Rooftop photovoltaic potential in data-scarce regions using open-source geospatial information: Scenario-based approach for a Mediterranean coastal city

Rooftop photovoltaic (PV) systems can support urban decarbonization, yet data scarcity limits city-scale assessments in developing regions. This study develops an open-source 2D GIS workflow for five central districts of Antalya, Turkiye, using Overture Maps building footprints. A three-stage procedure combined spatial and size filters, geometry and land-use screening, and targeted image inspection to retain the continuous urban building stock while excluding rural settlements, greenhouses, and structures smaller than 60 m 2 . Planimetric agreement was evaluated against 60 manually delineated reference structures totaling 93,070 m 2 . Structure-level mean absolute percentage error was 8.48%, whereas total-area relative error was 0.13%; leave-one-out values ranged from −0.43% to +0.96%, demonstrating sensitivity to sample composition. After screening, 130,458 structures with 38.97 million m 2 of gross rooftop footprint remained. Because verified building-use labels and citywide 3D height data were unavailable, rooftop availability was represented by literature-supported utilization factors of 0.20 and 0.50, incorporating obstructions, access, setbacks, layout constraints, and inter-row spacing. With annual plane-of-array irradiation of 2084.62 kWh m −2 , module efficiency of 21%, and a performance ratio of 0.75, technical generation was estimated at 2.56 to 6.40 TWh yr −1 . Under an average-grid displacement factor of 0.430 tCO₂ MWh −1 , this corresponds to 1.10 to 2.75 MtCO₂ yr −1 . These estimates represent screening-level technical potential, not building-specific deployment forecasts, because source-data omissions, inter-building shading, structural suitability, permitting, and grid constraints remain unresolved. The workflow separates footprint error, preprocessing decisions, and rooftop-utilization uncertainty, providing a transparent screening framework for data-scarce cities.