Spring melt pond fraction (fp) has been shown to influence September sea ice extent and, with a growing need to improve melt pond physics in climate and forecast models, observations at large spatial scales are needed. We present a novel technique for estimating fp on sea ice at high spatial resolution from the Sentinel-1 satellite during the winter period leading up to spring melt. A strong correlation (r = -0.85) is found between winter radar backscatter and fp from first-year and multiyear sea ice data collected in the Canadian Arctic Archipelago (CAA) in 2015. Observations made in the CAA in 2016 are used to validate a fp retrieval algorithm, and a fp prediction for the CAA in 2017 is made. The method is effective using the horizontal transmit and receive polarization channel only and shows promise for providing seasonal, pan-Arctic, fp maps for improved understanding of melt pond distributions and forecast model skill.