WorldPop Showcases Population Modelling at Global Statistics Summit

The SSPM team’s work tackles a critical challenge: while satellite imagery helps identify where people live, its accuracy can be compromised by environmental factors like dense forests, cloud cover, snow, and rugged terrain. In some cases, natural features can even be mistaken for buildings. To overcome this, Dr Nnanatu and his colleagues have developed an advanced statistical model that not only estimates population density but also accounts for the chances that satellites may miss (or misidentify) settlements.

By combining these factors into a single, cohesive model, the researchers significantly improved the accuracy of population mapping. In simulation studies and real-world application in the Democratic Republic of Congo, their approach reduced estimation bias by more than 40%, a major step forward in supporting development planning, humanitarian response, and public health efforts. The work is due to be submitted for publication soon; the link below describes earlier research using a slightly different model to calculate population estimates in Papua New Guinea.

In addition to presenting his team’s research, Dr Nnanatu also chaired a high-profile session titled “Statistical Learning Methods for Feature Selection and Model Tuning,” which drew praise for its line-up of expert speakers and insightful discussion.