RESUME / ABSTRACT  

WIND: Analog forecasting of ceiling and visibility using fuzzy logic and case-based reasoning

by

Bjarne Hansen

Fuzzy logic and case-based reasoning are two basic techniques used to develop applications of artificial intelligence. We describe a system, WIND-2, that combines fuzzy logic and case-based reasoning to produce forecasts of airport cloud ceiling height and visibility (C&V). Knowledge about meteorological and temporal features that experienced forecasters use to construct analogous climatological scenarios is encoded in a fuzzy similarity measure. The fuzzy similarity measure is used to locate k-nearest neighbors from airports' historical databases. These nearest neighbors are adapted to produce probabilistic time-evolution forecasts of C&V. Experiments show that the WIND-2 system produces relatively accurate forecasts compared to the benchmarks of persistence and official forecasts. It takes about one second for WIND-2 to produce a forecast. The approach used contrasts with other approaches being researched and developed by the Aviation Weather Research Program in the US.

Related website:

http://iweb.cmc.ec.gc.ca/~armabha/papers