Clearing up how cloud droplets freeze. (Outside-In).

A fresh look at old experimental data is threatening to overturn a longstanding theory about how water droplets freeze within clouds.

Suspended water droplets can remain liquid even when they and the air that surrounds them have temperatures far below the normal freezing point, says Azadeh Tabazadeh, an atmospheric chemist at NASA's Ames Research Center in Mountain View, Calif. Data collected in recent years show that clouds as cold as -37.5[degrees]C can still contain many supercooled droplets. Such droplets freeze solid almost instantly if they bump into each other or are otherwise disturbed.

Familiar as clouds are, the behavior of their constituent droplets remains only partly understood. This conceptual fog is particularly thick for conditions in which there aren't many particles in the air, says Tabazadeh.

Most scientists have long assumed that a tiny globule of pure, supercooled water, when disturbed, begins to freeze around an icy seed that suddenly forms inside it. According to this scenario, the time needed to freeze a given volume of water--say, 1 liter dispersed into a fine mist--is independent of the size of the individual droplets because the formation of a seed particle is a chance event.

But the results of recent laboratory experiments, when combined with information garnered from tests conducted as many as 30 years ago, don't back up that scenario. Together, the data indicate that the time needed to freeze a given volume of supercooled water varies drastically--by a factor of up to 100,000--according to droplet size, says Tabazadeh. She and her colleagues report their analysis in an upcoming Proceedings of the National Academy of Sciences.

This extreme variation makes sense if freezing begins at the surface of the drops, not at the cores. Dividing a given volume of water into a large number of small droplets yields more total surface area than if the volume is split into a small number of large drops, Tabazadeh explains. The freezing rate would then depend on the surface area.

The laws of thermodynamics also argue against ice nuclei forming inside supercooled droplets, says Howard Reiss, a physical chemist at the University of California, Los Angeles and a coauthor of the new report.

When water molecules begin to assemble into ice crystals, they release large amounts of latent heat. If that process occurred in the center of a supercooled droplet, the heat would remain trapped within the globule, slowing the freezing process. But if crystallization begins at the droplet's surface, latent heat can more easily transfer to the surrounding air. In this case, the droplets are so cold that heat released internally as the crystallization proceeds probably wouldn't melt the developing ice, says Tabazadeh.

Lightning, rainfall, and other meteorological phenomena vary with the ratio of water droplets and ice particles in clouds, says Lawrence Hipps, a meteorologist at Utah State University in Logan. Linking freezing rates of clouds to those atmospheric and other climate processes is one of the most unreliable areas in current climate simulations, Hipps adds. "It's important to understand how clouds operate if we ever expect to model them," he says.