Advancing seismic processing of the Skaftá jökulhlaup in 2015, Iceland

Glacier lake outburst floods (GLOFs) are a recurring hazard in Alpine environments and can occur in orogenies with steep topography and short subglacial flood paths. In Iceland, the topography can be more flat and longer distances can be travelled below the ice. There, the locally named jökulhlaups can propagate tens of kilometers before emerging to the surface. It is difficult to monitor remote places, especially water movement below extensive icecaps. Permanent monitoring stations on ice are difficult to maintain. Off the ice, the networks are too sparse. By contrast, temporarily installed seismic arrays provide a tool to locate the flood front and issue early warnings of the subsequent flood of areas below the glacier. This is possible, as the flood-associated seismic signals migrate significantly during the event.

Seismic array analysis is in general a suitable method to locate weak, distant seismic events and observe floods. However, locations of long-lasting, emergent signals such as tremor are difficult to analyze. Due to the lack of clear onsets, travel time estimates are usually not possible. The location depends on other methods and the azimuthal station coverage. From 30 September to 3 October 2015, a flood drained from the eastern Skaftá cauldron in Iceland, reaching a peak discharge of 3000 m³/s in the Skaftá river. The seismic data were analysed using classical fk-analysis (Eibl et al. 2020). Here we advance the seismic processing including Pyrocko-based modules cake and parstack, and matched-field processing that allows to mute for example dominant noise sources. This helps to address the following questions: Can we detect signals of the flood along the uppermost part of the flood path? Can we detect the seismic signal during the subglacial flood propagation on an array which is dominated by river noise? Can we enhance the characteristics when correlating in the time domain? Our goal is to refine the location of the seismic tremor to enhance our understanding on the tremor generation and flood propagation, but also to test whether the analysis is stable enough to apply it to other floods of similar systems e.g. Grimsvötn.