Effect of permanent stress on the time-dependent earthquake probability: implication for seismic hazard assessment in the African continent

Abstract

SUMMARYEarthquake hazard assessment is the first step towards implementing prevention, preparedness and response or faster recovery actions to reduce the risk of seismic disasters. In this paper, we present a comprehensive study on present-day seismicity in terms of the estimated waiting time and conditional probability in Africa by 2022–2072 using four competing distribution models (Brownian passage time (BPT), gamma, lognormal and Weibull). We also investigate how much Coulomb stress change $\ ( {\Delta {\rm CFF}} )$ induced by previous earthquakes that occurred in neighbouring active sources should revise the probability of occurrence at the location of the next events. We analyse large earthquakes with moment magnitude, ${M}_{\rm w} \ge 6.0$, collating data from the Global Centroid Moment Tensor and from several published literature that list fault plane solutions of large African earthquakes since 1900. We assume that the data set is stationary and consists of independent events. First, the model's parameters are estimated and the results of the statistical analysis of the interevent times show clear evidence of quasi-periodic recurrence behaviour for large earthquakes ${M}_{\rm w} \ge $ 6.0 in different seismotectonic regions in Africa. Next, a comparison among the distribution models is performed with the aim of selecting the most suitable one. The results in terms of the maximum likelihood criterion and its extension (Akaike Information Criterion) indicate that, in general, the BPT and Weibull models had similar fits to each other, and performed slightly better than gamma and lognormal models. Then, we use Gaussian random distributions to treat parameter uncertainties (e.g. aperiodicity, maximum expected magnitude, slip rate and mean recurrence time) of the distribution models associated with each seismotectonic region. From repeated Monte Carlo draws, we assess uncertainties of the 50-yr conditional probability values for the next earthquake obtained from two distribution models (BPT and BPT + $\Delta {\rm CFF}$) related to the 50th percentile. The results of the BPT distribution indicate very high chances of future earthquakes in the study region where the conditional probability of a large earthquake reaches 99.5, 95.6, 83.1 and 82.2 per cent for the western branch of the East African Rift System (EARS), northwest Africa, the Afar region, and the eastern branch of EARS, respectively. Taking into account the effect of stress change from interacting sources (BPT + $\Delta {\rm CFF}$), these probabilities are slightly modified to 99.8, 98.4, 89.9 and 87.3 per cent for the western branch of EARS, northwest Africa, the Afar region and the eastern branch of EARS, respectively. These marginal increases suggest that the estimated effect of the earthquake interaction introduced by the coseismic slip of previous earthquakes on neighbouring active sources is minor if compared with the uncertainties affecting the renewal models used for the basic time-dependent conditional probability assessment.