Research On Hydrological Simulation And Watershed Management Model Under Uncertainty

Under the coupling impacts of climate change and anthropogenic activities, watershed systems can no longer cope with changing environmental conditions in time, a lot of serious environmental problems and natural hazards have occurred in the watersheds around the world. For example, water pollution is aggravating, and floods and droughts have become more frequent. Watershed models provide efficient tools for integrated studies of the major physical, socioeconomic, and political aspects of watersheds. However, complex watershed systems and changing environmental conditions make uncertainties be incorporated within the watershed modeling studies, and as such, these uncertainties should be carefully considered, reflected and analyzed in the watershed models. As for hydrologic modeling, uncertainties exist in model inputs, parameters and structure. To perform hydrological simulations has to quantify and evaluate these uncertainties, and moreover, efficient approaches for dealing with them should be proposed to reduce the simulation uncertainty. Also, uncertainties in inputs and parameters should be effectively reflected in watershed management models. Consequently, this dissertation undertook the hydrological simulation and watershed management model under uncertainty through analyzing the trends in the precipitation and streamflow time series, applying the SLURP hydrologic model to simulate the hydrological processes in the Xiangxi River watershed, and investigating the effects of watershed subdivision level on the simulations as well as the uncertainty of parameters and discharge simulations. Based on these, a distributed reservoir watershed hydrologic model was developed to deal with uncertainties in precipitation inputs. To demonstrate the watershed management under uncertainty, two kind of watshed models, i.e. fuzzy constrained eco-friendly reservoir operation model and inexact land use allocation model, were developed in this dissertation. Main research works and innovative achievements were introduced as follows.(1) Multiple trend analysis methods were simultaneously applied to test and interpret the trends in precipitation and streamflow time series in the Xiangxi River watershed. The possible trends in various precipitation and streamflow statistics, the beginning of trends, and change points were identified in the Xiangxi River watershed. Mostly, streamflow time series exhibited a downward trend, and the negative trend in the monthly precipitation of September was also found to be statistically significant. The trends in streamflow mainly started in1970s, but the statistic characteristics of streamflow time series significantly changed after1988. As a result, the relation between precipitation and streamflow was accordingly changed, and anthropogenic activities, especially for the small hydropower construction were found to be mainly responsible for such changes in the Xiangxi River watershed.(2) The SLURP hydrologic model was applied to simulate the hydrological processes in the Xiangxi River watershed, and impacts of watershed subdivision level on the modeling efficiency, water balance components and runoff contribution from various land types were also examined. To increase the watershed subdivision level would increase the modeling efficiency, but the improvement of hydrograph simulations was limited. However, the simulations of water balance components and runoff contributions from various land types were found to vary a lot. In addition, fine watershed subdivision would decrease the sensitivity of parameters, and the difference became slight under fine subdivisions. Hence, fine watershed subdivisions would help the simulations approach stable.(3) A Markov Monte Carlo based formal Bayesian method was employed to evaluate the parameter uncertainty and simulation uncertainty of the SLURP model with various watershed subdivision level. The parameter uncertainty was found to vary a lot under different subdivision level, but to increase the watershed subdivision level would help to reduce the daily discharge simulation uncertainty. However, there was a threshold, after which the reliability of hydrologic model could not be improved through increasing the watershed subdivision level. The obtained findings would support the practical application of hdyrological models.(4) A chance-constrained control based distributed reservoir watershed hydrologic model was developed. Through representing the rainfall intenstiy with a stochastic distribution, the developed hydrologic model was expected to handle the uncertainty in precipitation inputs at the temporal scale, to take the scaling effects of precipitation inputs into hydrological modeling processes, and to improve the simulation of overland surface flow generation. Compared to the SLURP model, both modeling efficiency and performance was significantly improved, especially for the runoff deviation, which was effectively controlled at a satisfactory level.(5) Inexact reservoir operation of a cascade reservoir system and optimal land use management models were formulated to demonstrate the consideration of uncertainties in watershed management models. The inexact reservoir operation model introduced the fuzzy downstream ecological demands into the operation model as constraints, and a fuzzy constrained self-adaptive algorithm was proposed to solve the inexact model. As a result, the electric generation of the cascade system was maximized, and meanwhile, the tradeoff between electric generation and downstream ecological flow was effectively balanced. The inexact land use management model considered uncertainties expressed as interval, fuzzy and stochastic distribution, and employed the two-stage decision theory into programming framework for soil erosion control. After the solution of the inexact programming model, the maximum net benefit of watershed system was finally achieved under uncertainty.