Implementing an intervisibility analysis model on a parallel computing system

Abstract

Parallel computing will likely play an important role in integrating spatial environmental models for large-scale systems. Intervisibility analysis with error simulation in a digital elevation model is used to illustrate an approach to developing parallel models, and to demonstrate some benefits of high-performance computing. Analyzing the structure of the application problem ensures an appropriate match between problem and parallel system implementation. Data communication is the most important computational issue in this application. The line of sight algorithms and data distribution strategies for two intervisibility models run on the Connection Machine-2 are described and partial program listings are appended. Results of this study show that intervisibility may be overestimated when the effects of DEM error are ignored.