Epistemological Foundations and Critical Methodological Advances in African Urban Geodemography: Developing Fit-for-Purpose Population Estimation Methods for Nigerian Cities

Official population statistics for Nigerian cities are beset by data quality limitations rooted in census methodology controversies, rapid informal growth dynamics, and persistent political interference, generating fundamentally uncertain demographic baselines that undermine urban planning, infrastructure investment, and humanitarian response across the country. This dissertation advances the epistemological foundations and methodological toolkit for African urban geodemography by developing original fit-for-purpose population estimation methods for Nigerian cities that transcend the limitations of decennial census-dependent approaches. A theoretical framework integrating critical cartography, epistemology of spatial knowledge, and measurement theory critically evaluates the assumptions underlying conventional dasymetric population mapping when applied to Nigerian urban realities. Three original methodological innovations are developed and evaluated: a Deep Building Morphology Population Model that extracts population estimates from convolutional neural network analysis of building footprint shape and density in WorldView-3 imagery; a Multi-Source Ensemble Fusion Method that systematically combines satellite-derived estimates, mobile phone CDR data, and nighttime light brightness into a Bayesian posterior population distribution; and a Conflict-Adjusted Seasonal Flux Model that accounts for population mobility patterns in displacement-affected northern cities. Cross-validation against enumerated reference populations from eight purposively selected Nigerian city areas yields mean absolute percentage errors of 11.4 percent for the Deep Building Morphology model and 8.7 percent for the Multi-Source Ensemble approach. The dissertation constitutes an original epistemological and methodological advance in Nigerian urban population science with direct applicability to sub-Saharan African urban geodemographic practice. Keywords: urban geodemography, population estimation, Nigeria, deep learning, fit-for-purpose methods.