Figure 6: Evaluation of clumped isotope derived temperature and δ18Owater for locations with dual materials. A. Clumped-isotope derived temperature reconstructions using material specific calibrations (top row; black frame) and composite freshwater calibration (top row; gray frame). Bottom panel shows residuals for both the material specific calibrations and composite calibrations (semi-transparent symbols), along with the average difference in temperature between the two archives (black and gray numbers and lines for material specific and composite calibration, respectively). Clumped isotope derived temperatures are denoted as TΔ47, while independently observed water temperatures are denoted as Tobservations. Sites are labeled on plot: D: Daija Co, C: Cuona Lake, W: Wulungu Lake, GSL NA: North Arm, Great Salt Lake, GSL SA: South Arm, Great Salt Lake. We find a reduction of temperature residuals in most cases using a material specific calibration. We also find more realistic temperature estimates when applying material specific calibrations, given each lake’s individual setting.  B. Reconstructed water δ18O values using material specific (top row, black frame) and composite (top row, gray frame) calibrations. Clumped isotope based reconstructions of water δ18O and independently measured δ18O values are denoted as water δ18Oreconstructed and water δ18Oobserved, respectively. The residuals from measured values are displayed on the bottom plot, and residuals between different archives are denoted by black and gray bars and values for the material specific and composite calibrations, respectively.
3.5 Applications to Paleoclimate Reconstructions
3.5.1 Paleoclimatology of Lake Surprise, CA
We applied the calibration relationships derived within this study to reconstruct terrestrial paleo-hydroclimate variables, using samples of tufa that are Last Glacial Maximum (LGM; 23,000-19,000 years ago) and deglacial (19,000-11,000 years ago) in age from Lake Surprise, California (Egger et al., 2018; Ibarra et al., 2014; Santi et al., 2019; Santi et al., 2020). In the original study from Ibarra et al. (2014), the authors used pollen data to estimate temperature changes during the LGM and deglacial. These estimates were used within an isotope mass balance model to derive evaporation and precipitation rates at Lake Surprise. Santi et al. (2020) expanded on this work by providing further constraints on water temperatures (and air temperatures, by use of a transfer function) and δ18Owater for the same sample set and used these updated values within an isotope mass balance modeling framework to derive new estimates of evaporation and precipitation rates. Here, we applied our biologically-mediated and composite calibrations and the recently published calibration by Anderson et al. (2021) to data from tufa samples for Lake Surprise from Santi et al. (2020). Data used from Lake Surprise is identical to the original publication and was reprojected into the I-CDES reference frame following current best practices and standardization procedures (Bernasconi et al., 2021; Upadhyay et al., 2021). Detailed methodology can be found in Santi et al. (2020) and equations used for this analysis can be found in the Supplementary Information.
New estimates of Δ47-temperatures, water δ18O, evaporation, and precipitation rates are shown in Supplementary Figure 1, and reported in Supplementary Table 5, 6, and 7 using the material-specific and composite calibrations derived in this study, and the Anderson et al. (2021) calibration, respectively. Updated temperature predictions using our biologically-mediated calibrations generally result in cooler water temperatures throughout the LGM and deglacial at Lake Surprise than the original publication (Table 4). Water temperatures derived using the Anderson et al. (2021) calibration are 1.2 and 1.7°C cooler than the material-specific calibration, during the deglacial and LGM, respectively.  Estimates derived using the general calibration are warmer than the material specific calibration, but slightly cooler than the original publication, with water temperatures estimated to be 12.5 ± 1.1°C during the LGM and 11.4 ± 0.7°C during the deglacial. Average LGM water temperatures of 11.5 ± 1.4°C are estimated using the biologically-mediated calibration, which results in temperatures that are 1.2°C cooler than the originally reported value (Santi et al., 2020) of 12.8 ± 1.4°C. Using the same water-to-air temperature transfer function (Hren & Sheldon, 2012) as used in Santi et al. (2020) to translate water temperatures to LGM MAAT, we find a 3.0°C difference between the material-specific values and published values (1.8 ± 1.9°C and 4.8 ± 1.3°C, respectively). The composite calibration estimates similar temperatures to the original publication (3.2 ± 1.5°C), while the Anderson et al. (2021) calibration estimates much cooler temperatures during the LGM (-0.5 ± 1.7°C). Our deglacial MAAT estimates follow a similar trend, with estimates from Santi et al. (2020) resulting in the highest MAAT estimates, followed by our composite calibration, the biologically-mediated calibration, and the Anderson et al. (2021) calibration (Table 4).
MAAT estimates were used to calculate annual evaporation rate, using a modified Penman equation (Linacre, 1993; Mering, 2015; Santi et al., 2020). Therefore, shifts in temperature will dictate the direction and magnitude of changes in estimated lake evaporation rate. When using temperatures derived from our material-specific calibrations, we observe a 91 mm difference in evaporation in our reanalyzed data (981 ± 60 mm/yr) and the original publication (1072 ± 56 mm/yr) during the deglacial. Our estimates during the LGM are 50 mm/yr lower than estimates from Santi et al. (2020), with our results suggesting 1103 ± 115 mm/yr and 1153 ± 100 mm/yr, respectively. Our derived weighted evaporation rate, which takes into account free evaporation over the lake surface and evapotranspiration over land, differs significantly from Santi et al. (2020), with a 46 mm/yr decrease and 41 mm/yr increase difference between the new and old estimates for the LGM and deglacial, respectively (Table 4). Since we observe general decreases in temperature using the Anderson calibration, overall, weighted and unweighted evaporation values are lower than those derived in the original publication, while the agreement in temperature from the composite calibration derived evaporation estimates are similar to the original publication (Supplementary Tables 6 and 7).
Evaporation estimates are used within the hydrological modeling framework of Santi et al. (2020) to reconstruct precipitation that incorporates clumped isotope derived MAAT and δ18Owater and includes basin hypsometry derived from modern topography and watershed delineations (Broecker, 2010; Ibarra et al., 2014, 2018; Jones et al., 2007; McGee et al., 2018). We also incorporate an ice-volume correction for δ18Owater during the LGM and deglacial based on sample age, following the approach used by Tripati et al. (2014) and estimate anomalies with a revised value for modern annual precipitation rates to account for snowfall. In Santi et al. (2020), modern annual precipitation values used to calculate anomalies used a snow water equivalent of 1:10 mm. However, snow densities vary depending on temperature and location. To account for these changes in snow density, we utilized 10 years of SNOTEL snow density data (USDA Natural Resources Conservation Service, 2022) at two stations within the Surprise basin (Cedar’s Pass and Dismal Creek) to assess typical modern snow water equivalent in Surprise Valley. The average snow density for the area was 33%, and we combined this value with average snowfall amounts to integrate the contribution derived from snow into our modern precipitation measurements, for an updated modern precipitation value of 630 mm/year (Desert Research Institute; https://www.dri.edu/western-regional-climate-center/), a 102 mm/year increase.
These updated calculations using the material-specific calibration indicate that during the LGM, precipitation rates were 10% lower than modern, and precipitation increased during the deglacial to values that were 8% lower than today (Table 4). Estimates of precipitation rates using all calibrations are similar, with estimates for the LGM ranging from 551 to 565 mm/yr (Table 4). We observe a slight increase in precipitation rate estimates when applying all calibrations during the deglacial, when Lake Surprise reached its maximum extent, with values using the original published calibration resulting in estimates of 567 mm/yr, which is 16, 1, and 8 mm/year lower than estimates derived from our material-specific, composite, and the Anderson-derived values, respectively. Overall, this supports the original findings of Ibarra et al. (2014) and those of Santi et al. (2020) based on clumped isotopes that temperature and evaporation were likely dominant controls on Lake Surprise’s transgression and regression, and that increased evaporation relative to precipitation contributed to the eventual disappearance of Lake Surprise.