The shape and extent of the surface water polygons were derived from a machine learning algorithm used to identify surface water. The algorithm was based on raster inputs from the Sentinel-1 and -2 satellites. The machine learning algorithm was able to identify surface water with an accuracy 99.6%. When each pixel was classified into water or land the water pixels were converted to polygons. These polygons were then cleaned and attributed with: names, permanent/recurring, depth, and volume. For more detail on the methodology please refer to the Boreal Surface water inventory – technical documentation (available for download below).