The power of present-day observations to constrain uncertainty in future climate projections

Background

The standard approach in climate modelling is to build and then “tune” a model to agree well with present day observations, and then to assume that if a model simulation looks like the real world today then that model will reliably simulate future climate change. This assumption turns out to be incorrect, and the problem leads to large but often undiagnosed “spread” (or uncertainty) in climate projections. For example, if five modelling teams tuned their models to agree equally well with present-day observations of cloud properties, the simulated cloud properties in the different models would diverge when simulating clouds in a 2050 climate. The cause of this problem is called “equifinality” – the phenomenon where, in a highly complex model, it is possible to adjust the model in multiple ways to agree equally well with observations (different model parameterizations, different parameter settings, etc.). This is fine if you are just trying to simulate present-day clouds, but these different models can respond differently to climate change. This is a fundamental challenge that may limit the extent to which we can understand and ultimately reduce uncertainty in climate projections.

PhD Opportunity

This PhD project aims to quantify the scale of the problem of climate model equifinality and then to find ways to address it. You will focus on the aerosol and cloud component of the UK Earth System model (UKESM) for which equifinality has been exposed as a challenge in previous studies. Aerosols are a major cause of climate change historically and in the future because they cause a large radiative forcing. Clouds are important because they cause a large climate feedback, amplifying future warming. Both have large uncertainty. To expose and quantify the importance of equifinality you will use perturbed parameter ensembles and emulators to generate essentially millions of model variants that can be evaluated against observations. You will also use extensive aerosol and cloud observations from in situ platforms and satellite data to identify the observationally equally plausible model variants, then explore how these diverge in simulations of future climate. You will also explore optimal ways of reducing the problem of equifinality, for example by developing ways to observationally constrain the model’s processes and responses rather than just state variables (like cloud thickness) that have been used so far.

Applicant profile

This PhD would suit most science- or maths-trained graduates.

More information

A recent article (in review) summarising perturbed parameter ensembles and the challenge of equifinality is https://egusphere.copernicus.org/preprints/2025/egusphere-2025-4341/.

Equifinality was first exposed in a global atmospheric model in a study of aerosols. See https://www.pnas.org/doi/10.1073/pnas.1507050113