Reservoir Real-time Operation By Coupling Distributed Hydrological Simulation And Ensemble Precipitation Prediction

The accurate high-resolution forcing data are usually difficult to obtain. The uncertainty of forecast information hindered the process of real-time reservoir operation. The aim of this theis is:1) to assess the applicability of a global dataset in basin scale;2) to handle the forecast uncertainty by using ensemble approach;3) to develop an ensemble prediction-based reservoir optimization system. Main results and conclusions include:(1) A hydrological model (The Water and Energy Budget-based Distributed Hydrological Model, WEB-DHM) was established in a semiarid river basin (Fengman basin). The water (discharge) and energy (Land Surface Temperature, LST) cycles is simulated from2000to2006. It was concluded that the WEB-DHM is able to predict water and energy fluxes accurately over the Fengman basin by comparing with measured streamflows and MODIS/Terra (Moderate Resolution Imaging Spectroradiometer/Terra) LSTs.(2) Because accurate high resolution atmospheric forcing data is difficult to obtain for most cases, the applicability of GLDAS in basin scale water and energy budget study is evaluated. Main results include:1) the GLDAS is of high quality for daily and monthly precipitation, Tair, monthly Rlw.d, while it overestimates monthly Rsw.d,2) the GLDAS/Noah agrees well with the verified WEB-DHM and JRA-25in terms of LST, upward shortwave and longwave radiation. While the net radiation, evapotranspiration, latent and sensible heat fluxes modeled by GLDAS/Noah are larger than WEB-DHM and JRA-25simulations in wet seasons;3) the basin-integrated discharges and evapotranspiration can be reproduced reasonably well by WEB-DHM fed with GLDAS forcing except linear corrections ofRsw.d?(3) A multi-objective (the upstream and downstream safety, and water use) real-time reservoir operation system is developed by using deterministic rainfall forecast obtained from Japan Meteorology Agency (JMA). The efficiency of the WEB-DHM model is improved by63%comparing with the original WEB-DHM model. The SCE-UA (Shuffled Complex Evolution developed at The University of Arizona) is applied to optimize the multi-objective. The dynamic penalty function is applied to solve the multi-constraint. The results show that the system is able to optimize the objectives by evaluating the system for the three flood events (2001,2004and2005).(4) The ensemble precipitation generation method proposed by Saavedra et al,[2010] is improved in order to describe the uncertainties in NWPs. First, the precipitation forecast error is described by using the normalized evaluation index. Second, the definition of QPF perturbation weight is simplified by using mathematical functions instead of using proposed zones and a look up table. The ensemble QPFs generated by EPROS are comparable to that obtained from JMA by measuring their performances using CRPS (Continuous Ranked Probability Score) and RH (Rank Histogram) on2004and2005.(5) An Ensemble Prediction based Reservoir Optimization System (EPROS) is presented in this study by coupling ensemble hydrological prediction with reservoir real-time optimization. The EPROS system has been evaluated on Fengman reservoir for the flood events in2004and2005. For both events, the ensemble based streamflow predictions described the uncertainties of single prediction effectively by generating multiple (e.g.,30-member) streamflow sceneries (water levels, reservoir inflows). The system’s capability was evaluated under critical situations. The system is not sensitive to the ensemble sizes. The system is of high efficiency and easy to operate. It can provide reference for practical operation.Finally, the summary and further directions are given.