Studies On Watershed Hydrological Analysis And Hydrological Forecasting

Watershed hydrological analysis and hydrological forecasting are the two mian issues in the field of hydrology, which play a supporting role in water conservancy projects planning and construction, water resources optimal allocation and safe and efficient sustainable utilization. Due to the special geographical position and the climatic conditions of our country, the temporal-spatial water resources distribution is extremely uneven. Meanwhile, as strongly influenced by the impact of climate change and human activity, profound changes have taken place in watershed hydrological processes and temporal-spatial distribution law of water resources, intensifying the complexity of the watershed hydrological characteristics and water insecurity situation. Especially the large-scale water conservancy projects, inter-basin water transfer project, and other human activities produce a significant influence on the hydrological system, causing hydrological system deviates from the evolution law of the natural conditions, making watershed hydrological system more complex. These facts then propose higher requirements on watershed hydrological analysis and hydrological forecasting.In this thesis, we focus on the advanced theory and method of watershed hydrological analysis and hydrological forecasting. The current research could provide important scientific and theoretical support for the evolution situation analysis and optimal utilization of water resources. Therefore, it has profound theoretical significance and great engineering practical value for both reducing flood/drought disaster losses and achieving the sustainable utilization of water resources. Relevant achievements of the thesis can be used for reference by river basin administrative agencies, and have a good prospect of engineering application. The main achievements and innovations of the current paper are as follows:(1) In order to overcome the limitation that univariate trend analysis can't test the changing trend of hydrological events, we introduce the multivariate Mann-Kendal method to test the multivariate trends of annual peak, annual maximun 7 day volume and annual minimum monthly runoff, annual minimum 3 monthly runoff at the control hydrological stations in the upper reaches of the Yangtze River basin. The results indicate that flood processes overall show a decreasing trend, while low runoff processes as a whole show an increasing trend. The results also indicate that multivariate Mann-Kendal test can detect the integrated trends of joint variables, whereas univariate tests can only detect the univariate trends. Therefore, it is recommended to jointly apply univariate and multivariate trend tests in order to get a precise result.(2) Residual-based measures are commonly used criteria to provide a quantitative assessment of the differences between simulated and observed streamflow and thereby guide model calibration. However, these criteria basically cannot provide the modeler with enough power to make a meaningful comparative evaluation of the hydrological consistency of simulated results. In this regard, we develop a multi hydrological signature-based calibration methodology to achieve high hydrological consistency using residual-based measures as well as many hydrological signatures, with particular interest in extracting the distribution structure of flow time series by IED index. Results indicate that the proposed multi hydrological signature-based calibration method can provide superior results, improving hydrological consistency tremendously. In addition, the method discretizes the continuous hydrological signature values to effectively alleviate the dominance resistance, which also makes it possible for calibration process to include more objective functions.(3) Deterministic forecasts can be of limited value if they are not accompanied by information about the intrinsic level of forecast uncertainty. Prediction Intervals are commonly used to quantify the accuracy and precision of a forecast However, traditional ways to construct PIs typically require strong assumptions about data distribution and involve a large computational burden. Here, we improve upon the recent proposed LUBE method and extend it to a multi-objective framework. The proposed methods are demonstrated using a real-world flood forecasting case study for the upper Yangtze River Watershed. Results indicate that the proposed methods are able to efficiently construct appropriate prediction intervals with similar PICP but narrower width. By replacing the previous width indices with PIARW index, we are able to more completely utilize the available information about the target variable. Further, the PIARW is better suited to streamflow forecasting as it is based on a calculation of the relative width. In addition, the multi-objective implementation of LUBE makes it much easier for a decision maker to obtain a feasible PI solution that has an appropriate probability of coverage and width to suit the intended application, dramatically reducing the effort associated with trial-and-error implementation of the single-objective approach.(4) To improve the accuracy of long-term streamflow forecasting, this thesis adopts the partial mutual information method to select appropriate inputs from past observed rainfall, streamflow and teleconnection climatic factors for establishing long-term streamflow forecasting model and thereafter achieving high accurate streamflow forecasting. Partial mutual information can measure the linear and nonlinear correlation between input variables and variable of interest. In addition, it can effectively avoid selecting reductant variables. The standard function tests show that the partial mutual information method is effective for both time series model and nonlinear model; also the variables are selected in correspondence with their correlation order with variable of interest. Meanwhile, application of the partial mutual information method to the case study of forecasting streamflow in the Jinsha River Basin indicates that the partial mutual information method outperforms the linear correlation method for selecting appropriate input variables for streamflow forecasting model.