ARTS: a scalable data set for Arctic Retrogressive Thaw Slumps

Climate warming significantly affects high latitudes, leading to changes in permafrost dynamics. Retrogressive thaw slumps (RTS), the thawing and subsequent slumping of permafrost in susceptible terrain, are a notable example of extreme events that alter Arctic landscapes and accelerate carbon feedbacks. The extent and distribution of RTS are not well documented because of the remote location of RTS and the time-consuming nature of delineating them by hand via satellite imagery. Many studies have developed their own digitized RTS datasets to test deep learning methods for RTS identification, yet there is no comprehensive dataset of delineated RTS to allow researchers to use all existing datasets, limiting the scale and scope of further deep learning studies. This project addresses this gap by compiling 24 manually digitized RTS datasets into a single format; the Arctic Retrogressive Thaw Slump dataset (ARTS). We verified each dataset, compiled metadata, and created a system for indexing RTS and noted relationships between any overlapping RTS features. The ARTS dataset is designed to be easily contributable, with protocols in place to add more digitized RTS datasets, and flexible, with the ability to parse the dataset for relevant metadata based on study needs. We are using this compiled dataset to assess the spatial distribution of RTS, identify locations where RTS are absent, and pinpoint areas in the Arctic that lack RTS representation.