Improved ANS Lightning Predictors Using Additional Surface Wind and ElectricField Data

摘要

Because of the destruction by lightning of Atlas-Centaur 67 and its communicationsatellite payload on 27 March 1987, new launch commit criteria with respect to lightning were imposed by NASA and the Air Force for missile launches from the national ranges. These criteria are very conservative and restrict the available launch windows, especially during summer months at Kennedy Space Center (KSC) and Cape Canaveral Air Force Station (CCAFS) in Florida. In an effort to expand the launch windows while maintaining safety, we show that neural networks can be trained to generate spatio-temporal maps of predicted probabilities of lightning over the CCAFS/KSC complex. Input data used for training and testing the neural networks include the five minute averages from all 53 wind sensors, the Total Area Divergence product calculated by Watson, the occurrence of lightning strikes as recorded by magnetic direction finders, and most recently, the electric field mill data. Training the neural network lightning predictor with wind data spanning two days of data and the divergence product increased the PoD to 0.65.