I hate clouds. Not because they sometimes bring rain but because they are hard. Clouds come in all shapes and sizes: wispy, high cirrus, puffy cumulus, the low, gray stratocumulus layers that blanket gloomy days. This great diversity makes it difficult to predict how clouds will react worldwide as the earth's atmosphere changes.

Climate scientists like me know from reams of data that the earth will warm up this century and beyond. But we are struggling to pin down how much hotter it will get: Perhaps another one degree Celsius? Or two degrees, or three or four? The answer depends largely on clouds. Climate change is affecting the distribution of clouds in the atmosphere, which could actually help slow down global warming—or speed it up. Knowing this outcome would be tremendously helpful in guiding actions the world takes today and tomorrow.

Large teams of experts have developed more than 20 sophisticated climate models, tested against extensive climate data. All the models show the earth warming in response to ongoing greenhouse gas emissions, but for years they stubbornly disagreed on clouds. That is starting to change. Simulations of cloud effects are beginning to converge. Satellite data and other observations are revealing how changing cloud cover is altering the planet. Do the new insights give us hope or extend our fears?

Big feedbacks or small

Imagine the earth just before the Industrial Revolution. Humans on six continents have cut down forests for pastures and towns. Yet the concentration of carbon dioxide in the atmosphere has been stable at about 280 parts per million (ppm) for thousands of years. Then the internal-combustion engine arrives. Fast-forward to the late 1900s, and CO2 concentrations have soared. The shock reverberates throughout the entire planetary system. The troposphere that holds the air we breathe is warming. By 2017 the CO2 concentration is above 400 ppm. The continents are heating up. The shallow oceans are, too. The circulation of air and water vapor in the atmosphere is beginning to change. As current trends continue, atmospheric CO2 levels double the preindustrial values by midcentury. More heating occurs. Finally, after several hundred years, the planet reaches a new equilibrium at a higher temperature.

The planetary response to carbon dioxide doubling is called equilibrium climate sensitivity, or ECS. Every climate model tells us that ECS is larger than zero: we should expect some warming. But the degree of warming they predict ranges from approximately two to 4.5 degrees C—from significant to catastrophic.

The models do not align, largely because they disagree on what clouds will do in the future. A better handle on clouds will allow us to narrow that range and make a much tighter prediction. But zeroing in on the influence of clouds is hard for two reasons. First, warming affects different types of clouds differently. Second, cloud changes affect warming in different ways.

This two-way interaction is called a feedback. Certain climate feedbacks are well understood. Sea ice, for example, is bright white and therefore reflects most of the sun's rays back into space, but as it melts it reveals darker water that does not reflect as much sun. That warms the air more, which melts more reflective ice and exposes more dark ocean, which reflects even less sun—and the feedback cycle builds, accelerating global warming. We understand this building, or “positive,” feedback well, and most models are in reasonable agreement about how it can affect climate change.

Understanding cloud feedback is more complicated. Like archivists in a natural history museum, climate scientists have created a rough taxonomy of clouds, grouping them by distinguishing features. Two basic properties are their height above the earth's surface and their opacity. Low clouds can be relatively transparent, like scattered puffs on a sunny day, or opaque, like uniform blankets of coastal fog. High clouds, too, can range from wisps through which almost all sunlight passes to the towering anvils that blacken skies during a thunderstorm.

This taxonomy is useful because it highlights the main ways in which clouds warm or cool the planet. Some clouds enhance the greenhouse effect. They trap heat rising from the earth and reradiate some of it toward outer space; the planet would be colder without them. Clouds in the cold upper reaches of the atmosphere are particularly effective in this regard.

Clouds cover more than 70 percent of the earth on a typical day, strongly influencing climate. Here they blanket the Pacific Ocean. Credit: Getty Images

Other clouds have the opposite effect: they prevent sunlight from reaching the earth's surface in the first place, keeping the planet cool. This effect is pronounced in low, thick clouds. In our current climate, that influence is larger than the cloud greenhouse effect. In fact, the net cooling of clouds today is immense, roughly five times greater than the warming of CO2 doubling.

This means that even small changes to cloud cover can have big impacts. Add more high, transparent clouds that let sunlight through but keep heat in, and the earth warms up. Add more low, opaque clouds that keep sunlight out, and it cools down. Cloud migration matters, too; redistributing reflective clouds from sunny tropical and subtropical latitudes to the cold, dark poles diminishes their cooling effect. Altitude is important as well; lifting high clouds even higher to colder upper reaches of the atmosphere increases their greenhouse effect. A warmer world may also change the ratio of ice crystals to water droplets in cold clouds, making them moister and thicker and therefore more efficient at blocking incoming sunlight.

None of these effects occurs in isolation, which is why models struggle. Some show feedbacks that are strongly positive—they amplify warming significantly. Some show feedbacks that are weakly negative—they slow warming slightly. The models that predict the strongest positive feedback end up predicting ECS at the high end of the range between two and 4.5 degrees C.

It is also no surprise that models do not simulate clouds well because clouds are simultaneously large and small. They are formed by tiny water droplets and ice crystals, yet they typically cover more than 70 percent of the earth at any given time. In programming a computer model, we must make a choice: zoom in and explicitly simulate the turbulent motions of droplets that make each cloud form and dissipate in a small area, or simulate the large-scale motions of rising and sinking air that distribute water vapor around the planet. We cannot do both, because it takes too much computing power to carefully track the behavior of every water droplet in the entire atmosphere at every moment.

We therefore try to combine small and large scales, knowing there will be compromises involved. A global climate model tries to find simplified parameters that describe the aggregate behavior. We develop these parameters based on the physics of the atmosphere and test and improve them through comparisons with finer-scale models run over small areas of the globe.

Still, there is no perfect way to mix large and small. But can we improve?

Changing forces

Let's tackle the first challenge: high clouds. Measurements give us good reasons to believe that climate change will literally reshape the atmosphere, pushing ever higher the boundary between the troposphere—the lower atmosphere where weather occurs—and the stratosphere right above it. High clouds, we suspect, will rise along with the rising boundary.

Mark Zelinka, a scientist at Lawrence Livermore National Laboratory, has thought a lot about the implications of this rise. As the planet warms from CO2, Zelinka says, it tries to cool itself by losing energy in the form of infrared radiation toward space. If high clouds remain at their typical altitude, they would warm in lockstep with the atmosphere and, in so doing, increase the amount of heat they lose to space. Zelinka and others think, however, that high clouds will rise so they can remain at nearly the same temperature they seem to prefer now. As a result, they would not radiate as much of the increasing heat energy to space, and that energy would instead further warm the atmosphere. This is a positive feedback: by rising more and more, high clouds further reduce a warming planet's ability to cool itself off.

Next, what about low clouds? Climate models seem to agree that a warmer world means fewer low clouds. Yet Mark Webb, a climate scientist at the Met Office—the U.K.'s national weather service—and the Cloud Feedback Model Intercomparison Project, knows it is more complicated than that. (It is either an interesting quirk or a reflection of our field's lack of diversity that so many cloud experts are named Mark.) Webb says he and colleagues are debating why a warmer planet might have fewer low clouds. The mechanism seems to depend on the way moist air in low clouds is diluted by the convection or turbulence of drier air above. Conventional models, Webb says, do not have the computing power to represent these local processes directly and end up approximating them differently. Various models show larger or smaller changes in low cloud cover, but crucially, most of them project reductions. Fewer low clouds means less sunlight is reflected to space—another positive feedback that amplifies warming.

There is yet another effect to consider. The atmosphere's overall circulation is largely driven by the differences in sunlight and temperature between the equator and the poles. Warm tropical air rises, cooling as it does. Once it is high in the sky, it starts moving laterally toward the colder poles. Along the way, it cools sufficiently to sink back down to the surface, at around 30 degrees latitude, warming and drying as it descends. On the ground, we get rainy climates under the tropical band of air that rises and sheds water as it cools, and we get desertlike climates under the bands where air descends.

Climate change will shift this pattern. The northern high latitudes will warm faster than the tropics, a phenomenon known as Arctic amplification, which reduces the temperature difference between the poles and the equator. The reduction, already under way, changes everything. Most important, perhaps, is that the tropics expand, pushing the rainy and dry bands toward the poles. One effect on land is that marginal zones—the Mediterranean, the Sahel, the American Southwest—will likely become drier. Indeed, satellite observations that I recently analyzed with Céline Bonfils of Lawrence Livermore show that precipitation patterns are shifting just as predicted. If clouds follow the migration, then decks of reflecting clouds may be pushed from lower to higher latitudes, where the incoming sunlight is weaker, reducing their cooling effect compared with their long-standing position over the tropics.

One more complicating factor must be figured into improved climate models: a warmer world can change the makeup of clouds. Clouds contain tiny water droplets and ice crystals. Thick, low clouds tend to be more watery and are more opaque than thin, high clouds, which tend to be more icy. In a warmer world, more ice in high clouds turns liquid, making them more opaque, blocking more incoming sunlight. Thawing ice clouds will become “juicier,” Zelinka says, providing a negative feedback—an important cooling check on warming.

A tighter prediction

The changing nature of clouds makes it seem even more difficult for models to narrow the anticipated rise of global temperatures—to reduce that two- to 4.5-degree C range. But one powerful data set matters more than any other: the history of what has already happened.

We have been measuring clouds almost since we began putting weather satellites into orbit in the 1980s. We can compare our models with actual observations to make our models better. Some of the older satellite measurements can be problematic, though. Cameralike instruments in Earth-observing satellites can find clouds by looking down for white objects against dark backgrounds, but they strain to differentiate between different white things, particularly icy clouds above snowy ground. Moreover, high clouds can obscure changes in lower cloud cover.

Credit: Jen Christiansen

Our observations have improved greatly in the past decade, however, thanks in large part to the A-train. NASA's Afternoon Constellation, or A-Train, is a collection of six Earth-observing satellites that fly in formation, burning fuel to keep a stable orbit. Two of them, CloudSat and CALIPSO, provide invaluable information. CloudSat uses radio waves that can easily penetrate high, thin clouds to measure low, thicker clouds. It can also tell if a cloud is raining or snowing. CALIPSO uses laser-based radar, known as lidar, to image clouds. It can tell whether clouds are made of ice crystals or liquid droplets.

Together these satellites have enhanced our understanding of cloud cover and given us hints about how clouds may change in the future. For example, the observations seem to support the notion that high clouds will rise higher as the planet warms, reducing the planet's ability to cool itself. And a recent study has shown that only some high clouds contain more water and less ice than anticipated. That means that the negative feedback associated with clouds becoming juicier may not be as strong as we had earlier thought.

CloudSat and CALIPSO were launched in 2006, so their data records are too short for us to detect climate change effects against the background of natural climate variability. To add perspective, scientists are patching together older observations from systems designed to monitor short-term weather trends. Two efforts of note are the International Satellite Cloud Climatology Project and the Pathfinder Atmospheres–Extended project. Unfortunately, says Mark Richardson of NASA's Jet Propulsion Laboratory, various weather satellites investigated by the projects were designed differently and took data at different times of day. Still, there are clues in these records if you know where to look. In a 2015 study, Zelinka and I gave it a try.

We began by asking a simple question: Where, in the observations, are the cloudiest and clearest latitudes on the earth? As expected, we found peak cloudiness in the tropics. Cloud cover was also relatively high in narrow bands in the midlatitudes, where storms are driven by the prevailing winds. In the subtropical “desert” latitudes, high atmospheric pressure led to dry, sunny conditions that impeded cloud formation—the clearest bands.

We then looked to see if the locations of the cloudiest and clearest latitudes changed over the course of the long-term weather satellite record, from 1984 to 2009. What we found was remarkable: the cloudiest midlatitudes and clearest subtropical latitudes were being pushed toward the poles, exactly as the models told us they would. Moreover, each of the independent data sets agreed that changing atmospheric circulation patterns were dragging cloud patterns toward the poles. By comparing this with climate models run in modes that do not include human emissions, we established that the changes were too large to be attributable to natural variability alone. And the changes were larger than scientists had predicted.

The implications are troubling. If decks of low, reflecting clouds are shoved too far toward the poles, then their cooling power will be substantially reduced: they will block weak, temperate sunlight instead of intense, tropical sunlight. This migration would constitute a strong positive feedback and indicate a higher climate sensitivity.

A subsequent study led by Joel Norris of the University of California, San Diego, that took into account known discrepancies in the satellite record found a poleward shift in the cloud patterns, too. These data also suggested that high clouds may be rising. Scientists are debating the significance of these changes and whether they can be attributed to greenhouse gas emissions, waning of particulates spewed into the atmosphere by the 1991 Mount Pinatubo eruption or natural climate variability, or some combination of these factors. But one thing is clear: the long-term observations do not show any indication that clouds will slow down warming.

The clouds won't save us

The picture that is emerging from the observations is becoming clearer. High clouds are rising, and cloud patterns are generally shifting toward the poles. Both trends would accelerate planetary warming. Short-term observations suggest that reductions in tropical clouds will block less sun, thereby enhancing warming, and that thawing clouds may be a weaker check on warming than we had previously thought. There is little here to comfort us.

So do we now think that clouds will steer warming closer to the upper end of the ECS range? Equilibrium climate sensitivity is a theoretical quantity. It describes the eventual climate response to the swift doubling of carbon dioxide in the atmosphere—an artificial scenario that gives us a very real way to explore. Increased CO2 is not theoretical, however; the doubling will happen by midcentury if nations worldwide continue on their present course. More satellite observations, higher-resolution models and creative, up-and-coming scientists will help us pin down the answer to how much hotter the planet will become.

In the meantime, work is needed on another, more relevant quantity that has also stubbornly refused to budge: the 50 percent of the U.S. population that does not accept the fact that humans are changing the climate. Ultimately, if CO2 emissions continue unabated, the earth will warm a lot. Clouds, it seems, will make matters worse and, at minimum, will do nothing to alleviate the problem. That task falls to us.