The 1 km Surface Soil Moisture Dataset: Evaluation over Entire Africa in 2004-2012
Hegyiová, A.¹; Doubková, M.²; Sabel, D.¹; Wagner, W.¹; Bauer-Marschallinger, B.^1
1Vienna University of Technology, AUSTRIA; ²ESA, ITALY

A soil moisture product at 1 km spatial resolution has been derived from the ENVISAT ASAR GM data series (2004-2012) over the entire African continent within the scope of the ESA's TIGER NET and SHARE projects. The processing chain applied the change detection algorithm to over 18 000 ASAR GM backscatter scenes. The resulting 1 km Surface Soil Moisture (SSM) product extents over an area of more than 30 million km2, has a temporal resolution of 4 to 7 days and enables evaluation of the soil moisture algorithm over a variety of land cover types and climatic regions. Of special interest is the evaluation over arid environments, where prior studies demonstrated negative correlation between active and passive coarse resolution soil moisture retrievals.

First, the product was evaluated using coarse resolution soil moisture datasets available from the passive (AMSR-E) and active (METOP-ASCAT) microwave sensors as well as from the modelled soil moisture from the GLDAS-NOAH and ERA-Interim models. Both absolute (RMSE) and relative (R) evaluation measures were computed over the entire continent using the 1km SSM product spatially averaged to 5 km resolution and always the nearest pixel of the coarse resolution products. Second, the ability of the dataset to demonstrate spatio-temporal soil moisture patterns was demonstrated.

The results demonstrate a good agreement of the 1km SSM product to the satellite derived (AMSR-E, METOP-ASCAT) as well as to the modelled (GLDAS-NOAH, ERA-interim) coarse resolution soil moisture datasets over areas with an evident soil moisture variation. The average Pearson correlation coefficient between the ASAR and GLDAS-NOAH modelled data reached 0.53 over the region south of 15°N. In the desert region (north of 15°N) significant positive correlation values (R=0.38) were computed between the ASAR GM and ASCAT soil moisture dataset. Contrariwise, significant negative correlation values (down to -0.75) were computed over the same region between the ASAR GM and AMSR-E as well as between the ASAR GM and two modelled datasets. The physical reasons for this unusual backscatter behaviour over the desert areas has not yet been determined and needs to be further investigated.

Assimilation of the coarse resolution (~25 km) remotely sensed soil moisture products have demonstrated to improve the predicted parameters such as run off [1] or to study the land-atmospheric feedbacks [2]. Higher spatial resolution of the remotely sensed data would also allow a higher spatial resolution of the models, which is important especially for the regional use.

References:

[1] Crow W. T., and D. Ryu (2009). A new data assimilation approach for improving runoff predictions using remotely sensed soil moisture retrievals. Hydrol. Earth Syst. Sci., 13, 1-16, doi:10.5194/hess-13-1-2009

[2] Taylor, C.M., De Jeu, R.A.M., Guichard, F., Harris, P.P., Dorigo W.A. (2012), Afternoon rain more likely over drier soils, Nature, 489,282–286, doi:10.1038/nature11377.