Maintaining the ecological health of streams is vital for sustainable water resources management. Streamflow is a primary factor influencing the structure and function of ecological communities. A quantitative understanding of how stream biota respond to variation in streamflow is required for stream bioassessment. This dissertation focuses on quantifying relationships between streamflow regime and stream macroinvertebrate richness and composition. The contribution comprises statistical models that predict stream macroinvertebrate class from streamflow regime and predict streamflow regime from watershed attributes, and a tool that helps derive watershed attribute variables used in these models. The dissertation is a collection of three papers. In the first paper 12 variables were used to represent streamflow regime at 543 sites in the western US. Principal component analysis (PCA) and K-means clustering were used to obtain statistically independent factors and streamflow regime classes. We examined the relationship between these characterizations of streamflow and macroinvertebrate richness and composition at 63 of the 543 sites where there was also biological data. This analysis identified specific aspects of the streamflow regime that were useful in predicting macroinvertebrate richness and composition and that have potential application in classification-based bioassessment and management. A regional-scale study such as this requires tools for efficiently delineating watersheds and deriving their attributes. Paper two presents a multiple watershed delineation tool that addresses issues such as a) incorrectly positioned outlets and b) large Digital Elevation Models. This tool has capabilities to delineate stream networks with the threshold that determines drainage density being objectively determined so that the resulting networks adhere to geomorphological stream network laws. It also derives a suite of geomorphological watershed attributes that were used in prediction models in paper three. In paper three, we developed statistical models to predict streamflow regime class from watershed attributes. Four popular statistical methods were used and the uncertainty associated with class predictions for each method was quantified. Paper three also identified the watershed attributes that were most important for discriminating streamflow regime classes.

High rates of soil loss and high sediment loads in rivers necessitate efficient monitoring and quantification methodologies so that effective land management strategies can be designed. Constructing a sediment budget is a useful approach to address these issues. Quantifying a sediment budget using classical field-based techniques, however, is labour intensive and expensive for poorly gauged, large drainage basins. The availability of global environmental datasets in combination with GIS techniques provides an opportunity for studying large basins. Following this approach, a framework is presented for constructing sediment budgets for large, data-sparse drainage basins, which is applied to the mountainous upper Indus River basin in northern Pakistan. The methodological framework consists of five steps: (1) analyzing hydro-climatological data for dividing the drainage basin into characteristic regions, and calculating sediment yields; (2) investigation of major controls on sediment yields; (3) identification and mapping of sediment source areas by spatially distributed modelling of erosional processes; (4) spatially distributed modelling of sediment yields; and (5) carrying out the sediment budget balance calculation at the basin outlet. Further analysis carried out on the Indus data has enabled a better understanding of sediment dynamics in the basin.<p> Analysis of the available hydro-climatological data indicates that the basin can be subdivided into three characteristic regions based on whether runoff production and subsequent sediment generation is controlled by temperature (Region 1, upper, glacierized sub-basins), precipitation caused by the monsoon and western disturbances (Region 3, lower sub-basins), or a combination of the two (Region 2, middle reach sub-basins). It is also demonstrated that contrary to the conventional model, the specific sediment yield increases markedly with drainage area along the Indus River. An investigation of major controls on specific sediment yield in the basin indicates that percent snow/ice cover is a major land cover control for specific sediment yield. Spatially distributed erosion modelling predictions indicate that 87% of the annual gross erosion takes place in the three summer months with greatest erosion potential concentrated in sub-basins with high relief and a substantial proportion of glacierized area. Lower erosion rates can be explained by the arid climate and low relief on the Tibetan Plateau, and by the dense vegetation and lower relief in the lower monsoon sub-region. The model predicts an average annual erosion rate of 3.2 mm/a or 868 Mt/a. Spatially distributed sediment yield predictions made with coupled models of erosion and sediment delivery indicate that the Indus sub-basins generally show an increase of sediment delivery ratio with basin area. The predicted annual basin sediment yield is 244 Mt/a and the overall sediment delivery ratio in the basin is calculated as 0.28. The long-term mean annual sediment budget, based on mass balance, is characterized by a gross erosion of 762.9, 96.7 and 8.4 Mt, and a gross storage of 551.4, 66.1, and 6.5 Mt in the upper, middle, and lower regions of the basin, respectively. The sediment budget indicates that the major sources of eroded sediment are located in the Karakoram, in particular in the Hunza basin. Substantial sediment storage occurs on the relatively flat Tibetan Plateau and the Indus River valley reach between Partab Bridge and Shatial. The presented framework for sediment budget construction requires relatively few data, mostly derived from global datasets. It therefore can be utilized for other ungauged or poorly gauged drainage basins of the world.

High rates of soil loss and high sediment loads in rivers necessitate efficient monitoring and quantification methodologies so that effective land management strategies can be designed. Constructing a sediment budget is a useful approach to address these issues. Quantifying a sediment budget using classical field-based techniques, however, is labour intensive and expensive for poorly gauged, large drainage basins. The availability of global environmental datasets in combination with GIS techniques provides an opportunity for studying large basins. Following this approach, a framework is presented for constructing sediment budgets for large, data-sparse drainage basins, which is applied to the mountainous upper Indus River basin in northern Pakistan. The methodological framework consists of five steps: (1) analyzing hydro-climatological data for dividing the drainage basin into characteristic regions, and calculating sediment yields; (2) investigation of major controls on sediment yields; (3) identification and mapping of sediment source areas by spatially distributed modelling of erosional processes; (4) spatially distributed modelling of sediment yields; and (5) carrying out the sediment budget balance calculation at the basin outlet. Further analysis carried out on the Indus data has enabled a better understanding of sediment dynamics in the basin.<p> Analysis of the available hydro-climatological data indicates that the basin can be subdivided into three characteristic regions based on whether runoff production and subsequent sediment generation is controlled by temperature (Region 1, upper, glacierized sub-basins), precipitation caused by the monsoon and western disturbances (Region 3, lower sub-basins), or a combination of the two (Region 2, middle reach sub-basins). It is also demonstrated that contrary to the conventional model, the specific sediment yield increases markedly with drainage area along the Indus River. An investigation of major controls on specific sediment yield in the basin indicates that percent snow/ice cover is a major land cover control for specific sediment yield. Spatially distributed erosion modelling predictions indicate that 87% of the annual gross erosion takes place in the three summer months with greatest erosion potential concentrated in sub-basins with high relief and a substantial proportion of glacierized area. Lower erosion rates can be explained by the arid climate and low relief on the Tibetan Plateau, and by the dense vegetation and lower relief in the lower monsoon sub-region. The model predicts an average annual erosion rate of 3.2 mm/a or 868 Mt/a. Spatially distributed sediment yield predictions made with coupled models of erosion and sediment delivery indicate that the Indus sub-basins generally show an increase of sediment delivery ratio with basin area. The predicted annual basin sediment yield is 244 Mt/a and the overall sediment delivery ratio in the basin is calculated as 0.28. The long-term mean annual sediment budget, based on mass balance, is characterized by a gross erosion of 762.9, 96.7 and 8.4 Mt, and a gross storage of 551.4, 66.1, and 6.5 Mt in the upper, middle, and lower regions of the basin, respectively. The sediment budget indicates that the major sources of eroded sediment are located in the Karakoram, in particular in the Hunza basin. Substantial sediment storage occurs on the relatively flat Tibetan Plateau and the Indus River valley reach between Partab Bridge and Shatial. The presented framework for sediment budget construction requires relatively few data, mostly derived from global datasets. It therefore can be utilized for other ungauged or poorly gauged drainage basins of the world.

A flow duration curve (FDC) defines the relation between the flow amount of any time (daily, yearly, or another time) and its frequency. Moreover, FDCs are used in many water resources projects. However, the ungauged basins or limited amount of gauging in a basin is a common problem. Therefore, regional FDCs are needed to be developed in ungauged basins. Oltu basin has been chosen as the study area, which is located in the north-eastern part of Turkey in &Ccedil / oruh Basin. Two parametric approaches and a statistical approach have been applied to develop regional flow duration curves (FDCs) in Oltu Basin. Parametric approaches cover two different models, namely Model Kocatepe, which is a five parameter model depending on the regression analysis between discharge having certain probability of occurrences and geomorphologic and climatic factors / Model Quimpo, which is a two parameter model proposed by Quimpo. Lognormal distribution has been used in the statistical approach. Several performance indices have been evaluated to decide on if the model dependable or not. As a result of these analysis, it is concluded that, Model Quimpo gives good results in small basins, whereas, Model Kocatepe is effective in large areas. Statistical approach is not an appropriate method to use while regionalizing FDCs in Oltu basin.The analysis performed for short-term duration has revealed that 5-years record lengths of discharges are enough to develop a dependable FDC compared to regional FDC. The validation results and the performance indices are presented with the analysis results.

Στον ελληνικό χώρο, για έναν μεγάλο αριθμό λεκανών δεν υπάρχουν διαθέσιμες μετρήσεις απορροής, γεγονός που καθιστά δύσκολη την υδρολογική τους προσομοίωση για τον προσδιορισμό των συνιστωσών του υδρολογικού ισοζυγίου. Μια τέτοια περίπτωση αποτελούν οι λεκάνες απορροής των ποταμών του Πείρου και του Παραπείρου. Στους ποταμούς αυτούς κατασκευάζεται ένα σύστημα έργων που αποτελείται από ένα φράγμα εκτροπής στη θέση Βαλμαδούρα στον ποταμό Πείρο, έναν ταμιευτήρα στη θέση Αστέρι στον Παραπείρο και έναν συνδετικό αγωγό. Στόχος της κατασκευής των έργων αυτών αποτελεί η κάλυψη των υδρευτικών αναγκών της πόλης των Πατρών, της Βιομηχανικής Περιοχής και των κοινοτήτων της Β.Δ. Αχαϊας. Στη μελέτη των έργων αυτών (Υ.ΠΕ.ΧΩ.Δ.Ε., 1997) ο υπολογισμός του υδρολογικού ισοζυγίου στις λεκάνες που αντιστοιχούν στις θέσεις των έργων βασίστηκε σε απλοποιητικές παραδοχές. Συγκεκριμένα θεωρήθηκε ότι το νερό αποθηκεύεται μόνο στην εδαφική ζώνη και δεν γίνεται διάκριση ανάμεσα σε επιφανειακή και υπόγεια απορροή. Στην παρούσα εργασία για τον ακριβή και αναλυτικό προσδιορισμό των συνιστωσών του υδρολογικού ισοζυγίου χρησιμοποιήθηκε το υδρολογικό μοντέλο βροχής–απορροής ENNS (Nachtnebel, 1993), η χρήση του οποίου απαιτεί την εύρεση τιμών των παραμέτρων που υπεισέρχονται στις εξισώσεις των διαδικασιών που περιγράφουν το μοντέλο. Λόγω ελλείψεως μετρήσεων απορροής για τις λεκάνες του Πείρου και του Παραπείρου, για την εύρεση των τιμών των παραμέτρων του μοντέλου εφαρμόστηκε μια μεθοδολογία διερεύνησης της υδρολογικής ομοιότητας των δύο λεκανών με άλλες λεκάνες, για τις οποίες υπάρχουν διαθέσιμες μετρήσεις απορροής. Πιο συγκεκριμένα χρησιμοποιήθηκαν οι λεκάνες απορροής του Γλαύκου στον Νομό Αχαϊας, του Αχελώου που αντιστοιχεί στην θέση του φράγματος της Μεσοχώρας, του ποταμού Ερύμανθου, παραποτάμου του Αλφειού και του ποταμού Myakka στην Florida των Η.Π.Α.. Η διαδικασία που ακολουθήθηκε είναι η εξής: Αρχικά πραγματοποιείται η ρύθμιση του μοντέλου για τις τέσσερις λεκάνες για τις οποίες υπάρχουν μετρήσεις απορροής με στόχο (α) να βρεθούν οι τιμές των παραμέτρων που επηρεάζουν το αποτέλεσμα της προσομοίωσης και (β) να διαπιστωθεί το εύρος διακύμανσης των τιμών αυτών των παραμέτρων. Λόγω του σημαντικού εύρους διακύμανσης αυτών των παραμέτρων, πραγματοποιείται η κατάταξη των λεκανών του Πείρου, του Παραπείρου και των τεσσάρων λεκανών, για τις οποίες υπάρχουν διαθέσιμες μετρήσεις απορροής, σύμφωνα με κάποιους φυσιογραφικούς και κλιματικούς παράγοντες που επιλέχθησαν, ως προς την αναμενόμενη επιφανειακή απορροή. Στη συνέχεια, για τις λεκάνες για τις οποίες υπάρχουν διαθέσιμες μετρήσεις απορροής γίνεται έλεγχος εάν οι τιμές των παραμέτρων του μοντέλου είναι συμβατές με την κατάταξη των λεκανών σύμφωνα με τους φυσιογραφικούς και κλιματικούς παράγοντες. Επειδή προκύπτει ότι αυτό δεν συμβαίνει, πραγματοποιείται επαναρρύθμιση του μοντέλου αναζητώντας συνδυασμούς παραμέτρων, οι οποίοι να επαληθεύουν την κατάταξη των λεκανών. Τέλος με βάση τα αποτελέσματα της διερεύνησης της υδρολογικής ομοιότητας των λεκανών επιλέγονται οι τιμές των παραμέτρων του μοντέλου για την υδρολογική προσομοίωση των λεκανών του Πείρου και του Παραπείρου. Τα αποτελέσματα που προέκυψαν από την υδρολογική προσομοίωση του Πείρου και του Παραπείρου με την χρήση του μοντέλου συγκρίθηκαν με τα αποτελέσματα της μελέτης του Υ.ΠΕ.ΧΩ.Δ.Ε. (Μάρτιος 1997). Τα συμπεράσματα που προκύπτουν από την παρούσα εργασία είναι τα εξής: (α) Από την διερεύνηση της υδρολογικής ομοιότητας των λεκανών του Πείρου και του Παραπείρου με τις λεκάνες του Γλαύκου, του Ερύμανθου, της Μεσοχώρας και του ποταμού Myakka, Η.Π.Α. διαπιστώθηκε ότι η επαναρρύθμιση του μοντέλου που πραγματοποιήθηκε για τις τέσσερις λεκάνες, για τις οποίες υπαρχουν μετρήσεις απορροής, οδήγησε σε βελτίωση των αποτελεσμάτων της ρύθμισης (μείωση των αποκλίσεων της μετρημένης και προσομοιωμένης απορροής). Επίσης, επειδή ο λόγος της επιφανειακής απορροής προς την βροχή για τις τέσσερις λεκάνες επαληθεύει την κατάταξη σύμφωνα με τους φυσιογραφικούς και κλιματικούς παράγοντες, προκύπτει ότι η διαδικασία που ακολουθήθηκε αποτελεί μια δυνατότητα αντιστοίχισης των παραμέτρων του μοντέλου σε κριτήρια που προκύπτουν από αυτούς τους παράγοντες. (β) Από την σύγκριση των αποτελεσμάτων της μελέτης του Υ.ΠΕ.ΧΩ.Δ.Ε. με τα αποτελέσματα του μοντέλου αποδεικνύεται ότι το μοντέλο δίνει πιο ρεαλιστικές τιμές για τις συνιστώσες του υδρολογικού ισοζυγίου των λεκανών του Πείρου και του Παραπείρου από την μελέτη. Οι λόγοι είναι ότι (α) σύμφωνα με την κατάταξη των λεκανών βάσει των φυσιογραφικών και κλιματικών παραγόντων οι λεκάνες του Πείρου και του Παραπείρου παρουσιάζουν υδρολογική ομοιότητα με την λεκάνη του Γλαύκου, για την οποία το μοντέλο δίνει μια ρεαλιστική προσέγγιση της συνολικής απορροής (άμεσης και βασικής) και (β) η βασική απορροή που υπολογίστηκε με το μοντέλο παρουσιάζει πολύ μικρότερη μεταβλητότητα στον χρόνο έναντι της άμεσης απορροής λόγω της αποθηκευτικής ικανότητας του εδάφους. (γ) Από τα αποτελέσματα της υδρολογικής προσομοίωσης των λεκανών του Πείρου και του Παραπείρου με την χρήση του μοντέλου προκύπτει ότι τα ποσοστά της άμεσης απορροής προς την βροχόπτωση των λεκανών του Πείρου και του Παραπείρου σύμφωνα με το μοντέλο δεν ήταν τα αναμενόμενα, παρόλο που τα αποτελέσματα του μοντέλου κρίθηκαν ικανοποιητικά. Οι πιθανοί λόγοι είναι κάποια μειονεκτήματα που παρουσιάζει η μεθοδολογία διερεύνησης της υδρολογικής ομοιότητας των λεκανών και τα οποία κρίνεται ότι εάν ληφθούν υπόψη μπορούν να οδηγήσουν σε βελτίωση της μεθόδου. / In Greece, there is a significant number of ungauged basins, for which the hydrologic simulation is not possible and as a result calculating the components of the water balance is infeasible. The basins of the rivers Peiros and Parapeiros were selected as a case-study. These rivers are involved in a current project; the construction of a diversion dam in the position Valmadoura of Peiros River, a reservoir in the position Asteri of Parapeiros River and a conjection pipe. The aim of this project is to support the water supply of the city of Patras, the Industrial Region and the communities of the North-west Achaias. The study of this project conducted by the Hellenic Ministry of Environment, Physical Planning and Public Works in 1997 deals with the calculation of the water balance in these basins, based on simplified principles, assuming that the water is stored in the soil layer and there is no separation of the total runoff in surface flow and base flow. In the present study, the use of the rainfall-runoff model ENNS (Nachtnebel, 1993) allowed the exact and analytical quantification of the water balance. The use of this model requires the estimation of a number of parameters involved in the governing equations of the model. The lack of runoff measurements in Peiros and Parapeiros basins preclude the determination of the required parameters. A study of the hydrologic similarity of these two basins with a number of gauged basins gave solution to the problem. The chosen basins are: (a) Glaukos River basin in the prefecture of Achaias, (b) Mesoxora basin, a sub-basin of Acheloos River, (c) Erymanthos basin, a sub-basin of Alfeios River and (d) Myakka River basin, Florida of U.S.A. The procedure adopted is divided in the following steps: Firstly, the model ENNS was established for the basins mentioned previously in order to determine (a) the value of the parameters affecting the simulation result and (b) the range of their values. The significant variation led to the classification of the six basins, according to physiographic and climatic characteristics. The criterion used in the classification is the expected surface runoff to the amount of rainfall. Secondly, a compatibility check between the values of the parameters of the model and the basins classification was performed. The unsatisfying result compelled the reestablishement of the model for the four gauged basins with an aim to obtain the appropriate parameters in accordance with the basins classification. Finally, the values of the ENNS parameters, regarding Peiros and Parapeiros basins, were derived from the study of the hydrologic similarity of the basins. The results of the hydrologic simulation of Peiros and Parapeiros were compared to the results of the study conducted by the Hellenic Ministry of Environment, Physical Planning and Public Works in 1997. The conclusions of the present study are: (a)The setup of the ENNS model taking into account the hydrologic similarity study of the six basins improved the results of the model (decrease of the deviations between the measured and simulated discharge values). Moreover the ratio of the surface runoff to the rainfall for the four gauged basins verifies the basins classification. Hence, a satisfying degree of correlation between the model parameters and the physiographic and climatic basin characteristics, is obtained. (b)Comparing the obtained results to those of the Hellenic Ministry of Environment, Physical Planning and Public Works study in 1997, the present model gives a more realistic view of the water balance. The reasons are: (a) according to the basins classification, Peiros and Parapeiros basins are similar to Glaukos basin, for which the model gives a realistic approach of the total runoff (surface and base flow) and (b) the base flow calculated by the model shows a smaller variability, contrary to the surface flow due to the storage capacity of the soil. (c)The results of the hydrologic simulation of Peiros and Parapeiros basins showed that the percentage of the surface runoff to the amount of rainfall for the two basins is not compatible to the classification (smaller than 36.5% of Glaukos basin). However, the results of the model are satisfactory (accordance of the annual simulated values of actual evapotranspiration to the values calculated using the Turc method). The possible reasons are some disadvantages in the basins hydrologic similarity study that must be taken into consideration in order to improve the method.

The study was aimed at developing a simple methodology for flow prediction in ungauged basins using existing data resources. For this purpose, the streamflow measurements across the Kentucky River Basin located in Kentucky, USA were obtained from United States Geological Survey (USGS) archive. The flow transferring characteristics of the subbasins of the Kentucky River Basin were obtained by combining downstream and upstream stream gauges. The flow transferring function thus derived were related to watershed, channel and flow characteristics of the subbasins by multiple regression analysis. The gauge pairs were divided into two classes of subbasins representing Upper and Lower Kentucky, which were characterized mainly by the geology of the watersheds. The regression models corresponding to the two groups of subbasins were applied to example gauge pairs to evaluate the efficiency of the proposed model to predict streamflow in downstream channel. The estimated hydrographs agreed with the observed hydrographs with the performance efficiency of greater than 90%. The proposed method was tested for its applicability in first-order streams in the Goose Creek, a tributary to the Kentucky River. The overland flow component for the first-order streams was determined using TOPMODEL with topography, soil and climatic factors as inputs. The overland flow was routed to the Goose Creek outlet using the transfer function obtained from measured flow records. The simulated hydrographs were reproduced with 80% accuracy when compared with the observed hydrographs. The flow prediction of first-order ungauged streams was automated by the back-calibration algorithm. The algorithm is supported by the Shuffled Complex Evolution - University of Arizona algorithm for its optimization routine. The back-calibration procedure optimizes each first-order stream with the aid of the flow transferring function. The back-calibration procedure was imbedded in a Visual Basic.NET environment to automatically predict flow on a daily time scale and predicted was published on the internet using ESRI Arc Internet Mapping Server (ArcIMS). The project thus provides daily streamflow estimation for streams on a first-order level on every day basis, which will facilitate flow prediction of streams regardless of the size of the watersheds.

L’estimation de l’aléa hydrologique en sites non jaugés présente un enjeu important pour la gestion des risques. La complexité du phénomène réside à la fois dans la nécessité d’avoir une approche multivariée (estimation de caractéristiques multiples des crues : durées, périodes de retour) qui propose une extrapolation raisonnable des événements. SHYREG est une méthode basée sur la simulation de scénarios de crues, qui présente ces avantages. Évaluée lors du projet ANR ExtraFlo, elle présente de bonnes performances en justesse et en stabilité lorsqu'elle est calée sur des données locales de débits. Cette méthode vise à estimer des débits de crue en tout point du territoire. Elle doit donc pouvoir être appliquée en site non jaugés.Le travail de thèse présenté ici se focalise sur le transfert de la méthode vers le non jaugé en s’intéressant aux valeurs des débits simulés mais aussi à leur cohérence. Tout d’abord, une révision du calage a permis de s’assurer de la cohérence des débits simulés le long d’un cours d’eau. Ensuite, l’application d’un large panel de méthodes de régionalisation a permis de déterminer que la régionalisation devait s’appuyer à la fois sur la structure spatiale et sur les caractéristiques physiographiques des bassins. Finalement, une méthode qui régionalise SHYREG simultanément à son calage a été retenue. Une comparaison avec d’autres approches régionalisées a mis en évidence la qualité du modèle développé. / Flood hazard estimation in ungauged sites presents a major challenge for risk management. The complexity of the phenomenon arises from both the need for a multivariate approach (estimation of different flood characteristics: peak flow, volume, duration ...), and for an approach which offers a reasonable extrapolation of extreme events. The SHYREG method is based on the simulation of flood scenarios and presents these benefits. It has been evaluated during the ANR ExtraFlo project. It showed good performance in both accuracy and stability when calibrated against local discharge data. However, weaknesses have been identified when implemented in ungauged sites.The objective of the present thesis is to develop the method in order to improve the SHYREG performances in ungauged sites. Two kinds of modifications were implemented. First, the calibration of the method in gauged sites was reviewed. The main idea was to integrate more data and to take more into account the coherence between simulated discharges in different sites. Then, diverse regionalisation schemes extracted from the scientific literature were considered. Their application demonstrated the necessity to exploit information from nearby sites and the physical properties of the catchments. Finally, a version which realises the regionalisation simultaneously to the calibration has been selected. Its comparison with other method showed the quality of this new version of SHYREG.

En Europe, les crues éclair affectent principalement les régions méditerranéennes et montagneuses. Les bassins concernés sont le plus souvent de petite taille. Anticiper ce type d'événements demeure un exercice ardu parsemé de multiples difficultés. Cette thèse s'inscrit dans le cadre d'un projet commun à IRSTEA et Météo-France (projet RHYTMME) et ses deux objectifs sont :l'adaptation de modèle hydrologique de la méthode AIGA aux régions montagneuses, en tenant compte des caractéristiques de cet environnement ;l'évaluation du nouveau modèle sur des bassins de petite taille, qui sont réellement non-jaugés.Pour atteindre le premier objectif, le modèle est complexifié (prise en compte de la neige), testé sur 118 bassins versants jaugés et régionalisé. Quant au second but, des relevés de dégâts, issus de la base de données des services de Restauration des Terrains en Montagne (RTM), sont utilisés afin d'évaluer le modèle sur 123 très petits bassins dans des conditions aussi proches que possible de la réalité du terrain.Pour réaliser ce second objectif, une méthode d'évaluation est développée, fondée sur des statistiques de contingence, illustrant la concomitance entre dégâts observés et dépassement de seuil des débits simulés. Il est proposé un graphe renseignant le taux de détection (POD) et le taux de succès (SR) pour différents seuils simultanément à l'instar des courbes de ROC. Le principal résultat est qu'une évaluation sur bassins jaugés est une première étape nécessaire mais pas suffisante. En effet, les travaux illustrent le fait que les meilleurs modèles sur les bassins jaugés ne le sont pas forcément sur les petits bassins non-jaugés, sujets aux crues éclair. / In Europe, flash floods primarily occur in the Mediterranean and mountainous areas. The concerned basins are often small and ungauged with a short lag time. Anticipating such events is a tricky task with many difficulties. This thesis is part of a common project between Irstea and Météo-France (RHYTMME project) with two objectives : the adaptation of the hydrological model of AIGA method to the mountainous areas, taking into account the environment characteristics the evaluation of the new model on small basins, which are actually ungauged. To achieve the first objective, the model is complicated by the integration of a snow modelling on 118 gauged basins. Then the model is regionalized. The second objective is to use flood reports from the mountain area restoration services database (services de Restauration des Terrains en Montagne (RTM)). The model is evaluated with these reports on 123 very small basins under conditions as close to as possible the reality. To overcome the second objective, an evaluation method is developed, based on contingency statistics, illustrating the coincidence between observed damages and threshold crossing by simulated flows. A graphic with the probability of detection (POD) according to the success rate (SR) is introduced for different thresholds. Thereby, a multi-threshold approach is used to compare hydrological models like ROC-curves. The main result, highlighted by this thesis is an evaluation on gauged basins is a necessary first step but not sufficient. Indeed, the works illustrate that the decided compromises on the large gauged basins don?t lead automatically to the best performances on the small basins occurring flash floods.

Les ressources en eau sont très variables dans l'espace et dans le temps. Leur estimation à l'échelle des bassins versants requiert des modèles hydrologiques dont l'utilisation est contrainte par la complexité des bassins et une difficulté de mesure (caractéristiques du bassin et forçages météorologiques). Sur la plupart des bassins une chronique de débit est disponible à l'exutoire du bassin, on parle alors de bassin versant jaugé. Ces mesures peuvent être utilisées pour analyser le fonctionnement des bassins et optimiser les paramètres des modèles hydrologiques. En revanche, pour les bassins versants non jaugés, le fonctionnement du bassin doit être envisagé de manière différente et l'utilisation du modèle doit se faire via une régionalisation de ses paramètres.Le principe des études de régionalisation est d'utiliser les bassins versants jaugés pour développer des méthodes applicables aux bassins versants non jaugés. Ces études couvrent un grand nombre de domaines d'application: régionalisation des signatures hydrologiques (débit moyen interannuel, régime, courbe des débits classés...), des étiages, des crues, de l'hydrogramme... Dans cette thèse, nous nous sommes intéressés à l'estimation de l'hydrogramme sur des bassins versants non-jaugés, via la régionalisation des paramètres des modèles hydrologiques.De nombreuses méthodes ont été développées pour résoudre ce problème et reposent en majorité sur la connaissance des jeux de paramètres sur des bassins jaugés, donneurs de paramètres, qu'il faut alors transférer vers des bassins versants non-jaugés, receveurs de paramètres. Ces méthodes font donc deux hypothèses fortes: (i) les paramètres ont été correctement identifiés par le calage sur les bassins versants jaugés et (ii) nous sommes capables d'identifier des bassins versants jaugés et non-jaugés similaires entre lesquels le transfert de paramètres est possible. Cependant, ces hypothèses ne sont pas toujours vérifiées et, dans ce cadre, le calage régional apparaît comme une option prometteuse pour la régionalisation des modèles hydrologiques. En effet, le calage régional recherche des valeurs de paramètres communes à plusieurs bassins et permet (i) de mieux contraindre ces paramètres lors du calage et (ii) de réaliser conjointement le calage et le transfert de ces paramètres à des bassins versants non-jaugés. L'enjeu de cette thèse est d'évaluer le potentiel du calage régional pour régionaliser les paramètres du modèle hydrologique GR6J. Sur un large échantillon de bassins, nous avons en particulier étudié si le calage régional peut être une solution compétitive au regard des méthodes existantes.Dans un premier temps, nous nous sommes attachés à étudier le modèle en contexte jaugé. Nous avons mené une étude de la sensibilité du modèle à ses paramètres pour définir leur rôle fonctionnel et avons testé huit stratégies de paramétrisation.Nous nous sommes attachés à identifier le jeu de paramètres hydrologiquement optimal, c'est-à-dire un jeu de paramètres indépendant de la stratégie de calage et donc transférable à des bassins versants non jaugés. Dans un second temps, nous avons proposé une méthode innovante appelée “ calage groupé ”, basée sur le calage régional, pour relier les paramètres de GR6J à des caractéristiques mesurables du paysage. Cette méthode présente l'avantage de prendre en compte les compensations au sein du jeu de paramètre et de permettre une interprétation des valeurs de paramètres au regard des processus hydrologiques. / Water resources are variable in time and space. Hydrological models are required to assess these resources and have to cope with catchments complexity and measurement limits. Most of the catchments are gauged which means that streamflow measurements are available at the catchment outlet. In particular, streamflow measurements can be used to analyse the catchment hydrological behavior and calibrate model parameters. However for ungauged catchments, the hydrological behavior must be define by other ways and model parametrisation must be assessed through regionalization instead of calibration.Regionalization studies rely on gauged catchments to developp methodologies applicable to ungauged catchments. They cover a large range of applications from hydrological signatures predictions (mean streamflow, regime, flow duration curves ...) to low-flow or flood predictions. The present work aims at estimating streamflow hydrographs in ungauged catchments by regionalizing hydrological model parameters.Several methods have been developped to solve this problem and they mostly rely on the knowledge of parameter sets on gauged catchments to find the best way to transfer the parameters sets between catchments. Gauged catchments are then considered as donor catchments while ungauged catchments are receiver: the challenge of transfert-based methods being to assign the correct donnor to each receiver. However, these methods make two strong hypothesis: (i) parameters sets calibrated on gauged catchments are reliable and (ii) similar gauged and ungauged catchments can be accurately matched during the transfer. However, those hypothesis do not always seem to hold true. In this context, regional calibration appears as a promising way for hydrological model regionalization since regional calibration look for common parameter sets on several catchments. This way, parameters are known with less uncertainties on gauged catchment and parameters regionalization is joined with calibration. The challenge of this PhD thesis is to develop and apply a regionalization scheme, based on regional calibration, to the GR6J hydrological model. On a large dataset, we have studied if regional calibration is a competitive solution to regionalize hydrological models compared to existing methodologies (i.e. transfer-based methods).As a prelude to the regionalization study, we paid attention to analyse the model on gauged catchments. In particular, we analysed the model sensitivity to its parameters to define the functionnal role of each parameter. Eight model parametrisation schemes were tested to identify the hydrologically optimal parameter set, i.e. a parameter set that is only catchment-dependant and do not depend on calibration strategy. Secondly we present an inovative method called “group calibration”, based on regional calibration, to link model parameters and hydrological descriptors. This method is fit for taking into account compensations between model parameters, allows the hydrological interpretation of the obtained relationships and is competitive to existing methods. Last, we conceptualize the regionalization methods to better undestand how to best value regional informations and increase model performance in extrapolation. We show that all methods set a ceilling and that space for improvement do not lie in the correct matching of donnor and reciever catchments but in the identification of more easily transferable informations.

La connaissance des dynamiques hydrologiques et du transport chimique associé est indispensable à la mise en place de stratégies de gestion de l’eau et du territoire, mais elle est difficile à consolider pour tous les exutoires de bassins versants porteurs d’enjeux car ils sont souvent « non-jaugés ». A partir d’un hydrogramme observé à l’exutoire d’un bassin jaugé, l'inversion d’un modèle hydrologique à base géomorphologique rend accessible la variable pluie nette en pied de versant. Cette variable indépendante d'échelle est ensuite transposée vers un bassin non-jaugé similaire et reconvoluée en hydrogramme à son exutoire. Cette méthode de "transposition d'hydrogramme", est appliquée pour la première fois en régime pluvio-nival dans 21 bassins Québécois. Le succès de la simulation pour le couple de bassins « jaugé » et « non-jaugé » est conditionné par la similarité hydrologique, soit l’ensemble des caractéristiques assurant un fonctionnement proche entre bassins versants. Nous testons si la composition chimique de l’eau est susceptible de révéler cette similarité à l’aide d’une analyse de type spectral développée sur l’ouest de la France, à l’échelle synoptique. Des groupes de bassins versants se démarquent par une homogénéité / hétérogénéité spatiale des faciès de concentrations, et de l’aire à partir de laquelle est observée la stabilisation de ce faciès. Ceci suggère que la composition chimique des rivières est informative d’un fonctionnement hydrologique similaire. Cette hypothèse est ensuite validée grâce à la transposition avec succès sur six bassins de méso-échelle des hydrogrammes et des relations concentrationdébit (CQ), les meilleurs résultats étant obtenus lorsque le couple respecte une similarité hydrochimique. Enfin, le couplage des hydrogrammes et des relations CQ a permis de prédire efficacement, grâce au principe de similarité, les flux en bassin versant non-jaugé, au pas de temps de 15 minutes. / The knowledge of hydrological and hydrochemical dynamics is essential for policy makers to develop appropriate policies of remediation, but this goal is hard to achieve in « ungauged catchments ». Through the inversion of a geomorphology-based hydrological modelling, the net rainfall entering the stream network is assessed from a gauged catchment. This net rainfall is scale independant and can be convoluted in another (ungauged) catchment. This method is applied for the first time in a pluvio-nival context on 21 catchments from Québec. Success of the simulation depends of the hydrological similarity between gauged and ungauged catchments, which is defined as the ensemble of caracteristics insuring a close functionning. We test how riverwater chemical composition reveals this similarity through a synoptic spectral analysis developped on western France. Groups of catchments are identified by their spatial homogeneity/heterogeneoty of the hydrochemical facies, and the area at which hydrochemical stabilisation occurs. This result suggest that riverwater chemical composition is informative of a hydrological similarity. This hypothesis is successfully tested on six mesoscale catchments, by transposing hydrographs and concentration discharge relatioships (CQ), best transpositions are associated to similar catchments. Finally, the coupling of hydrological simulation to CQ curves allowed the prediction of hydrochemical fluxes in ungauged catchments with a 15 minutes timestep.