Statistical Relationship Between the Decrease of Major Seismicity and Drought in Southern California Since 1900

Pier Luigi Bragato

doi:10.3389/feart.2021.790412

Abstract

The seismicity in Southern California significantly decreased over the last decades. The decrease went in parallel with the reduction of meteoric groundwater recharge, which is a well-known factor capable of affecting seismicity. In this work the existence of a systematic statistical relationship was investigated by comparing the time density of Mw ≥ 5.7 earthquakes since 1900 with the time series of the Palmer Drought Severity Index (PDSI), an indicator of soil moisture roughly correlated with groundwater recharge. Given the non-stationarity of the two signals, the formal comparison was performed using both binomial logistic regression and cointegration testing. The analysis showed a significant statistical relationship, with peaks of seismicity 8 years behind those of PDSI. This finding suggests the hypothesis that groundwater recharge might affect earthquakes at a multi-year time scale. Proving this theory requires accurate measures and hydrogeological modeling, which is behind the scope of this work. Nonetheless, according to previous studies, the observed time lag might be explained by the slow propagation of pore pressure from the surface to the seismogenic volume. The ongoing trend towards an arid climate, made more evident by the recent severe droughts, might have contributed to the earthquake reduction of the last decades. The connection is particularly evident in the Salton Trough, with possible implications for the interpretation of its paleoseismicity.

Highlights

Seismic activity is caused for the largest part by the tectonic movements, with consequent accumulation of stress in the crust and its release through faulting
The solution here adopted was to test the correlation between Palmer Drought Severity Index (PDSI) and earthquakes by binomial logistic regression (Venables and Ripley, 2002) and counterchecking the result by cointegration testing (Engle and Granger, 1987), a technique dealing with nonstationary time series
There are fewer earthquakes in drought periods, an effect that is evident in the last decades

Summary

INTRODUCTION

Seismic activity is caused for the largest part by the tectonic movements, with consequent accumulation of stress in the crust and its release through faulting. The solution here adopted was to test the correlation between PDSI and earthquakes by binomial logistic regression (Venables and Ripley, 2002) and counterchecking the result by cointegration testing (Engle and Granger, 1987), a technique dealing with nonstationary time series To apply the latter method, the series of earthquake origin times was replaced by their smoothed time density sampled each year, introducing a dependence on the smoothing parameter discussed in the following. Cointegration was checked taking the time density of earthquakes as Y(ti) and PDSI as X(ti), using the KPSS test of stationarity (5% significance level) and ordinary least square regression for the estimation of the linear model in Equation 2 Tested in this way, the null hypothesis of cointegration between seismicity and PDSI was not rejected at a significance level of 5% (Table 1, data set number 1). The first check concerned the variability of the smoothed time series and its consequences for cointegration

12 Salton Trough

DISCUSSION AND CONCLUSION

Findings

DATA AVAILABILITY STATEMENT