Monitoring and projecting population distribution patterns in Kuwait using nighttime lights imagery and mosques data

Estimating urban population distribution is essential for data-driven planning, especially in developing nations. While nighttime light (NTL) data from DMSP/OLS show a strong correlation with human activity, their effectiveness is hindered by saturation and overglow, thereby limiting precise population mapping. To address this, we developed the Mosque-Adjusted NTL Index (MANTLI), integrating geolocated mosque data with NTL to enhance precision across Kuwait at a 500-meter grid scale. We evaluated MANTLI against four benchmark models (NTL, VANUI, EANTLI, and WorldPop) using residual spatial autocorrelation (Moran’s I) and correlation coefficients. The results confirm MANTLI’s superiority: only MANTLI and EANTLI exhibited random residual patterns (Moran’s I = 0, p > 0.05), indicating minimal spatial bias, whereas other methods showed significant clustering (p < 0.01). Moreover, MANTLI demonstrated the strongest correlation with census data, surpassing WorldPop (r = 0.57), while reducing the root mean square error (RMSE) by 12% compared to baseline NTL. The measure captured intricate residential brightness gradients between urban centres and peripheries, highlighting subtle spatial variability in density gradients. Collectively, MANTLI offers a reliable, high-precision proxy for pixel-level population estimation in arid urban environments, facilitating detailed research of settlement trends with significant policy implications.