Technical Comments

Response to Comments on “Reconciliation of the Devils Hole climate record with orbital forcing”

See allHide authors and affiliations

Science  21 Oct 2016:
Vol. 354, Issue 6310, pp. 296
DOI: 10.1126/science.aaf8679


Winograd and Coplen question the thorium-230 distribution model proposed to explain the age bias observed with increasing depth during Termination II. We have evaluated both criticisms and find that all samples display virtually identical fabrics, argue that the modern setting is not analogous to the conditions during Termination II, and reiterate the robustness of our age models. Our conclusions remain unchanged.

In their Comments on our Report (1), Winograd (2) and Coplen (3) question the hypothesis that an increase in 230Th concentration with depth in the Devils Hole aquifer [DHaq, encompassing Devils Hole cave (DH) and Devils Hole no. 2 cave (DH-2)] caused erroneously old ages in deeper samples across Termination II (TII). Winograd’s argument is based on the accuracy of the 230Th ages produced for the shallow samples, whereas Coplen’s argument is founded on a “challenging” depth gradient in 230Th needed to maintain the age offset.

δ18O time series produced over a depth range of 32.1 m within DHaq demonstrate a linear (R2 = 0.97) increase in age with depth for TII (1). Using well-understood ocean chemistry (46), Moseley et al. (1) and others (7, 8) proposed a model for DHaq whereby 230Th produced in the water column is scavenged and transported downward by settling particles. Continuous reversible exchange between dissolved 230Th and 230Th adsorbed onto particles causes an increase in concentration with depth (4). Scavenging (9) of dissolved 230Th onto calcite at the cave walls thus results in increasingly erroneously old 230Th ages with depth. Similar to the oceans, it is possible that changes with time in the total sediment flux affected the radionuclide flux (5, 9) and intensity of scavenging (9), hence providing a mechanism for the greater age biases observed during glacial terminations (1).

Winograd and Coplen do not dispute that an excess of 230Th exists in the DHaq [in (10), 230Th ages are corrected for excess 230Th]; in contrast to our model, however, Winograd argues that effective mixing and active flow in DHaq precludes an increase in 230Th concentration with depth (2), whereas Coplen questions the steepness of the 230Th gradient (3).

Winograd’s qualitative arguments are based on (i) temperature surveys in DH that suggest that convection occurs in winter and stratification in summer along with “horizontal groundwater flow” (11); (ii) low vertical thermal gradients characteristic of upward flow of geothermal water within fault zones; (iii) chemical homogeneity between 5 and 37.5 m depth (12); (iv) the difference between air and water temperatures; and, finally, (v) a statement about water transit times. Several lines of reasoning argue against these points.

Temperature measurements to –34 m in DH indicate that cold surface waters sink during the winter months, causing vertical convective mixing (11). This finding, however, is hardly relevant for our model because the roof collapse that opened DH to the surface occurred ~60,000 years ago (12); thus, modern winter conditions in DH are not analogous to the conditions that existed during TII. DH-2, which remains relatively enclosed with only two small (several m2) openings, provides the closest modern analog for TII conditions; however, detailed studies of the water column have not yet been conducted. Observations in DH during the summer, when the air temperature is closer to that of the water, are thus the closest available analog to TII conditions, and these indicate very precisely the presence of a stable, stratified water column with a temperature gradient of 0.005°C m−1 (11) (i.e., no convective mixing).

Winograd et al. (10) estimated the transit time of water recharging ~80 km away and discharging at DH to be less than 2000 years. Based on these numbers, the transit time between the caves (5 years) was calculated to be within 230Th age uncertainties (1); hence, differences in age for a specific event are unlikely to be caused by leads and lags in the arrival of the water. The transit time between the two caves may, however, be variable, and had it ever exceeded 230Th age uncertainties (1), then the δ18O time series for DH, which is downstream, should lag DH-2 [note that the flow direction is incorrect in (2)]. In fact, the data demonstrate the exact opposite (1). Winograd (2) implies that the flow velocity within the two caves is different from (faster than?) his average for the aquifer (10) and that the DH and DH-2 water bodies are not isolated standpipes. In such “active flow,” he postulates that 230Th would be displaced, preventing its accumulation with depth.

Assuming that Winograd’s (10) transit time is correct, the average flow velocity in the groundwater system is on the order of n × 10−6 m s−1, which is two orders of magnitude slower than the terminal velocity of settling fine-silt particles (13). Lateral laminar advective flow may thus deflect trajectories of settling particles in the direction of flow but cannot disturb the vertical 230Th gradient. Furthermore, the presence of “dirty” calcite on the footwalls of both caves is evidence that the hydrodynamic nature of the DHaq does not prevent the settling of particles (and hence the accumulation of 230Th with depth). Similarly, oceans are not stagnant water bodies either, yet both shallow and deep waters maintain a 230Th gradient (46).

Coplen outlines a case in which he assumes that sample DH-2spl collected at –21 m relative to the present water table (r.w.t.) is the “heretofore accepted Devils Hole time series” (3). Justification for why DH-2spl was chosen as the correct chronology over the other three samples deposited at +2.1 (DH2-D), +0.8 (DH2-E), or –30 (DH-11) m r.w.t. is, however, absent. We question the stance that DH-2spl yields the “correct” chronology across TII on the basis that the glacial-interglacial transition in this δ18O time series leads local dripstone records by ~8 thousand years (ky), of which the latter are in agreement with orbital forcing (1416). Because the DHaq and local dripstones received the same Pacific-sourced moisture during TII, the timing of δ18O shifts at each site should be in agreement.

In Moseley et al. (1), age models for samples DH2-D, DH2-E, and DH2spl were constructed in OxCal (17, 18) from high-resolution (100 μm) stable isotopes and a high density of precise [3 per mil (‰)] 230Th ages in stratigraphic order, which had been analyzed using the most up-to-date techniques (19, 20). Age-model indices (21) were ~100% in all cases. In keeping with previous discussions (22), and thus use of sample DH-11, which has a TII midpoint δ18O value of –15.7‰ (Vienna Pee Dee belemnite) (22), Moseley et al. assigned the TII midpoint in each newly analyzed sample as the first point in the age model with a δ18O value of –15.7 ± 0.08‰ (23). We consider our age modeling approach and allocation of TII-midpoint ages to be clear, logical, and reasonable.

Under the assumption that DH-2spl is the correct chronology, Coplen states that the timing of TII recorded in shallower cores DH2-D and DH2-E is too young (3). Using “smoothed” versions [(figure 1 in (3)] of the age models of Moseley et al. (1), Coplen shifts the TII midpoints by up to 1600 years (Table 1) and calculates an age disparity of ~2.5 ky m−1 from the two shallowest cores (excluding the two deeper cores) (Fig. 1). The method of construction of Coplen’s alternative age models is unclear; plus, we find no reason why a different approach to age modeling should be taken.

Table 1 Ages for the TII midpoint in samples from different depths/elevations in DH and DH-2. Uncertainties are ± 2 SD.
View this table:
Fig. 1 The ages of the TII midpoint in samples at different depths/elevations relative to the present water table.

Uncertainties are ± 2 SD. Linear trend lines are fitted through the original data set [red, n = 4 230Th ages (1)] and Coplen’s newly proposed data set [blue, n 230Th ages = 2 (3)].

Moseley et al. (1) demonstrated a strong linear correlation (R2 = 0.97) between age and depth that encompassed all four samples and yielded an age bias-depth gradient of 0.29 ky m−1 (Fig. 1). Similarly, a linear trend line fitted to Coplen’s four TII-midpoint ages (R2 = 0.95) would yield a gradient of 0.33 ky m−1. Consequently, the age bias-depth gradient (~0.3 ky m−1) is more realistic when all four samples across a depth range of 32.1 m are taken into consideration, rather than just two samples over a depth range of 1.3 m (2.5 ky m−1).

Coplen (3) suggests that analysis of additional higher-elevation cores from +4.5 and +5.5 m r.w.t. should yield TII-midpoint ages of 124.6 and 122.1 ky, respectively, based on the 2.5 ky m−1 gradient (3). Indeed, we do intend to analyze additional cores from higher elevations above the present water table, although such work will take considerable time, effort, and resources. Likewise, a detailed study examining the cause for deviation of δ18O values between shallow and deep samples at discrete time periods, such as 110 thousand years ago (ka) (3), reaches beyond the scope of this Response.

The suggestion that we analyze the 230Th concentration with depth in DH-2 is indeed something that we have considered. Although it would be an interesting exercise, this must await a major effort to collect such water without introducing contamination from the nearby walls of the cave. We have, however, measured the 234U concentration of shallow DH-2 water (Table 2) and find that it is 2.2 times that of seawater (24); thus, all other factors being equal, one would expect 230Th concentrations in DH-2 water to be higher than a marine equivalent. Furthermore, as discussed, the modern water may not be analogous to TII water; thus, given that the age bias is only observed during glacial terminations, measurements on today’s aquifer may not provide a suitable indication of the conditions that were present when the age-offset was created.

Table 2 U concentration and composition of DH-2 water.

The samples were collected in February 2016, ~0.2 and 1.3 m below the modern water table. “Filtered” indicates that water samples were passed through a 0.2-μm pore filter to remove particles >0.2 μm. Uncertainties are ± 2 SD.

View this table:

In summary, although several details of the DHaq reversible 230Th scavenging model require further study, arguments based on the hydrodynamic nature of the aquifer (2) and the 230Th gradient (3) do not render the model invalid.

Winograd (2) furthermore proposes that the ages reported in (1) may represent “mixed ages” of older fast-growing calcite and younger calcite that subsequently filled pore spaces. He speculates that fast calcite growth, leading to higher porosity, was due to CO2 outgassing in the upper ~5 m of the water column during TII. In support of this contention, he notes that samples displaying younger ages for TII formed much closer to the water table and that the water table was declining during the glacial-interglacial transition. Winograd rightly suggests that his assumption can be tested by thin-section examination (2).

Examination of the calcite fabrics that were deposited across TII in cores at +2.1, +0.8 and –21 m r.w.t. reveals virtually identical petrographic features (Fig. 2) generally consistent with previous observations (25). Mammillary calcite in all three samples has a compact fabric, comprising composite columnar crystals, each being a “bundle” of multiple rod-shaped (aspect ratio >20) crystallites. Primary growth zonation observed under epifluorescence indicates that the mammillary calcite surface advanced as multiple rhombohedral crystal terminations of crystallites 50 to 150 μm in size. Both the columnar crystals and the crystallites are tightly packed, and intervening pore spaces are absent. Aqueous fluid inclusions are very rare, and typically occur where several adjacent columnar crystals meet. The former show equant three-dimensional shapes and commonly preserve the morphology of cleavage rhombohedra (i.e., no sign of recrystallization).

Fig. 2 Thin sections of samples at +2.1, +0.8, and –21 m relative to the present water table (transmitted light; nicols partially crossed).

Calcite that was deposited during TII is located approximately within the bounds of the red dashed lines. The respective δ18O curve (yellow) for each sample is superimposed on top of the thin section in its approximate position and shows an increase of ~2‰ (not to scale) across TII. The boundary between mammillary calcite (subaqueous) and folia (formed at the water table) in DH2-D and DH2-E is emphasized by a solid white line. The calcite fabric deposited during TII is identical to that deposited before TII and in the Eemian and does not vary between samples.

There is thus no evidence of increased porosity in any of our samples across TII. The contention of Winograd (2) regarding “fast” calcite growth and development of porosity in mammillary calcite forming at shallow depth in DHaq is not supported by petrographic evidence.

Furthermore, although the mammillary calcite in the shallow cores was, indeed, deposited on average slightly faster than its deeper counterpart during TII, the growth rates show no obvious relation to the water table decline (1). In fact, deposition seems to have slowed down rather than accelerated across TII, as would be expected if the deposition rate were controlled by progressively higher CO2 outgassing in response to a declining water table. This suggests that CO2 outgassing was not the primary factor controlling the kinetics of mammillary calcite deposition. This conclusion is consistent with previous findings (12) that “rate dependence on PCO2 is small at CO2 partial pressures in the range of that at Devils Hole.”

References and Notes

Acknowledgments: This work was supported by the Austrian Science Fund (FWF) project no. FP263050 to C.S. and T 710-NBL to G.E.M., and in part by NSF grants 1337693 to R.L.E. and 1602940 to R.L.E. and H.C. This research was conducted under research permit numbers DEVA-2010-SCI-0004 and DEVA-2015-SCI-0006 issued by Death Valley National Park. We thank J. Wallraf for preparation of the thin sections and K. Wilson and R. Freeze for assistance in the field.
View Abstract

Navigate This Article