| At present, the HSPF model has been successfully applied in hydrology and water quality simulation. While it has been widely used in foreign countries, it has relatively less applications in the domestic. Therefore there exists a great space to explore for the HSPF modelThis paper aims to explore the simulation principle of the HSPF hydro logic model and the hydrological manual calibration. Nash coefficient ENS and deterministic coefficient are used as the calibration criteria, which show good results; however, the parameters of the HSPF model are so many that artificial parameter calibration process is complex and needs a lot o f time and manpower. Moreover, the results of hydrological calibration have great influence on further water pollution simulation results, therefore parameters’automatic calibration is particularly important for the HSPF hydrologic model. So next the PEST model single objective and multi-objective automatic calibration are used, respectively and better results are obtain than manual calibration; the PEST software can be used to perform automatic calibration of the model, in which fewer iterations are cost to converge to optimal value and objective values and weights can be changed as required to get the wanted results.At the end of this paper, the uncertainty of the model is analyzed. The hydrological model is the simulation of the environment. As there exists uncertainty in the simulation of the environment, the model input may not be completely accurate and there is random error, the uncertainty of hydrological model can universally exist. The uncertainty of hydrological model has an important influence on hydrological forecasting and decision optimization, so the analysis of the HSPF model uncertainty is necessary.This paper takes the watershed of the Qingshan Reservoir as an example. Through the HSPF model runoff simulation, automatic calibration and uncertainty analysis, the following results are obtained:(1) The hydrological mechanism of HSPF model is studied in detail in this paper, including the process of runoff generation and confluence process;(2)This paper introduces the HSPF model modeling methods and data production methods, processes the needed map data with ArcGIS, and does basic hydrological data processing using software such as WDMUtil;(3)Through the parameter estimation, the QingShan Reservoir runoff simulation model is constructed. The sensitive parameters of runoff simulation are also selected, including INTFW, LZSN, AGWRC, BASETP, IRC, UZSN, LZETP, and INFLT, varying from large sensibility to small sensibility. The general steps and methods of manual calibration are introduced and the Castle Peak Reservoir Watershed data is also simulated which yields good results. The Nash coefficients ENS of the monthly runoff, daily runoff, hour runoff of calibration period are 0.91, 0.81 and 0.74, respectively. Meanwhile, those three Nash coefficients of the verification period are 0.93,0.87,0.80, respectively. Those results indicate that the HSPF software has good applicability in Castle Peak Reservoir.(4) Next, the principle and usage of the PEST auto calibration is researched. The PEST method is used to realize single objective parameter calibration and multi objective parameter calibration. The results show that multi objective analysis can provide better simulation results of the rainstorm In this paper, the flood analysis is therefore performed on the results of multi objective parameter calibration, which shows that 6 out of the 7 floods events meet the results, with the other meeting the requirements of hydrological forecast.(5)The joint probability distribution function of QingShan Reservoir model parameters and rainfall observation relative error is established based on Bayesian theory. Then the posterior distribution samples of model parameters and rainfall error are obtained. The uncertainty of the model parameters and prediction are also studied. |
PDF Full Text Request