Accurate human population data is required at a high spatial resolution for a variety of applications, such as planning service delivery, assessing ideal locations of facilities and designing effective vaccination programmes. In many circumstances sufficient data may not be available to successfully fulfill these goals. Examples of this include situations where a full census is not possible, e.g. within insecure regions, or at post-censal time points where population counts are outdated. Using population data for small, well-defined areas (microcensus samples), along with ancillary geospatial covariates, it is possible to predict population across the whole of a study area using spatial statistical models common to ecological studies. Microcensus population data are often collected across seasons and years, making it essential to account for temporal effects in statistical inferences. The key temporal changes for human populations include increased fertility or survival, urban growth (which is often rural to urban migration) and seasonal work migration.
Results/Conclusions
Even if the overall quantity of population change is similar across these different scenarios, we know that the resultant dynamics will differ because of their spatial- and age- specificity. Here we investigate the role of resultant transient dynamics that arise from these different forms of population change and assess the best population estimation methods that account for the impact of transients on population counts. We do this by comparing the performance of spatio-temporal statistical modelling techniques and matrix modelling approaches in estimating population counts for ‘true’ simulated population data. We evaluate how data availability impacts the effectiveness of each approach’s performance and give country specific examples of their implementation.