Synergy between Soil Moisture from Passive (SMOS) and Active (RADARSAT-2) Microwave over Berambadi, India
Tomer, Sat Kumar1; Al Bitar, Ahmad1; Muddu, Sekhar2; Merlin, Olivier1; Bandyopadhyay, Soumya3; Maity, Saroj3; Corgne, Samuel4; Kerr, Yann1

The current study presents the comparison and analysis towards merging the soil moisture retrieved from passive (SMOS) and active (RADARSAT-2) satellites over the Berambadi watershed, South India. SMOS (Soil Moisture and Ocean Salinity) satellite from ESA has a passive microwave L-Band sensor providing acquisition at i40 km resolution and less than 3 days temporal resolution. RADARSAT-2 is an active microwave sensor from (CSA) operating in C-Band at a decametric spatial resolution and 24 days temporal resolution. Both the satellites are all-weather satellites. The retrieval of soil moisture from microwave satellite is impacted by soil roughness and vegetation etc. The impact of soil roughness and vegetation is less on SMOS compared to RADARSAT-2 as the former operates in passive mode at L-Band.

Twenty four images of RADARSAT-2 and SMOS-L2UDP soil moisture product, along with extensive field data collected in the field campaigns during 2010-2012 in the framework of the ongoing AMBHAS (Assimilation of Multi-satellite data at Berambadi watershed for Hydrology And land Surface experiment) project were used in the study. A non parametric algorithm based on the CDF transformation method was developed to retrieve the soil moisture from RADARSAT-2 backscatter coefficient at a spatial resolution of 100 m. This calibration and validation was performed using 12 images for both. The developed algorithm provided a good estimate of the surface soil moisture with a RMSE of 0.05 m3 m−3.

Then the validated RADARSAT-2 soil moisture maps were upscaled to compare with the SMOS data. Eight upscaling strategies were considered, taking into account the surface heterogeneity in terms of texture (clay sand), surface cover (forest, land cover) and SMOS mean antenna pattern. The strategies were based on the linear combination of the different parameters. Significant differences were observed between the eight strategies. The RMSE and coefficient of determination of the different strategies varied between 0.06-0.09 m3 m−3 and 0.3-0.9 respectively. The best comparisons with a RMSE of 0.06 m3 m−3 and a coefficient of determination of 0.7 were obtained for upscaling strategies that include land cover effect.

This result was used in the development of a downscaling procedure to merge the spatial information from RADARSAT-2 with the temporal dynamics of SMOS acquisitions. In order to implement this method the persistence of the spatial patterns in the RADARSAT-2 soil moisture map were evaluated by inspecting the spatiotemporal correlation coefficient across the two years, which was approximately 0.55. The impact of rainfall and farming activities were also taken into consideration in the analysis of the spatial heterogeneity.

This study shows the potential synergy between the use of active/passive microwave soil moisture retrievals for spatial and temporal down-scaling of soil moisture. This study also exhibits the potential of SMAP (Soil Moisture Active Passive) mission with active L-band from NASA due to launch in 2015.