Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #304546

Title: Evaluation of SMOS soil moisture products over the CanEx-SM10 area

item DJAMAÏ, N. - Universite De Sherbrooke
item MAGAGI, R. - Universite De Sherbrooke
item GOITA, K. - Universite De Sherbrooke
item HOSSEINI, M. - Universite De Sherbrooke
item Cosh, Michael
item BERG, A. - University Of Guelph
item TOTH, BRENDA - Environment Canada

Submitted to: Journal of Hydrology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/8/2014
Publication Date: 1/1/2015
Publication URL:
Citation: Djamaï, N., Magagi, R., Goita, K., Hosseini, M., Cosh, M.H., Berg, A., Toth, B. 2015. Evaluation of SMOS soil moisture products over the CanEx-SM10 area. Journal of Hydrology. 520:254-267.

Interpretive Summary: The Soil Moisture Ocean Salinity mission was launched in 2009 to provide global coverage of surface soil moisture as well as salinity estimates over the oceans. For the soil moisture products, ground validation requires in situ networks of soil moisture instrumentation to provide sufficient spatial and temporal coverage. In association with in situ networks, supporting field campaigns can provide a valuable opportunity to verify the in situ resources and give some estimate of the errors present across the network. The Canadian Field Experiment in 2010 (CANEX) deployed field samplers and airborne equipment to monitor soil moisture at large scales over networks near Kenaston, Saskatchewan and the northern boreal forest region of Saskatchewan. Good correlations were found between the in situ resources and the satellite estimates, but there were significant biases, varying in magnitude, depending on the version of the satellite data used for comparison. Other comparisons of alternative satellite products and the in situ networks were performed and similar results were found. The results of this study indicate that the SMOS products have been improving with each iteration of the product and it performs better than products from previous satellite missions. There is significant difficulty in estimating soil moisture in the boreal region of Saskatchewan, likely because of the dense vegetation cover. This research is useful for climate change scientists and weather agencies who wish to include satellite estimates of soil moisture into their models.

Technical Abstract: The Soil Moisture and Ocean Salinity (SMOS) Earth observation satellite was launched in November 2009 to provide global soil moisture and ocean salinity measurements based on L-Band passive microwave measurements. Since its launch, different versions of SMOS soil moisture products processors have been developed. The purpose of this study is to evaluate the processor versions 309, 400, and 501 by comparing them to a) soil moisture measurements from the Canadian Experiment for Soil Moisture in 2010 (CanEx-SM10) and from networks of permanent and temporary stations, and b) other existing satellite-based soil moisture products (AMSR-E/NSIDC, AMSR-E/VUA soil moisture, and ASCAT relative soil moisture). Rainfall data were used during the analysis in order to understand the episodic variability of soil moisture. The analysis included both agricultural site (Canadian Prairies) and forested site (Boreal Ecosystem Research and Monitoring Sites; BERMS), and considered separately the SMOS ascending and descending modes. An improvement in SMOS soil moisture estimation was observed from the processor versions 309 to 501. In general, there is little difference between the processor versions 400 and 501, and they are found to be more correlated to ground measurements than the processor version 309. For the agricultural site, all three SMOS processor versions underestimated the soil moisture, but to varying degrees depending on the overpasses mode. For the ascending overpass, the three processor versions have a high bias of approximately -0.12 m3/m3, with respect to the measured ground data. For the descending overpass, however, a good improvement in the algorithms was observed for the three processor versions. Thus the maximum bias for the measured ground data went from -0.12 m3/m3 for processor version 309 to -0.06 m3/m3 for processor version 501, and the soil moisture error seem to be less dependent on the absolute soil moisture for the two last versions. Highest correlation coefficients with ground measurements were obtained with SMOS processor version 501 (R=0.58), ASCAT (R=0.55), and AMSR-E/NSIDC (R=0.54) products for ascending overpasses. For descending overpasses AMSR-E/NSIDC (R=0.82) is better correlated to ground measurements followed by SMOS (R=0.64) and ASCAT (R=0.55). However, AMSR-E/VUA appears weakly correlated with ground truth for both overpasses. Despite the good correlation found with ground data, the temporal evolution of AMSR-E/NSIDC data became stable with the vegetation growth and presented a weak sensitivity to rainfall. Over the forested site, due to the more dense and complex vegetation cover, SMOS data were less correlated with the in situ data (R =0.40, for processor version 501) than for the agricultural Kenaston site. Soil moisture values from the ascending overpass were closer to the ground measurements (-0.03=bias=0 m3/m3) than the estimates from the descending overpasses (0.06=bias=0.09 m3/m3). The latter were generally overestimated, especially before the active vegetation period where the bias was greater than 0.12 m3/m3 obtained with prototype 501. ASCAT presented correlation coefficients to ground data comparable to those obtained by SMOS (version 501), whereas lower correlation coefficients were obtained with AMSR-E-NSIDC and mainly with AMSR-E/VUA data.