多传感器联合反演高分辨率降水方法综述

精确测量具有强烈时空变异性的降水,是水文气象学颇具挑战的科学研究目标之一。基于多传感器联合反演降水(Multi-sensor Precipitation Estimation,MPE)的方法已成为卫星反演降水的主流趋势。首先介绍MPE方法的定义与分类,回顾MPE方法的历史发展阶段及研究现状;然后介绍主要的MPE算法,包括TRMM多卫星降水分析算法(TMPA)、气候预测中心算法(CMORPH)、全球卫星降水制图算法(GSMa P)、美国海军研究实验室联合算法(NRLB)和神经网络降水算法(PERSIANN);对比这5种主要算法的优缺点和反演精度(PERSIANN精度范围为-56%~200%,其他产品为-67%~10%),指出存在的主要问题,并且评价不同类型MPE算法的性能;最后结合目前存在的问题探讨MPE方法研究发展趋势。

Multi-Satellite Retrieval of High Resolution Precipitation: An Overview

Precipitation is a basic output flux of atmospheric process and a driving force of hydrological process. Accurate observation of precipitation with highly spatial and temporal variability has long been a challenging scientific goal in the field of hydrometeorology. Multi-sensor Precipitation Estimation (MPE) has been the mainstream trend for retrieving precipitation. And it has been a unique way of providing global high accuracy and High Resolution Precipitation Products (HRPPs). This paper describes the definition and classification of MPE, and briefly summarizes the development and status of its history. The development of MPE can be divided into two parts based on the year 1997. The commonly used MPE algorithms to produce global HRPPs include TRMM Multi-satellite Precipitation Analysis (TMPA) algorithm, climate prediction center morphing (CMORPH) algorithm, Global Satellite Mapping of Precipitation (GSMaP) algorithm, Naval Research Laboratory Blended (NRLB) algorithm and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network (PERSIANN). Then, the existing problems are put forward through comparing assets and liabilities and accuracy of the five algorithms. The MPE can be roughly categorized into two methodologies: adjustment-based techniques (TMPA and NRLB) and the motion-based techniques (CMORPH and GSMaP). The adjustment-based techniques have the longest data record, but inherently rely upon an assumption of indirect relationship between IR temperatures and rainfall rates. The motion-based techniques can provide rain rate at desired intervals. One disadvantage of this approach, however, is that the cloud tops detected by the IR imagery can move at speeds different than the precipitation features below them, and precipitation may not be properly accounted for. At present, no one algorithm performs best in any regime. HRPPs algorithms generally tend to perform best in the convective situations during summer but dropped off considerably when moving into winter and higher latitudes with varied orography. PERSIANN overestimates heavy rainfall (200%) while underestimates rainfall (56%) in the mountains. The other four HRPPs underestimate rainfall ranging from 3 to 7 mm/d(10%~67%). For future development, advanced and/or new MPE algorithms will be proposed with analyzing existing algorithms. Furthermore, the Global Precipitation Measurement (GPM) mission will be improved and extend the TRMM measurement to high latitudes, with a more frequent sampling and higher sensitivity to light and heavy rainfalls. In addition, more focus will be taken on quantitatively evaluating accuracy of HRPPs.