Modélisation de la densité de la population dans le district de la Vallée du Bandama (Centre de la Côte d’Ivoire) : apport de la géomatique et l’indice de « Vegetation - Temperature-Light-Population (VTLPI) »

Population size in geographical areas is a crucial indicator for urban planning and assessing access to various basic social services. In most countries, and particularly in Côte d'Ivoire, conventional population censuses take place every ten years, limiting their usefulness for regular statistics. Satellite imagery offers an alternative to traditional survey data for generating socio-economic and socio-demographic indicators, providing a new perspective for regular statistics. A statistical model was built to forecast population density in Côte d'Ivoire and the Bandama Valley in particular, by comparing the performance of the Random Forest and XGBoost algorithms. This model used indices derived from night-time imagery (NTL), NDVI and LST, to construct a new VTLPI index to model population density. The model performed well, accurately predicting population density at the Bandama Valley district level. The results indicate that the population density estimated by the VTLPI is close to the field survey data. Using the census data as a reference, the mean absolute error at District level is 0.54 and the mean square error is 264 persons/km2 respectively. In addition, the coefficient of determination R² for the Random Forest and XGBoost algorithms are 0.84 and 0.87 respectively. The results show that our VTLPI-based model provides a better estimate of population density in the district, and allows us to characterize more spatial variation at the 250-meter grid level at the local level. The resulting population density provides better data on population exposure for natural disaster risk and loss assessment and other related applications