In order to deal with the chaotic nature of weather phenomena in the field of weather forecast, one tries to run simulations with many, slightly varied initial conditions. Based on the resulting ensemble of simulation results one can judge the reliability of the forecast or estimate the probability of the occurrence of certain weather events. During routine use it is desirable to automatically identify clearly visualize interesting weather phenomena in the ensemble. However, this presupposes that the essential structures of the weather phenomena can be adequately characterized and identified in huge data sets.
In this interdisciplinary project, meteorologists will continue to develop their methods for the detection and analysis of weather phenomena, so that the computer scientists can design efficient algorithms to sift through the extensive simulation data to the weather conditions being sought. Here, the minimization of external memory accesses and the parallelization of the process will play a major role. After a careful object-like analysis of the problem formulation, we will work on a meaningful and clear visualization.
Visit the INSIGHT web site