Refining the alkenone-pCO2 method: constraining algal physiology using Bayesian inference
Abstract
The alkenone-CO2 method has been widely used to estimate atmospheric CO2 levels over the Cenozoic. It is based on the concept that the stable carbon isotopic fractionation during photosynthesis of coccolithophores is controlled by aqueous CO2 concentrations and a physiology parameter 'b' determined by factors such as cell size, growth rate, cell membrane permeability, and the possible operation of carbon concentrating mechanisms. Recent studies have improved the accuracy of alkenone-CO2 estimates by better constraining 'b'. For example, the "size rule" expressed by the overall correlation between phytoplankton size and their growth rate was used to calculate 'b'. Importantly, coccolithophore cell size can be estimated from nannofossils preserved in marine sediment, providing constraints on the physiology of ancient alkenone-synthesizers. However, numerous possible relationships exist linking cell size to growth rate and linking cell volume to cell carbon content. It is difficult to determine which one(s) of these relationships should be used for the alkenone-CO2method. Here, we intend to solve this issue by using advanced statistics and new data from the late Pleistocene. By coupling the values of the parameters involved in these relationships, and newly measured/previously reported geochemical and micro-paleontological data covering the past 800 thousands of years, we built a Bayesian statistical model to estimate the probability distribution of these parameters via a Markov Chain Monte Carlo (MCMC) sampling approach. The calculated CO2 using Bayesian-determined parameters agree well with the ice-core record, validating our statistical model. Also, distributions of these parameters do not vary with different MCMC sampling methods, the number of the training samples, or different time intervals (glacial or interglacial) that the training data sets are sampled from. This highlights the robustness of the newly constructed probability distributions of these parameters and boosts our confidence in using the Bayesian approach. Applying this new set of parameters beyond the last 800 thousands of years ago, the newly established Pleistocene alkenone-CO2 estimates agree well with atmospheric CO2 levels derived from older Antarctic blue ice.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFMPP12D0667L