How many people live in a neighbourhood, a village, or a city block? It sounds like a simple question, but for millions of people around the world, the answer is still unclear. Accurate, local population data is essential for everything from planning health services and responding to disasters to allocating resources fairly. Yet traditional national censuses, the backbone of population statistics, are expensive, slow, and usually conducted only once every ten years. By the time results are available, they can often already be out of date.
The consequence is serious: large numbers of people may remain effectively “invisible” to planners and policymakers. This makes it harder, for example, to achieve the UN’s Sustainable Development Goals, around half of which depend on timely, detailed population data.
A recently updated research preprint led by Dr Attila Lazar reviews how a new generation of techniques is helping to close these gaps. Known as “bottom-up” population mapping, these approaches are not meant to replace national censuses, but to complement them – making population data more current, detailed, and responsive.
At the heart of this approach is a smart combination of limited demographic data and rich geospatial information. On the demographic side, researchers draw on sources such as small, rapid surveys in selected areas, partial census counts or listings from household surveys, and data collected during routine health campaigns, like the distribution of mosquito nets. While none of these sources is complete on its own, together they provide valuable clues about who lives where.
The real game-changer has been advances in satellite imagery and machine learning. Today, near global datasets exist that map individual building footprints, including their size and even estimated height. These details are powerful indicators of how populations are distributed across landscapes.
All of this information is fed into advanced computer models, including Bayesian, AI and machine learning methods, to estimate population numbers and characteristics – often down to areas as small as 100 meters across. These methods are already being used in practice, supporting national planning in countries like Mali, Papua New Guinea and South Sudan, and helping Burkina Faso and Colombia fill gaps left by traditional census efforts.
Challenges remain, especially in hard-to-reach rural areas and informal urban settlements, and there can be political hesitation around using modelled data, as well as capacity gaps in being able to implement such approaches within national statistical offices. Still, with continued investment, collaboration, and trust-building with national institutions, these technologies offer a powerful way to make the invisible visible – ensuring no one is left out of the data that shapes our world.
