Remotely Sensed Soil Moisture over Australia from AMSR-E

Soil moisture can significantly influence atmospheric evolution. However the soil moisture state predicted by land surface models, and subsequently used as the boundary condition in atmospheric models, is often unrealistic. New remote sensing technologies are able to observe surface soil moisture at the scales and coverage required by numerical weather prediction (NWP), and there is potential to improve modelled soil moisture, and ultimately atmospheric forecasts, through assimilation of this remotely sensed data into NWP models. Remotely sensed soil moisture is currently derived over Australia from passive microwave brightness temperatures from the Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E), on NASA's Aqua satellite. In collaboration with NASA, the Vrije Universiteit Amsterdam (VUA) are producing soil moisture separately from C-band (6.92 GHz) and X-band (10.65 GHz) AMSR-E data. Due to radio frequency interference (RFI) in the C-band microwave frequencies over north America, NASA also produce a soil moisture product from X-band data only, using a different algorithm. The three soil moisture products identified above are assessed in this paper by comparison to i) in-situ soil moisture timeseries from the Murrumbidgee Soil Moisture Monitoring Network (MSMMN), and ii) spatial patterns of antecedent precipitation. Specifically, the three products are assessed to determine their relative performance, and whether any are sufficiently accurate for use in NWP models. The benchmark against which they are assessed is the current scheme used to initialise soil moisture in the Australian NWP model: the Limited Area Prediction System (LAPS). Within the microwave spectrum, lower frequencies are theoretically better suited to sensing soil moisture, and a priori the soil moisture derived from C-band AMSR-E data was expected to be superior to that from X-band data. However, this analysis has revealed only a minimal difference between the VUA-NASA soil moisture derived from C-and X-band data. Both had realistic spatial behaviour, and a high correlation to the MSMMN soil moisture timeseries, and both outperformed the soil moisture currently used in LAPS. In contrast, the soil moisture derived at NASA from X-band AMSR-E data is persistently very low, and has a low correlation with the MSMMN data. It does not offer an improvement over the current soil moisture used in LAPS. This poor performance is believed to be due to the algorithm used in the retrieval, rather than the use of higher frequency brightness temperatures, since the VUA-NASA product based on X-band data performed comparatively well. This analysis concludes that the VUA-NASA soil moisture derived from AMSR-E C-band data is the most appropriate for use in assimilation experiments with the Australian NWP system, since it had the best performance, and lower microwave frequencies are theoretically favoured for sensing soil moisture. However, in regions where RFI prevents the use of C-band data, the VUA-NASA X-band product could be used, since it performed comparably in this assessment.