Advanced methods for climate and health attribution Summer School

Developing analytical skills in Bayesian inference and climate attribution for environmental health applications.  

Climate change is one of the greatest challenges of our time and its effects on human health are already being felt. As we enter an era where climate change is increasingly recognised as a major threat to public health, there is a growing demand for researchers who can assess the health impacts of climate-related hazards. This summer school is essential for researchers keen to progress in this field. Participants will learn how we can scientifically link climate change to health outcomes using advanced methods from environmental epidemiology and climate attribution science.

The week will consist of two modules:

Module 1: Probabilistic climate attribution

is a two-day workshop designed for researchers in Climate Change and Environmental Health who are interested in learning about attribution science but have limited or no prior experience.

Module 2: Bayesian models for climate and environmental health

Is a three-day workshop aimed at researchers in Environmental and Climate Health who are keen to explore Bayesian modelling for environmental and climate epidemiology.

This summer school is designed for postgraduate students and researchers at any stage in their career who are interested in climate and health attribution and are keen to learn advanced statistical methods.

Module 1: Probabilistic climate attribution

Lecturers

Fredi Otto, Imperial College London, London, UK

Clair Barnes, Imperial College London, London, UK

Course description

Attribution science is a relatively recent branch of climate science, evaluating the extent to which anthropogenic climate change has altered the likelihood and intensity of extreme weather events around the world. This short course provides participants with the knowledge and tools to conduct attribution analysis for extreme weather events, with a particular focus on heatwave events. By the end of the course, students will have the basic skills to conduct their own attribution analyses and calculate factual and counterfactual temperatures to be used in mortality attribution. Topics covered will include:

• Methods in attribution science
• Visualising weather and climate data
• Defining a meteorological event
• Software tools for probabilistic attribution
• Interpreting attribution results

The training includes lectures and hands-on computer labs using real data, with time to discuss your own research questions.

Requirements

To participate, you should:

• Be familiar with conda or similar package managers for installing python and R packages
• Be familiar with spatial/temporal data and common distributions (e.g., normal, Poisson) - helpful but not required
• Bring your own laptop, all exercises will be done using Jupyter Notebook

Software You’ll Be Introduced To

• Python for NetCDF data - xarray, xclim, cartopy
• R

Module 2: Bayesian models for climate and environmental health

Lecturers

Garyfallos Konstantinoudis, Imperial College London, London, UK

Robbie Parks, Columbia University, New York, US

Course description

In environmental and climate epidemiology, spatial and spatiotemporal methods are increasingly used with high-resolution environmental and health data, which allows detailed insights into associations between exposures and outcomes. In this context, Bayesian methods provide a natural setting to incorporate uncertainty and prior knowledge, borrow information between neighbouring units, and create hierarchical structures.  However, appropriate usage of these methods requires careful consideration and knowledge of building technical models, which can be intimidating to the uninitiated user. This course provides a practical and approachable introduction to:

• The basics of Bayesian inference
• Hierarchical models
• How to choose priors
• Working with different types of data (spatial, temporal, continuous, categorical)
• Software tools for Bayesian analysis
• Real-world examples and how to apply these methods to your own research
• Bayesian distributed lag non-linear models

The training includes lectures and hands-on computer labs using real data, with time to discuss your own research questions.

Requirements

To participate, you should:

• Know the basics of R and RStudio (e.g., installing packages)
• Be familiar with spatial/temporal data and common distributions (e.g., normal, Poisson), helpful but not required
• Bring your own laptop, all exercises will be done using RStudio Cloud

Software You’ll Be Introduced To

• BUGS / OpenBUGS / WinBUGS
• R-NIMBLE
• R-INLA

Presenters