Predicting Slowdowns in Decadal Climate Warming Trends With Explainable Neural Networks

The global mean surface temperature (GMST) record exhibits both interannual to multidecadal variability and a long‐term warming trend due to external climate forcing. To explore the predictability of temporary slowdowns in decadal warming, we apply an artificial neural network (ANN) to climate model data from the Community Earth System Model Version 2 Large Ensemble. Here, an ANN is tasked with whether or not there will be a slowdown in the rate of the GMST trend by using maps of ocean heat content (OHC) at the onset. Through a machine learning explainability method, we find the ANN is learning off‐equatorial patterns of anomalous OHC that resemble transitions in the phase of the Interdecadal Pacific Oscillation in order to make slowdown predictions. Finally, we test our ANN on observed historical data, which further reveals how explainable neural networks are useful tools for understanding decadal variability in both climate models and observations. Plain Language Summary Long‐term observations reveal that Earth's average temperature is rising due to human‐caused climate change. Along with this warming trend are also variations from year‐to‐year and even over multiple decades. This temperature variability is often tied to regional patterns of heat in the deep ocean, which can then modulate weather and climate extremes over land. In an attempt to better predict temperature variability on decadal timescales, we use a machine learning method called artificial neural networks and data from a climate model experiment, which was designed to compare climate change and variability. Here, our artificial neural network (ANN) uses maps of ocean heat to predict the onset of temporary slowdowns in the rate of global warming in both the climate model and in real‐world observations. We then use a visualization technique to find which areas of ocean heat that the ANN is using to make its correct predictions, which are found to be mainly across the Pacific Ocean. In agreement with recent research, our study finds that new data science methods, like machine learning, can be useful tools for predicting variations in global climate. Key Points An artificial neural network predicts the onset of slowdowns in decadal warming trends of global mean surface temperature Explainable AI reveals the neural network is leveraging tropical patterns of ocean heat content anomalies to make its predictions Transitions in the phase of the Interdecadal Pacific Oscillation are frequently assoc