On the potential of offshore sensors and array processing for improving seismic event detection and locations in the North Sea

Abstract

SUMMARY The capability to detect and locate seismicity in the North Sea is necessary to ensure the safety of future CO2 storage operations. Currently, the event location capabilities are hampered by a one-sided network configuration, as only sensors deployed on the Norwegian mainland are providing good signal-to-noise ratios. However, applying array processing techniques to groups of offshore sensors deployed at oil and gas reservoirs can improve the signal-to-noise ratio at such sites, which generally suffer from high noise contamination originating from platforms or seismic shooting. In this study we test the feasibility of array processing to sensors deployed at the Grane oil field to improve earthquake event locations. Out of the 3400 nodes installed, we have access to two data subsets: (i) continuous data from 10 sensors spread out over the whole field and (ii) short segments of data from 30 sensors selected with the purpose to apply array methods. As the average interstation distance between the 10 sensors is 6 km, the recorded wavefield is not coherent and traditional array processing methods are not applicable. To optimize the usage of the 10-sensor offshore data for event location purposes, we develop a new method for incoherent array processing. We compute a characteristic function based on the kurtosis of the seismic traces prior to a frequency–wavenumber analysis. This method successfully determines slowness and backazimuth for almost all the evaluated earthquakes in this study. Using common coherent array processing, the 30 sensors show superior detection capabilities over the 10 sensors and will therefore decrease the detection threshold for seismic events in the region significantly if continuous data are available. We conclude that offshore sensors at Grane can be included as a part of a system for passive seismic monitoring in the North Sea. We recommend using the 30-sensors data for this purpose. However, as we only have access to continuous data from the 10 sensors in this study, we found that they are a good substitute when the 30-sensor data are not available.