Steering smog prediction

Abstract

The use of computational steering for smog prediction is described. This application is representative for many underlying issues found in steering high performance applications: high computing times, large data sets, and many different input parameters.

After a short description of the smog prediction model, its visualization and steering are described. The amount of computation needed to solve the governing transport equations is alarmingly high. The user has a large number of options for the display of various aspects of the simulation, and also for the interactive control of its input data.

Smooth animation is very important to monitor the evolution of pollutants and for a responsive feedback to parameter changes. Here a performance of least 15 frames per second is required. We discuss techniques that allow the user to steer the numerical solver, such that an optimal tradeoff between computation speed and accuracy can be made.