A Parallel Adaptive Wavelet-based Mesh Refinement for Global Atmospheric Chemical Transport Modeling

Wavelet-based Adaptive Mesh Refinement (WAMR) algorithm that dynamically matches the local computational grid resolution to the local numerical solution behavior allows efficient use of required computational resources. In this work we apply WAMR method to the numerical simulation of global atmospheric chemical transport. Numerical modeling of such problems presents enormous computational difficulties associated with simulating a wide range of time and spatial scales and a large number of reacting chemical species. It was shown that static grid approach applied to computational simulation of global atmospheric Chemical Transport Models (CTMs) often results in poor spatial resolution that introduces large numerical errors into the calculations.

WAMR method is a three-dimensional parallel adaptive technique that employs fine grid only in the regions where small spatial structures occur without requiring small grid-spacing throughout the entire domain. Therefore the method allows the computational grid to adapt dynamically to the structures in the spatial distribution of chemical compounds as they evolve in time. The direct control of numerical error through appropriate grid refinement criteria allows good spatial resolution of localized fine scales in the numerical solution while significantly increasing computational speed and decreasing storage. The multilevel adaptive grid is represented by a dynamic data structure that allows rapid access to individual points, fast inter-grid operations and grid rearrangement. The developed data structure supports execution of parallel algorithm that uses run-time partitioning and load-balancing scheme for the adaptive grid.

The adaptive method has been tested for a variety of problems including numerical simulation of traveling pollution plumes. Numerical tests show that the method efficiently captures fine structures of chemical species spatial distributions due to much better local resolution. These fine structures cannot be resolved by conventional global CTMs that use fixed-resolution grids. Therefore WAMR method has significant advantages over the conventional computational methods based on static grids in terms of accuracy, convergence and computational cost. It is demonstrated that the developed wavelet-based adaptive technique is capable to accurately simulate multi-scale atmospheric chemical transport problems that are difficult or impossible to simulate with standard approach based on static grid presently used by the majority of global CTMs.

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