Complexity in microseismic phase identification: full waveform modelling, traveltime computations and implications for event locations within the Groningen gas field
|Reference:||Geophyshical Journal International Year: 2019 Number:217 Page: 620-649|
|ISBN / DOI:||doi: 10.1093/gji/ggz017|
Determining accurate microseismic event locations at the Groningen gas field in the Netherlands has important implications for understanding the ongoing induced seismicity and its associated seismic hazard. To improve the depth constraint of the microseismicity, downholemonitoring arrays have been deployed in the central region of the Groningen field. The observed seismicity at these receivers is characterized by significant complexity in the waveforms, due to the high velocity contrasts that exist and the acquisition geometry. Reliably identifying and picking phases for use in earthquake location algorithms is therefore particularly challenging.
Using a well constrained and highly detailed 3-D velocity model, we show how full waveform modelling can be used to understand the causes of the observed phase complexity. By identifying the different modelled phase arrivals in detail, we look to identify any systematic changes in waveform complexity with source location, to aid phase identification of the recorded downhole data.
Theoretical traveltimes are often the foundation of earthquake location algorithms. We highlight the associated problems of their computation for the downhole monitoring setup for the Groningen model by complementing the full waveform simulations with traveltime computations and their associated ray paths from both an eikonal solver and using thewavefront construction method. We highlight large inconsistencies in the traveltimes, demonstrate the limitations and sensitivity of ray tracing in a layered velocity model, and show how the theoretical traveltimes do not equate with the dominant phase arrivals observed within the modelled waveforms. Finally, we propose an approach based on computing P- and S-wave traveltimes directly from the full waveform modelling, such that phase picking is based on an amplitude threshold rather than individual phase identification, which can also be adjusted for any given moment tensor.