Abstract

Water resources availability and its variability is one of the most pressing global problems. Hydrological models are useful to understand the water balance of a basin, providing information for water resource forecast, assessment, and management. The effectiveness of the models in estimating the freshwater space-time availability and variability, however, depends on concurrent and explicitly modelling of all water budget components instead of a single component estimation and optimization. The whole water budget modelling at basin scale requires a combined solution from hydrological and spatial information tools, in-situ and remote sensing data. The present dissertation describes an effort to improve estimation of each water budget component, and water budget closure at various spatial and temporal scales, by combining JGrass-NewAge model system, GIS spatial toolbox, in-situ and remote sensing data. JGrass-NewAge is a system which deploys modern informatics to facilitate models maintainability and reproducible research. It integrates advanced GIS features and the Object Modelling System version 3 infrastructures, which allow for a component-based modelling experience. This means that JGrass-NewAGE is not actually a model, but a set of elements (the components) that can be combined just before runtime to produce various modelling solutions. Topics like calibration of processes, the interpolation forcing and the assessment of forecasting errors can therefore be faced with consistent and solid approaches. In this context also the use of some remote sensing resources can be inserted appropriately and with new techniques. For all the analysis, two significantly different basins, in terms of size and hydrological processes, are considered as case studies. These are Posina river basin in northeast Italy (small size basin) and Upper Blue Nile basin(large size basin) are used as case study. The uDig Spatial Toolbox (uST) GIS infrastructure that is used for generating the hydromorphological parameters is described in the second chapter. A large number of tools are included in uST for terrain analysis, river network delineation, and basin topology characterisation. In addition, the geomorphological settings necessary to run JGrass-NewAGE are shown. The third chapter studies the effect of spatial discretisation and the hillslope size on basin responses. The possible epistemic uncertainty exerted by the use of sub basin spatial discretisation of topographic information in the semi-distributed hydrological modelling has been studied. The use of different spatial representation in hydrological modelling context has been also studied by comparing JGrass-NewAGE with a model configuration called PeakFlow. The latter is an implementation of the geomorphological unit hydrograph based on the width function. The experiment indicates that the Peak-Flow model, with a more accurate spatial representation, reproduce the storm events slightly better than the JGrass-NewAGE model. In the fourth chapter, the thesis set-up JGrass-Newage modelling solution for the estimation of hydrological modelling inputs (rainfall, snow, temperature data) and estimates them, as well as with their errors. Regards to the meteorological forcings (mainly temperature and precipitation), in Posina river basin where there are relatively dense meteorological stations, the effects of different interpolation schemes were evaluated. Since the hydrological processes from rainfall is different from snowfall, a new method of separating rainfall and snowfall was introduced using MODIS imagery data. In the fifth chapter, JGrass-NewAGE was used to estimate the whole set of water balance components. For evapotranspiration (ET) estimation, the Priestley-Taylor component of JGrass-NewAGE is used. In order to calibrate its parameter a new method based on the water budget was implemented. This method uses two different hypothesis on available data (budget stationarity "Budyko hypothesis", and local proportionality of actual evapotranspiration to soil moisture availability). Finally the spatial and temporal dynamics of water budget closure of Posina river basin is presented. The sixth chapter concerns about the inputs data, particularly precipitation, for water balance modelling in a region where ground-based gauge data are scarce. Five high-resolution satellite rainfall estimation (SRE) products were compared and analysed using the available rain gauge. The basin rainfall is investigated systematically, and it was found that, at some locations, the difference in mean annual rainfall estimates between these SREs very high. In addition to the identification of the best performing products, the chapter shows that a simple empirical cumulative distribution (ecdf) mapping bias correction method can provide a means to improve the rainfall estimation of all SREs, and the highest improvement is obtained for CMORPH. In the seventh chapter, using the capability of JGrass-NewAGE components and different remote sensing data, the spatio-temporal water budget of Upper Blue Nile basin is simulated. The water budget components (rainfall, discharge evapotranspiration, and storage) were analysed for about 16 years at daily time step using the modelling solution and remote sensing data set. For the verification of the approaches followed, wide ranges of remote sensing data (MODIS ET product MOD16, GRACE, and EUMETSAT CM SAF cloud fractional cover) are used.