Slum Detection and Sub-Pixel Slum Cover Fraction Mapping with AlphaEarth Foundations: A Representation Learning Evaluation Across 12 Global Cities

Pixel-level slum mapping remains constrained by limited cross-city generalisation, the lack of continuous estimates of intra-pixel slum coverage, and weak comparability across geographic contexts. AlphaEarth Foundations (AEF), which provides globally consistent 64-dimensional annual surface embeddings at 10 m resolution, offers an analysis-ready representation for scalable and lightweight slum monitoring. However, its applicability to slum mapping remains unclear because slum conditions are only indirectly observable from Earth observation data and are jointly shaped by built-form and socioeconomic processes. We systematically evaluate AEF for binary slum classification and sub-pixel slum cover fraction estimation across 12 cities and 69 city–year pairs from 2017 to 2024, using high-resolution GRAM-derived pseudo-masks as supervisory labels. The evaluation spans four training strategies, two protocols (random split and 3 × 3 spatial block cross-validation), six auxiliary feature configurations, and five baseline models, complemented by representation-level analyses (PCA, SHAP) and full-AOI mapping. Five findings emerge. (1) Same-city cross-year training is optimal under both protocols (median spatial F1 = 0.616, R2 = 0.466); temporal expansion outperforms cross-city transfer, indicating city-scale representational drift. (2) Regression R2 is driven primarily by zero/non-zero boundary discrimination: positive-pixel R2 is consistently negative across all cities, revealing limited capacity to model intra-pixel variation in sub-pixel slum cover fraction at 10 m. (3) PC36 is consistently top-ranked across tasks; classification saturates at k = 32 while regression remains unsaturated at k = 64. (4) POI features yield the largest density gain (ΔR2 = +0.064). (5) For six cities meeting dual-task usability thresholds, full-AOI inference across 2017–2024 preserves slum cluster structure (mean SSIM = 0.926). The study delineates the capabilities and complementarity needs of foundation-model embeddings for slum monitoring.