Imagine you were asked to make a map of the whole world that counted all the people in it. Everyone you know, and everyone you don’t know. Where would you begin?
This is the task laid out for researchers who create today’s global population datasets.
Historically, most population maps have been based on specific censuses or statistical surveys, wherein a government attempts to count and locate all the citizens in its territory. Way back when, this was accomplished entirely through a process of humans counting other humans: the Census Act of 1879 required local census enumerators to “visit personally each dwelling house in his sub-division, and each family therein, and each individual living out of a family in any place of abode.”
Today, our census enumeration largely takes place through mail-in or digital surveys, but even our most recent 2020 Census still employed Census Bureau “door knockers” to aid in getting a complete a count as possible.
In the United States, these counts of people are aggregated to what we term “census geographies.” These are specially crafted sets of geographic boundaries used to aggregate population at scale. The Census Bureau uses census geographies to count, map, and visualize large numbers of people while still protecting individual privacy.
These local census geographies evolved over the course of the late 19th and early 20th centuries – from 1890’s experimental ‘sanitation districts’ to the early 20th century’s experimental ‘districts’ within major cities – and have emerged as today’s block, block group, tract, county, and state level geographies.
A national census is a great starting point to map the residents of one nation, but each national census has its own methodology, timing, and aggregation their counts to different (and changing) geographies. Some nations don’t have a regularly occurring census at all. How then do scientists begin to count everyone, everywhere, all at once?
The short answer is: they don’t. Scientists and geographers use complex spatial models to estimate the total number of people across the globe, and they do this in different ways. Some datasets seek to transform national surveys or censuses “top down” into population grids through areal weighting, while others build an estimation “bottom up” through complex algorithmic processing of input layers, which can include things like topography, slope, highways, known settlement areas, and satellite imagery. How to estimate population is a challenging ongoing topic of scholarly research and debate: methodologies for these datasets are typically peer-reviewed and can be found in in scientific journals.
The resulting global population rasters can be directly visualized in GIS software for cartographic effect (as seen above) or can be used as inputs to help answer questions through spatial analysis. These datasets help us answer basic foundational questions about the world: for example, how many people will be affected by a 6 inch sea level rise? Our answer to this question requires good geographic models of not just climate and oceans, but of the changing distribution of people throughout the world. These datasets are often the best resource we have when trying to understand how world events will impact populations. Amid the ongoing war in Ukraine, LandScan (developed by the Oak Ridge National Laboratory) released Russia and Ukraine High Definition population datasets, which provide populations estimates dated to 2021 at a 100 meter resolution. Both LandScan datasets, alongside WorldPop (globally for 2000-2020) are currently available for download online as part of the Geography & Map Division’s digital geospatial collections.