Downscaling of CCI soil moisture product based on apparent thermal inertia: Considering the effect of vegetation condition

This paper presents a downscaling method for coarse spatial resolution satellite soil moisture products based on apparent thermal inertia (ATI) with different vegetation cover. The study area locates in Naqu, an area with ~300 km × 300 km in the central Tibetan Plateau (TP), where soil moisture and land surface temperature were routinely collected at all 57 stations. ESA CCI soil moisture products (version 4.2) were used for downscaling. The apparent thermal inertia (ATI) was derived from the MODIS LST products was used as high spatial resolution proxy for SM downscaling. Because the CCI SM productrs were underestimated in the study area, a modified Q-Q adjustment method was applied to correct the bias of the CCI SM. Because the statistical relationship between SM and ATI changes with vegetation cover, the MODIS EVI product was used to present the vegetation cover over the study area. A logarithm regression model was built to present the relationship between the bias-corrected CCI SM and spatially aggregated ATI and EVI at 25km resolution, and was applied to 1km spatial resolution to get the downscaled 1km SM products. The downscaled 1km resolution SM data were evaluated using the in situ SM observations. The overall accuracy of downscaled SM is high, with R = 0.553 bias =0.009 m³/m³. The accuracy of downscaled SM at 1km spatial resolution are coherent in different vegetation cover fractions. The main uncertainties of the downscaled SM is due to the effect of short-term precipitation on the temporal variation of SM, which may not be presented by the ATI.