PREDICTION OF THE SPATIAL-DISTRIBUTION OF THE MODIFIED MERCALLI INTENSITY USING NEURAL NETWORKS

This paper presents the development of an adaptive, non-parametric for ecast model for the direct prediction of the spatial distribution of t he Modified Mercalli Intensity (MMI) corresponding to an earthquake sc enario. The model is based on recent advances in neural networks compu tation, and is constructed through supervised learning using historica l earthquake and regional geological data as training sets. A MMI fore cast model for moderate earthquakes with magnitudes between 6 and 7 wa s developed based on data from the Loma Prieta, Coalinga and Morgan Hi ll earthquakes. For these data sets, the neural networks forecast mode l is shown to have excellent data synthesis capability; multiple sets of data can be encapsulated by a relatively simple network architectur e. Limited comparison of forecasts made by the neural networks model a nd conventional models demonstrates that improved accuracy can be achi eved. Implementation and operational advantages of the neural networks approach such as general input features, minimum preconceived knowled ge of the data sets, the ability to learn and to adapt incrementally a nd the autonomous and automatic synthesis of the structure underlying the data sets, have been illustrated.