Seismic Monitoring and Characterization of the 2012 Outburst Flood of the Ice-Dammed Lake A.P.Olsen (NE Greenland)

Since the Zackenberg Research Station (ZRS) in NE-Greenland was established in 1995, regular floods of the adjacent Zackenberg River have been observed. The floods result from the sudden discharge of a marginal, ice-dammed lake at the pre-dominantly cold-based A.P. Olsen Ice Cap about 35 km inland. The lake filling usually starts with the melting season in May/June and ends with the flood sometime after early July. The run-off water from the lake discharges through the subsurface of the adjacent Argo glacier. The actual migration paths and depth of the water within the glacier are unknown until it re-appears at the glacier terminus at a distance of 4 km to the ice-dam. In spring 2012 a surface seismic monitoring network was installed on Argo glacier in 2-3 m boreholes near the lake to acquire continuous data for the whole fill- and drain cycle from start of May to end of November. The network comprises 3 stations with three-component sensors and 2 stations designed as tripartite arrays with vertically oriented sensors. The maximum interstation distance is 1.2 km. Microseismic event detection and localization is facilitated by the homogenous seismic structure of the ice and the extremely high S/N ratio of the borehole installations. An initial detection based on an STA/LTA algorithm and event assocator results in order-of-magnitude 100,000 seismic events. These events are generally attributed to the opening of surface crevasses due to the presence of weak body waves and strong surface wave energy, interpreted to be Rayleigh waves with dominant frequencies around 1-4 Hz. Time-lapse cross-correlations of the ambient seismic noise field reconstruct the surface waves travelling between the stations. Weekly stacks of the cross-correlations are stable, and show a distinct change correlated with the outburst flood. Apparent surface wave velocities increase slightly several weeks prior to the outburst event, which itself is characterized by a decrease in the correlation amplitude. After the outburst event, the velocities decrease abruptly and take several weeks to rebound to the pre-outburst value. We explore the change of the apparent velocities, which can stem from both real in-situ medium changes and from the spatio-temporal variation of the ambient noise source distribution.