Small area population forecasting using geospatial big datasets and national census in low and middle income countries

非洲的人口预计在未来40年内将翻一番，这将导致城市化的高速发展，并因此引发环境、社会经济和健康方面的变化。小的地理区域细节对于规划和跟上这些变化至关重要。尽管小区域预报很重要，但仍未得到充分探索，这限制了包括纳米比亚在内的中低国家应用循证解决方案的潜力。小区域预测的主要数据来源包括人口和住房普查、调查和行政数据。尽管普查提供了完整的人口统计以及生育率、死亡率和移民数据，但这些操作通常每十年才进行一次。国家统计局通常进行预测，以产生普查间估计数;然而，这些都是在非常粗略的空间尺度上进行的。因此，很难在更精细的地理尺度上捕捉人口趋势和模式，特别是通过移徙。卫星制图可以捕捉到地理上不均衡的人口增长，这通常很难使用标准投影模型进行预测和解释（Wardrop等人，同样，移动的电话数据提供了捕捉时间和空间移民模式季节性变化的机会。我提出了一项研究，旨在评估地理空间指标如何提高人口预测模型的准确性。该研究将以2011年人口普查为基线，探索不同的预测方法，并使用自上而下的模型将其分解为小区域，并将结果与2021年人口普查数据进行验证，以确定最佳估计。

Africa's population is forecasted to double over the next 40 years, contributing to a high rate ofurbanisation, and as a consequence, this trigger environmental, social-economic and health variations.Small geographic area details are crucial to plan and be abreast of these changes. Despite theirimportance, small area forecasts remain under-explored which limit the potential of low and middlecountries including Namibia to apply evidence-based solutions. Primary data sources for small areaforecasts include population and housing censuses, surveys and administrative data. Although thecensus provides a full count of the population as well as data on fertility, mortality and migration, theseoperations are usually done only every ten years. National statistics offices typically generate projectionsto produce inter-censual estimates; however, these are done at very coarse spatial scales. Hence,making it challenging to capture demographic trends and patterns at finer geographic scales, especiallythrough migration. Mapping from satellites can capture the geographically uneven population growththat is often difficult to predict and account for using standard projection models (Wardrop et al., 2018).Likewise, mobile phone data offer opportunities to capture seasonal variation over time and space inmigration patterns. I am proposing a study that aims to assess how geospatial metrics can improve theaccuracy of population projection models. The study will explore different projection methods withcensus 2011 as a baseline and use top-down models to disaggregate to small areas, and results will bevalidated against 2021 Census data to identify the best estimates.