The use of mechanistic simulation models for tactical management of annual crops has been successful in many areas. The implementation of regional strategies or the policy-level use of mechanistic models has been little explored. One of the traditional problems with the use of mechanistic models for large-scale management has been the amount of computer run-time required by physiologically detailed models. Faster platforms and decreases in the price of storage are quickly removing that limitation. There remains however, the problem of the type of data required by a mechanistic simulation model. This data is often exhaustive in its precision and expensive to collect. We have addressed the problem of scale by maintaining the resolution and granularity (and thereby mechanistic richness and relevance) of the basic model architecture while changing the required input and the range of the output. This extended abstract describes: 1) the characterization of regions using multi-layered, geo-referenced, production-relevant data, 2) the use of cluster analysis to discretize decision management regions, 3) the initialization of a simulation model from cluster data, and finally 4) the use of modified model output for strategic and policy-level decision-making and analysis.