Predicting Air Quality in a Production-Quality Grid Environment

Air Quality Forecasting (AQF) is a new computational discipline that attempts to predict atmospheric pollution, especially high levels of ozone. The application is complex, incorporating weather models, emissions processing and chemical transport models at multiple levels of refinement; it poses substantial computational and storage requirements. Deployment in a grid is one way in which timely and reliable production of forecast results may be ensured. We have extensively studied an AQF application based upon a community Air Quality model in order to determine its development and execution requirements as well as the ability of current grid technology to satisfy all phases of preprocessing, model execution, and storage and retrieval of observational and generated simulation data. A production-quality campus grid is being built at the University of Houston using up-to-date grid software to support this application. In this chapter, we discuss AQF and its computational needs, current grid-building software, and our experiences using it to build the campus grid. We describe the shortcomings of existing grid middleware that were identified during the course of this work and present our efforts to augment available software and to overcome some of these problems, with a focus on the user environment, resource management and authentication issues.