxx

properties It is dopplerised and dual-polarised but trans-mits horizontally polarised radiation only and detects hor-izontally and vertically polarised reflections providing thelinear depolarisation ratio (LDR) of the hydrometeorsWith a 200 ns pulse duration and a 07 beam width it

reaches a very fine spatial resolution of 30 m for radialvelocity Additional data measured in the absence ofclouds are available to determine radial velocity whenclear-air returns are sufficient (eg caused by insects orother floating materialdebris dust and Bragg scattering

Table 1a Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement heights(m AGL)

Surface energy exchange and near-surface observations

Energy balance stations(Different

manufacturers)

2

Temperature 10 min (1 s) 3Humidity 10 min (1 s) 33-d wind 10 min (005 s) 4Humidity 10 min (005 s) 4

Temperature 10 min (005 s) 4Air pressure 10 min (1 s) 01Precipitation 10 min 1

Radiation temperatureof surface

10 min (1 s) 0

Solar and reflected 10 min (1 s) 3irradiance

Long wave incomingand outgoing radiation

10 min (1 s) 3

Sensible and latent 30 min (005 s) 4heat and momentum

fluxesSoil heat flux 10 min (1 s) 005Soil moisture 10 min 4 variable depths

Soil temperature 10 min 4 variable depths

Flux stations(Different

manufacturers)

7

Temperature 10 min (1 s) 1 2 4Humidity 10 min (1 s) 1 2 43-d wind 10 mm (005 s) 5

Temperature 10 min (005 s) 5Air pressure 10 min (1 s) 01

Sensible heat and momentum fluxes 30 min (005 s) 5

Scintillometer (Scintec) 1 Sensible heat flux 1 min 2ndash3

20-m tower(Different

manufacturers)

1

Temperature 10 min (1 s) 05 1 2 4 8 16 19Humidity 10 min (1 s) 05 1 2 4 8 16 19Wind speed 10 min (1 s) 05 1 2 4 8 16 19

Wind direction 10 min (1 s) 20Temperature 10 min (005 s) 20

Soil temperature 10 min (1 s) 002 004 008 016032 064

3-d wind 10 min (005 s) 20Sensible heat andmomentum fluxes

30 min (005 s) 20

Mobile towers (Young) 3

Temperature 10 min 2Humidity 10 min 23-d wind 003 s 4

Temperature 003 s 4Air pressure 10 min 1Precipitation 10 min 1

Sensible heat andmomentum fluxes

30 min (003 s) 4

636 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

from refractive-index fluctuations BANTA et al 2013)The cloud radar is able to scan at all azimuths fromvertical direction down to the 45 zenith angle Typicallya time resolution of 10 s is used in the vertical staremode Additional instrument specifications are given byTRAUMNER et al (2011)

To measure the cloud base height (CBH) the cloudradar is complemented by a ceilometer (CHM15k Jenop-tik) The ceilometer is a lidar providing uncalibratedbackscatter at a spatial resolution of 15 m From these

raw data CBHs intrusion depths and other propertiesof up to three different cloud layers can be determined

Volume filling information on precipitation is gath-ered by a mobile X-band radar (Meteor 50DX Gematro-nik) This radar is dopplerised and able to measurepolarisation properties of the hydrometeors It transmitsradiation polarised at 45 whereas the received signalsare split up into the horizontal and vertical polarisationthus providing polarisation properties such as differentialreflectivity (ZDR) and specific differential phase (KDP)

Table 1b Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments The measurement range column includes the minimal distance spatial resolution and maximum distance x stands for avariable resolution

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement range (m)

Mean and turbulent atmospheric conditionsRadiosonde system

(Graw)2 Temperature 2ndash8 (1 s) 0x15000

Humidity 2ndash8 (1 s) 0x15000Wind velocity and direction 2-8 (1 s) 0xl5000

Microwave radiometer(Radiometer Physics)

1 Temperature 10 s 050120012002005000500040010000

Humidity 10 s 0200200020004005000500080010000

IWV 10 s ndashLWP 10 s ndash

Infrared temperature 10 s ndash

GPS receiver (JAVAD) 1 IWV 15 min ndash

Sodar (Scintec) 1 Wind velocity and direction 30 min 30101000Vertical velocity variance 30 min 30101000

Wind lidar WindTracer(Lockheed Martin)

2 Radial velocity 01ndash1 s 375x12000Aerosol backscatter 01ndash1 s 375x12000

Wind velocity and direction scan dependent 75508475

Wind lidar Windcube(Leosphere)

1 Radial velocity 16 s 4020600Wind speed and direction 10 min (7 s) 4020600

Measurements of clouds and precipitation propertiesCloud camera (Mobotix) 2 Hemispheric photo 2 min ndash

Cloud radar (Metek) 1 Radial velocity 1ndash10 s 1503014460

Ceilometer (Jenoptik) 1 Cloud base heights 1 min 1501515000

X-band radar (Gematronik) 1 Precipitation 5 min 125250100000

Micro rain radar (Metek) 1 Precipitation Drop size distribution 1 min 1 min 10010032001001003200

Disdrometer (KIT) 1 Precipitation Drop size distribution 1 min 1 min 1 1

Disdrometer (Distromet) 1 Precipitation Drop size distribution 1 min 1 min 01 01

Rain gauge (EML) 2 Precipitation 1 min 1

height dependent

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 637

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

in phase with each other may combine to strengthen thediurnal cycle of winds and form convergence zones dur-ing daytime (KOTTMEIER et al 2000)

The lsquolsquoKITcubersquorsquo was developed for state-of-the-artobservations the objective being to contribute to theunderstanding of the process chain from convection ini-tiation via cloud formation to precipitation It is amobile observation platform consisting of various in-situand remote sensing systems Most of the measurementdevices are preferably concentrated at one supersiteAdditional instruments are distributed within an areaof meso-c scale extension to exploit the synergy ofcomplementary measurements and to cover the spatialextension of convection-related phenomena KITcubewas implemented first during the international projectHyMeX (Hydrological cycle in the MediterraneaneXperiment DROBINSKI et al 2014 DUCROCQ et al2013) on the French island of Corsica in theMediterranean

During the late summer and autumn the westernMediterranean is often affected by heavy precipitation(DUCROCQ et al 2008) mainly convective in nature(DOSWELL et al 1998) Both the orography and thesea surface influence the formation and evolution of theconvective systems in the region (HOMAR et al 1999BUZZI and FOSCHINI 2000 ROTUNNO and FERRETTI2001) To improve the understanding of the initiationand development of heavy precipitation events in theMediterranean region the HyMeX project was initiatedUnder this project a special observation period (SOP)was conducted in autumn 2012 to provide comprehen-sive data sets for process studies and model evaluationalike Due to the processes involved in convective precip-itation observations on different scales were carried outDuring the SOP the KITcube was positioned at one ofHyMeXrsquos supersites on the Corsican Island (Fig 1)

Two main scientific goals were pursued with thedeployment of KITcube at this position (i) Corsica unitesboth features necessary to cause orographically inducedconvection or to modify it ie a landmass of sufficientsize (about 80 km by 180 km) and a high mountain ridgeof up to 2700 m above mean sea level (MSL) The KIT-cube investigated the influence of the island on initiationand modification of isolated convection (ii) For most ofthe synoptic conditions under which severe precipitationevents occur in the Ligurian Sea the Corsican Island ison the upwind side (LAMBERT et al 2011) ThusKITcube was well-positioned to provide upstream condi-tions for the intense precipitation events affecting conti-nental south-eastern France and northern and centralItaly

The aim of this paper is to introduce the KITcubeobservation platform present its deployment on Corsicaduring the HyMeX field campaign in 2012 and to dem-onstrate its observation capabilities to investigate convec-tive processes For the latter a time section from theintense observation period (IOP) 12a (11 October2012) was selected

2 The mobile observation platformKITcube

KITcube consists of measurement systems which provideinformation about the energy exchange at the surface tur-bulence and mean conditions in the convective boundarylayer (CBL) and whole troposphere respectively andabout cloud and precipitation properties The differentmeasurement systems of KITcube are listed in Table 1including eg temporal resolutions and measurementranges of the main derived parameters Details are givenbelow

21 Surface energy exchange andnear-surface observations

The energy balance stations measure meteorological vari-ables (temperature humidity wind speed and directionair pressure precipitation) radiation temperature of thesurface solar and reflected irradiance long-wave incom-ing and outgoing radiation soil heat sensible heat latentheat and momentum fluxes (KALTHOFF et al 2006) aswell as the moisture and temperature in the soil The tur-bulent fluxes are measured with an ultrasonic anemome-terthermometer and a fast infrared hygrometer Forpost-processing of turbulent fluxes the eddy-correlationsoftware package TK31 described by MAUDER andFOKEN (2011) is used

Seven additional flux stations provide temperaturehumidity (both at three levels) wind speed and wind

Figure 1 Deployment of KITcube devices on the Corsican Islandmain deployment sites (yellow) mobile towers (white) GPSstations operated by IGN and ACTIPLAN (magenta) and additionalGPS stations deployed by GFZ (red) For details see section 3

634 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

direction momentum and sensible heat fluxes The sta-tions deliver information on the spatial variability ofthe sensible heat fluxes Additionally a scintillometer(SLS20 Scintec) measures the sensible heat flux for spa-tial averages along the laser beam with an averaging timeof 1 minute The transmitting laser beam and receiverare typically separated by a distance of about 100 m to200 m

A 20 m high meteorological tower provides windspeed temperature and humidity profiles as well as winddirection and sensible heat and momentum fluxes at20 m above ground level (AGL) Soil temperatures aremeasured at six different depths

Battery-powered mobile towers that are deployedflexibly in the investigation area measure temperaturehumidity wind speed and wind direction air pressureprecipitation and sensible heat and momentum fluxes

22 Mean and turbulent atmosphericconditions

To obtain vertical profiles of temperature humidity windspeed and wind direction in the troposphere two radio-sonde systems (DFM-09 Graw) are available Dependingon height the 1 s raw data are averaged over 2 s (betweenground and 3 km AGL) 4 s (3 km and 10 km) and 8 s(gt 10 km) During IOPs radiosondes are typicallylaunched at 2- to 3-hour intervals The sounding data arealso used to derive significant levels and convection-related parameters such as the convective condensationlevel (CCL) lifting condensation layer (LCL) level of freeconvection (LFC) convective inhibition (CIN) and con-vective available potential energy (CAPE)

A ground-based scanning microwave radiometer(humidity and temperature profiler HATPRO) designedby Radiometer Physics measures the sky brightnesstemperature at 12 frequencies distributed within the22-30 GHz and 51-59 GHz bands With statistical retri-evals provided by the University of Cologne (LOHNERTand CREWELL 2003 CREWELL and LOHNERT 2003)these sky brightness temperatures are inverted to obtaintemperature and humidity profiles up to 10 km heightUsing additional boundary layer scans vertical tempera-ture profiles are obtained in 50 m steps close to the sur-face and in steps of 200 m to 400 m in the freetroposphere The system also provides integrated watervapour (IWV) and liquid water path (LWP) data alongthe line of sight 360 azimuth scans at fixed elevationangles carried out on a regular basis provide informationabout spatial humidity distribution The temperature ofcloud base is determined with an upwards directed infra-red thermometer

Additionally a JAVAD GPS receiver combined withsurface observations of temperature and pressure is usedto determine IWV with a temporal resolution of 15 minThe calculations are performed by GFZ Potsdam based

on algorithms described by DICK et al (2001) andGENDT et al (2004)

To determine the mean horizontal and vertical veloc-ity and vertical velocity variance a phased array sodar(MFAS Scintec) is used Operation frequencies in the2 kHz band and multi-frequency operation allow for aspatial resolution of 10 m and a maximum measurementrange of up to 1000 m AGL Backscatter data can beused for mixing height detection

One 2 lm and another 16 lm heterodyne wind lidarwith beam width of 75 mm at laser exit (WindTracerLockheed Martin) are part of KITcube to measure aerosolbackscatter and calculate radial velocity (TRAUMNER

et al 2011) Applying the vertical stare mode immedi-ately yields vertical velocity with a time resolution of1 s from about 375 m AGL to the top of the boundarylayer and partly above (depending on the aerosol concen-tration) The measurements are mainly restricted to thecloud-free atmosphere because wind lidars only partlypenetrate clouds The uncorrelated noise of velocitycould be estimated to be less than 02 m s1 for a highsignal to noise ratio (SNR) range Both lidars havetwo-axis scanners and yield profiles of the horizontalwind velocity via the velocity-azimuth display (VAD)technique (eg BROWNING and WEXLER 1968)Synchronised monitoring of the two lidars allow fordual-Doppler measurements (STAWIARSKI et al 2013)The CBL depth can be estimated using the profiles ofthe vertical velocity variance

In order to cover the range between the surface andthe lowest WindTracer measurement heights a third windlidar (Windcube 8 Leosphere) with a wavelength of154 lm is used which measures the wind profile from40 m AGL up to about 600 m AGL with averaging timesof 16 s to 10 min and a vertical range resolution of 20 mIn normal operation mode this small wind lidar has afixed elevation angle of 752 and an azimuth scannerBy means of a 4-point stop and stare plane position indi-cator (PPI) scan applying the VAD algorithm the windprofile is calculated Running a special operation modeby using a motorised tilted table allows for continuousvertical stare measurements for the direct detection ofvertical velocity In combination with a WindTracer thisleads to a full vertical coverage of vertical velocity fromthe ground up into the entrainment zone

23 Measurements of clouds andprecipitation properties

Clouds and precipitation properties are measured byremote sensing techniques as well as by in-situ observa-tions Every two minutes two cameras with fish eyelenses take photos of the upper hemisphere from whichthe cloud type and cloud cover can be estimated

A 355 GHz scanning cloud radar (MIRA36-SMetek) is implemented in the KITcube to measure cloud

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 635

eschweizerbart_xxx

properties It is dopplerised and dual-polarised but trans-mits horizontally polarised radiation only and detects hor-izontally and vertically polarised reflections providing thelinear depolarisation ratio (LDR) of the hydrometeorsWith a 200 ns pulse duration and a 07 beam width it

reaches a very fine spatial resolution of 30 m for radialvelocity Additional data measured in the absence ofclouds are available to determine radial velocity whenclear-air returns are sufficient (eg caused by insects orother floating materialdebris dust and Bragg scattering

Table 1a Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement heights(m AGL)

Surface energy exchange and near-surface observations

Energy balance stations(Different

manufacturers)

2

Temperature 10 min (1 s) 3Humidity 10 min (1 s) 33-d wind 10 min (005 s) 4Humidity 10 min (005 s) 4

Temperature 10 min (005 s) 4Air pressure 10 min (1 s) 01Precipitation 10 min 1

Radiation temperatureof surface

10 min (1 s) 0

Solar and reflected 10 min (1 s) 3irradiance

Long wave incomingand outgoing radiation

10 min (1 s) 3

Sensible and latent 30 min (005 s) 4heat and momentum

fluxesSoil heat flux 10 min (1 s) 005Soil moisture 10 min 4 variable depths

Soil temperature 10 min 4 variable depths

Flux stations(Different

manufacturers)

7

Temperature 10 min (1 s) 1 2 4Humidity 10 min (1 s) 1 2 43-d wind 10 mm (005 s) 5

Temperature 10 min (005 s) 5Air pressure 10 min (1 s) 01

Sensible heat and momentum fluxes 30 min (005 s) 5

Scintillometer (Scintec) 1 Sensible heat flux 1 min 2ndash3

20-m tower(Different

manufacturers)

1

Temperature 10 min (1 s) 05 1 2 4 8 16 19Humidity 10 min (1 s) 05 1 2 4 8 16 19Wind speed 10 min (1 s) 05 1 2 4 8 16 19

Wind direction 10 min (1 s) 20Temperature 10 min (005 s) 20

Soil temperature 10 min (1 s) 002 004 008 016032 064

3-d wind 10 min (005 s) 20Sensible heat andmomentum fluxes

30 min (005 s) 20

Mobile towers (Young) 3

Temperature 10 min 2Humidity 10 min 23-d wind 003 s 4

Temperature 003 s 4Air pressure 10 min 1Precipitation 10 min 1

Sensible heat andmomentum fluxes

30 min (003 s) 4

636 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

from refractive-index fluctuations BANTA et al 2013)The cloud radar is able to scan at all azimuths fromvertical direction down to the 45 zenith angle Typicallya time resolution of 10 s is used in the vertical staremode Additional instrument specifications are given byTRAUMNER et al (2011)

To measure the cloud base height (CBH) the cloudradar is complemented by a ceilometer (CHM15k Jenop-tik) The ceilometer is a lidar providing uncalibratedbackscatter at a spatial resolution of 15 m From these

raw data CBHs intrusion depths and other propertiesof up to three different cloud layers can be determined

Volume filling information on precipitation is gath-ered by a mobile X-band radar (Meteor 50DX Gematro-nik) This radar is dopplerised and able to measurepolarisation properties of the hydrometeors It transmitsradiation polarised at 45 whereas the received signalsare split up into the horizontal and vertical polarisationthus providing polarisation properties such as differentialreflectivity (ZDR) and specific differential phase (KDP)

Table 1b Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments The measurement range column includes the minimal distance spatial resolution and maximum distance x stands for avariable resolution

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement range (m)

Mean and turbulent atmospheric conditionsRadiosonde system

(Graw)2 Temperature 2ndash8 (1 s) 0x15000

Humidity 2ndash8 (1 s) 0x15000Wind velocity and direction 2-8 (1 s) 0xl5000

Microwave radiometer(Radiometer Physics)

1 Temperature 10 s 050120012002005000500040010000

Humidity 10 s 0200200020004005000500080010000

IWV 10 s ndashLWP 10 s ndash

Infrared temperature 10 s ndash

GPS receiver (JAVAD) 1 IWV 15 min ndash

Sodar (Scintec) 1 Wind velocity and direction 30 min 30101000Vertical velocity variance 30 min 30101000

Wind lidar WindTracer(Lockheed Martin)

2 Radial velocity 01ndash1 s 375x12000Aerosol backscatter 01ndash1 s 375x12000

Wind velocity and direction scan dependent 75508475

Wind lidar Windcube(Leosphere)

1 Radial velocity 16 s 4020600Wind speed and direction 10 min (7 s) 4020600

Measurements of clouds and precipitation propertiesCloud camera (Mobotix) 2 Hemispheric photo 2 min ndash

Cloud radar (Metek) 1 Radial velocity 1ndash10 s 1503014460

Ceilometer (Jenoptik) 1 Cloud base heights 1 min 1501515000

X-band radar (Gematronik) 1 Precipitation 5 min 125250100000

Micro rain radar (Metek) 1 Precipitation Drop size distribution 1 min 1 min 10010032001001003200

Disdrometer (KIT) 1 Precipitation Drop size distribution 1 min 1 min 1 1

Disdrometer (Distromet) 1 Precipitation Drop size distribution 1 min 1 min 01 01

Rain gauge (EML) 2 Precipitation 1 min 1

height dependent

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 637

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

direction momentum and sensible heat fluxes The sta-tions deliver information on the spatial variability ofthe sensible heat fluxes Additionally a scintillometer(SLS20 Scintec) measures the sensible heat flux for spa-tial averages along the laser beam with an averaging timeof 1 minute The transmitting laser beam and receiverare typically separated by a distance of about 100 m to200 m

A 20 m high meteorological tower provides windspeed temperature and humidity profiles as well as winddirection and sensible heat and momentum fluxes at20 m above ground level (AGL) Soil temperatures aremeasured at six different depths

Battery-powered mobile towers that are deployedflexibly in the investigation area measure temperaturehumidity wind speed and wind direction air pressureprecipitation and sensible heat and momentum fluxes

22 Mean and turbulent atmosphericconditions

To obtain vertical profiles of temperature humidity windspeed and wind direction in the troposphere two radio-sonde systems (DFM-09 Graw) are available Dependingon height the 1 s raw data are averaged over 2 s (betweenground and 3 km AGL) 4 s (3 km and 10 km) and 8 s(gt 10 km) During IOPs radiosondes are typicallylaunched at 2- to 3-hour intervals The sounding data arealso used to derive significant levels and convection-related parameters such as the convective condensationlevel (CCL) lifting condensation layer (LCL) level of freeconvection (LFC) convective inhibition (CIN) and con-vective available potential energy (CAPE)

A ground-based scanning microwave radiometer(humidity and temperature profiler HATPRO) designedby Radiometer Physics measures the sky brightnesstemperature at 12 frequencies distributed within the22-30 GHz and 51-59 GHz bands With statistical retri-evals provided by the University of Cologne (LOHNERTand CREWELL 2003 CREWELL and LOHNERT 2003)these sky brightness temperatures are inverted to obtaintemperature and humidity profiles up to 10 km heightUsing additional boundary layer scans vertical tempera-ture profiles are obtained in 50 m steps close to the sur-face and in steps of 200 m to 400 m in the freetroposphere The system also provides integrated watervapour (IWV) and liquid water path (LWP) data alongthe line of sight 360 azimuth scans at fixed elevationangles carried out on a regular basis provide informationabout spatial humidity distribution The temperature ofcloud base is determined with an upwards directed infra-red thermometer

Additionally a JAVAD GPS receiver combined withsurface observations of temperature and pressure is usedto determine IWV with a temporal resolution of 15 minThe calculations are performed by GFZ Potsdam based

on algorithms described by DICK et al (2001) andGENDT et al (2004)

To determine the mean horizontal and vertical veloc-ity and vertical velocity variance a phased array sodar(MFAS Scintec) is used Operation frequencies in the2 kHz band and multi-frequency operation allow for aspatial resolution of 10 m and a maximum measurementrange of up to 1000 m AGL Backscatter data can beused for mixing height detection

One 2 lm and another 16 lm heterodyne wind lidarwith beam width of 75 mm at laser exit (WindTracerLockheed Martin) are part of KITcube to measure aerosolbackscatter and calculate radial velocity (TRAUMNER

et al 2011) Applying the vertical stare mode immedi-ately yields vertical velocity with a time resolution of1 s from about 375 m AGL to the top of the boundarylayer and partly above (depending on the aerosol concen-tration) The measurements are mainly restricted to thecloud-free atmosphere because wind lidars only partlypenetrate clouds The uncorrelated noise of velocitycould be estimated to be less than 02 m s1 for a highsignal to noise ratio (SNR) range Both lidars havetwo-axis scanners and yield profiles of the horizontalwind velocity via the velocity-azimuth display (VAD)technique (eg BROWNING and WEXLER 1968)Synchronised monitoring of the two lidars allow fordual-Doppler measurements (STAWIARSKI et al 2013)The CBL depth can be estimated using the profiles ofthe vertical velocity variance

In order to cover the range between the surface andthe lowest WindTracer measurement heights a third windlidar (Windcube 8 Leosphere) with a wavelength of154 lm is used which measures the wind profile from40 m AGL up to about 600 m AGL with averaging timesof 16 s to 10 min and a vertical range resolution of 20 mIn normal operation mode this small wind lidar has afixed elevation angle of 752 and an azimuth scannerBy means of a 4-point stop and stare plane position indi-cator (PPI) scan applying the VAD algorithm the windprofile is calculated Running a special operation modeby using a motorised tilted table allows for continuousvertical stare measurements for the direct detection ofvertical velocity In combination with a WindTracer thisleads to a full vertical coverage of vertical velocity fromthe ground up into the entrainment zone

23 Measurements of clouds andprecipitation properties

Clouds and precipitation properties are measured byremote sensing techniques as well as by in-situ observa-tions Every two minutes two cameras with fish eyelenses take photos of the upper hemisphere from whichthe cloud type and cloud cover can be estimated

A 355 GHz scanning cloud radar (MIRA36-SMetek) is implemented in the KITcube to measure cloud

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 635

eschweizerbart_xxx

properties It is dopplerised and dual-polarised but trans-mits horizontally polarised radiation only and detects hor-izontally and vertically polarised reflections providing thelinear depolarisation ratio (LDR) of the hydrometeorsWith a 200 ns pulse duration and a 07 beam width it

reaches a very fine spatial resolution of 30 m for radialvelocity Additional data measured in the absence ofclouds are available to determine radial velocity whenclear-air returns are sufficient (eg caused by insects orother floating materialdebris dust and Bragg scattering

Table 1a Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement heights(m AGL)

Surface energy exchange and near-surface observations

Energy balance stations(Different

manufacturers)

2

Temperature 10 min (1 s) 3Humidity 10 min (1 s) 33-d wind 10 min (005 s) 4Humidity 10 min (005 s) 4

Temperature 10 min (005 s) 4Air pressure 10 min (1 s) 01Precipitation 10 min 1

Radiation temperatureof surface

10 min (1 s) 0

Solar and reflected 10 min (1 s) 3irradiance

Long wave incomingand outgoing radiation

10 min (1 s) 3

Sensible and latent 30 min (005 s) 4heat and momentum

fluxesSoil heat flux 10 min (1 s) 005Soil moisture 10 min 4 variable depths

Soil temperature 10 min 4 variable depths

Flux stations(Different

manufacturers)

7

Temperature 10 min (1 s) 1 2 4Humidity 10 min (1 s) 1 2 43-d wind 10 mm (005 s) 5

Temperature 10 min (005 s) 5Air pressure 10 min (1 s) 01

Sensible heat and momentum fluxes 30 min (005 s) 5

Scintillometer (Scintec) 1 Sensible heat flux 1 min 2ndash3

20-m tower(Different

manufacturers)

1

Temperature 10 min (1 s) 05 1 2 4 8 16 19Humidity 10 min (1 s) 05 1 2 4 8 16 19Wind speed 10 min (1 s) 05 1 2 4 8 16 19

Wind direction 10 min (1 s) 20Temperature 10 min (005 s) 20

Soil temperature 10 min (1 s) 002 004 008 016032 064

3-d wind 10 min (005 s) 20Sensible heat andmomentum fluxes

30 min (005 s) 20

Mobile towers (Young) 3

Temperature 10 min 2Humidity 10 min 23-d wind 003 s 4

Temperature 003 s 4Air pressure 10 min 1Precipitation 10 min 1

Sensible heat andmomentum fluxes

30 min (003 s) 4

636 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

from refractive-index fluctuations BANTA et al 2013)The cloud radar is able to scan at all azimuths fromvertical direction down to the 45 zenith angle Typicallya time resolution of 10 s is used in the vertical staremode Additional instrument specifications are given byTRAUMNER et al (2011)

To measure the cloud base height (CBH) the cloudradar is complemented by a ceilometer (CHM15k Jenop-tik) The ceilometer is a lidar providing uncalibratedbackscatter at a spatial resolution of 15 m From these

raw data CBHs intrusion depths and other propertiesof up to three different cloud layers can be determined

Volume filling information on precipitation is gath-ered by a mobile X-band radar (Meteor 50DX Gematro-nik) This radar is dopplerised and able to measurepolarisation properties of the hydrometeors It transmitsradiation polarised at 45 whereas the received signalsare split up into the horizontal and vertical polarisationthus providing polarisation properties such as differentialreflectivity (ZDR) and specific differential phase (KDP)

Table 1b Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments The measurement range column includes the minimal distance spatial resolution and maximum distance x stands for avariable resolution

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement range (m)

Mean and turbulent atmospheric conditionsRadiosonde system

(Graw)2 Temperature 2ndash8 (1 s) 0x15000

Humidity 2ndash8 (1 s) 0x15000Wind velocity and direction 2-8 (1 s) 0xl5000

Microwave radiometer(Radiometer Physics)

1 Temperature 10 s 050120012002005000500040010000

Humidity 10 s 0200200020004005000500080010000

IWV 10 s ndashLWP 10 s ndash

Infrared temperature 10 s ndash

GPS receiver (JAVAD) 1 IWV 15 min ndash

Sodar (Scintec) 1 Wind velocity and direction 30 min 30101000Vertical velocity variance 30 min 30101000

Wind lidar WindTracer(Lockheed Martin)

2 Radial velocity 01ndash1 s 375x12000Aerosol backscatter 01ndash1 s 375x12000

Wind velocity and direction scan dependent 75508475

Wind lidar Windcube(Leosphere)

1 Radial velocity 16 s 4020600Wind speed and direction 10 min (7 s) 4020600

Measurements of clouds and precipitation propertiesCloud camera (Mobotix) 2 Hemispheric photo 2 min ndash

Cloud radar (Metek) 1 Radial velocity 1ndash10 s 1503014460

Ceilometer (Jenoptik) 1 Cloud base heights 1 min 1501515000

X-band radar (Gematronik) 1 Precipitation 5 min 125250100000

Micro rain radar (Metek) 1 Precipitation Drop size distribution 1 min 1 min 10010032001001003200

Disdrometer (KIT) 1 Precipitation Drop size distribution 1 min 1 min 1 1

Disdrometer (Distromet) 1 Precipitation Drop size distribution 1 min 1 min 01 01

Rain gauge (EML) 2 Precipitation 1 min 1

height dependent

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 637

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

properties It is dopplerised and dual-polarised but trans-mits horizontally polarised radiation only and detects hor-izontally and vertically polarised reflections providing thelinear depolarisation ratio (LDR) of the hydrometeorsWith a 200 ns pulse duration and a 07 beam width it

reaches a very fine spatial resolution of 30 m for radialvelocity Additional data measured in the absence ofclouds are available to determine radial velocity whenclear-air returns are sufficient (eg caused by insects orother floating materialdebris dust and Bragg scattering

Table 1a Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement heights(m AGL)

Surface energy exchange and near-surface observations

Energy balance stations(Different

manufacturers)

2

Temperature 10 min (1 s) 3Humidity 10 min (1 s) 33-d wind 10 min (005 s) 4Humidity 10 min (005 s) 4

Temperature 10 min (005 s) 4Air pressure 10 min (1 s) 01Precipitation 10 min 1

Radiation temperatureof surface

10 min (1 s) 0

Solar and reflected 10 min (1 s) 3irradiance

Long wave incomingand outgoing radiation

10 min (1 s) 3

Sensible and latent 30 min (005 s) 4heat and momentum

fluxesSoil heat flux 10 min (1 s) 005Soil moisture 10 min 4 variable depths

Soil temperature 10 min 4 variable depths

Flux stations(Different

manufacturers)

7

Temperature 10 min (1 s) 1 2 4Humidity 10 min (1 s) 1 2 43-d wind 10 mm (005 s) 5

Temperature 10 min (005 s) 5Air pressure 10 min (1 s) 01

Sensible heat and momentum fluxes 30 min (005 s) 5

Scintillometer (Scintec) 1 Sensible heat flux 1 min 2ndash3

20-m tower(Different

manufacturers)

1

Temperature 10 min (1 s) 05 1 2 4 8 16 19Humidity 10 min (1 s) 05 1 2 4 8 16 19Wind speed 10 min (1 s) 05 1 2 4 8 16 19

Wind direction 10 min (1 s) 20Temperature 10 min (005 s) 20

Soil temperature 10 min (1 s) 002 004 008 016032 064

3-d wind 10 min (005 s) 20Sensible heat andmomentum fluxes

30 min (005 s) 20

Mobile towers (Young) 3

Temperature 10 min 2Humidity 10 min 23-d wind 003 s 4

Temperature 003 s 4Air pressure 10 min 1Precipitation 10 min 1

Sensible heat andmomentum fluxes

30 min (003 s) 4

636 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

from refractive-index fluctuations BANTA et al 2013)The cloud radar is able to scan at all azimuths fromvertical direction down to the 45 zenith angle Typicallya time resolution of 10 s is used in the vertical staremode Additional instrument specifications are given byTRAUMNER et al (2011)

To measure the cloud base height (CBH) the cloudradar is complemented by a ceilometer (CHM15k Jenop-tik) The ceilometer is a lidar providing uncalibratedbackscatter at a spatial resolution of 15 m From these

raw data CBHs intrusion depths and other propertiesof up to three different cloud layers can be determined

Volume filling information on precipitation is gath-ered by a mobile X-band radar (Meteor 50DX Gematro-nik) This radar is dopplerised and able to measurepolarisation properties of the hydrometeors It transmitsradiation polarised at 45 whereas the received signalsare split up into the horizontal and vertical polarisationthus providing polarisation properties such as differentialreflectivity (ZDR) and specific differential phase (KDP)

Table 1b Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments The measurement range column includes the minimal distance spatial resolution and maximum distance x stands for avariable resolution

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement range (m)

Mean and turbulent atmospheric conditionsRadiosonde system

(Graw)2 Temperature 2ndash8 (1 s) 0x15000

Humidity 2ndash8 (1 s) 0x15000Wind velocity and direction 2-8 (1 s) 0xl5000

Microwave radiometer(Radiometer Physics)

1 Temperature 10 s 050120012002005000500040010000

Humidity 10 s 0200200020004005000500080010000

IWV 10 s ndashLWP 10 s ndash

Infrared temperature 10 s ndash

GPS receiver (JAVAD) 1 IWV 15 min ndash

Sodar (Scintec) 1 Wind velocity and direction 30 min 30101000Vertical velocity variance 30 min 30101000

Wind lidar WindTracer(Lockheed Martin)

2 Radial velocity 01ndash1 s 375x12000Aerosol backscatter 01ndash1 s 375x12000

Wind velocity and direction scan dependent 75508475

Wind lidar Windcube(Leosphere)

1 Radial velocity 16 s 4020600Wind speed and direction 10 min (7 s) 4020600

Measurements of clouds and precipitation propertiesCloud camera (Mobotix) 2 Hemispheric photo 2 min ndash

Cloud radar (Metek) 1 Radial velocity 1ndash10 s 1503014460

Ceilometer (Jenoptik) 1 Cloud base heights 1 min 1501515000

X-band radar (Gematronik) 1 Precipitation 5 min 125250100000

Micro rain radar (Metek) 1 Precipitation Drop size distribution 1 min 1 min 10010032001001003200

Disdrometer (KIT) 1 Precipitation Drop size distribution 1 min 1 min 1 1

Disdrometer (Distromet) 1 Precipitation Drop size distribution 1 min 1 min 01 01

Rain gauge (EML) 2 Precipitation 1 min 1

height dependent

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 637

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

from refractive-index fluctuations BANTA et al 2013)The cloud radar is able to scan at all azimuths fromvertical direction down to the 45 zenith angle Typicallya time resolution of 10 s is used in the vertical staremode Additional instrument specifications are given byTRAUMNER et al (2011)

To measure the cloud base height (CBH) the cloudradar is complemented by a ceilometer (CHM15k Jenop-tik) The ceilometer is a lidar providing uncalibratedbackscatter at a spatial resolution of 15 m From these

raw data CBHs intrusion depths and other propertiesof up to three different cloud layers can be determined

Volume filling information on precipitation is gath-ered by a mobile X-band radar (Meteor 50DX Gematro-nik) This radar is dopplerised and able to measurepolarisation properties of the hydrometeors It transmitsradiation polarised at 45 whereas the received signalsare split up into the horizontal and vertical polarisationthus providing polarisation properties such as differentialreflectivity (ZDR) and specific differential phase (KDP)

Table 1b Measurement systems included in KITcube and main derived parameters typical settings and measurement ranges ofinstruments The measurement range column includes the minimal distance spatial resolution and maximum distance x stands for avariable resolution

Device (Manufacturer) Quantity Derived parameters Temporal resolution ofderived parameters (raw data)

Measurement range (m)

Mean and turbulent atmospheric conditionsRadiosonde system

(Graw)2 Temperature 2ndash8 (1 s) 0x15000

Humidity 2ndash8 (1 s) 0x15000Wind velocity and direction 2-8 (1 s) 0xl5000

Microwave radiometer(Radiometer Physics)

1 Temperature 10 s 050120012002005000500040010000

Humidity 10 s 0200200020004005000500080010000

IWV 10 s ndashLWP 10 s ndash

Infrared temperature 10 s ndash

GPS receiver (JAVAD) 1 IWV 15 min ndash

Sodar (Scintec) 1 Wind velocity and direction 30 min 30101000Vertical velocity variance 30 min 30101000

Wind lidar WindTracer(Lockheed Martin)

2 Radial velocity 01ndash1 s 375x12000Aerosol backscatter 01ndash1 s 375x12000

Wind velocity and direction scan dependent 75508475

Wind lidar Windcube(Leosphere)

1 Radial velocity 16 s 4020600Wind speed and direction 10 min (7 s) 4020600

Measurements of clouds and precipitation propertiesCloud camera (Mobotix) 2 Hemispheric photo 2 min ndash

Cloud radar (Metek) 1 Radial velocity 1ndash10 s 1503014460

Ceilometer (Jenoptik) 1 Cloud base heights 1 min 1501515000

X-band radar (Gematronik) 1 Precipitation 5 min 125250100000

Micro rain radar (Metek) 1 Precipitation Drop size distribution 1 min 1 min 10010032001001003200

Disdrometer (KIT) 1 Precipitation Drop size distribution 1 min 1 min 1 1

Disdrometer (Distromet) 1 Precipitation Drop size distribution 1 min 1 min 01 01

Rain gauge (EML) 2 Precipitation 1 min 1

height dependent

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 637

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

The transmitted peak power is roughly 70 kW and thebeamwidth is 13 Pulse widths between 05 ls and2 ls lead to spatial resolutions down to 75 m The radarreaches the total upper half-space down to an elevation of-2 The horizontal range of the radar is about 100 km

AMicro rain radar (MRR2Metek) measures drop sizedistributions at 30 different heights with a layer spacing ofup to 200 m Mostly this device is operated with a layerspacing of 100 m and a temporal resolution of 1 min Thisradar (mean power 50 mW 3 beam width frequency24 GHz) measures the reflected power as a function ofthe Doppler shift Assuming negligible vertical air motionthe drop size distribution is determined in each layer fordroplets between 02 mm and 6 mm in diameter

An optical disdrometer (Parsivel KIT) and a mechan-ical disdrometer (Joss-Waldvogel Distromet) as well as atipping bucket rain gauge (EML) complete the equipmentused for in-situ precipitation measurements All thesedevices have a temporal resolution of 1 min

24 The KITcube control centre

To fully use the synergetic effects of the different instru-ments for process studies and to guarantee a high level ofdata quality and security a sophisticated instrument anddata management is implemented The control centre ofthe KITcube housed in a swap body (EN 284) collectsall the data from the different instruments either via opti-cal fibres WLAN or satellite transmission so that thedata are available in near real-time The control centreallows for a coordinated scan strategy to ensure simulta-neous monitoring of the different scanning instruments inthe same direction or volume eg combined range heightindicator (RHI) scans or PPI scans with the two lidarsmicrowave radiometer and cloud radar in or perpendicu-lar to the prevailing wind direction or observations in thevertical stare mode to investigate CBL features like up-and downdraughts and related turbulent characteristicslike profiles of the vertical velocity variance in the CBL

To ensure a high data availability online monitoringof the states of the instruments (alert indication) is imple-mented in KITcube which allows the operator in the con-trol centre to react immediately to instrument failuresThe collected data are archived on a data file serverand stored in an additional database which is connectedto a graphic user interface the so-called advanced dataextraction infrastructure (ADEI) ADEI does not onlyserve as a graphic interface but also allows for exportingdata in different file formats To guarantee a high level ofdata security the data are stored on a back-up server andmirrored on the main KITcube file server at KIT inKarlsruhe

3 KITcube deployment during HyMeX

The aims of KITcube during HyMeX were twofold ie(i) to allow for the investigation of convection initiated

or modified by the Corsican Island and (ii) to providethe upstream conditions for most of the intense precipita-tion cases affecting continental south-eastern France andnorthern and central Italy To reach these goals the fol-lowing configuration was selected two main deploymentsites were chosen namely the KITcube control centre atCorte (369 m MSL) in the Tavignano valley in the centreof Corsica and a site on the east coast (San Giuliano39 m MSL) (Fig 1) The instrumentation deployed atboth sites is listed in Table 2 The Corte site wasequipped to capture the atmospheric conditions and evo-lution of moist convection in the centre of the island andthe San Giuliano site was set up to cover coastal tomaritime atmospheric conditions

In order to determine the propagation speed of the seabreeze front in the Tavignano valley one of the main val-leys draining into the centre of Corsica three mobile tow-ers (Aleria Casaperta Pont Genois) were installedbetween the coast and Corte (Fig 1)

Five additional JAVAD GPS receivers includingmeteorological surface stations (wind temperaturehumidity air pressure) were distributed in the KITcubearea by GFZ Potsdam at Vergio Rusio La PortaBravone and San Giuliano (Fig 1) to enhance the spatialresolution of the existing GPS network operated by theInstitut Geographique National (IGN) and ACTIPLANcompany

Finally the Dornier 128 aircraft D-IBUF from TUBraunschweig which is equipped with in-situ sensorsfor standard meteorological variables and turbulence(CORSMEIER et al 2001) was stationed at Solenzara(Fig 1) It was operated on specific IOP days to measurewind temperature and humidity profiles in the west and

Table 2 Observation systems of KITcube deployed at the two mainsites on the Corsican Island

Corte4330 N 0917 E 369 m MSL

San Giuliano4227 N 0952 E 39 m MSL

Energy balance station Energy balance station2 flux stations 2 flux stations20-m towerScintillometerRadiosonde system Radiosonde systemMicrowave radiometer Microwave radiometerGPS receiver

SodarDoppler wind lidar (WindTracer)Doppler wind lidar (Windcube)Cloud radarCeilometerCloud camera Cloud cameraDisdrometersMicro rain radarRain gauge Rain gauge

X-band radar

638 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

east of Corsica conditions along east-west flight tracksacross the island and momentum heat and moisturefluxes over the sea

4 Observations by KITcube fromIOP 12

The IOP 12 of HyMeX (from 11 to 12 October) includeddays when Catalonia Corsica Tuscany and central Italywere affected by heavy thunderstorms with precipitation(DUCROCQ et al 2013) A period from IOP 12a(11 October) is analysed in the following

41 Synoptic conditions

In the morning of 11 October Corsica was ahead of asmooth trough (Fig 2a) The trough axis and the associ-ated surface low passed the island during the subsequentnight The passage of the inherent cold front was associ-ated with embedded thunderstorm cells and heavy precip-itation Ahead of the trough moist air was transportedeastwards with the large-scale westerly flow (Fig 2b)Around noontime convection indices derived from ra-diosounding data indicated low convective instabilityand hardly any convective inhibition At 500 hPa how-ever large-scale weak lifting was present (Fig 2a) andfavoured convection In the moist air mass convective

precipitating cells developed over Corsica between1000 UTC and 1800 UTC while more stratiform precip-itation prevailed over the sea (Fig 3) The evolution ofthe boundary layer conditions thermally-induced circula-tion systems and prefrontal precipitating cells over theisland on this day could be investigated in detail withKITcube

42 Conditions on the Corsican Island

Over Corsica spatially different atmospheric conditionsdeveloped in the morning At San Giuliano on the eastcoast the day began cloud-free so that global radiationat about 0830 UTC reached up to 500 W m-2 (Fig 4)Consequently due to the onset of the sensible and latentheat flux temperature and specific humidity increasedafter 0600 UTC and at 0830 UTC reached about 23 Cand 12 g kg1 respectively Caused by the land-sea tem-perature contrast the north-westerly nocturnal landbreeze was replaced by southerly to south-easterly windsat about 0800 UTC At Rusio a station on a north-eastwards exposed slope (Fig 1) weak slope winds setin at 0600 UTC already ie right after the temperatureincreased (Fig 4)

At Corte some passing upper-level clouds caused areduced radiation between 0600 and 0800 UTC already(Fig 5) Until about 0840 UTC the global radiationincreased and reached up to 450 W m2 Between sun-rise and 0840 UTC the temperature increased by about8 C and reached 18 C This resulted in the onset of amoderate south-easterly upvalley wind at 0800 UTCAll three thermally driven wind systems ndash sea breezeslope wind and upvalley wind ndash persisted at least untilthe first precipitation occurred after 1000 UTC (Fig 4and 5)

From radiosondes and remote sensing data the depthsof the different wind systems were deduced At SanGiuliano a south-easterly wind extended up to about500 m AGL and a more southerly wind prevailed aboveup to a height of 1300 m AGL at 0900 UTC (Fig 6a)Above this layer a large-scale westerly flow was foundAt around 1400 UTC when a precipitation systemoccurred above the north-eastern part of Corsica(Fig 3) the westerly flow temporary penetrated downto the ground

At Corte three wind layers were present at 0700 UTCndash a weak downvalley wind of about 200 m depth anelevated easterly flow between 300 m and 700 m AGLand the westerly large-scale flow above (Fig 6b) The ori-gin of the elevated easterly flow could not be explained bythe KITcube observations but was probably part of a low-level large-scale flow around the Corsican Island as wassuggested by the analysis of the synoptic data After theonset of the upvalley wind the coupled upvalley windand the elevated easterly flow led to a south-easterly flowof 1000 m depth These two layers persisted until

Figure 2 (a) COSMO-EU analysis of 500 hPa geopotential heightin gpdm (isolines) and vertical velocity (colour-coded) and(b) specific humidity (colour-coded) and horizontal wind vectorsat 700 hPa level on 11 October at 1200 UTC

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 639

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

about 1400 UTC when associated with convective show-ers westerly wind replaced the upvalley wind whilethe elevated easterly flow below the clouds still remained

(Fig 5 and 6b) The different wind layers were also foundin the wind lidar data when applying the VAD algorithmfor horizontal wind field calculation (not shown)

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1200 UTC

dBZ0 20 40 60

San GiulianoCorte

Longitude (deg)

Longitude (deg)

Longitude (deg)

Latit

ude

(deg)

Latit

ude

(deg)

Latit

ude

(deg)

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1400 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

Hei

ght (

km M

SL)

0

5

10

Height (km MSL)5 10

1700 UTC

dBZ0 20 40 60

San GiulianoCorte

9 95 10 105

415

42

425

43

1200 UTC

1700 UTC

1400 UTC

Figure 3 (left) Cloud composites (MeteoFrance) and (right) maximum constant altitude plan position indicator (MAX-CAPPI) of radarreflectivity on 11 October at 1200 UTC 1400 UTC and 1700 UTC (top to bottom)

640 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

Fig 7 presents the combination of cloud radar radio-sonde ceilometer and wind lidar data for the Corte siteHigh specific humidity values (gt 8 g kg1) were found inthe layer of the upvalley wind and elevated easterly flow(Fig 7a) The spatial IWV distribution as measured bythe scanning microwave radiometer indicated that thehigh humidity advected by these wind systems fromthe sea was transported further upwards by the slopewinds to the mountain crests west of Corte (not shown)In the morning until about 1100 UTC a significanthumidity increase was found in a layer between 1 kmand 3 km AGL (Fig 7a) which was also present in theCOSMO-EU analysis (Fig 2) This humidity increasecaused by large-scale advection contributed mainly tothe strong increase of IWV in the morning (Fig 8)The time series of IWV show that a first IWV maximumof more than 34 kg m2 occurred on the western coast(Osani) at 1000 UTC while in the centre (Corte) andon the eastern coast (Santa-Lucia-di-Moriani) a first peakwas reached about one hour later

When the air mass with higher humidity arrived partlyprecipitating clouds formed The cloud radar reflectivitydata (Fig7a) show that the convective clouds extendedup to about 5 km to 6 km AGL Between 0900 UTCand 1100 UTC the CBH decreased from about 5 km to17 km AGL The CBH was determined from ceilometerdata (Fig 7b) and aerosol backscatter data of the windlidar (detected using a threshold in aerosol backscatterof ndash72 dB) (Fig 7c) The ceilometer data show severalabrupt decreases in CBH eg at 1000 UTC 1120 UTCand 1300 UTC which are not found in the CBH data fromthe wind lidar Comparison with cloud radar reflectivitydata (Fig 7a) shows that this is due to the onset of rainand not caused by a change in CBH

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)

0

200

400

600

Air t

empe

ratu

re

(deg C)

10

15

20

25

6

8

10

12

14

Win

d sp

eed

(m s

minus1)

0

1

2

3

Win

d di

rect

ion

(deg)

0

90

180

270

360

11 Oct (UTC)

Prec

ipita

tion

(mm

hminus1

)

0300 0600 0900 1200 1500 1800

0

2

4

6

San Giuliano

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San GiulianoRusio

San Giuliano

Figure 4 Global radiation and precipitation at San Giuliano andtemperature specific humidity horizontal wind speed and winddirection at San Giuliano and Rusio on 11 October

0

200

400

600

10

15

20

25

6

8

10

12

14

0

1

2

3

0

90

180

270

360

0

5

10

15

11 Oct (UTC)0300 0600 0900 1200 1500 1800

0

2000

4000

6000

ParsivelJossminusWaldvogel

LCL radiosondeLCL surfaceCCL radiosondeCBH lidar

Glo

bal r

adia

tion

(W m

minus2)

Spec

ific

hum

idity

(g

kgminus1

)Ai

r tem

pera

ture

(deg C

)W

ind

spee

d (m

sminus1

)W

ind

dire

ctio

n (deg

)Pr

ecip

itatio

n (m

m h

minus1)

Hei

ght

(m M

SL)

Figure 5 Global radiation temperature specific humidity hori-zontal wind speed and wind direction precipitation LCL CCLand CBH at Corte on 11 October

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 641

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

Comparison of the observed CBH with CCL and LCLcalculated from surface and radiosounding data at Corte(Fig 5) shows that the observed CBH was much higher

than the LCL and CCL On this day the temperature inthe upvalley wind layer and elevated easterly flow wasstably stratified throughout the day (Fig 7c) so that theconvection temperature was not reached at the groundThis indicates that the clouds were not locally initiatedbut orographically triggered over the mountain crestswest of Corte and then transported eastwards with thelarge-scale westerly flow This also corresponds to therain radar measurements (Fig 3) during four periodswith precipitation between about 1000 UTC and 1800UTC the precipitating cells in the radar image for thefirst time appeared over the Corsican mountains

The vertical velocity measured by the cloud radarshowed some weak dry convective activity before 0930UTC (Fig 7b) During cloudy periods precipitation couldbe identified apart from high reflectivity due to high valuesof negative vertical velocity of falling raindrops Theabrupt change in the vertical velocity from gt 1 m s1

to about 3 m s1 at a height of about 3 km AGL marksthe melting layer and separates falling snow above fromraindrops below as obvious from the comparison withthe vertical temperature distribution (Fig 7b) In the layerswith rain the vertical velocity of the ambient air could notbe calculated The cloud radar data showed that upwardmotion was encountered in a few areas in the clouds onlyndash mainly at the cloud tops

The vertical velocities w from the two wind lidars(WindTracer Windcube) are combined in Fig 7c Asalready seen in the cloud radar data weak dry convectionwas present in the morning before the first precipitationoccurred Below the precipitating clouds the manufac-turerrsquos built-in dominant-peak algorithm calculated nega-tive vertical velocity To determine the velocity a windlidar measures the Doppler shift of the backscattered lightinduced by the particles in the atmosphere The particlesare considered to have no air speed but to drift with thewind only Typically this results in a Gaussian-shapedpeak in the Doppler spectrum with a maximum at themost prevalent velocity (which is D2-weighted) and awidth characterising the turbulent condition During spe-cial atmospheric events like rain and snow fall or in

RadiosondeSodar

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

10 m sminus1

10 m sminus1

Radiosonde

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

1000

2000

3000

(a)

(b)

Figure 6 (a) Horizontal wind vectors derived from radiosondes(black) and sodar (grey) at San Giuliano and (b) horizontal windvectors derived from radiosondes (black) at Corte on 11 October

22

4

46

6

8

8

10

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

dB minus60

minus40

minus20

0

20

minus13minus9minus5minus1371115

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

296300304

308312316

11 Oct (UTC)

Hei

ght (

m A

GL)

0300 0600 0900 1200 1500 18000

2000

4000

6000

minus2

0

2CBH

m6 mmminus310 m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 7 (a) Specific humidity (isolines) cloud radar reflectivity(colour-coded) and horizontal wind vectors (arrows) (b) temper-ature (isolines) vertical velocity (colour-coded) CBH (dotted) and0 C level (thick solid line) and (c) potential temperature (isolines)vertical velocity (colour-coded) and CBH (dotted) at Corte on 11October Vertical velocity is calculated from cloud radar in (b) andwind lidar data in (c) CBH is determined from ceilometer in (b) andwind lidar data in (c) Temperature specific humidity and windvectors are from radiosonde data

11 Oct (UTC)

IWV

(kg

mminus2

)

0300 0600 0900 1200 1500 1800

20

25

30

35

40

West coast (Osani)Centre (Corte)East coast (SantaminusLuciaminusdiminusMoriani)

Figure 8 Time series of IWV from operational GPS stations on thewest coast (Osani) in the centre (Corte) and on the east coast(Santa-Lucia-di-Moriani) of Corsica on 11 October

642 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

clouds more than one velocity range may exist In thesecases more than the dominant peak in the Doppler spec-tra may occur as is illustrated by the example in Fig 9To account for this situation a specially designed auto-matic algorithm was developed to distinguish betweentwo potential peaks A Gaussian-shaped double peak

I weth THORN frac14 A1exp w w1

r1

2

thorn A2exp w w2

r2

2

was fitted to the vertical velocity spectrum between1105 m s1 and 126 m s1 using a least-square-method The parameter A12 are the values of the intensi-tiesrsquo peaks w12 are the positions of the centers of thepeaks and the standard deviations r12 control the widthsof the two bell-shaped curves This double peak-fit algo-rithm is quite sensitive to the used initial values For thisreason a routine calculates different initial values If theexpected values of the two peaks are separated by lessthan 2 m s1 or the width of at least one peak is narrowerthan 1 m s1 the algorithm uses a single-peak procedureagain to determine only one velocity The algorithm repro-duces 985 of the velocities determined by the manufac-turerrsquos built-in dominant peak-finding algorithm In aboutone third of all cases with an SNR ratio of more than-6 dB a second considerable peak was determined bythe double-peak algorithm

On 11 October two intensity peaks occurred in theclouds in the aerosol layer as well as in the aerosol-freelayer in between Fig 10 presents the results for a shortertime period around 1000 UTC of the whole precipitationevent (Fig 7) The vertical velocity derived from thedominant peak in the Doppler spectrum is shown in

Fig 10a In the clouds (CBH 3000 m AGL) thetwo peaks probably resulted from small cloud dropletsor snow (Fig 10b) and bigger already falling raindrops(Fig 10c) In the aerosol layer (lt 1000 m AGL) onepeak resulted from aerosols and thus represents the ver-tical velocity of the ambient air (Fig 10b) whereas thesecond was from the falling raindrops (Fig 10c) In thelayer between the aerosol and the cloud layer one peakwas caused by falling raindrops (Fig 10c) It is more dif-ficult to interpret the lower velocities (Fig 10b) theresults need further investigations The skill of the doublepeak algorithm is particularly evident in the aerosol layerbetween 0955 UTC and 1000 UTC and 1005 UTC and1015 UTC where the velocity of the falling raindropsand the vertical velocity of the ambient air could be sep-arated while the vertical velocity of the raindrops wasaround -3 m s1 the velocity of the ambient air wasaround 0 m s1

5 Discussion and conclusions

The KITcube was designed and constructed to investigatemeso-c processes related to convection initiation cloud

minus30 minus20 minus10 0 10 20 30

0

01

02

03

04

05

06

Doppler velocity (m sminus1)

Inte

nsity

(arb

itrar

y un

its)

minus579 m sminus1

minus112 m sminus1

Figure 9 Doppler spectrum with double peak at the 3rd lidar rangegate (466 m AGL) at Corte at 1005 UTC on 11 October Solid anddashed grey lines mark the Gaussian double peaks derived from theDoppler vertical velocity distribution (black dots) For details seesection 42

minus0 0

4

-4 -4

-4 -4

-4 -4

4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

minus0 0

4 4

8 8

12 12 12

16 16

11 Oct (UTC)

Hei

ght (

m A

GL)

0955 1000 1005 1010 10150

2000

4000

minus4

minus2

0

2

4

CBH

CBH

CBH

m sminus1

m sminus1

m sminus1

(a)

(b)

(c)

Figure 10 Rain episode as observed by Doppler wind lidar(a) vertical velocity derived from the dominant peak in the Dopplerspectrum (b) vertical velocity derived from the peak classified asvertical velocity of the ambient air (heights lt1000 m AGL)raindrops of smaller size (1000 m AGL lt heights lt 3000 m AGL)and droplets or snow (heights gt 3000 m AGL) (c) vertical velocityderived from the peak classified as velocity of the falling raindropsat Corte on 11 October Solid lines are temperature isolines fromradiosoundings CBH (dotted) is determined from wind lidar dataFor details see section 42

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 643

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

formation and precipitation The system was applied firstduring the HyMeX field campaign in autumn 2012 Todemonstrate KITcubersquos capability a period from theintensive operation period 12a (11 October) was selectedhere This period included orographically induced cloudformation and precipitation The synergetic use ofKITcubersquos different in-situ and remote sensing systemsprovides a comprehensive overview of the mesoscalemeteorological phenomena on that day It enablesinsight into processes from varying perspectives andallows for the verification of parameters derived from dif-ferent observation systems Main results from HyMeXIOP 12a are

The vertically high-resolved observations provided anoverview of the complex wind field consisting of ther-mally-induced circulations and large-scale flow Themeasurements also revealed that the deep south-easterlyflow at Corte was a combination of the upvalley windand an elevated easterly flow (Fig 6)

The combination of radiosonde microwave radiome-ter and GPS data proved that the strong increase of IWVwas mainly caused by large-scale humidity transportfrom the west in a layer between 1 km and 3 kmAGL The cloud radar observations showed that the pre-cipitating clouds formed in this moist westerly flow

As regards CBH detection a comparison of ceilome-ter and wind lidar data with infrared temperaturesmeasured by the microwave radiometer and cloud radarreflectivity data proved that the CBH measured by theceilometer was temporarily too low The ceilometerdetected the arrival of shower lines as CBH (Fig 7b)and was expectedly unable to detect the correct CBH dur-ing precipitation events

By comparing CCL and LCL values estimated fromsurface and radiosounding data with CBH it was con-cluded that the clouds at Corte were not locally initiatedbut orographically triggered over the mountain crests inthe west and then advected to the east This was con-firmed by rain radar data which showed that the precipi-tating cells for the first time appeared over the Corsicanmountains on the upwind side of Corte

Comparison of the vertical measurement ranges of thecloud radar and wind lidars under cloud-free conditionsshowed that the heights were quite different eg afterthe onset of the upvalley flow but before the onset ofthe first precipitation (0800 UTC ndash 0930 UTC) The aer-osol layer of about 1000 m depth which was found in thelidar measurements was clearly bound to the height ofthe coupled moist upvalley and elevated easterly flow(Fig 7) Backscatter from cloud radar however reachedup to about 1600 m AGL ie into the large-scale wes-terly flow Because of different wavelengths the windlidar and cloud radar saw different particles in cloud-freeatmospheres and thus provided data in various layersAdditionally the cloud radar requires a smaller concen-tration of scatterers to provide sufficient backscatter data

Detailed analysis of the vertical velocity spectra fromthe wind lidar partly allowed for the determination of ver-

tical velocities of different scatterers In clouds these areprobably caused by co-existing falling raindrops andsnow or cloud droplets In aerosol layers during rainevents the vertical velocity of falling raindrops couldbe distinguished from the vertical velocity of the ambientair In the layer between the aerosol and cloud layer thetwo detected vertical velocities might be produced by twoclasses of falling raindrops During the precipitationevent around 1000 UTC the raindrops sizes ranged from04 mm to 3 mm (Fig 11) However it is not clear whichthreshold of drop size separates the two vertical velocityranges Vertical velocities of the falling raindrops calcu-lated with the cloud radar resulted in higher values (com-pare Fig 7b and c) which is due to the strongerweighting of raindrops with greater diameters (D2-weighted)

Apart from the mesoscale processes in Corsica theKITcube data provided valuable information for theoverall HyMeX goals ie the analysis of high precipita-tion events over southern France and northern and centralItaly An additional benefit for the investigation of meso-scale processes results from supplementary Dornier 128aircraft measurements during HyMeX (DUCROCQ et al2013) Several HyMeX dry and moist convection epi-sodes are presently being investigated (eg ADLER andKALTHOFF 2013 KALTHOFF and ADLER 2013) Theyconfirm KITcubersquos capability of analysing CBL-lowertroposphere interactions on multiple scales

Acknowledgments

This work is a contribution to the HyMeX programmeWe would like to thank Veronique DUCROCQ fromMeteoFrance and Evelyne RICHARD from the Universityof ToulouseCNRS for their support in performing themeasurements of KITcube on Corsica during HyMeXAntoine PIERI from the fire brigade at CorteUniversityof Corsica and Olivier PAILLY from INRA in San Giuli-ano for their help and for hosting us during the nearlythree-monthsrsquo campaign We are also grateful to thewhole IMK team for their commitment to deployingthe KITcube The authors thank IGN and ACTIPLANfor operating the permanent GPS stations in CorteSanta-Lucia-di-Moriani and Osani as well as IGNSGN for processing these data operationally and

11 Oct (UTC)

Dro

p si

ze (m

m)

0300 0600 0900 1200 1500 18000

1

2

3

mminus3 mmminus1

0

1000

2000

Figure 11 Drop size distribution measured with the Parsiveldisdrometer at Corte on 11 October

644 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

Olivier BOCK (LAREG) for the screening of GPS dataand conversion of ZTD into IWV using the methodologydescribed in BOCK et al (2007) This work was also sup-ported by the Atmospheric Observatory CORSICA(httpwwwobs-mipfr) Finally we acknowledge provi-sion of cloud composites by MeteoFrance

References

ADLER B N KALTHOFF 2013 Boundary-layer character-istics over complex terrain observed by remote-sensingsystems during HyMeX 32nd ICAM 2013 ndash Book ofabstracts 5 Kranjska Gora 3-7 June 2013

BANTA RM 1990 The role of mountain flows in makingclouds In BLUMEN W Atmospheric processes overcomplex terrain ndash Meteor Monogr 45 229ndash283

BANTA RM CM SHUN DC LAW WO BROWN RFREINKING RM HARDESTY CJ SENFF MJ POST LSDARBY 2013 Observational techniques Sampling themountain atmosphere ndash In Mountain weather researchand forecasting Recent progress and current challengesCHOW FK SF DE WEKKER BJ SNYDER (Eds)ndash Springer Atmospheric Sciences Springer 409ndash530DOI 101007978-94-007-4098-3_8

BEHRENDT A S PAL F AOSHIMA M BENDER A BLYTHU CORSMEIER J CUESTA G DICK M DORNINGER CFLAMANT P DI GIROLAMO T GORGAS Y HUANG NKALTHOFF S KHODAYAR H MANNSTEIN K TRAUMNERA WIESER V WULFMEYER 2011 Observation of convec-tion initiation processes with a suite of state-of-the-artresearch instruments during COPS IOP8b ndash Quart J RoyMeteor Soc 37 81ndash100

BOCK O M-N BOUIN A WALPERSDORF J-P LAFORE SJANICOT F GUICHARD A AGUSTI-PANAREDA 2007Comparison of ground-based GPS precipitable watervapour to independent observations and numerical weatherprediction model reanalyses over Africa ndash Quart J RoyMeteor Soc 133 2011ndash2027 doi 101002qj185

BROWNING KA R WEXLER 1968 The determination ofkinematic properties of a wind field using Doppler radarndash J Appl Meteor 7 105ndash113

BROWNING K A BLYTH P CLARK U CORSMEIERC MORCRETTE J AGNEW D BAMBER C BARTHLOTTL BENNETT K BESWICK M BITTER K BOZIERB BROOKS C COLLIER C COOK F DAVIES B DENYT FEUERLE R FORBES C GAFFARD M GRAYR HANKERS T HEWISON N KALTHOFF S KHODAYARM KOHLER C KOTTMEIER S KRAUT M KUNZ DLADD J LENFANT J MARSHAM J MCGREGOR JNICOL E NORTON D PARKER F PERRY M RAMATSCHIH RICKETTS N ROBERTS A RUSSELL H SCHULZ ESLACKGVAUGHAN JWAIGHT RWATSONAWEBBAWIESER 2007 The convective storm initiation projectndash Bull Amer Meteor Soc 88 1939ndash1955

BRYAN GH JC WYNGAARD JM FRITSCH 2003Resolution requirements for the simulation of deep moistconvection ndash Mon Wea Rev 131 2394ndash2416

BUZZI A L FOSCHINI 2000 Mesoscale meteorologicalfeatures associated with heavy precipitation in the South-ern alpine region ndash Meteor Atmos Phys 72 131ndash146

CORSMEIER U R HANKERS A WIESER 2001 Airborneturbulence measurements in the lower troposphere onboardthe research aircraft Dornier 128ndash6 D-IBUF ndash MeteorolZ 10 315ndash329

CORSMEIER U N KALTHOFF C BARTHLOTT F AOSHIMAA BEHRENDT P DIGIROLAMO M DORNINGER JHANDWERKER C KOTTMEIER H MAHLKE S MOBBSE NORTON J WICKERT V WULFMEYER 2011 Processesdriving deep convection over complex terrain A multi-scale analysis of observations from COPS-IOP9c ndash QuartJ Roy Meteor Soc 137 137ndash155

CREWELL S U LOHNERT 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartII Sensor accuracy and synergy ndash Radio Sci 38 8042doi 1010292002RS002634

DICK G G GENDT C REIGBER 2001 First experiencewith near real-time water vapor estimation in a GermanGPS network ndash J Atmos Sol-Terr Phys 63 1295ndash1304

DOSWELL C AC RAMIS R ROMERO S ALONSO 1998A diagnostic study of three heavy precipitation episodes inthe western Mediterranean region ndash Wea Forecast 13102ndash124

DROBINSKI P V DUCROCQ P ALPERT E ANAGNOSTOUK BERANGER M BORGA I BRAUD A CHANZY SDAVOLIO G DELRIEU C ESTOURNEL N FILALIBOUBRAHMI J FONT V GRUBISIC S GUALDI VHOMAR B IVANCAN-PICEK C KOTTMEIER V KOTRONIK LAGOUVARDOS P LIONELLO MC LLASAT WLUDWIG C LUTOFF A MARIOTTI E RICHARD RROMERO R ROTUNNO O ROUSSOT I RUIN S SOMOTI TAUPIER-LETAGE J TINTORE R UIJLENHOET HWERNLI 2014 HyMeX a 10-year multidisciplinaryprogram on the Mediterranean water cycle ndash Bull AmerMeteor Soc doi 101175BAMS-D-12-002421

DUCROCQ V O NUISSIER D RICARD C LEBEAUPIN SANQUETIN 2008 A numerical study of three catastrophicprecipitating events over southern France II Mesoscaletriggering and stationarity factors ndash Quart J R MeteorSoc 134 131ndash145

DUCROCQ V I BRAUD S DAVOLIO R FERRETTI CFLAMANT A JANSA N KALTHOFF E RICHARD ITAUPIER-LETAGE P-A AYRAL S BELAMARI A BERNEM BORGA B BOUDEVILLAIN O BOCK J-L BOICHARDM-N BOUIN O BOUSQUET C BOUVIER J CHIGGIATOD CIMINI U CORSMEIER L COPPOLA P COCQUEREZEDEFER JDELANOE PDIGIROLAMOADOERENBECHERP DROBINSKI Y DUFOURNET N FOURRIE JJ GOURLEYL LABATUT D LAMBERT J LE COZ FS MARZANOGMOLINIE AMONTANI GNORDMNURET KRAMAGEB RISON O ROUSSOT F SAID A SCHWARZENBOECKP TESTOR J VAN BAELEN B VINCENDON M ARANJ TAMAYO submitted HyMeX-SOP1 the field campaigndedicated to heavy precipitation and flash flooding in thenorthwestern Mediterranean ndash Bull Amer Meteor Socdoi 101175BAMS-D-12-002441

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 645

eschweizerbart_xxx

GENDT G G DICK C REIGBER M TOMASSINI Y LIU MRAMATSCHI 2004 Near real time GPS water vapormonitoring for numerical weather prediction in Germanyndash J Meteor Soc Japan 82 361ndash370

HOMAR V C RAMIS R ROMERO S ALONSO JAGARCIA-MOYA M ALARCON 1999 A case of convectiondevelopment over the Western Mediterranean Sea a studythrough numerical simulations ndash Meteor Atmos Sci 71169ndash188

KALTHOFF N B ADLER 2013 Dependency of convectionon environmental conditions over complex terrain duringHyMeX ndash 32nd ICAM 2013 Book of abstracts 46Kranjska Gora 3ndash7 June 2013

KALTHOFF N M FIEBIG-WITTMAACK C MEIszligNER MKOHLER M URIARTE I BISCHOFF-GAUszlig E GONZALES2006 The energy balance evapo-transpiration and noc-turnal dew position of an arid valley in the Andes ndash J AridEnviron 65 420ndash443 doi 101016jjaridenv200508013

KALTHOFF N B ADLER C BARTHLOTT U CORSMEIERS MOBBS S CREWELL K TRAUMNER C KOTTMEIERA WIESER V SMITH P DI GIROLAMO 2009 The impactof convergence zones on the initiation of deep convectionA case study from COPS ndash Atmos Res 93 680ndash694

KALTHOFF N K TRAUMNER B ADLER S SPATH ABEHRENDT A WIESER J HANDWERKER F MADONNAV WULFMEYER 2013 Dry and moist convection in theboundary layer over the Black Forest ndash a combinedanalysis of in situ and remote sensing data ndash Meteorol Z22 445ndash461 doi 1011270941-294820130417

KHODAYAR S N KALTHOFF J WICKERT U CORSMEIERCJ MORCRETTE C KOTTMEIER 2010 The increase ofspatial data resolution for the detection of the initiation ofconvection A case study from CSIP ndash Meteorol Z 19179ndash198

KHODAYAR S N KALTHOFF J WICKERT C KOTTMEIERM DORNINGER 2013 High-resolution representation ofthe mechanisms responsible for the initiation of isolatedthunderstorms over flat and complex terrains Analysis ofCSIP and COPS cases ndash Meteor Atmos Phys 119 109ndash124

KOTTMEIER C P PALACIO-SESE N KALTHOFF U CORS-

MEIER F FIEDLER 2000 Sea breezes and coastal jets insoutheastern Spain ndash Int J Climatol 20 1791ndash1808

KOTTMEIERCNKALTHOFFCBARTHLOTTUCORSMEIERJ VAN BAELEN A BEHRENDT R BEHRENDT A BLYTH RCOULTER S CREWELL P DI GIROLAMO M DORNINGERC FLAMANT T FOKENM HAGEN C HAUCK H HOLLERH KONOWM KUNZ H MAHLKE S MOBBS E RICHARDR STEINACKER T WECKWERTH A WIESER 2008 Mech-anisms initiatingdeep convectionover complex terrain duringCOPS ndashMeteorol Z 17 931ndash948 doi 1011270941-294820080348

LAMBERT D M MALLET V DUCROCQ F DULAC FGHEUSI N KALTHOFF 2011 CORSiCA a Mediterraneanatmospheric and oceanographic observatory in Corsicawithin the framework of HyMeX and ChArMEx ndash AdvanGeosci 26 125ndash131

LOHNERT U S CREWELL 2003 Accuracy of cloud liquidwater path from ground-based microwave radiometry PartI Dependency on cloud model statistics ndash Radio Sci 388041 doi 1010292002RS002654

MAUDER M T FOKEN 2011 Documentation and instruc-tion manual of the eddy covariance software package TK3ndash Work Report University of Bayreuth Dept of Micro-meteorology 46 University of Bayreuth Bayreuth ISSN1614-8916 58 pp

MILLER MA A SLINGO 2007 The ARM Mobile Facilityand its first international deployment Measuring radiativeflux divergence in West Africa ndash Bull Amer Meteor Soc88 1229ndash1244

QIAN J-H 2007 Why precipitation is mostly concentratedover islands in the maritime continent ndash J Atmos Sci 651428ndash1441

ROTUNNO R R FERRETTI 2001 Mechanisms of intenseAlpine rainfall ndash J Atmos Sci 58 1732ndash1749

STAWIARSKI C K TRAUMNER C KNIGGE R CALHOUN2013 Scopes and challenges of dual-Doppler lidar windmeasurements ndash an error analysis ndash J Atmos OceanTechnol 30 2044ndash2062 doi 101175JTECH-D-12-002441

TAYLOR C D PARKER N KALTHOFF MA GAERTNER NPHILIPPON S BASTIN PP HARRIS A BOONE FGUICHARD A AGUSTI-PANAREDA M BALDI PCERLINI L DESCROIX H DOUVILLE C FLAMANTJ-Y GRANDPEIX J POLCHER 2011 New perspectives onland-atmosphere feedbacks from the African monsoonmultidisciplinary analysis ndash Atmos Sci Lett 12 38ndash44

TRAUMNER K J HANDWERKER J GRENZHAUSER AWIESER 2010 A synergy approach to estimate propertiesof raindrop size distributions using a Doppler lidar andcloud radar ndash J Atmos Ocean Technol 27 1095ndash1100

TRAUMNER K C KOTTMEIER U CORSMEIER A WIESER2011 Convective boundary-layer entrainment shortreview and progress using Doppler lidar ndash Bound-LayerMeteor 141 369ndash391

WECKWERTH TM 2000 The effect of small-scale moisturevariability on thunderstorm initiation ndash Mon Wea Rev128 4017ndash4030

WECKWERTH TM DB PARSONS SE KOCH JAMOORE MA LEMONE BB DEMOZ C FLAMANT BGEERTS J WANG WF FELTZ 2004 An overview of theinternational H2O project (IHOP_2002) and some pre-liminary highlights ndash Bull Amer Meteor Soc 85253ndash277

WILSON JW RD ROBERTS 2006 Summary of convectivestorm initiation and evolution during IHOP Observationaland modeling perspective ndash Mon Wea Rev 134 23ndash47

WILSON JW RE CARBONE JD TUTTLE TD KEENAN2001 Tropical island convection in the absence ofsignificant topography Part II Nowcasting storm evolu-tion ndash Mon Wea Rev 129 1637ndash1655

WULFMEYER V A BEHRENDT C KOTTMEIER UCORSMEIER C BARTHLOTT GC CRAIG M HAGEND ALTHAUSEN F AOSHIMA M ARPAGAUS H-SBAUER L BENNETT A BLYTH C BRANDAU C

646 N Kalthoff et al KITcube ndash a mobile observation platform Meteorol Z 22 2013

eschweizerbart_xxx

CHAMPOLLION S CREWELL G DICK P DI GIROLAMOM DORNINGER Y DUFOURNET R EIGENMANN RENGELMANN C FLAMANT T FOKEN T GORGAS MGRZESCHIK J HANDWERKER C HAUCK H HOLLER WJUNKERMANN N KALTHOFF C KIEMLE S KLINK MKONIG L KRAUSS CN LONG F MADONNA S MOBBSB NEININGER S PAL G PETERS G PIGEON E RICHARD

MW ROTACH H RUSSCHENBERG T SCHWITALLA VSMITH R STEINACKER J TRENTMANN DD TURNER JVAN BAELEN S VOGT H VOLKERT T WECKWERTH HWERNLI A WIESER M WIRTH 2011 The Convective andOrographically-induced Precipitation Study (COPS) thescientific strategy the field phase and research highlightsndash Quart J Roy Meteor Soc 137 3ndash30 doi 101002qj752

Meteorol Z 22 2013 N Kalthoff et al KITcube ndash a mobile observation platform 647