Crevasse field response to glacier dynamic change in Greenland

Crevassing provides visual information regarding gradients in ice flow velocity (i.e. strain rates), with relevance for multiple processes occurring at Greenland marine terminating margins, in the ice sheet interior, and for its peripheral glaciers and ice caps. Mapping crevasse distribution and relating them to ice dynamics can provide critical understanding for iceberg calving, ice damage enhanced glacier flow and glacier detachment in valley glacier and ice cap settings. Substantial effort has previously been exerted in structural glaciology to both map crevasses and relate their sizes and distribution to the dynamics of individual glaciers, though this has frequently involved time consuming manual mapping making large temporal/spatial scale investigations impractical. Building on recent work towards the automation of crevasse mapping, we present a new, highly flexible, methodologically simple automated approach for crevasse identification from top-of-atmosphere (TOA) Sentinel-2 optical satellite imagery. This has been developed to be computationally light, and unlike other thresholding based methods does not require standardisation of reflectance values through surface reflectance correction of imagery. The approach is implemented within the Google Earth Engine platform, meaning that the method has the potential to be applied rapidly and at scale for near-real time monitoring.

Our new approach allows rapid characterisation of the response of crevasse fields (and therefore glacier stress/strain environments) to glacier dynamic change. In this presentation we evaluate the efficacy of this approach against previously developed automated methods of crevasse mapping (namely Gabor filtering and digital elevation model based approaches). We conduct initial exploration into: whether analysis of crevasse fields allow identification of precursor signs of marine terminating glacier destabilisation; how crevasse fields evolve in response to observed terminus retreat; and if key summary statistics that result from the analysis can be related glacier calving styles. Through comparisons results generated by different methods we also highlight the strengths and weaknesses of each in characterising these signals.