SMOS soil moisture downscaling based on back propagation neural network with MODIS LST and EVI

Soil Moisture and Ocean Salinity (SMOS) is the first L-band passive microwave mission dedicated to soil moisture (SM) monitoring but has coarse resolution. SMOS SM was downscaled by Back propagation neural network (BPNN) and MODIS LST and EVI, and was evaluated by intensive 56 stations in-situ obtained from the mesoscale Tibetan Plateau Soil Moisture/Temperature Monitoring Network (SMTMN) during 2010-2012 over Naqu. It showed that downscaled rmse and bias were close for BPNN and regression, downscaled r was significantly improved by BPNN. In terms of r, it was improved greatly by downscaling comparing with original SMOS evaluations. Besides, descending of SMOS (D_SMOS) got better downscaled SM than ascending of SMOS and evaluation of downscaled D_SMOS (r=0.849, rmse=0.056m
³
/m
³
and bias=0.016 m
³
/m
³
) was nice in SMTMN scale. The fine SM BPNN downscaled by SMOS can be used for regional or local studies.