High resolution population estimates are essential for government planning, development projects, and public health campaigns, but countries where this information is most needed are often where recent census data are least available. We present a Bayesian modelling framework that combines information from limited microcensus surveys with GIS and remote sensing data to estimate population sizes for every 100 m grid cell nationally and to properly account for uncertainty in these estimates. Microcensus surveys enumerate all people in relatively small survey areas (e.g. three hectares each) that are randomly located throughout a study area. We will demonstrate the approach using results from 1,141 microcensus survey areas in Nigeria to map the population nationally, including areas that are inaccessible due to conflict. Important GIS predictors in the model included maps of school densities, household sizes, settlement types identified from satellite imagery (e.g. urban, rural, non-residential), and disaggregated population estimates from the last census in 2006.
Results/Conclusions
We mapped the Nigerian population at high resolution nationally, mapped uncertainty, and provided population totals for local government areas and states. This work has immediate applications for polio eradication efforts, yellow fever vaccination campaigns, and basic planning for government services, including future census efforts. One challenge was making results easily accessible and understandable. We present a user interface that provides easy access to customized model results designed to maximize uptake of results by governments, NGOs, and other stakeholders. In the future, our modelling approach can be customized to maximize information gleaned from microcensus datasets in other countries. When appropriate data are available it is possible to incorporate hierarchical sub-models to account for observation error associated with microcensus surveys, settlement mapping measurement error, and temporal autocorrelation. These concepts draw heavily on robust literature from theoretical ecology and conservation biology. Open exchange of ideas between ecologists and human demographers benefits both research areas, providing ecologists outreach strategies to increase impact of statistical models in government planning, and providing human demographers/geographers in developing countries sophisticated modelling tools common in ecology where data limitations are the rule rather than the exception.