Mapping of Bottomfast Lake Ice in the Northwest Territories Via Data Mining of Synthetic Aperture Radar Image Time Series

Abstract

Changes in climate, warming temperatures and increased precipitation are impacting surface water resources in the Northwest Territories, Canada. Satellite remote sensing is an important tool to monitor variability in lake surface area, but monitoring depth is challenging. The distribution of bottomfast ice within a lake provides an indicator of depth and previous research shows that as lake ice develops and becomes bottomfast it exhibits a distinct signature when observed using multi-temporal Synthetic Aperture Radar (SAR) data. This research proposes an efficient computational technique for identifying bottom-fast ice across lakes in the Northwest Territories using multi-temporal SAR backscatter images and applies a function called dynamic time warping (DTW), which provides a shape-based similarity metric for time series data. We used backscatter profiles from surveyed lakes with known bottomfast ice to generate a DTW similarity metric on a pixel by pixel basis for a set of lakes. The similarity metric was used to categorize ice status as bottomfast or floating ice with 89.1% accuracy. DTW is an effective technique to map bottomfast ice using SAR time series and has potential to address limitations of other approaches where certain ice structures over deep lakes can produce backscatter responses similar to bottomfast ice.