Disentangling causal interactions in last ice age model simulations

There is currently no consensus on the causal linkages between the climate system processes that together shaped Earth's ice age climate. Processes in the North Atlantic (NA) region, including meltwater release or sea-ice dynamics, are considered a key influence on the iconic Dansgaard-Oeschger (D-O) oscillations evident in a wide range of paleoclimate reconstructions. Another prominent feature of the glacial climate system is the 'bipolar see-saw', the idea that reduced heat transport to the NA due to weakening of the Atlantic Meridional Overturning Circulation (AMOC) during D-O cold events caused warming in the Southern Hemisphere (SH). Recent Earth system model simulations have been able to generate spontaneous millenial-scale (D-O) oscillations, with CO2 acting as a control on the climate transitions, which raises the prospect of improved mechanistic understanding of the climate system. However, extracting causal-mechanistic explanations from these numerical simulations remains a challenge, due to nonlinearity and noise, and standard approaches typically rely on analysing leads and lags.

Supervisors

Main supervisor: Bjarte Hannisdal, GEO-UiB
Co-supervisor: Jo Bendryen, GEO-UiB

Project description

The goal of this project is apply new methods to quantify causal relationships from time series of model output to test different causal hypotheses for D-O oscillations. Analyses will be done using the JuliaDynamics software ecosystem, featuring tools for inferring networks of causal relationships from time series. The data will come from published output of 8,000 year integrations of the Community Earth System Model (CESM) with glacial boundary conditions that reproduce the structure of D-O events, from which we can extract time series of the key variables hypothesized to influence AMOC strength (e.g. North Atlantic Oscillation, salt transport, Antarctic Bottom Water, sea ice fraction, surface freshwater flux). These time series will be analyzed using our software packages ComplexityMeasures.jl and Associations.jl, to infer the network of causal relationships, estimating the strength, directionality, and lag time of causal connections among the variables.

Field-, lab- and analysis work

For the project the student will use published model output from Community Earth System Model simulations. Causal analyses will be done using open-source high-performance software.

Proposed course plan

GEOV222 Paleoclimatology (10 ECTS - fall)
GEOV230 Glacial geology and geomorphology (10 ECTS - fall)
GEOV231 Marine Geological Field-and Laboratory Course (10 ECTS - spring)
GEOV302 Data analysis in Earth Science (10 ECTS - spring)
GEOV342 The Geochemical Toolbox (10 ECTS - spring)
GEOV324 Polar palaeoclimate (5 ECTS - fall)

Last updated: 23.06.2026