Predicting European cities’ climate mitigation performance using machine learning

Abstract

Although cities have risen to prominence as climate actors, emissions’ data scarcity has been the primary challenge to evaluating their performance. Here we develop a scalable, replicable machine learning approach for evaluating the mitigation performance for nearly all local administrative areas in Europe from 2001-2018. By combining publicly available, spatially explicit environmental and socio-economic data with self-reported emissions data from European cities, we predict annual carbon dioxide emissions to explore trends in city-scale mitigation performance. We find that European cities participating in transnational climate initiatives have likely decreased emissions since 2001, with slightly more than half likely to have achieved their 2020 emissions reduction target. Cities who report emissions data are more likely to have achieved greater reductions than those who fail to report any data. Despite its limitations, our model provides a replicable, scalable starting point for understanding city-level climate emissions mitigation performance.

The increasing emission divide between cities of the Global North and Global South: Towards adjustable mitigation scenarios at the city level

“…However, mitigation measures have been adopted in wealthier Western countries in recent decades, and they have achieved incomparable progress compared with those elsewhere (Jabareen, 2015b). Hsu et al (2022) assessed the mitigation performances of nearly 50,000 local and municipal actors in the European Union from 2001 to 2018, reporting that 84% of cities participating in transnational climate governance reduced emissions over that period. On average, participating cities reporting emissions data have higher annualised per capita reduction than cities without reported emissions.…”

“…Moran et al (2022) presented a new CO 2 emissions inventory for all 116,572 municipal and local-government units in Europe, containing 108,000 cities at the smallest scale used. Hsu et al (2022) assessed the mitigation performance of nearly 50,000 local and municipal actors in the European Union from 2001 to 2018.…”

Machine Learning Predicts the Methane Clumped Isotopologue (12CH2D2) Distributions Constrain Biogeochemical Processes and Estimates the Potential Budget

“…In XGBoost, the database is divided into a training set ( n = 127), a testing set ( n = 40), and validation set ( n = 12). The scaling of the delineation is fixed, and the difference lies in the specificity of the XGBoost model in training . After the initial training was completed, the Tune function was used to determine the Hyperparameter Optimization process with the appropriate parameters (RF: ntry = 2, ntress = 223; SVM: cost = 4, gamma = 2.72, epsilon = 0.1, sigma = 0.29, support vectors = 78), where the SVM shows a large improvement (Figure S3).…”

“…We evaluated several ML models, including random forest (RF), support vector machines (SVM), and extreme gradient boosting (XGBoost), that represent typical bagging and boosting algorithms (details in Text S3), which are widely used for regression studies. − The multivariate linear regression (MLR) model, a commonly used traditional regression approach, has been evaluated and compared with three ML models here.…”

“…The scaling of the delineation is fixed, and the difference lies in the specificity of the XGBoost model in training. 39 After the initial training was completed, the Tune function was used to determine the Hyperparameter Optimization process with the appropriate parameters (RF: ntry = 2, ntress = 223; SVM: cost = 4, gamma = 2.72, epsilon = 0.1, sigma = 0.29, support vectors = 78), where the SVM shows a large improvement (Figure S3). The RF, SVM, and XGBoost models were mainly developed via the R packages "randomForest", "e1071", and "xgboost".…”