A Data-Driven Exploration: Providing Early Feedback on Socioeconomic Mitigation Strategies for Climate Change

Abstract

The current rate of climate change is unprecedented in millennia and represents one of humanity’s most significant issues for the 21st century, with serious consequences for both civilization and the environment if no action is taken. To reduce the future impact of changes, coordinated international efforts such as those outlined in the Paris Agree- ment are required. However, mitigation and adaptation efforts should not only be extensive, but also be precise in order to align international socioeconomic development with temperature targets as well as other societal and envi- ronmental sustainability goals. In order to inform about the future climate, extensive efforts through institutions such as the Intergovernmental Panel on Climate Change (IPCC) have resulted in extensive scenario-based research with Earth system models (ESMs). This work forms the knowledge basis for information provided to the worlds decision makers. Currently, established methods for near-term feedback on mitigation efforts are mainly based on the global surface air temperature (GSAT) variable alone. By these methods, a clear separation of socioeconomic pathways does not emerge before 20 to 30 years after emission separation due to the internal variability of the climate system. Here, we use a machine learning approach to create a separation of the climatic response from the socioeconomic development pathways based on ESM output data. This policy feedback strategy has not been described previously. Using 40 realizations of ACCESS-ESM1.5 under SSP1-2.6 and SSP5-8.5, where emission starts to differ in 2015, we estimate that a classification accuracy above 80 % is attainable by the appropriate feature-set/model combinations as early as 2026 based on the mean accuracy across 50 random states. However, the uncertainty of estimated accuracy is greatly reduced towards 2030–2040, indicating that real-world applications are not yet attainable. Our findings suggest that classification models trained on ESM-forecasts have the potential to become a powerful tool for providing early feedback on how the climate system responds to mitigation efforts.