Statistical Adjustment of Satellite Microwave Monthly Rainfall Estimates over Amazonia

Abstract

A technique is described for quantifying nonlinear adjustment to the half-degree monthly rainfall estimates over land, derived from the Special Sensor Microwave Imager (SSM/I) observations. The technique uses a function that represents the distortion between the probability distributions of SSM/I and rain gauge half-degree monthly rainfall. The proposed adjustment procedure is assessed and evaluated with a 10-yr period (1988?97) of SSM/I observations over the northern South America region (15°N?15°S, 80°?35°W), which includes the Amazon basin. The rain estimates are derived from the National Aeronautics and Space Administration?s Goddard Profiling (GPROF) algorithm instantaneous rain-rate retrievals, averaged in half-degree areas and aggregated into monthly accumulations. Monthly rain accumulations from a network of 650 rain gauges distributed across the Amazon basin and the state of Ceara in northeastern Brazil are used for calibration and validation, respectively. Assessment of the adjustment relationship with the validation dataset shows an overall 45% GPROF?gauge root-mean-square (rms) difference reduction with respect to no adjustment, which is due mainly to elimination of the mean bias, and a respective 10% increase in the GPROF?gauge correlation. The rms difference between 5° gridbox monthly rain averages of adjusted GPROF and rain gauges is 23% of the mean rain, and the corresponding correlation coefficient is 0.94.