A Gaussian process state-space model for atmospheric CO2 and sea surface temperature index reconstruction from boron isotope and planktonic δ18O proxies

It often occurs in practice that only a small number of observations are given for reconstructing past climate events in the field of paleoclimatology. State-space models can overcome such scarcity by giving priors to those hidden states to make them correlated to one another. Inferring multiple events simultaneously from various proxies to exploit their mutual dependency is another option. Here we present a Gaussian process state-space model to reconstruct both atmospheric CO2 and sea surface temperature index from boron isotope and planktonic delta O-18 proxies.