Lagrangian Time Series of Arctic Sea Ice Melt from High Resolution Optical Satellite Imagery

The Arctic Ocean is shifting from an ice cover dominated by multi-year ice to one that melts and refreezes each season. As a result, processes governing ice albedo, particularly melt pond formation, are changing. There are still many unanswered questions about the nature of pond formation processes. These can only be addressed by time series observations of melt pond evolution, such as high resolution (submeter-scale) satellite imagery. Simply collecting such observations requires addressing the challenge of tracking and re-imaging individual drifting ice floes. We have developed a workflow that allows efficient tracking of moving sea ice targets with submeter scale optical imagery from Planet SkySat satellites. Here we present the results of tracking dozens of sea ice buoys as they drifted across the Arctic Ocean in the summers of 2020 and 2021 and the 2020 MOSAiC drift. This time series of optical imagery observes the evolution of sea ice at a scale only rarely captured before, permitting us to examine regional variations in melt,test hypotheses about pond behavior, and examine the generality of conclusions made at MOSAiC. Each image in the time series is processed with machine learning tools to calculate the areal fraction that belongs to one of 3 surface categories: 1) ice and snow, 2) melt ponds and submerged ice, and 3) leads and open ocean. The image processing results provide a new look at melt pond fraction evolution along Lagrangian drift tracks and reveal new insights into the differences between multi-year and seasonal ice covers.