PNNL

The Science 

Earth system models (ESMs) often divide land into large grid cells, but real landscapes vary over short distances as a result of differences in topography, vegetation, and weather patterns. This makes it difficult to represent important small-scale land surface variability that influences runoff, snow accumulation, and energy exchange. Researchers tested a method that breaks down each model computational unit into smaller topography-based subgrid units (TGUs) defined according to elevation and terrain structure. They created TGUs across the continental United States using both grid-based and watershed-based model units at four comparable spatial scales. Comparing how each approach captured variations in elevation, slope, vegetation cover (using the normalized difference vegetation index), and surface hydrometeorology—including precipitation, temperature, and snow water equivalent—they found that watershed-based TGUs captured small-scale land surface variability more effectively than grid-based TGUs for several key features that influence land surface processes.

The Impact 

Across spatial scales, this study found that watershed-based TGUs better represented patterns in topographic slope, vegetation cover, and hydrometeorology compared with grid-based TGUs, although the two approaches performed similarly for elevation. At the finest scale, watershed-based TGUs better matched observed precipitation, temperature, and snow water equivalent at most SNOwpack TELemetry (SNOTEL) sites. These findings show that watershed-based land discretization can improve how ESMs represent realistic land surface variability without requiring high grid resolution. This approach can also improve how land model performance is evaluated using site observations and help strengthen simulations of snow, water, and land-atmosphere interactions, supporting more realistic land surface representation in ESMs.

Summary 

This study evaluated how well two land discretization approaches—the use of grid- and watershed-based computational units—captured land surface heterogeneity when each was further discretized into topography-based subgrid units (TGUs). TGUs were derived across the continental United States at four comparable spatial scales using consistent parameters and assessed using statistical metrics for elevation, slope, vegetation cover (NDVI), and hydrometeorological variables (precipitation, air temperature, and snow water equivalent).

The results showed that watershed-based TGUs were more consistent at capturing heterogeneity linked to slope, land cover, and surface hydrometeorology across spatial scales, while both approaches were similar for elevation. When evaluated against site observations at the finest spatial scale, watershed-based TGUs better reproduced precipitation, temperature, and snow water equivalent at most SNOTEL sites, indicating potential advantages of watershed-based subgrid structure for improving model realism and evaluation.

Text was initially generated using artificial intelligence and subsequently reviewed, refined, and validated by experts at Pacific Northwest National Laboratory.

Contact 

Sally McFarlane, Earth and Environmental System Modeling Program, sally.mcfarlane@science.doe.gov

L. Ruby Leung, Pacific Northwest National Laboratory, ruby.leung@pnnl.gov

Funding 

This research is supported by the Office of Science of the Department of Energy as part of the Earth System Model Development program area through the Energy Exascale Earth System Model (E3SM) project. Pacific Northwest National Laboratory is operated by Battelle for the Department of Energy under contract DE-AC05-76RL01830. The reported research used the Department of Energy’s Biological and Environmental Research Earth System Modeling program’s COMPY computing cluster, located at Pacific Northwest National Laboratory. The authors thank the two reviewers for their insightful comments, questions, and suggestions. Notice: This manuscript was authored by Battelle Memorial Institute under contract DE-AC05-76RL01830 with the Department of Energy.