Development of Fully Implicit Techniques for Ocean and Atmosphere Models
In simulations of the Earth's climate one often has to integrate very long time scales. Most ocean and atmosphere models are restricted to explicit or semi-implicit time stepping schemes, forcing them to take very restricted time steps. This project aims at facilitating fully implicit time integration as a complementary approach in climate models. We will integrate various previous efforts (robust parallel linear solvers, Jacobian-free Newton-Krylov) to develop state-of-the-art fully implicit ocean and atmosphere models. Benefits of the development of implicit models include the rapid computation of steady states, the possibility to do bifurcation analysis and an efficient way of studying parameter sensitivity and other model uncertainties by data assimilation.
The project is carried out at the Dept. of Mathematics, Uppsala University, by Jonas Thies, CIM, and supervised by Gunilla Kreiss (Scientific Computing, Dept. of Information Technology, Uppsala University) and Anna Rutgersson (Air, Water and Landscape Sciences, Earth Sciences, Uppsala University).