Figure 5: Comparison of reconstructed values of temperature and
δ18Owater from material-specific and
composite calibrations from this study to observations. Use of a
composite Δ47-temperature calibration yields less
accurate and precise results. A. Comparison of measured temperature
(Tobs.) to Δ47-derived temperature
(TΔ47) values using the material-specific (top row, blue
frame) and composite freshwater calibration (top row, black frame).
Bottom panel shows a comparison of temperature residuals
(reconstructed-observations) using the composite and material-specific
calibrations. Values derived using the composite regression are
represented using black squares and material-specific calibrations are
represented using blue circles. Horizontal black bars represent the mean
and values at the bottom of each dataset show the mean value and
standard deviation for residuals along with the RMSE for each dataset
using the respective calibrations. Our results show improvement in RMSE
of reconstructed temperatures for biogenic carbonates and micrites when
applying the material specific calibration compared to the composite
calibration. Temperature residuals for biologically mediated carbonates
are similar between the composite and material specific calibration,
however, there is increased spread when applying the material specific
calibrations, reflected in the elevated RMSE relative to the composite
calibration. Similar to biologically mediated carbonates, the spread in
temperature residuals for travertines, and thus RMSE, increases, when
using material specific calibrations. However, we note that the overall
mean of the residuals is reduced for travertines when applying a
material specific calibration relative to the composite calibration. B.
Comparison of measured δ18Owater(δ18Oobs.) to
Δ47-derived δ18Owatervalues (δ18Orecon.) using material
specific (top row, blue frame) and composite freshwater calibration (top
row, black frame). δ18Owater values
are calculated using temperatures derived using our composite and
material-specific calibrations between Δ47 and
temperature, and oxygen isotope mineral-water fractionation factors from
Kim and O’Neil (1997) (calcite) or Kim et al. (2007) (aragonite). Bottom
panel displays δ18Owater residuals
(reconstructed-observations) using the composite and material-specific
calibrations. Our results show improvement in RMSE for
δ18Owater reconstructions for biogenic
carbonates, micrites, and travertines.
δ18Owater RMSE for
biologically-mediated carbonates is similar for both the composite and
material specific calibration, however, we note that the mean of the
residuals is reduced, and closer to zero, in the material specific
calibration.
3.3 Reconstructing water δ18O
In addition to providing thermodynamic constraints on the temperature of
formation of carbonates, Δ47 measurements can be paired
with carbonate oxygen isotope ratios to directly calculate source water
δ18O. We compare our clumped-isotope derived estimates
of source water δ18O derived with our temperature
predictions (δ18Ow-reconstructed) to
measured modern freshwater δ18O
(δ18Ow-measured) values to evaluate if
this method can accurately reconstruct the isotopic composition of the
water in which the carbonate precipitated. For
δ18Ow-measured, we synthesized any
available published measurements for the water bodies examined to
compare to reconstructed values.
δ18Ow-measured data is available for
86 of the 108 sites examined in this study (Supplementary Table 4).
Although some sites had long-term measurements of water body oxygen
isotope composition, some of the measurements were single point
measurements, and thus may not fully represent temporal variability.
We used the equation of Kim & O’Neil (1997) for calcite and Kim et al.
(2007) for aragonite in order to constrain the relationship between
formation temperature, δ18Ocarbonate,
and δ18Owater. We observe a positive
relationship between measured and clumped-isotope derived
δ18Ow-reconstructed derived from the
composite freshwater calibration in this study (p = <0.0001;
r2 = 0.7935) and material-specific reconstructed
temperatures (p <0.0001; r2 = 0.8267).
Figure 4c shows comparison of the residuals for reconstructed
δ18Ow using both the material specific
and composite calibrations developed within this study. RMSE was reduced
using the material specific calibration relative to the composite
calibration for biogenic carbonates (material specific RMSE = 1.5‰,
composite RMSE = 1.8‰), micrites (material specific RMSE = 1.4‰,
composite RMSE = 1.7‰), and travertines (material specific RMSE = 2.1‰,
composite RMSE = 2.3‰). However, for biologically mediated carbonates,
the composite calibration slightly outperforms the material specific
calibration (material specific RMSE = 1.4‰, composite RMSE = 1.5‰),
although the mean of the residuals is closer to zero.
Overall, material-specific calibrations perform better at reconstructing
δ18Ow-measured (Figure 5). Out of the
87 samples from sites with measured
δ18Owater values, 33 samples fall
within ±1‰ of hydrographic data when using the material-specific
calibrations, a 7% improvement from using the composite calibration
(Fig. 5). δ18Ow-reconstructed values
for biologic samples generally recover
δ18Ow-measured within 2‰. However,
this method yields, for micrites, lower
δ18Owater values than observations for
more enriched δ18Ow-measured values,
which could be due to either kinetic effects and/or changes in surface
water chemistry during carbonate precipitation events (Fig. 5). If the
latter, it is unlikely to be due to evaporative enrichment of
δ18Owater which would produce the
opposite trend, but it may arise from changes in carbonate chemistry.
Biologically-mediated carbonates and travertines show a positive offset
from the 1:1 line, overestimating
δ18Owater relative to the measured
value, that may also arise from pH related effects on isotopic
fractionation or kinetic isotope effects (e.g., Beck et al., 2005;
Tripati et al., 2015).
3.4 Comparison of multiple materials at individual sites
Five sites in this study contain a combination of two different types of
materials analyzed. Three sites in China (Daija Co, Cuona Lake, and
Wulungu Lake) have micrite (Li et al., 2021) and freshwater mollusks
(this study), and both the North Arm and the South Arm of the Great Salt
Lake in Utah contain modern microbialite and ooid samples (Pace et al.,
2016). Comparing results from different carbonate types provides another
method of assessing overall calibration performance, therefore we
compare clumped isotope-derived water temperature estimates using both
material-specific and our composite freshwater calibrations at these
sites where different materials are present.
Figure 6 shows increased agreement between different sample types for
the same location for temperature and water δ18O in a
majority of cases when a material-specific calibration is used. The
lakes in China containing dual-materials are terminal lakes where
authigenic carbonate precipitation is expected to occur in the later
part of summer, when temperatures are most elevated and carbonate
supersaturation occurs in the surface waters (Hren & Sheldon, 2012).
Both Daija Co and Cuona Lake are high elevation lakes (>4.5
km) where the monthly average air temperature doesn’t exceed 0°C until
May, thus, temperature requirements for mollusk and micrite growth are
met during similar conditions, yielding similar estimated temperatures
between both archives when using the material-specific calibrations.
However, applying our composite calibration to samples in these lakes
results in higher calcification temperatures projected for mollusks
relative to the micrite. We find more of a disparity between
Δ47-derived and directly measured temperature values
when taking into account material type at Wulungu Lake, an inland, low
elevation and high latitude (47°N) lake with a large range of
intra-annual temperatures (~36°C), however, our
estimates of water temperature are in agreement with typical seasonality
of precipitation from the two archives given the lake’s setting. Our
results using a material specific calibration suggest that micrite
precipitated during a more narrow interval, with higher temperatures and
δ18O values than the mollusk samples, consistent with
temperature-induced carbonate precipitation and evaporative enrichment.
The temperatures recorded by mollusks at the same lake using the
material specific calibrations suggest that the shells may be reflecting
dominant calcification during comparatively cooler temperatures in
spring or early summer.
Modern microbialite and ooid samples from the Great Salt Lake yield
similar results, with a 1.5°C difference between archives in both the
North and South Arms using the material specific calibration. Both
archives in the Northern and Southern arms of the lake are consistent
with summer water temperatures when taking into account material type.
Furthermore, the separation of the North and South arms of the Great
Salt Lake by a causeway results in a more restricted Northern arm with
less freshwater input from rivers and more evaporation (Gwynn, 2007).
Our Δ47-derived temperature data also shows that our
carbonate samples from the Great Salt Lake record more elevated heating
in the North Arm relative to the South Arm when using the material
specific (0.6°C) compared to the composite (0.1°C) calibrations. Further
study is needed to fully understand the differences in calcification in
between archives in the same settings, but these findings may support
the use of a material-specific calibration when developing
reconstructions from multiple types of carbonates for ancient freshwater
systems.