Refining temperature reconstructions with the Atlantic coral Siderastrea siderea

• Sr/Ca and δ18O in Siderastrea siderea are reproducible within analytical precision.
• Sampling off the corallite wall of S. siderea produces higher (colder) Sr/Ca values.
• Horizontal and vertical sampling paths within the colony produce similar coral Sr/Ca.
• Linear extension rates and δ13C in S. siderea are not reproducible among colonies.
• No growth-related effects in Sr/Ca for extension rates > 2.1 mm year− 1 in S. siderea.

Developing coral-based temperature reconstructions for a particular coral species requires determining the optimal sampling path orientation and resolution for geochemical analysis to avoid sampling artifacts and to increase reproducibility. Furthermore, a robust coral archive should have high intracolony and intercolony reproducibility for determining the common environmental signal. Here we assessed sampling path orientation and sampling resolution for Siderastrea siderea colonies within the Dry Tortugas National Park in the Gulf of Mexico (24°42′N, 82°48′W) to determine the optimal sampling protocol and to assess reproducibility of coral Sr/Ca, δ18O, and δ13C. We identified a sampling artifact due to extracting samples from the coral columella resulting in cold bias up to 5.2 °C in coral Sr/Ca. We found no shift to higher coral Sr/Ca values (i.e., colder) for years with a 50% reduction in average extension rate (2.1 mm year− 1) or for sampling along paths up to 70° off the vertical axis of the colony. Our sampling resolution comparison (1900–1993) indicated that the resolution of ~ 6 samples year− 1 used in a previous study for coral Sr/Ca and δ18O may not capture seasonal extremes and thus produces muted seasonal cycles, but that resolution is not biased towards one season. Reproducibility or average deviations, assessed using absolute differences (AD) and root mean square (RMS), among the monthly resolved coral Sr/Ca records for intracolony to intercolony comparisons were within 2σ of our analytical precisions. Average deviations were reduced by 19 to 61% when assessing interannual variability (36-month smoothed and mean annual) suggesting that subannual dating uncertainties (i.e., assigning a coral Sr/Ca value to a particular month) were the largest source of error in our monthly resolved coral Sr/Ca reconstruction. Similarly, coral δ18O was reproducible within 2σ of our analytical precision (AD = 0.10‰ and RMS = 0.07‰); however, coral δ13C and linear extension records were not reproducible. Our assessment of coral geochemical variations from multiple S. siderea colonies suggests this species is suitable for paleoclimatic reconstructions, including subfossil corals and microatoll colonies that grow laterally.