Biophysical characterization of summer Arctic sea-ice habitats using a Remotely Operated Vehicle-mounted Underwater Hyperspectral Imager

The impact of a rapidly shifting sea-ice cover on climate, ecosystem processes and biophysical habitat properties is not yet fully understood, particularly in the central Arctic Ocean, due to a lack of spatially representative observations. From June to July 2020 during the year-long Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC, leg 4) in the Transpolar Drift we deployed an underwater hyperspectral imager (UHI) mounted on a remotely operated vehicle (ROV) to characterize the biophysical properties of different sea-ice habitats. We conducted UHI surveys along two transects: i) under level first-year sea ice (FYI), which had a mean sea-ice draft of 1.4 m and was composed of primarily level FYI but also had a relatively shallow ridge (keel depth ∼2.6 m); and ii) under the flank of a ridge, named Jaridge, with a mean ice draft of 1.7 m, which was composed of a mix of level ice and thicker ridge blocks with over 3 m draft. We present a new unsupervised bio-optical quantification algorithm for hyperspectral surveys, the relative ice algal biomass index (RBI), using spectral mixture analysis (SMA). We compare this method to the supervised machine learning habitat classification algorithm, Support Vector Machine (SVM). The RBI showed good agreement to literature-based normalized difference indices (NDI) and PCA analyses, which confirm the RBI as a reliable unsupervised index for ice algal biomass. Our biophysical characterization of the two surveyed regions showed an association of sea-ice algal biomass with sea-ice ridge features. Our surveys also indicate that ice algal spatial distribution may be influenced by ice melt rates, and the formation of under ice meltwater layers and false bottoms. With high spatial coverage (> 100 m) at microscale resolution (∼ cm) we documented large spatial variability of summer Arctic sea-ice algae biomass and different patterns between adjacent but different coverage of ice habitats. We further demonstrate the need for improved understanding of sea-ice algal spatial variability as a complementary tool for sea-ice biogeochemical sampling using destructive ice core sampling.