As greenhouse gases build up in the atmosphere, the planet gets hotter, land-based ice melts, and warming ocean water expands. The result is sea level rise. A lot of scientific effort has gone into projecting how quickly and how high ocean waters will rise. But the ocean is only half the equation when it comes to the threat of sea level rise, and an update to the land side of the equation has dramatically increased the estimated number of people at risk.

“I think the whole climate science community has been very focused on understanding future sea levels because that's what's changing,” said Ben Strauss, CEO of Climate Central on Living Lab Radio Tuesday. “But, as we were so focused on that side of the equation, we missed a little bit paying attention to the quality of elevation data on land.”

Elevation data can come from several sources. Laser-based measurements, typically taken from a plane or helicopter, are the most accurate, but also most expensive. This kind of data is available for much of the U.S., western Europe, and Australia.

But for much of the rest of the world, elevation is likely to be based on more affordable — but also less accurate — radar measurements taken from satellites.

“When that radar strikes the Earth, its footprint is pretty large,” Strauss explained. “It sees buildings and treetops, as well as the ground, and blends them into a kind of average elevation that we found, on average, is more than six feet above the actual elevation of the ground.”

Six feet is a huge error. With sea level projected to rise between two and six feet by 2100, it could be the difference between dry and submerged.

And Strauss said the stakes are particularly high because of how many people live in potentially vulnerable coastal areas.

“We did a little calculation and looked at the first 10 feet or so above high tide and took an average and found that globally there are about 3 million people per vertical inch,” Strauss said. “So, when you hear about 1 or 2 inches of sea level rise, it may not sound like a lot, but our population — our development — is so concentrated in the very lowest elevation bands on our coasts that every inch means about 3 million people.”

Strauss said that he and his peers and colleagues studying sea level rise have known that the elevation data in widespread use had significant failings.

“But it was the best we had available,” he said, “and no one imagined that they could improve global elevation data without, say, sending a new satellite into orbit.”

Several years ago, Strauss and his colleague, Scott Kulp, realized they might be able to use machine learning to infer elevation in coastal areas where lots of trees and buildings were likely obscuring true ground level.

They developed an algorithm and tested it on coastlines with laser elevation data. They found that they could reduce the error from feet to inches. That’s the good news.

The bad news is that applying that algorithm to parts of the world with radar elevation data dramatically increases — under some conditions, triples — the estimated number of people living in areas likely to be flooded or submerged in coming decades. The increases are most notable in major cities in southeast Asia, such as Bangkok and Mumbai.

“I wouldn't say that we were surprised,” Strauss said, “but we were certainly sobered and there was a certain shock.”