- © 2016 by the Seismological Society of America
Recent damaging earthquakes in New Zealand ruptured faults that were not known to be active. We analyzed New Zealand historical moderate‐to‐great magnitude earthquakes since 1845 (Mw 6–8.2) to estimate the level of completeness of earthquake fault sources in the National Seismic Hazard Model (NSHM) and the paleoseismic record. Our analysis assumes that the historical earthquakes are representative of the paleoseismic record and, due to the small number of events (45 Mw≥6), is qualitative. We find that about half of all historical earthquakes Mw≥7.0 ruptured faults that, based on today’s state of knowledge of active‐fault locations, would not have been identified as active prior to the event. The majority of historical events on faults previously not identified as active were Mw<7.3 and either did not displace the ground surface or were located in areas where the rates of erosion or burial exceed fault‐slip rates. Incompleteness of active‐fault sources in the present NSHM is the greatest for earthquake fault sources with long recurrence intervals of ≥10,000 yr. These inferred unidentified active faults will, in many cases, be located in low strain‐rate areas, where they may make an important contribution to the seismic‐hazard budget.
The goal of probabilistic seismic‐hazard analyses (PSHAs) such as the New Zealand National Seismic Hazard Model (NSHM; Stirling et al., 2012) is to infer the sizes and rates of future moderate‐to‐great magnitude earthquakes, primarily using a combination of paleoseismic records and historical seismicity (instrumental and noninstrumental; e.g., Wesnousky et al., 1984; Wesnousky, 1986; Stirling, McVerry, et al., 2002; Stirling, Rhoades, et al., 2002; McCalpin, 2009; Stirling et al., 2012). In PSHA, paleoseismic data derived from active faults provide information on the location, sense of displacement, magnitude, and frequency of moderate‐to‐great magnitude earthquakes (e.g., Mw>5.5) in fault source models, although the uncertainties and incompleteness of these data can be significant (e.g., Stein and Newman, 2004; McCalpin, 2009; Nicol et al., 2011, 2016; Cox et al., 2012). The incompleteness of these active‐fault earthquake sources and the importance of background seismicity models have been highlighted recently by moderate and large damaging earthquakes in New Zealand, which ruptured active faults that were not previously known to exist (e.g., 2010 Mw 7.1 Darfield and 2011 Mw 6.3 Port Hills events; Beavan et al., 2011; Gledhill et al., 2011; Quigley et al., 2012). These earthquakes raise questions about how many other undiscovered active faults have the potential to produce future damaging earthquakes. In this article, we examine the completeness of active‐fault earthquake sources in the NSHM of New Zealand, which are dependent on the quality and quantity of the paleoseismic record. To aid our analysis, we adopt the following definitions that are consistent with application in the NSHM:
Active fault: A fault that has ruptured the ground surface or produced surface deformation in the Late Quaternary (i.e., last 125,000 yrs, except in the Taupo rift, where a younger definition of ∼25,000 yrs is used; e.g., Langridge et al., 2016).
Identified active‐fault earthquake source: A modeled earthquake fault source in the NSHM capable of accommodating future earthquakes. The earthquake source is located and parameterized using paleoseismic information from a mapped active fault(s).
Unidentified active‐fault earthquake source: An earthquake fault source capable of generating future earthquakes not incorporated into the NSHM. These earthquake sources are not identified as active because they do not produce resolvable active Late Quaternary deformation of the ground surface. Unidentified active‐fault earthquake sources may be faults mapped in bedrock but not associated with evidence of Late Quaternary prehistoric earthquakes or faults unmapped in the geology and not known to exist. In both cases, there is neither a mapped active‐fault trace nor information on the faults’ prehistoric earthquake history.
We estimate how many unidentified active‐fault earthquake sources have the potential to generate future damaging earthquakes in New Zealand by assuming that historical earthquakes are representative of the geological record and by calculating the number of missing sources in the NSHM. In addition, we examine historical earthquakes to assess what factors influence the likelihood that these active faults will be included as earthquake sources in the NSHM. We assess the completeness of the NSHM earthquake source model and the paleoseismic record that underpins the model by determining whether or not moderate‐to‐great magnitude earthquakes (Mw 6–8.2) from the post‐1845 New Zealand historical seismicity catalog occurred on identified active‐fault earthquake sources (Figs. 1–3; see Fig. 1 caption for explanation of the catalog period and Table 1 for details of events). Because of the short 171‐yr duration of the historical sample compared with the recurrence interval of ground‐rupturing earthquakes on most faults (e.g., >500 yrs) and the moderate rates of seismicity (one onshore shallow event of Mw≥6 every ∼4 yrs), robust statistical analysis of the data is not possible, and the conclusions are considered qualitative. About half of all historical earthquakes (Mw≥7.0) ruptured active faults that have unambiguous geomorphic expression and, based on today’s state of knowledge of active‐fault locations, would have been identified as active prior to the occurrence of the event (i.e., they would represent active‐fault earthquake sources in the NSHM). Historical earthquakes on unidentified active‐fault earthquake sources either did not displace the ground surface or were located in areas where the rates of erosion or burial exceed fault‐slip rates. The results are preliminary and will likely be refined by similar studies in other regions of active faulting worldwide.
EARTHQUAKE DATA AND ANALYSIS
Historical seismicity provides an important dataset for constraining future seismic hazard. The New Zealand historical seismicity catalog for moderate‐to‐great magnitude events has been compiled by numerous people over the last 50 yrs (e.g., Eiby, 1968; Dowrick and Smith, 1990; Downes, 1995; Doser et al., 1999; Doser and Webb, 2003; Dowrick and Cousins, 2003; Downes and Dowrick, 2015) and contains earthquakes up to Mw 8.2 recorded since 1840. We mainly analyze 45 moderate and great magnitude historical events since 1845 (see Fig. 1 caption for explanation of time interval sampled), with 25 Mw 6–6.4, 8 Mw≥6.5–6.9, and 12 Mw≥7 (see Fig. 3 and Table 1 for summary of Mw≥6.5 events analyzed); these data are primarily from Downes and Dowrick (2015). The poorly located 1843 Mw 7.6 western Hawkes Bay earthquake has been excluded from analysis (for further information on this event, see Downes and Dowrick, 2015), although it is retained in the figures for completeness. Post 1943, the locations, magnitudes, dimensions, and displacements of earthquakes were determined instrumentally, whereas prior to 1943, locations were also partly estimated using modified Mercalli‐felt intensities of ground shaking and, for some of the larger events (e.g., Marlborough 1848, Wairarapa 1855, North Canterbury 1888, Buller 1929, Hawkes Bay 1931; Fig. 3), by analysis of ground surface rupture (e.g., Cowan, 1990; Hull, 1990; Grapes et al., 1998; Berryman and Villamor, 2004; Schermer et al., 2004; Mason and Little, 2006; Rodgers and Little, 2006). The magnitude–time plot in Figure 1 shows fewer Mw<7 events prior to ∼1900 and supports an increasing magnitude of completeness (Mc) back through time (dotted line, Fig. 1). The earthquake catalog is believed to be approximately complete for magnitudes of ≥6.5 since 1840, ≥5 from 1943, and ≥4 after 1964 (Dowrick and Cousins, 2003). These Mc values are here assumed to be correct, although they appear to be poorly defined prior to 1943.
Because of data uncertainties and to avoid ambiguity in the results, only shallow earthquakes (≤25 km focal depth) located onshore (or may have ruptured faults that extend onshore) were considered. The focal depths are mainly from Downes and Dowrick (2015) and, for preinstrumental events, have been estimated from a range of observations, including the distribution of aftershocks and the thickness of the seismogenic crust. The selected focal depth cutoff is imposed to include only events that have the potential to rupture the ground surface, and using cutoffs ranging from 20 to 30 km does not modify the results for Mw≥7 events, which all have listed focal depths of <20 km. In the eastern North Island, three shallow events at depths of 16–25 km may have ruptured the overriding Australian plate, the subduction interface, or the subducting Pacific plate (Fig. 3; 1863 Dannevirke Mw 7.5, 1881 Palmerston North Mw 6.5, and 1917 Wairarapa Mw 6.6 events). We performed two calculations that account for the uncertainty in whether these three historic earthquakes occurred in the overriding plate and have the potential to rupture the ground surface. In the first calculation, the events are assumed to rupture the overriding plate with no surface rupture, whereas in the second calculation, they are assumed to rupture the subduction thrust or the subducting plate and are excluded from the analysis. Whether these events are categorized as upper plate, lower plate, or interplate does not significantly influence the general conclusions of this article.
The locations of some of the events presented in Figure 1 are shown in Figure 2b (Mw≥6) and Figure 3 (Mw≥6.5). Although detailed analysis of these locations is beyond the scope of this article, it is clear that the events are not uniformly distributed across New Zealand. The majority of historical onshore earthquakes ruptured faults in the eastern North Island and the northern South Island (Fig. 2b). Similarly, onshore historical large‐magnitude earthquakes were not uniformly distributed in time. Of particular note is a cluster of Mw≥6.5 events in the 25 yrs following 1917 (Fig. 1). Many of the events in the 1920s–1940s are also clustered in space, which may reflect increases in stress magnitudes and triggered slip induced by the 1929 Mw 7.7 Buller and 1931 Mw 7.8 Napier earthquakes (compare to Stein et al., 1997; Steacy et al., 2014).
New Zealand and global datasets of Mw≥5 events have been utilized here to examine the magnitude dependence of ground‐surface rupture (Fig. 4). Twenty Mw≥6.5 and twelve Mw≥7 events were also used to determine how many historical large‐magnitude events ruptured identified active‐fault earthquake sources (i.e., earthquake fault sources that are based on active faults detected at the ground surface and that would have been, with the present state of knowledge of active‐fault locations, modeled as a fault source in the NSHM even if the fault had not ruptured historically; Fig. 3). In making this assessment, knowledge of the existence of active faults due to detailed investigations arising directly from a historical earthquake has been discounted. For the most part, identified active‐fault earthquake sources are based on active faults that have scarps formed during prehistoric surface rupture(s). Historical earthquakes on unidentified active‐fault earthquake sources either ruptured bedrock faults that, based on the present state of knowledge of active‐fault locations, would not have been known to be active (e.g., the 1929 Mw 7.7 Buller event), or occurred where no bedrock fault would have been mapped (e.g., 2010 Darfield Mw 7.1 event). Determining the likelihood that a historical earthquake ruptured an identified active‐fault earthquake source provides information about the completeness of earthquake sources in the NSHM and associated paleoearthquake record. Factors that contribute to the completeness of the active‐fault earthquake sources include the relationships between magnitude and the probability of surface rupture and the role of surface process (erosion and deposition) in fault‐scarp preservation. These factors are examined using historical earthquakes (e.g., Figs. 1, 2b, and 3) and their comparison to paleoearthquakes estimated from the active‐fault earthquake source model in the NSHM (Stirling et al., 2012; Langridge et al., 2016; Fig. 2a).
ACTIVE‐FAULT EARTHQUAKE SOURCE IDENTIFICATION
The locations, names, and magnitudes of shallow historical events Mw≥6.5 are presented in Figure 3 and Table 1. Of these historical earthquakes, 14 ruptured faults that would not have been identified as active in the absence of the event (white‐filled rectangles in Fig. 3). A further six earthquakes ruptured identified active‐fault earthquake sources (i.e., ruptured known active faults) and are indicated by the gray‐filled rectangles in Figure 3. In addition, events where the primary fault ruptured the ground surface (thick black rectangle borders) or could have occurred on the subduction thrust and deeper (dotted gray rectangle borders) have been differentiated in Figure 3. These historical data suggest that whether moderate‐to‐great earthquakes occur on identified active‐fault earthquake sources is magnitude dependent. Figure 3 shows that unidentified active‐fault sources outnumber identified active‐fault sources, particularly at Mw≥6.5 to <7, with 80%–90% (i.e., up to seven events) of these events on unidentified active‐fault earthquake sources. The proportion of earthquakes on unidentified active‐fault earthquake sources decreases significantly for Mw≥7 historical events and is 58% (7 of 12) if the possible subduction‐related Mw 7.5 Dannevirke event is included in the calculations or 55% (6 of 11) with exclusion of the Dannevirke event. The percentage of Mw≥7.5 events on unidentified active‐fault earthquake sources decreases to ∼20%–30% (1 of 5, or 2 of 6 events), suggesting that the majority of these earthquakes rupture the ground surface and produce mappable active‐fault scarps. The conclusion that earthquakes Mw>7.5 can occur on fault sources not identified as active is illustrated by the 1929 Mw 7.7 Buller earthquake. The 1929 Buller earthquake ruptured a bedrock fault that accrued several kilometers of reverse displacement in the last 20–30 my (Ghisetti et al., 2016); however, in the absence of the 1929 surface rupture, there is no evidence that this fault is active (Berryman, 1980). Therefore, using the terminology defined in this article, the 1929 earthquake occurred on an unidentified active‐fault earthquake source that, in the absence of the 1929 event, would not have been incorporated as a fault source in the NSHM.
Incomplete sampling of active‐fault earthquake sources may reflect a combination of censoring effects, including subresolution deformation, surface erosion or deposition processes, and nonsurface rupture of some earthquakes. These sampling artifacts and processes influence the recognition of active faults and are likely to affect the detection of moderate‐magnitude earthquakes more severely than large‐magnitude events, as can be inferred from the historical data. Fault scarps produced by surface rupture can be removed by erosion or buried during sedimentation. Figure 3 indicates that all of the large to great historical onshore earthquakes on faults mapped as active ruptured the ground surface (i.e., in the figure, all gray‐filled rectangles have bold black borders). However, not all historical surface‐rupturing earthquakes occurred on a fault identified as active, even though slip on these faults could have produced surface scarps during the Late Quaternary (e.g., 1929 Mw 7.7 Buller and 2010 Mw 7.1 Darfield events; Berryman, 1980; Hornblow et al., 2014). In such cases, the active faults may not have been identified because they ruptured the ground surface in regions where the rates of erosion or burial were greater than the rates of fault displacement. For example, the Greendale fault ruptured alluvial plains in the 2010 Darfield earthquake (see Fig. 2a for fault location), but was not mapped prior to this event because the penultimate surface rupture occurred during alluvial sedimentation, which eroded and then buried the fault scarp that formed 20,000–30,000 yrs ago (Hornblow et al., 2014). Preservation of active‐fault traces (and accordingly the definition of active‐fault earthquake sources) is most likely to be poor in regions where the regional strain rates and fault displacement rates are low relative to the rates of surface process, as might occur on alluvial plains and in mountainous areas. The relatively low number of identified active faults in the Southern Alps (Fig. 2a), for example, may partly reflect the high rates of erosion in this part of the South Island (Cox et al., 2012).
PROBABILITY OF SURFACE RUPTURE
Historical earthquakes do not always rupture the ground surface with measureable displacement, and these active faults (and the characterization of associated active‐fault earthquake sources) will be difficult to identify from surface observations (Wells and Coppersmith, 1993; Lettis et al., 1997; Hecker et al., 2013; Nicol et al., 2016). The probability of resolvable surface rupture is dependent on earthquake magnitude and a range of additional factors, including fault type, rock properties (including the thickness of poorly consolidated sedimentary cover rocks), tectonic setting, and the resolution of the available topographic data pre‐ and postearthquake. Relationships between the probability of surface rupture and magnitude are shown in Figure 4 for New Zealand (this study) and global (Wells and Coppersmith, 1993, 1994; Berryman et al., 2001) historical earthquakes. Because earthquakes recorded prior to the routine use of instrumental data could be biased toward surface‐rupturing events (e.g., Wells and Coppersmith, 1993), we have also plotted global data between 1954 and 1994 from table 1 of Wells and Coppersmith (1994). Although the Wells and Coppersmith dataset is dated and dominated by earthquakes of different scaling from New Zealand’s (e.g., Stirling, Rhoades, et al., 2002), examination of their table 1 is useful for the purposes of our analysis. Independent of the source publication, the duration of the sample, or the mode of recording, all global compilations primarily plot within the light gray polygon in Figure 4 and display a near‐linear positive relationship between the probability of surface rupture and magnitude; the global curves are also similar to global relationships proposed by Lettis et al. (1997) and Hecker et al. (2013). In general, fewer historical New Zealand earthquakes appear to have ruptured the ground surface for a given magnitude than global events (compare light and dark gray polygons; Fig. 4). This difference may derive from a number of factors including (a) the under recording of surface ruptures in New Zealand (particularly in the nineteenth century when the country was widely forested and the population sparse), (b) the small number of earthquakes in the New Zealand sample, and/or (c) a disproportionate number of New Zealand events that do not produce surface rupture because, for example, they are associated with subduction, surface folding, or a thick seismogenic crust.
Irrespective of the cause(s) of the discrepancy between the two datasets, a number of conclusions can be drawn from Figure 4. First, there is a small chance (e.g., <0.1 for global data) that earthquakes of Mw≥7.5 will produce fault slip that does not displace the ground surface and may not be identified as an active fault (and therefore may not be used in the characterization of active‐fault earthquake sources). Large to great events that do not rupture the ground surface may be particularly important along the subduction margins in northeast and southwest New Zealand. Second, there is a positive relationship between the probability of surface rupture and magnitude, with probability decreasing at ∼0.3–0.5 per earthquake magnitude unit.
Insights into the factors that may locally influence the relation between surface rupture and magnitude are provided by the New Zealand historical earthquake catalog. In New Zealand, surface‐rupturing earthquakes Mw<7 are restricted to normal faults in the Taupo rift (see Fig. 2a for location), where the crust is thin and hot and the vast majority of earthquakes have focal depths of <12 km and estimated maximum magnitudes of Mw≤6.8 (Villamor and Berryman, 2001; Hurst et al., 2002). In regions dominated by reverse and strike‐slip faulting, historical surface‐rupturing earthquakes are Mw>7, although slip on reverse faults at depth can produce recordable folding and differential vertical deformation of the ground surface (e.g., Beavan et al., 2011). Although the NSHM has 10 potential sources of non‐surface‐rupturing faults (Stirling et al., 2012), the inclusion of these sources is insufficient to match the pattern in the historical record, suggesting that blind faults make a contribution to active‐fault earthquake sources capable of generating Mw≤7 events not resolvable in the near‐surface geology (for further discussion, see Lettis et al., 1997). If these historical observations also generally apply to prehistorical events, then strike‐slip and reverse fault scarps are most likely to have been formed during events of Mw≥7. This suggestion contrasts with the >50 reverse and strike‐slip active‐fault earthquake sources in the present NSHM (with active‐fault scarps formed during surface rupture) that have estimated maximum magnitudes of Mw<7. This discrepancy might partly arise because the magnitudes of prehistorical events are calculated from fault lengths that are likely to be underestimated.
RECURRENCE INTERVALS AND ACTIVE‐FAULT SAMPLING
The relationships between active‐fault preservation and slip rate (a consequence of earthquake slip and recurrence interval) have been tested here by categorizing historical New Zealand Mw≥6.5 earthquakes into short (<1250 yrs), intermediate (1250–10,000 yrs), and long (>10,000 yrs) recurrence interval faults using the available literature (Fig. 5). The data are presented in a stick plot of magnitude versus time, which discriminates different recurrence interval faults (see filled circles in the legend and the figure caption for derivation of recurrence intervals), identified active‐fault earthquake sources (black sticks), unidentified active‐fault earthquake sources (gray sticks), and possible subduction events (horizontal arrows). Faults identified as active that ruptured the ground surface in the historical period under consideration here (i.e., since ∼1845) exclusively have recurrence intervals in the short and intermediate recurrence interval classes, with most (four of six) events on the highest slip rate and lowest recurrence interval faults (Fig. 5). In contrast, many of the unidentified active‐fault earthquake sources that ruptured the ground surface in the historical period have long recurrence intervals, consistent with the notion that fault scarps are most likely to be eroded or buried when fault slip rates are low and elapsed times since the last event are long (both characteristics being typical of long recurrence interval faults). Given the small number of events, it cannot be determined if fault preservation and recurrence are related to magnitude. Figure 5 may suggest a weak correlation between magnitude and fault preservation, with many of the largest magnitude events on faults identified as active possibly because longer rupture lengths with greater displacements of the ground surface are more likely to be preserved. Despite the possible relationship between magnitude and preservation, large‐magnitude events can and do rupture faults of long recurrence interval, as indicated by the 1929 Mw 7.7 Buller event (Fig. 3).
Undersampling of long recurrence interval active faults appears to be reflected in the NSHM active‐fault earthquake source model. To examine the role of recurrence interval in the preservation and identification of active‐fault earthquake sources, we have plotted a nationwide recurrence interval for onshore active‐fault earthquake sources against magnitude for the NSHM and historical earthquake datasets (Fig. 6). Individual fault earthquake sources and historical earthquakes have been grouped into one of three recurrence interval classes (<1250 yrs, 1250–10,000 yrs, and >10,000 yrs), which are plotted separately in Figure 6a,b, and c, respectively. In each graph, the nationwide recurrence intervals are for all events greater than or equal to a given magnitude. The numbers at the top left of the graphs are recurrence intervals for all events Mw≥7. Because of uncertainties in the nationwide recurrence intervals (see error bars in Fig. 6), mean values within a factor of two for NSHM and historical data are here considered to be comparable. For individual faults with recurrence intervals of <10,000 yrs, the relationships between the nationwide recurrence interval and earthquake magnitude are similar for paleoseismicity derived from NSHM active‐fault earthquake sources and historical earthquakes (Fig. 6a,b). These observations suggest that the rates of earthquakes on short and intermediate recurrence interval active‐fault earthquake sources are comparable for the NSHM and historical records. In contrast, the relationships between nationwide recurrence intervals and earthquake magnitudes differ significantly between the NSHM and historical earthquakes for long recurrence interval active‐fault earthquake sources (Fig. 6c), with recurrence for the NSHM active‐fault earthquake sources being significantly longer (∼180–900 yrs vs. 40 yrs for Mw≥7). The implication of this discordance is that the proportion of moderate‐to‐large earthquakes on long‐recurrence faults is significantly higher in the historical record than for the national active‐fault earthquake source model. To account for these differences, it could be that either the historical record is not representative of the long‐term seismicity (e.g., >10,000 yrs) and/or that active‐fault earthquake sources in the NSHM undersample long recurrence interval faults. Using the recurrence interval values in Figure 6c, we estimate that an additional ∼140 active‐fault earthquake sources capable of generating Mw≥7 earthquakes are required in the NSHM to approximately achieve a match with the historical earthquake catalog. Because these potential earthquake sources are distributed along the length of the plate‐boundary zone and rupture infrequently (i.e., on average, every >10,000 yrs), they may only accommodate a small proportion (e.g., <10%) of the total strain budget across New Zealand.
IMPLICATIONS FOR SEISMIC‐HAZARD ASSESSMENT
The active‐fault earthquake source model in New Zealand appears to be incomplete. Despite the incompleteness of the paleoseismic record, these data provide important constraints for seismic hazards when the limitations of the data are appropriately accounted for (and tested) in the models (Stein et al., 2012). The present study enhances the information base for the background seismicity model of the present NSHM by providing the number of earthquakes in the existing background model that are expected to be produced by unidentified active‐fault earthquake sources. The completeness of active‐fault earthquake sources in the NSHM rises as a function of increasing magnitude and decreasing recurrence interval. Based on today’s state of knowledge of active‐fault locations, 10%–20% of Mw≥6.5 to <7 and ∼50%–60% of Mw≥7 historical earthquakes ruptured active‐fault earthquake sources that would have been identified (and detected) prior to the event. Incompleteness of active‐fault earthquake sources in the NSHM is greatest for earthquake sources with long recurrence intervals of ≥10,000 yrs. Historical earthquakes of Mw≥6.5 on faults with recurrence intervals of ≥10,000 yrs are at least four times more frequent than forecasts based on NSHM active‐fault earthquake sources with the same magnitude and recurrence interval ranges. Unidentified active‐fault earthquake sources occur in cases where evidence of surface rupture is removed by erosion or burial, or the earthquakes did not displace the ground surface.
To account for moderate‐to‐large magnitude earthquakes (Mw≥5) that do not occur on identified active‐fault earthquake sources and have not been directly incorporated into the NSHM as active‐fault earthquake sources, the background seismicity model predicts the magnitude and frequency of earthquakes according to the Gutenberg–Richter relationship up to a magnitude of 7.2 (Stirling et al., 2012). Questions remain about whether the background seismicity model adequately accounts for large‐magnitude earthquakes on unidentified active faults over timescales longer than the instrumental seismicity record. The background model adds about 50% to the fault source model at Mw≥7, which is comparable to the incompleteness estimated here from historical earthquakes. Despite the general accordance of our observations with the background seismicity in the NSHM, a case can be made based on Figure 4 to increase the maximum magnitude in the model up to Mw 7.5–7.8, in line with an earlier version of the NSHM (Stirling, McVerry, et al., 2002).
Recent advances in geophysical techniques for identifying surface and subsurface fault deformation (e.g., geodetic surveys, light detection and ranging [lidar], Interferometric Synthetic Aperture Radar, and seismic reflection profiles) provide optimism that with time we will increase the number of identified active‐fault earthquake sources in the NSHM and the paleoseismic information on these faults. However, it seems unlikely that we will ever be able to identify all of the potential sources of future moderate‐to‐large magnitude earthquakes from geological and geophysical investigations. Presently, background seismicity models are used to account for the incompleteness of fault sources but do not estimate the precise locations of potential earthquakes. Detailed discussion of preferential distribution of background seismicity is beyond the scope of this article; however, consideration might be given as to whether the locations of unidentified active‐fault earthquake sources can be constrained by geological faults that have not been mapped as active. The 1929 Mw 7.7 Buller event on the White Creek fault and 1868 Mw 7.2 Cape Farewell event on the Whakamarama fault indicate that future earthquakes will likely rupture some faults mapped in bedrock that are not considered to be active. Thousands of these bedrock faults with no definitive evidence of Late Quaternary earthquake activity (i.e., ≤25 ky in the Taupo rift and ≤125 ky everywhere else) have been mapped throughout New Zealand and could be included as modeled active‐fault earthquake sources in the NSHM. Further analysis is required to determine under what circumstances geological faults not explicitly identified as active should be included as active‐fault earthquake sources in the NSHM (e.g., bedrock faults optimally oriented for failure in the present stress regime; compare to Sibson et al., 2011), to examine how inclusion of these bedrock faults influences the hazard calculations, and to assess whether the inclusion of bedrock faults materially improves the hazard estimates.
The completeness of paleoseismic data from active faults can vary between tectonic domains. The probability of surface rupture for moderate‐magnitude events is noticeably higher in the normal‐faulting domain of the Taupo rift, where historical events with magnitudes in the mid 5s are reported to have ruptured the ground surface (Downes and Dowrick, 2015), than elsewhere in New Zealand. In regions of reverse and strike‐slip faulting, surface‐rupturing earthquakes of Mw<7 appear to be rare, and the background seismicity model may be required to account for the majority of Mw<7 events. Similarly, analysis of recurrence intervals suggests that the completeness of the paleoseismic record may decrease in areas where the regional strain rates are low and the recurrence intervals of active faults are long (e.g., >10,000 yrs). Such areas of low strain rates might include the western and northern North Island, and the northwest and southeast South Island. Further analysis of the available data is required to constrain better regional variations in completeness of the paleoseismic record. This analysis may provide additional information on the location, orientation, and recurrence behavior of unidentified active‐fault earthquake sources.
DATA AND RESOURCES
All data used in this article came from published sources listed in the references.
This research was funded by New Zealand Earthquake Commission (EQC) and the New Zealand Natural Hazards Research Platform. Hannu Seebeck prepared the maps in Figure 2. We thank James Dolan, Julian Lozos, Seth Stein, and an anonymous reviewer for thorough and constructive reviews of the article.