Soil Hydraulic Properties Modeled from Meter to Kilometer Scales Based on in Situ and SMOS Soil Moisture Data

Monday, 15 December 2014
Peter J Shellito, University of Colorado at Boulder, Boulder, CO, United States, Eric E Small, Univ of Colorado Boulder, Boulder, CO, United States and Ethan D Gutmann, National Center for Atmospheric Research, Boulder, CO, United States
A primary objective of land surface models (LSMs) is to accurately represent soil moisture to calculate realistic latent and sensible heat fluxes for use in atmospheric models. Soil hydraulic properties (SHPs) control the movement of soil water and thus evapotranspiration. Parameter values are typically selected using look up tables based on soil texture. This approach limits parameter values to those from small-scale laboratory measurements, and thus precludes the possibility that effective parameter values should be different at the scale of the model.

We calibrate four SHP parameters in the Noah LSM using 0-5 cm soil moisture measured at different scales: (1) individual soil moisture probes; (2) the average of ~20 in situ soil moisture probes distributed across a watershed; and (3) passive microwave-based soil moisture estimates from SMOS. All data are from three different USDA watersheds in Oklahoma. The DREAM algorithm is used to define the posterior distribution of each parameter conditioned on the following objective function: RMSE between the uppermost model level soil moisture (0-10 cm) and observed soil moisture. By comparing texture-based parameters with calibrated distributions from different scales and measurement techniques, we address the following: (1) How different are the calibrated parameters compared to those based on soil texture alone? (2) How do the parameters affect simulated soil moisture and fluxes? (3) Are parameter sets physically consistent?

Parameters found via calibration differ from texture-based values and improve the simulation of soil moisture for intervals not included in the calibration. For example, calibration to in situ and SMOS data reduced model error by 65% and 59%, respectively, when simulated values were compared to in situ data. Optimized simulations influence surface fluxes as well, changing average daily latent heat fluxes by up to 67% and 50 W/m2. Most parameters values fall within the range of laboratory measurements.