Where should fine-resolution heterogeneity be captured in land surface models?

Ryan Edward Hull

Climate change and variability and its associated impacts are influenced by the interplay between land and atmospheric processes. Quantifying these effects is one of the roles of land surface models, such as macroscale hydrologic models, which requires an understanding of the underlying physics as well as access to appropriate input data and adequate computational tools. Currently, large-scale hydrologic simulations are limited by the availability of both high-resolution gridded datasets and computational power so there is a need to critically consider when and where to apply high-resolution models. Despite the inherent nonlinearity of most hydrologic processes, their complex interactions can sometimes lead to nearly linear response curves. These areas of linear response can be upscaled with minimal increase in the overall error as long as lateral interactions, such as soil moisture redistribution between fine-scale spatial elements, is not a critical factor for the aggregate response. In this study we address three questions: Is it possible to aggregate over large climate gradients? Does accounting for fine-scale lateral redistribution impact watershed-scale response over climate gradients? And how can we use this information to inform upscaling efforts. First, we ran a series of simulations using a hydro-ecological model (RHESSys), independently scaling temperature (from -5 to +5 ˚C) and precipitation (25 - 175%). Then we computed a linear regression at each grid cell for evapotranspiration (ET) and net surface and subsurface flow (subsurfacein + runoffin - subsurfaceout - runoffout; Q) over different sized ranges of climate perturbations and used the square of Pearson's correlation coefficient as a measure of linearity. To test the impact of explicit lateral connectivity between grid cells, we repeated the above analysis with lateral redistribution disabled and compared the results. Finally, we generated "upscalability" recommendations per bioclimatic zone according to the distribution of linearity scores and found that whether upscaling is possible for a given response variable depends on the size of the climate gradient and the process being studied. These results have applications for upscaling existing models and datasets, and identifying the appropriate process scale when initializing new models.

Where should fine-resolution heterogeneity be captured in land surface models?

Files and links (1)

Abstract

Metrics

Details