Abstract:
A sensitivity analysis is conducted to investigate the contribution of forcing rainfall uncertainty with respect to model uncertainty in the simulation of soil moisture by integrating the NASA Catchment Land Surface Model (CLSM), forced with hydro-meteorological data. The research focuses on the Oklahoma region, which presents a good coverage by weather radars, satellite rainfall products and in-situ soil moisture measurements (Mesonet stations). The study employs high-resolution satellite rainfall fields derived from the NOAA-Climate Prediction Center morphing product and rain gauge-calibrated radar rainfall fields (considered as reference rainfall). Different sources of uncertainty are depicted: errors in the model input (i.e. incorrect rainfall estimates from sensor observations), and errors in the land surface model parameters. Specifically, forcing uncertainty is introduced using the satellite rainfall error model (named SREM2D) developed by Anagnostou and Hossain (2006), which generates an ensemble of satellite rain fields from high-accuracy 'reference' rain fields. The perturbed precipitation fields are propagated through CLSM to produce soil moisture ensembles. Errors in model physics and parameters are evaluated through two different approaches: (1) by perturbing model parameters using the generalized likelihood uncertainty estimation (GLUE) technique, and (2) by directly adding randomly generated noise to the model prognostic variables. The study compares standard normal deviates of Mesonet observations with standard normal deviates of soil moisture ensembles obtained by perturbing parameters and by perturbing prognostics. This comparison shows that the model parametric uncertainty captures the ground measurements in both cases. However, when comparing the two modeling uncertainty methods, the ensemble envelope is wider when the GLUE framework is adopted, that translates into a better encapsulation of the reference (Mesonet observations). Next, the combination of forcing uncertainty and modeling uncertainty (considering the two approaches) is studied and compared to the uncertainty introduced by only perturbing the forcing rainfall in terms of both surface and root zone soil moisture predictions. In order to quantify the uncertainty associated with each experiment, exceedance and uncertainty ratios are computed.