English Français |
Chibuzor Christopher Nnanatu, University of Southampton
Ortis Yankey, University of Southampton
Anaclet Dzossa, Institut National de la Statistique du Cameroun
Thomas Abbott, University of Southampton
Assane Gadiaga, University of Southampton
Attila Lazar, University of Southampton
Andrew J. Tatem, University of Southampton
High-resolution population data are important alternatives to outdated or coarse census projections which lacked the level of granularity required for effective developmental and healthcare policy planning and implementation at small area units. They are usually raster products with very fine spatial resolutions of around 100m by 100m produced by integrating population data (e.g., Microcensus, household survey) with satellite-based settlement data (e.g., building footprints) and geospatial covariates (e.g., nighttime lights) using advanced statistical models. Here, we capture the potential effects of spatial autocorrelation on population density and distribution motivated by household-based population datasets in Cameroon using a robust bottom-up population modelling scheme based on the integrated nested Laplace approximation and stochastic partial differential equations frameworks. Our methodology was successfully applied to produce small area population estimates in Cameroon, and simulation study results indicated that it was robust over different number of areal unit-observation coverage combinations thus, an important population modelling development.
Presented in Session 99. Computational approaches to population studies in Africa