Comprehensive evaluation of multi-satellite precipitation products with a dense rain gauge network and optimally merging their simulated hydrological flows using the Bayesian model averaging method
Comprehensive evaluation of multi-satellite precipitation products with a dense rain gauge network and optimally merging their simulated hydrological flows using the Bayesian model averaging method
复制标题
利用贝叶斯模型平均法对密集雨量网多卫星降水产品进行综合评价并优化融合其模拟水文流量
DOI：
10.1016/j.jhydrol.2012.05.055
复制
发表时间：
2012-07-25
影响因子：
6.4
通讯作者：
Ma, Mingwei
中科院分区：
文献类型：
--
作者：
Jiang, Shanhu;Ren, Liliang;Ma, Mingwei
利用密水流域（9972 km ~ 2）的密集雨量站网，对TMPA 3B 42 V6、TMPA 3B 42 RT和CMORPH三种常用卫星降水产品进行综合评价，并采用贝叶斯模型平均法将其模拟的水文流量与半分布式新安江模型进行优化融合。初始卫星降水数据的比较表明，再分析的3B 42 V6与雨量计观测的偏差为-4.54%，而两个近实时卫星数据集（3B 42 RT和CMORPH）分别低估了42.72%和40.81%的降水。由于模型参数首先以雨量计数据为基准，3B 42 V6的径流模拟性能也是三种产品中最优的，而两种近实时卫星数据集产生了恶化的偏差和Nash-Sutcliffe系数（NSCE）。尽管如此，当模型参数由每个单独的卫星数据重新校准时，两个近实时卫星产品的径流模拟的性能得到了显着改善，从而表明需要对近实时卫星输入的水文模型进行具体校准。此外，当最佳合并的径流强迫的两个近实时卫星降水产品和所有三个卫星降水产品使用贝叶斯模型平均法，得到的径流序列进一步改善，变得更加强大。总之，目前三种最先进的卫星降水产品已显示出在水文研究和应用方面的潜力。基准，重新校准，并在目前的工作中所描述的流域尺度上的径流模拟的最佳合并方案将有望成为未来利用卫星降水产品在全球和区域水文应用的参考。2012爱思唯尔有限公司版权所有。
This study first focuses on comprehensive evaluating three widely used satellite precipitation products (TMPA 3B42V6, TMPA 3B42RT, and CMORPH) with a dense rain gauge network in the Mishui basin (9972 km(2)) in South China and then optimally merge their simulated hydrologic flows with the semi-distributed Xinanjiang model using the Bayesian model averaging method. The initial satellite precipitation data comparisons show that the reanalyzed 3B42V6, with a bias of -4.54%, matched best with the rain gauge observations, while the two near real-time satellite datasets (3B42RT and CMORPH) largely underestimated precipitation by 42.72% and 40.81% respectively. With the model parameters first benchmarked by the rain gauge data, the behavior of the streamflow simulation from the 3B42V6 was also the most optimal amongst the three products, while the two near real-time satellite datasets produced deteriorated biases and Nash-Sutcliffe coefficients (NSCEs). Still, when the model parameters were recalibrated by each individual satellite data, the performance of the streamflow simulations from the two near real-time satellite products were significantly improved, thus demonstrating the need for specific calibrations of the hydrological models for the near real-time satellite inputs. Moreover, when optimally merged with respect to the streamflows forced by the two near real-time satellite precipitation products and all the three satellite precipitation products using the Bayesian model averaging method, the resulted streamflow series further improved and became more robust. In summary, the three current state-of-the-art satellite precipitation products have demonstrated potential in hydrological research and applications. The benchmarking, recalibration, and optimal merging schemes for streamflow simulation at a basin scale described in the present work will hopefully be a reference for future utilizations of satellite precipitation products in global and regional hydrological applications. 2012 Elsevier B.V. All rights reserved.