Hsu, A., Wang, X., Tan, J. et al. Predicting European cities’ climate mitigation performance using machine learning. Nat Commun

Objective:

• Employ a ML-driven approach to estimating and evaluating the mitigation performance of nearly all local and municipal actors in EU and UK

Case:

• EU + UK

Methodology:

• XGBoost
• SVM
• Random forest
• Feature selection: recursive feature elimination
• Time series: interrupted time series modelling

Data Source

• Self-reported emissions inventory
• Climate action policy
• ODIAC
• Building energy: NASA MERRA
• Emission
• GDP
• Population

Findings:

• Strong correlation between reported emission inventory and stationary fossil-fule CO2 emission

• EUCoM cities have on average, likely reduced annual per capita emissions from 2001 to 2018; 53% of external cities are likely to have experienced a negative trend in emission reductions

• Cities self-reporting emission inventory data are likely to have achieved greater average emissions reductions

Coding Reference:

• R