Just an idea: Sobiech et al compared TerraSAR-X (X-Band), Envisat-ASAR (C-Band) and ALOS (L-Band) for soil mositure estimation in the Lena Delta. You can find the abstract here: http://meetingorganizer.copernicus.org/EGU2011/EGU2011-607-1.pdf
For this reason you may consider to use Sentinel-1 (C-Band), ALOS-2 or Radarsat-2 data. These sensors offer high resolution less than 30 meter (depends on the acquisition mode). Most promising for the estimation of soil mositure with SAR - as far as I know - is the application of "long" wavelengths, like L-Band.
Two satellite systems are specifically dedicated to the measurement of soil moisture:
- ESA's Soil Moisture and Ocean Salinity (SMOS), launched on 2 November 2009 with an expected lifetime of 3 years but still operational today: http://www.esa.int/Our_Activities/Observing_the_Earth/SMOS
The spatial resolution of this passive sensor is 35 km at the centre of the field of view.
- NASA's Soil Moisture Active Passive (SMAP), launched 31 January 2015 with an expected lifetime of 3 years:
http://smap.jpl.nasa.gov/
Three products are expected from this instrument, at spatial resolutions of 3, 9 and 36 km respectively.
Both of these instruments offer products at rather coarser spatial resolution than what you are asking for, but they make a best effort to focus on soil moisture. The next--crucial--question to ask is "What accuracy do you need on soil moisture estimates?"
The SAR instruments mentioned by Tobias target other applications and may perhaps be used successfully to retrieve soil moisture under some conditions, but you'll need to verify whether those apply to your area of interest. In any case, the main issue is that active microwave signals are sensitive to various surface properties, most notably the dielectric constant (a characteristic that is dependent on water) and the roughness of the surface materials (which does not depend on the presence or quantity of water). So, to retrieve soil water from microwave measurements you need fairly sophisticated models capable of taking into account these various processes. It's a non trivial task.
Your best bet is probably to combine models with observations, using other sources of information as proxies to spatially distribute the satellite measurements obtained at coarser resolutions.