淮河上游流域分布式水文模拟的不确定性研究

由于降水分布的时空不均匀性、分布式水文模型的复杂性及其参数率定的敏感性等原因，分布式水文模型的数值模拟存在很大不确定性。本项目考虑流域降水资料输入、参数、模型结构、产品输出等方面的不确定性，全面研究这些不确定性对分布式水文模拟的影响。基于VIC分布式水文模型，在淮河上游流域进行径流模拟和不确定分析。卫星和站点融合定量降水估计产品可以减少卫星降水估计产品的误差。本项目比较和分析4套最新经过偏差订正的卫星和站点融合定量降水估计产品，包括中国气象信息中心新发展的CMORPH站点降水融合产品，研究这些卫星融合降水估计产品和中国站点格点降水估计产品对该流域模拟的影响。考虑了模型的不确定性，对选取的参数进行参数率定和不确定性分析，并研究使用不同训练数据的率定对模拟的影响。同时，采用不同参数组合和不同降水输入进行水文模拟集合预报，并利用概率预报和贝叶斯平均方法的后处理方法，来提高水文模拟预报精度。

Because of the heterogeneity of precipitation distributions in space and time, complex model structure, sensitivities in calibration processes, streamflow simulations by a distributed hydrologic model always pose great uncertainties. This study takes account into the uncertainties of input precipitation, parameter calibration, model structure， and output produtcs in hydrologic modeling, and investigates the impacts of these uncertainties on streamflow simulations by a distributed hydrologic model. The distributed Variable Infiltration Capacity (VIC) model is used to simulate streamflow and analyze uncertainties over the upper Huaihe River basin. Satellite-gauge quantitative precipitation estimate (QPE) products may reduce the errors in near real-time satellite precipitation estimates by combining rain gauge data. This study comprehensively evaluates four of the latest satellite-gauge QPEs, including the recent CMORPH satellite-gauge merged product developed at the National Meteorological Information Center (NMIC) of the China Meteorological Administration (CMA). The four satellite-gauge QPEs are statistically evaluated over the upper Huaihe River basin, compared to the China Gauge-based Daily Precipitation Analysis newly developed at CMA/NMIC. The selected parameters are calibrated considering the model error, and uncertainty analysis is conducted based on streamflow simulations. Also, the impact of calibration processes by using different training data on streamflow simulations is analyzed. The ensemble forecasts are implemented using different parameter sets and input precipitation forcing data. The probabilistic steamflow forecasts and the postprocessing method by Bayesian average method are used to improve the accuracy of streamflow outputs.