A dynamic grid adaptation (DGA) technique is used to numerically simulate tracer transport at meso- and regional scales. A gridpoint redistribution scheme is designed to maximize heuristic characteristics of a “good” grid. The advective solver used in conjunction with the DGA is the multidimensional positive definite advection transport algorithm (MPDATA). The DGA results for regional tracer transport are compared against results generated using the leapfrog as well as MPDATA advection schemes with uniformly spaced, static grids. Wind fields for all tracer transport algorithms are provided by the fifth-generation Pennsylvania State University–NCAR Mesoscale Model (MM5). A mesoscale-sized test case with idealized initial condition and wind field clearly shows qualitatively and quantitatively the advantage of using the dynamic adaptive grid, which is a marked reduction in numerical error. These results are further corroborated by more realistic test cases that used NCEP–NCAR reanalysis data from 6–11 March 1992 to set initial and boundary conditions for (i) a mesoscale-sized, 24-h simulation with an idealized initial tracer field, and (ii) a regional, 5-day simulation with water vapor field initialized from the reanalysis data but then treated as a passive tracer. A result of interest is that MPDATA substantially outperforms the leapfrog method with fourth-order artificial dissipation (central to MM5) in all of our test cases. We conclude that with dynamic grid adaptation, results with approximately the same accuracy as a uniform grid may be obtained using only a quarter of the grid points of the uniform grid MPDATA simulations. Compared to results generated using the leapfrog method on a uniform grid, the DGA does even better.