Local and Global Sea Level During the Last Interglacial: A Gaussian Process Approach
Abstract
Although its warmth was caused primarily by variations in Earth's orbital forcing, the Last Interglacial (LIG) stage (ca. 130-115 ka), with polar temperatures ~3-5 C warmer than today, serves as a partial analog for low-end future warming scenarios. Multiple indicators suggest LIG global sea level was higher than at present; based upon a small set of local sea level indicators, the IPCC Fourth Assessment Report inferred an elevation of ~4-6 m. Sea level is often mistakenly viewed as a simple function of changing global ice volume. This viewpoint neglects local variability, which arises from several factors, including the distortion of the geoid and the elastic and isostatic effects of lithospheric deformation by shifting ice masses. Accurate reconstruction of past global and local sea levels therefore requires integrating global data sets of local sea level indicators. Such attempts have rarely been made for periods predating the Last Glacial Maximum. To assess the robustness of the IPCC's global estimate and search for patterns in local sea level reflective of meltwater source, we have compiled a more comprehensive database that includes a variety of local sea level indicators, among them coral terraces, raised beaches, and sedimentological facies transitions, from sixty-five localities, as well as a global sea level record derived from oxygen isotopes. We generate a global synthesis from these data using a novel statistical approach based upon Gaussian process modeling. In Gaussian process modeling, as in the closely related geospatial technique of kriging interpolation, an estimate of a field value -- in this case, sea level -- at any arbitrary point of interest is derived from the data using a covariance function. We find an appropriate covariance function through a training process that employs pseudo-proxy data generated from a simple gravitational sea-level model under many plausible ice sheet histories. We use a Markov Chain Monte Carlo approach to account for geochronological errors. Analysis of synthetic histories suggest that the current data set is capable of resolving LIG global sea level with a 95% confidence interval of about ±2 m. Due to uncertainties in dating, it is unable to resolve robustly a short-lived mid-LIG lowstand like that proposed by several authors. We will present estimates of global and local sea level history during the LIG derived from this statistical approach.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2008
- Bibcode:
- 2008AGUFMPP43B1521K
- Keywords:
-
- 1641 Sea level change (1222;
- 1225;
- 4556);
- 4556 Sea level: variations and mean (1222;
- 1225;
- 1641);
- 4936 Interglacial