A Self-Calibrating Real-Time GOES Rainfall Algorithm for Short-Term Rainfall Estimates

Abstract

Estimates of precipitation from satellite data can provide timely information about rainfall in regions for which data from rain gauge networks are sparse or unavailable entirely and for which radar data are unavailable or are compromised by range effects and beam blockage. Two basic kinds of satellite-based estimates are available. Infrared data from geostationary satellite platforms such as the Geostationary Operational Environmental Satellite (GOES) can be used to infer cloud-top conditions on a continuous basis, but the relationship between cloud-top conditions and the rate of rainfall below can vary significantly. Microwave radiances are related more directly to precipitation rates, but microwave instruments are limited to polar-orbiting platforms, resulting in intermittent availability of estimates. A number of authors have made efforts to combine the strengths of both by using the microwave-based estimates to adjust the GOES-based estimates, mainly for long-term precipitation estimates at coarse spatial resolution. The self-calibrating multivariate precipitation retrieval (SCaMPR) technique represents an approach for doing the same for fine timescales and short time periods. This algorithm first selects an optimal predictor for separating raining from nonraining pixels, calibrates it to raining and nonraining areas from a Special Sensor Microwave Imager (SSM/I) algorithm, and then selects an optimal rain-rate predictor and calibrates it to the SSM/I rain rate for the raining pixels via linear regression. The performance of SCaMPR compared favorably with the autoestimator (AE) technique and GOES multispectral rainfall algorithm (GMSRA) when compared with rain gauge data for three cases. The linear correlations between the estimates and rain gauge observations were similar, but SCaMPR exhibited significantly less bias than did AE and GMSRA.