ZEUS: A Knowledge-Based Expert System that Assists in Predicting Visibility at Airbases.

Artificial Intelligence AI knowledge-based expert system KBES was developed to demonstrate the feasibility of using this approach to assist forecasters in predicting local visibility. The study developed a knowledge structure for handling rules and data needed to advise on advective and radiation fog formation processes. A review of alternatives for developing the needed software led to selection of the EXSYS, AI programming shell for developing and running a KBES on PC-compatible computers. The KBES for advising forecasters is called Zeus and was developed in three versions to be used at each of three airbases Dover, Seymour Johnson, and Fort Bragg. The system was enthusiastically used at the bases for a 2-mo period as reported in a user survey. The Zeus forecast results from user interactions and from a historical data evaluation that produced overall skill scores of 0.35 and 0.38, respectively. The skill scores were significantly better than for forecasters at one of the bases during the user test period. A drawback of the system is a tendancy to underpredict the frequency of occurrences of low-visibility categories. The good performance and the high user acceptance demonstrates a successful proof-of-concept. Further development of the approach is recommended. Author