One reason climate scientists have been able to confidently determine that humans are responsible for modern warming is that they have more than just weather records to work with. There are many places where a human cause can be identified if you know how to dust for fingerprints. For example, while the lower atmosphere warms, the stratosphere is actually cooling. That’s what you expect when greenhouse gases—rather than the Sun—are behind the warming.

A new study led by Lawrence Livermore National Laboratory’s Ben Santer looked for fingerprints in a new place: the seasonal cycle of temperatures. The ideal tool for analyzing this is the global temperature record produced by satellites, which began their watch in 1979. That means they don’t go back nearly as far as weather-station records, but the dataset is now long enough to be useful for studies like this.

Hot and cold
While everyone uses the same satellites, several different groups actually maintain separate satellite temperature datasets. This is because the measurements are far from straightforward, and a ton of work goes into all the necessary processing to spit out temperature maps. As a result, the different datasets don’t always line up perfectly with each other—or with those analyzed with previous versions of their processing algorithm. So in this study, the researchers used the most recent two versions of three different datasets.

Each dataset tracks a few different layers of the atmosphere. One record covers the lower troposphere—the first 10 kilometers above the Earth’s surface. Although that’s the closest thing to the surface temperatures we live in, it’s also the trickiest measurement to get right. A cleaner record centers on the middle troposphere, a bit higher up.

By tracking the difference between the coldest winter months and warmest summer months—the magnitude of the annual cycle of seasons—some interesting regional patterns pop out. (You can see this in the image at the top of the page.)

If you average together the mid-latitude stripe of the Northern Hemisphere, there is a larger seasonal temperature swing than in the Southern Hemisphere, because there is a much greater area of land (which warms and cools to greater extremes than the ocean does). But this seasonal cycle has also increased measurably since 1979—more so in the Northern Hemisphere—as a result of summer temperatures in the atmosphere rising faster than winter temperatures.

In the tropics, the seasonal cycle is basically unchanged. Near the poles, on the other hand, the satellites show that the seasonal temperature swing has decreased.

There is an interesting exception that doesn't show this pattern: the University of Alabama at Huntsville dataset, which is run by Roy Spencer and John Christy (two of the small handful of vocal contrarian scientists who reject or downplay human-caused climate change). That dataset looks totally different in the Antarctic. In fact, the weirdness in the Huntsville data only appears in its latest version update and seems to reflect a problem handling the transition between first-generation and second-generation satellites. There is a history of the managers of the other major satellite datasets discovering errors in the Huntsville algorithm, and this seems to add another example.

Models and reality
So what do these altered seasonal patterns mean? To answer that question, Santer and the other researchers used a collection of major climate-model simulations. That includes long simulations of an unchanging (pre-Industrial Revolution) climate, as well as simulations of the human-caused warming through 2016. That allowed them to analyze the seasonal patterns climate models predict for a warming world, as well as the extent to which these seasonal patterns can vary naturally.

The researchers found that the models predict nearly this exact seasonal pattern—greater seasonal swings in the mid-latitudes (especially for the Northern Hemisphere), little change in the tropics, and smaller swings in the Antarctic. The models also correctly predict a smaller seasonal cycle around India and Southeast Asia, which deviates from the general mid-latitude trend.

The one mismatch is in the Arctic, where the satellite data shows a stronger decline in the seasonal cycle than the average model predicted. About one-third of the models show a decline, but the rest did not. The models have generally underestimated the loss of Arctic sea ice, and it’s possible that is the key here, since sea ice loss is an important driver of the seasonal cycle change.

It’s us, version 857

To test how strongly the satellite-observed changes point to human-caused warming, the researchers used a signal-to-noise analysis. Using the pre-Industrial Revolution simulations to estimate the range of natural variability, the analysis showed that the observed seasonal patterns in the middle troposphere have clearly emerged above the noise. That makes them a clear manifestation of the patterns predicted by the models. It’s even clearer than the change in global average temperatures, in fact, because seasonal patterns across entire latitude bands won’t vary much in a stable climate.

While they were at it, the researchers repeated these analyses for a more direct measure of climate change: the annual average of temperatures, analyzed at locations around the world. They write, “We find here that, for annual mean [mid-troposphere temperatures], the estimated [signal-to-noise] ratios exceed 4.4 for temperature changes over the 38-year satellite record. This translates to odds of roughly 5 in 1 million of obtaining the annual mean [signal-to-noise] ratios by natural variability alone.”

Overall, they conclude, “The best explanation for these results is that basic physics and basic physical mechanisms are driving the large-scale changes in [seasonal patterns]. For tropospheric temperature, a human-caused signal is now evident in the seasonal cycle itself.”

That’s just one more reason scientists can be sure that humans are driving climate change.