Machine learning pinpoints what drives coal power across BRICS economies

Teresa Odom — Tue, 27 Oct 2026 07:07:01 +0000

A new study in Theoretical and Applied Climatology applies two machine-learning methods, Random Forest and XGBoost, to identify the factors that drive coal-based electricity generation across the BRICS economies. The work, published in October 2026, uses carbon intensity, natural resource rents, and institutional governance indicators as predictors of coal-fired power output. Both models achieved strong predictive accuracy, according to the authors, Mei Yang, Yang Yang, and Na Li.

Their feature importance analyses ranked carbon intensity and oil-rent dependency as the dominant climatologically significant factors shaping coal power generation in the group, which includes Brazil, Russia, India, China, and South Africa. The finding that oil-rent dependency ranks alongside carbon intensity suggests that fossil-fuel revenue structures, not only emissions profiles, are entangled with coal electricity production in these economies. Residual diagnostics, which examine the errors left over after a model's predictions, indicated regime-specific heterogeneity across the BRICS members.

In practical terms, the models behaved differently across the individual countries, pointing to structural differences that a single uniform policy approach may not capture. The authors argue that this heterogeneity means emission-reduction strategies must account for the distinct structural characteristics of each BRICS economy rather than treating the bloc as a homogeneous unit. The study is an assessment of drivers within the group, not an evaluation of specific policy interventions or their outcomes.

The machine-learning framework presented in the paper is offered as a robust statistical approach for evaluating climate-policy scenarios in fossil-fuel-dependent economies. Because the analysis is predictive and correlational in nature, the identified feature importances describe which variables carry the most explanatory weight in the models, rather than demonstrating causal effects of any single driver on coal generation. The full article, including the figures referenced in the study, appears in Volume 157 of the journal as article number 757.

The abstract is available as a preview of subscription content, with full access requiring institutional or personal subscription.

Subject of Research: Climate drivers of coal-fired power in BRICS: a random forest and XGBoost assessment

Article Title: Climate drivers of coal-fired power in BRICS: a random forest and XGBoost assessment

Article References: Yang, M., Yang, Y., & Li, N. (2026). Climate drivers of coal-fired power in BRICS: a random forest and XGBoost assessment. Theoretical and Applied Climatology, 157(11), Article 757. https://doi.org/10.1007/s00704-026-06690-1

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06690-1

Keywords: Climate, drivers, coal-fired, power, BRICS, random, forest, XGBoost, assessment, scientific research

influence of oil-rent dependency on coal generation – Science

Machine learning pinpoints what drives coal power across BRICS economies