Digital surface model generation using optimal RADARSAT-2 images

Remote Predictive Mapping can be used to map geological potential to help with the logistical challenges of mineral exploration. Elevation information is necessary for this purpose, but current elevation data for Northern Canada needs to be improved. This paper presents a method for creating stereoradargrammetric digital surface models (DSM) using Radarsat-2 (R-2) images. The study site is a 23 000 km² area located east of Great Slave Lake in the Northwest Territories. First, a filtering method was developed to provide vertical accuracy for the DSM from ICESat elevation points. Then, from the 148 wide-ultrafine R-2 images, pairs were identified for stereorestitution. Sub-DSM were assembled after denoising and lake flattening. The filtering method led to a vertical accuracy increase of 1.05 m evaluated with current data. The generated DSM showed better resolution (15 m vs 20 m) and better vertical accuracy (5 m vs 6 m) than available elevation sources.