The emergence of regional record-breaking summer heat predictability

Extreme heat is one of the most dangerous consequences of human-caused climate change, striking with a frequency and intensity that is accelerating every year. While scientists have long understood that a warming planet makes hot days more likely, predicting exactly when and where a specific, record-shattering summer will occur has remained a formidable challenge. This difficulty arises because weather is a chaotic mix of predictable long-term trends and unpredictable short-term fluctuations. For decades, the scientific community has debated how much of a heat event is driven by the steady hand of global warming versus the random wiggles of natural climate variability. The core question has been whether we can reliably forecast these deadly extremes years in advance, giving communities enough time to prepare, or if the inherent chaos of the atmosphere makes such long-range warnings impossible.

A team of researchers has now developed a new way to answer this question, demonstrating that the ability to predict record-breaking summers is not just a future possibility but is already emerging in many parts of the world. By using advanced computer learning tools, they found that combining the steady rise in global temperatures with specific patterns in ocean surface temperatures allows for accurate forecasts of extreme heat five to ten years before they happen. This approach outperforms traditional climate models in many regions, suggesting that we are entering an era where the timing of our hottest summers can be anticipated with significant confidence.

The researchers, Emily Gordon from the University of Auckland and Noah Diffenbaugh from Stanford University, focused on a specific type of prediction: determining whether a summer in the next five or ten years will be hotter than any summer recorded in the recent past. To do this, they trained a type of artificial intelligence known as a convolutional neural network. Think of this system as a highly skilled pattern recognizer that learns by studying vast amounts of historical data. The computer was fed two main types of information: the average temperature of the entire globe and the changing patterns of sea surface temperatures across the world's oceans. The goal was to see if the computer could learn to spot the specific combination of a warming planet and favorable ocean conditions that leads to a record-breaking summer.

When the team tested this method against real-world data from the last few decades, the results were striking. The computer learning model successfully identified the moments when a record-breaking summer was likely to occur, often predicting the event years in advance. For instance, in regions like Eastern Europe, the model detected the conditions that would lead to a new heat record around the early 2000s, correctly anticipating that the previous record would be broken. It also correctly predicted that no new record would be set in the 2010s, only to forecast another record-breaking summer arriving around 2024. This ability to distinguish between a hot year and a truly historic one proved that the model was capturing real signals rather than just guessing.

Crucially, the researchers found that this machine learning approach was often more accurate than the standard dynamical models used by climate scientists, which rely on complex physics equations to simulate the atmosphere and oceans. In many areas, including Northern Africa, the Mediterranean, East Asia, and parts of Australia and South America, the computer model predicted the onset of record heat with greater skill than the traditional models. The traditional models often struggled to capture the specific internal variations of the climate system that, when combined with global warming, trigger these extreme events. The machine learning model, by contrast, learned to recognize the subtle precursors in the ocean and atmosphere that signal a coming heatwave.

The study introduces a concept they call "predictability emergence." This describes the point at which global warming becomes strong enough that, when combined with natural ocean patterns, record-breaking heat becomes predictable. In the past, the random noise of natural climate variability was so loud that it drowned out the signal of global warming, making it impossible to forecast specific extreme events years ahead. However, as the planet continues to warm, the signal of that warming has grown louder. The researchers found that in many regions, this signal has already become strong enough to allow for reliable predictions. In some places, the ability to predict a record-breaking summer within a five-year window has already emerged, while in others, it is expected to become possible within the next decade or two if current warming trends continue.

The team also explored how far into the future these predictions could reach. They found that extending the forecast window from five years to ten years significantly improved the ability to predict record heat. This is simply because a longer time frame increases the statistical likelihood of an extreme event occurring. In many parts of the Northern Hemisphere, the model suggests that if we look ten years ahead, we can already skillfully predict the arrival of a new heat record, provided the ocean conditions are favorable. Even in regions where the warming signal is not yet dominant, the combination of a longer forecast window and the steady rise in global temperatures means that predictability is likely to emerge soon.

One of the most important findings of the study is that this predictability relies on both the long-term warming trend and the short-term fluctuations of the oceans. The researchers tested what would happen if they removed the ocean data from the model, leaving it with only the global temperature trend. They found that without the ocean information, the model's ability to predict specific record-breaking summers dropped significantly. This confirms that while global warming sets the stage for extreme heat, the specific timing and location of these events are still heavily influenced by the natural variability of the oceans. The machine learning model succeeds because it learns to read the script written by both the warming planet and the shifting seas.

The researchers also looked at how these predictions might change in the future. By simulating different levels of global warming, they estimated that for most land regions, the ability to predict a record-breaking summer within five years will become possible within the next 10 to 30 years, assuming the current rate of warming continues. In some areas, this threshold has already been crossed. The study suggests that as the planet warms further, the window for reliable prediction will widen, offering more time for communities to prepare for the inevitable arrival of extreme heat.

While the results are promising, the authors are careful to note the limitations of their work. The machine learning models were trained on data from a single climate simulation, and different models might yield slightly different results. Additionally, the models are better at predicting seasonal averages than specific, short-lived heatwaves, though the two are closely linked. The researchers also point out that if future ocean patterns evolve in ways that have never been seen before, the current predictions might need to be adjusted. However, the core finding remains robust: the combination of human-caused warming and natural ocean variability is creating a new era of predictability for extreme heat.

This work represents a significant shift in how scientists approach climate forecasting. Instead of viewing extreme events as purely chaotic and unpredictable, the study shows that they contain a predictable signal that grows stronger as the planet warms. By using machine learning to decode the complex relationship between the oceans and the atmosphere, researchers have found a way to look further into the future than ever before. This does not mean that the climate has become fully predictable, but it does mean that we are gaining the ability to anticipate the most dangerous heat events with a level of confidence that was previously out of reach. For communities around the world, this emerging ability to forecast record-breaking summers could be a vital tool for adaptation, offering a crucial window of time to protect lives and infrastructure from the escalating threat of extreme heat.