Mars had water—until it didn’t. Scientists thinks that about four billion years ago, the planet had substantial amounts of liquid water on its surface, enough to form rivers, lakes, seas, and even oceans—and perhaps also to support life. But something happened in the following billion years, triggering the loss of this water from the surface until all that was left was the cold, dry wasteland of a world that we see today. Why and how that happened remains somewhat of a mystery. “We don’t exactly know why the water levels decreased and Mars became arid,” says Eva Scheller of the California Institute of Technology.

In recent years, results from NASA’s Mars-orbiting MAVEN spacecraft suggested the driver of this water depletion may have been atmospheric loss. Long ago, for reasons unknown, Mars lost its strong magnetic field, exposing the planet to atmosphere-eroding outbursts from the sun. As a result, much of Mars’s air escaped to space, presumably carrying away most of the planet’s water with it. But in a new paper published this week in the journal Science, Scheller and her colleagues argue this process alone cannot explain Mars’s modern-day aridity. Instead they say that a substantial amount of the planet’s water—between 30 and 99 percent—retreated into the crust, where it remains today, in a process known as crustal hydration.

“That loss [to space] would have to be very large to explain the loss of all of Mars’s water,” said Bethany Ehlmann of Caltech, a co-author on the study, in a press briefing at this week’s virtually hosted Lunar and Planetary Science Conference , where the research was presented. “We realized we needed to pay attention to the evidence from the last 10 to 15 years of Mars exploration in terms of the nature of water in the Martian crust.”

Using this swathe of evidence from a variety of Mars missions, the team found that the rate of atmospheric loss today was not enough to explain the disappearance of all Mars’s water.

Additionally, the observed ratio of deuterium to hydrogen in the Martian atmosphere—an important clue in working out its watery past—was also not consistent with all of the planet’s water being lost to space. Whereas hydrogen is light enough to easily slip away from a planet’s gravitational grip, the element’s heavier isotope deuterium cannot. Thus, a relative dearth of deuterium in the atmosphere today suggests that less water may have been lost in this way than was thought. An alternative explanation was needed.

Crustal hydration—in which water is incorporated into the crystalline structure of minerals—is a natural choice for that explanation. And in fact, it was previously proposed as an important mechanism for Martian water loss. Various lines of evidence convincingly show that the process must have occurred at certain points in the planet’s history. For example, results from a neutron spectrometer instrument on NASA’s Mars Odyssey spacecraft, which arrived at the planet in 2001, showed that, “basically everywhere, the crust had at least 2 percent water,” Ehlmann says. “In the equator, that’s water in soils and rocks.” Later findings from NASA’s Mars Reconnaissance Orbiter corroborated those results, mapping hydrated minerals on the surface of Mars. “It became very clear that it was common, and not rare, to find evidence of water alteration,” she adds.

This crustal-hydration scenario would not mean Mars hides a liquid-water wonderland in its subsurface. Rather, because the water would be locked in minerals, the Martian crust could be especially enriched in clays and hydrated salts. The fact that, on Earth, this process has not robbed us of our oceans may be linked to plate tectonics, which allow the rock-locked water to be efficiently released through volcanic activity. On a planet free of plate tectonics such as Mars, however, this water would remain trapped.

If Mars’s current rate of atmospheric loss is the same as it was long ago, then the figure of crustal hydration is likely closer to the 99 percent estimate, Scheller says. “But where we get uncertainty is what the atmospheric structure of Mars was like [in the past],” she says. “There are different elements that can make that loss rate to space become quite high.” One possible way is Martian dust storms, which can dramatically increase the loss rates, says Paul Mahaffy, director of the Solar System Exploration Division at NASA’s Goddard Space Flight Center and a principal investigator on instruments on the Curiosity rover on the surface of Mars and on MAVEN. During a global dust storm, he says, “a year’s worth of hydrogen from water could be lost in just 45 days. So the history of water loss over time [on Mars] is complex and not full constrained.”

No matter how high the loss rate was, however, a “significant amount of water would have been going into the crust,” Scheller says—likely more than half the planet’s total. The team estimates that Mars would have lost between 40 and 95 percent of its water via this process in the planet’s Noachian period, which stretches from 4.1 billion to 3.7 billion years ago. But even later in Mars’s history, bursts of volcanic activity could have recycled some of the subsurface moisture, potentially giving the planet’s habitability a much-needed boost. “You may have episodic habitability,” says Michael Meyer, lead scientist of NASA’s Mars Exploration Program at NASA headquarters in Washington, D.C. “The real question is what these [volcanic] rates were. We think water was available 3.5 billion years ago. What about three billion years ago?”

Understanding how and when Mars lost its water is therefore crucial to knowing if life could have existed there—and for how long. “The persistence of surface water could be highly relevant to the possible emergence and existence of life on Mars,” Mahaffy says. Current and future missions could help us better answer the question. One such effort is the International Mars Ice Mapper mission, a collaboration among NASA, Japan, Canada and Italy with a proposed launch later this decade. “Although it’s designed to look for water itself, it can give you [subsurface] layers,” Meyer says. “And if you’re able to identify what the layers are, you can do some volume calculations.”

Meanwhile NASA’s Perseverance rover, which landed on Mars last month, could also provide useful results on how extensive hydrated minerals are at its landing site, Jezero Crater. More importantly, it will collect samples that could help delve into this problem further once they are brought back to Earth next decade. “We can measure the deuterium-to-hydrogen ratio in the water in those,” Meyer says. “That will help us sort out what ancient parts of Mars [were like].”