Determination of the Source Parameters of the June, 2020 Earthquake in Accra, Ghana and its Implication on Seismic Hazard Studies

Abstract

Southern Ghana has been plagued by small-to-moderate size earthquakes for decades, leading to the highest seismicity rate in West Africa. However, the underlying tectonic origins of these seismic activities have remained shrouded in uncertainty. On June 24, 2020, a significant seismic event occurred in Accra, Ghana, presenting a unique opportunity to investigate the earthquake dynamics of the region. Here, we leverage locally recorded data from the Ghana (GH) Digital Seismic Network and the GEOFON (GE) seismic network, situated within an epicentral distance of 200 km. Our research commences by estimating P and S wave arrival times, derived from continuous raw seismic data spanning from June 17th, 2020 to July 4th, 2020. We automatically identify the seismic phases by utilizing PhaseNet, an advanced machine learning-based method. Subsequently, we employ the Rapid Earthquake Association and Location (REAL) program to correlate these seismic phases with specific events. From the initial catalog produced, we curate a selection of events with high signal-to-noise ratios (SNR) to serve as templates, enabling us to apply the sliding-window cross-correlation (SCC) method for detecting smaller earthquakes that prove challenging to identify using conventional methodologies. We then employ Non-Linear Location (NLLoc) algorithm for precise event locations. Through the implementation of NLLoc, we successfully obtained highly accurate locations for 40 events based on PhaseNet detections and 25 more based on SCC detections. The distribution of these earthquake locations aligns with historical seismic events in Ghana, clustering along the northeast to southwest trending Akwapim fault zones, consistent with the fault plane solution of the mainshock. The epicenter of the mainshock was located at latitude 5.610°N and longitude 0.290°W. Focal mechanism and depth solutions for the mainshock were constructed using the cut-and-paste (CAP) method. The mainshock occurred at a depth of approximately 3 km with a magnitude (Mw) of 3.58, rupturing along a normal fault, in alignment with the Akwapim fault zones. The outcomes of this study reaffirm the intraplate nature of the underlying tectonic setting and contributes to a more comprehensive understanding of the characteristics of seismicity in this area, thereby leading to enhanced seismic risk management and preparedness efforts.

Keywords: machine learning, phase picking, non-linear earthquake location, template matching, focal mechanism inversion, cut-and-paste method