Adaptive atmospheric modeling is a relatively young discipline in the wide area of atmosphericsciences. Many obstacles – mainly of technological character – hindered theintroduction of adaptive modeling techniques into atmospheric simulation software. In recentyears, however, a number of approaches has shown up. One of the main reasons forthe recent success is the introduction of sophisticated optimization on all levels.In this work space-filling curves are used on several levels of algorithmic abstraction inorder to optimize an atmospheric modeling tool. For dynamic load balancing or irregularmeshes which rapidly change during the computation, space-filling curve based partitioningproves to be beneficial. Furthermore, space-filling curve induced indexing can help toreorder the unknowns such that data locality is maintained. Finally, the reordering leads tobetter behavior of ILU based preconditioned system solvers.These techniques have been used in PLASMA, a parallel adaptive atmospheric model forglobal studies of climate variability. PLASMA utilizes the grid generation and managementtool amatos with built in space-filling curve support.