Networks of small, often mobile, polarimetric radars are gaining popularity in the hydrometeorology community due to their rainfall observing capabilities and relative low purchase cost. In recent years, a number of installations have become operational around the globe. The problem of signal attenuation by intervening rainfall has been recognized as the major source of error in rainfall estimation by short-wavelength (C-, X, K-band) radars. The simultaneous observation of precipitation by multiple radars creates new prospects for better and more robust attenuation correction algorithms and, consequently, yields more accurate rainfall estimation. The University of Iowa hydrometeorology group's acquisition of a network of four mobile, polarimetric, X-band radars has resulted in the need for a thoughtful evaluation of the instrument. In this work, we use computer simulations and the data collected by The University of Iowa Polarimetric Radar Network to study the performance of attenuation correction methods in single-radar and network-based arrangements. To support the computer simulations, we developed a comprehensive polarimetric radar network simulator, which replicates the essential aspects of the radar network rainfall observing process. The simulations are based on a series of physics- and stochastic-based simulated rainfall events occurring over the area of interest. The characteristics of the simulated radars are those of The University of Iowa Polarimetric Radar Network. We assess the correction methods by analyzing the errors in reflectivity and rainfall rate over the area of interest covered by the network's radars. To enable the implementation of the attenuation correction methods to the data collected by The University of Iowa Polarimetric Radar Network, we first developed a set of utilities to assist with efficient data collection and analysis. Next, we conducted a series of calibration tests to evaluate the relative calibration and channel balance of the 2 network's radars. Finally, in an attempt to verify the results obtained via computer simulations, we applied the set of attenuation correction algorithms to the data collected by The University of Iowa Polarimetric Radar Network.

L'objectif de cette thèse était d'apporter une approche géophysique non linéaire à l'hydrologie urbaine. Elle a visé l'étude de la mise à l'échelle et de l'intermittence de la non-linéarité des précipitations, la réalisation d'une méthode de prévision stochastique à très court terme ("nowcast"), ainsi que son application aux processus hydrologiques dans les environnements (semi-) urbains. La partie modélisation hydrologique globale concerne la vallée de la Bièvre, zone semi-urbanisée de 110 km2 dans le sud-ouest de la région parisienne. Par conséquent, trois études différentes ont été réalisées dans cette zone à l'aide de deux modèles hydrologiques : le modèle conceptuel semi-distribué InfoWorks CS appliqué sur tout le bassin versant de Bièvre ; et le modèle physique complètement distribué Multi-Hydro, développé à l'École des Ponts ParisTech, appliqué sur deux sous-bassins versants de la Bièvre. Les principaux objectifs étaient de mieux comprendre les impacts de la variabilité spatio-temporelle des données pluviométriques en utilisant deux produits (les données radar bande-C de Météo-France avec une résolution de 1 km x 1 km x 5 min, et les données radar DPSRI band-X de l'ENPC à une résolution de 250 m x 250 m x 3.41 min) comme entrées pour les modèles, et d'identifier les capacités de chaque modèle pour traiter des données à une meilleur résolution, telles que la bande-X. Ensuite, les résultats obtenus démontrent que la fiabilité des simulations hydrologiques dépend intrinsèquement des caractéristiques des données pluviométriques. De plus, les données du radar bande-X pourraient mesurer des pics de précipitations plus élevés et le modèle complètement distribué était plus sensible à une meilleure résolution des données pluviométriques. Par la suite, des données de pluie provenant des radars météorologiques situés à des endroits complètement différents (Brésil, France, Japon) ont été analysées et comparées statistiquement afin d'améliorer la compréhension générale du comportement scalant des précipitations. De plus, le théorème d'intersection a été appliqué pour mettre en évidence les impacts de la variabilité spatiale d'un réseau virtuel de pluviomètres, qui a été généré en considérant l'emplacement des centres de masse de chaque sous-bassin versant de la vallée de la Bièvre. Ainsi, il a été possible d'identifier que la fractalité du réseau virtuel a conduit à une perte d'information importante des champs de pluie, biaisant leurs statistiques. Cela indique que le processus commun (largement retrouvé dans la littérature) de calibration des données radar à l'aide de pluviomètres devrait correctement prendre en compte cette fractalité. Enfin, une nouvelle approche de prévision stochastique immédiate a été proposée, à l'aide du modèle des multifractals universels (UM) en cascades continues. Cette méthode a été appliquée aux données des radars pluviométriques de la région amazonienne brésilienne et de Paris. Bien qu'il soit encore en développement et nécessite quelques améliorations, les premiers résultats obtenus avec ce modèle de prévision présenté ici sont vraiment encourageants et corroborent une fois de plus le besoin de données à haute résolution spatio-temporelle pour faire face aux crues soudaines / The focus of this thesis was to bring a nonlinear geophysical approach to urban hydrology. It aimed the study of rainfall non-linearity scaling and intermittency, achieving a stochastic very short-range forecast (nowcast) method, as well as its application to hydrological processes in (semi-) urban environments. The overall hydrological modelling part concerned the Bièvre Valley, which is a 110 km2 semi-urbanized area in the southwest of Paris region. Therefore, three different studies were performed within this area using two hydrological models: the conceptually-based semi-distributed model InfoWorks CS over the total Bièvre catchment, and the physically-based fully-distributed model developed at École des Ponts ParisTech called Multi-Hydro over two sub-catchments. The main goals were to better understand the impacts of spatio-temporal variability of rainfall data by using two products (the Météo-France C-band radar data with a resolution of 1 km x 1 km x 5 min; and the ENPC DPSRI X-band radar data at a 250 m x 250 m x 3.41 min resolution) as input to the models, and to identify the capacities of each model to deal with better resolution data, such as the X-band one. Then, the obtained results demonstrate that the reliability of the hydrological simulations are intrinsically dependent on rainfall data features. Moreover, the X-band radar data could measure higher peaks of rainfall rates and the fully-distributed model was more sensitive to better resolution rainfall data. Afterwards, different weather rainfall radar data from completely different sites (Brazil, France, Japan) were statistically analysed and compared in order to improve the general comprehension of rainfall scaling behaviour. In addition, the Intersection Theorem was applied to highlight the impacts of spatial variability of a virtual rain gauge network. The latter was generated by considering the location of each Bièvre Valley sub-catchment mass centre. Thus, it was possible to identify that the fractality of the virtual network led to an important information loss of the rainfall fields, biasing their statistics. This indicates that the common process (largely found in literature) of radar data calibration using rain gauges should be properly take into account this fractality. Finally, a new stochastic nowcast approach was proposed, using the continuous in scale cascade Universal Multifractals (UM) model. This method was applied to weather rainfall radar data from the Brazilian Amazon region and Paris. Although it is still under development and needs some improvements, the first results obtained with this forecast model presented here in this thesis are really encouraging and once more corroborate to the need of high spatio-temporal resolution data to cope flash floods

L'objectif principal de cette thèse était de parvenir à un outil de gestion fiable des bassins de rétention d'eaux pluviales en utilisant les données radar en bande X. Il s’est avéré que cela nécessite plusieurs développements de recherche. Le cas d’étude considéré comprend un bassin de 10000 m3 situé en Val-de-Marne et construit en aval d'un bassin versant urbain de 2.15 km2. Il assure un double rôle de traitement des eaux pluviales et de prévention des inondations par stockage du volume. Opérationnellement les modes de gestion associés à chacun de ces objectifs sont antagonistes si bien qu’une gestion prédictive a été mise en place ; exploitation routinière en mode anti-pollution et basculement vers le mode anti-inondation en cas de besoin. Il doit se faire sur la base d’une connaissance sûre de la situation pluvieuse prévue à court terme. Une façon courante de répondre aux besoins opérationnels de la gestion prédictive est de mettre en place un système d’alerte basé sur l’utilisation des données radar. Le système CALAMAR par exemple, repose sur l’utilisation des données radar brutes à mono polarisation du réseau radar de Météo-France; traitées avec des méthodes de conversion classiques Z-R et une calibration avec des pluviomètres. Cependant, la fiabilité de ce système fait débat, notamment vis-à-vis de la qualité de la mesure radar obtenue. Une nouvelle méthodologie de comparaison de produits radar a été développée au cours de cette thèse. Elle repose sur le cadre théorique des multifractals et permet une comparaison de la structure et de la morphologie des champs de précipitations dans l'espace et le temps à travers les échelles. Cette méthode a d'abord été appliquée sur les produits CALAMR et Météo-France, puis, pour confirmer certains des résultats, sur les premières données d’un radar bande X, acquis par l’Ecole des Ponts ParisTech dans le cadre du projet Européen RainGain et fournissant des mesures de précipitations à des échelles plus fines (jusqu’à 100m en espace et 1 min en temps). Les résultats obtenus mettent en évidence non seulement l'influence cruciale des méthodes de traitement des données brutes sur la variabilité spatio-temporelle à travers les échelles, mais permettent également de prédéfinir les conditions dans lesquelles la calibration CALAMAR peut aggraver la qualité des mesures. Elles seraient très difficiles à détecter par les méthodes classiques largement répandues, n’impliquant qu’un nombre très limité de pixels radar (seulement ceux correspondants aux pluviomètres au sol). Des extensions de la méthodologie proposée ouvriront de nouveaux horizons pour la calibration des données de pluie. Alors que la littérature scientifique, notamment autour expériences TOMACS au Japon et CASA aux Etats-Unis, souligne l’importance opérationnelle d’une mesure de pluie plus détaillée grâce au radar en bande X, son impact sur les performances des modèles hydrologiques fait encore débat. Les recherches antérieures, basée pour la plupart sur des modèles conceptuels, ne sont pas concluantes. Ainsi pour dépasser ces limites, nous avons utilisé deux modèles impliquant des approches de modélisation différentes : CANOE (semi-distribué et conceptuel) et Multi-Hydro (distribué et à base physique ; développé à l’ENPC). Une version opérationnelle de CANOE et une nouvelle configuration plus fine améliorant considérablement la sensibilité du modèle à la variabilité de la pluie ont été utilisées. Plusieurs développements ont été apportés à Multi-Hydro, y compris une optimisation de sa résolution, ce qui améliore grandement l'ensemble de ses fonctionnalités. Il ressort de ce travail qu’en prenant en compte la variabilité spatio-temporelle des précipitations à petite échelle, la performance des modèles hydrologiques peut être augmentée jusqu'à 20%.Nous pensons que cette thèse a contribué à la mise au point de nouveaux outils opérationnels, fiables ayant la capacité de prendre en compte les données en bande X haute résolution / The main goal of this thesis was to achieve a reliable management tool of storm water storage basins using high resolution X-band radar. It turned out that it required several research developments. The analysed case study includes a retention basin of 10000 m3 located in Val de Marne county downstream of a 2.15 km2 urban catchment. It has a twofold goal: storm water decontamination and flood protection by volume storage. Operationally the management strategies associated with these two aims are conflicting; hence, a predictive management has been set up: a routine exploitation of the basin in the anti-pollution mode, and a switch to the flood protection mode when needed. It should be based a reliable knowledge of short-term rainfall forecasts. A common way to respond to operational needs of the predictive management is to set up a warning system based on the use of radar data. For example, the CALAMAR system relies on the use of single-polarization raw radar data, coming from Meteo-France radar network, being processed with the conventional Z-R conversion methods followed by a calibration with rain gauge. However, the reliability of such warning systems has been subject to debate, often due to a questionable quality of the resulting radar rainfall estimates, compared to local rain gauges. Therefore a new methodology for more meaningful comparison of radar rainfall field products was developed during this PhD project. Being rooted to the multifractal theory, it allows a comparison of the structure and the morphology of rainfall fields in both space and time through scales. It was initially tested on CALAMAR and Meteo-France rainfall products before being applied for results confirmation on initial data from a X band radar, acquired by Ecole des Ponts ParisTech in the framework of the European project RainGain and providing data at higher resolution (up to 100 m in space and 1 min in time). The obtained results not only highlight the crucial influence of raw data processing on the scaling behaviour, but also permit to pre-define the conditions when the CALAMAR optimization may worsen the quality of rainfall estimates. Such conditions would be very difficult to detect with widely used conventional methods, which rely on a very limited number of radar pixels (only those containing rain gauges). Further extensions of the proposed methodology open new horizons for the rainfall data merging. While the scientific literature, notably around the TOMACS experiment in Japan and CASA one in the United States, highlights the operational benefits of higher resolution rainfall measurements thanks to X-band radars, its impact on the performance of hydrological models still remains a subject of debate. Indeed previous research, mainly based on conceptual models remains inconclusive. To overcome these limitations, we used two models relying on two very distinct modelling approaches: CANOE (semi-distributed and conceptual) and Multi-Hydro (fully distributed and physically based research model developed at ENPC). An operational version of CANOE and a new much finer configuration, which increases the sensitivity of the model to spatio-temporal variability of small-scale rainfall, were used. Several extensions of the Multi-Hydro were developed, including an optimization of its resolution, which greatly improves its whole functionality. It appears from this work that by taking into account the spatial and temporal variability of small-scale rainfall, the performance of hydrologic models can be increased up to 20%.Overall, we believe that this dissertation contributes to the development of new, reliable, operational tools to use in their full extent the high-resolution X-band data

[pt] A principal meta desta tese é estudar os efeitos da chuva nos enlaces operando na faixa de micro-ondas e comprimentos de ondas milimétricas. Para realizar este estudo, é considerado o modelo de chuva que considera um meio de chuva realista composto por um conjunto de gotas com a relação formato-tamanho proposta por Chuang e Beard, uma distribuição de tamanho das gotas dada por de Wolf, o índice de refração complexo da água para uma frequência e temperatura dada sugerido por Ray e uma distribuição de orientação dos eixos de simetria da partícula. O Extended Boundary Condition Method (EBCM) foi aplicado ao modelo descrito para determinar a atenuação, depolarização e espalhamento devidos à chuva. O desenvolvimento foi validado com sucesso por intermédio de comparações de seus resultados com os correspondentes disponíveis na literatura. O modelo de chuva realista foi utilizado em duas aplicações diferentes. Na primeira, foi estudada a interferência devida à chuva entre enlaces de telecomunicações sem fio operando em frequências de ondas milimétricas em ambientes urbanos. Outra aplicação envolve a determinação da taxa de precipitação por intermédio de radares meteorológicos (em particular, radares banda-X). Considerando que seu custo é relativamente baixo e sua resolução elevada, os radares em banda-X estariam entre as melhores opções para monitorar eventos meteorológicos. Entretanto, são susceptíveis à atenuação devida a gases atmosféricos e chuva ao longo dos enlaces, que impedem que a taxa de precipitação seja estimada diretamente a partir da potência recebida correspondente a uma determinada posição. Desta forma, um modelo de chuva realista foi implementado para calcular a seção reta de retroespalhamento e estimar a atenuação específica por intermédio do EBCM em cada um dos volumes existentes entre o radar e a posição selecionada. Este desenvolvimento permite a correção dos efeitos da atenuação existente no enlace formado entre estas duas posições. / [en] The main goal of this research is to study the rain effects on microwave and millimeter wave radio links. Thus, the rain-induced attenuation, depolarization and scattering are studied. To carry out this study, a realistic rain model is proposed, which consider a realistic rain medium composed by a cluster of raindrops with the shape-size relation proposed by Chuang and Beard, a raindrop size distribution given by de Wolf, index of refraction of water for a given temperature and frequency suggested by Ray and a distribution of the orientation angle of the symmetry axis. The realistic rain model is evaluated with two different applications of systems operating at microwave and millimeter wave frequencies. One of the applications involves wireless telecommunication systems, which are strongly affected by the presence of precipitation. To design an efficient radio communication system, the realistic rain model is applied for the analysis and quantification of rain-induced effects on links operating at millimeterwave frequencies in urban environments. Another application involves weather radars (X-band radars in particular). Considering their relatively low cost and high resolution, X-band radars would be among the best options to monitor meteorological events. However, they are susceptible to attenuation by fog, snow or rain. To solve this problem, a realistic and improved rain model is implemented to compute backscattering cross sections and estimate rain attenuation at each range gate. The proposed method is evaluated using radar data provided by the CASA OTG X-band (lambda equal a 3cm) radar located in Mayaguez, Puerto Rico, and X-band radar METEOR 50DX –Selex located in Belém, Brazil.

The climate adaptation of our cities and the optimization of our technical systems with regards to weather sets high demands on the availability and the processing of weather data. The possibility to forecast disturbances of influent flow rate to wastewater treatment plants allow control systems counteract these disturbances before they have a harmful effect on the treatment processes. These forecasts can be made by different models A neural network models complex patterns between different data sets through a multi-layered structure containing a large amount of transformation functions. The aim of this project was to examine how the complex neural network performed compared with a simpler linear regression model when forecasting wastewater flow using high resolution X-band rain radar data. The study also investigated to what extent X-band rain radar data contributes to the performance of the model. The performance was evaluated at rain flow periods only. Wastewater flow data were provided by Avedøre wastewater treatment plant in Copenhagen operated by BIOFOS. The X-band rain radar data was provided by HOFOR. The neural network was developed by Informetics on the TensorFlow platform. This project concluded that the neural network and the linear regression model performed equally well at predicting when a rain flow period began. The neural network was more accurate at predicting the flow rate while the linear regression was better at approximating the accumulated flow over an entire rain flow period. Using additional rain data up to 30 km within the radar station location in comparison with using data only from within the catchment indicated a 20 to 30-minutes improvement of possible lead time. A conceivable lead time when forecasting the sewer flow to Avedøre wastewater treatment plant was estimated to be around 4 hours. / Det föreligger höga krav på tillgänglighet och bearbetning av väderdata för att kunna optimera tekniska system i förhållande till väder och klimat. Att kunna förutsäga ändrat inkommande flöde till avloppsreningsverk möjliggör för kontrollsystem att kunna motverka negativa konsekvenser på reningsprocesserna på grund av det ändrade flödet. X-band radardata kan användas för att prognoser av flöden med hjälp av olika modeller.Ett neuralt nätverk, reproducerar komplexa mönster mellan olika dataset genom en struktur med flera lager och en mängd överföringsfunktioner.  Målsättningen med det här projektet var att utvärdera hur ett komplext neuralt nätverk presterar jämfört med en enklare regressionsmodell i att förutsäga avloppsflöde med hjälp av högupplöst X-band radardata. I projektet undersöktes också hur tillgång av olika radardata kunde bidra till modellens prestanda. Modellerna utvärderades endast under regnflödesperioder. Data över avloppsflödet som användes i projektet kom från Avedøre avloppsreningsverk i Köpenhamn. Reningsverket drivs av BIOFOS. Radardata kom från HOFOR. Det neurala nätverket som användes har utvecklats av Informetics på plattformen Tensorflow. Slutsatser som kunde dras i projektet var att det neurala nätverket och den linjär regressionsmodellen var lika bra på att förutsäga när en regnflödesperiod startade. Det neurala nätverket kunde förutsäga det momentana flödet bättre än regressionsmodellen, medan det omvända gällde för att uppskatta den totala flödesvolymen under en hel regnflödesperiod. Genom att använda ytterligare regndata, upp till 30 kilometer från radarstationen, jämfört med att endast använda data från avrinningsområdet kunde en 20–30 minuters förbättring av den möjliga prognostiden påvisas. En tänkbar prognostiden för att förutsäga avloppsflödet till Avedøre avloppsreningsverk visades ligga omkring 4 timmar.