- R 100%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| heatwave_greenspace_analysis.R | ||
| README.md | ||
Small Accessible Urban Parks Mitigate Heat-related Mortality
Overview
This repository contains the statistical analysis code for our paper:
Small accessible urban parks mitigate heat-related mortality
He, C., Yin, P., Chen, R., et al.
Key Finding: Urban green space configuration—not just total coverage—plays a crucial role in reducing heatwave-related mortality. Cities with more dispersed, accessible green spaces and regular boundaries show significantly lower mortality risks during heatwaves.
What this code does
This analysis examines how different green space landscape patterns modify the relationship between heatwaves and mortality across 265 Chinese cities (2013-2019). We use:
- Stage 1: Distributed lag non-linear models (DLNM) to quantify city-specific heatwave-mortality associations
- Stage 2: Random-effects meta-analysis to assess how five landscape metrics modify these associations
- Key metrics: PLAND (coverage), SPLIT (dispersion), LSI (boundary regularity), PD (patch density), LPI (largest patch size)
Requirements
Software
- R (≥ 4.0.0)
- Required packages:
mgcv,dlnm(≥2.4.7),mvmeta(≥1.0.3),splines,dplyr
install.packages(c("mgcv", "dlnm", "mvmeta", "splines", "dplyr"))
Data
You will need two datasets:
-
Daily mortality and meteorological data (
ccches_heatwave.csv)- Columns:
code_city,death,temp,rh,dow,time,season, heatwave indicators
- Columns:
-
City-level green space metrics (
city_metadata_greenspace.csv)- Columns:
code_city,PLAND,PD,LPI,LSI,SPLIT(categorized as 1/2/3 for Low/Medium/High)
- Columns:
Advanced Usage
Sensitivity Analyses
# Test different lag periods
results_lag14 <- analyze_heatwave_mortality(data, lag_days = 14)
# Test different degrees of freedom
results_sens <- analyze_heatwave_mortality(data, df_time = 6, df_humidity = 6)
Stratified Analyses
# Age-specific analysis
data_young <- data_main %>% mutate(death = death_5_64)
results_young <- analyze_heatwave_mortality(data_young)
data_old <- data_main %>% mutate(death = death_65plus)
results_old <- analyze_heatwave_mortality(data_old)
# Climate zone-specific analysis
data_temperate <- subset(data, code_city %in% temperate_cities)
results_temperate <- analyze_heatwave_mortality(data_temperate)
Results
Output File
heatwave_greenspace_results_all.csv contains:
| Column | Description |
|---|---|
Heatwave_Definition |
92.5%_2day / 92.5%_3day / 95%_2day / 95%_3day |
Landscape_Metric |
PLAND / PD / LPI / LSI / SPLIT |
Category |
Low / Medium / High |
RR |
Relative risk (point estimate) |
Lower_CI |
Lower 95% confidence interval |
Upper_CI |
Upper 95% confidence interval |
P_value |
Significance of difference vs. Low category |
Interpretation
- RR > 1: Increased mortality during heatwaves
- P < 0.05: Significant difference between categories
- Compare High vs Low categories for each metric
Methods
Statistical Models
Stage 1 - City-specific GAM:
death ~ crossbasis(heatwave, lag=10) + factor(dow) + ns(rh, 4) + ns(time, 12)
- Family: quasi-Poisson
- Crossbasis: B-spline (var, 4 df) + natural spline (lag, 4 df, log-knots)
- Controls: day of week, humidity (4 df), time trend (4 df/year)
Stage 2 - Meta-analysis:
mvmeta(coef ~ 1, vcov, method = "reml")
- Random-effects model
- REML estimation
- Stratified by landscape metric categories
Citation
If you use this code, please cite:
@article{he2024greenspace,
title={Small accessible urban parks mitigate heat-related mortality},
author={He, Cheng and Yin, Peng and Chen, Renjie and others},
journal={[Journal Name]},
year={2026},
doi={[DOI]}
}
Contact
Lead Author:
- Cheng He: chenghe@hsph.harvard.edu
For code-related questions, please open an issue on GitHub.
License
This code is provided for research and academic purposes. For commercial use, please contact the authors.