Peter Hailey
Research Subjects
Peter Hailey has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
Anne M Khisa
1 shared publicationKaty Sheen
1 shared publication- Matt Fortnam
Senior Research Fellow University of Exeter United Kingdom
1 shared publication Nancy Balfour
1 shared publicationSteven Squires
1 shared publicationTheo Economou
1 shared publication
Latest Publications
- Open Access
- 17 August 2026
- Published by Elsevier in SSRN Electronic Journal
A Narrative Review of Machine Learning Applications for Prediction and Projection in the Climate-Health Domain
Predicting how a changing climate will affect human health is one of the more consequential, and more difficult, forecasting problems in applied climate science: target variables are noisy, predictors span both climate and non-climate domains, and the underlying systems are nonstationary and only partially observed. Machine learning (ML) is increasingly used to generate predictions and projections of climate-related health outcomes, potentially enabling earlier intervention, but its effective application in this domain is challenging and often undertaken by researchers without a strong background in either ML or the climate-health field. We provide a narrative review of how ML is applied across the domain, giving particular attention to aspects common to climate prediction more broadly: forecast horizon, non-stationarity, spatial and temporal resolution, and the quantification of predictive uncertainty. Most applications forecast one of three outcomes: healthcare utilisation, disease incidence/spread, or malnutrition. Predictors are heterogeneous, often not public, and noisy with substantial missing data; pre-processing, model choice and uncertainty reporting all vary widely, and predictive performance ranges from encouraging to disappointing. We review the health indicators, climate and non-climate inputs, and ML approaches used across the literature, together with key aspects, challenges and future directions for producing uncertainty-aware, actionable climate-health predictions.
- 17 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Narrative Review of Machine Learning Applications for Prediction and Projection in the Climate-Health Domain
Predicting how a changing climate will affect human health is one of the more consequential, and more difficult, forecasting problems in applied climate science: target variables are noisy, predictors span both climate and non-climate domains, and the underlying systems are nonstationary and only partially observed. Machine learning (ML) is increasingly used to generate predictions and projections of climate-related health outcomes, potentially enabling earlier intervention, but its effective application in this domain is challenging and often undertaken by researchers without a strong background in either ML or the climate-health field. We provide a narrative review of how ML is applied across the domain, giving particular attention to aspects common to climate prediction more broadly: forecast horizon, non-stationarity, spatial and temporal resolution, and the quantification of predictive uncertainty. Most applications forecast one of three outcomes: healthcare utilisation, disease incidence/spread, or malnutrition. Predictors are heterogeneous, often not public, and noisy with substantial missing data; pre-processing, model choice and uncertainty reporting all vary widely, and predictive performance ranges from encouraging to disappointing. We review the health indicators, climate and non-climate inputs, and ML approaches used across the literature, together with key aspects, challenges and future directions for producing uncertainty-aware, actionable climate-health predictions.