A neural network for hail size prediction

The National Severe Storms Laboratory has developed an algorithm for computing a number of attributes believed to be relevant for the detection/prediction of hail. The attributes include Doppler radar-derived parameters and other quantities representing near storm environment. In this paper, the development of a neural network for severe hail detection/prediction is outlined. Special emphasis is placed on the role of local minima in determining the optimal architecture of the neural network via bootstrapping.