Where should fine-resolution spatial heterogeneity be captured within Earth System Models?
Abstract:Land-atmosphere interactions impact the environment in many ways, such as through partially driving our climate system, and in changing the availability and usability of our natural resources. Earth System Models (EaSMs) are being used increasingly to explore these coupled dynamics from watershed to global scales. However, many EaSMs do not adequately represent landscape-scale spatial heterogeneity that influences land surface response, as relatively coarse resolution simulations are necessitated by computational limitations. Research is needed to understand which types of spatial heterogeneity, over which biomes and climate types, should be represented such that an EaSM accurately captures the aggregate land surface response to a changing climate.
Spatial heterogeneity in a landscape arises due to differences in model forcings; in underlying soil, vegetation, and topographic properties that control moisture, energy and nutrient fluxes; and in land surface responses that arise due to spatially-organized connections. While our long-term goal is to understand how each of these sources should be represented in an EaSM, in this study we focus first on parameter heterogeneity. We apply the Regional Hydro-Ecological Simulation System (RHESSys), a distributed process-based model that was originally developed for catchment-scale applications. We explore the functional form of the hydrologic response of a RHESSys “patch” (a 200-400 m element with homogenous landscape parameters) to an invoked change. According to scale transition theory, a linear response makes it is possible to upscale (or aggregate) the model resolution without biasing the model response.
We perform RHESSys simulations for more than 500 individual catchments within the Willamette and Yakima River basins in the Pacific Northwest region of the U.S. Each catchment was imposed with incremental perturbations of temperature and precipitation. The response curves for hydrologic variables such as evapotranspiration, soil moisture, and runoff are tested for linearity using the Pearson correlation coefficient. A cluster analysis is conducted and compared to biome and climate classifications to identify landscapes where upscaling is possible without biasing the aggregate hydrologic response.