A combined classification scheme to characterise river ice from SAR data.

Rivers and streams are key elements in the terrestrial re-distribution of water. This is particularly so in winter, when the overland flow is minimal. An ice cover has a significant impact on rivers, as it affects the discharge capacity, can modify the ecosystem and microclimate, cause flooding, restrict navigation and impact hydropower generation. In the recent years, BC Hydro has been looking at the potential of RADARSAT-1 data (C-band, HH polarization) to classify river ice types and to determine ice cover characteristics for safe and efficient dam operations. The preliminary results of this project have confirmed that SAR data could provide valuable information about the ice front location and the ice conditions upstream and downstream of this point. This paper shows that a combination of texture and backscattering images improved the discrimination of ice cover types during freeze-up. RADARSAT-1 and ASAR data showed comparable backscattering profiles for 500 m averaged river reaches and produced comparable ice maps. However, in comparison with RADARSAT-1 fine beam images, the relatively coarse resolution of the ASAR data further reduces the sensitivity of the classification by missing details and incorrectly classifying mixed pixels. The addition of the vertical polarization did not add significant skill to the classification.