Significance of VP/VS ratio in locating earthquakes of a long-duration swarm in the western coast of India

The earthquake swarm in Palghar, western India, is a natural laboratory to understand long-duration seismogenesis. However, it requires an accurate estimation of the key parameters, such as earthquake locations and the velocity model of the region. Two separate studies performed by two organisations in India, the CSIR-National Geophysical Research Institute (NGRI) and the National Center for Seismology (NCS) reported significantly differing results from the first 8–12 months of earthquake monitoring in Palghar. CSIR-NGRI reported the depth of earthquakes in the range of 4–15 km, whereas NCS reported a much shallower depth down to 7 km only. We investigate the reason for this depth discrepancy by analysing datasets from the networks of both these organisations. We find that no available velocity models are able to reconcile the depth difference between the two datasets. Selection criteria generally adopted to determine a good velocity model, such as travel-time residuals and location errors, also failed to identify a suitable velocity model. Several synthetic tests helped us to identify that the network geometry and source-station separation are contributing factors to the observed depth difference. However, the main reason for the discrepancy was found to be inconsistent velocity models used for location at both networks. The main objective of our work is to estimate a velocity model which minimises the depth discrepancy between both datasets. In this regard, we synthetically generated hundreds of models with different V_P and V_P/V_S ratios, followed by a grid search of V_P and V_P/V_S ratios, which harmonises the location from the two datasets. The best model consists of a V_P lying between 5.55–5.85 km/s and a V_P/V_S ratio between 1.77–1.81. For this velocity model, there is a greater overlap of the hypocentral clouds from both networks, and about 80% of the earthquakes of the CSIR-NGRI network were located within depths ≤ 7 km. We find that location errors vary seasonally, which is also related to the V_P/V_S of the crust. Since earlier studies have also found a strong correspondence between rainfall and seismicity in this region, we infer that the high V_P/V_S ratio might have been caused by the rainfall saturation of cracks in the Palghar seismicity zone.