By the end of this century, rising oceans will almost certainly flood the lands where tens of millions of people live as accelerating climate change warms the waters and melts ice sheets.
But precise estimates of the vulnerable populations depend on precise measurements of the planet’s topography, to understand just how close to sea level communities have settled.
A new study that seeks to correct for known errors in earlier elevation models finds that researchers might have been undercounting the number of people exposed to rising tides by hundreds of millions. That’s three to four times more people than previously projected, depending on the specific scenarios.
If these higher estimates prove correct, it will dramatically increase the damages and casualties from sea-level rise, swell the costs of adaption efforts like constructing higher seawalls, and escalate mass migration away from the coasts.
But it measured upper surfaces—say, the tops of skyscrapers or trees—rather than the bare ground. That skewed estimates of the elevation, particularly in densely vegetated or populated areas.
In a study published in Nature Communications on Tuesday, researchers at Climate Central used more reliable lidar data to train a form of machine learning to reduce the “vertical bias” in the NASA model.
The new analysis found that about 110 million people are already living on land that falls below daily average high tides today, compared with an estimated 28 million people under the earlier models.
By 2100, that figure could grow to 190 million people versus 50 million, assuming the world pumps out a relatively low amount of additional greenhouse-gas emissions. Under high estimates of both climate emissions and sea-level rise, some 340 million people would see their homes consistently inundated, compared with 94 million under the earlier model.
These numbers grow by tens of millions more when they include those living on land below annual flood levels as well as high tides.
Update: This story was updated to correct the researchers' affiliation.