Contributi allo sviluppo di modelli idrologici accoppiati previsionali e Montecarlo

A careful management of the environment and of water resources requires modern eco-hydrological models which allow the evaluation of flood events and of the response natural systems, especially in a context of accelerated environmental changes due to human activities and climate change. Modelling such complex dynamics requires mathematical tools, such as modern hydrological response models, rainfall stochastic models and meteorological models. In the present thesis the research activities have been focused on coupling physically-based hydrological models and models which prescribe relevant hydro-meteorological forcings on the basis of predictions (meteorological models) or probabilistic representations of statistical features of climate variables (weather stochastic generators). A geomorphological model of the hydrological response applied to a number of relevant case studies is here developed. The use of a reliable flood forecasting model is extremely beneficial to prevent extreme events because of its capability to predict the behaviour of a hydrological system under different initial and boundary conditions. The hydrological model implemented constitutes a robust basis for the study and the development of coupled meteorological and hydrological models. Special attention is devoted to the analysis of high-resolution rainfall fields which are forecasted by a LAM meteorological model (ETA model); spatial and temporal analyses are carried out to test the main characters of precipitation forecasts. Hydrological, ecological and water resources applications need as forcing input weather time series with a suitable spatial and temporal detail. Measured ground-based meteorological data are often inadequate, particularly because of their limited length which may lead to a mis-representation of extreme events. Stochastic weather generators are a viable technique for simulating time series consistent with the statistical characters of observed data. In this thesis a new stochastic model is developed, based on a multivariate autoregressive framework. The objective is the generation of a coherent set of hydro-meteorological variables including maximum and minimum air temperature, maximum and minimum air humidity, atmospheric pressure, daily mean wind velocity and direction and the meteorological tide component for a given geographic location. These variables are required to force models applied to coastal hydrological systems, where the presence of a tidal component is important in determining the response to ordinary and extreme events (e.g. as in the important case of the Venice Lagoon). However, the daily resolution is not detailed enough to describe the meteorological tidal level fluctuations in usual hyrologic applications; a down-scaling technique is thus applied in order to obtain hourly time series of simulated tidal levels. The results highlight the model capability to reproduce the main statistics both at daily and hourly resolution. Moreover the stochastic generator is able to well reproduce observed extreme events frequencies.