Urban Ecological Drivers of CO₂ Intensity in Arid Cities: A Multi-Scale, VHR-based Analysis of Baghdad and Erbil

Abstract

Urban areas in arid zones face distinct emissions challenges from rapid growth, high cooling demand and sparse vegetation cover, which limits the natural absorption of CO₂ and strains urban ecology. Using an integration of VHR satellite imagery with local socioeconomic and infrastructural data, this study presents a framework that measures carbon dioxide emission intensities across the Zayouna/Mansour district (Baghdad) and Erbil Municipality Center (Erbil), two major Iraqi cities situated in an arid environment. A combined bottom-up and top-down framework with Random Forest regression was used to identify localized emission hotspots and the ecological determinants of CO₂ emissions, drawing on Sentinel-2-derived NDVI, Landsat 8/9-derived LST, VIIRS-derived night light intensity, WorldView/Pléiades-derived building density, OpenStreetMap road density and surface albedo at 250 × 250 m grid resolution over 2020–2024. Results reveal spatially concentrated emission hotspots near major road arteries and high-density residential and commercial zones, while green areas show lower emission intensities, confirming vegetation's role as a CO₂ sink in urban areas. The spatial emission model performs well in estimating each city's footprint (R² = 0.78, RMSE = 0.008 kg CO₂/m²). Panel data and fixed-effect models identify traffic density as the primary driver of CO₂ emissions in Baghdad (coefficient = 0.040, p < 0.001, R² = 0.85), while energy consumption dominates in Erbil (coefficient = 0.028, p < 0.001, R² = 0.80). Vegetation cover (NDVI) is negatively related to emission intensity in both cities (Baghdad: −0.095, Erbil: −0.080; both p < 0.001). These findings offer actionable guidance for climate-smart urban planning and customized emission-reduction policies in data-scarce arid cities, and the methodology is transferable to other data-poor urban areas developing smart city strategies.