Earth and Planetary Science Letters 301 (2011) 52–64

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e p s l

Deriving fault-slip histories to test for secular variation in slip, with examples from
the Kunlun and Awatere faults
Ryan D. Gold ⁎, Eric Cowgill
Department of Geology, University of California, Davis, CA 95616, USA

a r t i c l e i n f o a b s t r a c t

Article history: Although offset and age data from displaced landforms are essential for identifying earthquake clusters and thus
Received 12 March 2010 testing whether faults slip at uniform or secularly varying rates, it is not clear how the uncertainties in such
Received in revised form 7 October 2010 measurements should be propagated so as to yield a robust fault-slip history (i.e., record of fault displacement over
Accepted 7 October 2010
time). Here we develop a Monte Carlo approach for estimating the distribution of geologically reasonable fault-slip
Available online 30 November 2010
histories that fit the offset and age data from a population of dated and displaced landforms. The model assumes
Editor: T.M. Harrison that the landforms share common faulting histories, the offset and age constraints are correct, and the fault has not
reversed shear sense. Analysis of the model results yields both a precise average slip rate, in the case where a linear
Keywords: fit is applied to the data, and a best-fit fault-slip history, in the case where the linear constraint is removed. The
slip rate method can be used to test for secular variation in slip because the uncertainty on this best-fit history is quantified.
fault-slip history By applying the method to previously published morphochronologic data from faulted late Quaternary terrace
secular variation in slip risers along the Kunlun fault in China and the Awatere fault in New Zealand, we have assessed the extent to which
morphochronology our modeled average slip rates match previously reported values and the data support previous interpretations of
Kunlun fault
uniform slip rate. The Kunlun data set yields average slip rates of 8.7 +3.6/−2.1 mm/yr and 5.1+ 1.6/−1.2 mm/yr
Awatere fault
(68.27% confidence), for the central and eastern reaches of the fault, respectively, both of which match previously
Monte Carlo
published slip rates. Our analysis further indicates that these fault reaches have both slipped uniformly over the
latest Quaternary. In contrast, analysis of data from the Saxton River site along the Awatere fault reveals a mid-
Holocene deceleration in slip rate from 6.2 +1.6/−1.4 mm/yr to 2.8+ 1.0/−0.6 mm/yr. This result contradicts
previous interpretations of uniform slip along the Awatere fault. The Monte Carlo method we present here for
quantifying fault-slip histories using the offset and age data from a population of faulted landforms provides an
important tool for distinguishing temporally uniform from secularly varying fault slip.
Published by Elsevier B.V.

1. Introduction changing surface loads such as those from glaciers or large lakes
(Hampel and Hetzel, 2006; Hampel et al., 2009; Hetzel and Hampel,
Determining the extent to which temporal clustering of surface 2005). In addition, measuring the constancy of strain release along
rupture occurs on a single fault is essential for understanding active faults guides seismic hazard evaluation (e.g., Knuepfer, 1987;
earthquake and fault mechanics (Ben-Zion, 2008). Characterizing such Lave et al., 2005; Weldon et al., 2004).
clustering is important because measured slip histories provide the An individual earthquake cycle for a locked fault is conceptually
observational basis for testing a number of mechanical models of the based on elastic rebound theory (Reid, 1910). In the cycle, strain
earthquake cycle. Such models predict temporal variations in the rate of accumulates across a locked fault until a critical threshold is reached, at
fault slip due to processes such as co-dependent slip along conjugate which time the stored energy is released in an earthquake (Fig. 1).
faults (e.g. Ambraseys, 1971; Hubert-Ferrari et al., 2003; Peltzer et al., Postseismic relaxation follows the earthquake, during which transient
2001) and fault strands (Bennett et al., 2004; Dolan et al., 2007; Nicol deformation deviates from the preseismic rate (e.g., Savage and Svarc,
et al., 2006), post-seismic stress recycling in the non-seismogenic layer 2009; Thatcher, 1983). The cycle then repeats, with strain accumulation
(Kenner and Simons, 2005), metamorphic reactions and grain size and release via a later earthquake. Other modes of fault behavior include
reductions generated by initial events that lead to reduced thresholds fault creep, in which movement across a fault is semi-continuous
for future slip events (Oskin et al., 2008), climatically-driven changes in through time (e.g., Steinbrugge et al., 1960; Wesson, 1988) and silent
crustal water content (Chéry and Vernant, 2006), or the response to earthquakes or transient creep (e.g., Melbourne et al., 2005; Ozawa et al.,
2002) in which seismic moment, often in subduction zones, is released
⁎ Corresponding author. Geologic Hazards Science Center, U.S. Geological Survey,
over periods of days to months.
Golden, CO 80401, USA. Our goal in developing the methods we present here is to resolve
E-mail address: rgold@usgs.gov (R.D. Gold). the history of fault slip as it is averaged over tens to hundreds of surface-

0012-821X/$ – see front matter. Published by Elsevier B.V.

doi:10.1016/j.epsl.2010.10.011
 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64 53

50
morphochronology offers several advantages, particularly at longer
A maximum
displacement timescales. First, slip history reconstructions require measurements of
Displacement (m)

minimum both fault offset and rupture timing. While paleoseismic trenches
age maximum typically yield robust temporal records of rupture, it is often challenging
age
minimum
to directly measure the corresponding displacements (Weldon et al.,
displacement 2002). In contrast, the cumulative surface offsets recorded by faulted
landforms are easier to determine. Second, morphochronologic records
have the potential to extend geologic records of fault slip to longer
0 timescales (tens or hundreds of thousands of years) because displaced
50 landforms can be preserved for time intervals well in excess of those
B available in typical paleoseismic studies. To access these same time-
Displacement (m)

accelerated scales using traditional paleoseismic techniques requires very deep

strain release
trenches (McCalpin, 1996b) or sites at which faulting is widely
distributed so that individual events can be distinguished (Weldon
d-t envelope
(measurement) et al., 2002). Furthermore, at these long timescales paleoseismic
average slip rate trenching techniques are vulnerable to missing events due to periods
fault slip history of non-deposition (McCalpin, 1996b). Additionally, longer time periods
short-term slip rate
0 correspond to larger total offsets, which can result in the juxtaposition of
0 20
Time (ka) non-correlative stratigraphic packages across the fault that can be
difficult to interpret and may result in either the omission or duplication
Fig. 1. Hypothetical records of faulting with identical average slip rates, but of paleoseismic records (Weldon et al., 1996).
(A) temporally steady and (B) secularly varying slip histories (modified from Wallace, Recent developments in topographic surveying (e.g., Haugerud et al.,
1987). In plot (B) the millennial slip rate is matched by the slip history integrated over a
small number (2–3) of seismic cycles. In plot (B) secular variations in the frequency and
2003) and geochronology (e.g., Gosse and Phillips, 2001; Murray and
offset associated with coseismic slip lead to local slip rates integrated over several Olley, 2002; Sharp et al., 2003) have facilitated the collection of
paleoseismic events that do not match the slip rate averaged over the whole record. To morphochronologic data. Likewise, the uncertainties associated with
distinguish between these scenarios, a sufficient number of displacement–time (d–t) such measurements are also increasingly understood (e.g., Cowgill,
measurements with adequate precision are required. These measurements are
2007; Gold et al., 2009 #1260; Zechar and Frankel, 2009). Morpho-
graphically represented by d–t envelopes, the bounds of which are defined by
minimum and maximum constraints on offset and age, respectively. chronologic data are typically analyzed by determining average rates
using linear fits to populations of age-offset determinations from
multiple landforms (e.g., Mason and Little, 2006b; Mériaux et al., 2004,
rupturing earthquakes, or over similar time spans on a creeping fault. In 2005; Van der Woerd et al., 2002; Zechar and Frankel, 2009). In some
particular, we seek to discern secular variation in fault slip at millennial cases, comparison of geodetically constrained rates of fault slip with
time scales, which is temporal variation in the rate of fault slip when the millennial average rates derived from morphochronology has revealed
slip rate is averaged over multiple surface-rupturing events. To do so, we discrepancies interpreted to reflect variable rates of strain release
use displaced and dated landforms to constrain fault-slip histories, (Chevalier et al., 2005; Frankel et al., 2007; Knuepfer, 1992; Oskin and
which record the timing and magnitude of fault displacement through Iriondo, 2004; Oskin et al., 2008; Peltzer et al., 2001). Only a limited
time (Friedrich et al., 2003; Knuepfer, 1992; Wallace, 1987). For a fault number of studies have used populations of morphochronologic
that does not display secularly varying slip, the slopes of the average slip observations to construct fault-slip histories to quantify slip-rate
rate and slip history are nearly identical when averaged over a small variation (e.g., Bennett et al., 2004; Friedrich et al., 2003; Knuepfer,
number (e.g., 2–3) of surface ruptures (Fig. 1A). We refer to this 1992; Wallace, 1987; Weldon and Sieh, 1985).
behavior as “constant slip rate,” by which we mean that when averaged Here we present a new Monte Carlo method for analyzing age and
over several earthquake cycles, the average slip rate is constant through offset measurements from faulted landforms to evaluate the temporal
time. In contrast, the average rate and slip history diverge for faults constancy of slip along a fault. We begin in Section 2 by describing the
displaying secular variation in slip (Fig. 1B). Monte Carlo method, which identifies a distribution of fault-slip
Paleoseismologists have documented secular variation in slip at histories that fit a population of dated and displaced landforms that
millennial timescales along the San Andreas Fault (~5 kyr record) are assumed to share a common slip history. This distribution of
(Weldon et al., 2004), in the eastern California shear zone and Los histories yields 1) an average slip rate for the interval spanned by the
Angeles basin (~12 kyr record) (Dolan et al., 2007), and within the Dead morphochronologic data and 2) a fault-slip history from which we can
Sea Graben (~50 kyr record) (Marco et al., 1996). However, most discriminate uniformly slipping versus temporally varying scenarios.
paleoseismic records are limited to the last few millennia and the most In Section 3, we compare the Monte Carlo method with previous
recent 2–4 earthquakes (e.g., Mason and Little, 2006b; McCalpin, 1996b; approaches to the analysis of morphochronologic data. We then
Sieh, 1978a,b; Washburn et al., 2001; Weldon et al., 2004). An alternative review uncertainties in morphochronologic data in Section 4 and the
method for determining “deep time” fault records that extend tens to considerations associated with integrating measurements from along
hundreds of thousands of years is to use morphochronology, in which the fault strike in Section 5. In Section 6 we present two case studies in
offsets and ages of landforms that are cut by the fault are used to which we use the method to analyze previously published morpho-
reconstruct the history of fault slip (Tapponnier et al., 2001a). Such chronologic data from faulted Quaternary terrace risers along the
faulted landforms include fluvial terrace risers, stream channels, alluvial Kunlun fault (northern Tibet, China) and the Awatere fault (South
fans, shorelines, marine terraces, and glacial moraines. Geometric Island of New Zealand). The case studies allow us to assess the extent
components of such landforms can be used to define piercing lines that to which the modeled average slip rates that we obtain match the
can be projected into a fault plane to define piercing points. Importantly, previously reported values, and to examine if previous claims of
an average slip rate since the time at which a feature formed can be uniform slip rate are supported by the data.
obtained by measuring both the offset (d) recorded by a set of piercing
points and the age (t) of the landform that defines them. Such rates 2. Modeling approach
typically integrate offsets produced by multiple surface ruptures.
Although morphochronology and traditional paleoseismic trenching We investigate two questions: first, can multiple displacement-time
techniques can be used in tandem to reconstruct faulting records, (d-t) points be combined to reduce uncertainty in the average slip-rate
54 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64

measurement, assuming that the rate is uniform over time? Second, can can pass by reducing the “conceptually permitted slip history area” to
multiple d–t data be combined to test for secular variation in slip, and in that of the “model slip history area,” as shown on Figure 2D. The latter
particular, can the statistical significance of possible variations be space is bounded by lines linking the outermost corners of the
quantified? To address these questions, we use a Monte Carlo method to trimmed d–t envelopes. Limiting the model space in this way forces
identify the range and quantify the distribution of slip histories that are the slip histories to take the shortest possible paths between a set of
permitted by populations of d–t data. Because the focus of this section is points selected from within the individual trimmed d–t envelopes,
to present the Monte Carlo method, we restrict our analysis to examples and as a result, the slip histories will be as linear as possible. Thus, the
with boxcar uncertainties, as justified in Section 4 below, but we note model is conservative in identifying secular variation in slip: although
that the method is generalizable to any form of uncertainty (e.g., a suite of d–t observations may permit secular variation, such as a
Gaussian, trapezoid, probability distribution function (PDF), etc.). To history that follows the outer edge of the conceptually permitted area
quantify the uncertainties in individual age and offset measurements, on Figure 2C, the model will only identify variations in slip rate that
we bracket them using minimum and maximum bounds provided by are required by the d–t data.
geologic and geochronologic observations. Graphically, these bounds
are represented as envelopes on displacement–time plots. 2.3. Modeling slip histories

2.1. Model assumptions To fit a suite of slip histories to a population of trimmed d–t

envelopes, we use a Monte Carlo method that employs the following
1. The faulted landforms share a common slip history; i.e., after steps in a Matlab computational code:
adjacent features formed, they experienced the same cumulative
1. Construct d–t envelopes—D–t envelopes are constructed from
displacement history.
primary age and offset determinations, which have been critically
2. The shear sense along a fault has been uniform for the time interval
reviewed for inclusion in the model. We define the maximum age
spanned by the model; i.e., on a displacement–time plot, fault-slip
bracket using the upper bound on age, plus the 2-sigma (2σ) value.
histories must have slope ≥ 0.
Similarly, for the minimum age bracket we use the lower bound on
3. The d–t envelopes contain the true age and offset of their corre-
age, minus the 2-σ value. The displacement brackets are defined
sponding markers. In detail we use the simplest model possible, in
using the minimum and maximum for displacement (e.g., for an
which all points within a d–t envelope are equally likely to be the true
offset determination of 50 ± 10 m (2-σ), minimum and maximum
value with zero probability that the value lies outside of the two-
values of 40 and 60 m are used, respectively).
dimensional boxcar (e.g., Sutherland et al., 2006).
2. Trim—Bounds on a set of input d–t envelopes are trimmed using
the no-negative slope rule, to reduce search area and improve
2.2. Trimming to resolve permissible slip history areas
computational efficiency. For situations where d–t envelopes
overlap so that trimming would create complex polygons, the d–t
Together, assumptions 1 and 2 lead to a key conclusion: relation-
envelopes are not trimmed because such polygons are difficult to
ships between d–t envelopes reduce uncertainty and thus limit the
sample uniformly.
range of possible slip histories. This point is illustrated using a
3. Random point—For each slip history iteration, one point is selected
synthetic dataset presented in Figure 2, in which three faulted
in each trimmed d–t envelope using a uniformly distributed
landforms are represented by d–t envelopes, numbered 1–3 in order
pseudorandom number generator so that calculations can be
of decreasing age. If these features share a common slip history, then
reproduced and tested (Bevington and Robinson, 2003). The line
older features must have offsets that are equal-to-or-larger-than
connecting the resulting set of points is one candidate slip history.
those of younger ones. The green line in Figure 2B shows a permissible
4. Check slope—The candidate slip history is queried to determine if it
slip history, in which features 1 and 2 experienced the same
has a non-negative slope. Computationally, this is accomplished by
magnitude of cumulative offset. In contrast, the red line in Figure 2B
evaluating the degree to which two variables (e.g., offset and age)
shows an impermissible history, which requires the fault to reverse
can be described by a monotonic function. For a slip history defined
shear sense so that feature 2 can accumulate more cumulative offset
by points that can be connected by line segments with slope
than feature 1. Importantly, the general conclusion is that portions of
values ≥ 0, the candidate history is accepted as being geologically
d–t envelopes 1 and 2 are incompatible with one another (Fig. 2C).
possible. Candidate d–t sets for which one or more line segment
This relationship is similar to that used to refine radiocarbon ages,
slopes b0 are rejected because they require reverse shear sense for
where the relative stratigraphic positions of samples are used to
a fault. For datasets with overlapping d–t envelopes that could not
reduce the range of the probability density functions for the calibrated
easily be trimmed (step 2), invalid locations within the d–t
sample ages (e.g., Biasi et al., 2002; Bronk Ramsey, 1995, 2001).
envelopes are identified by this analysis because they result in
If slip histories cannot have negative slopes, then each d–t envelope
negatively sloping slip histories.
has two fields through which the fault-slip history cannot pass (Fig. 2C).
5. Iterate—Steps 3 and 4 are repeated until a finite number of slip
These fields originate from the upper-left and lower-right corners of the
histories are accepted. The analyses presented here use 1000
envelope, which are the points defined by the maximum offset with the
histories because testing indicated larger numbers of accepted
minimum age and the minimum offset with the maximum age. When
histories do not significantly change the solution.
multiple d–t boxes are considered together, these fields can overlap
with the d–t envelopes from other markers and thereby trim the Consideration of the origin in the slip history model is not
envelope. By reducing the d–t envelopes, such trimming tightens the straightforward because at present day, a fault is at an unknown point
bounds on the range of d–t space through which the slip histories are within the earthquake cycle (e.g., Weldon and Sieh, 1985). We make
permitted to pass. This space is shown as the “conceptually permitted the simplifying assumption that the origin can be treated as data point
slip history area” on Figure 2C. Note that this trimming process also in the slip history simulations (d = 0 and t = 0 for each simulation)
provides a check on the internal consistency of a set of d–t envelopes, because doing so: 1) is consistent with the observation of zero offset
because incompatible data points will have no area of overlapping d–t at present day; 2) anchors the average slip-rate line near the origin; 3)
space. So this check also serves to test the assumption that a set of d–t removes the need for additional inputs (predicted time to next
data shares a common slip history. earthquake and anticipated future displacement); and 4) does not
Our ultimate goal is to test for changes in slip rate. Thus, we further significantly influence the model results. In more detail with regard to
limit the area of d–t space through which the modeled slip histories the slope, using the origin as a data point is necessary for defining the
 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64 55

Fig. 2. Explanation of the Monte Carlo slip history analysis method applied to a synthetic dataset. (A) D–t envelopes representing constraints on displacement and age for three
faulted features, numbered in order of decreasing age. (B) Permissible (green line) and impermissible (red line) slip histories through the d–t envelopes. The permissible slip history
represents a scenario in which features 1 and 2 experienced the same magnitude of total displacement, with lesser displacement recorded by feature 3. In contrast, the impermissible
slip history records a scenario in which feature 2 has experienced a larger magnitude of cumulative displacement than has the older feature 1. The implication for this type of
relationship is that the fault experienced reverse shear during the interval between deposition of features 1 and 2, which is geologically implausible. (C) Trimmed slip history space
that prevents the occurrence of negative-slope slip histories, conceptually permitted slip history area, and trimmed d–t envelopes. (D) Result from Monte Carlo model of 1000 slip
histories through the synthetic dataset. (E) A geometric mean linear least squares regression and 68.27% and 95.45% confidence bounds about the median (black lines) is calculated
from the slip histories, which provides a measure of the average slip rate over the duration of fault history recorded by the markers. In addition, a median slip history and the
associated nonparametric 68% and 95%-confidence envelopes are calculated, which provide a means to discriminate intervals of secular variation in slip for a given dataset. The slip
histories are only permitted to span the model slip history area, the bounds of which are defined by the range of sloping lines that can connect the d–t envelopes.

slope for slip histories where d–t data are clustered over limited 2.4. Statistical analysis
displacement and time intervals (e.g., many measurements between
50–60 m offset and 5–6 ka) and lack sufficient spread to define a We analyze the suite of accepted model slip histories in two ways.
realistic average slip rate. We also considered the possibility of 1. Average slip rate—In this case we explicitly assume uniform slip rate
treating the origin as a d–t envelope, the bounds of which would be and compute a linear regression through the modeled points that
defined by the recurrence interval and average slip-per-event define individual slip histories to yield an average slip rate with its
associated with surface-rupturing earthquakes (Weldon and Sieh, associated uncertainty. Specifically, a linear slope for each accepted
1985). This treatment would add a d–t envelope in negative d and t slip history is measured using a geometric mean linear least squares
space (i.e., the future) to account for potential future slip that is stored regression (Ricker, 1973), also known as a reduced major axis
as potential energy, which will be released in the next earthquake. For regression (Kermack and Haldane, 1950), by fitting a line for each
a major continental strike-slip fault (e.g., the San Andreas Fault), the population of the simulated d–t points that define a slip history.
box would have bounds on the order of 1–10 m in displacement and We use this regression technique because misfit is minimized in both
0.1–1.0 kyr in time (e.g., Mason and Little, 2006b; Washburn et al., the x and y dimensions, rather than only minimizing misfit in the
2001; Weldon et al., 2004). We chose not add such a d–t envelope y-dimension, as is accomplished with a standard least squares linear
because it requires additional knowledge about the detailed rupture regression. To determine the average slip rate and associated
behavior of the fault, and more importantly, it is too small to have a uncertainty, we calculate the median of the population of individual
significant effect on slip histories derived from d–t envelopes slope determinations (i.e., individual slip rates) and report the 68.27%
spanning timescales ≥ 10 kyr, which are the focus here. and 95.45% confidence bounds about the median (Fig. 2E). This system
56 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64

of reporting adheres to the protocol described by Zechar and Frankel d–t measurements for a suite of markers. In contrast, the other
(2009). Note that the line representing the linear slip rate is not forced approaches (e.g., Bennett, 2007; Bird, 2007) use a derivative of those
through the origin. Instead, treatment of the origin in the simulation as raw data (e.g., inferred instantaneous slip-rate versus time or a single
a data point (d= 0 and t =0 for each simulation) anchors the average rate with uncertainty), which limits the ability to graphically examine
slip-rate line near the origin, but not directly at the origin because we the influence that an individual datum exerts in controlling an
don't know that the origin represents the “best” fit with regard to inferred slip-rate record. Second, the Monte Carlo results communi-
future fault behavior in the next earthquake. For example, if the linear cate which parts of the slip history are tightly constrained by
fit were forced through the origin and then there was an earthquake observational data and which parts are not. This is difficult to assess
tomorrow, the solution would be poor. For situations where the best- using the methods presented by either Bennett (2007) or Bird (2007).
fit slip-rate line is very far from the origin, the result may indicate that Third, the Monte Carlo approach quantifies the history by yielding
a fault that has not experienced a uniform slip rate. both mean slip history and an associated uncertainty envelope. Bird's
2. Slip history—In this case we remove the assumption of uniform slip (2007) approach only leads to a graph of the PDF for the long-term
rate and compute a best-fit slip history. In particular we quantify rate. Fourth, the Monte Carlo approach implements knowledge that
the distribution of the accepted slip histories by contouring the d–t points are often not isolated data by recognizing that younger
model histories to compute a median history and nonparametric features must have experienced displacements that are equal-to-or-
68%- and 95%-confidence envelopes from this distribution less-than older features in a suite of faulted features, and then using
(Fig. 2E). The slip history contouring is accomplished by discretiz- this relationship to reduce the uncertainties on individual data points.
ing each slip history into horizontal time intervals that are much Fifth, the Monte Carlo approach allows a check of internal consis-
less than the width of the entire record (for the case studies below, tency of a population of d–t points (i.e., some pairs of boxes are
we use an interval of 100 years). The median, 68% and 95%- fundamentally incompatible with one another). This concept is
confidence bounds are defined through analysis of the population illustrated in Figure 2 and explained in detail below. And sixth, our
of slip histories by counting vertically through each column at the approach encourages users to review and critically evaluate input
specified horizontal intervals. Importantly, by defining the range of data because of the need to define bounding boxes, whereas users of
possible offset-age combinations for a fault-slip history, this second Bird's (2007) approach may input data with the expectation that
approach can be used to test for secular variation in slip. Such outlier will be determined using statistical methods. Furthermore,
variations are quantified by applying a piecewise version of the Bird's approach relies on sufficiently large datasets that outliers
linear regression measurement technique described above over (displacement and age) can be identified via statistical analysis. But as
intervals between two or more d–t envelopes. Bird (2007) notes, a very small number of faults in the U.S. can be
described by this analysis because the datasets are too sparse.
3. Comparison to other approaches
4. Treatment of uncertainties
A handful of studies have sought to use populations of morpho-
chronologic observations to quantify temporal variation in slip using A number of factors contribute to the range of values permitted by
three different approaches. In the first, which we follow here, offset the minimum and maximum bounds on the offset and age for a
and age constraints from faulted geologic markers are analyzed on given marker and thus the dimensions of a d–t envelope. The most
displacement–time diagrams (e.g., Fig. 1), so that slope variations significant uncertainties result from projection of landforms into a
between measurements can be used to identify changes in slip rate fault zone (Mason and Little, 2006b); the difficulty of directly dating
over time (e.g., Friedrich et al., 2003; Knuepfer, 1992; Wallace, 1987; some landforms, resulting in the need to bracket their ages using
Weldon and Sieh, 1985; Weldon et al., 1996). A second approach samples from adjacent dateable units (e.g. Cowgill et al., 2009;
integrates offset and age data with a smoothing algorithm to recover Harkins and Kirby, 2008); accounting for potential cross-fault
instantaneous slip rate as a function of time (e.g., Bennett, 2007; diachroneity of the correlated landform (Gold et al., 2009; Harkins
Bennett et al., 2004), and the third evaluates the extent to which a and Kirby, 2008); accounting for inheritance and detrital age signals
single slip rate statistically satisfies data from a population of faulted (Anderson et al., 1996; Repka et al., 1997; Rymer et al., 2002; Scharer
landforms (Bird, 2007). In this latter approach, when a single rate fails et al., 2007); and measurement and calibration errors associated with
to satisfactorily explain the data, the population either contains Quaternary geochronometers such as 14C, TCN, TSL, OSL, and U-series
outlier data or records secular variation in slip. dating techniques (e.g., Gosse and Phillips, 2001; Reimer et al., 2004).
Each of these approaches is well suited for particular applications. Because of our focus on bracketing the age and offset measure-
For example, a goal of the method proposed by Bird (2007) is to ments, we conservatively assume uniform probability within the d–t
characterize the average slip rate, while accounting for the earthquake envelope between the bracketing measurements. Specifically, the d–t
cycle. Bird's method has the advantages that it can be applied generally envelopes are boxcars, the edges of which are defined by the
(i.e., from a few earthquakes to all of the slip-rate data from a fault) and minimum and maximum bounds at the 2-σ confidence level on the
that it can account for a wide variety of different uncertainty models offset (d) and age (t) measurements for a given marker (Fig. 1A) (e.g.,
(e.g., Gaussian, boxcar, PDF, etc.). The method presented by Bennett Knuepfer, 1992). While more complex expressions of uncertainty are
(2007) is well suited for comparing geodetic, paleoseismic, and geologic possible (Bird, 2007; Zechar and Frankel, 2009), here we employ
rates. But while both the methods presented by Bennett (2007) and Bird boxcars because our main goal is to develop the Monte Carlo approach
(2007) provide a means for quantitatively comparing rates measured and we recognize that the method for simulating slip histories can be
using different methods and sampling different time scales, and thus can generalized for whatever uncertainty distribution is most appropriate
reveal time-varying rates, they reveal only limited information about for a given set of data. We also use boxcars because the uncertainties
the internal structure of the slip history (i.e., when rates changed, how on the individual age or offset measurements tend to be much smaller
long they lasted, and the data points that capture that event). than size of the bracket, which results from uncertainties in the site
Here we have built upon the first approach because it offers several reconstruction, such as how a landform projects into the fault or how
advantages for testing for temporal variability in rate along a single the geochronologic data relate to the true age of the marker.
fault and for determining the ages, magnitudes, and durations of such As the following demonstrates, use of boxcar uncertainties is
rate changes if they are present. First, it yields an estimate of the particularly appropriate for the relatively large brackets that typically
history of fault slip. This result is presented as a graphical display characterize age-offset measurements from laterally faulted fluvial
of the full range of possible offset-rate histories permitted by the terrace risers. Such risers are escarpments separating upper and lower
 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64 57

fluvial terraces and they are important due to the frequency with Figure 4A, which shows a displacement–time plot with a d–t envelope
which they are used to determine rates of strike-slip faulting (e.g., for a single marker with uncertainties of ±Δd and ±Δt. For the case
Cowgill, 2007; Harkins and Kirby, 2008; Knuepfer, 1992; Lensen, where the uncertainties are derived from bracketing the age and
1968; Mason et al., 2006; Mériaux et al., 2004, 2005; Van der Woerd offset values, as opposed to being Gaussian or some other form, then
et al., 1998; Weldon and Sieh, 1985). Because such risers are generally the data permit maximum (Rx) and minimum (Rn) slip rates that are
difficult to date directly, the ages of the flanking terraces are typically given by the slopes of the lines passing through the origin and the
used either as a proxy for, or to bracket, the age of riser formation (e.g., upper-left and lower-right corners of the d–t boxcar, respectively.
Cowgill, 2007; Harkins and Kirby, 2008; Knuepfer, 1992; Lensen, More specifically, Rx = (d + Δd) / (t − Δt) and Rn = (d − Δd) / (t + Δt).
1968; Mason et al., 2006; Mériaux et al., 2004, 2005; Van der Woerd Thus, the uncertainties set a minimum level at which different slip
et al., 1998; Weldon and Sieh, 1985). Figure 3 illustrates how upper/ rates can be resolved using a single marker. To quantify this level, we
older and lower/younger terrace abandonment ages (depicted as define a sensitivity threshold (T) as T = Rx/Rn so that T ≥ 1. Using the
Gaussian curves) bracket the formation age of a faulted terrace riser. relative uncertainties in offset (ud = Δd/d) and age (ut = Δt/t), it can
We assume no knowledge regarding the timing of riser formation
between abandonment of the bracketing terraces. Thus, the age of the
riser is most accurately described by a PDF with uniform probability in A
the interval between the terrace abandonment ages, flanked by Rx
T=
Gaussian tails corresponding to the uncertainties in the bracketing age Rn
determinations (Fig. 3). Note that when the terrace age uncertainties
are small relative to the age difference between the bracketing
Rx

Displacement (m)
terraces, the PDF will approach the shape of a boxcar. Data from 42
faulted terrace riser along the Altyn Tagh, Kunlun, and Awatere faults
indicate that the boxcar captures all but 0.8% of the area under the PDF d
with Gaussian tails if the 2-σ age uncertainties are used to define the Rn
edges of the d–t envelope (Fig. 3). This exercise considers only the
uncertainty associated in the age-dimension of a displacement-time
history. However, displacement measurements are prone to similar
uncertainties, especially when cross-fault riser diachroneity is
accounted for (Gold et al., 2009; Harkins and Kirby, 2008). Thus, we
employ the most conservative approach in which we assume that any t
offset-age combination within the d–t boxcar is uniformly probable.
Any individual d–t envelope permits a range of rates, resulting in a
minimum rate difference that it can resolve. Consideration of the Age (ka)
precision (i.e., d–t envelope size) available from such single markers
0.4
indicates they provide limited resolution for distinguishing between
B

5.
0
anything but extreme differences in slip rate. Thus, it appears that

4.
5
multiple markers are required to reduce the uncertainty on average

4.
0
Relative uncertainty in displacement

slip rates and to derive slip histories from which variations in slip can

3.
0.3
be distinguished. To illustrate this point, in the following we

5
quantitatively evaluate the minimum difference in slip rate that can
3.
0
be resolved using a single landform given typical uncertainties in
morphochonologic data. The problem is illustrated graphically in
2.
5

0.2
2. 9
0
1.

Lower Terrace
for terrace ages
density function

1.
8
Normalized

1. 6
probability

Upper Terrace
7
1.

Riser
0.1
1.

Formation
5
1.
4
1.
3
1.
2
1.
1

Boxcar 0.0
density function

0.0 0.1 0.2 0.3 0.4

for riser age
Probability

Gaussian Relative uncertainty in age

Area of tails
not considered Fig. 4. Plots illustrating how uncertainties in morphochronologic data define a minimum
0.8 % difference in slip rate that can be resolved using a single marker. (A) Schematic
displacement–time plot. Gray box shows constraints from a single offset landform for
Age (kyr) which the age (t) and displacement (d) lie within uncertainties of± Δt and ± Δd,
respectively. In the case where the uncertainties are derived from bracketing (see
Fig. 3. Illustration of riser age-bracketing approach using boxcar and Gaussian Section 3), slopes Rx and Rn define the maximum and minimum slip rates that are
uncertainties. An upper- and lower-terrace-abandonment age brackets an interval permitted by the single marker. The ratio of these rates defines a sensitivity threshold (T),
during which the riser formed. In the boxcar treatment, the riser is assumed equally which quantifies the minimum difference in slip rate that can be resolved. (B) Plot
likely to have formed at any time in the interval between the 2-σ error bounds on the showing the dependence of the sensitivity threshold (T) on the relative uncertainties in
terrace abandonment ages (e.g., for a riser bound by terraces abandoned at 4.0 ± 0.1 ka offset (ud = Δd/d) and age (ut = Δt/t). Lines are contours of equal T value and range from
(2-σ) and 5.0 ± 0.1 ka, respectively, the boxcar spans the interval from 3.9 to 5.1 ka.). In T = 1.1 to T = 5 (i.e., uncertainties permit a 5-fold difference in the maximum and
the Gaussian treatment, the riser is assumed to have formed in the interval between the minimum rates permitted by the d–t envelope). Note change in contour interval at T = 2.0.
terrace abandonment ages, where the “tails” are represented by a Gaussian description Gray box delineates range of relative uncertainties typical for slip-rate studies, and
of the uncertainty associated with the terrace abandonment age. Analysis of 42 faulted indicates that d–t measurements from a single marker can only resolve between grossly
terrace risers from along the Altyn Tagh, Kunlun, and Awatere faults indicates that only different slip rates because the uncertainties are too large to yield useful T values (e.g.,
0.8% of the areas under the Gaussian tails are not considered in the boxcar treatment. T b 1.5).
58 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64

be shown that the sensitivity threshold is only a function of the single-event slip-distributions average to a roughly uniform cumula-
relative uncertainties: tive offset over a relatively small number of earthquakes (e.g., 2–3)
along the considered fault reach.
1 + ut + ud + ud ut We suspect it is reasonable to integrate d–t measurements from
T= :
1−ud −ut + ud ut single fault segments that are up to ~ 250 km long and lack splaying,
joining, or overlapping with other structures because far-field strain
To illustrate how these relative uncertainties affect the sensitivity rates should be uniform over the timescales we are considering (e.g.,
threshold, Figure 4B shows contours of equal T on a plot of ud versus Bird, 2007). For strike-slip faults, surface rupture lengths for large
ut. As expected, T increases with both increasing ud and ut. In addition, (Mw N 7) earthquakes typically range from 50 to 200 km (Wells and
the negative slopes of the contours indicate the uncertainties trade off Coppersmith, 1994). Although coseismic offset can vary significantly
with one another, so that large uncertainty in one measurement can along the rupture (e.g., Xu et al., 2006), they appear to remain grossly
be compensated by small uncertainty in the other. But most constant over lengths of 35–132 km (Haeussler et al., 2004; Mason
importantly, Figure 4B also indicates that d–t measurements from a and Little, 2006a; Sieh, 1978c; Xu et al., 2006). In addition, because
single marker are unlikely to discriminate between end-member slip morphochronologic data generally record multiple events, they may
rates of interest because the uncertainties required to obtain relatively average out the slip variations that occur between individual ruptures.
low threshold values (e.g., T b 1.5) require uncertainties that are However, integration of measurements should be undertaken with
generally difficult to obtain. For example, relative uncertainties in age caution along fault reaches that are punctuated by geometric
and offset are both typically on the order 0.1 to 0.2 (i.e., 10 to 20%), complexities such as restraining and releasing bends, or intersections
yielding threshold values of 1.5 to 2.2 as shown by the gray box on with other faults. Both rupture length and coseismic slip are strongly
Figure 4B. controlled by the presence of such complexities (e.g., Haeussler et al.,
2004; Wesnousky, 2006). Finally, we advocate undertaking a
5. Integration of measurements along fault strike sensitivity analysis for populations of data integrated over large
distances along fault strike that indicate slip-rate variability. For these
The analysis presented in Figure 4 illustrates that multiple d–t situations, we recommend applying the Monte Carlo analysis to
measurements are needed to reduce uncertainty in measurements of subsets of d–t datasets as a function of position along strike to test for
fault slip, but an analysis of the number of faulted landforms reported significant slip history variability. If this analysis does not reveal slip-
from individual slip-rate sites that have been used to determine rates rate variability, then integration of measurements is reasonable. If
of strike-slip faulting within the Indo-Asian collision indicates that the there is variability, it may result from either along strike or temporal
average number of faulted markers at a single site is 1.9 (Table 1). variability. These scenarios can be differentiated by evaluating if the
Consideration of b2 markers is insufficient to resolve secular variation slip rate and slip history results vary as a function of position along
in slip. Thus, obtaining a sufficient number of d–t envelopes to fault strike. We recognize that integrating displacement–age data
measure the constancy of slip rate is a potential challenge to our from multiple sites along fault strike is not always a reasonable
approach. One way to increase the number of d–t envelopes is to assumption. But we note that the legitimacy of this assumption is a
compile data for a single fault from multiple sites along strike (e.g., separate question from the robustness of the modeling approach; i.e.,
Bird, 2007). However, the approach presented above assumes that all the problem is not with the model itself but rather the conditions
markers used in the model experienced a common history of fault slip. under which it can be applied accurately.
Thus, an essential question is: over what fault length do sites
experience common slip histories? The slip history analysis focuses 6. Case studies
on offsets that have accumulated over multiple earthquakes. Thus, we
expect to be able to compile d–t data from along fault strike as long as To evaluate the extent to which the Monte Carlo model yields
millennial offset gradients are small (e.g., b10%) and any variations in average slip rates that reproduce previously published values, here we

Table 1
Compilation of the number of faulted geomorphic markers preserved at morphochronologic study sites along several first-order strike-slip faults within the Indo-Asian collision.

Fault Site Reference Number of markersa

Kunlun Fault Site 1 van der Woerd et al., 1998, Geology; van der Woerd et al., 2002, GJI 3
Kunlun Fault Site 2 van der Woerd et al., 2000, GRL; van der Woerd et al., 2002, GJI 2
Kunlun Fault Site 3 van der Woerd et al., 2002, GJI 2
Kunlun Fault Ninzha He/Site 4 van der Woerd et al., 2002, GJI 2
Kunlun Fault Xiadawu/Site 5 van der Woerd et al., 2002, GJI 3
Kunlun Fault Hongshui Gou Li et al., 2005, EPSL 3
Kunlun Fault Site 1, Ken Mu Da Kirby et al., 2007, Tectonics 3
Kunlun Fault Site 2, Quor Goth Qu Kirby et al., 2007, Tectonics 1
Kunlun Fault Deng Qin Harkins and Kirby, 2008, GRL 2
Karakorum Manikala Chevalier et al., 2005, Science 2
Karakorum TangtseValley Brown et al., 2002, JGR 1
Xianshuihe Fault Gelu Allen et al., 1991, GSAB 1
Xianshuihe Fault Laohekou Allen et al., 1991, GSAB 1
Xianshuihe Fault Longdenba Allen et al., 1991, GSAB 2
Haiyuan Fault Site 1/Majia Wan Lasserre et al., 1999, JGR 1
Haiyuan Fault Site 2/Xuanma Wan Lasserre et al., 1999, JGR 1
Haiyuan Fault Xiying He Lasserre et al., 2002, JGR 1
Altyn Tagh Faultb Cherchen He Mériaux et al., 2004, JGR; Cowgill, 2007, EPSL 4
Altyn Tagh Fault Tuzidun Gold et al., 2009, JGR 1
Altyn Tagh Fault Yuemake Cowgill et al., 2009, Geology 1
a
Coseismic displacements from the most recent and penultimate events are not included in this compilation.
b
We did not include the number of faulted geomorphic features reported by Mériaux et al. (2004) at Sulamu Tagh nor features reported by Mériaux et al. (2005) at Aksay along
the Altyn Tagh Fault because the displacement-age relationships in these studies are the subject of ongoing debate (e.g., Zhang et al., 2007).
 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64 59

apply the approach to existing morphochronologic datasets from the (92.2–94.7°E longitude), four study sites spanning a 230-km-long
Kunlun and Awatere faults. We also examine whether the reported reach of fault are well documented. From west to east these are
data are consistent with uniform or secularly varying slip. We have Hongshui Gou (Li et al., 2005), Site 2 (Van der Woerd et al., 2000,
selected the Kunlun fault because existing datasets have been 2002), Site 1 (Van der Woerd et al., 1998, 2002), and Site 3 (Van der
interpreted to suggest that the central and eastern segments of this Woerd et al., 2002) (Fig. 5). Offsets at these sites range from 3 to
fault system slip at uniform but different rates. In addition, these data 110 m and all dated surfaces are ≤ 14.9 ka. Along the eastern segment
provide an opportunity to evaluate the effect of integrating d–t of the Kunlun fault, five laterally faulted terrace risers ranging in age
constraints from along fault strike. For our second case study, we from 3.3 to 28.8 ka are offset 31 to 115 m at the Deng Qin (Harkins and
selected the Saxton River site along the Awatere fault, because it Kirby, 2008) and Ken Mu Da sites (Harkins and Kirby, 2008; Kirby
provides d–t constraints for six faulted late Quaternary terrace risers, et al., 2007), which are separated by b3 km along fault strike (101.4°E
which is a large number of markers for a single site (Table 1). Also, longitude).
data from this site have been contrastingly interpreted to reflect both
temporally uniform (Mason et al., 2006) and temporally varying slip
(Knuepfer, 1992) and the purpose of our analysis is to examine the 6.1.2. Reanalysis of morphochronologic data from the Kunlun fault
extent to which the new slip history analysis method can discriminate While a total of 10 faulted terrace risers are reported for the four
between these two scenarios. sites along the central Kunlun fault, only 7 are flanked by upper and
lower terraces that have both been dated and thus have a bracketed
age. We consider only these 7 risers. For the eastern Kunlun fault, the
6.1. Case study #1: The Kunlun fault
original slip-rate calculations were made using the bracketing
approach so we consider all 5 features. The d–t envelopes defined
The active, left-lateral Kunlun fault extends ~1600 km, between 87°
by the faulted markers are presented in Figure 6 and the offset and age
and 105°E longitude across northeastern Tibet (Tapponnier and Molnar,
data are provided in Table DR1 in the supplemental materials.
1977) (Fig. 5). The slip rate along this first-order fault system plays a
The analysis of the central Kunlun fault yields a uniform slip rate of
central role in resolving the importance of strike-slip faulting in
8.7 + 3.6/−2.1 mm/yr (68.27% confidence) (Fig. 6A, Table 2), which
accommodating Indo-Asian convergence (e.g., Tapponnier et al.,
overlaps the previously reported slip-rate measurements of 10.0 ±
2001b). Furthermore, the fault is adjacent to several additional first-
1.5 mm/yr (Li et al., 2005) to 11.7 ± 1.5 mm/yr (Van der Woerd et al.,
order strike-slip structures including the Altyn Tagh, Haiyuan, and
2002). The rate we find here is the first reported for the central Kunlun
Xianshuihe faults. Thus, inter-comparison of slip records between
fault to use riser ages that are bracketed using upper- and lower
these faults can be used to determine the extent to which they share
terrace ages, which contributes to the larger uncertainty relative to
co-dependent slip histories. Late Quaternary slip-rate studies in the
that previously reported. In addition we find no evidence of variable
central portion of the Kunlun fault yield rates of 10.0 ± 1.5 mm/yr (Li et
slip rate because both the 68.27% and 95.45% error bounds on the
al., 2005) to 11.7 ± 1.5 mm/yr (Van der Woerd et al., 2002). Slip rates
linear regression defining the uniform slip rate entirely contain the
appear to decrease eastwards along strike to 5 ± 2 mm/yr (Harkins and
best-fit history (median) derived from contouring as well as the 68%-
Kirby, 2008; Kirby et al., 2007) along the eastern segment of the fault,
and 95%-confidence envelope (Fig. 6A). For the eastern Kunlun fault,
before dying out completely (Kirby et al., 2007). These late Quaternary
our analysis yields a linear slip rate of 5.1 + 1.6/–1.2 mm/yr (68.27%),
slip rates are consistent with the geodetically determined rates of 8–
which again overlaps within error with the previously reported slip
12 mm/yr (Chen et al., 2004; Kirby et al., 2007; Zhang et al., 2004) and
rate of 5 ± 2 mm/yr (Fig. 6B, Table 2) (Harkins and Kirby, 2008; Kirby
3–5 mm/yr (Kirby et al., 2007) for the central and eastern portions of the
et al., 2007). While there is some scatter in the range of possible slip
fault, respectively.
histories through the data, the best-fit contoured slip history
(median) falls entirely within the 68.27% error bounds defined for
6.1.1. Morphochronologic data from the Kunlun fault the linear rate. There is more disparity between the 95.45% linear
A number of morphochronologic studies have been undertaken error bounds and the shape of the 95%-confidence envelopes defined
along the length of the Kunlun fault, primarily using faulted terrace from the eastern Kunlun analysis, though in general those regions
risers (Harkins and Kirby, 2008; Kirby et al., 2007; Li et al., 2005; Van of poorest confidence defined by the 95% envelope correspond to
der Woerd et al., 1998, 2000, 2002). In the central portion of the fault the 95.45% linear error bound. In conclusion, for both the central and

Fig. 5. Overview of the first-order strike-slip faults within the Indo-Asian collision (left) and geologic map of the main trace of the active, left-lateral Kunlun fault (right) and the
location of morphochronologic slip-rate sites along the central (black) and eastern (green) fault reaches. Abbreviations are as follows: KKF, Karakorum Fault; GLF, Gozha-Longmu Co;
ATF, Altyn Tagh Fault; KF, Kunlun fault; HF, Haiyuan Fault; XSF, Xianshuihe Fault; JL, Jiali Fault; MFT, Main Frontal Thrust. Fault locations simplified from Taylor and Yin (2009).
60 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64

Fig. 6. Slip history analysis applied to morphochronologic data from the Kunlun fault. (A) Central and (B) Eastern Kunlun fault morphochronologic d–t envelopes (left), slip history
paths (central) and slip history analysis results (right). D–t envelopes are defined from published morphochronologic datasets of faulted terrace riser along the central (Li et al.,
2005; Van der Woerd et al., 1998, 2000, 2002) and eastern Kunlun fault (Harkins and Kirby, 2008; Kirby et al., 2007). This analysis represents the first time that these markers from
the central Kunlun fault have been considered using a upper and lower terrace age bracketing approach (Cowgill, 2007). In general, the average slip rate and slip history yield
compatible results, suggesting that the respective segments of the Kunlun fault have experienced uniform slip rates over the late Quaternary.

eastern reaches of the Kunlun fault, the slip history analysis (1) determined right-lateral motion across the entire Marlborough fault
reproduces a linear slip rate that closely matches previously published system is ~45 mm/yr (Knuepfer, 1992 and references therein), which is
slip rates and (2) produces a fault-slip history that closely match the broadly compatible with the relative motion between the Pacific and
linear slip rates. These results are consistent with previous inter- Australian plates of ~39 mm/yr (DeMets et al., 1990, 1994).
pretations (Kirby et al., 2007; Van der Woerd et al., 2002) that the The Saxton River site is located at the intersection of the Saxton River
Kunlun fault has maintained a uniform left-lateral slip rate through with the eastern segment of the Awatere fault. This site is an ideal
the late Quaternary of ~10 mm/yr and ~ 5 mm/yr along the central candidate for applying the slip history analysis developed above because
and eastern segments, respectively. it preserves a relatively large number (5–7) of faulted landforms, the d–t
constraints for which have been contrastingly interpreted to indicate
6.2. Case study #2: The Awatere fault, NZ that the fault has either substantially decelerated in the late Holocene
(Knuepfer, 1992) or maintained a uniform slip rate over the duration of
The active, right-lateral Awatere fault is part of the greater Alpine the late Quaternary (Mason et al., 2006). Thus, reanalysis of the d–t data
shear system, which forms the boundary between the Australian and may help clarify this dispute.
Pacific plates. It is one of four principal fault strands within the 150 km
wide Marlborough strike-slip system on the northeastern part of the 6.2.1. Previous work at the Saxton River site
South Island of New Zealand (e.g., Little et al., 1998) (Fig. 7A). From north Knuepfer (1992) combined offset measurements made by Lensen
to south, these are the Wairau, Awatere, Clarence, and Hope faults. The (1973) with terrace age data determined from calibrated pebble
NE-striking Awatere fault has a total length of 175 km. The geodetically weathering rind thicknesses and soil morphology (Knuepfer, 1988) to

Table 2
Slip-rate results from Monte Carlo analysis, assuming uniform slip-rate solution.

Fault Time interval Median slip rate 68.27% interval 95.45% interval
(ka) (mm/yr)
(+) (−) (+) (−)

Kunlun Fault, central segment (92.2–94.7°E) 14.9 to present 8.7 3.6 2.1 7.2 2.9
Kunlun Fault, eastern segment (101.4°E) 28.8 to present 5.1 1.6 1.2 3.2 2.0
Awatere Fault, Saxton Rive site ~14.5 to present 5.6 1.0 0.8 1.9 1.6
Awatere Fault, Saxton Rive site (Early Holocene) ~14.5 to 3.5–5.8 6.2 1.6 1.4 3.5 2.3
Awatere Fault, Saxton River site (Late Holocene) 3.5–5.8 to present 2.8 1.0 0.6 2.5 1.0
 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64 61

Fig. 7. Analysis of the morphochronologic data from faulted terrace risers at the Saxton River site along the Awatere fault. (A) Site location map along the Awatere fault, and adjacent
right-lateral fault systems (fault positions as reported on the New Zealand Active Faults Database: http://data.gns.cri.nz/af/index.jsp, accessed May 30th, 2009). (B) Comparison of
d–t envelopes derived from faulted terrace risers at the Saxton River site as originally presented by Knuepfer (1992), Mason et al. (2006), and as they are defined using bracketed
upper and lower terrace abandonment age approach. Results from slip history analysis including (C) model path and points, (D) uniform average slip rate, (E) fault-slip history, and
(F) two discrete slip-rate intervals in the late Quaternary. Importantly, panel (F) depicts a statistically significant factor of 2.2 slip-rate deceleration in the mid-Holocene.
62 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64

conclude that the site records a factor of 2.5 decrease in slip rate, from constraints on offset and age for multiple faulted landforms. These
9.4 mm/yr from 10 to 4 ka to a rate of 3.8 mm/yr from 4 ka to present. constraints can be represented as d–t envelopes on plots of
The site was later reevaluated by Mason et al. (2006), who identified displacement versus time. Importantly, uncertainties associated
two additional piercing points, used GPS-surveys to report new offset with individual markers are reduced by analyzing the relationships
measurements, and obtained new OSL ages to date the two oldest between faulted landforms with common histories. For example,
terraces. From these new data Mason et al. (2006) concluded that the older features must always have cumulative offset that is greater-
fault has slipped at a uniform rate of 5.6 ± 0.8 mm/yr over the past than-or-equal-to the cumulative offset recorded by younger features,
15 ka, and has not experienced significant secular variations in slip assuming constant shear sense. Our Monte Carlo modeling approach
over the late Quaternary. provides a means to measure the range of slip histories permitted by
these trimmed d–t envelopes, with the requirement that each history
6.2.2. Reanalysis of morphochronologic data from the Saxton River site must maintain a positive slope. Analysis of the distribution of the slip
We modeled slip histories based on the data reported by Knuepfer histories yields both a measure of the average slip rate for a
(1992) and Mason et al. (2006), after bracketing the ages of the offset population of d–t envelopes and the slip history to distinguish secular
terraces with the upper and lower terrace ages. We used the offset variation in slip. We have restricted our analysis to boxcar age and
values reported by Mason et al. (2006). For the age values, we used displacement uncertainties, but future efforts can expand upon this
OSL ages (Mason et al., 2006) for the oldest terraces, weathering rind approach using alternative forms of uncertainty (e.g., Gaussian,
ages for the intermediate terraces (Knuepfer, 1988, 1992), and a trapezoid, PDF, etc.).
combined terrace rind and 14C age for the youngest terrace (Knuepfer, This analysis is illustrated using published morphochronologic
1988, 1992; McCalpin, 1996a) (Table DR2). A more detailed discus- data from laterally faulted terrace risers along the Kunlun and
sion of our treatment of these data is presented in the supplemental Awatere faults. The time span over which the laterally faulted terrace
materials, and includes a discussion of how the OSL and rind ages risers were interpreted to have formed was bracketed by adjacent
compare for common terraces. From these constraints, we con- upper and lower terrace abandonment ages. This is the first bracketed
structed d–t envelopes (Fig. 7B), following the same approach treatment of data from the central Kunlun fault and Saxton River site
adopted by Knuepfer (1992) and Mason et al. (2006), although along the Awatere fault.
these authors used only the age uncertainties as determined for the Previous work along the Kunlun fault suggests that the central and
lower surface terrace ages in contrast to the larger age uncertainties eastern reaches of this fault have slipped at uniform rates through the
that stem from the bracketing approach we employ here. Figure 7B late Quaternary, which provides a benchmark test for this analytical
shows a comparison of the d–t envelopes as originally presented by method. Application of this slip history analysis to the Kunlun fault
Knuepfer (1992) and Mason et al. (2006) with the ones we use here. datasets yields linear slip-rate measurements of 8.7 + 3.6/−2.1 mm/yr
Results of the slip history analysis are summarized in Figure 7C–F. A and 5.1 + 1.6/−1.2 mm/yr (68.27% confidence), for the central and
linear regression applied to the modeled slip histories yields a slip rate eastern reaches of the Kunlun fault, respectively. Similarly, the slip
5.6 + 1.0/−0.8 mm/yr (68.27% confidence) from 14.5 ± 1.5 ka to pres- histories determined from the model are nearly indistinguishable from
ent (Fig. 7D, Table 2). However, removal of the uniform-rate assumption the linear average slip rates for both fault reaches, which is interpreted
suggests that the fault decelerated in the mid-Holocene. This deceler- to reflect uniform slip rate in the late Quaternary.
ation is graphically depicted by the shallow slope defined for the median Previous work at the Saxton River site along the Awatere fault on the
slip history and 68% and 95% envelopes since formation of the T3/T4–T5 South Island of New Zealand has produced conflicting reports regarding
riser to the present, compared to the earlier interval (Fig. 7E). The the temporal variation of slip rate in the late Quaternary. Application of
deceleration is quantified via a piecewise linear regression applied to the the slip history analysis to this dataset suggests that while the d–t
modeled slip histories, which records a slip rate of 6.2 + 1.6/−1.4 mm/ constraints can be satisfied via a uniform slip rate, two intervals of
yr (68.27% confidence) from 14.5 ± 1.5 ka to 3.5–5.8 ka and a slip rate of statistically distinct slip rate are identified: from 14.5 to 3.5−5.8 ka an
2.8 + 1.0/–0.6 mm/yr (68.27% confidence) from 3.5–5.8 ka to present average slip rate of 6.2 + 1.6/−1.4 mm/yr and a lower slip rate of 2.8 +
(Fig. 7F, Table 2). The two slip-rate values differ by a factor of 2.2 and are 1.0/−0.6 mm/yr (68.27% confidence) from 3.5–5.8 ka to present. This
statistically distinct at the 68.27% confidence level. Importantly, the two analysis confirms previous reports of slip-rate deceleration along the
slip-rate values are not distinguishable at the 95.45% confidence level Awatere fault in the mid-Holocene (Knuepfer, 1992). In summary, the
(Table 2). Furthermore, we note that the uniform-rate solution (Fig. 7D) results from this paper demonstrate that by relating age and offset
does not pass near the origin, which we interpret to support our constraints from faulted markers and quantifying the distribution of slip
conclusion that the Awatere fault has not slipped uniformly. In contrast, histories, temporally constant and secularly varying fault-slip scenarios
the piecewise linear regression in the deceleration model (Fig. 7E) can be distinguished in “deep time.”
passes quite near the origin, supporting our interpretation that this
model better satisfies this d–t dataset. Acknowledgements
Application of the bracketed upper and lower terrace age approach
to the data from the Saxton River provides the most conservative d–t This work was supported by NSF grant EAR-0610107 from the
envelope bounds on the Awatere fault-slip history. Thus, it follows Tectonics Program and the East Asia and Pacific Program in the NSF
that the slip history defined from this treatment provides the most Office of International Science and Engineering. We thank J. D. Zechar,
conservative bounds on the range of possible slip histories con- P. Bird, and R. Weldon for comments that significantly improved the
strained for these data. Given the wide age bounds associated with the paper. We are grateful for the computer programming assistance
bracketing approach, a uniform slip rate can satisfy these data. But provided by S. Andrews and N. Conjeepuram and also for statistical
consideration of the fault-slip history shows a statistically significant consultation provided by N. Willits of the UC Davis Statistical
deceleration in slip rate in the mid-Holocene, consistent with the Laboratory. Conversations with D. Perkins, E. Burkett, and A. Forte
interpretation reported by Knuepfer (1992). contributed to the development of this analysis. An early version of
the manuscript benefited from reviews from P. Gold and A. Reuther.
7. Conclusions
Appendix A. Supplementary data
We have shown that evaluating the extent to which strain release
varies along a given patch of fault over the timescale of thousands to Supplementary data to this article can be found online at doi:10.1016/j.
tens of thousands of years can be accomplished by integrating epsl.2010.10.011.
 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64 63

References Kermack, K.A., Haldane, J.B.S., 1950. Organic correlation and allometry. Biometrika 37,
30–41.
Allen, C.R., Zhuoli, L., Hong, Q., Xueze, W., Huawei, Z., Weishi, H., 1991. Field study of a Kirby, E., Harkins, N., Wang, E., Shi, X., Fan, C., Burbank, D., 2007. Slip rate gradients
highly active fault zone: the Xianshuihe fault of southwestern China. Geol. Soc. Am. along the eastern Kunlun fault. Tectonics 26, TC2010.
Bull. 103, 1178–1199. Knuepfer, P.L.K., 1987. Changes in Holocene slip rates in strike-slip environments. In:
Ambraseys, N., 1971. Value of historical record of earthquakes. Nature 232, 375–379. Crone, A.J., Omdahl, E.M. (Eds.), Directions in Paleoseismology, U.S. Geological
Anderson, R.S., Repka, J.L., Dick, G.S., 1996. Explicit treatment of inheritance in dating Survey Open-File Report 87-673. Denver, CO, pp. 249–261.
depositional surfaces using in situ 10Be and 26Al. Geology 24, 47–51. Knuepfer, P.L.K., 1988. Estimating ages of Late Quaternary stream terraces from analysis
Bennett, R.A., 2007. Instantaneous slip rates from geology and geodesy. Geophys. J. Int. of weathering rinds and soils. Geol. Soc. Am. Bull. 100, 1224–1236.
169, 19. Knuepfer, P.L.K., 1992. Temporal variations in latest quaternary slip across the
Bennett, R.A., Friedrich, A.M., Furlong, K.P., 2004. Codependent histories of the San Australian–Pacific plate boundary, Northeastern South Island, New-Zealand.
Andreas and San Jacinto fault zones from inversion of fault displacement rates. Tectonics 11, 449–464.
Geology 32, 961–964. Lasserre, C., Morel, P.-H., Gaudemer, Y., Tapponnier, P., ryerson, F.J., King, G.C.P.,
Ben-Zion, Y., 2008. Collective behavior of earthquakes and faults: continuum-discrete Métivier, F., Kasser, M., Kashgarian, M., Liu, B., Lu, T., Yuan, D., 1999. Postglacial left
transitions, progressive evolutionary changes, and different dynamic regimes. Rev. slip rate and past occurrence of M ≥ 8 earthquakes on the western Haiyuan fault,
Geophys. 46, 70. Gansu, China. J. Geophys. Res. 104, 17633–17651.
Bevington, P.R., Robinson, D.K., 2003. Data Reduction and Error Analysis for the Physical Lasserre, C., Gaudemer, Y., Tapponnier, P., Meriaux, A.-S., Van der Woerd, J., Yuan, D.,
Sciences, 3rd. McGraw-Hill, New York. Ryerson, F.J., Finkel, R., Caffee, M., 2002. Fast late Pleistocene slip rate on the Leng
Biasi, G.P., Weldon, R.J., Fumal, T.E., Seitz, G.G., 2002. Paleoseismic event dating and the Long Ling segment of the Haiyuan fault, Qinghai, China. J. Geophys. Res. B Solid
conditional probability of large earthquakes on the southern San Andreas fault, Earth Planets. 107, 1–15.
California. Bull. Seismol. Soc. Am. 92, 2761–2781. Lave, J., Yule, J.D., Sapkota, S., Basant, K., Madden, C., Attal, M., Pandey, R., 2005. Evidence
Bird, P., 2007. Uncertainties in long-term geologic offset rates of faults: general for a Great Medieval Earthquake ( 1100 A.D.) in the Central Himalayas, Nepal.
principles illustrated with data from California and other western states. Geosphere Science 307, 1302–1305.
3, 577–595. Lensen, G.J., 1968. Analysis of progressive fault displacement during downcutting at the
Bronk Ramsey, C., 1995. Radiocarbon calibration and analysis of stratigraphy: the OxCal Branch River terraces ,South Island, New Zealand. Geol. Soc. Am. Bull. 79, 545–556.
program. Radiocarbon 37, 425–430. Lensen, G.J., 1973. Guidebook for excursion A10. Field-trip guide for Ninth Congress of
Bronk Ramsey, C., 2001. Development of the radiocarbon calibration program OxCal. International Union for Quaternary Research. University of Canturbury, Christch-
Radiocarbon 43, 355–363. urch, New Zealand, pp. 15–18.
Brown, E.T., Bendick, R., Bourlès, D.L., Gaur, V., Molnar, P., Raisbeck, G.M., Yiou, F., 2002. Li, H., Van der Woerd, J., Tapponnier, P., Klinger, Y., Xuexiang, Q., Jingsui, Y., Yintang, Z.,
Slip rates of the Karakorum fault, Ladakh, India, determined using cosmic ray 2005. Slip rate on the Kunlun fault at Hongshui Gou, and recurrence time of great
exposure dating of debris flows and moraines. Journal of Geophysical Research 107, events comparable to the 14/11/2001, Mw 7.9 Kokoxili earthquake. Earth Planet.
2192–2205. Sci. Lett. 237, 285–299.
Chen, Q., Freymueller, J., Wang, Q., Yang, Z., Xu, C., Liu, J., 2004. A deforming block model Little, T.A., Grapes, R., Berger, G.W., 1998. Late Quaternary strike slip on the eastern part
for the present-day tectonics of Tibet. J. Geophys. Res. 109. of the Awatere fault, South Island, New Zealand. Bull. Geol. Soc. Am. 110, 127–148.
Chéry, J., Vernant, P., 2006. Lithospheric elasticity promotes episodic fault activity. Earth Marco, S., Stein, M., Agnon, A., Ron, H., 1996. Long-term earthquake clustering: a
Planet. Sci. Lett. 243, 211–217. 50,000-year paleoseismic record in the Dead Sea Graben. J. Geophys. Res. 101,
Chevalier, M.-L., Ryerson, F.J., Tapponnier, P., Finkel, R., Van der Woerd, J., Haibing, L., 6179–6191.
Qing, L., 2005. Slip-rate measurements on the Karakorum fault may imply secular Mason, D.P.M., Little, T.A., 2006a. Refined slip distribution and moment magnitude of
variations in fault motion. Science 307, 411–414. the 1848 Marlborough earthquake, Awatere Fault, New Zealand. NZ J. Geol.
Cowgill, E., 2007. Impact of riser reconstructions on estimation of secular variation in Geophys. 49, 375–382.
rates of strike–slip faulting: revisiting the Cherchen River site along the Altyn Tagh Mason, D.P.M., Little, T.A., 2006b. Refinements to the paleoseismic chronology of the
Fault, NW China. Earth Planet. Sci. Lett. 254, 239–255. eastern Awatere Fault from trenches near Upcot Saddle, Marlborough, New
Cowgill, E., Gold, R.D., Chen, X., Wang, X.-F., Arrowsmith, J.R., Southon, J.R., 2009. Low Zealand. NZ J. Geol. Geophys. 49, 383–397.
Quaternary slip rate reconciles geodetic and geologic rates along the Altyn Tagh Mason, D.P.M., Little, T.A., Van Dissen, R.J., 2006. Rates of active faulting during late
fault, northwestern Tibet. Geology 37, 647–650. Quaternary fluvial terrace formation at Saxton River, Awatere fault, New Zealand.
DeMets, C., Gordon, R.G., Argus, D.F., Stein, S., 1990. Current plate motions. Geophys. J. Bull. Geol. Soc. Am. 118, 1431–1446.
Int. 101, 425–478. McCalpin, J.P., 1996a. Tectonic geomorphology and Holocene paleoseismicity of the
DeMets, C., Gordon, R.G., Argus, D.F., Stein, S., 1994. Effect of recent revisions to the Molesworth section of the Awatere Fault, South Island, New Zealand. NZ J. Geol.
geomagnetic reversal time scale on estimates of current plate motions. Geophys. Geophys. 39, 33–50.
Res. Lett. 21, 2191–2194. McCalpin, J.P., 1996b. Field techniques in paleoseismology. In: McCalpin, J.P. (Ed.),
Dolan, J.F., Bowman, D.D., Sammis, C.G., 2007. Long-range and long-term fault Paleoseismology, International Geophysics Series, pp. 33–84.
interactions in Southern California. Geology 35, 855–858. Melbourne, T.I., Szeliga, W.M., Miller, M.M., Santillan, V.M., 2005. Extent and duration of
Frankel, K.L., Dolan, J., Finkel, R., Owen, L.A., Hoeft, J.S., 2007. Spatial variations in slip the 2003 Cascadia slow earthquake. Geophys. Res. Lett. 32, 4.
rate along the Death Valley–Fish Lake Valley fault system determined from LiDAR Mériaux, A.-S., Ryerson, F.J., Tapponnier, P., Van der Woerd, J., Finkel, R., Xiwei, X.,
topographic data and cosmogenic 10Be geochronology. Geophys. Res. Lett. 34. Xiwei, X., Caffee, M., 2004. Rapid slip along the central Altyn Tagh Fault:
Friedrich, A.M., Wernicke, B.P., Niemi, N.A., Bennett, R.A., Davis, J.L., 2003. Comparison morphochronologic evidence from Cherchen He and Sulamu Tagh. J. Geophys.
of geodetic and geologic data from the Wasatch region, Utah, and implications for Res. 109, 23.
the spectral character of Earth deformation at periods of 10 to 10 million years. J. Mériaux, A., Tapponnier, P., Ryerson, F.J., Xiwei, X., King, G., Van der Woerd, J., Finkel, R.,
Geophys. Res. 108, 23. Haibing, L., Caffee, M., Zhiqin, X., Wenbin, C., 2005. The Aksay segment of the
Gold, R.D., Cowgill, E., Arrowsmith, J.R., Gosse, J., Wang, X., Chen, X., 2009. Riser northern Altyn Tagh fault: tectonic geomorphology, landscape evolution, and
diachroneity, lateral erosion, and uncertainty in rates of strike-slip faulting: a case Holocene slip rate. J. Geophys. Res. 110, 1–32.
study from Tuzidun along the Altyn Tagh Fault, NW China. J. Geophys. Res. B Solid Murray, A.S., Olley, J.M., 2002. Precision and accuracy in the optically stimulated
Earth Planet. 114, 24. luminescence dating of sedimentary quartz: a status review. Geochronometria 21,
Gosse, J.C., Phillips, F.M., 2001. Terrestrial in situ cosmogenic nuclides: theory and 1–16.
application. Quatern. Sci. Rev. 20, 1475–1560. Nicol, A., Walsh, J., Berryman, K., Villamor, P., 2006. Interdependence of fault
Haeussler, P., Schwartz, D.P., Dawson, T.E., Stenner, H.D., Lienkaemper, J.J., Sherrod, B., displacement rates and paleoearthquakes in an active rift. Geology 34, 865–868.
Cinti, F.R., Montone, P., Craw, P.A., Crone, A.J., Personius, S.F., 2004. Surface rupture Oskin, M., Iriondo, A., 2004. Large-magnitude transient strain accumulation on the
and slip distribution of the Denal and Totschunda faults in the 3 November 2002 M Blackwater fault, Eastern California shear zone. Geology 32, 313–316.
7.9 Earthquake, Alaska. Bull. Seismol. Soc. Am. 94, S23–S52. Oskin, M., Perg, L., Shelef, E., Strane, M., Gurney, E., Singer, B., Zhang, X., 2008. Elevated
Hampel, A., Hetzel, R., 2006. Response of normal faults to glacial-interglacial shear zone loading rate during an earthquake cluster in eastern California. Geology
fluctuations of ice and water masses on Earth's surface. J. Geophys. Res. 111. 36, 507–510.
Hampel, A., Hetzel, R., Maniatis, G., Karow, T., 2009. Three-dimensional numerical Ozawa, S., Murakami, M., Kaidzu, M., Tada, T., Sagiya, T., Hatanaka, Y., Yarai, H.,
modeling of slip rate variations on normal and thrust fault arrays during ice cap Nishimura, T., 2002. Detection and monitoring of ongoing aseismic slip in the Tokai
growth and melting. J. Geophys. Res. 114. Region, Central Japan. Science 298, 1009–1012.
Harkins, N., Kirby, E., 2008. Fluvial terrace riser degradation and determination of slip rates Peltzer, G., Crampé, F., Hensley, S., Rosen, P., 2001. Transient strain accumulation and
on strike-slip faults: an example from the Kunlun fault, China. Geophys. Res. Lett. 35, 6. fault interaction in the Eastern California shear zone. Geology 29, 975–978.
Haugerud, R.A., Harding, D.J., Johnson, S.Y., Harless, J.L., Weaver, C.S., Sherrod, B.L., 2003. Reid, H.F., 1910. On mass-movements in tectonic earthquakes. In: Lawson, A.C. (Ed.),
High-resolution lidar topography of the Puget Lowland, Washington—a bonanza The California earthquake of April 18, 1906. Carnegie Institute of Washington, D.C.
for earth science. GSA Today 13, 4–10. Reimer, P.J., Baillie, M.G.L., Bard, E., Bayliss, A., Beck, J.W., Bertrand, C.J.H., Blackwell, P.G., Buck,
Hetzel, R., Hampel, A., 2005. Slip rate variations on normal faults during glacial– C.E., Burr, G.S., Cutler, K.B., Damon, P.E., Edwards, R.L., Fairbanks, R.G., Friedrich, M.,
interglacial changes in surface loads. Nature 435, 81–84. Guilderson, T.P., Hogg, A.G., Hughen, K.A., Kromer, B., McCormac, G., Manning, S., Ramsey,
Hubert-Ferrari, A., King, G., Manighetti, I., Armijo, R., Meyer, B., Tapponnier, P., 2003. C.B., Reimer, R.W., Remmele, S., Southon, J.R., Stuiver, M., Talamo, S., Taylor, F.W., van der
Long-term elasticity in the continental lithosphere; modeling the Aden Ridge Plicht, J., Weyhenmeyer, C.E., 2004. IntCal04 terrestrial radiocarbon age calibration,
propagation and the Anatolian extrusion process. Geophys. J. Int. 153, 111–132. 0–26 cal kyr BP. Radiocarbon 46, 1029–1058.
Kenner, S., Simons, M., 2005. Temporal clustering of major earthquakes along Repka, J.L., Anderson, R.S., Finkel, R.C., 1997. Cosmogenic dating of fluvial terraces,
individual faults due to post-seismic reloading. Geophys. J. Int. 160, 179–194. Fremont River, Utah. Earth Planet. Sci. Lett. 152, 59–73.
64 R.D. Gold, E. Cowgill / Earth and Planetary Science Letters 301 (2011) 52–64

Ricker, W.E., 1973. Linear regressions in fishery research. J. Fish. Res. Board Can. 30, Van der Woerd, J., Ryerson, F.J., Tapponnier, P., Meriaux, A.S., Gaudemer, Y., Meyer, B.,
409–434. Finkel, R.C., Caffee, M.W., Zhao, G., Xu, Z., 2000. Uniform slip-rate along the Kunlun
Rymer, M.J., Seitz, G.G., Weaver, K.D., Orgil, A., Faneros, G., Hamilton, J.C., Goetz, C., 2002. Fault: implications for seismic behaviour and large-scale tectonics. Geophys. Res.
Geologic and paleoseismic study of the Lavic Lake Fault at Lavic Lake Playa, Mojave Lett. 27, 2353–2356.
Desert, Southern California. Bull. Seismol. Soc. Am. 92, 1577–1591. Van der Woerd, J., Tapponnier, P., Ryerson, F.J., Meriaux, A.-S., Meyer, B., Gaudemer, Y.,
Savage, J.C., Svarc, J.L., 2009. Postseismic relaxation following the 1992 M7.3 Landers Finkel, R.C., Caffee, M.W., Zhao, G., Xu, Z., 2002. Uniform postglacial slip-rate along
and 1999 M7.1 Hector Mine earthquakes, southern California. J. Geophys. Res. 114, the central 600 km of the Kunlun fault (Tibet), from 26Al, 10Be and 14C dating of
B01401. riser offsets, and climatic origin of the regional morphology. Geophys. J. Int. 148,
Scharer, K.M., Weldon II, R.J., Fumal, T.E., Biasi, G.P., 2007. Paleoearthquakes on the 356–388.
Southern San Andreas Fault, Wrightwood, California, 3000 to 1500 B.C.: a new Wallace, R.E., 1987. Grouping and migration of surface faulting and variations in slip
method for evaluating paleoseismic evidence and earthquake horizons. Bull. rates on faults in the Great-Basin Province. Bull. Seismol. Soc. Am. 77, 868–876.
Seismol. Soc. Am. 97, 1054–1093. Washburn, Z., Arrowsmith, J.R., Forman, S.L., Cowgill, E., Wang, X., Zhang, Y., Chen, Z.,
Sharp, W.D., Ludwig, K.R., Chadwick, O.A., Amundson, R., Glaser, L.L., 2003. Dating 2001. Late Holocene earthquake history of the central Altyn Tagh Fault, China.
fluvial terraces by 230Th/U on pedogenic carbonate, Wind River Basin, Wyoming. Geology 29, 1051–1054.
Quatern. Res. 59, 139–150. Weldon, R., Sieh, K., 1985. Holocene rate of slip and tentative recurrence interval for
Sieh, K., 1978a. Prehistoric large earthquakes produced by slip on the San Andreas Fault large earthquakes on the San Andreas fault, Cajon Pass, southern California. Geol.
at Pallett Creek, California. J. Geophys. Res. 83, 3907–3939. Soc. Am. Bull. 96, 793–812.
Sieh, K., 1978b. Slip along the San Andreas fault associated with the great 1857 Weldon, R., McCalpin, J.P., Rockwell, T.K., 1996. Paleoseismology of strike-slip tectonic
earthquake. Bull. Seismol. Soc. Am. 68, 1421–1448. environments. In: McCalpin, J.P. (Ed.), Paleoseismology, International Geophysics
Sieh, K.E., 1978c. Slip along the San Andreas fault associated with the great 1857 Series, pp. 271–329.
earthquake. Bull. Seismol. Soc. Am. 68, 1421–1448. Weldon, R.J., Fumal, T.E., Powers, T.J., Pezzopane, S.K., Scharer, K.M., Hamilton, J.C., 2002.
Steinbrugge, K.V., Zacher, E.G., Tocher, D., Whitten, C.A., Claire, C.N., 1960. Creep on the Structure and earthquake offsets on the San Andreas Fault at the Wrightwood,
San Andreas fault. Bull. Seismol. Soc. Am. 50, 389–415. California, paleoseismic site. Bull. Seismol. Soc. Am. 92, 2704–2725.
Sutherland, R., Berryman, K., Norris, R., 2006. Quaternary slip rate and geomorphology Weldon, R.J., Scharer, K.M., Fumal, T.E., Biasi, G.P., 2004. Wrightwood and the earthquake
of the Alpine fault: implications for kinematics and seismic hazard in southwest cycle; what a long recurrence record tells us about how faults work. GSA Today 14, 4–10.
New Zealand. Bull. Geol. Soc. Am. 118, 464–474. Wells, D.L., Coppersmith, K.J., 1994. New empirical relationships among magnitude,
Tapponnier, P., Molnar, P., 1977. Active faulting and tectonics in China. J. Geophys. Res. rupture length, rupture width, rupture area, and surface displacement. Bull.
82, 2905–2930. Seismol. Soc. Am. 84, 974–1002.
Tapponnier, P., Ryerson, F.J., Van der Woerd, J., Mériaux, A.S., Lasserre, C., 2001a. Long- Wesnousky, S.G., 2006. Predicting the endpoints of earthquake ruptures. Nature 444,
term slip rates and characteristic slip: keys to active fault behaviour and 358–360.
earthquake hazard, Comptes Rendus de l'Academie des Sciences Series IIA. Earth Wesson, R.L., 1988. Dynamics of Fault Creep. J. Geophys. Res. 93, 8929–8951.
Planet. Sci. 333, 483–494. Xu, X., Yu, G., Klinger, Y., Tapponnier, P., Van der Woerd, J., 2006. Reevaluation of surface
Tapponnier, P., Xu, Z., Roger, F., Meyer, B., Arnaud, N., Wittlinger, G., Yang, J., 2001b. rupture parameters and faulting segmentation of the 2001 Kunlunshan earthquake
Oblique stepwise rise and growth of the Tibet Plateau. Science 294, 1671–1677. (Mw7.8), northern Tibetan Plateau, China. J. Geophys. Res. 111. doi:10.1029/
Taylor, M., Yin, A., 2009. Active structures of the Himalayan–Tibetan orogen and their 2004JB003488.
relationships to earthquake distribution, contemporary strain field, and Cenozoic Zechar, J.D., Frankel, K.L., 2009. Incorporating and reporting uncertainties in fault slip
volcanism. Geosphere 5, 199–214. rates. J. Geophys. Res. 114.
Thatcher, W., 1983. Nonlinear strain buildup and the earthquake cycle on the San Zhang, P.-Z., Shen, Z., Wang, M., Gan, W., Burgmann, R., Molnar, P., Wang, Q., Niu, Z., Sun,
Andreas fault. J. Geophys. Res. 88, 5893–5902. J., Wu, J., Hanrong, S., Xinzhao, Y., 2004. Continuous deformation of the Tibetan
Van der Woerd, J., Ryerson, F.J., Tapponnier, P., Gaudemer, Y., Finkel, R., Meriaux, A.S., Plateau from global positioning system data. Geology 32, 809–812.
Caffee, M., Zhao, G., He, Q., 1998. Holocene left-slip rate determined by cosmogenic Zhang, P.-Z., Molnar, P., Xu, X., 2007. Late Quaternary and present-day rates of slip along
surface dating on the Xidatan segment of the Kunlun fault (Qinghai, China). the Altyn Tagh Fault, northern margin of the Tibetan Plateau. Tectonics 26, 1–24.
Geology 26, 695–698.