Bridging the Soil Water Scale Gap in Land-Atmosphere Interactions

Treatment of subgrid-scale variability has been a perennial concern in calculating the fluxes of radiation, momentum, heat, water and trace gases such as CO2 between the lower atmosphere and the land surface in atmospheric models. Modelers frequently use grid-averaged quantities in the calculation of fluxes for grid squares with size on the order of 1 to 100 km or more, although vegetation, terrain and soil wetness can all vary on the scale of a few meters. The calculation of grid-scale flux becomes a real problem whenever the dependent variable, e.g. evapotranspiration (ET), is non-linearly dependent on a spatially-varying variable, such as soil moisture.

It has been shown that that the lower atmospheric forcing (downwelling radiation, temperature, humidity, wind speed) is relatively homogenous over a grid square at a given time and acts on a surface of varying vegetation density, slope and soil wetness. Simple area-averaging schemes can be used to aggregate vegetation density and slope with little impact on the accuracy of calculated fluxes of heat and moisture. Soil wetness, with its greater spatial heterogeneity, has been found to be more problematic.

A dynamic binning technique was previously developed that captured the spatial variability of soil moisture using a small number of wetness "bins" which are used to support a single grid-averaged calculation of evapotranspiration. This grid-averaged flux is then deconvolved to update the bin contents and thereby realistically model the resulting changes in the soil moisture distribution. This idea was previously demonstrated in a simple "toy" soil moisture-evaporation model (Sellers et al., 2007). The approach has recently been implemented in a land model (SiB3) that has a proven track record when evaluated against site-scale observations as well as when coupled to both GCM and mesoscale models.

We have incorporated the ‘bins' method for representing subgrid variability and grid-scale soil wetness constraint on ET into SiB3, and evaluate the results at multiple instrument sites across climatological and vegetation gradients. We find that we can reproduce, or, in many cases improve upon overall simulated surface flux (energy, moisture, CO2) when compared to control simulations, while improving fidelity to component processes. We're able to get the right answer for the right reason, where previous parameterizations sometimes require assumptions that may be unrealistic. Robustness in mechanistic realism provides confidence for simulations during meteorological extremes and/or change in climatic forcing.

Reference: Sellers, P.J., M.J. Fennessy, R.E. Dickinson, 2007: A numerical approach to calculating soil wetness and evapotranspiration over large grid areas. J. Geophys. Res., 112, D18106, doi:10.1029/2007/JD008781.