Semi-automated classification of river ice types on the Peace River using RADARSAT-1 synthetic aperture radar (SAR) imagery

Abstract

The winter regime of the Peace River in northern British Columbia and Alberta is a determining factor for the operation of the British Columbia Hydro and Power Authority Williston Reservoir. Therefore, fine beam RADARSAT-1 synthetic aperture radar (SAR) satellite images were acquired in winter and spring 2000 and 2001 and analyzed for ice cover types. Video footage of the ice conditions on the Peace River was obtained from aerial ice observations that were conducted simultaneously with the image acquisitions. The analysis of the images was done (i) visually and (ii) using an unsupervised Fuzzy K-means classification. To streamline the process in an operational environment, the unsupervised ice classification was semi-automated. The unsupervised classification broke the data into seven classes, which represent the major ice cover types observed on the Peace River. The spatial distribution of ice cover types as generated by the ice classification generally coincides well with air-truth information and backscatter signatures from SAR images. The location of the boundaries between the ice types appears to be accurate, but it may not be precise. The ice maps can be used for ice monitoring, decision-making purposes, or the verification of ice models.Key words: river ice, remote sensing, RADARSAT-1, Peace River, semi-automated classification.