Normally, I don’t trust the weather forecast to know the difference between sunshine or showers even a few days out. But a new study just published in Nature Geoscience indicates that meteorologists are close to being able to predict heat waves on the East Coast up to 50 days before they occur. The advance warning could help big cities, utilities companies, transit systems, and agricultural operations prepare for extreme temperatures — and, hopefully, prevent serious disruptions like power blackouts.

“There are many possible health, economic and other impacts of these hot temperatures and advanced warning can only help,” as climate scientist Jonathan Overpeck of the University of Arizona told the Associated Press.

But how is it even possible? So glad you asked. Basically, the study’s authors found a strong correlation between water temperatures in the north-central Pacific Ocean (the area just above Hawaii) and summer temperature spikes in the eastern United States. When the southern part of that Pacific region gets really hot, and the northern part is really cold, you have what scientists call the Pacific Extreme Pattern. Looking back into records of summer weather conditions from 1982 to 2015, the study’s authors found that when PEP was high, the eastern U.S. (everything east of the Mississippi River, including Iowa and minus Florida, according to the AP) tended to experience extreme heat.

The study’s authors concluded that the strong correlation “suggests potential for long-lead predictability.” Meaning when scientists observe those ocean temperatures, they can say with some certainty that a high-pressure system is about to park itself over the East Coast. That would certainly be useful, given the record-breaking heat the U.S. has experienced in recent years — and will likely continue to experience as heat waves become more frequent and intense in a warming world.

Meteorologists will start using these findings to predict heat waves starting in May, so keep your eyes peeled for something that quintuples your 10-day forecast — and is maybe that much more reliable, too.

See: