An Apparent Thermal Inertia Based Trapezoid Model for Downscaling ESA CCI Soil Moisture Products

Existing long-term Soil Moisture (SM) products are relatively coarse in spatial resolution, limiting their applications in heterogeneous scales. Various spectral information derived from optical satellite data, such as the Land Surface Temperature-Vegetation Parameter (LST-VP), have been widely employed to detect spatiotemporal variability of SM under different regional hydrological scales. In this study, inspired by the concept of LST-VI space, an ATI-VP (Apparent Thermal Inertia-Vegetation Parameter) was proposed and assessed for downscaling the ESA CCI SM product from 25 km to 1 km. Different vegetation indices (including NDVI, EVI, NIRv, and MSAVI) and biophysical variables (LAI and fPAR) derived from MODIS satellites were first assessed as inputs of the ATI-VP space to estimate AVDI (Apparent Thermal Inertia/Vegetation Drought Index). The AVDI was then applied to the weight decomposition model for SM downscaling. Overall, LAI for the ATI-VP space achieved the best AVDI performance. The accuracy of SM estimation was validated using in situ SM collected from the Murrumbidgee Soil Moisture Monitoring Network (MSMMN). The results showed that the accuracy of the downscaled 1km SM (R = 0.637, bias = 0.038 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> /m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> ) was close to that of the CCI SM (R = 0.661, bias = 0.030 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> /m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> ). However, the downscaled SM data exhibited enhanced spatial detail compared to CCI SM data. Further analysis based on the time series SM indicated that both the CCI SM and the downscaled SM are in good agreement in terms of temporal evolution. The downscaling method shows high potential for application in SM mapping across semi-arid regions.