9503976 Dietterich The goal of this research is to develop and prototype methods for the automatic calibration and validation of computer models of complex systems. The platform for this research is an ecosystem model called MAPSS (Mapped Atmosphere-Plant-Soil System). The purpose of the MAPSS model is to predict the biosphere impact of global climate change. This project aims to improve the development, maintenance, calibration, and accuracy of MAPPS by combining and extending ideas from three areas of computer science: (a) declarative programming, (b) parallel computing, and (c) machine learning. Together the proposed methods will provide model developers with significant speedup through a more efficient representation and parallel implementation, a system for automatic model calibration, and a system for model validation designed to maximize predictive power. These methods will improve the efficiency of model development as well as help to ensure that the finished product is as accurate as is possible. Further, these methods should generalize beyond the MAPSS project, providing a set of tools for the development of many computer models of complex systems.
