Experiments in Shower-Top Forecasting Using an Interactive One-Dimensional Cloud Model

Abstract

Experiments were made in prediction of the elevation of warm season shower-tops, both prevailing and highest, using a one-dimensional cloud model run on a real-time minicomputer system. A forecaster inter-actively altered the initial temperatures and/or mixing ratios taken from 1200 GMT radiosondes over the eastern two-thirds of the United States. Subjective methods and numerical guidance were used to estimate upper air changes from morning to afternoon, but observed afternoon surface dewpoints were employed in the developmental work. A forecast based on the unaltered initial sounding was run as a control. Observed tops were taken from radar reports within 2° latitude boxes, fine-tuned somewhat by enhanced infrared satellite imagery. Development results show root-mean-square errors (RMSE) of less than 2.0 km can be achieved for both prevailing and highest tops if the surface dewpoint is specified accurately. Independent tests were consistent only for highest tops, and the RMSE increased to 2.54 km when forecasters had to predict the dewpoint. “Prevailing tops” are apparently difficult to distinguish from “highest tops” reliably in real-time conditions.