The new study, which was motivated in part by President Trump’s tweets about how a cold day in one particular location disproves global warming, uses statistical techniques and climate model simulations to evaluate how daily temperatures and humidity vary around the world. Scientists compared the spatial patterns of these variables with what physical science shows is expected because of climate change.
The study concludes that the spatial patterns of global temperature and humidity are, in fact, distinguishable from natural variability, and have a human component to them. Going further, the study concludes that the long-term climate trend in global average temperature can be predicted if you know a single day’s weather information worldwide.
Study co-author Reto Knutti, of ETH Zurich, said the research alters what we can say about how weather and climate change are connected.
“We’ve always said when you look at weather, that’s not the same as climate,” he said. “That’s still true locally; if you are in one particular place and you only know the weather right now, right here, there isn’t much you can say.”
However, on a global scale, that is no longer true, Knutti said. “Global mean temperature on a single day is already quite a bit shifted. You can see this human fingerprint in any single moment.
“Weather is climate change if you look over the whole globe,” he added.
The research uses the techniques applied in other “detection and attribution” studies that have sought to identify the signal of human-caused climate change in longer-term changes at the global level, such as the seasonal temperature cycle of the planet or heating of the oceans.
The authors, from research institutions in Switzerland and Norway, use machine learning to estimate how the patterns of temperature and moisture at daily, monthly and yearly time scales relate to two important climate change metrics: global average surface temperatures and the energy imbalance of the planet. Increasing amounts of greenhouse gases in the atmosphere are causing Earth to hold in more of the sun’s energy, leading to an energy surplus.
The researchers then utilized machine learning techniques to detect a global fingerprint of human-caused climate change from the relationships between the weather and global warming metrics, and compare it with historical weather data.
By doing this, scientists were able to tease out the signal of human-caused global warming from any single day of global weather observations since 2012. When looking at annual data, the human-caused climate signal emerged in 1999, the study found.
In what one outside expert, Michael Wehner of Lawrence Berkeley National Laboratory, deemed a “profoundly disturbing” result, the study found that the global warming fingerprint remained present even when the signal from the global average temperature trend was removed.
“This . . . is telling us that anthropogenic climate change has become so large that it exceeds even daily weather variability at the global scale,” Wehner said in an email. “This is disturbing as the Earth is on track for significantly more warming in even the most optimistic future scenarios.”
Stanford University climate scientist Noah Diffenbaugh, who was not involved in the study, said it advances our understanding of climate change’s effects.
“The fact that the influence of global warming can now be seen in the daily weather around the world — which in some ways is the noisiest manifestation — is another clear sign of how strong the signal of climate change has become,” he said in an email. “This study provides important new evidence that climate change is influencing the conditions that people and ecosystems are experiencing every day, all around the world.”
The research may provide a bridge between two approaches to detecting the human fingerprint on the changing climate. One of these techniques focuses on long-term trends, while another looks at regionally specific, shorter-term extreme weather events. Until this new study, there was no way to integrate these two specialties.
“Because it’s not possible to disentangle the fingerprint of climate change from natural internal variability for any particular extreme event, these studies use model simulations to estimate how the probabilities of such ‘class of events’ may have changed under anthropogenic climate change,” said study lead author Sebastian Sippel, of the Institute for Atmospheric and Climate Science at ETH Zurich.
“Our study could be seen also as linking these two sides of the same coin,” he said.
The study contains uncertainties, particularly when it comes to the accuracy of computer models in simulating various climate cycles. It also does not tease out the importance of other factors that influence the climate, such as land-use change and human-made and volcanic aerosols.
Knutti notes that the use of machine learning techniques, which can help tease out patterns in large data sets, can introduce uncertainties as well, although he’s confident those were minimized here.
While the new study does not attribute the climate change trends they found completely to human activities, Sippel said it’s unlikely there is another plausible explanation.
“We know from many other studies that the warming in the last 40 years is almost entirely human,” he said, adding that this is the subject of follow-up work.