BAYESIAN PARAMETER ESTIMATION IN A 1D MODEL OF PRECIPITATION AND EVAPORATION: COMPARISON OF MIDDLE MIOCENE AND MODERN CLIMATES USING PLANT LIPID DEUTERIUM (δD) MEASUREMENTS

Understanding ancient climates provides a tool for understanding our current climatic situation, and the changes that have been observed over the past hundred years. The Middle Miocene Climate Optimum, approximately 15 million years ago, represents a unique global warming period in Earth history when the high global temperature was accompanied by global atmospheric CO₂ concentration that was similar to present day. However, the global precipitation gradient during the time is poorly understood. Hydrogen isotopic signals (specifically molecular δD) from fossils and sediments offer insights into intrinsic precipitation data of ancient climates. In this work we develop a 1D phenomenological model that relates zonally averaged precipitation and evaporation to δD values, through a fractionation process. We apply Bayesian parameter estimation techniques, using Markov Chain Monte Carlo (MCMC), to fit this model to measured δD values. These values are obtained from n-alkanes extracted both from well-preserved Middle Miocene plant fossils and sediments, as well as modern samples, each taken from various latitudes across the Northern Hemisphere. From this model fit, we are able to infer the impact of the distribution of precipitation and evaporation on the climate throughout the Northern Hemisphere. In addition, the measured isotope data are able to constrain the magnitudes of the parameters of the model, and the Bayesian methods allow for the quantification of the uncertainty. We compare these magnitudes and uncertainties in both Middle Miocene and modern environments.