A BAYESIAN STATISTICAL MODEL FOR RECONSTRUCTING AND ANALYSING FORMER SEA LEVELS

Abstract

In order to understand the present we must first gain insight into the past. Therefore, to understand and have historical context for current rates of sea-level rise we need to be informed about past changes that have occurred. Sea-level reconstructions can provide this information by giving us insight into the magnitude and rates of past sea levels. We have produced sea-level reconstructions along the U.S Atlantic East coast using biological and geochemical sea-level indicators preserved in dated cores of salt-marsh sediment. I have developed statistical models that can help us to bridge the gap between the information held in these raw proxy data and a high-resolution sea-level reconstruction. Using a Bayesian framework for these models aids in the understanding and quantification of the uncertainty that is inherent in these data and the resulting records of former sea levels. I present A Bayesian transfer function modeling approach that is used to produce reconstructions of past sea level through the calibration of a biological proxy (e.g., foraminifera) into tidal elevation. The first step in the transfer function approach is building a model that captures the relationship between a biological proxy and tidal elevation in a modern environment. The second step uses this relationship to produce estimates of paleo-tidal elevation with uncertainty for each layer in a sediment core. Additional proxies (e.g., δ13 C) can be used to further constrain these estimates and potentially reduce uncertainty. Combining output from the Bayesian transfer function with a core chronology provides us with a reconstruction of relative sea level through time. With the aim of estimating rates of sea-level change, reconstructions are analyzed using an errors-in-variables integrated Gaussian process model. Ultimately, through the combination of these statistical models we can capture the continuous and dynamic evolution of rates of RSL change with a full consideration and propagation of available uncertainties. Results show that 20th century sea-level rise along the U.S. Atlantic coast is the highest it’s been in at least the last 15 centuries.

Date
Jan 14, 2021 1:00 PM
Event
Meeting of the Irish Mathematical Society
Location
Online
Niamh Cahill, PhD
Niamh Cahill, PhD
Assistant Professor, Statistics

My research interests include Bayesian Hierarcichal Modelling; Time Series Analysis; Climate Change; Family Planning