Epistemology of weather and climate prediction: What certainty can we have in the future of our environment?

Background

This project will use mathematics, meteorology and philosophy to explore the epistemology of weather and climate prediction. When we forecast the future of our environment, where does our “knowledge” come from? How can it be shared?

We understand the Earth’s climate to be a chaotic dynamical system. The revolution in AI-based forecasting forces us to revisit questions on predictability and the relationships between mathematics, science, data and social context. Criteria to demonstrate confidence in predictions, as well as assessment and communication of their uncertainties, are needed for the development of information that is trustworthy and adequate for decision making. The project will use interdisciplinary tools, from mathematical and social sciences to foundational philosophical analysis. Through collaboration with weather and climate prediction centres, and the development of our own prediction tools, the results will influence operational practice.

PhD opportunity

The key challenge will be to provide a philosophically grounded and scientifically informed evaluation of how markers of trustworthiness of information can be applied consistently to a range of climate services, across timescales from 0 hours to 100 years.

Our philosophical questions have practical importance. Should we trust a statistical forecast based on past data less than a forecast based on equations of physics? How can we communicate trust to users? What about users’ own knowledge, such as a farmer’s knowledge of the soil moisture, relative to knowledge from a scientist or an AI algorithm? We need a framework applicable across contexts, in order to inform decision-making for climate adaptation. This involves interdisciplinary understanding of the epistemology of future risk, dynamical systems theory, data-based methods including AI, the integration of disciplines informing risk, and markers of trustworthiness. It includes ethical aspects of knowledge production, such as the equitability of communication, and the role of power structures in the development of information.

To make progress, the candidate may develop case studies, or analogues, of canonical high-impact phenomena such as seasonal droughts or floods, considering how our knowledge derived from observed climate influences knowledge of climate change. We may create and analyse our own predictions, applying statistical and epistemological analyses of their products. We may also consider idealised systems, such as the Lorenz equations. The candidate can use the analogues to inform the development of general criteria to assess the quality, trustworthiness and credibility of tools that are currently used to produce projections across different timescales. How can we improve equity in the creation and communication of predictions? Can we create frameworks for the quality-control of information from increasingly diverse sources?