Flexible paleoclimate age-depth models using an autoregressive gamma process

Analysis

Why This Paper Matters

This paper addresses a fundamental challenge in paleoecology: constructing accurate age-depth models from radiocarbon dates. Traditional methods often assume constant or piecewise linear accumulation rates, which can be unrealistic for sediment cores with varying deposition. By introducing a Bayesian autoregressive gamma process, the authors provide a flexible framework that captures natural variability in accumulation rates while incorporating prior knowledge. The use of the t-walk MCMC algorithm ensures robust sampling even with complex posterior distributions, making the method practical for diverse datasets. The paper's high citation count (3835) reflects its adoption as a standard tool in the field.

Technical Contributions

The key innovation is the gamma autoregressive model for accumulation rates, which allows for smooth but non-constant changes with depth. This is combined with a semiparametric approach that can handle an arbitrary number of subdivisions, providing flexibility without overfitting. The integration of informative priors is crucial for constraining the model when data are sparse. The use of the t-walk MCMC sampler is a practical contribution, as it automatically tunes proposal distributions, reducing the need for manual tuning. Additionally, the Student-t model for outliers robustly handles anomalous radiocarbon dates, which are common in sediment cores.

Results

The paper presents two case studies: a peat core and a lake core. For the peat core, Bacon produces age-depth models that are more realistic than those from classical methods (e.g., linear interpolation or polynomial fitting), with narrower credible intervals and better fit to the data. For the lake core, the method successfully identifies and downweights an outlier date, leading to a more coherent chronology. Quantitative comparisons show that Bacon's posterior estimates have lower deviance information criterion (DIC) values compared to alternative approaches, indicating better model fit. The method also provides full posterior distributions for ages at any depth, enabling uncertainty propagation in subsequent analyses.

Significance

This work has had a transformative impact on paleoecology by providing a statistically rigorous and user-friendly tool for chronology building. The Bacon software (implemented in R) is widely used, enabling researchers to produce reproducible and objective age-depth models. The methodological innovations—autoregressive gamma processes, robust MCMC, and outlier handling—are applicable beyond paleoecology to any domain requiring flexible time series modeling with uncertain observations. The paper exemplifies how Bayesian methods can solve practical problems in the earth sciences, bridging the gap between statistical theory and applied research.