Testing the ability of AI-based regional models to capture changes in storm tracks using climates of the geologic past

Background

Before the Anthropocene, climate changed naturally over the course of thousands of years. However, without instrumental observations of it, we only have climate reconstructions from proxy records. These show interesting changes in the seasonal storm tracks driven by subtle variations in the Earth’s orbit. Previous global climate models, based on solely on physical laws, are uncertain about the storm track changes – partly because of systematic biases arising from the models’ coarse spatial resolution. Machine-learning offers the potential to overcome this resolution problem through ‘regional downscaling’. Yet this kind of machine-learning requires large training sets of observed weather conditions, which it then uses to extrapolate into a future with a changed climate. Simulating known past climate changes can be useful way to assess the validity of this extrapolation and to estimate the uncertainties associated with it.

PhD Opportunity

Future projections with accurate uncertainties are essential for climate adaptation decisions. These come from coordinated international efforts with an ‘ensemble’ of multiple state-of-the-art climate models. We can test the spread of this ensemble by exploring how well it captures the range of climates seen in past warm periods. The ongoing effort (CMIP7) includes several such experiments, such as simulating the climate of 6,000 years ago (the mid-Holocene) and of 127,000 years ago (the last interglacial). These times both saw seasonal variations in the amount of incoming solar radiation, which led to shifts in the mid-latitude storm tracks over Eurasia that can be detected in climate proxy reconstructions that come with their own uncertainties around dating and interpretation.

This project would start off by looking the set of simulations that are currently underway under the international Paleoclimate Modelling Intercomparison Project (PMIP), with a particular focus on the storm tracks. They would be compared against to proxy reconstructions to see if there is sufficient information to robustly discriminate between their representations. The second phase of the project would then explore the impact of increasing the spatial resolution on the ability to capture these past changes, through analysis of paired simulations performed by the USA’s National Center for Atmospheric Research (NCAR). These high-resolution simulations required substantial computing resources, and machine-learning offers the possibility of ‘downscaling’ to an even higher resolution for a fraction of the cost. The third and final part of this project will explore how appropriate such approaches are for past climates, by deploying the UK Met Office’s AI4Climate downscaling model and comparing its results to both the existing high-resolution simulation from NCAR and the lower-resolution PMIP experiments.

Applicant Profile

Students who excel at any sufficiently numerate subject are encouraged to apply; such as physics, mathematics, Earth sciences, geography, computer science or a related field. Prior education in climate or data science is not needed, although it would be helpful. An established interest in the topic is obviously vital, and a curiosity to learn across research disciplines. Some previous experience with scientific programming is required, and good oral and written communication skills are essential. Relevant professional experience and/or an MSc qualification would increase your competitiveness, but is definitely not a requirement.

Other information

You would join the Palaeoclimate Modelling Intercomparison Project community, of which Chris is a co-chair: https://pmip.lsce.ipsl.fr/about_us/overview

You can learn more about storm tracks and the global climate from the Met Office: https://weather.metoffice.gov.uk/learn-about/weather/atmosphere/global-circulation-patterns