A Paleoclimate Reconstruction on the Common Era (1-2000AD) was performed using a Hierarchical Bayesian Model from three types of data: proxy data from PAGES2k project dataset, HadCRUT4 temperature data from the Climatic Research Unit at the University of East Anglia, and external forcing data from several sources. Five data reduction techniques were explored with the purpose of achieving a parsimoneous but sufficient set of proxy equations. Instead of using the MCMC approach to solve for the latent variable, we employed an INLA algorithm that can approximate the MCMC results and meantime is much more computationally efficient than MCMC. The role of external forcings was investigated by replacing or combining them with a fixed number of BSplines in the latent equation. Two different validation exercises confirm that it is feasible to improve the predictive ability of traditional external forcing models.
10/14/2018 ∙ by Luis A. Barboza, et al. ∙ 0 ∙ share
Luis A. Barbozais this you? claim profile