An accurate simulation of global atmospheric chemical transport presents an enormous challenge. Chemical transport models (CTMs) combine chemical reactions with advection by a meteorologically predicted flow velocity. The resulting system of equations is extremely stiff, nonlinear and involves a large number of chemically interacting species. The difficulty of solving these equations imposes severe limitations on the spatial resolution of the CTMs implemented on a uniform or quasi-uniform grids. Relatively crude spatial resolution introduces significant numerical diffusion into the system. It was shown that this spurious diffusion significantly distorts the pollutant mixing and transport dynamics for typically used grid resolution.

Here we present dynamically adaptive multilevel Wavelet-based Adaptive Mesh Refinement (WAMR) method for numerical modeling of atmospheric chemical evolution equations that provides a significant reduction in computational cost, therefore address the described above numerical difficulties. WAMR allows a fine grid in the regions where sharp transitions and cruder grid in the regions of smooth solution behavior. Thus WAMR results in much more accurate solutions than conventional numerical methods implemented on a uniform or quasi-uniform grids. The algorithm allows to provide error estimates of the solution that are used in conjunction with an appropriate threshold criteria to adapt the non-uniform grid.

The method has been tested for a variety of benchmark problems including numerical simulation of traveling pollution plumes. It was recently discovered that pollution plumes in the remote troposphere can preserve their identity as well-defined structures for two weeks or more as they circle the globe. Present Global CTMs implemented for quasi-uniform grids are completely incapable of reproducing these layered structures due to high numerical plume dilution caused by numerical diffusion combined with non-uniformity of atmospheric flow.

The simulations show excellent ability of the algorithm to adapt to a solution having different scales at different spatial locations so as to produce accurate results at a relatively low computational cost.

This work is supported by a grant National Science Foundation under Award No. HRD-1036563.