Surprising seismological signals during the October 2015 Skaftá jökulhlaup

Understanding the spatiotemporal details in the occurrence of jökulhlaups, also referred to as glacier lake outburst floods (GLOFs), is important for improving early warning and forecasting future events. Jökulhlaups occur in many different glacier-related settings and differ in their characteristics depending on the natural conditions: From very rapid floods (minutes-hours) originating from moraine dammed lakes in steep valleys to gradual floods (days-weeks) from subglacial lakes such as the ones beneath Vatnajökull ice cap, in Iceland. Previous studies of the October 2015 Skaftá jökulhlaup suggested that several hours of early-warning is possible based on the generated seismic tremor. Here, for the first time, we looked into all three spatial components of GNSS and seismic array data, respectively. Previous studies have already analysed the seismic events (icequakes, tremor, other migrating transient events) in detail, yet only on the z component. We reprocessed all three components of the seismic array data using frequency-wavenumber -analysis (fk-analysis) and match field processing (MFP). Both techniques allow to locate distant signal sources, either by direction only (fk) or actual location (MFP). We specifically focused on the time period when the tremor source is moving with the flood front and found two unexplained seismic signals:

• A second migrating signal is visible on the lowermost part of the flood path 6 hours later than the passing of the first flood front.

  We compared this with a GNSS observations on top of the subglacial flood path and a hydrometric station 25 km downstream from the glacier margin in the affected Skaftá-river.

  After aligning the time series by the arrival of the pressure wave, the timing of the second seismic signal fits well with a 10 cm uplift of the glacier at the GNSS station; but also with a change in the rate of water level rise at the hydrometric station.

  We discuss this in the context of either explaining GNSS, hydrometric and seismological data individually or giving a hypothetical process that explains all three together. That could be a second intraglacial water lense draining, after the emptying of the lake deformed the overlaying glacier and connected the two water bodies. However, radio echo sounding survey over the source area in spring 2015 did not indicate a significant intraglacal water lense above the subglacial lake. The GNSS data may be cleared as noise artifact and the hydrometric data explained by flow of water out of the river course of Skaftá and onto porous lava fields between Sveinstindur, where the discharge of Skaftá is measured, and the glacier. Yet: The seismic signal then is left unexplained and open for discussion.

• Finally, 18 hours after the first pulse, we found a sudden deceleration in horizontal motion on the GNSS that coincided with a sudden increase in seismic signals originating at the glacier terminus. We discuss if what we see is actually the glacier stopping, after losing the flood lubrication.