Performance Assessment of a New Stationarity-Based Parameter Estimation Method with a Simplified Land Surface Model Using In Situ and Remotely Sensed Surface States

Abstract

his study evaluates the performance of a new stationarity-based method for parameter estimation of a simple coupled water and energy balance model using in situ and remotely sensed surface soil moisture [from Advanced Microwave Scanning Radiometer for Earth Observing System (EOS; AMSR-E)] and surface temperature [from a combined Moderate Resolution Imaging Spectroradiometer (MODIS) and AMSR-E product]. Parameter estimation is carried out using both direct calibration to measured surface fluxes (latent, sensible, and ground heat) and a recently published method based on enforcing stationarity of model-predicted surface state tendency terms. The latter stationarity-based method was developed for parameter estimation without knowledge of observed fluxes?that is, it requires only forcing terms (e.g., radiation, wind speed, air temperature) and surface states (moisture and temperature). In addition, the stationarity-based method can easily handle gaps in atmospheric forcing and surface state data, as it does not integrate over time to simulate fluxes. The evaluation is conducted at three AmeriFlux sites. Changing the data sources of surface states (in situ measured and remotely sensed data) leads to little degradation in estimating turbulent fluxes despite the relatively poor quality of satellite data at some sites. In all cases, direct calibration yields marginally better predictions than the stationarity-based method, with site-averaged root-mean-square errors for daily total energy fluxes approximately 5?6 W m?2 lower. However, direct calibration requires observed fluxes in the objective function, which imposes a strong limit on its application.