Soil moisture (SM) is a pivotal component of the Earth system, affecting interactions between the land and the atmosphere. Numerous applications, such as water resource management, drought monitoring, rainfall-runoff modelling and landslide forecasting, would benefit from spatially and temporally detailed information on soil moisture. The ESA CCI provides long-term records of SM, globally, and with daily temporal resolution. However, its coarse spatial resolution (0.25°) limits its use in many of the above-mentioned applications. The aim of this work is to downscale the ESA CCI SM product to 0.05° using machine learning and a set of static and dynamic variables affecting the spatial organization of SM at this scale. In particular, we employ land cover information from the Copernicus Global Land Service (CGLS) together with land surface temperature and reference evapotranspiration from the EUMETSAT Prototype Drought & Vegetation Data Cube (D&V DC). The latter facilitates the access to numerous satellite-derived environmental variables and provides them on a regular grid. Preliminary results against in-situ measurements across Europe obtained from the International Soil Moisture Network (ISMN) show that the downscaled SM preserves the high temporal accuracy of the ESA CCI SM while simultaneously increasing the spatial level of detail. Furthermore, spatial correlations against large in-situ networks (> 20 stations) suggest that the downscaled SM provides a better description of the spatial distribution of SM compared to the original ESA CCI product. We will also highlight the strengths of the proposed approach compared to other downscaled SM products and discuss some limitations and possible improvements.