IAP-26-025

Persistence, predictability and extremes in a changing climate

The impacts of human-caused climate change are now being experienced around the world through the increasing occurrence of extreme weather events, including heatwaves, droughts, and floods. These are often related to unusual conditions in the climate system that last several weeks to months – much longer than individual weather systems. For example, unprecedented heatwaves are commonly preceded by large soil moisture deficits arising from the prolonged absence of rain-bearing weather patterns. Summers which see the largest melting of the Greenland Ice Sheet are often accompanied by recurrent high-pressure weather systems, which bring strong sunshine and unusual warmth. Long periods with extremely high or low temperatures accompanied by little wind and/or sunshine can drastically reduce low-carbon electricity generation (Bloomfield et al. 2020), placing increasing pressure on electricity grids as renewables account for a growing share of generation. The wettest winters in Britain usually feature several weeks with a rapid succession of storms from the Atlantic, which can bring strong winds, devastating flooding and limit the ability of critical sectors such as energy and transport to recover between storms. These are examples of “persistence” in the atmosphere and its wide-ranging impacts on our environment.

To be able to predict these persistent events with sufficient lead-time, and to understand how they may be changing with climate change, requires a comprehensive understanding of the governing atmospheric dynamics. Yet such an understanding remains elusive. This leads to deficiencies in the representation of persistence of extreme weather events in models used for weather and climate prediction (Lee and Polvani 2026), which ultimately limits the ability to national critical infrastructure providers to prepare. Moreover, the rarity of extremes superimposed on the rapidly shifting background state of a changing climate means that the atmospheric, oceanic and land surface conditions required to generate extremes may not have even occurred in the current climate (Kay et al. 2025), leading to a reliance on potentially biased model simulations. Addressing these knowledge gaps therefore requires applying dynamical understanding to a systematic assessment of large numbers of model simulations.

The goal of this PhD project is to transform our understanding of the drivers and impacts of atmospheric persistence and its representation in models used operationally for prediction weeks to seasons to decades in advance. The project will yield results that can be incorporated into future model development and used operationally by forecasters. These outcomes will benefit the increasing number of sectors dependent on longer-term atmospheric forecasts, including the energy sector, transport, agriculture and water resource managers.

This project will use a hierarchy of tools and data, including simple and complex model simulations, analysis of increasingly large datasets of gridded observational and operational forecast data, and data-driven statistical methods, to addressing three key objectives: (1) Which is more important for persistence: local-scale processes or large-scale dynamics? (2) Does persistent weather always lead to persistent and extreme surface impacts? (3) What role does persistence play in the subseasonal-to-seasonal prediction of extreme events?

Initially, the project will focus on interrogating large ensembles of initialised prediction systems, including subseasonal prediction systems which contribute to the Subseasonal-to-Seasonal (S2S) database, seasonal prediction systems which contribute to the Copernicus Climate Change Service (C3S) archive, and the Met Office Decadal Prediction System (DePreSys; Dunstone et al. 2016). These ensembles will be analysed through the framework of large-scale weather regimes, focusing on Europe and North America (Lee et al. 2023). The large number of model realisations will enable a much better understanding of extremely rare and persistent configurations of the climate system, such as individual weather regimes that persist for over a month (Lee and Polvani 2026). Weather regimes have been previously shown to relate well to energy system extremes and therefore provide a key impact to examine (Bloomfield et al. 2020).

A next step will be to decompose the physical processes which set apart the most extreme events, considering the relative roles of large-scale dynamics versus local-scale processes. This will help develop understanding of the extent to which models differ in their treatment of extremes, including the role of biases. It is then anticipated that the student will subsequently design novel model experiments using the Met Office DePreSys system or the ECMWF OpenIFS model to test hypotheses developed through the project. These experiments could include nudging regions of the atmosphere, ocean or land surface toward observations, extreme ensemble members, or climatology, isolating their contributions to persistence. Examples of similar model experiments for a recent event that had a large impact on the energy sector are given in Mudhar et al. (2026).

Year 1

Develop understanding of atmospheric dynamics and climate science through an extensive literature review. Become familiar with handling and manipulating large ensembles of weather and climate prediction data. Begin to apply methods to quantify persistence within these data.

Year 2

Develop statistical frameworks to quantify persistence within observational and forecast data and link these to dynamical processes, extracting plausible extreme scenarios. Form hypotheses and test them in bespoke modelling experiments. Attend and present work at national workshops to network and share findings with the community.

Year 3

Using results obtained so far, quantify the role of persistence in extreme event prediction for a particular impact sector, considering the roles of local versus large-scale processes. Define the model dependency of the results through extensive data analysis. Present the key results at a major international conference, such as the European Geosciences Union annual meeting in Vienna.

Year 3.5

Write up and submit thesis, including manuscripts for publication.

The student will be trained in several aspects of physical climate science including atmospheric dynamics, ensemble prediction, climate modelling, climate impact modelling and climate change. Throughout the project, the student will be encouraged to closely monitor weather and climate events around the world for opportunities to lead research into their causes and predictability in the context of this project. The student will write up and publish results in academic journals, and attend conferences, workshops and training events to present their work and develop an international collaborative network. The student will also be trained in highly sought-after technical skills in computational modelling, high-performance computing, and quantitative ‘Big Data’ analysis.