Adaptive atmospheric modeling: Scientific computing at its best

Many diverse topics, including grid generation, computational geometry, graph partitioning, and the solution of partial differential equations are combined, to solve a problem related to ozone depletion in the Artic stratosphere. Obstacles to introducing adaptive methods in atmospheric modeling include, that the modeling software must be faster than real-time. The physical parameterizations of subgrid processes are usually optimized for a specific grid resolution, making it difficult to use locally refined and temporally changing grids. Numerical methods for solving partial differential equations use adaptively refined grids to increase the order of accuracy locally, thereby distributing the discretization error uniformly.