Technical Comments

Comment on "Absence of Cooling in New Zealand and the Adjacent Ocean During the Younger Dryas Chronozone"

See allHide authors and affiliations

Science  09 May 2008:
Vol. 320, Issue 5877, pp. 746
DOI: 10.1126/science.1152098


Barrows et al. (Reports, 5 October 2007, p. 86) presented cosmogenic exposure dates and data from an ocean sediment core that challenge evidence for glacier advance in New Zealand during the Younger Dryas event. We use modeling of geomorphic processes to argue that their cosmogenic exposure dates are inconclusive.

Barrows et al. (1) recently argued that the Waiho Loop moraine in New Zealand was deposited after the end of the Younger Dryas event, ∼11,600 years before the present (yr B.P.). They presented 24 new cosmogenic exposure dates, ranging from ∼5000 yr B.P. to ∼12,800 yr B.P., from ten boulders on the moraine. The authors estimated the age of the moraine by taking the error-weighted mean of the ages of their nine oldest boulders (2). This procedure yielded an age of 10,480 ± 240 yr B.P. (1 SD). The age of the Waiho Loop moraine had previously been estimated to be ∼12,800 yr B.P. by radiocarbon dating (3, 4). Because the new age estimate for the Waiho Loop moraine is much younger than this previous estimate, Barrows et al. concluded that “the Waiho Loop advance...was not a [Younger Dryas] event, as previously thought.”

We applaud Barrows et al. (1) for their investigation of the age of this moraine. As they note, the age of the Waiho Loop moraine is a critical test of various hypotheses explaining the climatic changes at the end of the last glaciation. Their use of multiple cosmogenic nuclides and multiple samples from each boulder enables assessment of their analytical work.

However, the age assignment for the Waiho Loop moraine given by Barrows et al. (1) is potentially biased. Barrows et al. noted that the scatter in their exposure dates is greater than can be explained by measurement error alone. This additional scatter is common in sets of exposure dates from moraines (5, 6) and is usually attributed to geomorphic processes (57). To address this scatter, many researchers remove a subset of the data and then take an average, as Barrows et al. did. However, application of the error-weighted mean assumes that the differences between the exposure dates and the true age of the moraine are normally distributed (8) and have a mean of zero. The known geomorphic processes that affect cosmogenic exposure dates change the mean of these differences, as well as their standard deviation (7). Consequently, the error-weighted mean may return a misleading estimate of the moraine's age if the dates have been affected by geomorphic processes.

Barrows et al. (1) acknowledged the effects that “former shielding...or earlier exposure” may have had on their dates; here, we consider these effects more explicitly. If the boulders sampled by Barrows et al. (1) were shielded from cosmic rays by a till cover, their exposure dates would tend to underestimate the moraine's age. Barrows et al. noted that one of their boulders had “...smaller blocks resting on top of it, perhaps indicating a former till cover.” Moreover, the Waiho Loop receives over 2 m of precipitation yearly (9). These observations suggest that till has been removed from the crest of the moraine, exposing fresh boulders to cosmic rays.

We present output from a simple numerical model that shows the impact of till shielding on the distribution of cosmogenic exposure dates. We updated the moraine degradation model of Hallet and Putkonen (5, 10, 11) to predict the distributions of 10Be exposure dates on moraines, including production at depth by muons (12). This model predicts that the distribution of 10Be exposure dates from a degrading moraine should have a peak close to the true age of the moraine and a long, heavy tail toward the young side of the distribution (Fig. 1). This result is robust over a range of initial moraine heights, initial moraine slopes, and topographic diffusivities. For this exercise, we considered only the 10Be dates, because the relationship between 36Cl dates and the true age of an eroding surface is complex (13). Although further modeling work may change the shape of this distribution somewhat, the conclusions we draw from this exercise are sound.

Fig. 1.

Comparison of eight 10Be exposure dates from the Waiho Loop moraine (black curves) (1) to the distribution of 106 exposure dates predicted by the updated degradation model (red curves) (5, 10, 11). The best fit between the model and the data occurs at a model age of 11,600 yr B.P. (dashed red line), corresponding to the end of the Younger Dryas. Our preferred age assignment is thus ∼1100 years older than that given in (1) (black, dashed line). This best fit was found by adjusting the modeled age of the moraine until both the sum of squared errors and the maximum vertical distance between the cumulative density curves (inset) were minimized. The probability density curve for the observed exposure dates was constructed by summing the Gaussian curves of the individual dates and normalizing the total curve to 1. We assumed that the moraine's initial height and slope were 50 m and 34°, respectively; topographic diffusivities are from (5). Although these diffusivities are probably too small for the wet West Coast of New Zealand (9), multiplying the diffusivities by 10 does not change the fundamental resemblance of the model results to the 10Be dates.

For an assumed moraine age of 11,600 yr B.P., at the end of the Younger Dryas, the model produces a distribution that resembles the 10Be exposure dates reported in (1) (Fig. 1). If the 10Be exposure dates of Barrows et al. are drawn from a parent distribution like that produced by the model, their error-weighted mean underestimates the age of the moraine. Instead, we suggest that the age of the moraine lies somewhere close to their oldest 10Be dates, possibly at the end of the Younger Dryas chronozone. Our preferred age assignment is thus ∼1100 years older than that favored by Barrows et al.

We also considered the possibility that the dates observed by Barrows et al. (1) are biased by earlier exposure rather than till shielding. Glaciers sometimes incorporate boulders with preexisting concentrations of cosmogenic nuclides in their moraines (7). Dates from these preexposed boulders tend to overestimate the ages of the moraines they rest upon. Modeling of this process produces distributions with a peak near the true age of the moraine and a long, heavy tail toward the old side of the distribution (11, 14). These distributions do not resemble the distribution of dates reported by Barrows et al. We conclude that earlier exposure is not primarily responsible for the spread in their dates, although their measured concentrations may include a small inherited component.

We do not intend to reinterpret the cosmogenic exposure dates presented by Barrows et al. (1) as requiring a Younger Dryas age for the Waiho Loop moraine. Other geologic processes not treated by our modeling may have influenced the observed exposure dates, and further calibrations of nuclide production rates may require reassessment of these data. Instead, we use this example to argue that geomorphic processes must be carefully considered before making age assignments for moraines from cosmogenic exposure dates. In particular, discarding some observations and taking an average of the rest can lead to biased age estimates.

References and Notes

Stay Connected to Science


Navigate This Article