4.1 Initial single bin inversion (Strategy 1)
First, we perform a single inversion of all the event spectra to solve for stress drops for all individual earthquakes in our dataset. We use the hybrid SNSS approach to calculate a single ECS to correct for the source spectra of all the earthquakes (no consideration of spatial varying attenuation), and use a constant shear wave velocity (assuming constant rupture velocity) to compute stress drops from corner frequencies and moments with Equation 4 (Figure 3).
To quantify any dependence of stress drop on seismic moment, we calculate the median stress drop in overlapping Mw bins of 1 Mw unit width centered at 0.5 unit intervals. We then use linear regression to find the best fitting trend and R-squared value (squared correlation coefficient) in two separate intervals with a relatively large number of events: (i) Mw0.75-1.75, in which stress drop increases with moment, and (ii) Mw1.75-2.75, in which there is negligible dependence of stress drop with moment, see Table 1 and Figures 3B and S6.
We use a similar approach to quantify any dependence of stress drop on depth, by calculating the median stress drop in 1 km depth intervals. The results of the single inversion with no correction for depth dependent velocity or attenuation indicate an increase in stress drop with depth, both between 1.5 and 6.5 km, and between 6.5 and 11.5 km (Table 1 and Figure 3A).