Dual-Doppler Radar and Surface Measurements of

1

14th International Conference on Wind Engineering – Porto Alegre, Brazil – June 21-26, 2015

Dual-Doppler Radar and Surface Measurements of Thunderstorm Outflow Winds

W. Scott Gunter1, John L. Schroeder

1,2

1Department of Geoscience, Texas Tech University, Lubbock, Texas, USA 2National Wind Institute, Texas Tech University, Lubbock, Texas, USA

[email protected], [email protected]

ABSTRACT: For most regions away from hurricane-prone coastlines, thunderstorms produce the majority of extreme winds

and wind-induced damage. While certain extreme wind - producing phenomena have been thoroughly investigated in numerical

models, little full-scale data exist to validate the conclusions of these studies. Using instruments that collect high resolution wind

measurements, Project SCOUT was designed to address this dearth in observations, seek out extreme wind producing

thunderstorms, and collect valuable wind profile and surface measurements within these events. Analysis of several cases

reveals a rapidly evolving wind profile through the course of the outflow. Additionally, differences in the driving meteorology

of each event translated to differences in the wind profile shape and evolution. High-resolution wind speed measurements at

2.25 m in height were used to investigate the turbulence properties of each event. Turbulence parameters varied substantially

between different regions a thunderstorm outflow and were typically maximized near the gust front. As with the wind profiles,

the driving meteorology significantly affected the wind speed time history and evolution of the surface turbulence parameters.

KEY WORDS: Thunderstorm wind profiles; Full-scale measurements; Surface Observations; Turbulence; Dual-Doppler.

1 INTRODUCTION

Numerical and laboratory experiments have demonstrated that thunderstorm outflow wind profiles are both time and space

dependent [1-5]. Many of these downburst simulations have shown that the greatest near-surface horizontal wind speeds within

thunderstorm outflows are associated with the passage of a horizontal vortex traveling radially outward from the downdraft

center. Recent numerical and observational studies also suggest that thunderstorm wind damage may be more commonly caused

by meso- and/or miso- scale circulations along outflow gust fronts of convective lines [6]. This research will summarize data

collected in the outflow of multiple thunderstorms types. The meteorological factors driving high wind events will be

investigated as well as the evolution of wind profiles and surface measurements through the course of different thunderstorm

outflow events.

2 INSTRUMENTATION AND METHODS

In order to contribute thunderstorm outflow observations, Project SCOUT (Severe Convective OUtflow in Thunderstorms)

was designed using Texas Tech University (TTU) observational assets to obtain collocated wind profile and surface

measurements in thunderstorms capable of producing damaging outflow winds. Specific types of thunderstorms that were

targeted include supercell thunderstorms, bow echoes, and other mesoscale convective systems (MCS).

2.1 Instrumentation

Surface measurements were collected in most outflow events with Texas Tech University's StickNet. This portable network of

24 tripods can be rapidly deployed in the path of severe thunderstorms (or hurricanes) and has been designed to withstand harsh

environments while taking quality measurements. Each tripod is equipped with instrumentation to measure standard atmospheric

variables (including wind speed and direction) at approximately 2.25 meters. StickNet platforms sample between 1 and 10 Hz,

providing high-resolution records of the phenomena in which they are deployed.

The TTUKa mobile Doppler radars not only provided meteorological context of an outflow event, but the combined data from

both systems also allowed for the computation of dual-Doppler wind speed and direction profiles. The TTUKa radar systems

nominally collect radial velocity data (the component of wind along the radial of the radar beam) with 15 m range resolution and

a 0.49° half-power beamwidth. The temporal resolution depends on the scanning strategy such that horizontal Plan Position

Indicator scans (PPIs; constant elevation, multiple azimuths) are performed at 30° s-1

and vertical Range-Height Indicators scan

(RHIs; constant azimuth, multiple elevations) are performed at 6° s-1

. As a single radar measurement reflects only the

component of the wind moving along the radial of the radar beam, single-Doppler radial velocity data cannot be used to describe

the full horizontal wind vector. However, there are established techniques to retrieve the full horizontal wind vector if two radars

2


are synchronously measuring approximately orthogonal components of the wind [7].

2.2 Dual-Doppler Methodology

If two radars perform RHI scans along azimuths separated by approximately 90°, then the intersection of the coordinated RHI

planes represents a vertical set of points where dual-Doppler wind syntheses are possible and wind speed and direction profiles

can be retrieved. A similar method was previously used with two Doppler lidars [8]. For this study, radial velocity data

corresponding to the range bin closest to the intersection point were mapped onto a vertical grid with 20 m resolution. The

gridded radial velocity profiles from each radar were then synthesized to obtain dual-Doppler wind speed and direction profiles.

In a thorough comparison with sonic and UVW anemometer data from a 200 m tower, this method was shown to be within 1 m

s-1

of the tower data above 50 m in altitude. Below 50 m, it was hypothesized that a lag in scatterer deceleration as compared to

the deceleration of the wind contributed to an observed 2.2 m s-1

overestimation of the mean dual-Doppler wind speed near the

surface [9].

3 SCOUT DATASET OVERVIEW

3.1 SCOUT TTUKa Radar Observations

Data collected during project SCOUT not only illustrate the differences in thunderstorm outflow structure, but also the effect

of structural differences on the wind speed and direction profiles. This can be seen in a comparison of single-Doppler RHI and

dual-Doppler wind profiles near the gust front of four events. The first event was a cluster of multicell thunderstorms sampled

near Syracuse, Kansas on 11 June 2011. The single-Doppler RHIs from Ka-1 display the outflow as outgoing (positive) radial

velocities that begin as a shallow wedge at the leading edge of the gust front (Figure 1a). At this time, the instantaneous dual-

Doppler wind speed profile differs from the standard logarithmic shape in the greatest wind speeds are found near the bottom of

the profile, but wind speeds are relatively weak. Additionally, the wind direction profile exhibits a layer of directional shear with

over 130° of turning between 50 and 750 m above ground level (AGL).

A squall line associated with a cold front produced the outflow in Figure 1b near Truscott, Texas. This particular outflow

displays the classic density current shape noted in earlier studies [10]. For this event, the outflow appears as negative (inbound)

radial velocities in the RHI image. The leading edge of the outflow was characterized by a large head extending above 2 km and

a pronounced, elevated nose (Figure 1b). As the leading edge propagates through the RHI intersection the lower portion of the

nose becomes evident in the wind speed profile near 500 m AGL. Unlike the Syracuse, Kansas event, there is very little

directional shear associated with the leading edge of this outflow event as seen in the wind direction profile.

The outflow in figure 1c was generated by a bow echo that produced substantial wind damage across West Texas on the

evening of 5 June 2013. The bow echo was intercepted approximately 64 km northwest of Lubbock, Texas, near the town of

Pep. The leading edge of the outflow winds, which appear as negative (or inbound) radial velocities from the perspective of Ka-

1 (Figure 1c), is similar to the wedge-shaped structure of the leading edge of the Syracuse, Kansas event. With this particular

outflow structure, outflow winds are experienced near the surface first, and then build upward through the profile. This

evolution contrasts to that of the Truscott, Texas event where outflow winds are experienced through the depth of the profile

almost simultaneously. At the time of Figure 1c, the outflow is approximately 1 km deep at the profile location. Above this

level, the dual-Doppler wind speed begins to drop and the wind direction begins to shift before the necessary removal of noisy

velocity data truncate the profiles. The radial velocities of Figure 1c also show the outflow to be much deeper downstream.

The rear flank downdraft (RFD) of a supercell thunderstorm was sampled in south-central Kansas on 11 May 2014 near the

town of St. John. RHIs from TTUKa1 show the outflow as negative radial velocities that extend up to 6 km in height (Figure

1d). The leading edge is similar to that of the Pep, Texas and Syracuse, Kansas events in shape, but the depth of the RFD

outflow appears to increase much faster. The instantaneous Dual-Doppler wind profiles at this time reveal little variation in wind

speed and direction below 500 m with wind speeds just over 20 m s-1

and wind direction approximately 250° . The wind speed

profile then becomes quite variable until the profile exceeds the height of the outflow around 1.6 km AGL. Above this level,

inflow air is most likely being sampled. The inflow wind speeds increase with height to near 30 m s-1

, while the wind direction

profile backs to approximately 170°. This supercell went on to produce a tornado approximately 45 minutes later near the town

of Sterling, Kansas.

The variability between thunderstorm outflow events is further demonstrated in the 10-minute mean profiles from multiple

events (Figure 2). While the stage of each outflow [11] and the underlying roughness characteristics differed, several of the

profiles from the first 10 minutes of the outflow display a jet of faster wind speeds above the surface (LKF, SYR, and PEP),

while other profiles are characterized by a general increase in wind speed with height (PSPR, TRUS, and STJ). Different

meteorological factors associated with these events are most likely driving the profile shapes. For instance, the circulation

around a horizontal vortex, similar to those seen in numerical models [1-5], increased the wind speed in the lower portion of the

SYR profile, while decreasing the wind speed in the upper portion of the profile. The impingement of a descending rear-inflow

jet is potentially responsible for the high magnitude of wind speeds in the PEP profile as well as the maximum just above 200 m.

Despite these differences, all of the 10-minute mean profiles in Figure 6 indicate the wind maximum was elevated. Other

averaging times were also investigated for several events. For instance, the 1-minute mean profiles of the SYR event (Figure 3)

demonstrate the evolution of the wind profile about the mean, but the wind max remained elevated through each time period.

3


While none of the events sampled were identical to a typical microburst often simulated in numerical models, this finding was

significant given that many of these simulations place a wind maximum in the lowest 50 m of the atmosphere.

Figure 1. RHI snapshots and inset wind speed and direction profiles from 4 selected outflow events: a) Syracuse, Kansas b)

Truscott, Texas c) Pep, Texas d) St. John, Kansas. Outbound radial velocities (m s-1

) are in blue, while inbound radial velocities

(m s-1

) are in brown. The vertical black line between 2 and 4 km represents the intersection point. Dual-Doppler wind speed and

direction profiles at the intersection point are inset.

Figure 2. Dual-Doppler profiles from the first ten minutes of the outflow of multiple events sampled during Project SCOUT.

4


Figure 3. Various dual-Doppler mean wind speed profiles from the SYR event. Mean profiles include 1-minute means from 5

different segments (1min Mean1-5), 10-minute means from 2 different segments (10min Mean1-2), and the mean profile over

the entire event (Event Mean)

3.2 SCOUT StickNet Observations

StickNet deployments focused on collecting data that could be used to evaluate the near surface turbulence within

thunderstorm outflows. Emphasis was primarily placed on deploying StickNet arrays in a manner that would allow for the

computation of longitudinal and lateral integral scales. This objective required a StickNet array deployed parallel and

perpendicular to the expected outflow wind direction to capture the longitudinal and lateral integral scales respectively. Also,

spacing between the probes was varied within each array to ensure that a wide range of length scales could be computed. A

secondary objective of each StickNet deployment was having at least 1 probe near or under the intersection of the coordinated

RHIs. To facilitate these objectives, road intersections were often targeted for both StickNet and TTUKa radar deployments.

The type of thunderstorm producing the outflow winds also factored in the deployment strategy. Large mesoscale convective

complexes (MCSs) were ideal in that they often provided adequate lead-time as well as the greatest probability of intercepting

severe winds with minimal threat to SCOUT participants and TTU property. For such a system, fine-scale StickNet arrays were

deployed in perpendicular directions about a road intersection (if available) with individual probe separations on the order of 25

m. From the fine-scale array, the probe separation was increased until the entire array spanned approximately 1 km. Supercell

thunderstorms, on the other hand, often provided the opposite experience. Lead times were short, storm evolution was complex,

and there was significant risk for the low probability of intercepting a narrow corridor of severe winds within the rear flank

downdraft (RFD). To account for these factors, a large-scale StickNet array with 483 m (0.3 mile) probe spacing was distributed

perpendicular to the expected motion of the RFD. After allowing more time for storm evolution, the StickNet probe mostly

likely to experience severe wind was identified, and a fine-scale array was distributed about that probe. The logistical difficulties

and safety considerations involved in supercell thunderstorms resulted in few successful supercell deployments. Despite these

challenges, multiple datasets were collected from a variety of thunderstorm types though the duration Project SCOUT.

As the radar data in the previous section suggest, the wind speed and direction time histories as measured by StickNet varied

considerable between events. Some thunderstorm outflows were similar to the wind ramp events examined in [12] and displayed

a sharp peak in the wind speed time history followed quickly by a rapid decrease (Figure 4a). Other events exhibited a rapid

increase to the peak wind speed and a much slower decrease in wind speeds after the passage of the gust front (Figure 4b). Still

others events involved a much slower increase and subsequent decrease in the outflow wind speeds as well as the existence of

multiple surges (Figure 4c). These events were similar those studied in [13]. Figure 4a in particular demonstrates the localized

yet significant impact that a miso-scale circulation along the leading edge of a gust front can have on outflow wind speeds.

During this event, multiple StickNet platforms recorded peak instantaneous values over 30 m s-1

while another weather station

and a 200 m tall instrumented tower located 2.3 and 2.6 km to the north recorded no such ramp in wind speed. Examination of

TTUKa radar data near the time of the peak wind speed reveals two small circulations: one of the west of the StickNet array and

one overtop the StickNet array as evidenced in both the reflectivity and radial velocity fields.

5


Figure 4. StickNet raw wind speed time histories from multiple events sampled during project SCOUT. Each time history represents one hour of data starting with inflow conditions.

4 SCOUT CASE STUDY: 5 JUNE 2013 HIGH WIND EVENT

Few of the sampled outflow events produced severe wind gusts (defined by the United States National Weather Service as a

wind gust exceeding 26 m s-1

) that were realized at the 2.25 m measurement height of the StickNet platforms. However, the bow

echo event on 5 June 2013 produced wide spread severe wind reports and wind damage across West Texas. Eleven West Texas

Mesonet stations [14] recorded a peak 3 second gust over 30 m s-1

, with three of those stations recording gusts over 35 m s-1

at

10 m. Damage to structures and significant tree fall were documented in and around Lubbock, Texas. The TTUKa radar and

StickNet teams were deployed in the northwestern corner of Hockley county, 64 km to the northwest of Lubbock. All ten towers

that were deployed recorded peak 3 second gusts over 30 m s-1

at 2.25 m. The deployment location was primarily characterized

by flat, open terrain in the form of recently plowed fields. The deployment was arranged such that several StickNet towers

would be close to the RHI intersection point (and thus the wind profile location) and collocated wind profile and surface

measurements could be compared. Of the ten towers deployed, StickNet 102 was the closest to the intersection point. The raw

10 Hz wind speed time history from StickNet 102 reveals a complex evolution after the passage of the gust front at 02:24 UTC

(Figure 5a) and is qualitatively similar to outflow time histories discussed in previous studies [13,15]. A peak 3 second wind

gust of 33 m s-1

from 246° was measured in association with the surge in wind speed just before 02:29 UTC. This peak occurred

326 seconds after the passage of the gust front. The main surge occurs near 02:31 UTC. The other nine probes recorded similar

features as indicated by the outflow statistics in Table 1.

Mean wind profiles associated with these regions were also computed. Given the design of the dual-Doppler scanning

strategy, groups of 10 RHIs were separated by 2 PPIs thus creating intermittent groups of 10 wind profiles. The ten radar

profiles (over approximately 75 seconds ) collected during each portion of the outflow were averaged to produced the mean

profiles in Figure 5b and 5c. During the initial surge, wind speed profile (I) shows a broad region of higher wind speeds between

150 and 500 m in height. The mean wind direction profile (I) veers with height up to approximately 500 m which is suggestive

of the mean height of the outflow during this time. The wind speed profile averaged over time period (II) shows a shallower

slope between 0 and 100 m than profile (I) indicating that wind speed shear has increased. Additionally, the wind speed is

relatively uniform above 200 m with values near 40 m s-1

. During the main surge, the wind speed profile (III) indicates higher

wind speeds below 800 m than the other two profiles. A jet centered at 200 m is also apparent in this region of the outflow. The

wind direction profile (III) is also relatively uniform through 1000 m.

Before investigating the turbulence characteristics of the wind produced in this event, a more objective method was

incorporated to define different regions of the outflow. As the regions described above (gust front, main surge, etc) were

associated with different wind directions, the 40-second mean wind direction was used to segregate the StickNet time histories

into 5 regions: the inflow, the gust front, region 1 of the outflow, a secondary gust front, and region 2 of the outflow (Figure 6).

While the gust front typically separates the inflow and outflow winds, TTUKa radar data indicate that the wind speed and

direction time history of this event were also influenced by a circulation situated along the gust front. The effect of the

circulation was to not only to slow the veering of the wind direction after the passage of the gust front (Figure 6a; R1), but also

6


to increase the wind speed on the southern edge of the circulation (Figure 6b; GF2). Turbulence statistics were computed for the

different regions of the outflow as defined by the 40-second mean wind direction. As exemplified by StickNet 102 (Figure 7),

running turbulence intensity (calculated as in Holmes et al. (2008) using a 40-second moving average) was typically greatest in

the gust front region (GF1). Surprisingly, mean TI values measured in the inflow were similar to those measured in outflow

region 2 (R2). The peak 3-second gust was also contained in this region for many probes. Lateral TI was also computed and was

almost uniformly greater in the two gust front regions. Given the sensitivity of lateral TI to changing wind direction, this result

is not surprising. Longitudinal integral scales were computed using 40-second segments and assuming signal stationarity over

that time period. The results indicate that mean integral scale values were uniformly smaller in inflow as compared to other

regions of the outflow for all probes. Values ranged from 15 to 24 m in the inflow and roughly doubled to between 31 and 46 m

in region 2 of the outflow. The highest integral scale values, including a peak of 63 m, were noted in the GF2 region for most

probes (Figure 8). Similar trends were noted in [13] despite the use of a different method and a longer averaging time to

compute longitudinal integral scales.

Table 1. Gust statistics from the 5 June 2013 high wind event. Tower 102 was the easternmost tower while tower 105 was the westernmost tower.

StickNet Peak 3 s Gust (m s-1

) Peak 3 s Gust WD (°) Tmax - Tmin (s)

102 32.97 246.46 326 103 32.88 243.03 307.8 104 32.93 240.25 336.4 106 32.87 245.18 304.2 107 34.53 296.79 488.6 108 34.21 301.20 580.3 109 30.77 300.52 449.2 112 32.09 267.07 375.2 111 36.63 304.38 388 105 31.71 268.50 381.6

Figure 5. StickNet 102 wind speed time history (a) with select regions enclosed by the black lines. Mean wind speed and direction profiles for the select regions are given in (b) and (c) respectively.

7


Figure 6. Time histories of the 40-second mean wind direction (a) and the 40 second mean wind speed (b) for the Pep, TX high wind event on 5 June 2013. The different regions of the time history discussed in the text are separated by color.

Figure 7. Comparison of running turbulence intensity and the longitudinal wind speed for StickNet tower 102 from the Pep, TX high wind event on 5 June 2013. The regions of the time history are labelled at the top of the plot as in Figure 6.

8


Figure 8. Variation of longitudinal integral scale values by storm region. The abscissa represents the StickNet towers in the order they were deployed from west (left) to east (right).

5 SUMMARY

Obtaining high-resolution full-scale observations of thunderstorm outflow winds was the main objective of Project SCOUT.

Project SCOUT continues to be successful in collecting innovative, high-resolution thunderstorm outflow data that are being

analyzed to further understand the structure and evolution of thunderstorm outflow wind profiles and surface turbulence. The

coordinated RHI method for dual-Doppler wind profiles was successfully verified in a comparison with data from a 200 m

meteorological tower in both precipitating and clear-air environments [9]. Most of the results of the clear-air validation dataset

echo what was seen in thunderstorm validation datasets of lesser duration. However, differences between the clear-air and

thunderstorm near-surface wind speed profile that were apparent were attributed to different scatterers in the atmosphere. Multiple outflow events and parent storm types were also discussed, in which significant variation was noted not only in the

shape and depth of the different gust fronts, but also in the shape of the wind speed and direction profiles. StickNet observations

of the 5 June 2013 high wind event in West Texas indicated a relatively slow ramp (> 5 minutes) to the peak gust as well as

multiple surges within the outflow. Dual-Doppler data illustrated that the shape of the wind speed and direction profiles vary

across the different outflow surges. Future endeavors include investigating integral scales computed through distance

correlations as well as trying to relate boundary layer features, such as those noted in [16], to integral scale trends in the

StickNet data. Additionally, the deployment of TTU observational assets for Project SCOUT has been optimized to include

smaller spacing between towers (to resolve the lateral dimension of the integral scales) and a quicker revisit time of dual-

Doppler RHIs. The results of these and future analyses have the potential to make a significant contribution to the current

understanding of the structure and dynamics of thunderstorm outflow winds. With this gain of knowledge, modifications can be

made to existing design codes to maximize positive thunderstorm wind-structure interaction at minimal cost.

ACKNOWLEDGMENTS

This research is supported by the National Science Foundation grant CMMI-1000160. The authors also thank Dr. Brian Hirth,

Jerry Guynes and Dr. Pat Skinner for help with the collection of radar and tower data and radar constuction and maintenance.

Additional conversations with Dr. Frank Lombardo and Rich Krupar greatly contributed to this work.

REFERENCES

[1] J. Kim and H. Hangan, Numerical simulations of impinging jets with application to downbursts, Journal of Wind Engineering and Industrial Aerodynamics,

95 (2007) 279-298.

[2] M.S. Mason, G.S. Wood and D.F Fletcher, Numerical simulation of downburst winds, Journal of Wind Engineering and Industrial Aerodynamics, 97 (2009)

523-539.

9


[3] W.E. Lin, L.G. Orf, E. Savory, C. Novacco, Proposed large-scale modeling of the transient features of a downburst, Wind and Structures, 10 (2007) 315-346.

[4] B.C. Vermeire, L.G. Orf, and E. Savory, A parametric study of downburst line near-surface outflows, Journal of Wind Engineering and Industrial

Aerodynamics, 99 (2011) 226-238.

[5] L. G. Orf, E. Kantor, and E. Savory, Simulation of a downburst-producing thunderstorm using a very high-resolution three-dimensional cloud model, Journal

of Wind Engineering and Industrial Aerodynamics, 104 (2012) 547-557.

[6] D. M. Wheatley, R. J. Trapp, and N. T. Atkins, Radar and Damage Analysis of Severe Bow Echoes Observed during BAMEX, Monthly Weather Review,

134 (2006) 791–806.

[7] R.P. Davies-Jones, Dual-Doppler radar coverage area as a function of measurement accuracy and spatial resolution, Journal of Applied Meteorology

18.9 (1979) 1229-1233.

[8] R. Calhoun, R. Heap, M. Princevac, R. Newsom, H. Fernando, and D. Ligon, Virtual towers using coherent Doppler lidar during the Joint Urban 2003

Dispersion Experiment, Journal of Applied Meteorology, 45 (2006) 1116-1126.

[9] W.S. Gunter, J.L. Schroeder, and B.D. Hirth, Validation of dual-Doppler wind profiles with traditional anemometry, Journal of Atmospheric and Oceanic

Technology, In Press.

[10] J. Charba, Application of a gravity current model to analysis of squall line gust front, Monthly Weather Review, 102 (1974) 140-156.

[11]R.M. Wakimoto, The life cycle of thunderstorm gust fronts as viewed with Doppler radar and rawinsonde data, Monthly Weather Review, 110,1060-1082.

[12] F.T. Lombardo, D.A. Smith, J.L. Schroeder, and K.C. Meta, Thunderstorm characteristics of importance to wind engineering, Journal of Wind Engineering

and Industrial Aerodynamics, 125 (2014) 121-132.

[13] K.D. Orwig and J.L. Schroeder, Near-surface wind characteristics of extreme thunderstorm outflows, Journal of Wind Engineering and Industrial

Aerodynamics, 95 (2007) 565-584.

[14] J.L. Schroeder, W.S. Burgett, K.B. Haynie, I. Sonmez, G.D. Skwira, A.L. Doggett, and J.W. Lipe, The West Texas Mesonet: A technical overview, Journal

of Atmospheric and Oceanic Technology, 22 (2005) 211 - 222.

[15] J.D. Holmes, H.M. Hangan, J.L. Schroeder, C.W. Letchford, and K.D. Orwig, A forensic study of the Lubbock-Reese downdraft of 2002, Wind and

Structures, 11 (2007) 137-152.

[16] S. Lorsolo, J. L. Schroeder, P. Dodge, and F. Marks Jr, An observational study of hurricane boundary layer small-scale coherent structures Monthly

Weather Review, 136 (2008) 2871–2893.

Documents

Dual-Doppler Radar and Surface Measurements of