Towards a better understanding of wind gusts: observations, processes, predictions and verification


Forecasting wind gusts may become the next major challenge in numerical weather prediction. With increasing computer power, operational NWP systems just entered the convective scale, allowing the model physics to simulate convective processes more explicitly. While this is very beneficial for precipitation forecasting, wind gusts are still a sub-grid scale phenomenon relying on crude parametrizations. Furthermore, wind gusts cause large socio-economic damages every year. Wind gust predictions are getting higher relevance and load, e.g. for transport, aviation, urban development or public weather warnings. Yet forecast verification exhibits exceptionally low skill for wind gust predictions compared to other meteorological variables, which might also be impacted by a very sparse observational network. The spatial variability of wind gusts is probably as large as that of precipitation, but the observational network is much less dense and no equivalent to the spatial coverage of radar derived precipitation exists.

This session welcomes contributions which lead to a better understanding of the physical processes that determine wind gusts, and novel ideas/methods to improve wind gust forecasting and warnings in the future. More specifically, contributions on the following topics are welcome:

- Observations: The development of novel measurement tools for wind gusts (e.g. WindLIDARS) and suggestions for an optimized observational network in the future. Descriptions of the spatio-temporal variations of gusts.

- Explicit modelling: Small-scale model simulations (e.g. LES simulations) are a prerequisite to explicitly resolve the processes leading to wind gusts. Beside a better understanding of the physical processes they can be utilized to improve empirical approaches to approximate wind gusts more accurately.

- Wind gust forecasting and warnings: Methods to obtain guidance for wind gusts forecasts and warnings from operational weather forecasts, e.g. using historical observations by statistical postprocessing or forecast assimilation techniques. Prediction uncertainty of wind gusts.

- Climate monitoring: Long-term data sets for wind gusts as well as techniques for spatial wind gust analysis which are necessary for climate change adaptation and mitigation strategies

- Evaluation: The high-resolution model simulations on the one side and a sparse observational network on the other side require novel ideas in the verification of wind gusts simulations and warnings.

Convener: Irene Suomi | Co-convener: Sabrina Wahl
Lightning talks
| Tue, 07 Sep, 14:00–15:30 (CEST)

Lightning talks: Tue, 7 Sep

Chairperson: Sabrina Wahl
Jenna Ritvanen, Seppo Pulkkinen, and Dmitri Moisseev

Thunderstorm gust fronts threaten human safety and property, especially in industries such as aviation and construction. The ability to predict the precise time and location of gust front arrivals would mitigate risk and reduce damage. 

Existing methods for nowcasting gust front locations are based on detecting the gust fronts from individual Doppler weather radars or scanning lidars. Even though these methods are locally effective, they have so far not been applied to large-scale radar mosaics to generate forecasts that could benefit society at large. To address this gap, an object-based method is proposed for nowcasting gust fronts by any number of ground-based Doppler weather radars.  

The gust fronts are first detected from the radar measurements and presented as objects consisting of spline curves. Given the one-dimensional geometry of the curves, existing object-based tracking methods, designed for tracking thunderstorms and based on two-dimensional polygons, cannot be applied to the gust front objects. Instead, a tracking method is formulated that matches multiple observations of the same gust front based on the location and length of the curves. The tracking considers possible splitting and merging of the gust front objects. After matching the gust front instances between consecutive timesteps, the location of the gust front is nowcast with a Kalman filter algorithm.  

The methodology is demonstrated with case studies of gust fronts related to mesoscale convective systems (MCS) in Finland. MCSs occur frequently in Finland during summer and cause significant wind and other storm-related damage. Spatially and temporally accurate forecasting of MCS events would aid preparedness and reduce the risk posed to society. The methodology presented in this work can be used to nowcast the gust front trajectory and thus increase preparedness especially for the wind damage related to MCS events. The methodology can also be combined with existing object-based methods for nowcasting convective storm cells, to create comprehensive hazard forecasting systems for thunderstorms.

How to cite: Ritvanen, J., Pulkkinen, S., and Moisseev, D.: Object-based tracking and nowcasting of gust fronts, EMS Annual Meeting 2021, online, 6–10 Sep 2021, EMS2021-44, https://doi.org/10.5194/ems2021-44, 2021.

A new LES-based system for short-range forecasting of near-surface wind gusts at airports
Helge Knoop and Björn Maronga
Santiago Gaztelumendi

Severe  weather  phenomena  impact  the  European  society  and economy  in  many  ways,  from disruption  in  various  sectors  and substantial  damages  in  infrastructure  to  human  and  economic losses.  Windstorms are amongst the most damaging natural hazards in Europe, with approximately 5 € billion of estimated annual losses in the EU.

Economic losses from meteo-climatic hazards depend on the nature of the meteo-climatic severe event and on the vulnerability of exposed human assets to the particular hazard. In  this  work  we  focus  on  wind  damages  that  are  produced  when different meteorological conditions  promoted  high wind gust that affect human assets in a particular area.

In the Basque Country, wind impact episodes can occur during all the year, on one hand during summer session related with severe storms and coastal trapped disturbance and on the other during winter period mainly related with relatively deep pressure system affecting the territory. The latter are by far the ones that generate the events with the greatest impact, both in terms of their spatial extension and the amount of damage generated.

Here we present a characterization study of economic losses in Basque Country due to high impact wind events during a ten years period from 2009 to 2018. For this purpose we analyzed the “non-typical cyclonic storm”   damage   data   provided   by   the Spanish   Insurance Compensation Consortium (CCS) for the Basque Country area. 71.846 accepted claims corresponding to 189 days, affecting 251 municipalities, with a total amount of 83.6 million euros for the ten years period studied. We analyzed those data considering their typology and their spatial and temporal distribution. In order to extract some useful conclusions for further impact modeling, we include surface wind characteristics registered in the Automatic Weather Station (AWS)  Basque network. The final objective of this study is  to  contribute  in  reducing  the  knowledge  gaps  at  the  interface  between  available  local wind prediction/analysis systems and impact observed in Basque Country as a consequence of wind severe events.

The damages are distributed in 93 episodes, made up of one or more consecutive days with paid claims. The 21 high-impact episodes (more than 50 claims) account for 43% of the days with 99% of the claims and damages. In the 10 days of extreme impact (more than 1000 claims), which represent 5% of the days, 90% of the claims and 89% of the damages are recorded. The seasonal distribution of damage shows a clear winter pattern. Although the damage are produced throughout the territory, with different degree of impact in nearly all the municipalities,  there is a certain concentration in the three Basque capitals with 22% of the claims and 23% of the economic losses.

How to cite: Gaztelumendi, S.: Analysis of wind damages during 2009-2018 period in Basque Country, EMS Annual Meeting 2021, online, 6–10 Sep 2021, EMS2021-226, https://doi.org/10.5194/ems2021-226, 2021.

Julian Steinheuer, Carola Detring, Frank Beyrich, Ulrich Löhnert, Petra Friederichs, and Stephanie Fiedler

In the Field Experiment on Sub-Mesoscale Spatio-Temporal Variability in Lindenberg (FESSTVaL, www.fesstval.de) various phenomena in the atmospheric boundary layer are investigated. One goal is to detect wind gusts from measurements of a Doppler wind lidar (DWL). DWL’s allow the determination of wind vector profiles with a high vertical resolution (∼ 30 m) and therefore are an attractive alternative to metorological towers.

However, obtaining wind gusts from DWL measurements is not trivial because a monostatic lidar provides only a radial velocity, i.e., only one component of a three-dimensional vector per individual beam. Measurements in at least three linearly independent directions are therefore necessary to derive the wind vector. These must be performed sequentially, which prolongs the time interval for determining the wind vector and therefore limits the time resolution of the derived wind vector. In order to retrieve wind gusts, wind maxima of a few seconds, one needs to operate the instrument in a quick scanning mode. In this presentation, we show results from different scanning modes and discuss the method for retrieving wind gusts. We tested various configurations with respect to their ability to detect gusts and mean winds at the Boundary Layer Field Site in Falkenberg in autumn 2019. The DWL configurations that measure different lines-of-sight with rapid temporal repetitions have a lower signal-to-noise ratio (SNR) but the highest chance of detecting gusts.

We have developed a new retrieval method that skips prior SNR filtering and instead iteratively removes a fixed number of measurements that do not match a least-squares-fit. The least-squares-fit is then recalculated on the reduced set of measurements, and if necessary, this step is repeated. With appropriate retrieval steps and iteration criteria, our results suggest that prior filtering can be omitted.

We present the results of our new retrieval for eight different DWL configurations consisting of double-beam swinging, step-stare modes, and continuous-scanning modes. The evaluation is done by a comparison of the minimum, maximum and mean wind speed at 90 m a.g.l. against the reference measurements of a sonic anemometer that is located nearby. Ongoing work is addressing further comparison of our retrieved wind variables with unmanned aerial vehicles from the FESSTVaL campaign in summer 2020.

How to cite: Steinheuer, J., Detring, C., Beyrich, F., Löhnert, U., Friederichs, P., and Fiedler, S.: New flexible retrieval for gusts and mean winds from Doppler wind lidars, tested for various scanning configurations, EMS Annual Meeting 2021, online, 6–10 Sep 2021, EMS2021-366, https://doi.org/10.5194/ems2021-366, 2021.

Carola Detring, Julian Steinheuer, Eileen Päschke, Ronny Leinweber, Markus Kayser, and Frank Beyrich

A central aspect of the Field Experiment on Sub-Mesoscale Spatio-Temporal Variability in Lindenberg (FESSTVaL, www.fesstval.de) is the investigation of wind gusts with Doppler lidar measurements. Compared to meteorological tower observations, they have the advantage of being able to probe higher altitudes of the atmosphere, they thus offer the possibility to record a vertical profile of wind gusts with a resolution of about 30 m in the atmospheric boundary layer. Nevertheless, it is difficult to capture wind gusts with these instruments as it is challenging to measure fluctuations of short duration with an instrument which needs a certain time for one complete measurement.

Based on the research of Suomi et al. (2017), different configurations were tested in a pre-campaign in autumn 2019 to identify a suitable measurement mode for Halo Photonics Stream Line Scanning Doppler LiDAR systems. Different lidars were operated in parallel to compare configurations against each other. A promising mode was tested during the FESST@MOL campaign in summer 2020 for a three month period. This is a continous scan mode (CSM) configuration that takes about 3.4 seconds per circulation and performs measurements in 10-11 directions.

The derived wind gusts and mean wind speeds are compared with high resolution sonic anemometer measurements at 90.3 m to verify the quality of the lidar measurements. In a first comparison good agreement is shown despite the different measuring principles. In addition, various parameters are tested to identify optimal thresholds that allow a reliable derivation of wind gusts.

In summer 2021 this fast CSM mode will be operated and further tested in the FESSTVaL campaign in parallel with UAS measurements. Moreover lidars will be installed at different locations to analyse the spatial characteristics of wind gusts with this scanning configuration.


Suomi, I., Gryning, S.‐E., O'Connor, E.J. and Vihma, T. (2017), Methodology for obtaining wind gusts using Doppler lidar. Q.J.R. Meteorol. Soc., 143: 2061-2072. https://doi.org/10.1002/qj.3059

How to cite: Detring, C., Steinheuer, J., Päschke, E., Leinweber, R., Kayser, M., and Beyrich, F.: First results of a promising Doppler lidar configuration to derive wind gusts within the FESSTVaL campaign, EMS Annual Meeting 2021, online, 6–10 Sep 2021, EMS2021-247, https://doi.org/10.5194/ems2021-247, 2021.

José Antonio Benavent-Oltra, Djordje Romanic, Milos Lompar, and Massimiliano Burlando

In this work, two-dimensional (2D) wind fields retrieved by SingleDop software using scanning Doppler lidar data are compared with anemometric measurements in Genoa (Italy). SingleDop is a software module based on the theoretical work described in Xu et al. (2006), which is intended to retrieve 2D low-level winds from either real or simulated Doppler radar data. The lidar used in this work is a three-dimensional (3D) scanning WindCube 400S lidar, developed by Leosphere (France), which scans the azimuthal range of 100º –250º, up to a maximum distance of 14 km in the radial direction, for 4 elevations corresponding to 2.5º, 5°, 7.5° and 10º from the horizontal. The anemometer used for comparison is located about 1.3 km (horizontally) from the Doppler lidar and provides the wind velocity with a sampling rate of 1 Hz.

The dataset analyzed is from November 2019 to June 2020. The total number of available lidar scans per day is ~420 for each elevation (2.5º, 5º, 7.5º and 10º). The 2D wind fields are retrieved by SingleDop for different de-correlated lengths (L= 10, 5 and 1 km). The overall number of measurements available for the comparison is therefore approximately 420 scans per day × 180 days × 4 heights × 3 L, which results in nearly 106 wind velocity values. The wind direction retrieved by SingleDop properly corresponds to the anemometric data with a  BIAS ~13º, RMSE ~40º and a circular correlation of 0.8. Concerning the wind intensity, the results obtained for L = 5 km show the best agreement with the anemometric measurements with a BIAS of 0.8 m/s, RMSE around 1.8 m/s and a correlation coefficient higher than 0.9. Both for wind direction and velocity, the BIAS and RMSE slightly increase with the elevation whereas the circular and linear correlations decrease, as expected due to the increasing distance between lidar and anemometric measurements.

The contribution of José Antonio Benavent-Oltra, Massimiliano Burlando and Djordje Romanic to this research is funded by the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant agreement No. 741273) for the project THUNDERR—Detection, simulation, modeling and loading of thunderstorm outflows to design wind safer and cost-efficient structures—through an Advanced Grant 2016. 

How to cite: Benavent-Oltra, J. A., Romanic, D., Lompar, M., and Burlando, M.: Comparison between the 2D wind fields retrieved by a scanning Doppler lidar and anemometric measurements, EMS Annual Meeting 2021, online, 6–10 Sep 2021, EMS2021-370, https://doi.org/10.5194/ems2021-370, 2021.


Supporters & sponsors