Major Earthquakes Occur Regularly on an Isolated Plate Boundary Fault

See allHide authors and affiliations

Science  29 Jun 2012:
Vol. 336, Issue 6089, pp. 1690-1693
DOI: 10.1126/science.1218959

The Sedimentary Life of Earthquakes

Estimating the hazards associated with possible large earthquakes depends largely on evidence of prior seismic activity. The relatively new global seismic networks installed to monitor earthquakes, however, have only captured the very recent history of fault zones that can remain active for thousands of years. To understand the recurrence of large earthquakes along the Alpine Fault in New Zealand, Berryman et al. (p. 1690) looked to the sediments near an old creek for evidence of surface ruptures and vertical offset. Along this fault segment, 24 large earthquakes seem to have occurred over the last 6000 years, resulting in a recurrence interval of ∼329 years. The activity is more regular than other similar strike-slip faults, such as the San Andreas Fault in California.


The scarcity of long geological records of major earthquakes, on different types of faults, makes testing hypotheses of regular versus random or clustered earthquake recurrence behavior difficult. We provide a fault-proximal major earthquake record spanning 8000 years on the strike-slip Alpine Fault in New Zealand. Cyclic stratigraphy at Hokuri Creek suggests that the fault ruptured to the surface 24 times, and event ages yield a 0.33 coefficient of variation in recurrence interval. We associate this near-regular earthquake recurrence with a geometrically simple strike-slip fault, with high slip rate, accommodating a high proportion of plate boundary motion that works in isolation from other faults. We propose that it is valid to apply time-dependent earthquake recurrence models for seismic hazard estimation to similar faults worldwide.

A long-standing acceptance of Reid’s (1) elastic rebound theory of earthquakes combined with the knowledge that tectonic plates move steadily over geological time scales has led to an appealing—but rarely demonstrated—idea that major earthquakes on plate boundary faults occur relatively regularly (25). In contrast, several studies have suggested that faults rupture randomly or produce temporal clusters of earthquakes in response to various complexities, including fault interactions (69). The increasing popularity of models of random or clustered earthquake recurrence may reflect the paucity of earthquake histories from geometrically simple, rapidly slipping, isolated plate boundary faults. Paleoseismology provides evidence for the timing, size, and location of past major earthquakes on faults over longer time periods than the historical record, improving understanding of fault behavior and enabling estimates of future earthquake occurrence to be made (5, 7, 9, 10).

We present a long earthquake record determined using paleoseismological techniques from the Alpine Fault in southwest New Zealand to assess the relationship between fault characteristics and patterns of earthquake recurrence. The Alpine Fault is ~850 km long and slips horizontally at average rates ranging from 14 mm/year in the north (11) to 31 mm/year in the south (12), with a subordinate amount of vertical slip (Fig. 1, A and B), making it one of the longest, straightest, and fastest-moving plate boundary transform faults on Earth. In the southwestern South Island, the Alpine Fault accommodates two-thirds of the relative motion between the Pacific and Australian tectonic plates (Fig. 1, A and B) (13, 14). The remainder is accommodated by distributed deformation across the width of the plate boundary in the South Island (1315). No major earthquake has occurred on the Alpine Fault since written records began (~170 years ago), but various lines of evidence indicate that the fault ruptures in large [moment magnitude (Mw) > 7] to possibly great (Mw > 8) earthquakes (16) and poses a substantial seismic hazard. Previous paleoseismic work has provided age constraints for the past four surface-rupturing earthquakes (1721), but this yields only three interseismic intervals with which to assess the Alpine Fault’s recurrence behavior. We extend the known record to a total of 24 major earthquakes over the past 8000 years at Hokuri Creek on the southern onshore section of the fault (Fig. 1C).

Fig. 1

Tectonic setting of the Alpine Fault and Hokuri Creek study site. (A) Australian-Pacific plate boundary through the South Island of New Zealand with plate motion rate after Sutherland et al. (13). (B) Detail of the Alpine Fault showing slip-rate estimates along the fault and other active faults in red (1113, 32). MFZ, Marlborough Fault Zone. Bathymetry is from the New Zealand Oceanographic Institute’s Chart Miscellaneous Series No. 74. (C) Aerial photograph of Hokuri Creek study area. Letters A to J show locations of the studied outcrops (fig. S1). Present outflow direction of both north and south branches of Hokuri Creek is to the southwest. The former outflow was via a gorge across the fault (edges marked by white dashed lines). Box “1” shows the area of a topographic survey where channels are offset (fig. S2).

The formation and exposure of a long, fault-proximal, sedimentary earthquake record at Hokuri Creek is the result of specific geomorphological conditions. Hokuri Creek used to flow across the Alpine Fault through what is now an abandoned gorge (Fig. 1C). Holocene sediments accumulated against the fault over a 20-ha area adjacent to the creek to a total section thickness of 18 m (fig. S1). Deposition ceased when the creek changed course and flowed along the fault instead of across it at ~1000 C.E., before the penultimate Alpine Fault earthquake. Rapid incision around the junction of the north and south branches of Hokuri Creek has since exposed the sediments (Fig. 1C) (22). The sedimentary section comprises decimeter-thick beds of alternating shallow-water peat and silt units. Geomorphological investigations reveal that this cyclic stratigraphy is bounded to the northwest by the main scarp of the Alpine Fault.

Geologic and paleoenvironmental investigations indicate that the mechanism for formation of the cyclic peat-silt stratigraphy depends on surface rupture of the Alpine Fault. At the Hokuri Creek locality, the uphill-facing scarp produced by the vertical component of slip on the Alpine Fault provides a natural dam to Hokuri Creek drainage. For example, in the most recent surface-rupturing earthquake, channels in the abandoned gorge were displaced by ~1 m vertically and 7.5 ± 0.5 m dextrally (fig. S2). Raising this dam periodically caused ponding against the fault scarp, with consequent deposition of catchment-derived silt. When drainage was reestablished across the scarp, autochthonous peat deposition resumed (Fig. 2). Raising the northwestern side of the fault in every surface-rupturing earthquake not only provides the mechanism for alternating between peat and silt deposition but also explains the intermittent creation of accommodation space required to sustain shallow-water sedimentation on the southeastern side of the fault for much of the Holocene. We interpret the appearance of silt in the stratigraphic section to mark the first sedimentation after an earthquake and every package of silt plus peat to represent the length of an earthquake cycle.

Fig. 2

Depositional model for fault-controlled sediment accumulation at Hokuri Creek. (A) Hokuri Creek flows across the fault scarp, and a wetland exists on the southeast side of the fault. In situ, organic sediment deposition occurs in wetlands and pond margins, forming peat. (B) Surface rupture of the Alpine Fault occurs (involving 7.5 ± 0.5 m dextral, and 1 m vertical offset). This displacement blocks the drainage of Hokuri Creek, causing a change in hydrological regime. Deposition on the southeast side of fault is dominated by transported, catchment-derived silt in shallow ponds. (C) Silt deposition continues, and the creek starts to reestablish drainage across the fault scarp. (D) Sedimentation on the southeast side of the fault returns to in situ peat formation.

Aggradation of catchment-derived sediment on floodplains, and in the form of beach ridges, at the time of past Alpine Fault earthquakes has been documented at various sites north of Hokuri Creek (18, 20, 23). These studies illustrate that Alpine Fault earthquakes are the dominant control on large-scale sedimentation events in catchments that cross the fault. Climate events have not caused equivalent features to be preserved at the same sites. The west coast of the South Island has a very high and relatively constant rate of rainfall (24), which is consistent with regular flushing of catchments. In addition, the timing of peat-silt transitions does not correlate to any regional proxies of paleo-precipitation (25, 26).

We developed a record of depositional events resulting from fault rupture using unit thickness, continuity, sedimentology, and paleoenvironment (27). Twenty-two silt-peat couplets, deposited between 6000 B.C.E. and 1000 C.E., are consistent with an earthquake origin. Although earthquake size cannot be measured directly at Hokuri Creek, we know from the inferred mechanism of formation that every earthquake was big enough to cause surface rupture with a vertical component of offset capable of altering site hydrology. Magnitude estimates for the two most recent events on the southern section of the fault ranged between Mw 7.6 and 8.3 (16). Eighty-two radiocarbon ages were used to produce our best estimate of the earthquake sequence (Fig. 3). The ages for event horizons were derived from bounding radiocarbon dates (usually multiple) on short-lived, fragile organic fractions such as single leaves and seeds. The two most recent earthquakes of the Alpine Fault chronology were adopted from paleoseismic trenches 100 km along strike to the northeast at Haast (21). One of these events is recognized as offsets in the abandoned gorge at Hokuri Creek, but it could not be dated in the stratigraphic section because deposition had ceased by this time. The third event identified at Haast (688 to 1066 C.E.) overlaps in time with the youngest event recognized in the Hokuri Creek stratigraphic section (730 to 921 C.E.). We assume, because of the overlap of these ages and similar large single-event displacements (7.5- to 9-m strike-slip) at both sites, that they represent a single earthquake, and we combined the two dates to generate the age of event HK1 in Fig. 3. Thus, our composite record (Fig. 3) covers the past 24 surface-rupturing events on the southern onshore section of the Alpine Fault between 6000 B.C.E. and the present day.

Fig. 3

Ages of surface-rupturing earthquakes on the Alpine Fault from Hokuri Creek and Haast. (A) Probability distributions for the earthquake ages modeled from bounding radiocarbon ages (table S1). Gray-shaded probability distributions are the events recorded at Haast. Calendar years in brackets after event names are the 95% ranges. The most recent event, Ha1, is shown with a small vertical bar at its inferred date of 1717 ± 5 C.E. (18). (B) Recurrence interval probability distributions for the events shown in (A). Horizontal bars under distributions indicate the 95% range. The black vertical line is the mean recurrence interval (329 years). The gray vertical line represents the elapsed time since the most recent earthquake (295 years).

We estimate a mean recurrence interval of 329 ± 68 years and a coefficient of variation (COV) of 0.33 for the 24-event data set (Fig. 3B). Kagan and Jackson (6) define completely random earthquake recurrences as having a COV of 1, whereas distributions with a COV < 1 are “quasi-periodic” and distributions with a COV > 1 are termed “clustering.” Given the number of intervals examined, earthquake recurrence on the southern onshore section of the Alpine Fault is clearly quasi-periodic. Quasi-periodic earthquake recurrence conforms with elastic rebound theory and the idea that faults rupture in response to a specific amount of strain accumulation (1, 4, 5). Comparison with previous paleoseismic studies on the southern onshore section of the Alpine Fault highlights the strength of a long earthquake record in determining mean recurrence intervals. For example, the three earthquakes identified at Haast (21) indicate an average recurrence time of about 485 years, and although the same Haast events are incorporated into the earthquake record at Hokuri Creek, the mean recurrence interval using 24 earthquakes is 329 ± 68 years. Unlike the Haast study, this value is consistent with the long-term average slip rate on this part of the fault (Fig. 1B), probably because sufficient variation in recurrence is encompassed by the 8000-year time span.

The southern onshore section of the Alpine Fault exhibits more regular or periodic behavior than other major transform faults (Fig. 4). For example, a 3000-year, 29-event record from the San Andreas Fault at Wrightwood in California exhibited quasi-periodic behavior (5) but with a COV of about 0.7. Recent reanalysis of the 10-event record at Pallett Creek near Wrightwood yielded a COV estimate of 0.60 (28). On the Jordan Valley Fault section of the Dead Sea Transform system, Ferry et al. (9) described clustered behavior from a 14,000-year, 12-event record in which recurrence intervals ranged from 280 to 1160 years. Marco et al. (7) estimated COVs for earthquake recurrence in the Dead Sea Graben ranging from 1 (random) to 1.75 (clustered).

Fig. 4

Recurrence patterns of large earthquakes on major transform faults from long geological records. The timing of earthquakes on the Alpine Fault in New Zealand (top line, data from this study) is the closest to periodic with a coefficient of variation (COV) of 0.33. The timing of earthquakes on the San Andreas fault in the United States (second line down) (5) is quasi-periodic, with a COV of 0.7. The timing of earthquakes on the Dead Sea Transform in the Middle East (lower two lines) ranges from random to clustered with COVs of 1 to 1.75 (7, 9).

In light of the periodicity exhibited by the Alpine Fault, we can identify characteristics that may be good indicators of quasi-periodic recurrence on other faults. In a global context, the Alpine Fault is a prime example of a transform plate boundary fault that has a simple structure (long, straight trace with few large step-overs), large total offset (~480 km), and a high slip rate (23 ± 2 mm/year at Hokuri Creek) (13, 29). However, perhaps the most important feature in this context is the lack of other major structures nearby for much of the length of the fault. In the southwestern South Island, one third of relative plate motion (13, 14) is taken up on structures other than the Alpine Fault, but these have little effect on regulating the timing of Alpine Fault earthquakes. Structures that accommodate residual motion in this part of the plate boundary have low slip rates and are distributed broadly across the margin (14, 15). The slightly less periodic (higher COV) San Andreas fault is known to be influenced by major faults nearby (30), and the aperiodic Dead Sea Transform is a more complicated system and slower slipping (9) than either the San Andreas or the Alpine Fault. Other faults that have similar characteristics to the Alpine Fault are sections of the North Anatolian in Turkey and the Denali Fault in Alaska.

Existing examples of long earthquake records suggest a continuum of recurrence behavior with more periodic recurrence on fast-moving, simple and smooth at seismogenic depth, isolated structures at one end of the spectrum (such as this Alpine Fault example) and aperiodic recurrence on low slip-rate, complex, rough, and networked structures at the other end. Thus, in the absence of long paleoseismic records, fault characteristics such as total slip, slip rate, geometric complexity, and possible interaction with other nearby major faults can enable the choice of appropriate statistical models for use in earthquake forecasting and hazard analysis. Regular earthquake recurrence can be considered an end-member of fault behavior. Our study highlights that the regularly repeating earthquake cycle is a realistic foundation on which to base earthquake forecasting and seismic hazard efforts, especially where a fault is acting in isolation to accommodate a high proportion of plate motion.

Supplementary Materials

Materials and Methods

Figs. S1 and S2

Table S1

References (3336)

References and Notes

  1. For example, Milford Sound (Fig. 1B) receives 6 to 7 m of rain annually, and ~186 days per year are wet (31).
  2. Materials and methods are available as supplementary materials on Science Online
  3. Acknowledgments: This research was funded by the Marsden Fund of the Royal Society of New Zealand. The Department of Conservation provided permits for us to work in the field area. D. Barker, T. Bartholomew, M. Hemphill-Haley, N. Litchfield, S. Marco, N. Palmer, D. Pantosti, and R. Van Dissen contributed to the field work. T. Bartholomew, T. Dutton, L. Ingham, G. Turner, and J. West carried out laboratory work. C. Prior and her team produced all the radiocarbon analyses. N. Litchfield, A. Nicol, R. Sibson, M. Stirling, R. Sutherland, J. Townend, and L. Wallace made comments that improved the manuscript. Data reported in this paper are provided in the supporting online material. K.R.B. initiated the research, led the fieldwork, and contributed to all aspects of the project, especially interpretation. U.A.C. contributed to preliminary fieldwork, produced the diatom-based paleoenvironmental interpretations, contributed to figures, and led the writing of the manuscript. K.J.C. was involved in all fieldwork, led development of the stratigraphic model, organized laboratory analyses, and contributed to figures and writing. G.P.B. contributed to fieldwork, interpretation, figures, and writing and produced the event sequence using Oxcal. R.M.L. and P.V. contributed to a large proportion of the fieldwork and to interpretations presented in the manuscript.

Stay Connected to Science

Navigate This Article