Can assimilating remotely-sensed surface soil moisture data improve root-zone soil moisture predictions in the CABLE land surface model?

Abstract

The ability to quantify soil moisture content over depths including the root zone is important for predicting key hydrological processes for a range of applications in agriculture, emergency planning, and weather prediction. Remote-sensing provides a large amount of spatially distributed information related to water balance quantities. This includes brightness temperature data from passive microwave sensors such as the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E), which is used to estimate surface soil moisture content. Such observations can add valuable information to hydrologic/land surface modelling when combined using data assimilation techniques. However soil moisture from sensors such as AMSR-E can only be retrieved for the top couple of centimetres of soil at most, so assimilating this information into the surface layer of a model needs to constrain soil moisture in deeper layers. This will depend partly on the accuracy of model structure in relating surface moisture dynamics to deeper soil profiles, and also on assumptions made about model errors that can affect how the assimilation algorithm adjusts the model. Here we assimilated AMSR-E soil moisture observations into a CSIRO Atmosphere Biosphere Land Exchange model (CABLE) simulation over the Yanco region in the Murrumbidgee catchment, Australia. We then examined the impact it had on deeper soil moisture profiles (0-30cm and 0-60cm) in CABLE compared to in-situ validation data for two sites. CABLE was not calibrated, however the AMSR-E data were rescaled to match CABLE's soil moisture time series for the year 2005 (where annual rainfall was representative of long term climatology in the study region). The latest remotely-sensed Leaf Area Index (LAI) product from MODIS was also used as a key model parameter input. Initial results after perturbing this data and other inputs, to represent model error for the assimilation process, indicated the stability of assimilation applied to CABLE is sensitive to large perturbations. From subsequent experiments, the impact on predictions over 0-30cm and 0-60cm depth ranges from assimilating AMSR-E surface soil moisture was minimal compared to CABLE predictions with no assimilation, even though there were clear impacts on surface moisture predictions. This study provides direction for more focused research into understanding and representing model error and sensitivity with the help of remotely-sensed information.