Comparison of SMOS Soil Moisture data with ground truth measurements over India
Calla, OPN; Gadri, Kishan Lal; Kalla, Abhishek; Sharma, Rahul; Agrahari, Sunil Kumar; Rathore, Gaurav
International Center for Radio Science, INDIA

Soil Moisture (SM) plays a vital role in land-atmosphere interactions. It is a key parameter in various agricultural, hydrological & meteorological models like crop growth simulation models, weather forecasting models, climate simulations, soil evaporations, rainfall runoffs and soil erosions. Despite its multi-disciplinary importance, a reliable and accurate estimation of surface soil moisture is not feasible using conventional point observations. However, satellite remote sensing can fill this gap. On Nov, 2009, European Space Agency (ESA) launched Soil Moisture and Ocean Salinity (SMOS) mission with an aim to provide global soil moisture and ocean salinity maps using multi-angular L-Band microwave brightness temperature data . The mission carries a novel 2-D interferometric L-Band radiometer MIRAS with multi-angular viewing capabilities. In near future, NASA's Soil Moisture Active Passive (SMAP) mission will provide surface soil moisture maps using both L-Band (1.4GHz) radiometer and radar. Thus, these satellites make surface soil moisture monitoring more feasible at global scale. However, to replace conventional point observation methodology with satellite observations, validation and calibration of satellite data is indispensable. The present paper states the results of comparison of SMOS L2 soil moisture (m3/m3) with ground truth data over four different test sites, each of 40x40 km2 area, in India. These test sites are having different climatic conditions, soil types and topographies. The test sites are located in different states of India, viz. Rajasthan, Gujarat, Kerala, and Tamil Nadu. Ground truth soil moisture data were collected during SMOS passes over these test sites. For ground truth data collection over each 40x40km2 test site area, 5 to 6 SMOS Discrete Global Grids were selected within the test site. Then, five focus farms of 1kmx1km area were selected within each DGG at which ground truth soil moisture data were collected using both gravimetric methods and soil moisture sensors. There were 9 gravimetric points(gv.pts) equally distributed within the focus farm at which soil samples were collected in air tied zip-locks bags for estimating in-situ soil moisture data using gravimetric methods. In addition, random soil moisture measurements were done within each DGG, using soil moisture sensors. Ground truth soil moisture data were collected during Mid-Oct, 2011 to Dec, 2011 over Rajasthan test site, during July-Aug, 2012 over Tamil Nadu and West Bengal test sites and during Oct-Nov-2012, over Kerala and Gujarat test sites. It must be noted that both the data estimated and measured using gravimetric method and soil moisture sensors are in volumetric percentage. Thus, ground truth data collected in one shot during the corresponding SMOS pass, are averaged over each SMOS soil moisture DGG (~15km) so as to compared with its single soil moisture value. Results are shown in figs. A1 to A4. Figs. A.1 to A.4 show the correlation between SMOS volumetric soil moisture (%) and ground truth volumetric soil moisture (%). Fig. A.1. represents all the data sets of SMOS soil moisture and its corresponding ground truth soil moisture with their correlation factor of 0.735. Fig. A.2 shows the same data with additional information of difference between SMOS volumetric soil moisture (%) and ground truth soil moisture (%). It must be noted from fig. A.2 that except for two outliers, having the magnitude of 'difference values' of 10.98% & 17.35%, all other data sets are having 'difference values' near to or below 6% (fig. A.4). However, when these data sets are plotted again in fig. A.3 by removing the two outliers, the correlation factor increased to 0.904. It must be clearly noted from fig. A.2 that for maximum data sets, the difference between SMOS and Ground Truth Volumetric Soil Moisture (%) is below 4%. Out of two outliers, for initial case (i.e. 10.98% observed over Gujarat test site), SMOS has underestimated the surface soil moisture and for the other case (17.35%), it has over estimated the surface soil moisture. The general cause for underestimation of surface soil moisture by SMOS is RFI at 1.4 GHz and for over estimation is ponding effects (over saturation of soil and creation of small ponds) during occurrence of rainfalls. Thus the present paper compares the SMOS soil moisture product with the ground truth measurements and validates its accuracy.