Les agrosystèmes sont soumis à de fortes hétérogénéités spatiales et temporelles, notamment du fait des pratiques agricoles. Les modèles de surface, qui servent à quantifier les échanges d'eau et d'énergie entre le sol, la végétation et la basse atmosphère, dits flux hydrométéorologiques, utilisent la plupart du temps des résolutions spatiales trop larges et une description des pratiques agricoles trop simple pour caractériser ces hétérogénéités, faute d'information spatialisée fiable et à fréquence temporelle suffisante pour paramétrer les simulations. Pourtant, réussir à simuler de manière plus réaliste les agrosystèmes à l'échelle du paysage, comme un bassin versant par exemple, est d'importance cruciale que ce soit pour gérer la répartition des ressources en eau ou évaluer les interactions entre pratiques agricoles et évolution climatique. La télédétection à haute résolution spatiale et temporelle, à l'image de la mission spatiale Sentinel-2 de l'ESA, permet de fournir des informations sur la surface terrestre à des résolutions inégalées (10 m, 5 jours) et sur l'ensemble du globe. Cette thèse visait donc à exploiter ce type de données dans un modèle de surface, le modèle SURFEX-ISBA développé par le CNRM, afin d'améliorer la représentation des pratiques agricoles et évaluer son impact sur les flux hydrométéorologiques à l'échelle du paysage. Le premier volet de la thèse avait pour objectif de représenter l'hétérogénéité spatio-temporelle des cultures, du fait des choix des dates de semis et de récolte ainsi que des rotations de culture, dans le modèle. Pour ce faire, j'ai exploité les produits issus des données du satellite optique Formosat-2 (8m, acquisitions programmées), sous la forme de cartes d'occupation des sols et de cartes multi-temporelles d'indice de surface foliaire (LAI) afin de simuler un agrosystème du Sud-Ouest de la France sur une zone de 576 km2. Afin de simuler de telles étendues en exploitant la haute résolution des produits satellite tout en limitant le temps de calcul, une approche de simulation par parcelle a été mise en place. / Agricultural practices generate strong spatial and temporal heterogeneities of the vegetation in agrosystems. Land Surface Models (LSMs), which simulate water and energy fluxes between soil, vegetation and atmosphere, use coarse spatial resolutions and very simplified agricultural practices representations. Therefore, they cannot characterize such heterogeneities. However, simulating agrosystems in a realistic way is of great interest to manage water resources at landscape scale, like a river basin, or study the interactions between climate evolution and agriculture. High resolution remote sensing, like the ESA's Sentinel-2 space mission, allows monitoring the Earth surface globally with unprecedented spatio-temporal resolution of 10 meters and 5 days. This Ph. D. thesis aimed to exploit such data in the SURFEX-ISBA LSM, developed by the CNRM, to represent agricultural practices in the hydrometeorological fluxes estimation at landscape scale. The first part of the thesis aimed at representing the spatial and temporal heterogeneities of the vegetation due to the choice of sewing and harvesting dates and crop rotations in the model. I used multi-temporal Leaf Area Index and annual land cover maps derived from the Formosat-2 remote sensing date (8m, tasking acquisitions). Simulations were performed on a 576 km2 agricultural plain in southwestern France. In order to keep the interest of high resolution while saving computation time, a plot scale simulation approach was used.

A characteristic of alpine drainage basins is the very sparse distribution of meteorological recording stations. This study models a contiguous distribution of microclimate and snowpack accumulation in the upper Oldman River basin. To accomplish this goal, gaps between weather recording stations are first filled using a modified MTCLIM climate simulation model in conjunction with the spatial analysis capabilities of the PAMAP geographic information system (GIS). The GIS provides terrain information such as elevation, slope, and aspect on a 100 metre grid as input into the microclimate simulator which, in turn, outputs daily meteorological conditions for a user-defined period of time. The estimation of snowpack accumaltion is achieved with another component of the model which makes use of the modelled microclimate to calculate daily accumulation and ablation on a grid point basis. Simulation results are returned to the GIS for display and spatial analysis. Discussion includes such thngs as the grouping of terrain variables and the derivation of an altitudinal precipitation profile, both of which are required for computational efficiency. While regression analysis indicates a very close relationship between observed and simulated temperature, precipitation is less successfully modelled at the daily time scale. Comparisons of simulated temperature with observed data resulted in an r2 + .94 and are therefore considered very reliable. Daily precipitation comparisons initially indicated a low correlation between observed and simulated data. However, when monthly totals are considered instead, r2 rises to 0.66. When snopack conditions are simulated for several snow pillows in the region, regression analysis with observed data producers r2 values as high as 0.896. / xi, 178 leaves : ill., maps ; 29 cm.

Les étiages extrêmes entraînent souvent des conséquences importantes sur de multiples secteurs socio-économiques. Les récentes études liées au changement climatique semblent indiquer que ces événements risquent de devenir plus sévères et plus fréquents au cours des prochaines décennies. Malheureusement, le peu de données hydrométéorologiques disponibles avant les années 1970 ne permet pas de remettre dans un contexte historique les derniers événements observés ni ceux projetés par les études d'impact. Ces travaux de thèse s'attachent ainsi à améliorer l'état des connaissances sur les étiages extrêmes historiques ayant touché le territoire français depuis la fin du XIXe siècle grâce à une reconstruction hydrométéorologique. Ils proposent aussi un nouveau cadre méthodologique pour l'étude de ces extrêmes.Des informations sur la situation synoptique atmosphérique depuis la fin du XIXe siècle ont récemment été mises à disposition de la communauté via des réanalyses globales étendues comme la Twentieth Century Reanalysis (20CR). Ces travaux introduisent la méthode SCOPE qui permet de reconstruire la météorologie locale sur l'ensemble de la France à partir de la réanalyse 20CR. Elle produit un ensemble de 25 chroniques météorologiques spatialement homogènes de précipitations, température et évapotranspiration sur la période 1871-2012 et sur une grille de 64 km² recouvrant la France. Ces séries constituent le jeu de données SCOPE Climate, qui est ensuite utilisé comme forçage d'un modèle hydrologique sur un large échantillon de plus de 600 bassins versants français faiblement anthropisés. Un ensemble de 25 reconstructions de débits journaliers, appelé SCOPE Hydro, est ainsi produit entre 1871 et 2012. SCOPE Climate et SCOPE Hydro ont tous deux montré de bonnes performances en comparaison de données indépendantes de leur construction.Des événements spatio-temporels d'étiage extrême sont finalement identifiés et caractérisés à partir des chroniques de débit de SCOPE Hydro. Ces événements sont tout d'abord définis localement comme une période où le débit se trouve en-dessous d'un seuil mixte, combinaison d'un seuil fixe et d'un seuil variable avec la saison. Un regroupement spatial des événements locaux au sein du même événement spatio-temporel d'étiage extrême est effectué à l'échelle de la France, suivant une méthode spécialement mise au point dans le cadre de ces travaux. A l'issue de ces étapes, un événement peut être étudié localement ou à l'échelle nationale, grâce à des caractéristiques d'étendue, de durée ou de sévérité. Cette méthodologie permet d'identifier des événements exceptionnels d'étiage extrême anciens et peu connus (1878, 1893, 1942-1949), ou relativement récents mais peu documentés (1972, 1978, 1985), en plus de ceux connus (1921, 1976, 1989-1990, 2011). L'étude de l'évolution de ces événements sur 140 ans montre une plus grande proportion du territoire français touchée par des étiages extrêmes après les années 1940. Une comparaison des événements reconstruits à des sources documentaires faisant mention des sécheresses passées montre une bonne cohérence entre reconstructions et réalité.Ce travail de thèse contribue aux questions scientifiques d'actualité, notamment dans le cadre de la variabilité climatique et du changement climatique d'origine anthropique et de leurs conséquences hydrologiques. Ces travaux proposent deux méthodes innovantes sur la descente d'échelle statistique multivariée et l'identification spatio-temporelle des événements d'étiage extrême. Ils ont par ailleurs produit deux jeux de données hydroclimatiques ensemblistes de référence utilisables pour tout type d'étude climatique et hydrologique requérant une profondeur historique importante. / Extreme low-flow events have significant consequences on numerous socio-economic fields. Hydroclimate projections for the 21st century suggest an increase in low-flow severity and frequency. Nevertheless, projected events as well as recent observed events can hardly be put into a sufficiently long historical perspective due to the lack of hydrometeorological data before the 1970s. This work proposes to improve the knowledge on past extreme low-flow events having affected France since the end of the 19th century thanks to a hydrometeorological reconstruction. It also provides a new methodological framework to study these extreme events.Information on the atmospheric synoptic situation since the end of the 19th century have recently been released to the scientific community through extended global reanalyses like the Twentieth Century Reanalysis (20CR). This work introduces the SCOPE method that provides local meteorological reconstructions on the entire France derived from the 20CR reanalysis. SCOPE produces a 25-member ensemble of spatially coherent meteorological series of daily precipitation, temperature and evapotranspiration over the 1871-2012 period and on a 64 km² grid covering France. This dataset, called SCOPE Climate, is then used as forcings to run continuous hydrological simulations over more than 600 near-natural French catchments leading to a 25-member ensemble of daily streamflow time series -- called SCOPE Hydro -- between 1871 and 2012. SCOPE Climate and SCOPE Hydro show a relatively high skill during validation experiments against independent data.Spatio-temporal extreme low-flow events are finally identified and characterised from SCOPE Hydro reconstructed series. The events are first locally identified based on deficit characteristics under a novel combination of a fixed threshold and a daily variable threshold. A spatial matching procedure at the scale of France is developed in order to spatially assemble local extreme events into the same spatio-temporal event. After these steps, an event can either be studied at the local or national scale through its spatial extent, duration or severity characteristics. This work identified past and little known exceptional extreme events (1878, 1893, 1942-1949) or recent but poorly documented events (1972, 1978, 1985) besides well-known events (1921, 1976, 1989-1990, 2011). The evolution of these events since 1871 shows that a greater proportion of the French territory is affected by extreme low-flow events since the 1940s. A good coherence is found between reconstructed events and documentary sources on historical droughts.This work contributes to timely scientific issues, especially within the context of climate change and its hydrological impacts. This work proposes two innovative methods on multivariate statistical downscaling and spatio-temporal identification of extreme low-flow events. It also produced two reference hydroclimatic datasets that may be used in any study requiring long hydrometeorological series.

Dans un contexte de variabilité climatique et de multiplication des usages de l'eau, la compréhension et la prévision de la variabilité des débits des cours d'eau est aujourd'hui un enjeu majeur pour améliorer la gestion des ressources en eau à l'échelle des bassins versants. En France, le bassin versant de la Durance (Alpes du Sud), lieu de multiples usages de l'eau (hydroélectricité, agriculture, alimentation en eau potable, loisirs), fait l'objet d'une attention particulière en ce qui concerne les impacts du changement climatique qui pourraient être importants du fait de sa situation géographique et de son régime partiellement nival, et remettre en question les équilibres en place permettant le partage de la ressource. Dans l'optique d'une meilleure anticipation de la variabilité hydrologique d'un bassin versant, bien connaître le passé est une étape fondamentale permettant à la fois une meilleure connaissance du fonctionnement hydrologique du bassin et une mise en perspective des projections hydro-climatiques futures. Cette thèse a pour objectif et résultat principal de faire progresser l'état des connaissances sur la variabilité hydrologique du bassin de la Durance à l'échelle du dernier siècle. Deux axes principaux ont été développés pour remplir cet objectif. Une première étape a été la recherche et la mise au jour d'un ensemble de longues séries hydrométriques concernant la Durance et ses affluents, permettant d'élever à 11 le nombre de séries centenaires de débits journaliers aujourd'hui disponibles sur ce bassin. Les nombreux documents accompagnant les données retrouvées nous ont par ailleurs permis de retracer, pour une partie de ces séries, l'évolution des méthodes utilisées pour les construire. Un processus de simulation des méthodes anciennes à partir de données horaires disponibles sur la période récente nous a permis de quantifier l'incertitude associée à ces méthodes et de mettre en évidence des biais importants causés par celles-ci sur certaines portions de séries. Une méthode de correction a été proposée et appliquée à plusieurs séries. Le deuxième axe de notre travail a consisté en la reconstitution de séries hydrologiques en différents points du bassin. À cette fin, nous avons présenté et appliqué une méthode originale appelée ANATEM de reconstitution de séries climatiques à l'échelle du bassin versant à partir de données climatiques de grande échelle (champs de pressions atmosphériques), combinées à des informations plus régionales (séries de précipitations ou de températures observées). Les séries climatiques ainsi reconstituées ont ensuite été utilisées en entrée d'un modèle hydrologique pour construire des séries de débits. Ce processus nous a permis d'obtenir une vingtaine de séries hydrologiques couvrant la période 1884-2010 sur le bassin de la Durance. La comparaison des reconstitutions hydro-climatiques avec les longues séries de débits observés aujourd'hui disponibles a permis de valider la chaîne de reconstitution sur une période de plus d'un siècle. Les séries observées et reconstituées illustrent finalement la variabilité hydrologique du bassin de la Durance qui se caractérise par une alternance de périodes sèches et humides à l'échelle de la décennie, ainsi que par une légère tendance à la baisse des débits. Ces longues séries de l'hydrologie du passé permettront de mettre en perspective les études prospectives sur les ressources en eau disponibles à l'échelle du siècle prochain. / Understanding and predicting hydrological variability is becoming a major issue to improve water resources management at the watershed scale, as climate variability and multipurpose use of water increase stress on the available resources. In France, the Durance watershed (South part of the Alps) is both characterised by numerous water-related activities such as hydropower, irrigation, water supply, tourism, and by a wide range of meteorological contexts ranging from mountainous to dry Mediterranean watersheds. As a consequence, this watershed appears as very sensitive to observed and projected climate variability, with an impact on water resources sufficient to question the current balance between users. In order to better forecast the Durance watershed hydrological variability, tracing back its past evolution is an essential step. Indeed, historical knowledge provides a better understanding of how the watershed works, and put into perspective hydro meteorological projections for the next century. The main goal of this Ph.D. thesis is then to improve our knowledge of the hydrological variability of the Durance watershed over the last century. Two main themes have been developed.The first step focused on historical research, bringing to light 11 centennial time-series of daily streamflow on the Durance watershed. Those data were quite well documented, allowing us to follow the evolution of the methods used to construct some of those time-series. Based on recent streamflow time-series, a simulation process allowed us to quantify the uncertainty associated to the methods used in the past, and to highlight the significant biases they carried on some periods of time. A correction process was then developed, leading to the partial revision of some of the time-series. A second step involved reconstructing hydrological time-series at different points of the watershed. An original method, called ANATEM, has been introduced and exhaustively applied to rebuild climatological time-series at the watershed scale. This method is based on the use of large scale climatological variables (atmospheric pressure fields) combined with regional scale observations (observed precipitation or air temperature). Those reconstructed climatological time-series were then prescribed in a rainfall-runoff model, allowing the computation of hydrological simulations on the 1884-2010 period. The comparison of the simulated data with our 11 centennial observed time series allowed us to validate our hydro-climatological reconstruction chain over more than a century. Finally, the observed and simulated time-series illustrate the climatological and hydrological variability of the Durance watershed. This variability is characterised by the succession of alternatively dry and humid periods lasting for ten to fifteen years, and by a slight trend to streamflow decrease. These long-term hydrological time-series will then put into perspective future investigations on water resources available over the next century.

Dans cette thèse, nous nous intéressons à la représentation et à la prise en compte des incertitudes dans les systèmes de prévision hydrologique probabilistes à moyen-terme. Ces incertitudes proviennent principalement de deux sources : (1) de l’imperfection des prévisions météorologiques (utilisées en intrant de ces systèmes) et (2) de l’imperfection de la représentation du processus hydrologique par le simulateur pluie-débit (SPQ) (au coeur de ces systèmes).La performance d’un système de prévision probabiliste s’évalue par la précision de ses prévisions conditionnellement à sa fiabilité. L’approche statistique que nous suivons procure une garantie de fiabilité à condition que les hypothèses qu’elle implique soient réalistes. Nous cherchons de plus à gagner en précision en incorporant des informations auxiliaires.Nous proposons, pour chacune des sources d’incertitudes, une méthode permettant cette incorporation : (1) un post-traitement des prévisions météorologiques s’appuyant sur la propriété statistique d’échangeabilité et permettant la prise en compte de plusieurs sources de prévisions, ensemblistes ou déterministes ; (2) un post-traitement hydrologique utilisant les variables d’état des SPQ par le biais d’un modèle Probit arbitrant entre deux régimes hydrologiques interprétables et permettant ainsi de représenter une incertitude à variance hétérogène.Ces deux méthodes montrent de bonnes capacités d’adaptation aux cas d’application variés fournis par EDF et Hydro-Québec, partenaires et financeurs du projet. Elles présentent de plus un gain en simplicité et en formalisme par rapport aux méthodes opérationnelles tout en montrant des performances similaires. / In this thesis, we are interested in representing and taking into account uncertainties in medium term probabilistic hydrological prediction systems.These uncertainties mainly come from two sources: (1) from the imperfection of meteorological forecasts (used as inputs to these systems) and (2) from the imperfection of the representation of the hydrological process by the rainfall-runoff simulator (RRS) (at the heart of these systems).The performance of a probabilistic forecasting system is assessed by the sharpness of its predictions conditional on its reliability. The statistical approach we follow provides a guarantee of reliability if the assumptions it implies are complied with. We are also seeking to incorporate auxilary information to get sharper.We propose, for each source of uncertainty, a method enabling this incorporation: (1) a meteorological post-processor based on the statistical property of exchangeability and enabling to take into account several (ensemble or determistic) forecasts; (2) a hydrological post-processor using the RRS state variables through a Probit model arbitrating between two interpretable hydrological regimes and thus representing an uncertainty with heterogeneous variance.These two methods demonstrate adaptability on the various application cases provided by EDF and Hydro-Québec, which are partners and funders of the project. Those methods are moreover simpler and more formal than the operational methods while demonstrating similar performances.

Depuis le développement des mesures satellites de nombreuses missions spatiales sont dédiées au suivi de l'atmosphère et de la surface terrestre. Ces travaux de thèse s'inscrivent dans le cadre de la mission Megha-Tropiques dédiée au cycle de l'eau et de l'énergie en zone tropicale. L'objectif est d'évaluer le potentiel des estimations de précipitation par satellite pour des applications hydrologiques en zone tropicale. Les Tropiques réunissent les plus grands fleuves du globe, mais ne bénéficient pas de réseaux d'observation in-situ denses et continus permettant une gestion intégrée efficace de la ressource et des systèmes d'alertes. Les estimations des précipitations issues des systèmes d'observation satellite offrent une alternative pour ces bassins peu ou pas instrumentés et souvent exposés aux extrêmes climatiques. C'est le cas du fleuve Niger, qui a subi une grande variabilité climatique depuis les années 1950, mais aussi d'importants changements environnementaux et hydrologiques. Depuis les années 2000, le Niger moyen connaît une recrudescence des inondations pendant la période de crue Rouge (engendrée par ses affluents sahéliens pendant la mousson). A Niamey, des niveaux record de hauteur d'eau et de période d'inondation ont été enregistrés en 2003, 2010, 2012 et 2013, engendrant de nombreuses pertes humaines et matérielles. Ces travaux analysent l'influence du forçage pluviométrique sur les inondations liées à la crue Rouge à Niamey. Une gamme de produits pluviométriques (in situ et satellite) et la modélisation hydrologique (ISBA-TRIP) sont combinés pour étudier : (i) l'apport des produits satellite pour diagnostiquer la crue Rouge récente, (ii) l'impact des caractéristiques des produits et de leurs incertitudes sur les simulations et enfin (iii) l'évaluation du rôle des précipitations, face aux changements de conditions de surface, dans l'évolution de la crue Rouge à Niamey depuis les années 1950. L'étude a mis en évidence l'impact des caractéristiques des estimations des précipitations (cumul, intensité et distribution spatio-temporelle) sur la modélisation hydrologique et le potentiel des produits satellites pour le suivi des inondations. Les caractéristiques des précipitations se propageant dans la modélisation, la détection des inondations est plus efficace avec une approche relative à chaque produit plutôt qu'avec un seuil absolu. Ainsi des produits présentant des biais peuvent être envisagés pour la simulation hydrologique et la détection des inondations. Le nouveau produit TAPEER de la mission MT présente un fort potentiel hydrologique, en 2012 et pour la zone d'étude. D'autre part, l'étude de la propagation de l'erreur associée à ces précipitations a mis en évidence, la nécessité de déterminer la structure du champ d'erreur pour l'utilisation d'une telle information en hydrologie. Enfin la modélisation a été utilisée comme levier pour décomposer les sensibilités de la crue Rouge aux variations des précipitations et des conditions de surface. Pour simuler les changements hydrologiques entre les périodes 1953-1982 et 1983-2012, les changements d'occupation du sol et d'aire de drainage doivent être pris en compte. Puis les variations des précipitations peuvent expliquer les changements majeurs décennaux et annuels entre les années 1983 et 2012. / Since the development of satellite based remote sensing in the 1970s, many missions have been dedicated to monitoring the terrestrial atmosphere and surfaces. Some of these satellites are dedicated to the Tropics with specific orbits. Megha-Tropiques (MT) is devoted to the water and energy cycle in the tropical atmosphere and provides an enhanced sampling for rainfall estimation in the tropical region. This PhD work was initiated within MT hydro-meteorological activities, with the objective of assessing the hydrological potential of satellite rainfall products in the Tropics. The world most important rivers lay in tropical areas where the in situ observation networks are deficient. Alternative information is therefore needed for water resource management and alert systems. The present work focuses on the Niger River a basin which has undergone drastic climatic variations leading to disasters such as droughts and floods. Since 1950, the Niger has been through 3 main climatic periods: a wet period (1950-1960), a long and intense drought period (1970-1980) and since 1990 a partial recovery of the rainfall. These climatic variations and the anthropic pressure, have modified the hydrological behaviour of the basin. Since 2000, the middle Niger River has been hit by an increase of floods hazards during the so-called Red flood period. In Niamey city, the highest river levels and the longest flooded period were recorded in 2003, 2010, 2012 and 2013, leading to heavy casualties and property damage. This study combines hydrological modelling and a variety of rainfall estimation products (satellite and in-situ) to meet several objectives: (i) the simulation of the Niamey Red flood and the detection of floods (during the recent period 2000-2013) (ii) the study of the propagation of satellite rainfall errors in hydrological modelling (iii) the evaluation of the role of rainfall variability, and surface conditions, in the changes of the Red flood in Niamey since the 50s. The global model ISBA-TRIP, is run with a resolution of 0.5° and 3h, and several rainfall products were used as forcing. Products derived from gauges (KRIG, CPC), pure satellite products (TAPEER, 3B42RT, CMORPH, PERSIANN) and mixed satellite products adjusted by rain gauges (3B42v7, RFE2, PERSIANN-CDR). This work confirms the hydrological potential of satellite rainfall products and proposes an original approach to overcome their biases. It highlights the need for documenting the errors associated with the rainfall products and the error structure. Finally, the hydrological modelling results since the 1950s have given a new understanding of the relative role of rainfall and surface conditions in the drastic increase of flood risk in Niamey.

There is ample growth in scientific evidence about climate change. Since, hydrometeorological processes are sensitive to climate variability and changes, ascertaining the linkages and feedbacks between the climate and the hydrometeorological processes becomes critical for environmental quality, economic development, social well-being etc. As the river basin integrates some of the important systems like ecological and socio-economic systems, the knowledge of plausible implications of climate change on hydrometeorology of a river basin will not only increase the awareness of how the hydrological systems may change over the coming century, but also prepare us for adapting to the impacts of climate changes on water resources for sustainable management and development. In general, quantitative climate impact studies are based on several meteorological variables and possible future climate scenarios. Among the meteorological variables, sic “cardinal” variables are identified as the most commonly used in impact studies (IPCC, 2001). These are maximum and minimum temperatures, precipitation, solar radiation, relative humidity and wind speed. The climate scenarios refer to plausible future climates, which have been constructed for explicit use for investigating the potential consequences of anthropogenic climate alterations, in addition to the natural climate variability. Among the climate scenarios adapted in impact assessments, General circulation model(GCM) projections based on marker scenarios given in Intergovernmental Panel on Climate Change’s (IPCC’s) Special Report on Emissions Scenarios(SRES) have become the standard scenarios. The GCMs are run at coarse resolutions and therefore the output climate variables for the various scenarios of these models cannot be used directly for impact assessment on a local(river basin)scale. Hence in the past, several methodologies such as downscaling and disaggregation have been developed to transfer information of atmospheric variables from the GCM scale to that of surface meteorological variables at local scale. The most commonly used downscaling approaches are based on transfer functions to represent the statistical relationships between the large scale atmospheric variables(predictors) and the local surface variables(predictands). Recently Support vector machine (SVM) is proposed, and is theoretically proved to have advantages over other techniques in use such as transfer functions. The SVM implements the structural risk minimization principle, which guarantees the global optimum solution. Further, for SVMs, the learning algorithm automatically decides the model architecture. These advantages make SVM a plausible choice for use in downscaling hydrometeorological variables. The literature review on use of transfer function for downscaling revealed that though a diverse range of transfer functions has been adopted for downscaling, only a few studies have evaluated the sensitivity of such downscaling models. Further, no studies have so far been carried out in India for downscaling hydrometeorological variables to a river basin scale, nor there was any prior work aimed at downscaling CGCM3 simulations to these variables at river basin scale for various IPCC SRES emission scenarios. The research presented in the thesis is motivated to assess the impact of climate change on streamflow at river basin scale for the various IPCC SRES scenarios (A1B, A2, B1 and COMMIT), by integrating implications of climate change on all the six cardinal variables. The catchment of Malaprabha river (upstream of Malaprabha reservoir) in India is chosen as the study area to demonstrate the effectiveness of the developed models, as it is considered to be a climatically sensitive region, because though the river originates in a region having high rainfall it feeds arid and semi-arid regions downstream. The data of the National Centers for Environmental Prediction (NCEP), the third generation Canadian Global Climate Model (CGCM3) of the Canadian Center for Climate Modeling and Analysis (CCCma), observed hydrometeorological variables, Digital Elevation model (DEM), land use/land cover map, and soil map prepared based on PAN and LISS III merged, satellite images are considered for use in the developed models. The thesis is broadly divided into four parts. The first part comprises of general introduction, data, techniques and tools used. The second part describes the process of assessment of the implications of climate change on monthly values of each of the six cardinal variables in the study region using SVM downscaling models and k-nearest neighbor (k-NN) disaggregation technique. Further, the sensitivity of the SVM downscaling models to the choice of predictors, predictand, calibration period, season and location is evaluated. The third part describes the impact assessment of climate change on streamflow in the study region using the SWAT hydrologic model, and SVM downscaling models. The fourth part presents summary of the work presented in the thesis, conclusions draws, and the scope for future research. The development of SVM downscaling model begins with the selection of probable predictors (large scale atmospheric variables). For this purpose, the cross-correlations are computed between the probable predictor variables in NCEP and GCM data sets, and the probable predictor variables in NCEP data set and the predictand. A pool of potential predictors is then stratified (which is optional and variable dependant) based on season and or location by specifying threshold values for the computed cross-correlations. The data on potential predictors are first standardized for a baseline period to reduce systemic bias (if any) in the mean and variance of predictors in GCM data, relative to those of the same in NCEP reanalysis data. The standardized NCEP predictor variables are then processed using principal component analysis (PCA) to extract principal components (PCs) which are orthogonal and which preserve more than 98% of the variance originally present in them. A feature vector is formed for each month using the PCs. The feature vector forms the input to the SVM model, and the contemporaneous value of predictand is its output. Finally, the downscaling model is calibrated to capture the relationship between NCEP data on potential predictors (i.e feature vectors) and the predictand. Grid search procedure is used to find the optimum range for each of the parameters. Subsequently, the optimum values of parameters are obtained from the selected ranges, using the stochastic search technique of genetic algorithm. The SVM model is subsequently validated, and then used to obtain projections of predictand for simulations of CGCM3. Results show that precipitation, maximum and minimum temperature, relative humidity and cloud cover are projected to increase in future for A1B, A2 and B1 scenarios, whereas no trend is discerned with theCOMMIT. The projected increase in predictands is high for A2 scenario and is least for B1 scenario. The wind speed is not projected to change in future for the study region for all the aforementioned scenarios. The solar radiation is projected to decrease in future for A1B, A2 and B1 scenarios, whereas no trend is discerned with the COMMIT. To assess the monthly streamflow responses to climate change, two methodologies are considered in this study namely (i) downscaling and disaggregating the meteorological variables for use as inputs in SWAT and (ii) directly downscaling streamflow using SVM. SWAT is a physically based, distributed, continuous time hydrological model that operates on a daily time scale. The hydrometeorologic variables obtained using SVM downscaling models are disaggregated to daily scale by using k-nearest neighbor method developed in this study. The other inputs to SWAT are DEM, land use/land cover map, soil map, which are considered to be the same for the present and future scenarios. The SWAT model has projected an increase in future streamflows for A1B, A2 andB1 scenarios, whereas no trend is discerned with the COMMIT. The monthly projections of streamflow at river basin scale are also obtained using two SVM based downscaling models. The first SVM model (called one-stage SVM model) considered feature vectors prepared based on monthly values of large scale atmospheric variables as inputs, whereas the second SVM model (called two-stage SVM model) considered feature vectors prepared from the monthly projections of cardinal variables as inputs. The trend in streamflows projected using two-stage SVM model is found to be similar to that projected by SWAT for each of the scenarios considered. The streamflow is not projected to change for any of the scenarios considered with the one-stage SVM downscaling model. The relative performance of the SWAT and the two SVM downscaling models in simulating observed streamflows is evaluated. In general, all the three models are able to simulate the streamflows well. Nevertheless, the performance of SWAT model is better. Further, among the two SVM models, the performance of one-stage streamflow downscaling model is marginally better than that of the two-stage streamflow downscaling model.

The thesis is concerned with development of efficient regional frequency analysis (RFA) approaches to estimate quantiles of hydrometeorological events. The estimates are necessary for various applications in water resources engineering. The classical approach to estimate quantiles involves fitting frequency distribution to at-site data. However, this approach cannot be used when data at target site are inadequate or unavailable to compute parameters of the frequency distribution. This impediment can be overcome through RFA, in which sites having similar attributes are identified to form a region, and information is pooled from all the sites in the region to estimate the quantiles at target site. The thesis proposes new approaches to RFA of precipitation, meteorological droughts and floods, and demonstrates their effectiveness. The approach proposed for RFA of precipitation overcomes shortcomings of conventional approaches with regard to delineation and validation of homogeneous precipitation regions, and estimation of precipitation quantiles in ungauged and data sparse areas. For the first time in literature, distinction is made between attributes/variables useful to form homogeneous rainfall regions and to validate the regions. Another important issue is that some of the attributes considered for regionalization vary dynamically with time. In conventional approaches, there is no provision to consider dynamic aspects of time varying attributes. This may lead to delineation of ineffective regions. To address this issue, a dynamic fuzzy clustering model (DFCM) is developed. The results obtained from application to Indian summer monsoon and annual rainfall indicated that RFA based on DFCM is more effective than that based on hard and fuzzy clustering models in arriving at rainfall quantile estimates. Errors in quantile estimates for the hard, fuzzy and dynamic fuzzy models based on the proposed approach are shown to be significantly less than those computed for Indian summer monsoon rainfall regions delineated in three previous studies. Overall, RFA based on DFCM and large scale atmospheric variables appeared promising. The performance of DFCM is followed by that of fuzzy and hard clustering models. Next, a new approach is proposed for RFA of meteorological droughts. It is suggested that homogeneous precipitation regions have to be delineated before proceeding to develop drought severity - areal extent - frequency (SAF) curves. Drought SAF curves are constructed at annual and summer monsoon time scales for each of the homogeneous rainfall regions that are newly delineated in India based on the proposed approach. They find use in assessing spatial characteristics and frequency of meteorological droughts. It overcomes shortcomings associated with classical approaches that construct SAF curves for political (e.g., state, country) and physiographic regions (e.g., river basin), based on spatial patterns of at-site values of drought indices in the study area, without testing homogeneity in rainfall. Advantage of the new approach can be noted especially in areas that have significant variations in temporal and spatial distribution of precipitation (possibly due to variations in topography, landscape and climate). The DFCM is extended to RFA of floods, and its effectiveness in prediction of flood quantiles is demonstrated by application to Godavari basin in India, considering precipitation as time varying attribute. Six new homogeneous regions are formed in Godavari basin and errors in quantile estimates based on those regions are shown to be significantly less than those computed based on sub-zones delineated in Godavari basin by Central Water Commission in a previous study.

The world’s freshwater resources are being placed under increasing pressure owing to growth in population, economic development, improved standards of living, agricultural intensification (linked mainly to irrigation), pollution and mismanagement of available freshwater resources. Already, in many parts of the Orange River Catchment, water availability has reached a critical stage. It has become increasingly evident that water related problems can no longer be resolved by water managers alone, owing to the problems becoming more interconnected with other development related issues, as well as with social, economic, environmental, legal and political factors. With the advent of climate change and the likelihood of increases in extreme events, water managers’ awareness of uncertainties and critical reflections on the adequacy of current management approaches is increasing. In order to manage water resources effectively a more holistic approach is required than has hitherto been the case, in which technological, social and economic development are linked with the protection of natural ecosystems and with dependable projections of future climatic conditions. To assess the climate risk connected with rural and urban water management, and to develop adaptive strategies that can respond to an increasingly variable climate that is projected into the future and help to reduce adverse impacts, it is necessary to make connections between climate related hazards, climate forecasts as well as climate change, and the planning, design, operation, maintenance, and rehabilitation of water related infrastructure. Therefore, adaptive water resources management (AWRM), which in essence is “learning by doing”, is believed to be a timely extension of the integrated water resources management (IWRM) approach as it acknowledges uncertainty and is flexible in that it allows for the adjustment of actions based on information learned about the system. Furthermore, it is suggested that climate risk management be imbedded within the AWRM framework. The objective of the research presented in this thesis is to develop techniques to integrate state-of-the-art climate projection scenarios – which forms part of the first step of the adaptive management cycle – downscaled to the regional/local scale, with hydro-climatic hazard determination – which forms part of the first step in the risk management process – in order to simulate projected impacts of climate change on hydro-climatic hazards in the Orange River Catchment (defined in this study as those areas of the catchment that exist within South Africa and Lesotho). The techniques developed and the results presented in this study can be used by decision-makers in the water sector in order to make informed proactive decisions as a response to projected future impacts of hydro-climatic hazards – all within a framework of AWRM. Steps towards fulfilling the above-mentioned objective begins by way of a comprehensive literature review; firstly of the study area, where it is identified that the Orange River Catchment is, in hydro-climatic terms, already a high risk environment; and secondly, of the relevant concepts involved which are, for this specific study, those pertaining to climate change, and the associated potential hydro-climatic impacts. These include risk management and its components, in order identify how hazard identification fits into the broader concept of risk management; and water resources management practices, in order to place the issues identified above within the context of AWRM. This study uses future projections of climate from five General Circulation Models, all using the SRES A2 emission scenario. By and large, however, where techniques developed in this study are demonstrated, this is done using the projections from the ECHAM5/MPI-OM GCM which, relative to the other four available GCMs, is considered to provide “middle of the road” projections of future climates over southern Africa. These climate projections are used in conjunction with the locally developed and widely verified ACRU hydrological model, as well as a newly developed hydro-climatic database at a finer spatial resolution than was available before, to make projections regarding the likelihood and severity of hydro-climatic hazards that may occur in the Orange River Catchment. The impacts of climate change on hydro-climatic hazards, viz. design rainfalls, design floods, droughts and sediment yields are investigated, with the results including a quantitative uncertainty analysis, by way of an index of concurrence from multiple GCM projections, for each of the respective analyses. A new methodology for the calculation of short duration (< 24 hour) design rainfalls from daily GCM rainfall projections is developed in this study. The methodology utilises an index storm approach and is based on L-moments, allowing for short duration design rainfalls to be estimated at any location in South Africa for which daily GCM rainfall projections exist. The results from the five GCMs used in this study indicate the following possible impacts of climate change on hydro-climatic hazards in the Orange River Catchment: · Design rainfalls of both short and long duration are, by and large, projected to increase by the intermediate future period represented by 2046 - 2065, and even more so by the more distant future period 2081 - 2100. · Design floods are, by and large, projected to increase into the intermediate future, and even more into the more distant future; with these increases being larger than those projected for design rainfalls. · Both meteorological and hydrological droughts are projected to decrease, both in terms of magnitude and frequency, by the period 2046 - 2065, with further decreases projected for the period 2081 - 2100. Where increases in meteorological and hydrological droughts are projected to occur, these are most likely to be in the western, drier regions of the catchment. · Annual sediment yields, as well as their year-to-year variability, are projected to increase by the period 2046 - 2065, and even more so by the period 2081 - 2100. These increases are most likely to occur in the higher rainfall, and especially in the steeper, regions in the east of the catchment. Additionally, with respect to the above-mentioned hydro-climatic hazards, it was found that: · The statistic chosen to describe inter-annual variability of hydro-climatic variables may create different perceptions of the projected future hydroclimatic environment and, hence, whether or not the water manager would decide whether adaptive action is necessary to manage future variability. · There is greater uncertainty amongst the GCMs used in this study when estimating design events (rainfall and streamflow) for shorter durations and longer return periods, indicating that GCMs may still be failing to simulate individual extreme events. · The spatial distribution of projected changes in meteorological and hydrological droughts are different, owing to the complexities introduced by the hydrological system · Many areas may be exposed to increases in hydrological hazards (i.e. hydrological drought, floods and/or sediment yields) because, where one extreme is projected to decrease, one of the others is often projected to increase. The thesis is concluded with recommendations for future research in the climate change and hydrological fields, based on the experiences gained in undertaking this study. / Thesis (Ph.D.)-University of KwaZulu-Natal, Pietermaritzburg, 2012.

MESCH / Department of Hydrology and Water Resources