Assessment of gridded population data in slums for SDG11 monitoring using community generated data

Abstract Modelled gridded population datasets are an attractive alternative to outdated/inaccurate census data for Sustainable Development Goal monitoring. SDG 11.1.1–percent of urban population in “slums,” informal settlements, or inadequate housing–is critical because 90% of global population growth over the next 25 years is expected in African and Asian cities alone, mainly in deprived areas. We use KnowYourCity Campaign data from 351 “slums” across 10 cities to assess accuracy of nine gridded population datasets in “slums.” We found the most accurate gridded population dataset estimated just 43% of “slum” dwellers. We further estimated SDG 11.1.1 in six cities with citywide slum boundaries, and found substantially fewer “slum” dwellers using gridded population estimates (1–37%) versus current survey-based approaches (52–69%). To make gridded population datasets suitable for SDG 11.1.1 monitoring, we propose three data/method innovations to improve urban estimates: auxiliary data about building morphology, hyper-fine scale outputs, and training data from "slums".