Knowledge Representation in an Expert Storm Forecasting System

Abstract

METEOR is a rule- and frame-based system for short-term (3-18 hour) severe convective storm forecasting. This task requires a framework that supports inferences about the temporal and spatial features of meteorological changes. Initial predictions are based on interpretations of contour maps generated by statistical predictors of storm severity. To confirm these predictions, METEOR considers additional quantitative measurements, ongoing meteorological conditions and events, and how the expert forecaster interprets these extra factors. Meteorological events are derived from interpreting human observations of weather conditions in the forecast area. To accommodate the large amounts of different types of knowledge characterizing this problem, a number of extensions to the rule and frame representations were developed. These extensions include a view scheme to direct property inheritance through intermingled hierarchies and the automatic generation of production system rules from frame descriptions on an as-needed basis for event recognition.