Simulation And Projection Of Flood Season Runoff In The Headwaters Of The Yellow River Basin

The headwaters of the Yellow River Basin(upstream Tangnaihai hydrological station),yielding an annual runoff of 20 billion m3,contribute approximately 38% of the total runoff of the whole basin,and play a significant role in water resources and social impacts.However,because of climate change and human activities in the region,the runoff has been demonstrating a trend of decreasing in the recent decades.The objective of this study is to analyze the trend of the flood season runoff in the headwaters region,by using a corresponding Digital Yellow River Integrated Model(noted as DYRIM hereafter)for runoff simulation and projection.Computation time is a major challenge when calibrating a large-scale model using optimization techniques.To overcome this difficulty,we proposed a double-layer parallel method for calibration of hydrological models using high-performance computing(HPC)systems.The method can be used for parallel computation within a single hydrological model as well as in hydrological models under different scenarios in the same time.By making full use of the computation resources of the HPC system,it exploits parallelisms in both the optimization process and the hydrological models,providing a simple access to calibration of the models and investigating the sensitivity of output to the parameters.The proposed method is one of the key techniques for runoff projection in large-scale hydrological simulation.To quantify the uncertainty of recorded rainfall and its effects on hydrological simulations,a bootstrap method using combinations of rainfall station subsets was adopted in this study.By comparing the uncertainties of rainfall and runoff,we examined responses of the hydrological simulation to spatial uncertainty and non-uniformity of the rainfall.The results indicates that output of the hydrological models is sensitive to the spatial distribution of the rainfall data.When the hourly rainfall is used for input and minutely runoff is demanded for output,the appropriate average area controlled by a single station in the midstream region of the Yellow River should be no more than 40 km~2,which suggests that the gridded rainfall data with a spatial resolution of 0.1°(10 km)can satisfy the demand of the simulation.To get better understandings of the trend of flood season runoff in the headwaters of Yellow River Basin,the DYRIM,developed with the extraction of global drainage networks,as well as the double-layer parallel system,is applied as the large-scale and high-resolution distributed hydrological model.The accuracy of multisource rainfall data in this study was compared and evaluated.The historical rainfall data from the CMORPH satellite merged with the ground data have high resolution and good precision.The predicted rainfall data from the several data sets of the Coupled Model Intercomparison Project 5(CMIP5)significantly overestimate annual and seasonal rainfall amounts,demanding the modification for runoff projection.The above two types of rainfall data are used to complete the simulation and prediction of flood season runoff in the headwaters region.For the runoff simulation from 2008 to 2012,the results show that the Nash-Sutcliffe efficiency coefficient(NSE)in both the calibration and validation period are good,and the total amounts of simulated runoff are satisfactory,which proves the feasibility of the DYRIM in the headwaters region.For the trend of the flood season runoff from history to future,the observed and simulated flood season runoff show a declining trend from 1956 to 2045 with the average annual reduction of 0.12%,about 2.0 billion m3/10 a.The predicted runoff appears to decrease more significantly from 2005 to 2045 with the average annual reduction of 0.25%,about 3.7 billion m~3/10 a.