PhD Scholarship - Machine learning for uncertainties of past ice-volume changes
Future projections of global sea-level rise are based on our understanding of changes in the past and these climate variations are preserved in the geological record in the form of proxies. An example are the calcite shells of benthic foraminifera found in marine sediment cores. Their oxygen isotope ratio, δ18O, records both changes in temperature and land-locked ice volume through time. However, the nature and potential time-dependence of this relationship between ice-volume and temperature and their uncertainty is not yet fully understood. This project seeks to address the role of proxy uncertainties inherent to the proxy formation, preservation, and analysis processes for our understanding of past ice-volume changes by use of Machine Learning methods.
Reconstructions of past climate changes and their uncertainty estimates can be derived by applying a transfer function that links a climate variable, such as bottom water temperature or ice volume changes, to paleoclimatic observations. The “LR04” stack of benthic δ18O records is one of the most widely used proxies from which we infer how ice-volume, sea level and global mean temperatures may have changed during the last 5.3 million years but such records come with different sources of uncertainty that we have to account for in a probabilistic way. The PhD project will integrate Machine Learning methods and Proxy System Models with various paleoclimate proxy records in order to develop a statistical framework that incorporates multiple sources of uncertainty associated with the proxy formation, preservation, and analysis processes. Such a probabilistic assessment of past temperature and ice-volume changes will allow us to make future sea-level projections that are well constrained by paleo-climate proxies and their uncertainties.
The PhD project combines data science and paleo-climate reconstructions and is thus highly interdisciplinary. The work builds on existing marine sediment core records and statistical methods, such as Bayesian inference and Machine Learning. The PhD project will focus on the Holocene period to establish a robust uncertainty estimate for the partitioning of the δ18O signal into an ice-volume and a bottom-water-temperature component for the major ocean basins, i.e., Atlantic, Pacific, Indian, and Southern Ocean. The time dependence of the partitioning and its uncertainty during the Pleistocene, Pliocene, and the Miocene will be tested in a probabilistic way to quantify the various sources of proxy uncertainty. The PhD combines statistical methods with geochemical, geological and paleoclimatic processes that act on many different time scales.
A fully funded PhD position is available at the Antarctic Research Centre (ARC), Victoria University of Wellington to use Machine Learning methods for improving our understanding of uncertainties of past ice-volume and global temperature inferred from geological proxy records. The successful candidate will be hosted by the ARC and collaborate with a large number of international scientists and graduate students from the ARC, GNS Science, and the Antarctic Science Platform. Being part of the new and exciting National Modelling Hub, the candidate has access to excellent technical support, high-performance computing, statistical modelling, climate modelling, and to collaborations with the University’s Data Science Programme.
The scholarship is open to applicants with an MSc degree and have an excellent grade/GPA.
- Applicants must have a MSc degree with an excellent grade/GPA.
- Applicants must have a background in one or more of the following fields:
- Data Science/Statistical Modelling/Applied Mathematics
- Marine sediment analysis
- Paleoclimate reconstructions/Earth Sciences
- Applicants will need to have basic programming skills (e.g. Python, R, Julia, R, Matlab)
In order to apply for this scholarship, applicants are required to send the following documentations to Dr Mario Krapp: firstname.lastname@example.org
- Cover letter (one page)
- Two references (who will be contacted by us)
The successful recipient will be selected by Dr Mario Krapp (email@example.com) and Prof Tim Naish (firstname.lastname@example.org), who will also be the student's principal doctoral advisors. Additional advisors may include A/Prof Robert McKay. Mario and Tim will assist the successful candidate with the Victoria University of Wellington graduate admission process.
For more information please email Dr Mario Krapp: email@example.com