Bayesian Inference of Absolute Ages for Chinese Neogene Mammalian Faunas

Determining the relative age of strata based on the evolutionary level of fossil mammalian faunas is a fundamental research method in Cenozoic biostratigraphy and chronostratigraphy. However, the inability to provide absolute ages from the faunas alone has long constrained the development of biostratigraphy. Against the backdrop of rapid advancements in geophysical dating techniques, biochronology is increasingly becoming marginalized. To address this, we have developed a new model based on the Bayesian Markov Chain Monte Carlo (MCMC) method, which requires only very limit absolute age information to infer the absolute age distribution of a sequence of faunas. The model follows Bayesian theorem, 𝑓𝑓(𝐴𝐴|𝐵𝐵) = 𝑓𝑓(𝐴𝐴) ∗𝑓𝑓(𝐵𝐵|𝐴𝐴)/𝑓𝑓(𝐵𝐵) , where A represents the sequence of faunas, and B represents all the mammalian taxa contained within these faunas. The model aims to infer the distribution of A through the distribution of B. It assumes that any mammalian taxon in B follows a normal distribution over its temporal range, and that the standard deviation of this distribution has a linear positive correlation with the total number of occurrences of that taxon across all faunas. 𝑓𝑓(𝐵𝐵|𝐴𝐴) represents the joint probability density of all taxa in B given the known age sequence of the faunas. Assuming the taxa are independent of each other, its value is the product of the probability densities of each taxon.𝑓𝑓(𝐴𝐴) represents the prior joint probability density of the fauna ages, obtained from the product of the prior probability densities for each individual faunal age. During the simulation process, we randomly assign initial ages to each fauna and perform sampling through MCMC iterations. After convergence to a stationary distribution, we obtain the estimated age distributions for each fauna. Using this method, we calculated the absolute age distributions for 145 mammalian faunas in China. Apart from 11 faunas from the late Paleogene and early Quaternary, the remaining 134 faunas cover the vast majority of reported Chinese Neogene faunas. Relying on age distributions of very limit taxa that are well-dated, and giving a very broad prior distributions for each fauna, the model yielded relatively reliable posterior age distributions for each fauna. Based on these results, we refined the existing Chinese Neogene Mammalian Unit (NMU) system into a sequence comprising 18 mammalian unit zones, detailed as follows: NMU1 ~23–22.5 Ma, NMU2 ~22.5–21.5 Ma, NMU3 ~21.5–19.6 Ma, NMU4a ~19.6–18 Ma, NMU4b ~18–16 Ma, NMU5 ~16–14.5 Ma, NMU6 ~14.5–13 Ma, NMU7a ~13–12.1 Ma, NMU7b ~12.1–11.2 Ma, NMU8 ~11.2–9.8 Ma, NMU9 ~9.8–8.7 Ma, NMU10a ~8.7–7.6 Ma, NMU10b ~7.6–6.8 Ma, NMU11 ~6.8–6.0 Ma, NMU12a ~6.0–5.3 Ma, NMU12b ~5.3–4.0 Ma, NMU13 ~4.0–2.6 Ma, NMU14 ~2.6–1.8 Ma (extending into the earliest Pleistocene). Each mammalian unit is supported by representative faunas. This method, capable of quantitatively estimating the age distributions of faunas and their bearing strata even in the absence of reliable absolute ages, provides a new constraining tool for addressing age controversies, such as those arising from paleomagnetic correlations. It holds the potential to unlock new avenues for biochronology from the classic research approach of mammalian fauna correlation.