Researchers have used deep learning to model more precisely than ever before how ice crystals form in the atmosphere. Their paperpublished this week in PNAS, hints at the potential to significantly increase the accuracy of weather and climate forecasting.
Researchers have used deep learning to predict how atoms and molecules behave. First, the models were trained on small simulations of 64 water molecules to help them predict how electrons in atoms interact. The models then replicated those interactions on a larger scale, with more atoms and molecules. It was this ability to accurately simulate electron interactions that allowed the team to accurately predict physical and chemical behavior.
“The properties of matter arise from the behavior of electrons,” says Pablo Piaggi, a researcher at Princeton University and lead author of the study. “Explicitly simulating what happens at that level is a way to capture a much richer physical phenomenon.”
This is the first time this method has been used to model something as complex as ice crystal formation, also known as ice nucleation. This is one of the first steps in cloud formation, where all precipitation comes from.
Xiaohong Liu, a professor of atmospheric sciences at Texas A&M University who was not involved in the study, says that half of all precipitation — whether it’s snow or rain or sleet — starts as ice crystals, which then grow and result in precipitation. If researchers could more accurately model ice nucleation, it could give a big boost to weather forecasting in general.
Ice nucleation is currently predicted based on laboratory experiments. Researchers collect data on ice formation under different laboratory conditions, and that data is fed into models to predict weather under similar real-world conditions. This method sometimes works well enough, but often ends up being imprecise due to the large number of variables involved in real-world weather conditions. If even a few factors vary between the lab and the real world, the results can be quite different.
“Your data is only valid for a specific region, temperature, or type of lab setting,” says Liu.
Predicting ice nucleation based on how electrons interact is much more accurate, but it is also computationally expensive. The researchers are required to model at least 4,000 to 100,000 water molecules, and even on supercomputers, such a simulation could take years. Even that could only model interactions for 100 picoseconds, or 10-10 seconds – not long enough to observe the ice nucleation process.
Using deep learning, however, the researchers were able to complete the calculations in just 10 days. The time duration was also 1000 times longer – still a fraction of a second, but enough to see nucleation.
Of course, more accurate ice nucleation models alone won’t make the forecast perfect, Liu says, since it’s only a small but critical component of weather modeling. Other aspects are also important – understanding how water droplets and ice crystals grow, for example, and how they move and interact under different conditions.
Still, being able to more accurately model how ice crystals form in the atmosphere would greatly improve weather predictions, especially those involving whether and how much rain or snow will fall. It could also help predict climate by improving the ability to model clouds, which affect the planet’s temperature in complex ways.
Piaggi says future research could model ice nucleation when substances like smoke are present in the air, potentially improving the model’s accuracy even more. Due to deep learning techniques, it is now possible to use electron interactions to model larger systems over longer periods of time.
“It basically broke new ground,” says Piaggi. “It already has and will have an even bigger role in simulations in chemistry and in our simulations of materials.”