An automated procedure to map breaking river ice with C-band HH SAR data

Integrating intensity and context for improved supervised river ice classification from dual-pol Sentinel-1 SAR data 

“…Even though the developed supervised classifier results in high classification accuracy, different ice stages (e.g., open water and sheet ice covered by wet snow) with similar feature characteristics remain a challenge, albeit less of a challenge after inclusion of texture features. Temporal patterns could further improve classification results, as was proposed by Van der Sanden et al (2021) as well. Figure 6 shows the typical backscatter behavior over one breakup season, subdivided into i) thermal breakup and ii) dynamic breakup patterns.…”

“…Rubble ice induces a much higher backscatter signal than sheet ice because its upper surface is much rougher. Van der Sanden et al (2021) provide further details on the breakup process and its implication for the backscatter.…”

“…First, most existing approaches use the backscatter intensity comprised in single-polarized C-band SAR data-this holds true for the classification of river ice under freeze-up, winter, and breakup conditions (e.g. Jasek 2003;Puestow et al, 2004;Gauthier et al, 2010;Floyd et al, 2014;Van der Sanden et al, 2021). The most commonly applied polarization is Horizontal-Horizontal (HH).…”

“…by training the algorithm. Without ground data and analyst intervention, unsupervised algorithms will not produce reliable classification results when one or more classes of interest are absent-unlike, well developed, supervised algorithms (Sobiech and Dierking, 2013;Van der Sanden et al, 2021).…”

The Ecology of River Ice

“…Ice formation, ice type, and breakup are major physical forces that can influence the ecology of rivers during the winter. Most studies that monitor river ice patterns focus on large rivers that can either be resolved from remote sensing (Beaton et al, 2019;Chaouch et al, 2014;Chu & Lindenschmidt, 2016;Cooley & Pavelsky, 2016;Huntington et al, 2003;Pavelsky & Smith, 2004;van der Sanden et al, 2021), from in situ data in accessible locations (Arp et al, 2020;Janowicz, 2010;Lacroix et al, 2005;Magnuson et al, 2000;Rokaya et al, 2018;Shiklomanov & Lammers, 2014;Smith, 2000), or estimated using physically based models (Lindenschmidt, 2017;Park et al, 2016;Shen et al, 1995). Ice cover can vary greatly across biomes, varying from frequent solid ice formation in Arctic streams and rivers to open water conditions in steep alpine streams, for example.…”