AERMOD sensitivity to terrain and land-cover representation at a coastal power plant in Northern Iran

AERMOD sensitivity to terrain and land-cover inputs was evaluated at the Shahid Salimi Power Plant in northern Iran. A controlled 2 × 2 design combined terrain grids with reported spacings of 12.5 and 30 m with standard and Sentinel-2 object-based land-cover representations. Emissions, source and receptor geometry, and meteorological observations were held constant. Scenario performance for NOx, SO 2 and CO 2 was compared using R 2 , root mean square error (RMSE) and mean absolute error. Replacing the 30 m grid with the 12.5 m grid reduced RMSE by 14.9–31.3% across paired comparisons. The 12.5 m grid with standard land cover produced the strongest agreement for NOx (R 2 = 0.73; RMSE = 12.30 µg/m 3 ) and CO 2 (R 2 = 0.72; RMSE = 105.48 µg/m 3 ), whereas the OBIA scenario performed best for SO 2 (R 2 = 0.66; RMSE = 3.65 µg/m 3 ). Terrain change improved all metrics, while land-cover responses depended on pollutant. Metric confidence intervals were unavailable, and native terrain products remained unverified. OBIA was illustrative because classification validation and parameter-transfer records were incomplete. Without documented background removal, CO 2 statistics remained descriptive and supported neither ambient validation nor health-risk assessment. Future work should verify spatial inputs and extend concurrent seasonal monitoring across the coastal plain and inland foothills.