Análisis estructural de modelos hidrológicos y de sistemas de recursos hídricos en zonas semiáridas

Résumé

ABSTRACT The development of hydrological modeling has experienced an extraordinary boost in recent decades. From the first conceptual models of the 60s of the last century until today, not only the number of developed models is countless, but their typologies have multiplied, their practical applications have been extended, and underlying theory to modeling has been the subject of substantial progress. This thesis will focus, firstly, in a specific type of those models, the so-called rainfall-runoff conceptual models of continuous water balance, simulating the hydrologic response of a basin, in order to carry out a massive application of several models to numerous Spanish basins, and study their structural behaviour. To do this, and after a general presentation of the hydrological cycle and its underlying processes from the perspective of these models, some of them commonly used are described in detail showing its basic principles, describing equations, parameterization used and application conditions. Then, the problem of structural analysis is discussed, including sensitivity of the models, data analysis and perturbations, uncertainties, identifiability problems, and validation of its operation. The description is self-contained and includes the mathematical developments needed for an understanding of the introduced concepts and techniques. To apply these techniques we developed a software that incorporates many of the models and techniques described, and allows a large-scale execution. The software interfaces and the results provided are briefly described. Using this software we proceeded to the practical application of the theory to an extensive set of Spanish basins, focusing perspective in singularities and differences of models, from a structural point of view, according to the degree of aridity of the considered basins. The concepts of regionalization and structural stability dependence on indicators are introduced, adopting as indicator the basin aridity. The practical applications are completed with the study through conceptual models of perturbations applied to the evaluation of effects of climate change on water resources. Following these purely hydrological issues, research is completed with a theoretical consideration of the possible effect of highly irregulars daily flow on the results of optimization and simulation models for water resources systems, with a monthly time step, and applying it to real case. Finally, we conclude with some considerations and possible extensions of the exposed techniques to semidistributed and daily new model under development. The analysis has shown the feasibility of the practical application of these methodologies, the possibility of improving the parametrization of the models, and many structural results for widely used models (structural identification matrix, covariances of parameter, eigenvalues, conditioned sensitivity, sensitivity ratio, etc.) that are in fact dependent on the aridity of the basins. Aridity not only determines the expected value and permissible range of parameters, useful information to narrow the a priori information required, but also it affects their levels of correlation and therefore structural identifiability. It has also been verified that the irregularity of daily flows, typical in semiarid areas, can affect the results of optimization and simulation models of water resources systems, due to invalidate the assumption of full monthly step diversion taken by these models. The importance of this effect depends on the configuration of the system and the flow characteristics at submonthly scales as daily. With the same objective of evaluating the effects of aridity, we analyzed the effect of intraday concentration of rainfall intensities on a daily scale infiltration, noting that the effect can be significant depending on the soil texture and initial water content. Finally we include some considerations for extend the techniques to other types of models such as daily or monthly semi-distributed using tessellation, line of great practical interest to enable water resource assessments on a large scale, in extensive territories with numerous gage stations. Model results can be contrasted by calibration with gauged watersheds, and parameter extrapolation can be carried out with the described techniques.