Assessing the Contribution of Sea Surface Temperature and Salinity to Coral δ18O using a Weighted Forward Model

Abstract:

The isotopic ratio of 16O/18O (δ18O) in coral carbonate skeletons is a robust, high-resolution proxy for sea surface temperature (SST) and sea surface salinity (SSS) variability predating the instrumental record. Although SST and δ18O-water (correlated to SSS) variability both contribute to the δ18O signal in the coral carbonate archive, the paucity and limited temporal span of SST and SSS instrumental observations limit the ability to differentiate respective SST and SSS contribution to each δ18O record. From instrumental datasets such as HadISST v.3, ERSST, SODA, and Delcroix (2011), we forward model the δ18O (“pseudoproxy”) signal using the linear bivariate forward model from Thompson 2011 (“pseudoproxy”= a1(SST)+a2(SSS)). By iteratively weighting (between 0 and 1 by 0.005) the relative contribution of SST and SSS terms to the δ18O “pseudoproxy” following Gorman et al. 2012 method, we derive the percent contributions of SST and SSS to δ18O at each site based on the weights that produce the optimal correlation to the observed coral δ18O signal. A Monte Carlo analysis of error propagation in the weighted and unweighted pseudoproxy time series was used to determine how well the weighted and unweighted forward models captured observed δ18O variance. Across the south-western Pacific (40 sites) we found that SST contributes from less than 8 to more than 78% of the variance. This work builds upon this simple forward model of coral δ18O and improves our understanding of potential sources of differences in the observed and forward modeled δ18O variability. These results may also improve SST and SSS reconstructions from corals by highlighting the reef areas whose coral δ18O signal is most heavily influenced by SST and SSS respectively. Using an inverse approach, creating a transfer function, local SST and SSS could also be reconstructed based on the site-specific weights and observed coral δ18O time series.