enyandwi7/Population-Housing-Structures-in-Rwanda-Inferred-from-Above-

In planning, practitioners mainly rely on administrative registers and census data for population statistics. Yet they need more than aggregated data: they require detailed information on conditions needing intervention, prioritising resorces where needs are most acute, and to provide solutions across various spatial scales, from individual blocks to large zones. To close the gap between aggregated data and spatially disaggregated, decision-relevant information, this study proposes a methodological framework that enriches population datasets with housing characteristics inferred from RS imagery. First Equation belwo was applied to disaggregate village-level population data to the scale of household units, utilizing variables such as the estimated average household size, building size, $P = P_0 + \left(\frac{A_i}{\sum_j A_j}\right) \cdot \Delta P_v$ where $P_i$ is the modelled population of unit i; P0 denotes the baseline population, estimated using the lower-bound of the official household size range; $A_i$ represents the building size of a probable household i; and $∆P_v$ is the residual population term that ensures population balance after deducting the sum of initial allocations from the total surveyed village population. This approach ensures that each building receives a reasonable population while maintaining consistency with village-level population totals. A housing wealth map, derived from remote sensing imagery using deep learning-based multi-class building extraction which extend beyond simple extraction of buildng shapes, was used to disagregate village population data. We collected data from district-level One Stop Centres across Rwanda. Given that the relevant administrative reporting structure is already established, our framework can be readily integrated into existing statistical system to generate timely and reliable urban data. In particular, village leaders are institutionally positioned to collect population nformation and report to high-up a …