Storm-wide precipitation retrievals

Abstract

The monitoring of precipitation is important for scientific purposes, such as the study of world weather patterns, the development of global precipitation maps, and the tracking of seasonal and diurnal variations in precipitation rates. Over time many observation methods have been used to estimate precipitation: rain gauges, ground based radar systems, and visible, infrared, and passive microwave sensors in orbiting satellites. This research project uses data from the Advanced Microwave Sounding Unit (AMSU-A and AMSU-B) which consists of passive microwave sensors that collect data in the opaque water vapor and oxygen microwave absorption bands. This data supports 15-km resolution global precipitation rate estimates. The goal of this research is to develop a computational method that will improve the accuracy of these precipitation estimates by including spatial information in the precipitation retrieval, which is currently pixel based. This spatial information, which consists of the precipitation rate at each pixel of the image, is used to divide the data into separate storms, where a storm is defined as a precipitation region that is separated from other regions by an area of low or zero precipitation. Once storms have been identified, a neural network is used to estimate the integrated precipitation rate over each storm using as input several feature vectors that characterize the initial storm-wide precipitation rate estimates. Then the estimate of integrated precipitation rate is used to adjust the precipitation values of the pixels that correspond to the storm.

(cont.) These methods have resulted in a decrease in the mean-square-discrepancy of the estimate of integrated precipitation rate, as compared to NEXRAD ground-based radar systems, by nearly a factor of two.