TGNet:Intelligent Identification of Thunderstorm Wind Gusts Using Multimodal Fusion

Thunderstorm wind gusts are small in scale,typically occurring within a range of a few kilometers.It is extremely challenging to monitor and forecast thunderstorm wind gusts using only automatic weather stations.Therefore,it is necessary to establish thunderstorm wind gust identification techniques based on multisource high-resolution observations.This paper introduces a new algorithm,called thunderstorm wind gust identification network(TGNet).It leverages multimodal feature fusion to fuse the temporal and spatial features of thunderstorm wind gust events.The shapelet transform is first used to extract the temporal features of wind speeds from automatic weather stations,which is aimed at distinguishing thunderstorm wind gusts from those caused by synoptic-scale systems or typhoons.Then,the encoder,structured upon the U-shaped network(U-Net)and incorporating recurrent residual convolutional blocks(R2U-Net),is employed to extract the corresponding spatial convective characteristics of satellite,radar,and lightning observations.Finally,by using the multimodal deep fusion module based on multi-head cross-attention,the temporal features of wind speed at each automatic weather station are incorporated into the spatial features to obtain 10-minutely classification of thunderstorm wind gusts.TGNet products have high accuracy,with a critical success index reaching 0.77.Compared with those of U-Net and R2U-Net,the false alarm rate of TGNet products decreases by 31.28%and 24.15%,respectively.The new algorithm provides grid products of thunderstorm wind gusts with a spatial resolution of 0.01°,updated every 10minutes.The results are finer and more accurate,thereby helping to improve the accuracy of operational warnings for thunderstorm wind gusts.

A Deep Learning Approach for Forecasting Thunderstorm Gusts in the Beijing–Tianjin–Hebei Region

Thunderstorm gusts are a common form of severe convective weather in the warm season in North China,and it is of great importance to correctly forecast them.At present,the forecasting of thunderstorm gusts is mainly based on traditional subjective methods,which fails to achieve high-resolution and high-frequency gridded forecasts based on multiple observation sources.In this paper,we propose a deep learning method called Thunderstorm Gusts TransU-net(TGTransUnet)to forecast thunderstorm gusts in North China based on multi-source gridded product data from the Institute of Urban Meteorology(IUM)with a lead time of 1 to 6 h.To determine the specific range of thunderstorm gusts,we combine three meteorological variables:radar reflectivity factor,lightning location,and 1-h maximum instantaneous wind speed from automatic weather stations(AWSs),and obtain a reasonable ground truth of thunderstorm gusts.Then,we transform the forecasting problem into an image-to-image problem in deep learning under the TG-TransUnet architecture,which is based on convolutional neural networks and a transformer.The analysis and forecast data of the enriched multi-source gridded comprehensive forecasting system for the period 2021–23 are then used as training,validation,and testing datasets.Finally,the performance of TG-TransUnet is compared with other methods.The results show that TG-TransUnet has the best prediction results at 1–6 h.The IUM is currently using this model to support the forecasting of thunderstorm gusts in North China.

The Evolution of Microphysical Structures and Cloud-to-Ground Lightning in a Deep Compact Thunderstorm over the Nanjing Area

In this study,we examine the dynamics and microphysical structures of a deep compact thunderstorm event driving cloud-to-ground(CG)lightning over the Nanjing area located within the Yangtze-Huai River Basin(YHRB)during the monsoon break period.The microphysical structures combined with the dynamics in the glaciated,mixed-phase,and warm-phase layers during the formative,intensifying,and mature stages of the thunderstorm were first investigated using C-band polarimetric radar and CG lightning observations.The results show that the mature phase of the thunderstorm produced a local cold pool,which collided with a southerly warm wind,resulting in a strong updraft.The strong updraft favored the lifting of raindrops to the mixed-phase region to form abundant supercooled liquid water and graupel.From the formative stage to the developing stage and further to the mature stage,increased ZH-and reduced ZDR-values within the mixed-phase region are found,especially within the strong updraft region(>5 m s^(-1)).This phenomenon suggests that supercooled raindrops evolved into large hydrometeors(graupel and hail),indicative of a strong riming process.The signatures within this region are consistent with a favorable environment for thunderstorm electrification and generate the most frequent lightning during the thunderstorm life cycle.

Seasonal Variability of Rainfall and Thunderstorm Patterns in Kenya

This paper presents an analysis of spatial and temporal variation of rainfall and thunderstorm occurrence over Ken-ya from January 1987 to December 2017.The meteorological data used were obtained from the Kenya Meteorological Department(KMD)for the same period.This included the monthly thunderstorm occurrences and rainfall amounts of 26 synoptic stations across the country.The characteristics of monthly,seasonal and annual frequency results were presented on spatial maps while Time series graphs were used to display the pattern for annual cycle,seasonal varia-tions and the inter-annual variability of rainfall amounts and thunderstorm occurrences.A well-known non-parametric statistical method Mann Kendall(MK)trend test was used to determine and compare the statistical significance of the trends.Thunderstorm frequencies over the Eastern,Central and Coast regions of the country showed a bimodal pattern with high frequencies coinciding with March-April-May(MAM)and October-November-December(OND)rainy sea-sons.Very few thunderstorm days were detected over June-July-August(JJA)season.The areas to the western part of the country,near Lake Victoria,had the highest thunderstorm frequencies in the country over the three seasons:MAM,JJAS and OND.The annual frequency showed a quasi-unimodal pattern.These places near Lake Victoria showed sig-nificantly increasing thunderstorm trends during the MAM and OND seasons irrespective of the rainfall trends.This shows the effects of Lake Victoria over these areas,and it acts as a continuous source of moisture for thunderstorm for-mation.However,most stations across the country showed a reducing trend of thunderstorm frequency during MAM and JJA seasons.The importance of these findings is that they could support various policy makers,and users of cli-mate information,especially in the agriculture and aviation industries.

Radar Echo and Lightning Characteristics Analysis on A Strong Thunderstorm Weather in Fuxin

Based on the radar data and lightning position indicator data of strong thunderstorm weather which happened in Fuxin on July 8,2007,the relationship between the lightning activity and the radar echo was analyzed.The results showed that Fuxin area located in the cross position of T-shaped trough and was affected by the cold air which continuously glided down.The corresponding warm front on the ground advanced southward and arrived here.It was the weather background of this thunderstorm weather.The position variation of lightning occurrence was closely related to the strong echo movement of squall line,and the velocity echo clearly reflected and predicted the movement tendency of the radar echo.

Cause Analysis on the Missing Report of First Thunderstorm Weather in Shenyang City in 2010

The first thunderstorm weather appeared in southern Shenyang on May 2,2010 and did not bring about severe lightning disaster for Shenyang region,but forecast service had poor effect without forecasting thunderstorm weather accurately.In our paper,the reasons for missing report of this thunderstorm weather were analyzed,and analysis on thunderstorm potential was carried out by means of mesoscale analysis technique,providing technical index and vantage point for the prediction of thunderstorm potential.The results showed that the reasons for missing report of this weather process were as follows:surface temperature at prophase was constantly lower going against the development of convective weather;the interpreting and analyzing ability of numerical forecast product should be improved;the forecast result of T639 model was better than that of Japanese numerical forecast;the study and application of mesoscale analysis technique should be strengthened,and this service was formally developed after thunderstorm weather on June 1,2010.

Analysis on Features of Spatial and Temporal Distribution of Thunderstorm in Shanghai

Based in 11 daily weather observation station data in Shanghai from 1971 to 2008,a careful research and analysis on the features of thunderstorms spatial and temporal distribution and thunderstorm movement in Shanghai was carried out by using the statistical software of SAS,the method of Mann-Kendall test and wavelets. The results showed that the average annual numbers of thunderstorms days were 26.1,and inter-annual thunderstorm variability was obvious,the annual number of thunderstorm days had a decreasing trend,its value of decreasing days was about-0.418 5 d/10 a. Mann-Kendall test showed that there was an abrupt change in 2000. The seasonal variation of thunderstorm in Shanghai was explicit. The period from March to September was the season when thunderstorm occurred most frequently,about 64.9% of the thunderstorms in a year took place in summer. The results from wavelets analysis showed that the variation cycle period of the annual number of thunderstorms days was about 3,5,12 and 20 years.

Analysis of Laws of Thunderstorm Activity in Hefei City

[Objective] The study aims to analyze characteristics of thunderstorm activity in Hefei City. [Method] Based on conventional ground observational data during1981-2010 and lightning location data in 2010-2013 in Hefei City, temporal and spatial variation of thunderstorm days were analyzed using statistical methods, and then the distribution laws of thunderstorm days were compared with the lightning location data. [Result] In Hefei City, multi-year average of thunderstorm days from1981 to 2010 was more in the south but less in the north, and annual distribution of thunderstorm days was extremely uneven. Moreover, there were obvious seasonal and monthly variation in thunderstorm days in Hefei City. Thunderstorm days were the most in summer, and monthly average of thunderstorm days in Hefei City had a peak in July. From 2010 to 2013, the monthly variation curves of total frequency of cloud-to-ground lightning and frequency of negative cloud-to-ground lightning in Hefei City had a peak each, and cloud-to-ground lightning was frequent in July and August, especially August. The frequency of cloud-to-ground lightning exceeded the average from 12：00 to 21：00. The maximum intensity of cloud-to-ground lightning in Hefei City varied greatly in different months, and it was the highest in July. There are certain differences between the two kinds of data in the distribution laws, so it is needed to combine data of lightning position indicator and long-term artificial observation data to study the detection efficiency of lightning position indicator. [Conclusion] The research can provide theoretical references for lightning protection and disaster reduction in Hefei City.

Temporal and Spatial Distribution and Influence Systems of Spring Thunderstorms in 2013 in Sichuan Province

Based on the lightning monitoring data, automatic and routine weather station observation data in spring （March-May） of 2013 of Sichuan Province, the corresponding relationship between the spatial distribution and the different regions, and the characteristics of atmospheric circulation and evolution of influence the sys- tem were analyzed and summarized. The results show that： the lightning and thunderstorm showed great regional differences in the spring of 2013 in Sichuan Province and the thunderstorm activity period was not the same in different areas. Because of the change of atmospheric circulation, the influence system from March to May corresponding to the thunderstorms in Sichuan tended to be volatile, also.

Three-Dimensional Numerical Simulations of the Effects of a Cold Water Surface on the Evolution and Propagation of Thunderstorms

The influences of large areas of semi-unbounded cold water surface on the evolution, propagation and precipitation production of thunderstorms are simulated by using a fully elastic three-dimensional numerical hailstorm model. Real sounding profiles for temperature, humidity and wind are employed. The model has successfully simulated the significant modification of the propagation path of thunderstorms near the cold water area. The path change can be either' along-bank' or' toward-bank', depending on the position of the storm system relative to convergence zone of the water-land circulation. The simulations also show that thunderstorms developing or propagating within the convergence zone of local circulation will be intensified and produce much heavier hail, whereas those over cold water surface or the air that has been cooled by the water will be strongly inhibited.The influence of the cold water surface on thunderstorm characters is largely dependent upon the direction and intensity of the low-level wind.

Thunderstorm and Lightning Activities over Western Pacific,Northern Indian Ocean and South China Sea Along with Their Adjacent Lands

The Lightning Imaging Sensor(LIS)and Radar Precipitation Feature(RPF)data are used to investigate the activities and properties of lightning and thunderstorms over a region including the Western Pacific,northern Indian Ocean and the South China Sea along with their adjacent lands.The lands feature significantly more frequent lightning flashes and thunderstorms than the oceans,especially the open oceans.The highest densities of lightning and thunderstorm occur over the Strait of Malacca and the southern foothills of the Himalayas.Over the ocean regions,the Bay of Bengal and the South China Sea are characterized by relatively frequent lightning and thunderstorm activities.Larger average spatiotemporal size and optical radiance of flashes can be found over the oceans;specifically,the offshore area features the most significant flash duration,and the open ocean area is characterized by the greatest flash length and optical radiance.The smallest average values of flash properties can be found over and around the Tibetan Plateau(TP).The oceanic thunderstorms tend to have a significantly larger horizontal extent than the continental thunderstorms,with the former and latter having the average area of the regions with radar reflectivity larger than 20 dBZ,generally over 7000 km^(2) and commonly below 6000 km^(2),respectively.The TP thunderstorms show the smallest horizontal extent.Meanwhile,the oceanic thunderstorms exhibit greater 20 dBZ but smaller 40 dBZ top heights than the continental thunderstorms.The average flash frequency and density of the oceanic thunderstorms are typically less than 5 fl min^(-1) and 0.3 fl 100 km^(-2) min^(-1),respectively;in contrast,the corresponding values of continental thunderstorms are greater.It is explored that the regions associated with strong convective thunderstorms are more likely to feature small-horizontal-extent and low-radiance flashes.

Temporal and Spatial Analysis of Thunderstorms in 46 Years in Henan Province

By dint of natural orthogonal function(EOF) decomposition,correlation and trend analysis methods,the temporal and spatial variation characteristics of thunderstorms in recent 46 years in Henan Province were analyzed.The result showed that the thunderstorms in Henan Province decreased gradually from the northwest region to the southeast region and the frequency of thunderstorms in the southeast area was relatively high.The thunderstorm intensity area was in its horizontal distribution.Thunderstorms acted relative actively in 60s and tended to dwindle in the end of 80s.While in recent years,the thunderstorms tended to increase and started act frequently.Since March to August in every year,thunderstorms multiplied in each region and decreased after September.The period between 16:00 to 20:00 was the high peak hours of thunderstorms every year.Thunderstorms distribution in Henan Province had pretty good consistence,increasing and decreasing at the same time.The annual variation of thunderstorms showed an unobvious decreasing tendency.

Charge Structure of a Summer Thunderstorm in North China:Simulation Using a Regional Atmospheric Model System

Electrification and simple discharge schemes are coupled into a 3D Regional Atmospheric Model System （RAMS） as microphysical parameterizations, in accordance with electrical experiment results. The dynamics, microphysics, and electrifi- cation components are fully integrated into the RAMS model, and the inductive and non-inductive electrification mechanisms are considered in the charging process. The results indicate that the thunderstorm mainly had a normal tripole charge structure. The simulated charge structure and lightning frequency are basically consistent with observations of the lightning radiation source distribution. The non-inductive charging mechanism contributed to the electrification during the whole lifetime of the thunderstorm, while the inductive electrification mechanism played a significant role in the development period and the mature stage when the electric field reached a large value. The charge structure in the convective region and the rearward region are analyzed, showing that the charge density in the convective region was double that in the rearward region.

Case Studies of Sprite-producing and Non-sprite-producing Summer Thunderstorms

Three summer thunderstorms in the eastern region of China were analyzed in detail using multiple data, including Doppler radar, lightning location network, TRMM （Tropical Rainfall Measuring Mission）, MT- SAT （Multi-Function Transport Satellite） images, NCEP （National Centers for Environmental Prediction） Reanalysis, and radiosonde. Two of the three storms were sprite-producing and the other was non-sprite- producing. The two sprite-producing storms occurred on 1 2 August and 2~28 July 2007, producing 16 and one sprite, respectively. The non-sprite-producing storm occurred on 29-30 July 2007. The major ob- jective of the study was to try to find possible differences between sprite-producing and non-sprite producing storms using the multiple datasets. The results showed that the convection in the 1-2 August storm was the strongest compared with the other storms, and it produced the largest number of sprites. Precipitation ice, cloud ice and cloud water content in the convective regions in the 1-2 August storm were larger than in the other two storms, but the opposite was true in the weak convective regions. The storm microphysical prop- erties along lines through parent CG （cloud-to-ground lightning） locations showed no special characteristics related to sprites. The flash rate evolution in the 1-2 August storm provided additional confirmation that major sprite activity coincides with a rapid decrease in the negative CG flash rate. However, the evolution curve of the CG flash rate was erratic in the sprite-producing storm on 27-28 July, which was significantly different from that in the 1 2 August storm. The average positive CG peak current in sprite-producing storms was larger than that in the non-sprite-producing one.

A PRELIMINARY STUDY ON THUNDERSTORM FORECAST WITH LS-SVM METHOD

The LS-SVM(Least squares support vector machine) method is presented to set up a model to forecast the occurrence of thunderstorms in the Nanjing area by combining NCEP FNL Operational Global Analysis data on 1.0°×1.0° grids and cloud-to-ground lightning data observed with a lightning location system in Jiangsu province during 2007-2008.A dataset with 642 samples,including 195 thunderstorm samples and 447 non-thunderstorm samples,are randomly divided into two groups,one(having 386 samples) for modeling and the rest for independent verification.The predictors are atmospheric instability parameters which can be obtained from the NCEP data and the predictand is the occurrence of thunderstorms observed by the lightning location system.Preliminary applications to the independent samples for a 6-hour forecast of thunderstorm events show that the prediction correction rate of this model is 78.26%,false alarm rate is 21.74%,and forecasting technical score is 0.61,all better than those from either linear regression or artificial neural network.

Statistical Analysis of Thunderstorms on the Eastern Tibetan Plateau Based on Modified Thunderstorm Indices

The Tibetan Plateau, with an average altitude above 4000 m, is the highest and largest plateau in the world. The frequency of thunderstorms in this region is extremely high. Many indices are used in operational forecasting to assess the stability of the atmosphere and predict the probability of severe thunderstorm development. One of the disadvantages of many of these indices is that they are mainly based on observations from plains. However, considering the Plateau＇s high elevation, most convective parameters cannot be applied directly, or their application is ineffective. The pre-convective environment on thunderstorm days in this region is investigated based on sounding data obtained throughout a five-year period（2006–10）.Thunderstorms occur over the Tibetan Plateau under conditions that differ strikingly from those in plains. On this basis,stability indices, such as the Showalter index（including SI and SICCL）, and the K index are improved to better assess the thunderstorm environments on the Plateau. Verification parameters, such as the true-skill statistic（TSS） and Heidke skill score（HSS）, are adopted to evaluate the optimal thresholds and relative forecast skill for each modified index. Lastly, the modified indices are verified with a two-year independent dataset（2011–12）, showing satisfactory results for the modified indices. For determining whether or not a thunderstorm day is likely to occur, we recommend the modified SICCLindex.

Lightning Nowcasting with an Algorithm of Thunderstorm Tracking Based on Lightning Location Data over the Beijing Area

A thunderstorm tracking algorithm is proposed to nowcast the possibility of lightning activity over an area of concern by using the total lightning data and neighborhood technique.The lightning radiation sources observed from the Beijing Lightning Network(BLNET)were used to obtain information about the thunderstorm cells,which are significantly valuable in real-time.The boundaries of thunderstorm cells were obtained through the neighborhood technique.After smoothing,these boundaries were used to track the movement of thunderstorms and then extrapolated to nowcast the lightning approaching in an area of concern.The algorithm can deliver creditable results prior to a thunderstorm arriving at the area of concern,with accuracies of 63%,80%,and 91%for lead times of 30,15,and 5 minutes,respectively.The real-time observations of total lightning appear to be significant for thunderstorm tracking and lightning nowcasting,as total lightning tracking could help to fill the observational gaps in radar reflectivity due to the attenuation by hills or other obstacles.The lightning data used in the algorithm performs well in tracking the active thunderstorm cells associated with lightning activities.

A data complementary method for thunderstorm point charge localization based on atmospheric electric field apparatus array group

Data loss or distortion causes adverse effects on the accuracy and stability of the thunderstorm point charge localization.To solve this problem,we propose a data complementary method based on the atmospheric electric field apparatus array group.The electric field component measurement model of the atmospheric electric field apparatus is established,and the orientation parameters of the thunderstorm point charge are defined.Based on the mirror method,the thunderstorm point charge coordinates are obtained by using the potential distribution formulas.To test the validity of the basic algorithm,the electric field component measurement error and the localization accuracy are studied.Besides the azimuth angle and the elevation angle,the localization parameters also include the distance from the apparatus to the thunderstorm cloud.Based on a primary electric field apparatus,we establish the array group of apparatuses.Based on this,the data measured by each apparatus is complementarily processed to regain the thunderstorm point charge position.The results show that,compared with the radar map data,this method can accurately reflect the location of the thunderstorm point charge,and has a better localization effect.Additionally,several observation results during thunderstorm weather have been presented.

The Characteristics of Cloud-to-Ground Lightning Activity with Severe Thunderstorm Wind in South and North China

The characteristics of cloud-to-ground(CG) lightning activity with severe thunderstorm wind(STW) in South and North China are analyzed using CG lightning data, radar data, and serious weather reports. The percentage of positive CG(PCG) flashes with STW in North China is larger than that in South China. STW takes place during the period when the total CG and PCG density is increasing fastest. STW also occurs close to the high-value center of CG and PCG density. In North China, the CG and PCG density in the grid of STW maximizes approximately 20 minutes after the STW occurs; while in South China, the PCG density and percentage of PCG in the grid of STW maximizes about 10 minutes before the occurrence of STW. The high-value centers of CG density and PCG density in North China move slightly faster than those in South China, which is opposite to the rate of increasing CG activity.

Observations of NO_2 and O_3 during Thunderstorm Activity Using Visible Spectroscopy

Simultaneous observations for the total column densities of NO2,O3 and H2O were carried on using the portable Spectrometer (438-450 nm and 400-450 nm) and the visible Spectrometer (544.4-628 nm) during premonsoon thunderstorms and embedded hail storm activity at Pune (18°32'N & 73°51'E),India.These observations confirm the fact that there is an increase in O3 and NO2 column densities during thunderstorms.The increase in O3 was observed following onset of thunderstorm,while the increase in NO2 was observed only after the thunder flashes occur.This implies that the production mechanisms for O3 and NO2 in thunderstorm are different.The observed column density of NO,value (1 to 3×1017molecules cm-2) during thunderstorm activity is 10 to 30 times higher than the value (1×10th molecules cm-2) of a normal day total column density.The spectrometric observations and observations of thunder flashes by electric field meter showed that 6.4×1025molecules/flash of NO2 are produced.The increased to-oil column density of ozone during thunderstorm period is 1.2 times higher than normal (clear) day ozone concentration.The multiple scattering in the clouds is estimated from H2O and O2 absorption bands in the visible spectral region Considering this effect the calculated amount of ozone added in the global atmosphere due to thunderstorm activity is 0.26 to 0 52 DU,and the annual production of ozone due to thunderstorm activity is of the order of 4.02×10 molecules/year The annual NO2 production may be of the order of 2.02×1035molecules/year.