Ice sheets have strong influence on the climate system. Numerical simulation provides a mathematical tool to study the ice dynamics in the past and to predict their contribution to climate change in the future. Large scale ice sheets behave as incompressible non-Newtonian fluid. The evolution of ice sheet is governed by the conservation laws of mass, momentum and energy, which is formulated as a system of partial differential equations. Improving the efficiency of numerical ice sheet modeling is always a desirable feature since many of the applications have large domain and aim for long time span. With such a goal, the first part of this thesis focuses on developing efficient and accurate numerical methods for ice sheet simulation.

A large variety of physical processes are involved in ice dynamics, which are described by physical laws with parameters measured from experiments and field work. These parameters are considered as the inputs of the ice sheet simulations. In certain circumstances, some parameters are unavailable or can not be measured directly. Therefore, the second part of this thesis is devoted to reveal these physical parameters by solving inverse problems.

In the first part, improvements of temporal and spatial discretization methods and a sub-grid boundary treatment are purposed. We developed an adaptive time stepping method in Paper I to automatically adjust the time steps based on stability and accuracy criteria. We introduced an anisotropic Radial Basis Function method for the spatial discretization of continental scale ice sheet simulations in Paper II. We designed a sub-grid method for solving grounding line migration problem with Stokes equations in Paper VI.

The second part of the thesis consists of analysis and numerical experiments on inverse problems. In Paper IV and V, we conducted sensitivity analysis and numerical examples of the inversion on time dependent ice sheet simulations. In Paper III, we solved an inverse problem for the thermal conductivity of firn pack at Lomonosovfonna, Svalbard, using the subsurface temperature measurements.