work · 2022 lancet global health geospatial modeling

mapping who cooks with solid fuels, and the child mortality attributed to it

Density plot of world population and under-five pneumonia deaths distributed across PM2.5 air-pollution exposure, 2000 vs 2018, split by country

Where the world’s people — and where children’s pneumonia deaths — fall across the air-pollution exposure scale, 2000 vs 2018, by country. Read the paper →

Billions of people cook with solid fuels — wood, charcoal, dung, coal — and the household air pollution that results is a leading cause of pneumonia death in young children. National averages hide enormous within-country inequality. This work mapped solid-fuel use, the pollution exposure it creates, and the under-five mortality attributed to it at 5 km resolution across 98 low- and middle-income countries, 2000–2018, with forecasts to the 2030 clean-cooking target.

5 km

resolution

countries

2.1M

survey observations

663

surveys

Context

Clean cooking is a Sustainable Development Goal (7.1) because the alternative kills: fine particulate matter from burning solid fuels indoors drives lower-respiratory infections, and children under five bear much of the mortality. Progress is usually tracked with national numbers, which average over districts that are worlds apart — a capital city and a rural highland can share a country and almost nothing else. To target the problem you have to see it at the scale it actually varies.

What this is

A two-stage model-based-geostatistics (MBG) pipeline that turns millions of scattered survey responses into continuous 5 km surfaces:

Assemble — 2.1 million household observations from 663 surveys, geo-located and collapsed to cluster-level solid-fuel-use rates, joined to dozens of satellite/geospatial covariates (nightlights, travel time, climate, land cover, population, ambient PM2.5).
Stack — a constrained ensemble of machine-learning submodels (GAM, gradient-boosted trees, penalized regression) fit with cross-validation; their predictions become the mean surface.
Geostatistics — a Bayesian hierarchical model adds a spatial Gaussian process (SPDE), country and sub-national random effects, fit with R-INLA — borrowing strength across space to estimate, with uncertainty, even where surveys are sparse.
Calibrate & aggregate — raked to national GBD estimates, aggregated to admin 0/1/2.
Forecast & burden — annualized-rate-of-change projections to 2030 vs. SDG 7.1; combined with ambient PM2.5 and relative-risk curves to estimate attributable under-five mortality.

What it shows

Solid-fuel use remains near-universal across much of sub-Saharan Africa and pockets of South and Southeast Asia, and the within-country gradients are stark. Projected forward, most countries are not on track to meet the 2030 clean-cooking target — and the resulting household air pollution is still associated with hundreds of thousands of child deaths a year.

Map of the modeled probability of attaining SDG 7.1 clean-cooking target by 2030, red where unlikely, blue where likely

Probability of attaining the 2030 clean-cooking target (SDG 7.1) — red where it’s unlikely on current trends.

What I owned

First author — the full analysis, from survey extraction through the modeled surfaces, forecasts, and burden estimates.
The stacked-ensemble + Bayesian geostatistical model and its covariate/validation design.
The 2030 SDG forecasts and the exposure-to-mortality burden chain.
The paper — published in Lancet Global Health.

What it changes

The deliverable wasn’t a model; it was a decision surface: where the problem is worst, where it’s moving, and where it won’t resolve without intervention.

See it

the paper in lancet global health analysis code on github

Attribution. Published as Frostad et al., Lancet Global Health (2022); funded by the Bill & Melinda Gates Foundation; built on IHME’s model-based geostatistics framework. Figures are from the paper.