Retrieval and analysis of remotely sensed rainfall for basin hydrologic modeling.

摘要

A satellite rainfall estimation technique (called IMRA) is designed to utilize infrared (IR) brightness temperatures (TBs) as the main input data. It uses Slope and Hessian techniques to determine the cloud-top temperature gradient for discriminating rain/no-rain pixels, and allows for adjustment of derived IR-rainfall estimates using microwave TBs and spatial filtering techniques. IMRA rainfall estimates for the Peace River Basin of Southwest Florida (USA) were assessed by comparing directly with gauge and radar rainfall data, and indirectly with the corresponding streamflow predicted by the SAC-SMA model. Generally, IMRA-Slope provided better rainfall estimates than IMRA-Hessian. The daily predicted streamflow using satellite rainfall estimates was comparable to that of radar and better than the gauge data reflecting the potential of satellite rainfall estimates in hydrologic modeling.; A Haar wavelet scheme was used to merge WSR-88D radar and gauged rainfall data in order to correct the underestimation of radar rainfall depths but at the same time maintain its original spatial variability as much as possible. The scheme was evaluated in terms of streamflow simulated by the semi-distributed, physics-based rainfall-runoff model (DPHM-RS) for the Blue River Basin of South Central Oklahoma (USA) driven by event-based, hourly rainfall data. The tests included the effect of radar data accuracy, radar rainfall spatial variability, model resolution, and the gauge-radar merging techniques (wavelet scheme versus the Statistical Objective Analysis (SOA) Scheme) on the streamflow simulated by DPHM-RS.; Radar rainfall data simulated more accurate runoff hydrographs than gauged data for convective storms but significantly under-estimated the observed hydrographs for stratiform storms. The data merging schemes (i.e., Wavelet and SOA) substantially reduced radar's under-estimation of observed streamflow hydrographs for stratiform storms, with the wavelet performing better than SOA. The influence of model resolution and spatial variability of rainfall on predicted streamflow was evident, which justifies the expensive and tedious effort to account for spatial variability of rainfall and other basin properties via either dense raingauge monitoring networks, or radar meteorology, or meteorological satellites, and distributed or semi-distributed hydrologic modeling.