AbstractHigh computational cost associated with numerical modeling of multiscale global atmospheric chemical transport (ACT) imposes severe limitations on the spatial resolution of fixed nonadaptive grids. Recently it has been shown that the interaction of numerical diffusion caused by the crude resolution with complex velocity field of atmospheric flows leads to large numerical errors. To address the described difficulties, the authors have developed a wavelet-based adaptive mesh refinement (WAMR) method for numerical simulation of two-dimensional multiscale ACT problems. The WAMR is an adaptive method that minimizes the number of grid points by introducing a fine grid only in the locations where fine spatial scales occur and uses high-order spatial discretization throughout the computational domain. The algorithm has been tested for several challenging ACT problems. Particularly, it is shown that the method correctly simulates dynamics of a pollution plume traveling on a global scale, producing less than 1% error with a relatively low number (~105) of grid points. To achieve such accuracy, conventional nonadaptive techniques would require more than three orders of magnitude more computational resources. The method possesses good mass conservation properties; it is shown that an error in the total pollutant mass does not exceed 0.02% for this number of points. The obtained results demonstrate the WAMR’s ability to achieve high numerical accuracy for challenging ACT problems at a relatively low computational cost.