Implementation of global soil databases in the Noah-MP model and the effects on simulated mean and extreme soil hydrothermal changes

Soil properties and their associated hydrophysical parameters represent a significant source of uncertainty in land surface models (LSMs), with consequent effects on simulated sub-surface thermal and moisture characteristics, surface energy exchanges, and turbulent fluxes. These effects can result in large model differences, particularly during extreme events. As is typical of many model-based approaches, spatial soil information such as the location, extent, and depth of soil textural classes is derived from coarse-scale soil information and employed largely due to its being readily availability rather than its suitability. However, the use of a particular spatial soil dataset can have important consequences for many of the processes simulated within an LSM. This study investigates model uncertainty in the Noah-MP model in simulating soil moisture (expressed as a ratio of water to soil volume, m3 m−3) and soil temperature changes, associated with two widely used global soil databases (STATSGO and SoilGrids). Both soil datasets produced significant dry biases in loam soils of 0.15 and 0.10 m3 m−3 during a wet and dry period, respectively. The spatial disparities between STATSGO and SoilGrids also influenced the simulated regional soil hydrothermal changes and extremes. SoilGrids was found to intensify drought characteristics – shifting low and moderate drought areas into the extreme and exceptional classifications – relative to STATSGO. Our results demonstrate that the coarse STATSGO performs as well as the fine-scale SoilGrids soil database, though the latter represents the soil moisture dynamics better. However, the results underscore the need for greater collaborative efforts to develop more detailed regionally derived soil texture characteristics and to improve pedotransfer function (PTF) parameterizations for better representations of soil properties in LSMs. Enhancing these soil property representations in LSMs is essential for improving operational modeling and forecasting of hydrological processes and extremes.