Many applications such as irrigation management and flood predictions require soil moisture (SM) estimates at fine-resolutions. Soil moisture estimates of current satellite missions are normally at coarse spatial resolutions, such as 25–36 km for Soil Moisture and Ocean Salinity (SMOS) SM data. Many spatial downscaling approaches have been developed to get finer resolution SM. Apparent thermal inertia (ATI), as a reliable approximation of thermal inertia (TI) that could be derived from daily fine-resolution optical/thermal remote sensing (RS) data, has been testified to be highly related to SM. However, the empirical relationship between ATI and SM varies with vegetation cover density. The main purpose of this study is to develop an ATI-based SM downscaling method that is free of vegetation cover effects. The study was conducted in Naqu areas in the central Tibetan Plateau (TP) in southwest China, where spatially intensive in-situ SM measurements were routinely obtained. European Space Agency’s Climate Change Initiative (ESA CCI) soil moisture combined product (version 4.2) was used for this study. The fine-resolution (1 km) ATI data were derived from the Collection 6 MODerate resolution Imaging Spectroradiometer (MODIS) products. A modified quantile-quantile (Q-Q) adjustment method was applied to correct the bias of the CCI SM data before downscaling. A non-linear regression model between bias-corrected CCI SM and fine resolution ATI was built to disaggregate the bias-corrected SM to spatial resolution of 1 km. In this model, Enhanced Vegetation Index (EVI) was introduced to globally and quantitatively adjust the vegetation effect in the relationship between SM and ATI. The bias-corrected CCI SM and the downscaled 1 km resolution SM were evaluated using the in-situ SM measurements obtained from the Multi-scale Soil Moisture and Temperature Monitoring Network on the central TP (CTP-SMTMN) in the study area. The results indicate that the proposed modified Q-Q adjustment method effectively reduced the bias of the original CCI SM data (bias = 0 m3/m3). The downscaled 1 km resolution SM shows fine scale spatial variability of soil moisture and preserves the accuracy (R = 0.552) and temporal variability of the bias-corrected CCI SM. The accuracies of the downscaled SM over areas with different vegetation cover show quite good consistency. The uncertainties of the proposed SM spatial downscaling method were comprehensively analyzed. The proposed ATI-based SM downscaling algorithm is applicable to generating representative and consistent high spatial resolution SM over areas with different vegetation cover density.