https://doi.org/10.1177/0959683618756802
The Holocene
1 –13
© The Author(s) 2018
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0959683618756802
journals.sagepub.com/home/hol
Introduction
Rock glaciers are ice-rich debris creeping with velocities of some
centimetres to some metres per year. Their genesis has been
largely discussed and two main modes of rock glacier formation
have been proposed (Barsch, 1988, 1996; Clark et al., 1998; Hae-
berli et al., 2006; Humlum, 1988; Whalley and Martin, 1992): (1)
periglacial rock glaciers resulting from the creep of ice-rich sedi-
ments in permafrost conditions and (2) glacigenic rock glaciers
deriving from the burial of small glaciers, mainly located in con-
tinental areas.
Their development and evolution is highly sensitive to cli-
matic parameters (Brazier et al., 1998; Humlum, 1998; Kääb
et al., 2007); rock glaciers constitute thus an important potential
as archive for the regional Late Glacial and Holocene palaeocli-
matology. Furthermore, rock glacier activity does exclude con-
current glacial activity. Their occurrence at specified location
potentially suitable for glaciers may, therefore, bear implications
and act as a proxy for limited extensions of glaciers within wider
regions at certain times (Böhlert et al., 2011a, 2011b). The general
palaeoclimatic potential of such locations within the transition
between glacial and periglacial process systems has recently been
highlighted by Matthews et al. (2017).
The dating of rock glaciers has been the subject of several
studies, indicating that in many cases, relict rock glaciers mainly
formed during the Late Glacial, whereas currently active rock gla-
ciers predominantly formed during the Holocene (e.g. Barsch,
1996; Scapozza, 2015). Dating methods applied (see Haeberli
et al. (2003) for a review) involve relative-age dating methods
such as, for example, the Schmidt-hammer (Kellerer-Pirklbauer
et al., 2008; Scapozza et al., 2014) or weathering-rind thickness
(Frauenfelder et al., 2005); numerical age-dating methods (radio-
carbon dating: Haeberli et al., 1999; terrestrial cosmogenic
nuclide dating (TCND): Böhlert et al., 2011a; Cossart et al.,
2010a; optical stimulated luminescence: Fuchs et al., 2013); or
age estimations based on current velocity fields measured with
aerial photogrammetry (Frauenfelder et al., 2005). These studies
generally reveal considerable age variations on global and hemi-
spheric scales. For instance, Böhlert et al. (2011a) provide evi-
dence of two phases of rock glacier formation immediately
following glacier retreat at the end of the Younger Dryas and
another subsequent phase starting between 10,000 and 6000 years
Age constraints of rock glaciers in the
Southern Alps/New Zealand – Exploring
their palaeoclimatic potential
Stefan Winkler
1
and Christophe Lambiel
2
Abstract
Two rock glaciers in the valley head of Irishman Stream in the central Ben Ohau Range, Southern Alps/New Zealand, have been investigated using the
electronic Schmidt-hammer (SilverSchmidt). Longitudinal profiles on both features reveal a consistent trend of decreasing R(Rebound)-values and, hence,
increasing weathering intensity and surface-exposure age on their numerous transverse surface ridges from rooting zone towards the front. Previously
published numerical ages obtained by terrestrial cosmogenic nuclide dating (TCND) allowed the calculation of a local Schmidt-hammer exposure-age
dating (SHD) age-calibration curve by serving as the required fixed points. Age estimates for the lowermost rock glacier surface ridges fall within the
early Holocene between 12 and 10.5 ka and indicate a fast disappearance of the Late Glacial glacier formerly occupying the valley head, followed by the
initiation of rock glacier formation around or shortly after the onset of the Holocene. Although it cannot be judged whether the rock glaciers investigated
were active within the entire Holocene or only repeatedly during multiple episodes within, their location and intact morphology exclude any substantial
glacial activity at Irishman Stream during the Holocene. This has considerable regional palaeoclimatic implications because it opens for the hypothesis that
climatic conditions during early Holocene were possibly comparatively dry and favourable for rock glacier initiation, but less so for glaciers. It would also
challenge the view that air temperature is the sole major climate driver of glacier variability in the Southern Alps. More work utilising the palaeoclimatic
potential of rock glaciers in the Southern Alps is advised.
Keywords
electronic Schmidt-hammer (SilverSchmidt), Holocene palaeoclimate, New Zealand, rock glaciers, Schmidt-hammer exposure-age dating (SHD),
Southern Alps
Received 6 October 2017; revised manuscript accepted 8 October 2017
1
Institute of Geography and Geology, Julius-Maximilians-Universität
Würzburg, Germany
2
Institute of Earth Surface Dynamics, University of Lausanne,
Switzerland
Corresponding author:
Stefan Winkler, Institute of Geography and Geology, Julius-Maximilians-
Universität Würzburg, Am Hubland, 97074 Würzburg, Germany.
Email: stefan.winkler@uni-wuerzburg.de
756802HOL
0010.1177/0959683618756802The HoloceneWinkler and Lambiel
research-article2018
Research paper
2 The Holocene 00(0)
ago in the European Alps. Early-Holocene ages of 10,000–8000
years are also reported by Kellerer-Pirklbauer et al. (2008) and
Kellerer-Pirklbauer (2008) from northern Europe, whereas a start
of formation just prior or during the Holocene Thermal Maximum
is given by Scapozza (2015) for the Swiss Alps. Younger ages
were, however, obtained by Cossart et al. (2010) who propose
late-Holocene ages of 2500 years in the French Alps. These pat-
terns likely reflect different regional palaeoclimatic developments
during the Holocene and underline their potential as climate
proxies.
Although not uncommon within the dryer eastern parts of the
Southern Alps, New Zealand, comparatively few previous studies
(e.g. Brazier et al., 1998; Jeanneret, 1975; Kirkbride and Brazier,
1995) have focused on the spatial distribution of rock glaciers or
their dynamics and evolution (see Sattler et al., 2016). The first
complete rock glacier inventory for the Southern Alps has only
been completed recently (Sattler et al., 2016) and neither a sys-
tematic monitoring programme nor detailed recent studies on
their age constraints have yet been undertaken. In our study area,
the Ben Ohau Range, rock glaciers have been incorporated in the
subdivision of Holocene deposits by Birkeland (1982) and their
stratification classification would suggest an early late-Holocene
age (3000–4000 years old), previous work of McGregor (1967)
even a somewhat younger age. But because these old studies
mainly apply relative-age dating methods and, for example, sig-
nificantly underestimate the ages of Late Glacial/early-Holocene
moraines studied by Kaplan et al. (2010), these old age estimates
seem unreliable (see below). Many rock glaciers in the Southern
Alps are located in cirques or valley heads formerly occupied by
glaciers during the Late Glacial. Obtaining information about the
timing of rock glacier initiation alongside age constraints for their
successive development and dynamics will, therefore, yield evi-
dence about the retreat of glaciers and a related decrease of glacial
activity at those sites to potentially help understanding the
regional palaeoclimatic conditions during the Late Glacial/Holo-
cene transition and the early Holocene.
It is a common view that glacier extension in the Southern
Alps has been continuously reduced during the Holocene caused
by long-term warming related to an underlying increase of solar
insolation following the orbital trend (Kaplan et al., 2013; Solo-
mina et al., 2015). There is, however, to date only fragmentary
reliable morphological evidence of the proposed larger early-
Holocene glacier expansion to confirm this (Kaplan et al., 2013;
Putnam et al., 2012), often explained by potentially efficient ero-
sional censoring sensu Kirkbride and Winkler (2012) in the highly
dynamic geomorphological environment. Model-based glacio-
logical studies postulating that regional Holocene glacier variabil-
ity very strongly depends on air temperatures (Anderson and
Mackintosh, 2006; Mackintosh et al., 2017) support such consid-
erations. The latter view is, however, challenged by work on gla-
cier fluctuations during the late 20th century CE and its conclusion
that atmospheric circulation patterns and precipitation changes
are equally important drivers for the maritime New Zealand gla-
ciers (Chinn et al., 2005; Clare et al., 2002). Furthermore, the
absence of widespread early-Holocene glacier advances would
correspond to the glacier chronology of mid-latitudinal mountain
glaciers in South America (cf. Kaplan et al., 2016; Menounos
et al., 2013; Strelin et al., 2014) and underline postulated telecon-
nections between both regions (Fitzharris et al., 2007). Chrono-
logical studies on rock glaciers in the Southern Alps could provide
an alternative way to assess this issue. Any formation of rock gla-
ciers at early stages during the Holocene would open for the
hypothesis that Late Glacial glaciers in the Southern Alps retreated
faster and were more significantly retracted immediately at the
onset of the early Holocene than previously anticipated. Keeping
in mind that rock glaciers required permafrost for their formation,
the early Holocene must (still) have been cold enough despite
comparatively moderately lowered air temperatures during the
Late Glacial (Doughty et al., 2013) and the following long-term
warming trend during the Holocene (Kaplan et al., 2013). There-
fore, retracted early-Holocene glaciers could provide a palaeocli-
matic signal for climatic conditions too dry for glaciers, thus
questioning the postulated strong dependency on temperature
conditions only. In many northern hemispheric regions, rock gla-
cier initiation following the mid-Holocene Thermal Optimum and
even as late as during the ‘Little Ice Age’ (LIA) (Janke et al.,
2013) reflects hemispheric climate patterns (Wanner et al., 2011).
By contrast, predominant rock glacier formation in the Southern
Alps taking place during the early Holocene would nicely under-
line different hemispheric patterns aligned with the respective
palaeoclimatic conditions.
But unlike single-age landforms related to comparatively
short-lived events (e.g. moraines formed during individual glacier
advances), obtaining age estimates for the initiation of rock gla-
cier formation and its subsequent development constitute a more
difficult methodological task. The characteristic rock glacier mor-
phology and dynamics require large sample sizes with any type of
surface-exposure dating because every single surface boulder
inhibits the potential of ongoing disturbance after its initial accu-
mulation. Following pilot studies of using the Schmidt-hammer
for (relative) dating of moraines on glacier forelands in Aoraki/Mt
Cook National Park (Winkler, 2005), Schmidt-hammer exposure-
age dating (SHD) has subsequently been successfully applied on
moraines as well as on (glacio)fluvial terraces in the Southern
Alps (Stahl et al., 2013; Winkler, 2009, 2014). In particular, the
widespread and comparatively homogeneous ‘greywacke’ of the
Torlesse composite terrane dominating east of the Main Divide
(Cox and Barrell, 2007) has proven to be suitable for Schmidt-
hammer studies, even indicating potential for the development a
regional age-calibration curve like Matthews and Owen (2010)
did in Southern Norway (Winkler and Corbett, 2014). But whereas
the Schmidt-hammer has already been applied on rock glaciers in
the European Alps, Iceland and Norway (e.g. Böhlert et al.,
2011b; Frauenfelder et al., 2005; Kellerer-Pirklbauer et al., 2008;
Matthews et al., 2013; Rode and Kellerer-Pirklbauer, 2011;
Scapozza et al., 2014), neither rock glaciers nor any other perigla-
cial landforms have previously been sampled in the Southern
Alps of New Zealand.
The primary aim of our study is to test whether SHD can suc-
cessfully be applied on rock glaciers in the Southern Alps and to
compare the results obtained with those of similar studies on rock
glaciers elsewhere. We will, furthermore, calculate a local SHD
age-calibration curve by using fixed points in the form of TCND
ages previously published by Kaplan et al. (2010) to gain age con-
straints for the rock glaciers investigated and enable interpretation
of their initiation and dynamics. Finally, by comparing this infor-
mation with established glacier chronologies, the potential of fur-
ther studies on rock glaciers as palaeoclimatic archives will be
discussed briefly.
Study area
The high-altitude valley head of Irishman Stream in the Ben
Ohau Range between Lakes Ohau and Pukaki is located roughly
30 km southeast of the Main Divide of the Southern Alps. It
contains a number of separate rock glaciers – two selected for
this study (Figures 1–3). Rock glacier 1 is a 400 m long land-
form composed of a succession of ridges and furrows. Altitudes
range from 1870 m a.s.l. at the front to 1990 m a.s.l. at the root-
ing zone (or head), which places this rock glacier below the sug-
gested lower limit of modern permafrost at c. 2000 m a.s.l.
(Brazier et al., 1998; Kirkbride and Brazier, 1995) and the mod-
elled permafrost distribution by Sattler et al. (2016). Whereas
rock glacier 1 is the lowermost rock glacier in the entire valley
Winkler and Lambiel 3
head, rock glacier 2 just reaches down to c. 2000 m a.s.l. It com-
prises two separate segments indicated by a prominent bend in
its flow line between ridges 3 and 4 at about 2025/2030 m a.s.l.
(Figure 1). The upper part trending relatively straight N–S
appears to have marginally overridden the lower part, indicating
a possible two-phase development.
The bedrock of the central Ben Ohau Range belongs to the
Rakaia terrane of the Torlesse composite terrane, mostly undif-
ferentiated dated as Permian to Triassic age (see Cox and Barrell
(2007) for details). The original quartzofeldsphatic sandstones
interbedded with silt and mudstones become more strongly meta-
morphosed and deformed towards the Alpine Fault northwest of
our study area, but east of the Main Divide, ‘greywacke’ domi-
nates. The ‘massive’ appearance of this rock type has previously
proven to be suitable for Schmidt-hammer measurements (Stahl
et al., 2013; Winkler, 2005). It weathers comparatively slowly
while changing its fresh greyish surfaces into pinkish colours.
This coincides with selective surface weathering and the common
thin quartz veins, avoided with Schmidt-hammer sampling, stick
out by up to 10 mm on Late Glacial boulders. But in general, the
surface morphology of boulders is minimal and of no concern for
the reliability of Schmidt-hammer measurements.
No detailed meteorological data are available for the study
area. Mean annual air temperatures (MAAT) for active rock gla-
ciers in the Southern Alps have been modelled to be positive and
among the highest reported on a global scale (see Sattler, 2016;
Sattler et al., 2016, for details). MAAT for Irishman Stream Val-
ley head should be around 0°C or slightly above at 2000 m a.s.l
and it has a southerly aspect. The high annual precipitation
likely exceeds 10,000 mm on western slopes and, thanks to an
‘overspill’ effect, is also typical for the Main Divide of the
Southern Alps itself and areas immediately to its east (Chater
and Sturman, 1998; Griffiths and McSaveney, 1983; Henderson
and Thompson, 1999). But it does not reach the central Ben
Ohau Range. Irishman Stream is located east of the 2000 mm
isohyet (Sattler, 2016; Tait et al., 2006). It seems too dry to sup-
port considerable Holocene glaciation, thus allowing the forma-
tion of rock glaciers practically absent in the more westward
parts of the Southern Alps.
Figure 1. Aerial orthophoto with contours of the Irishman Stream valley head and its location within New Zealand and the Ben Ohau Range,
respectively (inserts). Main contours show 50-m intervals, auxiliary ones 10-m intervals. The longitudinal profiles for Schmidt-hammer sampling
on the two rock glaciers studied are shown alongside individual sample sites of 2014 and 2016. The TCND control sites refer to the ages
published by Kaplan etal. (2010), complemented by ‘young’ control site No.6 (see text for further details; modified from Land Information New
Zealand’s online database http://www.linz.govt.nz).
Figure 2. Aerial photo taken 3 March 1960 (non-rectified)
depicting the detailed morphology of the rock glaciers of the
Irishman Stream valley head. The rock glaciers investigated (I
and II) are indicated (cf. Figure 1), also three presumed late-
Holocene/‘Little Ice Age’ moraines not investigated here in detail
(see text; cf. Birkeland, 1982; Kaplan etal., 2010).
4 The Holocene 00(0)
Methods
Schmidt-hammer measurements
Schmidt-hammer measurements were carried out in February
2014 and January 2016. Electronic Schmidt-hammers (Silver-
Schmidt) of the N-type configuration with calibrated impact
energy of 2.207 Nm for the plunger (Proceq SA, 2012) were cho-
sen to ensure high efficiency with acquiring substantial raw data in
limited time. Although not attempted in our study, Winkler and
Matthews (2014) have previously shown that Q-(velocity) values
obtained by electronic Schmidt-hammers are interconvertible with
R-(rebound) values of mechanical instruments, thus allowing com-
parison with existing data series. For simplification, all Silver-
Schmidt Q-values indicating compressive strength of the boulder
surfaces sampled when measured using an inbuilt sensor to record
the rebound velocity of the plunger independent of its impact
direction are called ‘R-values’ hereafter. This anticipates the man-
ufacturer’s decision to re-name impact data ‘R-values’ with the
improved version of the electronic Schmidt-hammer (Rock-
Schmidt; cf. Proceq SA, 2014; Winkler et al., 2016). All instru-
ments used during fieldwork were checked for their calibration to
the manufacturer’s specifications. Detailed data from related pre-
and post-fieldwork tests on the specific manufacturer’s test anvil
additionally allowed correcting the raw data for minor differences
(~R = 1) that otherwise simply would have fallen within the
acceptable range of tolerance for calibration.
Abundance of boulders suitable for sampling and character of
the landforms investigated (rock glaciers and moraines) called for
an established sampling design of single impacts on large num-
bers of boulders. With one impact each on 50 randomly selected
boulders and multiple samples (preferably 3 or more) on every
individual site, potential postdepositional disturbance in high-
alpine environments was also addressed, and preferable to designs
using multiple impacts on fewer boulders. Individual samples
were finally combined for every site after they had shown consis-
tency without outliers. They are reported in the accepted standard
format as mean R-values with 95% confidence (α = 0.05) follow-
ing Shakesby et al. (2006) and others using the equation:
Xtsn
±−()
1
(1)
where
X
= arithmetic mean, s = sample standard deviation, t = Stu-
dent’s t statistic and n = number of impacts (sample size). Because
the rock glacier sites were initially expected to resemble a diachron-
ous rather than a synchronous surface (i.e. with a considerable spread
Figure 3. Irishman Stream valley head photos taken on 10 January 2016 and 15 February 2014 (Figure 3d): (a) The western valley head
with the rock glaciers 1 and 2 investigated in this study (cf. Figures 1 and 2); (b) the eastern valley head with additional rock glaciers showing
morphological signs of activity. (c) Displays the Late Glacial moraines of the former Late Glacial Irishman Stream Glacier as seen towards the
south (outermost Late Glacial moraine indicated by arrows); (d) depicts a view from TCND fixed point 2 in northern direction towards rock
glaciers and the valley headwall. (e) The typical ridge-and-furrow surface morphology of rock glacier; (f) was taken near the rooting zone of
rock glacier 2.
Winkler and Lambiel 5
of exposure ages as revealed by their R-values), detailed histograms
were produced for all sites for further interpretation. Except Kol-
mogorov–Smirnov tests for normality, no further statistical analysis
was performed at this stage.
In 2014, sampling was primarily carried out on Late Glacial
and early-Holocene moraines ridges investigated by Kaplan et al.
(2010) that eventually were used as fixed points for the calcula-
tion of SHD age-calibration curves (see below). Apart from one
site that was re-sampled to ensure reproducibility, sampling in
2016 focused on the selected rock glaciers in the form of two
longitudinal profiles from front to their rooting zone. Every indi-
vidual transverse surface ridge along these profiles mapped with
a differential GPS was sampled and treated as individual site. To
avoid potential influence of micro-climatic and micro-weathering
differences, furrows on the rock glacier surface between those
ridges were avoided (except for one test site).
Calculation of Schmidt-hammer exposure-age
calibration curves
Reliable and well-documented TCND ages from Irishman Stream
valley head reported by Kaplan et al. (2010) provided sufficient
fixed points in the form of numerical age information required for
the construction of a local age-calibration curve and subsequent
application of SHD as described in detail by Matthews and Owen
(2010) or Shakesby et al. (2011). Because Kaplan et al. (2010)
applied the
10
Be-production rates of Putnam et al. (2010), the only
available regional calibration scheme available to date, no re-
calculation of their original data seemed necessary (see discus-
sion). Late Glacial and early-Holocene moraine ridges investigated
in their study have, thanks to the restricted catchment of Irishman
Stream, the same source of debris with a visible dominance of
dumped former supraglacial debris on their surface. Schmidt-
hammer data were obtained on these moraines easily identifiable
on the ground, supplemented by a ‘young’ fixed point in the form
of ‘fresh’ appearing rock fall boulders at the uppermost rooting
zone of rock glacier 2. This site was given a fixed age of 100 ±
100 years as best estimate. For our fixed points, we did not use the
moraine ages calculated by Kaplan et al. (2010) as arithmetic
means of all related TCND ages but only those individual samples
located on the actual part of the moraine ridges we sampled with
the SilverSchmidt (see Figure 1). An arithmetic mean for fixed
point ages was only calculated if more than one individual sample
was available for the part of the moraine we sampled (Table 1).
The underlying rational was to avoid any potential influence of
minor lithological and micro-climatic differences, or complica-
tions caused during the uninvestigated history of moraine forma-
tion. Importance of the latter was recently highlighted at nearby
Mueller Glacier by Rezninchenko et al. (2016) who presented
evidence that the use of an arithmetic mean of related TCND ages
may not necessarily be an appropriate way to calculate a reliable
date for moraine formation in the specific geomorphological
environment of the Southern Alps.
The procedure of calculating SHD calibration curves followed
recent practice established and explained in detail by Matthews
and Winkler (2011), Matthews et al. (2014) or Winkler et al.
(2016). In their fundamental work on an ideal study site, Shakesby
et al. (2011) confirmed that the R-value–age relationship is best
described by a linear function. Local SHD studies (Winkler, 2014;
Winkler and Corbett, 2014) came to the same conclusion and cor-
respond to linear weathering intensity-age trends applied with the
weathering-rind-thickness dating method in the region (Birke-
land, 1982; Chinn, 1981, and others). Even with the limited
spread of available fixed points preventing better local re-assess-
ment, initial testing also demonstrated better fit of linear versus
exponential functions and provided, together with those above-
mentioned results, sufficient confidence about the applied shape
of the SHD calibration curves that follow the standard equation
for linear regression:
ya bx=+
(2)
where y = surface age in years, x = mean R-value, a = intercept
age and b = slope of the calibration curve.
Using all available fixed points, the local calibration curve is
defined by the equation:
yx
= 59963 1037 862 0690
..
−
(3)
and subsequently applied for the calculation of SHD age estimates
presented here. Three additional calibration curves were calcu-
lated by excluding one or two of those fixed points representing
the ‘inner moraines belt’ of Kaplan et al. (2010). These alternative
calibration curves resulted in only insignificant improvements of
R
2
(Table 2) and could easily be artefacts of fewer points generally
giving better fit with linear equations. Their similarity would any-
way have resulted in similar age estimates within the related error
ranges. Even if the possibility of postdepositional disturbance
because of higher altitude of the ‘inner moraines belt’ located
within altitudinal range of the lowermost rock glacier fronts justi-
fied initial testing, their application was not followed up. Because
the TCND ages serving as fixed points were not collected as part
of this study, we honestly felt unable to judge the representative-
ness of individual samples provided by Kaplan et al. (2010) and,
therefore, decided not to exclude any.
Table 1. Fixed points for the SHD age-calibration curve at Irishman Stream.
Site R-value
a
n (boulders) TCND age
b
n (samples) Sample IDs
c
1 54.18 ± 1.84 100 13100 ± 339 6 06-31 – 33,06-35 – 37
2 55.04 ± 1.46 150 12800 ± 300 2 06-41/42
3 55.49 ± 1.28 200 12100 ± 300 1 06-47
4 57.32 ± 1.08 300 11900 ± 300 1 06-28
5 54.58 ± 1.06 350 11650 ± 490 2 06-26/27
6 69.47 ± 1.26 50 100 ± 100
d
– –
SHD: Schmidt-hammer exposure-age dating; TCND: terrestrial cosmogenic nuclide dating.
R-values and numerical ages of the selected fixed points for the SHD age-calibration curve of the Irishman Stream valley head (for position and altitude
cf. Figure 1; see also text and Kaplan etal. (2010) for the original TCND dataset and their sample IDs; Kaplan etal. used the
10
Be-production model of
Putnam etal. (2010), the only regional calibration scheme available for the Southern Alps to date).
a
Mean of R-values (SilverSchmidt) with 95% confidences (α = 0.05).
b
TCND ages calculated on the basis of results published by Kaplan etal. (2010). With more than one sample, the arithmetic mean has been calculated
(see text).
c
Original TCND sample IDs of Kaplan etal. (2010).
d
‘Young’ fixed point (see text).