Implementation and testing of variable soil depth in the global land surface model CLM4.5

Most land surface models currently have soil extents of constant depth of 2-5 m, as previously no good global depth to bedrock (DTB) datasets were available. We developed a global 1-km resolution DTB dataset separately for upland and lowland areas using global topographic, vegetation, and geologic data and tuned with the STATSGO soil data for uplands and DTB data from groundwater wells in several U.S. states for lowlands and valley bottoms. By using 1 km estimates of the fraction of upland and lowland areas, we derive a global terrestrial map of the weighted mean DTB.

Here, we use this information within the National Center for Atmospheric Research's Community Land Model version 4.5 (CLM4.5), which is a component of the Community Earth System Model version 1.2. We replaced the fixed soil domain of 10 layers with a bottom depth of 3.8 m to a soil domain that varies between five (bottom depth of ~0.5 m) and 14 layers (bottom depth of 28 m) depending on where the weighted mean DTB from the dataset lies within the current model node structure. In many lowland regions, the soil domain is extended, whereas in most mountainous regions, it is shallower than the current soil domain in CLM4.5. By including variable soil depths, we can explicitly model the unconfined aquifer when the water table is within the soil domain.

In an offline run of CLM4.5 forced using the near-surface atmospheric data, such an implementation impacts the simulated water budget in some regions. For instance, surface latent heat flux is relatively smaller in the southwestern U.S. and is also changed substantially in scattered pockets of the mid- to high northern latitudes. Also, subsurface drainage is reduced in certain mountainous areas like South China and northwestern North America during certain seasons, and runoff is increased in those same mountainous regions throughout the year. Across Amazonia, there is increased subsurface drainage and runoff in boreal spring. Also, since the heat capacity increases when the soil domain is extended, the annual cycle in deep soil temperatures is reduced in regions where the soil domain is deepened. This change in deep soil temperature when it occurs in the Arctic may significantly affect the representation of permafrost in the model.