4.3. Practical implications for coral-based SST reconstructions
Most if not all fossil corals are subject to some level of diagenetic alteration, motivating the development of an evidence-based protocol for acquiring accurate paleotemperature estimates from this archive. This is especially true because, except in the most extreme cases, diagenesis can escape detection via standard screening protocols. While alteration is often associated with older fossil corals, diagenesis routinely occurs in relatively young corals, such as the <100yr-old fossil coral featured in this study, and is also observed in cores collected from living colonies (e.g. Nothdurft et al. , 2007;Nothdurft and Webb , 2008; Rabier et al. , 2008;Nurhati et al. , 2011; Sayani et al. , 2011). As the composition of secondary cements is significantly different from that of coral aragonite, even light to moderate levels of alteration can produce significant errors reconstructions that rely on bulk analytical techniques. Moreover, given the scarcity of fossil corals, researchers often attempt to recover reconstructions from corals that grew during a given time period of interest, even if they have experienced moderate to significant alteration. Thus, given the pervasiveness of diagenesis and its substantial impact on accuracy of fossil coral Sr/Ca-based paleotemperature estimates, rigorous screening for diagenesis would ideally be combined with micro-scale SIMS Sr/Ca as an independent constraint on paleotemperature.
SIMS Sr/Ca of coral skeletons allows for the selective analysis of less altered skeletal material, but the cost- and labor-intensive nature of these analyses requires a sampling strategy that maximizes the accuracy of the resulting paleotemperature reconstructions with as few analyses as possible. Our results suggest that the most strategic use of SIMS Sr/Ca involves analysis of discrete multi-year windows to check the accuracy of longer bulk coral Sr/Ca timeseries generated by ICP-OES. In doing so, it is important to consider how to achieve the following goals with a minimum number of SIMS analyses: collecting enough measurements in a given time interval to overcome microscale compositional heterogeneity in coral aragonite, and collecting timeseries of SIMS measurements that will provide an apples-to-apples comparison to bulk Sr/Ca measurements. By sampling ~3-4 samples per month of coral growth, and applying a ~2.5 month running mean filter, we obtained monthly-resolved Sr/Ca with an analytical uncertainty of ±0.04mmol/mol or ±0.6˚C (1σ). While this is adequate for validating bulk Sr/Ca, it may not be sufficient for exploring paleotemperature changes without accompanying longer, bulk Sr/Ca timeseries. Increasing the number of SIMS Sr/Ca analyses per month of coral growth would indeed reduce the analytical uncertainty of SIMS-based Sr/Ca timeseries. For example, ~20 analyses/month would yield uncertainties as low as ±0.022mmol/mol (1sigma) or ±0.3˚C in smoothed SIMS Sr/Ca, assuming a Gaussian distribution, however this may not be practical in most cases. Additionally, paleo-SST reconstructions from fossil corals must also contend with intercolony offsets, such as the ±0.08mmol/mol or ±1˚C (1σ) offset observed in SIMS data from the three modern corals studied here, that make it difficult attribute differences in mean Sr/Ca values between two corals to climate change (e.g. Sayani et al. , 2019). As such, given that multiple corals are needed from a desired time interval to derive robust estimates of mean paleo-SST change, the most practical use of SIMS coral Sr/Ca is to validate the accuracy of bulk coral Sr/Ca timeseries.