Event Detection for Cryoseismology

Abstract:

Passive seismology is increasingly being used to understand active glaciers, motivated by the potential for seismology to inform the community on surface and hidden processes like crevassing and melt water drainage. An obstacle to this type of analysis is that seismic event detection is currently reliant on visual identification of events or time expensive trial and error methods. Automated seismic event detection techniques cannot be easily practiced because diverse seismic sources (e.g. brittle cracking, basal slip, calving) can result in signals of very different durations: in the range of a few seconds to a full day. The signals also vary greatly both in their absolute amplitudes and their signal-to-noise levels.

We present a novel approach for the automated detection of events in cryoseismology data sets that share these diverse attributes. We introduce a new event detection algorithm called multi-STA/LTA which we implement in the widely available ObsPy python package. This event detection algorithm runs multiple short-term and long-term average combinations at once to generate an event catalogue that captures the various durations and amplitudes of glacier events. Following an overview of the methodology, we optimize the free parameters in the algorithm using a Monte Carlo simulation of 100 pseudo-random physically-based waveforms. We find that the optimal values for short-term and long-term average are tightly constrained. In contrast, the number and range of short-term and long-term average combinations are strongly correlated with each other; values can be chosen from a wide spectrum along a line of best fit. We conclude by applying the event detection algorithm to a sample data set from Whillans Ice Stream in West Antarctica, demonstrating its potential for real data sets. The code will be made available for the benefit of the cryoseismology, and applied seismology, research communities.