An approach for mapping near-surface soil moisture at the watershed scale from RADARSAT-1 synthetic aperture radar (SAR) data was developed and tested on seven RADARSAT-1 SAR images acquired over the northern portion of the Ch¿teauguay River Basin in southwestern Quebec, Canada, dominated by agricultural and herbaceous fields. A soil surface roughness map was first retrieved from a SAR image by inverting an empirical backscatter model with known (or assumed) soil moisture. The resulting map was then used with the backscatter model to recover near-surface soil moisture for the remaining SAR images. Field campaigns were conducted concurrent to SAR image acquisitions to measure soil moisture and surface roughness in 24 fields. Good agreement was observed between watershed-scale soil moisture values and measurements averaged for all sampled fields, with a correlation coefficient of 0.96 and an RMS error of 2.2%. However, considerable scatter was found between observed and SAR-derived soil moisture estimates at the field scale. Although the generated maps reveal reasonable small-scale soil moisture variability, no definitive conclusions could be drawn as to whether or not the proposed approach can quantify soil moisture at the field or within field scale due to insufficient ground measurements. Furrows and herbaceous and crop vegetation, which are known to affect the radar signal, appeared to have little influence on the ability to retrieve soil moisture at the watershed scale for the images analyzed in this study.