Billions of years ago, Mars was a warm home to lakes and oceans. That is until these enormous liquid bodies on its surface vanished around 3 billion years ago. For years, scientists have assumed that this water disappeared into space when the planet’s atmosphere thinned out.

As it turns out, the water may not have gone up, up, and away. It might have actually gone in the opposite direction—underground. According to a new model developed by researchers at Caltech and published in Science today, between 30% and 99% of Mars’s ancient global water may still be found within the Martian crust.

“This work is standing on the shoulders of work that has gone for decades,” says Eva L. Scheller, a planetary geologist at Caltech and the lead author of the new study. “And more observational evidence has been leading us more and more toward thinking about the loss of water on Mars in new ways.”

Current estimates suggest Mars may have had between 100 and 1,500 meters global equivalent layer (m GEL) of water on its surface. (m GEL refers to a layer of 1 meter of water that would cover an even surface of the planet—Scheller says 1,000 m GEL is equivalent to roughly half the water of the Atlantic Ocean.) Even the lower end of this estimate is still plenty of water that potential life could have used to make a home for itself.

So learning how it disappeared is critical. If we know what happened, we could have a better understanding of what locations on Mars could have preserved evidence of any life that evolved during that time—and how current and future Mars missions could look for that evidence.

In most water loss models that assume atmospheric loss, the idea has been that UV radiation causes water high in the air to dissociate into hydrogen and oxygen. Both elements—but especially the lighter hydrogen molecules— escape the atmosphere and head into space. Scientists measure this hydrogen loss (using neutron detectors like the FREND instrument on ESA and Russia’s Trace Gas Orbiter) as a proxy for determining the rate of water loss on Mars over time.

However, there are two problems with this theory. For one, it doesn’t explain why TGO or other missions still detect so much water in the Martian crust. Second, the rate of hydrogen loss measured so far is too small to account for how much water we think Mars originally had. “It could only really account for the lower end of what most geologists think,” says Scheller.

At the same time, we now have a better understanding of how much water is buried within the Martian crust. Much of this is thanks in great part to rover missions like Curiosity that have studied Martian rocks directly, as well as lab analysis of meteorites from Mars that have landed on Earth. And all of that data has slowly led scientists to take more seriously the idea that the crust played a more significant part in the loss of water on Mars.

Now Scheller and her colleagues have come up with a new model that uses current data to examine whether the water could have gone underground instead.

This water would not have been sucked down into huge subterranean oceans. Instead, water molecules became incorporated into mineral structures like clays as a result of processes like weathering. The same happens here on Earth.

This process could account for anywhere between 30% and 99% of the total water loss in the planet’s first 1 to 2 billion years, according to the model. The atmospheric loss could make up the rest.

“It’s an extremely intriguing model,” says Joe Levy, a geologist at Colgate University, who wasn’t involved with the study. “Hydrated minerals and vein-forming minerals are almost everywhere we look on Mars. Runaway chemical weathering is a really provocative hypothesis to explain what happened to Mars’s water.”

A range of 30% to 99% is, of course, huge. That’s because we simply don’t know enough about the water content in the crust (least of all on a global scale), or what the ancient atmosphere of Mars looked like and to what extent it encouraged or limited atmospheric water loss. The model also attempts to take into account how geological activity in the ancient past (such as volcanism) could have affected these water loss mechanisms.

The model gives us new clues when it comes to Martian habitability. “The findings don’t just answer how Mars might have lost its water, but also when it lost its water,” says Scheller. The authors are certain the hydrated minerals in the crust are over 3 billion years old, which means Mars was potentially most habitable before that. Any search for evidence of ancient life would be best geared toward rocks that have been preserved from this earlier period.

Scheller suggests that both the Curiosity and Perseverance rovers may be able to look for samples within this time range. Perseverance in particular, whose mission is mainly dedicated to looking for evidence of Martian life, will explore a former lake bed that’s 3.8 billion years old. “It will be right there to investigate what might have been the mechanisms that caused water sequestration in these minerals in the crust,” says Scheller. Even if it cannot do the job on its own, it will capture samples that scientists could study for themselves in the lab.

Earth and Mars started out as very similar wet worlds but ended up taking drastically different paths. The loss of water to hydrated minerals in the crust isn’t unique to Mars; this happens on Earth all the time. But Earth benefits from the fact that its tectonic plates actively recycle its crustal rocks in a process that would release this water. Plus, it retained a thick atmosphere that kept the planet at the perfect temperature for life to evolve and thrive. Mars has no tectonic plates, and it hemorrhaged its atmosphere once its magnetic field shut down 4 billion years ago.

“Ultimately, this is the thing to keep in mind about habitability on terrestrial planets,” says Scheller. “It’s very fragile.”

According to