RAINSAT. A One Year Evaluation of a Bispectral Method for the Analysis and Short-Range Forecasting of Precipitation Areas

Abstract

RAINSAT uses under data to calibrate GOES visible and infra data in terms of probability of rain. It produces probability of rain maps and 3 h forecast probability of rain maps by extrapolation. An evaluation is made of RAINSAT probability of rain analyses and forecasts for the year 1985, with emphasis on the summer months, using both radar data and raingauge data for verification. In the daytime RAINSAT has skill in separating cloudy areas with near zero probability of rain from cloudy areas with a significant probability of rain. There is some skill in splitting the latter category into different probability levels. Using infrared data only. during day or night, results in a significant drop in skill. Forecasts show some skill out to 6 h. Season-to-season and within-season comparisons of monthly probability of rain relationships (PoRRs) derived from radar are made. Within-season variability is small, especially in summer and winter. There are large day-to-day variations in the occurrence of rain as a function of visible and infrared values. Regional variations in PoRRs are assessed in two ways: (1) Regional analyses based on a remote radar PoRR are verified against surface data. (2) Analyses for each region trained on local surface data are compared with those trained on the remote radar. Both approaches support the use of radar data to train the system in regions remote from the radar.