A Neural Network-based Device for Long Range, Local Area EO Weather Forecasts

An Adaptive Resonance Theory (ART) artificial neural network (ANN) Forecast Model for the Weather Impact Decision Aid (WIDA) is described. The Air Force has been involved in the development of electro-optical/infrared tactical decision aids (EOTDA) since the mid-1970s and has achieved significant operational success. However, their performance degrades beyond 24 hours due to lack of accurate long range forecast of the critical weather parameters used as inputs. Pilots and planners require precise forecasts of acquisition and lock-on ranges. A neural network model will be demonstrated which operates like the human forecaster with noisy and missing data but able to integrate much more of the existing data and to consider explicitly the nonlinearities of the atmosphere. The ART ANN Forecast Model will provide 24- to 72-hour forecasts of key weather and geophysical parameters for use in WIDA in support of long-range operations planning. The ART ANN will use state-of-the-art Very Large Scale Integration (VLSI), neural network technology for its implementation into existing research hardware/software (including WIDA related systems) and for its eventual deployment to the operational units. Our VLSI-based approach will allow one to scale the ART ANN-based WIDA support system to small, portable devices for field use.