Challenges in Understanding and Predicting Greenland Lake Drainage Events

To accurately predict ice flow, an ice-sheet model must resolve the complex spatio-temporal variability of the ice-sheet hydrologic system. For Greenland, this requires understanding rapid lake drainage events, by which moulins deliver water from supraglacial lakes to the ice-sheet base. Critical metrics include the drainage event location and its timing during the melt season. Here, we use multiple remote sensing datasets to investigate whether local principal strain rates control the dates of rapid supraglacial lake drainage events. We identify 359 rapid lake drainage events through a semi-automated analysis of MODIS and Landsat imagery, which we apply to Pâkitsoq, western Greenland, over nine summers (2006-2010 and 2013-2016). We compare these drainage dates to principal strain rates derived from InSAR (MEaSUREs and other products) and Landsat (GoLIVE and other products) satellite data over the same years. The InSAR-derived strain rates have lower uncertainties ( 0.01 yr-1) but capture only a wintertime average; the Landsat-derived strain rates have larger uncertainties ( 0.1 yr-1) but feature higher temporal resolution (≥16 days) and span the entire year, including the melt season. We find that locations with more-tensile wintertime strain rates are associated with earlier draining of supraglacial lakes in the subsequent summer. This is consistent with observations of lake drainage "clusters" or "cascades", where the perturbation from an initial lake drainage event is thought to trigger other lake drainages in the area. Our relation is not statistically significant, however, and any causality is complicated by a stronger correlation with more traditional metrics such as surface elevation and cumulative melt days. We also find that the Landsat-derived summertime strain rates, despite their higher temporal resolution, do not resolve the transient extensional strain rates known from GPS observations to accompany and/or incite rapid lake drainages. Our results highlight the current challenges in observing, at the regional scale, the causes of rapid lake drainage events, which must be better understood in order to parameterize surface-to-bed hydrological connections in ice-sheet models.