Inter-comparison of microwave satellite soil moisture retrievals over the Murrumbidgee Basin, southeast Australia

• Published in: Remote sensing of environment Vol. 134; pp. 1 - 11
• Main Authors: Su, Chun-Hsu, Ryu, Dongryeol, Young, Rodger I., Western, Andrew W., Wagner, Wolfgang
• Format: Journal Article
• Language: English
• Published: New York, NY Elsevier Inc 01-07-2013; Elsevier
• Subjects: AMSR-E; Animal, plant and microbial ecology; Applied geophysics; ASCAT; Biological and medical sciences; Earth sciences; Earth, ocean, space; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; Hydrology; Hydrology. Hydrogeology; Internal geophysics; Remote sensing; SMOS; Soil moisture; Teledetection and vegetation maps; Validation; Australia; Validation; SMOS; AMSR-E; Soil moisture; Remote sensing; ASCAT; microwaves; satellites; soil moisture; hydrology; Meteorological phenomenon; atmospheric precipitation; Earth Observing System; Australasia; Bias; Space remote sensing; Satellite observation; measurement sensor; roots; Scatterometer; Variance; correlation; radiometry; Radiometer
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2013.02.016

The use of satellite-based soil moisture retrievals for hydrologic, meteorological and climatological applications is advancing significantly due to increasing capability and temporal coverage of current and future missions. Characterisation of the relative skill of soil moisture products from different satellite sensors on a common spatial grid is crucial to achieve synergetic applications. This paper therefore evaluates three soil moisture products from AMSR-E (Advanced Microwave Scanning Radiometer — Earth Observing System), ASCAT (Advanced Scatterometer) and SMOS (Soil Moisture and Ocean Salinity) in absolute soil moisture units and on a common grid, against in-situ observations from southeast Australia. Before renormalisation, the three products yield correlations of 0.63–0.71 and a similar root-mean-square difference (RMSD) in the order of 0.1m3m−3, although showing different levels of error contributions from bias, variance and correlations. The results are compared with land and precipitation data to investigate the sensitivity of their errors to land surface features. Three renormalisation strategies – minimum–maximum matching, mean/standard-deviation (μ–σ) matching and cumulative distribution function (CDF) matching – are considered for correcting systematic differences between ground and satellite data. The renormalised satellite data is found to retain RMSDs of 0.04–0.06m3m−3 on average. The CDF method produces only marginal further improvements to correlations (0.67–0.75) and RMSDs compared to the μ–σ approach. The renormalisations by μ–σ and CDF methods also bring three products into better agreements with each other, but lead to strong correlations between RMSD and the dynamic range of in-situ soil moisture.
•Skills of AMSR-E and SMOS show sensitivity to land surface features.•Upscaling of ASCAT to a 0.25° grid improves its agreements with in-situ data.•The products show similar RMSDs on average but with different error profiles.•CDF matching is only marginally better than mean/standard-deviation matching.•RMSDs correlate with variance of in-situ data strongly after renormalisation.