Watershed Scale Observations of Snowmelt from Sentinel-1 SAR in Southern British Columbia, Canada

The annual melt of seasonal snow in Western North America provides critical sources of streamflow and soil moisture during warm summer months. Climate change induced shifts in snowmelt timing can affect wildfires, droughts, and floods, in addition to ecosystem health and phenology. As direct measurements are sparse, an increase in spatially distributed observations of the snowpack may help to better understand these shifts going forward. Synthetic Aperture Radar (SAR) is sensitive to the liquid water content of snow and has been successfully used to map wet snow in alpine regions. In this study we use a fusion of optical and radar imagery to estimate the time of snowmelt onset, snowpack saturation, and snowmelt duration for watersheds in southern British Columbia, Canada. We build time series of C-band SAR imagery in Google Earth Engine (GEE) for four ablation seasons from 2018 to 2021. We use SAR minima to define the inferred period at which the snowpack is saturated and begins to generate melt. The end of snowmelt, defined as the date a pixel becomes snow free, is determined from Landsat-8 and Sentinel-2 observations using an NDSI and NDFSI based classification system. Differencing the SAR runoff onset dates and the optical snow free dates provides an estimate of the duration of snowmelt. While this approach has limitations related to SAR snow wetness detection under dense forest cover and on steep slopes, data fusion estimates of snowmelt onset and duration agree with snow water equivalence records from automated hydrometeorological stations in the region. Ongoing research examines the impact of watershed characteristics and topography on SAR snowmelt detection, as well as the transferability of this method between climate regions. This approach can be used to better constrain hydrologic model parameters, adjust hydrologic model states in water supply or flood forecasting applications and in water resource applications.