Earth Observation for Surface Melt Monitoring over Antarctic Ice Shelves: Opportunities and Challenges

Surface meltwater is becoming an increasing driver for ice shelf disintegration and consequent mass loss from the AIS. In this regard, monitoring surface melt over Antarctic ice shelves can enhance our understanding of their stability. Earth Observation (EO) satellites provide decadal records of land dynamics over Antarctica, and have been applied in surface melt monitoring. Thereby, they hold a potential to monitor the spatiotemporal evolution of surface melt over the entire AIS. Among the wealth of EO satellites, scatterometer and radiometer observations are most frequently used for surface melt detection, followed by SAR and optical data. Most studies used observations from a single satellite to study surface melt, while specific sensor characteristics (e.g., spatial resolution, overpass time, penetration depth) largely influence the potential for detecting surface melt. Therefore, we compare differences in melt detection between radiometer, scatterometer, SAR and optical sensors to assess the opportunities and challenges in observing surface melt for different EO satellites. We apply state-of-the-art melt detection algorithms to radiometer (Special Sensor Microwave Imager/Sounder, SMMIS), scatterometer (Advanced Scatterometer, ASCAT), Synthetic Aperture Radar (SAR; Sentinel-1), and optical (Moderate Resolution Imaging Spectroradiometer, MODIS) data over the Larsen B+C and Amery Ice Shelves for the 2015-2020 melt seasons. We construct melt timeseries and spatial maps using the melt detection algorithms. and intercompare the spatiotemporal patterns of detected melt. Finally, we compare areas with different melt patterns with auxiliary data sets (i.e., Regional Atmospheric Climate Model (RACMO2), Digital Elevation Models (DEM), high resolution optical imagery). Our results show that the largest differences in detected melt between the EO satellites can be linked to physical properties of the surface, sensor properties and atmospheric conditions. Over the blue ice areas, MODIS indicates more surface melt than the other sensors, as they miss blue ice areas due to either a coarse spatial resolution or the applied detection algorithms. However, over the ice shelves, MODIS detects significantly less surface melt, which can be attributed to the very high cloud obstruction frequency over AIS. Based on this intercomparison, we discuss the opportunities and challenges for melt detection across the AIS regarding the choice of different sensors and the chosen melt detection algorithms. We conclude that merging observations from different satellites (e.g. using machine learning) would further strengthen our knowledge on the presence of surface melt across the AIS, since this combines the strengths of specific sensors based on their sensor characteristics and the area of interest.