[EN] Water is the foundation for all biological life on Earth and one of the basic links between the biosphere and atmosphere. It is equally fundamental for humans and nature (Tolba, 1982). In an environment of growing scarcity and competition for water, increasing the understanding of all fluxes of the water cycle lies at the heart of the scientific community's goals. Traditionally, water and vegetation have been considered as different systems. However, it is necessary to take a holistic approach which considers the question of the water cycle in an integrated manner by taking into account both: blue water and green water (Birot et al., 2011). Around this idea, the new discipline Ecohydrology emerged in the early 20th century and, from then; it has grown steadily as shown by the increasing number of research lines and scientific papers related to this new field. However, most of the current hydrological models includes the vegetation as static parameter and not as state variable. There are some exceptions taking explicitly the vegetation as state variable but in those cases, the models' complexity and parametrical requirements increase substantially. In practice, we have to deal against the 'data scarcity - high parametrical requirements' issue really often. To reduce that issue, two strategies can be applied: (1) simplification of the models' conceptual scheme and (2) increase of data availability by incorporating new sources of information. In this thesis, we explored the use of a distributed parsimonious ecohydrological modelling (with low parametrical requirements) calibrated and validated exclusively with remote sensing data. First, we used the parsimonious ecohydrological model proposed by Pasquato et al. (2015) in an experimental plot located in a semi-arid Mediterranean forest. The results in this previous stage suggested that the model was able to adequately reproduce the dynamics of vegetation as well as the soil moisture variations. In other words, it has been shown that a parsimonious model with simple equations can achieve good results in general terms. But, as long as we applied the model at plot scale, the challenging task to reproduce the spatial variation of the vegetation and water cycle remained. To explore the spatio-temporal variation of the vegetation and the water cycle, the distributed version of the parsimonious ecohydrological model used previously was applied in a basin located in Kenya, concretely in the Upper Ewaso Ngiro River basin. In order to explore the potential applicability of the satellite data, we calibrated the model using exclusively the NDVI provided by NASA. First of all, we had to deal with the fact that we were not calibrating the model with only one temporal series such as historical streamflow as usual. In fact, satellite data is composed by one temporal series per pixel. We had to identify how to use spatio-temporal (and not only temporal) data during models' calibration and validation. In that sense, unfortunately, there is still a deep lack in literature. A methodology based on the use of Empirical Orthogonal Function analysis was proposed and successfully applied. This experience provided amazing and promising results. The obtained results demonstrated that: (1) satellite data of vegetation dynamics contains an extraordinary amount of information that can be used to implement ecohydrological models in scarce data regions; (2) the proposed semi-automatic calibration methodology works satisfactorily and it allows to incorporate spatio-temporal data in the model parameterization and (3) the model calibrated only using satellite data is able to reproduce both the spatio-temporal vegetation dynamics and the observed discharge at the outlet point. It is important to highlight the positive consequences of this last result particularly in ungauged basins where the use of satellite data could be an alternative in order to obtain a proxy of the streamflow at outlet point. / [ES] El agua es la base de toda vida biológica en la Tierra y uno de los enlaces básicos entre la biosfera y la atmósfera. Es igualmente fundamental para los seres humanos y la naturaleza (Tolba, 1982). Tradicionalmente, el agua y la vegetación se han considerado como sistemas diferentes pero es claramente necesario tomar un enfoque holístico que considere la cuestión del ciclo del agua de una manera integrada, teniendo en cuenta tanto el agua azul como el agua verde (Birot et al., 2011). Alrededor de esta idea surgió la nueva disciplina llamada Ecohidrología a principios del siglo XX y desde entonces, no ha dejado de crecer tal y como demuestran el creciente aumento de líneas de investigación y publicaciones científicas relacionadas con este nuevo campo. Sin embargo, la mayoría de los modelos hidrológicos actuales incluye la vegetación como un parámetro estático y no como una variable de estado. Hay algunas excepciones que toman explícitamente la vegetación como variable de estado, pero en esos casos, la complejidad y el número de parámetros a determinar de los modelos aumentan sustancialmente. En la práctica, tenemos que hacer frente a la temible combinación de "escasez de datos - alto número de parámetros a determinar" con mucha frecuencia. Para reducir este problema, se pueden aplicar dos estrategias: (1) simplificar la complejidad conceptual de los modelos y así reducir el número de parámetros a calibrar, y/o (2) aumentar la disponibilidad de datos mediante la incorporación de nuevas fuentes de información. En esta tesis, hemos explorado el uso de un modelo ecohidrológico distribuido y parsimonioso (con pocos parámetros a determinar) que ha sido completamente calibrado y validado exclusivamente con datos de teledetección. En primer lugar, se utilizó el modelo ecohidrológico parsimonioso propuesto por Pasquato et al. (2015) en una parcela experimental situada en un bosque mediterráneo semiárido. Los resultados obtenidos en esta primera etapa de la tesis sugirieron que el modelo era capaz de reproducir adecuadamente la dinámica de la vegetación, así como las variaciones de humedad del suelo. En otras palabras, se pudo demostrar que un modelo parsimonioso con ecuaciones simples puede lograr buenos resultados en términos generales. Pero, como el modelo había sido aplicado a escala de parcela, todavía quedaba como tarea pendiente reproducir la variación espacial de la vegetación y del ciclo hidrológico. Para explorar la variación espacio-temporal de la vegetación y del ciclo del agua, se aplicó la versión distribuida del modelo ecohidrológico y parsimonioso utilizado previamente en una cuenca situada en Kenia. Con el fin de explorar la posible aplicabilidad de los datos de satélite, calibramos el modelo utilizando exclusivamente el NDVI proporcionada por la NASA. Se aplicó con éxito una metodología basada en el uso de la identificación de las funciones ortogonales empíricas (EOF por sus siglas en inglés). Esta última prueba proporcionó resultados prometedores: (1) los datos de satélite contienen una cantidad extraordinaria de información que puede ser usado para implementar modelos ecohidrológicos en regiones donde no se dispone de tal cantidad de información; (2) la metodología de calibración propuesta funciona satisfactoriamente y permite incorporar datos espacio-temporales en el proceso de parametrización del modelo, y (3) el modelo calibrado sólo con datos de satélite es capaz de reproducir tanto la dinámica espacio-temporal de la vegetación así como el caudal observado en el punto de desagüe de la cuenca. Es importante destacar las consecuencias positivas de este último resultado sobre todo en cuencas no aforadas, donde el uso de datos de satélite podría ser una alternativa para obtener una aproximación del recurso en el punto de desagüe. / [CA] L'aigua és la base de tota vida biològica a la Terra i un dels enllaços bàsics entre la biosfera i l'atmosfera. És igualment fonamental per als éssers humans i la naturalesa (Tolba, 1982). Tradicionalment, l'aigua i la vegetació s'han considerat com a sistemes diferents però és clarament necessari prendre un enfocament holístic que considere la qüestió del cicle de l'aigua d'una manera integrada, tenint en compte tant l'aigua blava com l'aigua verda (Birot et al., 2011). Al voltant d'aquesta idea va sorgir la nova disciplina anomenada Ecohidrología a principis del segle XX i des de llavors, no ha deixat de créixer tal com demostren el creixent augment de línies de recerca i publicacions científiques relacionades amb aquest nou camp. No obstant això, la majoria dels models hidrològics actuals inclou la vegetació com un paràmetre estàtic i no com una variable d'estat. Hi ha algunes excepcions que prenen explícitament la vegetació com a variable d'estat, però en aquests casos, la complexitat i el nombre de paràmetres a determinar dels models augmenten substancialment. En la pràctica, hem de fer front a la temible combinació de "escassetat de dades - alt nombre de paràmetres a determinar" amb molta freqüència. Per reduir aquest problema, es poden aplicar dues estratègies: (1) simplificar la complexitat conceptual dels models i així reduir el nombre de paràmetres a calibrar, i/o (2) augmentar la disponibilitat de dades mitjançant la incorporació de noves fonts d'informació. En aquesta tesi, hem explorat l'ús d'un model ecohidrològic distribuït i parsimoniòs (amb pocs paràmetres a determinar) que ha estat completament calibrat i validat exclusivament amb dades de teledetecció. En primer lloc, es va utilitzar el model ecohidrològic i parsimoniòs proposat per Pasquato et al. (2015) en una parcel·la experimental situada en un bosc mediterrani semi-àrid. Els resultats obtinguts en aquesta primera etapa de la tesi van suggerir que el model era capaç de reproduir adequadament la dinàmica de la vegetació, així com les variacions d'humitat del sòl. En altres paraules, es va poder demostrar que un model parsimoniòs amb equacions simples pot aconseguir bons resultats en termes generals. Però, com el model havia estat aplicat a escala de parcel·la, encara quedava com a tasca pendent reproduir la variació espacial de la vegetació i del cicle hidrològic. Per explorar la variació espai-temporal de la vegetació i del cicle de l'aigua, es va aplicar la versió distribuïda del model ecohidrològic i parsimoniòs utilitzat prèviament en una conca situada a Kenya. Al mateix temps, amb la finalitat d'explorar la possible aplicabilitat de les dades de satèl·lit, calibrem el model utilitzant exclusivament el NDVI proporcionat per la NASA. Es va aplicar amb èxit una metodologia basada en l'ús de la identificació de les funcions ortogonals empíriques (EOF per les seues sigles en anglès). Aquesta última prova va proporcionar resultats sorprenents i prometedors. De fet, els resultats obtinguts van demostrar que: (1) les dades de satèl·lit contenen una quantitat extraordinària d'informació que pot ser usada per implementar models ecohidrològics en regions on no es disposa de tal quantitat d'informació; (2) la metodologia de calibratge proposat funciona satisfactòriament i permet incorporar dades espai-temporals en el procés de parametrització del model, i (3) el model calibrat només amb dades de satèl·lit és capaç de reproduir tant la dinàmica espai-temporal de la vegetació així com el cabal observat en el punt de desguàs de la conca. És important destacar les conseqüències positives d'aquest últim resultat sobretot en conques no aforades, on l'ús de dades de satèl·lit podria ser una alternativa per obtenir una aproximació del recurs en el punt de desguàs. / Ruiz Pérez, G. (2016). On the use of satellite data to calibrate a parsimonious ecohydrological model in ungauged basins [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/72639

La prévision des crues est un exercice hydrologique complexe : les incertitudes y sont nombreuses, aussi bien dans le processus de modélisation hydrologique, dans la détermination de l'état initial du bassin versant avant le lancement de la prévision, que dans l'évolution des conditions météorologiques futures. Dans le cas des bassins versants non jaugés, où les observations de débits sont lacunaires voire absentes, ces incertitudes sont encore plus importantes, et le besoin de les réduire devient incontournable. Cette thèse s'intéresse à des méthodes simples et robustes qui peuvent apporter de l'information pertinente pour quantifier les incertitudes de prévision dans les bassins versants non jaugés. Le but est d'étudier la meilleure stratégie pour chercher l'information dans les bassins jaugés "donneurs", et pour la transférer vers le site non jaugé. Nous étudions les besoins pour mettre en place un modèle de simulation pluie-débit et pour effectuer une mise à jour du modèle de prévision en temps réel. Ces deux composantes de la prévision sont ainsi découplées dans notre approche. Cette thèse s'appuie sur une large base de données constituée d'environ 1000 bassins versants français, dont un jeu clé de 211 bassins versants qui permet la validation des approches développées. Elle s'appuie également sur une archive d'environ 4,5 années de prévisions d'ensemble de pluies, utilisées en forçage à la modélisation hydrologique journalière. La démarche adoptée consiste à intégrer les scenarios de transfert de l'information régionale disponible et les scenarios de la prévision météorologique d'ensemble dans un système de prévision orienté vers les bassins versants non jaugés. L'approche de prévision d'ensemble est ainsi généralisée à ce cas particulier de la prévision hydrologique. A travers plusieurs scénarios de débits futurs, nous cherchons à quantifier les incertitudes de prévisions dans les sites cibles non jaugés. Pour évaluer les différents scénarios des prévisions hydrologiques émis, un cadre de diagnostic d'évaluation des principales qualités d'un système de prévision d'ensemble, comprenant plusieurs critères numériques et graphiques, a été mis en place. Dans cette thèse, une attention particulière est prêtée aux attributs "fiabilité" et "précision" des prévisions. Nous proposons ainsi un nouveau critère graphique, nommé diagramme de précision d'ensemble. Ce critère permet notamment de mettre en valeur la qualité des prévisions qui ne sont pas forcément fiables, mais qui sont précises. Les résultats obtenus ont mis en évidence que la fiabilité des prévisions peut être améliorée sur un bassin versant non jaugé par l'utilisation de plusieurs jeux de paramètres issus des bassins versants voisins. Si la variabilité apportée par le voisinage géographique influe sur la dispersion des membres, et augmente ainsi la fiabilité des prévisions, la prise en compte des caractéristiques physiques, principalement de la surface des bassins versants, est apparue comme une alternative intéressante, influençant positivement aussi l'attribut précision des prévisions sur le site cible. De plus, il a été montré que la précision des prévisions d'ensemble sur le site non jaugé est améliorée par l'intermédiaire du transfert des bassins versants jaugés vers le site cible des corrections faites lors de la mise à jour sur les bassins voisins (mise à jour caractérisée ici par l'assimilation de la dernière observation de débit dans le modèle hydrologique, avant l'instant de prévision). Les différentes mesures de performance ont montré que la meilleure option pour améliorer la précision des prévisions serait de considérer les corrections effectuées sur le bassin le plus proche à chaque pas de temps de prévision. Le krigeage a également donné des résultats satisfaisants, marqués en plus par l'influence positive sur l'attribut fiabilité des prévisions. / Flood forecasting is a complex hydrological task: there are numerous uncertainties in the hydrological modelling process, in the determination of the initial catchment conditions before launching the forecast, and in the evolution of future weather conditions. In ungauged catchments, where streamflow observations are incomplete or absent, these uncertainties are even greater, and the need to reduce them becomes essential.This thesis focuses on simple and robust methods that can provide relevant information to quantify the uncertainty in ungauged catchments. The aim is to study the best strategy to search for information in gauged "donors" basins and to transfer it to the ungauged site. We investigate what information is needed to set up a rainfall-runoff model and to perform forecast updating in real time. These two components of a flood forecasting system are thus decoupled in our approach.This thesis is based on a large database of about 1000 French catchments, which includes a key set of 211 catchments that are used to validate the developed approaches. It also relies on an archive of about 4.5 years of ensemble forecasts of rainfall, which are used for hydrological modelling on a daily time step. The methodology adopted here integrates the scenarios of regional transfer of information and the scenarios of weather forecasting together in a forecasting system for ungauged basins. The approach of ensemble forecasting is thus generalised to this particular case of hydrological forecasting. Using several scenarios of future flows, we seek to quantify the predictive uncertainty in ungauged sites.To evaluate the flow forecast scenarios of the hydrological ensemble prediction system, a diagnostic framework with several numerical and graphical criteria is developed. Special attention is paid to the attributes of "reliability" and "accuracy" of the forecasts. We propose a new graphic criterion, named "diagram of ensemble accuracy". This criterion allows to highlight the quality of forecasts that are not necessarily reliable, but are accurate.The results show that forecast reliability in ungauged sites can be improved by using several sets of parameters from neighbour catchments. If on the one hand the variability brought by the information from the geographical proximity influences the spread of the ensemble forecasts, and thus improves forecast reliability, on the other hand taking into account the physical characteristics of the catchments, especially the surface, emerged as an interesting alternative, as it positively influences also the accuracy of the forecasts at the ungauged site.It is also shown that the accuracy of ensemble forecasts at ungauged sites can be improved with the transfer of updating information from gauged neighbour catchments (forecasting updating is here characterized by the assimilation of the last discharge observation in the hydrological model before the time of forecast). The updating information transferred to the ungauged site is the correction applied to the routing reservoir of the hydrological model. Different measures of forecast performance showed that the best option to improve forecast accuracy is to consider the corrections made at the closest gauged site. Kriging also gave satisfactory results, with additionally a positive impact also on the reliability of the ensemble flow forecasts.

For the process of social and economic development, hydropower energy has an important role such as being renewable, clean, and having less impact on the environment. In decision of the hydropower potential of a study area, the preliminary condition is the availability of the gages in the area. However, in Turkey, the gages in working order are limited and getting decreased in recent years. Therefore, the satellite based precipitation estimates has been gaining importance to predict runoff for ungauged basins. In this study, &Ccedil / oruh basin, which is located in the north-eastern part of Turkey, is selected to perform hydrologic modeling. The input precipitation data for the model are provided from the observations at meteorological stations and the Tropical Rainfall Measuring Mission (TRMM) satellite products (3B42 and 3B43). TRMM satellite is used to monitor and study the rainfall distribution. The precipitation radar on the TRMM is the first radar to make precipitation estimation from the space. Using both precipitation data, HEC-HMS, being well known hydrological model, is applied to the &Ccedil / oruh Basin for 2005 and 2003 water years. To distinguish the differences in the runoff simulations and water budget, comparisons are done with respect to flow monitoring stations. Statistical criteria show that model simulation results obtained from TRMM 3B42 products are promising in estimating the water potential in ungauged basins.

Demand for energy is constantly growing both in the world and in Turkey. Sustainable development being an important concept, development of small hydro power projects has been popular in recent years. Eastern Black Sea Basin in Turkey has a lot of small hydro power potential because of high amount of precipitation and existence of steep slopes. Since the amount of river runoff is the only parameter that is variable in order to determine the power potential, it is vital to estimate the project discharge in ungauged basins accurately that have hydro power potential. Projects discharges of hydro-power plants in ungauged basins have been calculated using conventional methods up to now. This study aims to introduce a statistical model in linear and multi-variate form using the topographical and morphological parameters derived from GIS and hydro-meteorological variables to estimate the specific flow duration curves of potential small hydro-power locations for the selected study areas in Eastern Black Sea Region namely Solakli and Karadere basins. As well as developing an annual regression model using the annual values of hydro-meteorological parameters / seasonal regression model (spring season) has also been developed by including the mean seasonal (spring) air temperature variable instead of snow covered area (SCA) in addition to basin parameters. By studying the spring model, effect of different variables from the annual model were tested and discussed with some recommendations for the future studies.

New and renewable energy resources are important in view of reduction of greenhouse gasses causing climate change and in eliminating of dependence on foreign sources in energy respects. Within this context, hydraulic energy is evaluated as one of the prior energy resources that should be utilized. Turkey has 26 basins and Eastern Black Sea Basin is one of the most feasible basins with a lot of small hydroelectric power plants. In the other hand, there is not enough number of discharge gauging stations in the basin. For that reason, up to now generally area ratio method has been used to estimate the project discharges of small hydroelectric power plants. Objective of this study is to estimate &ldquo / the project discharge&rdquo / which is corresponding to 5 flow percentiles (5%, 10%, 15%, 20%, 25%) depending on topographical, meteorological, hydrologic and soil-land cover parameters through developing a multilinear statistical model for Iyidere Basin as a part of Eastern Black Sea Basin. Perimeter of the basin, the ratio of the basin perimeter to the main stream length of the same basin, the drainage frequency, the mean slope of basin, v the mean annual precipitation and the curve number are the parameters that have been analysed for the multilinear statistical model. Principal Component Analysis, Multiple Regression Analysis and Stepwise Regression Analysis have been run for the data sets. For the computed discharges validation has been done. As a result of validation, it has been seen that the stepwise regression gives much closer discharge values to the observed values than the multiple regression results.

L'estimation d'indices d'étiage est d'une grande importance pour une meilleure connaissance de la ressource en eau disponible et en déduire des règles de gestion de cette ressource et des risques associés. Idéalement, ces indices sont calculés en sites jaugés à partir de longues chroniques de débits mesurés. En sites non jaugés, ces indices doivent être estimés. Cette thèse s'inscrit dans le projet d'amélioration de l'outil informatique " LoiEau " utilisé par les DREAL et les Agences de l'eau. Le principal objectif est de développer un modèle hydrologique simple et flexible pour estimer tous les indices d'étiage souhaités à des intervalles de temps journaliers ou supérieurs en sites non jaugés. La simplicité du modèle repose sur l'hypothèse forte que deux paramètres plus facilement régionalisables suffisent à estimer les indices avec une précision voulue. Cette précision dépend de la fonction objectif qu'il a fallu adapter aux faibles débits pour caler convenablement les paramètres du modèle sur des sites jaugés. Le modèle se veut flexible dans le sens où il est applicable à une large variété de bassins versants. Il a été comparé au modèle journalier à quatre paramètres GR4J et au modèle mensuel à deux paramètres LoiEau en s'intéressant aussi à la façon de régionaliser les paramètres. Des comparaisons ont également été réalisées avec une méthode d'estimation dite stochastique d'un indice d'étiage classique. Ces travaux montrent le potentiel du modèle à deux paramètres et l'avantage du pas de temps journalier, la variabilité temporelle des précipitations n'étant pas sans conséquence sur l'estimation de débits moyens en raison de la non-linéarité des modèles pluies-débits. / Estimating low-flow indices is of paramount importance to understand low flows and to manage water resources and risk assessments. These indices are derived from time-series of river discharges that are measured at gauged sites over long periods. At ungauged sites, the indices must be estimated. This research work is part of a project to improve “LoiEau”, a software package that is used by French regional environmental or water agencies. The main objective is to develop a simple and flexible rainfall-runoff model to simulate low-flow indices of ungauged sites at daily or longer time intervals. The model simplicity relies on the strong assumption that two free parameters are sufficient to provide accurate enough estimates of low-flow indices, yet making easier the regionalisation of models. The model accuracy depends on the objective function that is used to calibrate model parameters on gauged sites and had to be adapted to low-flow simulations. The model is flexible in the sense that it is designed to fit to a wide variety of catchments and hydro-meteorological behaviours. This model was compared with GR4J, a daily rainfall-runoff model which involves four parameters, and LoiEau, a monthly model. Comparisons were also carried out with a stochastic estimation method applied to a specific low-flow index. This research work shows the potential of the two-parameter model, but also the advantage of a daily time step to account for the temporal variability of precipitations, which is not without consequence on the assessment of average discharges due to the nonlinearity of rainfall-runoff models.

<p>Streamflow prediction in ungauged basins (PUB) is a process generating streamflow time series at ungauged reaches in a river network. PUB is essential for facilitating various engineering tasks such as managing stormwater, water resources, and water-related environmental impacts. Machine Learning (ML) has emerged as a powerful tool for PUB using its generalization process to capture the streamflow generation processes from hydrological datasets (observations). ML’s generalization process is impacted by two major components: data splitting process of observations and the architecture design. To unveil the potential limitations of ML’s generalization process, this dissertation explores its robustness and associated uncertainty. More precisely, this dissertation has three objectives: (1) analyzing the potential uncertainty caused by the data splitting process for ML modeling, (2) investigating the improvement of ML models’ performance by incorporating hydrological processes within their architectures, and (3) identifying the potential biases in ML’s generalization process regarding the trend and periodicity of streamflow simulations.</p><p>The first objective of this dissertation is to assess the sensitivity and uncertainty caused by the regular data splitting process for ML modeling. The regular data splitting process in ML was initially designed for homogeneous and stationary datasets, but it may not be suitable for hydrological datasets in the context of PUB studies. Hydrological datasets usually consist of data collected from diverse watersheds with distinct streamflow generation regimes influenced by varying meteorological forcing and watershed characteristics. To address the potential inconsistency in the data splitting process, multiple data splitting scenarios are generated using the Monte Carlo method. The scenario with random data splitting results accounts for frequent covariate shift and tends to add uncertainty and biases to ML’s generalization process. The findings in this objective suggest the importance of avoiding the covariate shift during the data splitting process when developing ML models for PUB to enhance the robustness and reliability of ML’s performance.</p><p>The second objective of this dissertation is to investigate the improvement of ML models’ performance brought by Physics-Guided Architecture (PGA), which incorporates ML with the rainfall abstraction process. PGA is a theory-guided machine learning framework integrating conceptual tutors (CTs) with ML models. In this study, CTs correspond to rainfall abstractions estimated by Green-Ampt (GA) and SCS-CN models. Integrating the GA model’s CTs, which involves information on dynamic soil properties, into PGA models leads to better performance than a regular ML model. On the contrary, PGA models integrating the SCS-CN model's CTs yield no significant improvement of ML model’s performance. The results of this objective demonstrate that the ML’s generalization process can be improved by incorporating CTs involving dynamic soil properties.</p><p>The third objective of this dissertation is to explore the limitations of ML’s generalization process in capturing trend and periodicity for streamflow simulations. Trend and periodicity are essential components of streamflow time series, representing the long-term correlations and periodic patterns, respectively. When the ML models generate streamflow simulations, they tend to have relatively strong long-term periodic components, such as yearly and multiyear periodic patterns. In addition, compared to the observed streamflow data, the ML models display relatively weak short-term periodic components, such as daily and weekly periodic patterns. As a result, the ML’s generalization process may struggle to capture the short-term periodic patterns in the streamflow simulations. The biases in ML’s generalization process emphasize the demands for external knowledge to improve the representation of the short-term periodic components in simulating streamflow.</p>

Avec le développement des mesures de pluie à hautes résolutions spatiales et temporelles , l’utilisation de modèles hydrométéorologiques distribués est désormais envisagée pour anticiper les phénomènes de crue soudaine sur les petits bassins versants non jaugés. Toutefois les approches développées jusqu’ici se sont généralement concentrées sur l’évaluation des phénomènes hydrologiques, laissant de côté la question de leurs impacts, qui dépendent fortement de la configuration du terrain et des enjeux qui y sont présents. Ce travail de thèse a permis de développer et tester une méthode d’évaluation directe de ces impacts à partir des sorties d’un modèle pluie-débit sur un territoire limité. La démarche mise en oeuvre repose sur un travail préalable d’analyse du territoire permettant d’une part d’évaluer les emprises submergées dans une large gamme de débits par une approche hydraulique simplifiée, puis de construire des relations univoques débit-enjeux pour chaque bief de cours d’eau. Ces relations permettent de produire des cartes d‘enjeux potentiellement touchés ,pouvant être actualisées régulièrement en cours d’événement. Deux études des cas ont permis de réaliser une première évaluation des performances de cette approche, du point de vue de la qualité d’estimation des emprises inondées, et du point de vue des impacts estimés à l’échelle d’un événement grâce à la comparaison avec des données d’assurance. Finalement, ces travaux ont permis de confirmer le potentiel de la méthode, dont les performances semblent en adéquation avec l’objectif visé : obtenir rapidement une première hiérarchisation des impacts occasionnés par les crues soudaines à l’échelle d’un grand territoire. / With the development of rainfall measurements at highspatial and temporal resolutions, the use of distributed hydrometeorological models is now considered to forecast flash floods on small and ungauged catchment areas. Current flashflood monitoring systems generally enable a real-time assessment of the potential flash-flood magnitudes. However they do not assess the potential impacts of flash-flood, which highly depends on the catchment areas configuration and on the importance of potentially affected assets. The purpose of this PhD research work was to develop and test a method which can be used to directly estimate the impacts of flash-floods, based on the outputs of a distributed rainfall-run off model. The approach is based on a prior analysis of the study area in order to assess the potential impact of different discharge levels on the flooded areas and to identify from geography database the associated buildings at risk. The aim is to build impact models on specific river reaches, using discharge versus impact graphs. The use of these impact models combined with a rainfall-run off model, has enabled us to compute maps of potential impacts, based on real time assessment of flood events updated every 15 minutes. This method was evaluated on two case studies looking at the accuracy and relevance of estimated impacts for each event – and comparing the outcomes to insurance losses data. This research work has helped to confirm the efficiency of this new combined method, which may become a useful tool to forecast large-scale effects of local impacts of flash-floods.

The conceptual rainfall-runoff model Hysim is used to estimate the flow in ungauged catchments in Scotland by Scottish Water. However, there are non-quantified uncertainties associated with the outcomes of the modelling strategy used. In order to identify and quantify these uncertainties it was necessary to use the framework of proxy-basin validation in order to evaluate the performance of different modelling strategies.   The proxy-basin validation test requires hydrologically analogous catchments for the evaluation of models, a Region Of Influence regionalisation method was used in order group selected catchments by Q95(%MF). Four groups of four catchments were established, which covered Q95(%MF) 5-7%, 7-9%, 9-11% and 11-13%.   The allocation of “donor catchment” and “target catchment” for each Q95(%MF) group was accomplished through discussion with Scottish Water with respect to existing Scottish Water modelled catchments. A single donor catchment and three target catchments were therefore indicated for each group.   Two modelling strategies were developed by the study; the first full transposition method used the entire optimised parameter-set from the donor catchment with the exception of the target catchment’s “catchment area” parameter. The second partial transposition method used the entire optimal parameter-set with the exception of the target catchment’s “interception storage”, “time to peak”, “rooting depth” and “catchment area” parameters.    It was found that the full transposition method had the least uncertainty associated its use for flow estimation when the parameter-set was derived from a donor catchment calibration that was excellent. Contrarily, it was found that the partial transposition model method had the least uncertainty associated with flow estimation for parameter-sets that were derived from a relatively poor donor catchment calibration.   Encouraged by this testing framework, this study has suggested the use of catalogue of donor parameter-sets that can be used to estimate flow for catchments that are hydrologically similar. This strategy of hydrological modelling has been recommended to improve existing Scottish Water Hysim methodology.

Atualmente, uma atividade que se tornou estratégica a nível nacional é o cultivo de espécies oleaginosas para o mercado alimentício e energético, em especial o plantio da palma de óleo (dendê) na região nordeste do estado do Pará, na Amazônia oriental. Esta cultura, assim como tem apresentado benefícios, como fixação do homem no campo, recuperação de áreas degradadas e redução da perda de solo, também tem apresentado riscos de ordem social e ambiental, como possíveis expropriações de terras e aumento do desmatamento e empobrecimento da diversidade ecológica. Mas se conhece pouco ainda dos impactos da expansão dessa cultura sobre o balanço hídrico e processos erosivos. Por isso, este trabalho estimou os impactos da expansão da cultura da palma de óleo na dinâmica de mudança de uso e cobertura da terra, bem como no escoamento superficial e na produção de sedimentos. Para isto, aplicou-se o modelo hidrossedimentológico Soil and Water Assessment Tool (SWAT) e o modelo de dinâmica espacial Conversion of Land Use and its Effects at Small region extent (CLUES) sobre as sub-bacias não monitoradas dos rios Bujaru (SBRB) e Mariquita (SBRM), com calibração do SWAT realizada a partir da técnica de regionalização de vazão por regressão não linear e medições em campo com o molinete hidrométrico. A princípio, as equações de regressão se apresentaram eficientes nas estimativas de dados de vazão para as sub-bacias, fundamentadas no bom resultado da calibração e validação sobre as estações reais. Nas áreas de palma de óleo, o modelo foi capaz de estimar com bom grau de eficiência a evapotranspiração nestas Unidades de Resposta Hidrológica da SBRB (1089,2 mm) e da SBRM (1093,1 mm), em relação a literatura e medidas em torre de monitoramento micrometeorológico. O modelo CLUE-S foi capaz de integralizar as variáveis explanatórias com as demandas agregadas e as características de elasticidade com o objetivo de gerar cenários futuros de uso e cobertura da terra, bem como modelar a palma de óleo nas sub-bacias, identificando as variáveis biofísicas como as principais forçantes de mudança de uso e cobertura da terra. As estimativas de escoamento superficial e produção de sedimentos apontaram para uma redução na SBRB e um aumento na SBRM entre os cenários de 2008, 2013 e o cenário projetado com o CLUE-S para 2023 em especial por razão da grande variação das áreas de vegetação secundária. As áreas de Palma de Óleo tiveram menor escoamento superficial e produção de sedimentos médio mensal do período mais chuvoso em ambas sub-bacias e em todos os cenários em relação as áreas de Agricultura Geral e Pastagem. Os resultados também mostraram a boa capacidade do uso integrado dos modelos SWAT e CLUE-S na geração de dados que contribuem para a análise do impacto ambiental da expansão da palma de óleo na região nordeste do estado do Pará, sendo também importante para o planejamento e gestão ambiental rural em bacias hidrográficas não monitoradas na Amazônia Oriental, pois demonstra a eficiência do método em proporcionar o aumento dos conhecimentos do comportamento hidrológico destas bacias em relação a dinâmica espacial de uso e cobertura do solo. / Nowadays, oilseed production for food and energy has become a strategic activity at national level in Brazil, particularly oil palm crops located in the Northeast of Pará State, Eastern Amazon. Oil palm crops have shown benefits such as keeping farmers on the land, recovering degraded areas and reducing soil loss. Conversely, it may also increase social and environmental risks linked to land tenure instability and land expropriation, deforestation and biodiversity losses. In such context, there is still a lack of knowledge concerning the impacts of such crop on the local water balance and erosion processes. Thus, this research estimated the impacts of increasing oil palm crops on land use and land cover change dynamics, as well as on runoff and soil erosion processes. To do so, it was applied the Soil and Water Assessment Tool (SWAT) and the spatial explicit framework Conversion of Land Use and its Effects at Small region extent (CLUE-S) over two ungauged sub-basins of Bujaru (SBRB) and Mariquita (SBRM) rivers. SWAT calibration was done by the regionalization streamflow method that adopts nonlinear regression and field measurements using a current meter. Initially, regression equations were effective in streamflow data estimation for the subbasins, this was based on effective calibration and validation results upon real stations. In the oil palm crop areas, the SWAT modeling was able to successfully estimate evapotranspiration on both hydrologic response unit of SBRB (1089,2 mm) and SBRM (1093,1mm) when compared to the literature and measures in micrometeorological monitoring tower. When applying CLUE-S model it was capable to integrate explanatory variables to scenario demands and elasticity parameters determining land use/cover change. Such integration allowed modelling oil palm spatial-temporal dynamics in current and future scenario demands within the two sub-basins SBRB and SBRM, as well as the identification of biophysical variables as the core drivers of land use/cover change. Runoff and sediment yield pointed out towards a decline in SBRB and an increase in SBRM in the current scenario between 2008 and 2013, as well as in the future scenario modelled using CLUE-S land use/cover change maps for 2023, particularly because of a large variation in the dynamics of secondary vegetation between the two sub-basins. Oil palm areas had smaller monthly average runoff and sediment yield in the rainiest period in both sub-basins and in current and future scenarios regarding agriculture and pasture areas. The results also show a suitable capability of integration between SWAT and CLUE-S models when generating data that contribute to the analysis of environmental impact of oil palm expansion in the Northeast of Para State. Such contribution is also relevant to the rural environmental planning and management in ungauged river-basins in Eastern Amazon, since the results found here demonstrate the efficiency of the method in providing an improved knowledge of the hydrological behavior of these basins concerning land use and land cover changes dynamics.