Paleo Data Assimilation of Coral δ 18 O Part 1: Best Practices and Uncertainties

Abstract Uncertainties in early 20th‐century observations limit understanding of the full range of internal climate variability and muddle understanding of climate system response to external forcing. Coral archives can provide high‐resolution information about historical sea‐surface temperatures (SSTs) and hydroclimatic change over the tropical oceans. Paleoclimate data assimilation (paleo‐DA) can extract information from these valuable records and generate gridded, full‐field reconstructions. However, reasonable choices in the paleo‐DA methodology can influence the reconstructed climate variability. Here, we outline the challenges and opportunities for using the oxygen isotope composition of corals (δ 18 O) to reconstruct temperature and hydrological cycle variability with paleo‐DA using the Last Millennium Reanalysis offline framework. Using a proxy network of coral δ 18 O and Sr/Ca that spans the 20th century, we evaluate the sensitivity of reconstructions to the choice of coral δ 18 O proxy system models (PSMs), the choice of instrumental sea surface salinity (SSS) products with the PSMs, and the choice of a climate model prior, ultimately testing 28 distinct reconstruction approaches. The choice of coral δ 18 O PSM and climate model prior can generate inconsistencies in the reconstructions. This influence is likely due to sparse observations of δ 18 O sw and SSS limiting “training” data, the nonstationary relationship between SSS and δ 18 O sw , and the strong covariability of SST and SSS across the global tropics. Given these limitations, we advise the use of a high‐resolution climate model prior, but the “best” configuration of PSM, SSS product, and prior is dependent on the question and region of interest.