We explored how citizen scientists can help map changes in Arctic landscapes. Using a web tool we created, more than 100 volunteers contributed the approximate center points of particular ground patterns called ice-wedge polygons in aerial images from Alaska and Canada. Our work shows that the data created by volunteers can be used to reconstruct ice-wedge polygon networks and provide valuable insights on the state of frozen ground in the Arctic.

Antarctic sea ice has changed rapidly in recent years. Here we developed a deep learning model trained by multiple climate variables for extended seasonal Antarctic sea ice prediction. Our model shows high predictive skills up to 6 months in advance, particularly in predicting extreme events. It also shows skillful predictions at the sea ice edge and year-to-year sea ice changes. Variable importance analyses suggest what variables are more important for prediction at different lead times.