Mathematics of Computational Science
Home > Hamiltonian-based numerical Methods for forced-dissipative climate prediction
Funded by: NWO
PhD: Bob Peeters
Supervisor: Onno Bokhove
Collaboration: Jason Frank (CWI, UvA)
In this project we assess the behavior of (idealized) climate models which have a Hamiltonian discretization in the limit of no forcing and dissipation. We believe that such type of discretizations lead to better climate predictions. This objective is investigated in two ways:
Difference in performance between Hamiltonian and conventional non-Hamiltonian based numerical discretizations for simplified low-order models.
Construction of a hydrostatic stratified model on the sphere based on a symplectic Hamiltonian particle-mesh method (HPM) and finite-element method (FEM).
Adiabatic ﬂow becomes two-dimensional when expressed in isentropic coordinates. This simplifies the Hamiltonian description and discretization of the hydrostatic flow (away from the boundaries, where isentropes may intersect).
Particle trajectories under small-amplitude wave flow in a static atmosphere, shown for three different isentropic (S) levels. Dots are computed particle vertical (Z) positions from the HPM-FEM code: HPM in horizontal direction, FEM in entropic direction.
2nd meeting of Wave-flow Interactions, a network in mathematics, Edinburgh 25-29 May 2009,presentation
EGU General Assembly Vienna, 2007, Hamiltonian-based numerical methods for forced-dissipative climate prediction