全球降水观测计划GPM是基于卫星族群的新一代高精度全球遥感降水对地观测计划，GPM对洪涝、干旱、滑坡、飓风等水文气象灾害的大尺度监测具有重要的应用价值。然而，GPM的星载多传感器联合反演特点导致水文用户不清楚多卫星降水误差的根本来源和产生机理，限制了GPM在水文领域应用的深入和广度。本项目拟借助GPM卫星族群上的星载雷达、被动微波和红外等传感器的轨道扫描，对新一代多卫星降水联合反演系统进行剖析解译，并结合地面观测实现不同传感器误差的独立溯源，揭示GPM联合反演误差的传感器来源和产生机理；采用神经网络非线性聚类算法，实现传感器反演误差的独立订正；通过GPM与分布式水文模型的集成模拟，解析微波和红外两类传感器反演误差的水文传递特性与规律，探明传感器误差对径流模拟的影响。研究成果可以拓展GPM在大尺度缺资料地区的水文应用，并为我国第一颗水循环观测卫星WCOM的遥感降水反演提供原型算法和技术参考。

The Global Precipitation Measurement Mission (GPM) is a constellation-based satellite mission specifically designed to provide a new generation of high-precisely observations of global rainfall at a fine resolution from space. As an international science mission with integrated application goals, the GPM will play an important role in monitoring the extreme hydrometeorological events, such as the floods, droughts, landslides, and hurricanes, etc. However, it is unclear for many hydrological users with regard to the fundamental source and mechanism of precipitation errors due to the unified retrievals of multi-sensors from a satellite constellation in the GPM systems. This unexpectedly limited the various research and operational applications of GPM in hydrological community. Our project aims to tracing the error sources of satellite precipitation retrievals from different GPM sensors including the dual-frequency Precipitation Radar (DPR) on the GPM core observatory, the Passive Microwave (PMW) imagers and radiometers aboard low-Earth-orbiting satellites, and the Infrared measurements from geostationary satellites. By analyzing the orbit scanning of these sensors, we will first identify the error sources for the two GPM-based multi-satellite precipitation retrieval systems (IMERG and GPM-GSMaP), and then reveal the generation mechanisms for these sensors’ errors. Next, a novel approach, the neural network self-organizing nonlinear algorithm, will be employed to completely correct the error structures and pattern features of precipitation retrievals for each sensor. Subsequently, we will further investigate the nonlinear characteristics of error propagation of different GPM sensors into hydrologic response by using the GPM precipitation estimates to drive the distributed hydrological model. The impacts of sensor errors on the simulated runoff will be quantitatively diagnosed. This study could not only extend the hydrological applications of GPM in those ungauged areas on large scales, but also provide prototype algorithms and technical references for the first water cycle observation satellite (WCOM) of China.