Abstract
In this thesis I investigate the recurrence of large earthquakes in low seismicity regions. A number of previous studies have found evidence that suggests large earthquakes recur less regularly on low activity-rate faults than their more active plate-boundary counterparts. This represents a departure from the quasiperiodic earthquake recurrence predicted by elastic rebound theory, and poses challenges for how the seismic hazard posed by faults in low seismicity regions should be characterised.
Here I firstly compile a global dataset of 80 long-term earthquake records from different tectonic regions. I characterise the variability of earthquake inter-event times in this dataset using the burstiness and memory coefficient measures. I find that the majority of Earth's well-studied faults shows weakly periodic and uncorrelated large earthquake recurrence, consistent with the expectations of elastic rebound theory. However, many low activity-rate faults show random or clustered earthquake recurrence, which cannot be easily explained by elastic rebound theory. This study is presented in Chapter 2 of this thesis.
I then use the Otago range and basin province of southern New Zealand as a case study of a low activity-rate fault system. Previous studies from this region have proposed that large earthquakes recur episodically, with individual faults going through active and quiescent phases, making the region an important case study of earthquake recurrence in low seismicity regions. I expand our knowledge of the distribution of earthquake inter-event times and fault slip rates in the region through the collection of new data.
I undertake the first paleoearthquake investigations of the Hyde Fault, a major fault in east Otago, finding evidence for four paleoearthquakes since ~65 ka. Optically stimulated luminescence dating constrains the timing of the earliest event (Event D) to around 47.2 ka (37.5 – 56.7 ka at 95% confidence), Event C to around 34.6 ka (24.7 – 46.4 ka at 95% confidence), Event B to around 23.5 ka (19.7 – 27.3 ka at 95% confidence) and the most recent event (Event A) to around 10.5 ka (7.9 – 13.1 ka at 95% confidence). The earthquake record for the Hyde Fault does not show evidence for episodic earthquake recurrence. This finding contrasts with previous paleoearthquake studies of two other faults within the fault system, the Dunstan and Akatore faults, that have suggested earthquakes recur episodically on these faults. The results from the Hyde Fault demonstrate that episodic earthquake recurrence is not ubiquitous in Otago, and suggest that diverse recurrence behaviours may coexist within the same fault system. This study is presented in Chapter 3 of this thesis.
I then obtain the first constraints on late Quaternary incremental fault slip rates in the Otago region using 10Be cosmogenic radionuclide dating of faulted alluvial fan surfaces on the Hyde and Dunstan faults. I determine a slip rate of 0.27 mm/yr (0.22 – 0.34 mm/yr at 95% confidence) for the Hyde Fault since 115 ka and a slip rate of 0.16 mm/yr (0.12 – 0.20 mm/yr 95% confidence) for the Dunstan Fault since 320 ka. Both faults show deviations from a constant linear slip rate through time, with slip rates varying by a factor of 2 – 4 over timescales of 15 – 80 kyr. Increases in slip rate are out of phase on the two faults, supporting a hypothesis that strain is shared within the fault system over timescales on the order of 10s – 100s kyr, although neither fault shows evidence for long periods of seismic quiescence. This study is presented in Chapter 4 of this thesis.
The diversity of recurrence behaviour evident in the paleoearthquake and slip rate records from the Hyde and Dunstan faults poses challenges for how the hazard posed by these faults should be characterised. To address this problem, I present a novel Bayesian method for developing time-dependent models of earthquake recurrence that does not rely on a priori assumptions of quasiperiodic earthquake recurrence. Using the additive property of the Brownian passage time distribution, I make inference on the model parameters jointly from paleoearthquake and cumulative fault offset data. Monte Carlo Markov Chain methods are used to sample the posterior distribution of model parameters, which are subsequently used to forecast earthquake probabilities. These forecasts fully account for uncertainties in the data and model parameters. The method is demonstrated for the Hyde and Dunstan faults using the data collected in Chapters 3 and 4, along with previously reported data. This study is presented in Chapter 5 of this thesis.
Together, the findings of this thesis show that large earthquake recurrence in low seismicity regions is a complex process, and does not conform to the patterns of quasiperiodic recurrence observed in many plate-boundary regions and supported by elastic rebound theory. While some low activity-rate faults show evidence suggesting episodic earthquake recurrence (e.g. the Dunstan Fault), others do not (e.g. the Hyde Fault). Because the physical processes controlling rates of strain accumulation and release in low seismicity regions are poorly understood, statistical models of earthquake recurrence should avoid strong a priori assumptions about the distribution of earthquake inter-event times. The Bayesian method for modelling earthquake recurrence that I present here uses a purely data-driven approach to account for aperiodicity. I argue that this data-driven approach is the most reasonable approach to take given our present understanding (or lack thereof) of the complex processes controlling earthquake recurrence in low seismicity regions.