Coral based reconstructions of past Sea Surface Temperature – Insights into uncertainty

The past decades have seen the development of a wide range of global gridded Sea Surface Temperature (SST) products documenting historic changes in SST based on in situ and remotely sensed data. However, there are many sources of uncertainty in these records and significant differences exist among them. This thesis uses the record of past climate contained within tropical corals to advance our understanding of these uncertainties. Specific aims of this thesis were to perform: an investigation into the relationship between the local temperature at the coral site and gridded SST, a quantification of the total uncertainty involved in coral derived SST, and finally the assessment of existing SST products in light of new coral-based estimates of historic SST change. Through intercomparisons among SST products with data from temperature loggers, the discrepancies between gridded and in situ temperatures were quantified with a focus on coral reefs. This offers insight into the limitations of existing coral reef monitoring efforts and informs the process of estimating the uncertainty in coral SST reconstructions. The process of deriving SST through coral proxies was then scrutinized, the uncertainty of each step was quantified, and the total uncertainty was estimated. The total uncertainty in a single coral SST timeseries was time dependent and for the median uncertainties ranged from 0.1 to 1.90 °C (2sd). A method to reconstruct regional SST from several corals was then applied to critically assess the accuracy of three major SST products. This independent assessment verified the warm bias of World War II (WWII) anomalies others have shown and revealed that at the beginning of the historical SST records, the variance in SST anomalies is likely overinflated. A maximum bias above 0.5 °C was estimated for ERSSTv5, while the biases of HadISST and HadSST4 were over 0.4 °C, during the WWII interval. The estimates of global and regional surface temperature trends of the SST products for 1870-1990 were also found to be overestimated by up to 0.04 °C/decade. It was shown that coral SST reconstructions can resolve issues in the historical records, especially in periods and areas of high uncertainty. Finally, the potential to further constrain the uncertainty of coral reconstructions was highlighted, which opens the door for better reconstructions, extending to the pre-industrial era, and across a wide range of paleoclimate studies