Abstract:
A sensitivity analysis was carried out to investigate the uncertainty in the simulation of soil moisture by integrating a land surface model, forced with hydro-meteorological data. The study seeks to address the characterization of two sources of uncertainty: 1) errors in the rainfall estimation from sensor observations; and, 2) the land surface model parametric error, which manifests as non-uniqueness in soil hydraulic parameters. The study is conducted in the Oklahoma region, which presents a good coverage by weather radars, multi-year satellite rainfall products and in-situ meteorological and soil moisture measurement stations. The land surface model that has been chosen is the NASA Catchment Land Surface Model (CLSM; Koster et al., 2000). The framework to characterize the parametric error is represented by the generalized likelihood uncertainty estimation (GLUE) technique. The forcing rainfall uncertainty is analyzed through an error model included in the NASA Land Data Assimilation System (LDAS), which is applied to satellite rainfall fields to obtain an ensemble of equiprobable realizations of precipitation. The perturbed precipitation fields are propagated through CLSM to produce multiple ensembles of soil moisture. This numerical experiment allows us to quantify the propagation of uncertainty from rainfall to soil moisture prediction, accounting for the two main error sources, i.e. rainfall forcing and land surface model parameterizations. In previous studies we have shown how satellite-rainfall error alone can impact soil moisture uncertainty. Results from this study will complement these initial findings to quantify the relative impact of rainfall vs modeling error and the combined uncertainty on the prediction of soil moisture.