This study addresses two main issues of scale in coupled atmospheric-hydrologic modeling and prediction. The first problem relates to the degree at which small (subgrid) scale variability of rainfall affects the accuracy of predicted water and energy fluxes from a coupled land-atmosphere model. The MM5/BATS coupled model is combined with a dynamical/statistical scheme for statistically downscaling rainfall. Through this coupled modeling framework, it is demonstrated that the nonlinear propagation of small-scale rainfall variability through the land-atmosphere system affects the variability and organization of predicted fluxes at all scales. It was found that including subgrid-scale rainfall variability affects the spatial organization of the storm system itself, surface temperature, soil moisture, and sensible and latent heat fluxes. These effects were found to occur at spatial scales much larger than the scale at which rainfall variability was prescribed illustrating the pronounced nonlinear spatial dynamics of the land-atmosphere system and its important role on hydrometeorological predictions. The results emphasize the importance of accounting for subgrid-scale rainfall variability even if the interest lies in larger-scale hydrometeorological predictions.

The second problem addressed in this work relates to the scale-dependency of nonlinear parameterizations used in coupled models. Many of the relationships used to describe land atmosphere interactions have been empirically parameterized for a specific range of scales. However, they are often applied at scales different than the ones they were intended for due to practical necessity. For illustration purposes, this study considers parameterizations which are explicit nonlinear functions of soil moisture and uses data from the 1997 Southern Great Plains Hydrology Experiment (SGP97) to quantify the spatial variability of soil moisture as a function of scale. Our analysis demonstrates that parameterizations involving variables with significant spatial variability at all scales (such as soil moisture) are scale-dependent, and unless adjustments are made to acknowledge this dependency, systematic biases may result in model-predicted water and energy fluxes. A methodology is presented by which the scale-dependency of parameterizations can be semi-analytically derived and used in the models to preserve predicted fluxes at any desired scale.