A Gaussian process state-space model for sea surface temperature reconstruction from the alkenone paleotemperature proxy

Abstract

Reconstructing past climate events relies on the relevant proxies and how they are related. Depending only on such relationships, however, could not be robust because only few proxy observations are usually available at each age. A state-space model employs a prior to make the hidden past climate events correlated with one another so that extreme inferences are precluded. Here, we construct a Gaussian process state-space model for reconstructing past sea surface temperatures from the alkenone paleotemperature proxy and apply the model to nine sediment cores with three different calibration curves and compare the results.

Keywords

• Alkenone paleotemperature proxy
• Gaussian process state-space model
• $ {\mathrm{U}}_{37}^{\mathrm{K}\prime } $ calibration curves