High-resolution population estimation using household survey data and building footprints.

Area of Science:

• Demography
• Geospatial analysis
• Statistical modeling

Background:

• National census data becomes outdated between censuses, hindering effective public policy and decision-making.
• Accurate and up-to-date population data is crucial for resource allocation and planning.
• The Democratic Republic of the Congo has not conducted a national census since 1984, creating a significant data gap.

Purpose of the Study:

• To develop a Bayesian hierarchical model for producing high-resolution, up-to-date population estimates.
• To estimate population totals and demographic breakdowns (age, sex) with associated uncertainty.
• To validate the model's accuracy using recent household surveys and building footprint data.

Main Methods:

• A Bayesian hierarchical model was employed, integrating recent household survey data with building footprint information.
• Population estimates were generated for grid cells of approximately 100m resolution.
• The model was applied across five provinces in the Democratic Republic of the Congo.

Main Results:

• The model demonstrated a strong fit, achieving an R-squared value of 0.79 for out-of-sample population total predictions at the microcensus-cluster level.
• Age and sex proportions were predicted with near-perfect accuracy at the province level (R-squared = 1.00).
• The study successfully produced population estimates with associated uncertainty measures.

Conclusions:

• Combining household surveys and building footprints is an effective strategy for high-resolution population estimation.
• This approach significantly improves population data accuracy in countries with outdated census information.
• The developed Bayesian model offers a valuable tool for contemporary demographic analysis and planning.