Forecasting Vehicle Emissions Using an Integrated COPERT- Artificial Neural Network Modeling Framework

Abstract

Urban vehicle emission modeling has traditionally relied on conventional regression methods that inadequately capture complex non-linear interactions among influencing variables. Moreover, the combined influence of fleet composition and local environmental conditions remains poorly understood. This study integrated COPERT-derived baseline passenger vehicle (PV) emission factors with an Artificial Neural Network (ANN) model to predict Addis Ababa’s city-specific PV emission levels. The framework also employed Polynomial Linear Regression (PLR) model to forecast PV fleet growth between 2005 and 2025 and to evaluate the associated environmental impacts from 2018 to 2025. The models utilized climate data, vehicle activity patterns, and PV registration records as key inputs. Results reveal that PV ownership in Addis Ababa has increased more than twentyfold over the past two decades. Baseline emission factors indicated substantial reductions in CO and NO_x emissions with higher Euro classification levels, although CO₂ emissions remain persistently high. The ANN-based predictions show a 25% increase in CO₂ emissions, while NOx emissions rose from 1.89 to 2.08 tons/year for gasoline and from 6.02 to 7.27 tons/year for diesel PVs. CO emissions peaked at 26.25 tons/year in 2021 before declining to 21.10 tons/year by 2025, following the ban on internal combustion engine PVs. The ANN model achieved high predictive accuracy, with R² values ranging from 0.96 to 0.99. Overall, the integrated COPERT–ANN framework offers a robust, data-driven approach for urban emission prediction, providing valuable insights to guide sustainable transport planning and emission mitigation in rapidly growing cities.