Introduction to SARWeather
Ólafur Rögnvaldsson – IMR/[email protected]
Logi Ragnarsson – IMR/[email protected]
Cloud Computing Education series
Acknowledgement• SARWeather is a joint research project led by IMR/Belgingur, in
collaboration with NOAA/ESRL, the University of Bergen, and the private companies GreenQloud and DataMarket. To ensure maximum usability for SAR operators, SARWeather is developed in close collaboration with ICE-SAR and the Civil Protection Department of the Icelandic Police.
• SARWeather was initially funded in part by grant number 550-025 (Vejrtjeneste for Søberedskab) from NORA and by the European Commission under the 7th Community Framework Programme for Research and Technological Development (GalileoCast). GalileoCast is managed by GSA, the European GNSS Supervisory Authority.
• Current development of SARWeather is funded in part by the Icelandic Technical Development Fund – RANNÍS
• Development related to the SUMO has in part been funded by the COST project ES0802
SARWeather – www.sarweather.com
Cloud Computing Education series
• Atmospheric Models• Global vs. regional• The WRF model• Importance of resolution
• Current Crisis Response System• SARWeather demonstration
• Use of observations from UAS’s• Conclusions and future work
Overview
SARWeather – www.sarweather.com
Cloud Computing Education series
Atmospheric models
Set of equations that describe the atmospheric flow and need to be integrated forward in time to produce a weather forecast
Cloud Computing Education series
SARWeather – www.sarweather.com
Global
Regional
Atmospheric models
Set of equations that describe the atmospheric flow and need to be integrated forward in time to produce a weather forecast
Cloud Computing Education series
SARWeather – www.sarweather.comRegional
Initial and boundary data come from global models
• Open source model developed in collaboration of• NCAR, NOAA, FSL, AFWA, NRL, Univ. of Oklahoma & FAA
• Five development teams with sixteen workgroups• More than 160 official developer• Additional development by academic and governmental
institutions in the US and abroad• Very large user community, excellent support
• Non-hydrostatic• Regional and global applications• Wide range of scales for both real time and idealized applications• Optional data assimilation; 3D-VAR, 4D-VAR, EnKFM, and FDDA• Has been modified for volcanic applications• Includes a dust module (re-suspension of dust/ash)
The WRF atmospheric modelCloud Computing Education series
SARWeather – www.sarweather.com
Importance of high resolutionCloud Computing Education series
Importance of high resolution
How does the weather model “see” this mountain?
Cloud Computing Education series
SARWeather – www.sarweather.com
Importance of high resolutionVatnajökull icecap, max height is 1675m. Top height of Mt. Öræfajökull is only 880m
True height of Mt. Öræfajökull is 2119m
Cloud Computing Education series
SARWeather – www.sarweather.com
At 1km resolution the max height is 2020m. Top height of Mt. Öræfajökull is now 1945m
Importance of high resolution
9 km resolution
10 15 20 25 30 35 40 45 m/s
Cloud Computing Education series
SARWeather – www.sarweather.com
Importance of high resolution
3 km resolution
10 15 20 25 30 35 40 45 m/s
Cloud Computing Education series
SARWeather – www.sarweather.com
Importance of high resolution
1 km resolution
Cloud Computing Education series
Simulation results are also affected by the choice of parameterization schemes
Rögnvaldsson, Ó., Bao, J.-W., Ágústsson, H., and Ólafsson, H.: Downslope windstorm in Iceland – WRF/MM5 model comparison, Atmos. Chem. Phys., 11, 103-120, doi:10.5194/acp-11-103-2011, 2011.
SARWeather – www.sarweather.com
Cloud Computing Education series
Importance of high resolution
How is the wind flow over this mountainous peninsula simulated at 9 and 1 km resolution?
Cloud Computing Education series
SARWeather – www.sarweather.com
Importance of high resolution
When model resolution is increased to 1 km the true complexity and strength of the wind field becomes apparent
Very high resolution – 1 km
Medium resolution – 9 km
SARWeather – www.sarweather.com
Cloud Computing Education series
Why not always use 1km resolution?
0204060801001201
10
100
1000
10000
100000
1000000
Grid size [km]
CP
U p
ow
er
= C
os
t
Need 1000-times more CPU power to simulate a 1 km resolution forecast than a 10 km one for the same region!
Cloud Computing Education series
SARWeather – www.sarweather.com
• You only need high very high resolution once in a while?• Computer clouds (e.g. Azure, EC2 and
GreenQloud) are starting to offer HPC service• Offers great scalability• Relatively cheap• And there is already a solution out there
What if
SARWeather – www.sarweather.com
Cloud Computing Education series
• Quality weather information help improve decision making
• Current CRS uses the WRF model and consists of a • Backend and Frontend
• Frontend is called SARWeather• Easy to use• Fast• Flexible model output and presentation• CF and ArcGIS compliant output files• Interactive and static maps
Crisis Response System
SARWeather – www.sarweather.com
Cloud Computing Education series
Simulated and observed surface windson 15 July 2009 at 15 UTC
WRF at a resolution of 500 m forced with ECMWF-data on model levels.
Observed surface winds in red
Mt. Esja
High resolution not always sufficient
Model simulates a see-breeze that is not seen in observations
Cloud Computing Education series
SARWeather – www.sarweather.com
SUMO and WRF
The SUMO (Small Unmanned Meteorological Observer) can measure winds, humidity, pressure, and temperature in a vertical profile up to a 4km height
SARWeather – www.sarweather.com
Can be operated in cold climates
Cloud Computing Education series
SUMO and WRF
This data can be assimilated with the WRF weather forecast
Cloud Computing Education series
SARWeather – www.sarweather.com
The SUMO-data is incorporated into the WRF-simulation, via obs-nudging
SUMO and WRF Cloud Computing Education series
SARWeather – www.sarweather.com
Simulated and observed surface windson 15 July 2009 at 13 UTC
WRF at a resolution of 500 m forced with ECMWF-data on model levels and SUMO data
Observed surface winds in red
Effects of additional observations
The flow structure is now in much better agreement with available observations
Cloud Computing Education series
SARWeather – www.sarweather.com
Simulated flow in N-S section across Mt. Esja
A major difference in flow pattern extending far above mountain top leve
23 km
Wind speed, ranging from 0 to 12 m/s
Effects not just at the surface Cloud Computing Education series
SARWeather – www.sarweather.com
Effects can be far reaching
CTRL - ObsNudge
Marius O. Jonassen, Haraldur Ólafsson, Hálfdán Ágústsson, Ólafur Rögnvaldsson, and Joachim Reuder (2012). Improving a high resolution numerical weather simulation by assimilating data from an unmanned aerial system. Accepted for publication in Monthly Weather Review
“Substantial improvements of winds, temperatures and humidity in the region are achieved”
SARWeather – www.sarweather.com
Cloud Computing Education series
Transmitting data from the field
SARWeather – www.sarweather.com
Three profiles used to nudge the forecast at times 11:55, 12:58 and 14:38 UTCSite altitudes ~ 1300 m.a.s.l.Profile heights ~ 2000 m.a.g.l.From 860hPa to 650/680hPa
One profile used for comparison at 16:30 UTC
Observations made over Myrdalsjokull ice cap in South Iceland on 17 May 2012Data can be transmitted via 3G mobile connection
Cloud Computing Education series
Transmitting data from the field
SARWeather – www.sarweather.com
Observations made over Myrdalsjokull ice cap in South Iceland on 17 May 2012
Obs in blackWRF - noSUMO in redWRF - withSUMO in blue
Profile used for nudging @ 14:38 UTC
Profile used for comparison @ 16:30 UTC
850 hPa
750 hPa
850 hPa
750 hPa
TemperatureWind-speed
Temperature
Wind-speed
-20°C -10°C 0°C 5m/s 10m/s
-20°C -10°C 0°C 5m/s 10m/s
Problems with low-level windspeed and a cold bias 2hrs after last profile
Cloud Computing Education series
The SUMO has been equipped with an optical dust sensor
Additional sensors
GP2Y1010AU0F is a dust sensor by optical sensing system:• An infrared emitting diode (IRED) and an
phototransistor are diagonally arranged into the device
• It detects the reflected light of dust in air• Especially effective to detect very fine particle• In addition it can distinguish smoke from house
dust by pulse pattern of output voltage
Saturation at about 500μg/m3
Cloud Computing Education series
SARWeather – www.sarweather.com
The SUMO dust sensor has been tested in France and Iceland
Preliminary results
SARWeather – www.sarweather.com
Cloud Computing Education series
The SUMO dust sensor has been tested in France and Iceland
Preliminary results
Sensor is now being calibrated and tested with ash from Mt. Eyjafjallajökull
AscendingDescending
Cloud Computing Education series
SARWeather – www.sarweather.com
Calibration – preliminary results• The sensor (BU) readings show high
sensitivity in the range 125 – 300. After that the sensitivity is rather low in the range 350 to 700.
• Using the meter for ash surveillance for jet aircrafts, the best thing would be to designate below 300 as “safe” but above 400 as “unsafe”
SARWeather – www.sarweather.com
Cloud Computing Education series
Various uses of model output
To correctly simulate distribution of pollutants, it is very important to correctly simulate the atmospheric conditions close to the source.
Data courtesy of Prof. Saji Hameed at the University of Aizu, Fukushima.
Cloud Computing Education series
SARWeather – www.sarweather.com
Various uses of model output
Output from SARWeather could also be used as input to coastal waves and circulation and storm surge models such as the coupled SWAN+ADCIRC system
Simulated maximum water level for hurricane Gustav, September 2008. Taken from www.adcirc.org
Cloud Computing Education series
SARWeather – www.sarweather.com
Misc. datasources (optional)
Current research and development Cloud Computing Education series
SARWeather – www.sarweather.com
Current research and development Cloud Computing Education series
SARWeather – www.sarweather.com
• Model resolution is important• Especially in the vicinity of complex terrain
• On-Demand CR system has been developed• Called SARWeather – www.sarweather.com • Data from SARWeather can be used as a driver for a number of
other modeling systems• Additional observations can improve the simulation
• Vertical profiles made by the SUMO, radiosondes or other means
• The SUMO is a low-cost system with many advantages• Proof of concept before investing in a more durable and
expensive UAS• Additional sensors are being added to the system• Is currently being integrated to the SARWeather CR system
Conclusions
SARWeather – www.sarweather.com
Cloud Computing Education series
• Continue improving usability• Re-occurring forecasts• Add output at different vertical levels• Add weather parameters for aviation• Visibility, icing potential, turbulence severity
• Better support for mobile devices• Location based services
• Add support for different modeling systems, e.g.• Dispersion models• Wave and storm surge models• Flash-flood modeling
• What would you like to see?• Interest in developing add-on solutions for SARWeather?
Future work, collaboration?
SARWeather – www.sarweather.com
Cloud Computing Education series