Many state-of-the-art GCMs and other coarse-resolution models use a “mosaic-of-tiles” approach for representing subgrid-scale land heterogeneity. In other words, each surface type (e.g., vegetation type; soil type) in a given grid element of these models is represented by a subgrid “patch” with a specific fractional coverage. These patches are then assumed to interact independently with the same atmosphere, i.e., the grid-resolved atmospheric variables of that grid element. While this approach improves grid-average quantities such as surface sensible and latent heat flux, it is only a stage in more realistically representing the coupling between the variable land surface and the atmosphere. More advanced stages would address the following: (i) representing the impact of subgrid-scale surface heterogeneity on subgrid-scale atmospheric dynamics and thermodynamics; (ii) given (i) above, correctly accounting for instances of non-zero spatial correlation between the subgrid-scale surface variability and the subgrid-scale atmospheric response induced by this heterogeneity (e.g., subgrid rain might fall preferentially over subgrid patches that are already wet, increasing the nonlinearity of the land-atmosphere coupling and introducing further feedback potential). Our goal is to provide an estimate of the errors introduced into the current generation of coarse resolution model simulations, none of which account for (i) and (ii) above.

Using a version of the Regional Atmospheric Modeling System (RAMS) that has been specially modified to mimic a mosaic-of-tiles GCM, we compare surface and atmospheric quantities simulated by this modified RAMS with output from similar simulations but conducted with unmodified RAMS at a high spatial resolution. We focus on case study days from July 1995 over the Oklahoma/Kansas portion of the GCIP study area. We find substantial differences between these two sets of simulations, particularly in the mid/upper convective boundary layer, and these differences can largely be attributed to the mesoscale dynamics that is captured by the high-resolution simulations but not by the “synthetic-mosaic” runs. In particular, days that exhibit enhanced mesoscale variability in surface fluxes, and hence more (and more intense) landscape-forced mesoscale circulations, show the greater differences. We also discuss the dependence of these results on model resolution. Based on this comparison, we estimate the differences in vertical heat and moisture fluxes required to explain the differences in spatially-averaged vertical profiles of potential temperature and specific humidity. From these flux differences, we create a dimensionless “flux adjustment” profile. We then apply this adjustment to the parameterized (subgrid-scale) vertical fluxes in several independent, coarse-resolution RAMS simulations; in other words, we modify the mosaic approach used in these simulations to account for the subgrid-scale dynamical impact of subgrid-scale surface variability. We examine the sensitivity of these simulations, with special attention to the resulting changes in convection, clouds, precipitation, and soil moisture, to this modification. We discuss the feedbacks resulting from this adjustment, as well as the implications for improving the parameterization of surface-atmosphere interactions in GCMs.