Spatio-temporal monitoring of non-regulatory settlements using remote sensing and machine learning: A case study of Bni Makada, Tangier, northern Morocco

Urban expansion is frequently accompanied by the proliferation of informal or non-regulatory constructions, which pose major challenges to sustainable land management. Remote sensing combined with machine learning (ML) provides a robust framework for monitoring such dynamics over space and time. This study investigates the spatio-temporal evolution of non-regulatory habitats in Bni Makada, Tangier (Morocco), using Sentinel-2 imagery (2017 and 2025), machine learning classifiers (ANN, SVM, RF), and ancillary datasets including the Microsoft Building Footprints and the official urban development plan. After spectral band selection and supervised classification, model performance was assessed using recall, precision, F1-score, overall accuracy, and Cohen’s kappa index. Results show that the ANN classifier outperformed other models (OA = 0.93; κ = 0.87), enabling reliable discrimination between built-up and non-built-up areas. Between 2017 and 2025, Bni Makada recorded a net increase of 0.68 km2 in built-up surface, with more than 10,000 constructions identified in non-regulatory zones. Although the relative share of informal units decreased slightly (from 39.6% to 37.2%), their absolute number rose (+268 constructions), reflecting persistent urban governance challenges. The integration of satellite imagery, ML, and planning data provides a transferable methodology for detecting, quantifying, and managing non-regulatory constructions, thereby offering actionable insights for urban planning authorities in Morocco and comparable contexts worldwide.