.png)
189 – Hybrid or Online Teaching and Quantitative Reasoning
A Bayesian Hierarchical Chronology Model for Time Series Analysis of Paleoenvironmental Data
Aaron Springford
Queen's University
Time series data consist of observations collected sequentially in time. The analysis of time series is well developed in cases where the sampling times are known and evenly spaced. However, in some cases we may not observe the times directly and are forced to treat them as latent. Common examples include paleoenvironmental studies based on core samples, such as paleolimnology. Paleolimnological core data consist of sediment samples collected at sequential depths. A subset of the samples are then dated using radio-isotopes or other methods. A chronology model that relates the observed time proxy (depth) to time is required before any time series analysis can be undertaken. Chronology models used in the literature tend to favour piecewise linear formulations, usually by interpolating between depths with available dating estimates. I present a chronology model based instead on modelling the accumulation of the core. A Bayesian hierarchical approach results in probabilistic chronology estimates that are robust to outliers common to radio-isotope analysis, and allows uncertainty in timing to be applied directly to subsequent time series analysis.