In the Asian and other monsoon regions of the world most of the severe weather observed is local or mesoscale in nature. Forecasting convective storms or mesoscale systems in the monsoon regions, especially in the tropics, has always been a challenging task to operational meteorologists. Maldives Islands, being situated in the tropical Indian Ocean, are affected by monsoon depressions and tropical cyclones. Thunderstorms and the passage of squall lines are well known sources of heavy rainfall. However, due to the lack of professional people and necessary equipment the weather systems around these islands are seldom studied. Therefore the aim of this thesis is to investigate whether the small islands can create sufficient perturbations in the mesoscale environment to result in the development of convective systems. In this regard, two numerical models, Weather Research and Forecasting model (WRF version 2.2.1) and Regional Atmospheric Modelling System (RAMS version 6.0) were used in this study. Two experiments were performed using the WRF model. In the first experiment, a case study was investigated where the selected day experienced heavy rainfall and thunderstorms. In the second experiment, the same case study was used but with the topographical and surface properties removed in order to investigate the influence of the island in modifying the mesoscale environment. All the experiments were initialized using the re-analysis data from NECP. WRF was able to predict the large scale synoptic features with reasonable accuracy when compared to the observations. Development of the boundary layer and the downstream advection of the temperature anomaly generated by the island were well represented. However, the magnitude of the effects was shown to be weak, probably due to the influence of large scale synoptic features. Even though the model was able to predict the large scale features and some of the mesoscale features, it did not predict any storm development and underestimated the precipitation. Therefore, it was decided to idealize the storm development using the RAMS model. RAMS model was used in a two-dimensional framework. The model was initialized horizontally homogenous using a single sounding and six simulations were performed. The simulation results clearly depicted that the small island can generate its own circulation and influence the mesoscale environment. The daytime heating of the island and the downstream advection of the temperature anomaly in a moist unstable atmosphere could trigger a thunderstorm later in the day. The storm becomes mature approximately 40-80 km offshore. This also suggests that triggering of a storm on one side of an atoll could influence the islands on the downstream side. Sensitivity of storm development to the thermodynamics showed that even with an unstable atmosphere, enough moisture in the lower and mid-troposphere is needed to trigger the storm. Sensitivity to the change of SST showed that convective development was suppressed with a drop of 1 oC. However, this needs further investigation. Assessment of sensitivity to the size of the island showed that the time of triggering of the storm was later and the scale of influence was smaller with a smaller island.

Neste trabalho foram estudadas as linhas de instabilidade (LI) que atingiram a RMSP, o desenvolvimento de tempo severo associados a estas e a possibilidade de utilizar alguns parâmetros de tempo severo na identificação destes eventos. Foram identificadas 185 linhas no período de 2002 a 2009 através da observação de imagens de satélite. Estas ocorrem durante todos os períodos do ano, embora a frequência seja maior nos meses de verão, em função da maior atividade convectiva nessa estação do ano. Foi identificado que 94 % das LI se deslocaram para leste enquanto apenas 6 % tiveram seu deslocamento para oeste. Testes sobre configurações mais adequadas para a simulação de LI com o modelo BRAMS indicaram que as parametrizações de convecção disponíveis no modelo não são capazes de representar os eventos de forma adequada, havendo a necessidade de aumento na resolução das simulações e forte dependência da parametrização de microfísica. Foi constatado que espaçamentos de grade de oito quilômetros são suficientes para uma boa representação dos eventos. Os testes também indicaram que, dentre as opções disponíveis no modelo, o uso da deformação Anisotrópica para o tratamento dos processos turbulentos é o mais indicado para a simulação de casos de LI. Os parâmetros convectivos mostraram-se bons indicadores de tempo severo na presença de sistemas de origem baroclínica, apresentando valores similares aos encontrados para o HN. Porém, em tempestades onde o efeito termodinâmico é o fator mais importante, como nas tempestades isoladas ou na presença de ZCAS os índices cinemáticos nem sempre são bons preditores de tempo severo. A análise dos parâmetros convectivos obtidos das radiosondagens lançadas na RMSP mostra que, em muitos casos observados de LI, não há valores indicativos de ocorrência de tempo severo, ressaltando a importância do uso de perfis obtidos da modelagem numérica, numa região de abrangência maior, para a obtenção desses índices. / A study about the squall lines that reached Metropolitan Area of São Paulo (MASP), the development of severe weather associated with these and the possibility of using some parameters in the identification of severe weather are presented. It was identified a total of 185 squall lines between 2002 and 2009 through the analysis of satellite images. These lines occur during all periods of the year, although the frequency of squall lines is higher in summer months, due to increased convective activity in this season. Also, it was identified that 94 % of the squall lines moved eastward while only 6% of then moved westward. Tests about the most appropriate configuration to the squall lines simulation with BRAMS model indicated that the convective parameterizations available in the model are not able to represent the events in an adequate manner, being necessary to use higher resolutions in the simulations, which creates a stronger dependence on microphysics parameterization. We note that grid spacing as low as eight kilometers are enough for a fair representation of the events. The tests also indicated that, among all the options available in the model for turbulent process the anisotropic deformation is the one that works better in simulating squall lines. The convective parameters proved to be good indicators of severe weather in the presence of baroclinic systems, with similar values to those found for the Northern Hemisphere. However in storms where the thermodynamic effect is the most important factor, as in isolated storms and in the presence of the Southern Atlantic Convergence Zone (SACZ), kinematics indices are not always good predictors of severe weather. The analysis of convective indices, obtained through radiosondes variables launched in the MASP, shows that in many squall line observed cases there is no indication of severe weather occurrence. This fact increases the importance of using the profiles provided by numerical modeling over a larger region to obtain such indices.

<p>This is a dissertation by Funing Li submitted to the Faculty of Purdue University in partial fulfillment of the requirements for the degree of Doctor of Philosophy.</p>

Thunderstorms are both dangerous as well as important rain-bearing structures for large parts of the world. The prediction of thunderstorm trajectories is however difficult, especially in tropical regions. This is largely due to their smaller size and shorter lifespan. To overcome this issue, this thesis investigates how well a neural network composed of long short-term memory (LSTM) units can predict the trajectories of thunderstorms, based on several years of lightning strike data. The data is first clustered, and important features are extracted from it. These are used to predict the mean position of the thunderstorms using an LSTM network. A random search is then carried out to identify optimal parameters for the LSTM model. It is shown that the trajectories predicted by the LSTM are much closer to the true trajectories than what a linear model predicts. This is especially true for predictions of more than 1 hour. Scores commonly used to measure forecast accuracy are applied to compare the LSTM and linear model. It is found that the LSTM significantly improves forecast accuracy compared to the linear model. / Åskväder är både farliga och livsviktiga bärare av vatten för stora delar av världen. Det är dock svårt att förutsäga åskcellernas banor, främst i tropiska områden. Detta beror till större delen på deras mindre storlek och kortare livslängd. Detta examensarbete undersöker hur väl ett neuralt nätverk, bestående av long short-term memory-lager (LSTM) kan förutsäga åskväders banor baserat på flera års blixtnedlslagsdata. Först klustras datan, och viktiga karaktärsdrag hämtas ut från den. Dessa används för att förutspå åskvädrens genomsnittliga position med hjälp av ett LSTMnätverk. En slumpmässig sökning genomförs sedan för att identifiera optimala parametrar för LSTM-modellen. Det fastslås att de banor som förutspås av LSTM-modellen är mycket närmare de sanna banorna, än de som förutspås av en linjär modell. Detta gäller i synnerhet för förutsägelser mer än 1 timme framåt. Värden som är vanliga för att bedöma prognosers träffsäkerhet beräknas för att jämföra LSTM-modellen och den linjära. Det visas att LSTM-modellen klart förbättrar förutsägelsernas träffsäkerhet jämfört med den linjära modellen.

Alabama, also known as the Dixie state, is no stranger to severe weather. Severe weather can occur during much of the year. Experienced local forecasters have long suspected that North and Central Alabama has its own tornado alley. Many of these forecasters have noticed storm tracks as well as tornado tracks to be similar to past historic events. Many questions have risen about the exact influential factors that cause convective initiation and tornadic development. For example the effects of terrain, water, and population on tornado climatology will be discussed in this study. The sometimes unreliable climatology of tornadoes will be addressed as well as the history of storm reporting. Tornado clusters were found and further explained regarding relationships with terrain, water, and population. Through this research, it is concluded that there are two distinct tornado regions that exist in North and Central Alabama.

Les fortes précipitations sont parmi les phénomènes météorologiques les plus dangereux pouvant causer des dégâts matériels, des blessés et des morts. Le programme de recherche HyMeX (Hydrological cycle of the Mediterranean eXperiment) s'intéresse à leur étude sur le bassin méditerranéen et plus particulièrement sa partie nord occidentale. Les travaux réalisés dans le cadre de cette thèse ont porté en particulier sur l'étude des mécanismes associés aux événements de fortes précipitations (High Precipitation Events, HPE) se produisant en Corse. Une climatologie des HPE en Corse sur une durée de 31 ans a été réalisée ainsi que l'étude détaillée de trois HPEs de l'automne 2012 pendant la campagne de mesures d'HyMeX. Ces trois cas d'études sont abordés par le biais de l'analyse des données et par celui de la modélisation. L'étude climatologique a montré que 173 HPEs (caractérisés par plus de 100 mm de précipitations en 24h) se sont produits en Corse sur la période 1985-2015. Ils sont principalement caractérisés par le fait qu'ils affectent plutôt la partie orientale de la Corse, plus particulièrement son orographie. Ces HPEs se produisent surtout de septembre à décembre avec un maximum en octobre. Une analyse en composantes principales a permis de classer ces événements en trois catégories. Les dépressions méditerranéennes chaudes d'automne, celles d'hiver froides, et une catégorie dite mixte associée aux dépressions atlantiques de grande échelle. Les précipitations les plus fortes sont observées quand l'orographie corse fait obstacle à un flux de sud-est chaud et humide. Les cas d'études présentés sont tous les trois différents en terme de mécanismes impliqués. Le cas du 4 septembre 2012 est associé à une dépression stationnaire donnant des précipitations sur toute la Corse avec un maximum sur le littoral et le relief de l'est de l'île. Celui du 31 octobre correspond à une dépression se déplaçant rapidement et induisant une évolution en plusieurs phases associée à un flux de basse couche initialement de sud-est tournant à l'ouest, associé à des précipitations d'abord convectives le long du relief oriental puis évoluant au fur et à mesure en pluies stratiformes sur l'ouest et le sud-est de l'île. Le dernier cas, du 23 octobre, est composé d'une ligne de cellules convectives résultant d'une convergence stationnaire au sud-est, sous le vent de la Corse. Les cellules convectives sont advectées vers l'île par le flux de sud-est de moyenne et haute altitude. Cette configuration permet la stationnarité de la ligne convective, provoquant un épisode de précipitations relativement court et très localisé. Les résultats de ce travail confirment que le modèle numérique Meso-NH permet de bien simuler ce type de phénomène avec une précision satisfaisante à une résolution horizontale de 2,5 km. Cependant, cette étude met également en évidence l'importance de la bonne représentation des conditions initiales. En outre, la distribution spatiale des précipitations dépend fortement de la représentation de l'orographie dans le modèle et de la résolution horizontale. Elle est améliorée quand on utilise une résolution de 500m. / Heavy precipitation is one of the primal meteorological reasons for property damage, injuries, and deaths. In the framework of the Hydrological Cycle of the Mediterranean (HyMeX) program, heavy precipitation is analyzed throughout the entire Mediterranean basin with a special focus on the northwestern Mediterranean. This work studies in particular the mechanisms of high precipitation events (HPEs) on Corsica. For this purpose, a 31 year (1985 - 2015) climatology of HPEs on Corsica is presented. In addition, three HPEs during autumn 2012 are analyzed in detail using observations and numerical modeling. A climatology of 173 events shows that the eastern half of Corsica, specifically the orography, is most affected by high precipitation events. The months from September to December, most of all October, are identified as most prone to heavy precipitation events over Corsica. A principal component analysis is used to classify the events into three categories, which correspond to warm autumn and cold winter Mediterranean cyclones as well as a mixed category which contains also larger scale Atlantic cyclones. The heaviest precipitation is observed when warm moist southeasterly flow encounters the Corsican orography. In addition, three case studies are presented, each with different mechanisms involved. A stationary cyclone on 4 September 2012 led to widespread precipitation over Corsica with the most intense rain observed over the east of the island, along the coast and the orography. On 31 October, a fast moving cyclone caused a multi-phase event, which was characterized by low level wind turning from southeast to west while precipitation gradually changed from convective along the orography in the east of the island to stratiform mainly over the west and southwest. The last event, 23 October 2012, was comprised of a line convective cells which formed over stationary lee side convergence southeast of Corsica. The convective cells were advected toward the island by the mid- and upper level southeasterly wind. These conditions allowed the convective line to remain stationary, resulting in a highly localized and relatively short event. The findings confirm that the numerical model Meso-NH is well capable of simulating such events with satisfactory precision at a grid spacing of 2.5 km. However, the studies also underline the importance of well captured initial conditions. Additionally, the spatial distribution of precipitation is highly dependent on the representation of the orography in the model as well as the horizontal grid spacing and is improved when using a horizontal grid spacing of 500 m instead.

Tempestades severas ocorrem na Região Metropolitana de São Paulo (RMSP) principalmente nos meses quentes e úmidos do ano. Nesta tese, os mecanismos de disparo da convecção profunda são estudados por meio de análises observacionais e simulações numéricas com o Advanced Regional Prediction System (ARPS). A metodologia proposta compreende o uso da parametrização microfísica fria na simulação dos processos físicos que levam à formação de nuvens cumulonimbus, sem o uso da parametrização de cúmulos nas grades de altíssima resolução espacial. Nos eventos estudados, as primeiras células de precipitação observadas e simuladas surgiram em razão da interação entre o escoamento atmosférico na camada limite planetária e a topografia local. As células secundárias foram geralmente mais intensas, uma vez que elas surgiram após o aquecimento diabático adicional. O mecanismo de disparo das células secundárias foi a corrente ascendente induzida pela propagação horizontal das frentes de rajada em baixos níveis da atmosfera das correntes descendentes das células primárias. As frentes de rajada tiveram velocidade de propagação horizontal típica de 6 m s-1. No evento de 02 de fevereiro de 2004, células convectivas profundas foram simuladas com alto grau de realismo no domínio da grade de 3 km de resolução espacial. Observou-se que, neste caso, a frente de brisa marítima pôde atuar como guia de ondas para a colisão entre duas frentes de rajada. A propagação da frente de brisa marítima para o interior do continente ocorreu em conjunção a um forte gradiente de vapor dágua nos níveis mais baixos da troposfera. As células convectivas profundas secundárias surgiram e se desenvolveram exatamente nesta zona de interface, a qual representa o contraste entre as diferentes massas de ar marítima e continental. No evento de 04 de fevereiro de 2004, na grade de 1 km de resolução, a análise objetiva com as medidas das estações de superfície na RMSP correspondente às 1800 UTC indicou a presença de uma ilha de calor urbana com até 4 oC de aquecimento diferencial entre a Capital e vizinhanças. O principal efeito da assimilação destas medidas foi a redução do NCL em até 80 hPa, o que favoreceu o disparo da convecção naquela área. / Severe thunderstorms occur in the Metropolitan Area of São Paulo (MASP) mainly in the warm and wet months of the year. In this work, the triggering mechanisms of deep convection are studied through observed data and numerical simulations with the Advanced Regional Prediction System (ARPS). The proposed methodology focuses in the use of microphysics parameterization of cold clouds to simulate physical process linked to the life cycle of thunderstorms. The cumulus cloud parameterization isnt used in high resolution numerical grids. In the real case studies, both observed and simulated, early convective cells developed as a consequence of the interaction between the planetary boundary layer atmospheric flow and the local topography. The secondary convective cells were generally strongest, once they developed after additional surface diabatic heating. The triggering mechanism of these secondary cells was the updraft induced by gust fronts generated by downdrafts of primary cells. The gust fronts had a typical horizontal propagation velocity of 6 m s-1. In the February 02 2004 event, deep convective cells were simulated with high degree of realism with a 3 km resolution grid. It was observed that, in this case, the sea-breeze front could act as a wave guide to the collision between two different gust fronts. In addition, the sea breeze front propagated to the continental area together with a strong low level water vapor gradient. The secondary deep convective cells arose and developed exactly on this interface zone, which represents the contrast between the oceanic and continental air masses. The interface zone was marked by a water vapor mixing rate of 14 g kg-1. In the February 04 2004 event, the objective analysis, made with some MASP´s surface stations measurements at 1800 UTC in the 1 km resolution grid, indicates the presence of an urban heat island with up to 4 oC of differential heating between São Paulo city and its neighboring area. The main effect in assimilating these surface measurements was the lowering of the lift condensation level up to 80 hPa, which favored the triggering of convection in that area.

En France les filtres plantés à écoulement vertical sont utilisés pour le traitement des surverses de déversoir d’orage (CSO en anglais). Ils sont efficaces pour le traitement des polluants ainsi que pour assurer un rôle de tampon hydraulique permettant de protéger les milieux hydriques superficiels. Ils ont la particularité de recevoir des effluents bruts dont la fréquence, l’intensité, la durée et les concentrations sont stochastiques. En conséquence de quoi l’optimisation du dimensionnement est particulièrement délicate. Dans ce travail, le dimensionnement est abordé suivant deux approches. D’une part en définissant les processus clefs pour une optimisation du dimensionnement et, d’autre part, en construisant un modèle dynamique simplifié (Orage)permettant de prendre en compte, sur le long terme, le caractère stochastique des évènements et les contraintes locales. Le cœur du modèle a été calibré et validé sur une large gamme d’évènements mesurées sur un site en taille réelle (site de Marcy l’Etoile). La qualité de la calibration a été validé aussi bien visuellement que statistiquement et la robustesse du modèle étudiée par une analyse de sensibilité (méthode de Morris). L’optimisation automatique du dimensionnement a nécessité la réalisation d’une boucle itérative également paramétrée et testée dans le cadre du site de Marcy l’Etoile et pour des cas théoriques. Les dimensionnements proposés ont été estimés réalistes. La définition des processus clefs du système de traitement, et le développement du modèle simplifié, a également été possible grâce à la simulation d’expérimentations sur colonnes par un modèle mécaniste (HYDRUS / CW2D). La paramétrisation des processus du modèle simplifié a été réalisée sur la base d’expérimentations détaillées sur le site de Marcy l’Etoile. Le site a été suivi pendant trois ans aussi bien par des prélèvements automatiques, des traçages, que des mesures en lignes (hydraulique, azote, teneur en eau). Parmi les processus calibrés, nous retiendrons i) l’importance du niveau de saturation de la base du filtre au début d’un évènement vis-à-vis des court circuits hydrauliques, ii) la définition des capacités d’adsorption de l’azote ammoniacale de différents matériaux (pouzzolane, mélange de sable et de zéolite) et iii), la définition des cinétiques de nitrification suivant la température et quantité d’azote ammoniacal adsorbé. Si les performances de l’ouvrage sont hautes pour les paramètres majeurs (MES, DCO, N-NH4), l’impact du dimensionnement, de la gestion et des facteurs environnementaux sur les performances et l’accumulation de boue a été étudié. De plus, quelques suivis ont été réalisés sur les métaux et les HAP. Ces mesures ont permis non seulement de calibrer le logiciel Orage mais aussi de proposer des modifications pour l’amélioration du dimensionnement. / Constructed wetlands for combined sewer overflow treatment (CSO CWs) are vertical flow filters in France. They effectively remove pollutants and mitigate hydraulic peaks, which protects natural waters. They receive unsettled flows with stochastic return periods, volumes and concentrations, making design optimization difficult. In the presented work, design is targeted from two sides. First, the scientific basis is laid down for the design-support software Orage. The tool is model-based and as such, it considers the stochasticity of CSOs site-specifically. Its core model has been calibrated and then verified by a range of tests. Simulations fit to measured data from a full-scale CSO site (Marcy l’Etoile). The goodness of fit is evaluated visually and statistically. Model robustness has been confirmed by a sensitivity analysis (Morris method). After this, the iterative shell has been parameterized and tested (the software element doing automatic optimization). The design proposals of the tool have been found realistic. The other approach to target optimization has been a detailed field research at Marcy l’Etoile. The site at Marcy l’Etoile has been monitored for three years, using automatic samplers and online probes. The filling of the porous media at event start proves and explains shortcutting behaviour. Adsorption capacities have been quantified for pozzolana and a sand and zeolite mixture. A temperature-dependent equation is calibrated to calculate nitrification rate. System efficiency is high for target pollutants (TSS, COD, NH4-N). Design, operational and environmental factors have been analysed to seek potential effects on removal performances and sludge accumulation. Additionally, PAHs and metals are indicated for a few selected events. The field results were essential to calibrate Orage and to see options for design improvements. The understanding of CSO CWs and the development of Orage was promoted also by simulating lab-scale columns using the process-based model package HYDRUS / CW2D.

On January 29-30, 2008 a squall line of thunderstorms moved through the Ohio Valley resulting in four deaths and one injury. Such events highlight the importance of accurate forecasting for public safety. Mesoscale Modeling plays an important role in any forecast of a potential squall line. The focus of this study was to examine the performance of several parameterization scheme combinations in the Weather Research and Forecasting Model version three (WRF) as they related to this event. These examinations included cloud microphysics (WRF Single-Moment 3-class, 6-class, and Goddard), cumulus parameterization (Kain-Fritsch and Bets-Miller-Janjic) and planetary boundary layer schemes (Yonsei-University and Mellor-Yamada-Janjic). A total of 12 WRF simulations were conducted for all potential scheme combinations. Data from the WRF simulations for several locations in south central Kentucky were analyzed and compared using Kentucky Mesonet observations for four locations: Bowling Green, Russellville, Murray and Liberty, KY. A fine model resolution of 1 km was used over these locations. Coarser resolutions of 3 km and 9 km were used on the outer two domains, which encompassed the Ohio and Tennessee Valleys. The model simulation performance was assessed using established statistical measures for the above four locations and by visually comparing the North American Regional Reanalysis dataset (NARR) along with modeled simulations. The most satisfactory scheme combination was the WRF Single-Moment 3-class Microphysics scheme, Kain-Fritsch cumulus parameterization scheme and Yonsei University scheme for the planetary boundary layer. The planetary boundary layer schemes were noted to have the greatest influence in determining the most satisfactory model simulations. There was limited influence from different selections of microphysics and cumulus parameterization schemes. The preferred physics parameters from these simulations were then used in six additional simulations to analyze the affect different initialization data sets have with regards to model output. Data sets used in these simulations were the Final Operational Analysis global data, North American Regional Reanalysis (3 and 6 hour) and the North American Mesoscale Model at 1, 3 and 6 hour timesteps, for a total of six simulations. More timesteps or an increase in model resolution did not materially improve the model performance.

From the Proceedings of the 1972 Meetings of the Arizona Section - American Water Resources Assn. and the Hydrology Section - Arizona Academy of Science - May 5-6, 1972, Prescott, Arizona / There have been many efforts in recent years to modify thunderstorms through cloud seeding. Collective cloud seeding efforts in Arizona before 1971 are reviewed and an operational convective cloud seeding program carried out in Arizona in the summer of 1971 is analyzed. The comprehensive Santa Catalina cloud seeding experiment (1957 to 1964) was a randomized seeding using silver iodide. Results of this experiment are uncertain as numerous interpretations are possible. Numerous individual experiments from 1966 to 1970 at flagstaff were conducted, with uncertain results. An intensive program of seeding individual cumulus clouds with silver iodide was carried out in the summer of 1971 in central and eastern Arizona. No statistically significant changes were noted. Results of the Catalina experiment imply that seeding decreased rainfall on and downwind from the target. Two other experiments were inconclusive. Nine figures show precipitation patterns.