Exploring a Data‐Driven Approach to Identify Regions of Change Associated With Future Climate Scenarios

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 4
• Main Authors: Labe, Zachary M., Delworth, Thomas L., Johnson, Nathaniel C., Cooke, William F.
• Format: Journal Article
• Language: English
• Published: Wiley 01.12.2024
• Subjects: climate change; climate mitigation; climate variability; large ensembles; machine learning; radiative forcing
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000327

A key consideration for evaluating climate projections is uncertainty in future radiative forcing scenarios. Although it is straightforward to monitor greenhouse gas concentrations and compare observations with specified climate scenarios, it remains less obvious how to detect and attribute regional pattern changes with plausible future mitigation scenarios. Here we introduce a machine learning approach for linking patterns of climate change with radiative forcing scenarios and use a feature attribution method to understand how these linkages are made. We train a neural network using output from the SPEAR Large Ensemble to classify whether temperature or precipitation maps are most likely to originate from one of several potential radiative forcing scenarios. Despite substantial atmospheric internal variability, the neural network learns to identify “fingerprint” patterns, including significant localized regions of change, that associate specific patterns of climate change with radiative forcing scenarios in each year of the simulations. We illustrate this using output from additional ensembles with sharp reductions in future greenhouse gases and highlight specific regions (in this example, the subpolar North Atlantic and Central Africa) that are critical for associating the new simulations with changes in radiative forcing scenarios. Overall, this framework suggests that explainable machine learning could provide one strategy for detecting a regional climate response to future mitigation efforts. Plain Language Summary There are several sources of uncertainties when considering future projections of climate change. This includes uncertainty related to natural climate variations, uncertainties related to biases and climate sensitivity among different models, and the uncertainty related to the trajectory of greenhouse gas emissions. We focus on this third source of uncertainty, which is typically considered by running a climate model with a range of scenarios that include varying amounts of greenhouse gases. Although comparing real‐world greenhouse gas levels with each climate scenario is a relatively simple task, it is harder to compare which climate scenario is most closely aligned with year‐to‐year patterns of weather and climate anomalies. In this study, we introduce a machine learning approach that learns to associate yearly maps of global temperature and precipitation with individual climate scenarios. We then compare how these future predictions of scenarios may change over time depending on the introduction of climate mitigation efforts and show regions that are particularly sensitive to this change. Our results indicate that starting mitigation efforts earlier leads to a lower emission scenario being predicted by the machine learning model in 2100 compared to delaying mitigation another 10 years, which also corresponds to reduced impacts from regional warming. Key Points A neural network applied to large ensembles can link annual mean maps of climate variables to a range of radiative forcing scenarios Information extracted from regional change patterns is used to distinguish between climate scenarios, even those with similar global warming Radiative forcing scenario classifications for the later 21st century are sensitive to a difference in the timing of mitigation by 10 years