Long-Term Forecasting of Transport Sector CO₂ Emissions in Data-Scarce Environments: An Optimized Hybrid Ensemble Framework

Description

This study utilizes data from 2000 to 2024 on transport-sector CO₂ emissions in Pakistan from the World development Indicators (WDI). Energy consumption data (coal, oil, gas, electricity, biofuel) come from the International Energy Agency (IEA). Transport activity data and vehicle fleet information are sourced from the Pakistan Economic Survey. All data sources are publicly available without access restrictions. Quarterly observations were generated from annual data via linear interpolation to enhance temporal detail.

Steps to reproduce

SARIMAX is implemented here as a multivariate model that incorporates past CO₂ emissions, time, and other relevant variables (e.g., energy sources, GDP growth, vehicle fleet). Conversely, LSTM is used as a multivariate forecasting model with the same explanatory variables and lagged emissions data. Prophet incorporates exogenous regressors selected for their statistical relevance. These time-series models effectively capture seasonality and nonlinearity relationship. The accuracy of the model is evaluated by RMSE, MSE, MAE and R² for a comprehensive evaluation. In the third stage, predictions from individual models are integrated using a Whale Optimization Algorithm (WOA)-based ensemble method. The WOA calculates the optimal weights for each forecasting model, thus enhancing overall accuracy and reliability. Finally, a sensitivity analysis is performed to assess how the independent variables influence CO₂ emissions in the transportation sector. Multivariate linear regression is used to explore the relationships between the explanatory variables and emissions, both separately and together. Overall, this approach provides useful information for identification of key drivers of emissions and useful insights for policy makers to formulate effective policies and strategies for sustainable transport and emission reduction in Pakistan.

Institutions

• Chang'an University

  Shaanxi, Xi'an

Categories

Licence