A hybrid approach to improving rainfall forecasts

High-resolution rainfall forecasting has important benefits, such as enabling flood prediction, yet little progress has been made toward developing an effective strategy. The hybrid approach presented here combines weather physics, statistics, and artificial neural networks. The strategy is able to draw on all available information, account for and use aspects of the domain physics that are better understood, and exploit the strengths of the available data-dictated tools