Replication materials for "Increasing risk of mass human heat mortality if historical weather patterns recur"

 This repository provides code and data for "Increasing risk of mass human heat mortality if historical weather patterns recur," published in Nature Climate Change, doi:10.1038/s41558-025-02480-1. You can read a preprint here: https://eartharxiv.org/repository/view/8375/

## Organization

The repository is organized into Scripts/, Figures/, and Data/ folders.

- Data: This folder includes intermediate and processed summary data that enable replication of most the figures and numbers cited in the text. Some of the files for the climate model projections, raw observational data, and damage estimates are quite large, so they are not provided here but are publicly accessible or can be easily reproduced with the provided code. Details are below.

- Scripts: Code required to reproduce the findings of our work is included in this folder. Scripts are written in Python/Jupyter and R. The Machine_Learning_Analysis subfolder within this folder contains the code necessary to reproduce the machine learning training and predictions, adapted from Trok et al. (2024), https://www.science.org/doi/10.1126/sciadv.adl3242. There is a separate README within this folder which contains more specific details for reproducing this component of the analysis should you wish to do so. 

- Figures: This is where figures will be saved if you run the scripts. There may be small differences between these figures and the final publication figures due to post-processing in Adobe Illustrator.

## Data

Much of the intermediate data required to reproduce the final figures and numbers in the paper are provided in the various folders within the Data directory, including the overall panel dataset that includes district-level mortality and temperature. However, much of the initial/raw data is too large to be hosted here. They are all publicly available at various locations:

- The E-OBS station-based observations are available here: https://surfobs.climate.copernicus.eu/dataaccess/access_eobs.php#datafiles

- The Eurostat mortality and population data are available here: https://ec.europa.eu/eurostat/data/database?node_code=demomwk. Weekly mortality is denoted by the `demo_r_mweek3` code and population is denoted by the `demo_r_pjanaggr3` code. 

- CMIP6 monthly temperature data (used for the adaptation analysis) is generally available from the Earth System Grid Federation (e.g., https://aims2.llnl.gov/search/cmip6/). Our analysis uses monthly temperature (tas_Amon) from the historical and SSP3-7.0 scenarios.

## Scripts

Each script performs one component of the analysis.

- Aggregate_EuropeWide_E-OBS_Data.py takes the raw E-OBS input data and aggregates it to district-level (NUTS) boundaries at the daily and weekly level. This script was run remotely on an HPC system and may require significant memory.

- Construct_Eurostat_Panel.R combines the E-OBS district-level data with matching weekly mortality and temperature data and produces a final panel dataset spanning 2000-2022.

- Europe_Mortality_Regressions.R performs the regression analysis to derive the exposure-response functions.
  
- Plot_Event_Timeseries.ipynb plots and sets the time periods of each event.

- Counterfactual_Events.py combines the E-OBS data and neural network predictions to create counterfactual versions of each historical event at varying levels of global warming. This script was run remotely on an HPC system and may require significant memory.

- CMIP6_EU_Scaling.py and CMIP6_Adaptation_Scaling.ipynb aggregate CMIP6 projections to the EU district level and calculate the scaling between global mean temperature and district-level temperature (for use in the adaptation analysis).

- Counterfactual_Event_Mortality.py performs the final calculations of each event's mortality at varying levels of global warming. 

- Figures.ipynb plots most of the figures shown in the main text and supplementary material.