Research questions:
(RQ1) How do 2D greenness (NDVI, land cover) and 3D greenness (green view index, tree canopy) at multiple scales relate to local heat exposure (land surface temperature, heat index/UTCI)? (RQ2) What are the associations between greenness and cardiovascular and mental health outcomes, and with biomarkers (CRP, HbA1c, lipids, creatinine, cystatin C)?
(RQ3) How much of any greenness benefit is mediated by reduced heat exposure?
(RQ4) Do effects vary by age, sex, socioeconomic status, urbanicity, and air pollution?
Objectives: (O1) Build high-resolution 2D/3D greenness layers for UKB participants. (O2) Estimate total and mediated effects of greenness on health risks and biomarkers using modern causal inference. (O3) Produce equity-aware, policy-ready metrics for heat adaptation.
Scientific rationale: Nature-based solutions can mitigate urban heat and improve cardiovascular and mental health via shading, evapotranspiration, and stress reduction. 2D NDVI captures vegetation quantity but misses street-level shade; 3D metrics better reflect human experience. Combining both with heat metrics in a large cohort enables robust inference on direct and heat-mediated pathways.
Methods overview: We will derive NDVI/land cover from satellites and compute 3D GVI/tree canopy from street-level imagery or LiDAR via deep learning. Exposures will be summarized in multiple buffers and windows. Heat exposure will be modeled from land surface temperature and reanalysis-derived indices. Health links will be assessed with survival models and machine learning (gradient boosting, random forests). Mediation will partition effects into heat-mediated and direct components. Confounding will be addressed using rich covariates (demographics, behaviors, SES) and spatial effects, with sensitivity analyses across buffers, windows, and modeling choices.