If there’s life — or evidence of past life — on Mars, it’s terrifyingly likely that we might accidentally miss finding it, maybe by just a few centimeters.
At a dry lakebed in Chile’s Atacama desert, SETI Institute astrobiologist Kimberley Warren-Rhodes and her colleagues recently found that communities of bacteria and other microscopic life tend to huddle in tiny patches of livable space, sometimes just a few square inches. These tiny environments are called microhabitats, and machine learning could help improve our odds of finding them on Mars.
Warren-Rhodes and her colleagues published their work in the journal Nature Astronomy.
A Little Slice of Mars on Earth
If you look at northern Chile’s Salar de Pajonales from the air, it looks strikingly similar to Mars Reconnaissance Orbiter photos of basins on Mars, whose salty surfaces are also laced with cracks and little ridges.
Pajonales was a lake, millions of years ago, but the water that once filled it has long since dried up, leaving behind a salty crust and some pockets of underground brine, nestled on the boundary between Chile’s Atacama Desert and its Altiplano, or high plains. The 64 square mile lakebed has a lot in common with some of the places scientists plan to search for evidence of ancient life on Mars: cold, salty, dry, and bombarded by ultraviolet rays because of its high altitude, Pajonales isn’t exactly a hospitable place, but it’s livable enough for ultra-hardy microbes.
But when Warren-Rhodes and her colleagues surveyed a chunk of the ancient lakebed, they found that only 9.2 percent of the area they’d searched actually hosted microscopic life, such as colonies of photosynthetic bacteria. There’s life in Pajonales, but if you just checked in a random spot in the lakebed, you’d have an 80.8 percent chance of missing it.
The Microbes Are in the Details
Since Pajonales is a decent analog for dry lakebeds on Mars, Warren-Rhodes and her colleagues wanted to figure out how Mars mission planners could improve their odds of finding life, or evidence of past life, in such a place. They explored the lakebed on foot and with UAVs, mapping where microbes lived — and they recorded their data at several scales, from the bird’s-eye view of a satellite like MRO or a helicopter like Ingenuity, to the view from a rover’s mast camera, to a zoomed-in scale of inches.
When looking at the environment from a few yards away, Warren-Rhodes and her colleagues noticed that bare, flat ground was almost always lifeless, but salt domes (mounds a few feet high) were home to life about 40 percent of the time. And patterned ground — mostly flat, but crisscrossed with small, ribbonlike bands of gypsum an inch or two high — was inhabited about half the time. Those two types of terrain, combined, only made up about a third of the lakebed, but they were the best places to look for life.
Even so, a hypothetical Mars mission that searched domes and patterned ground would still have just about a 50/50 chance of finding — or missing — traces of life, even in an inhabited place like Pajonales. That’s like flipping a coin to decide whether we discover alien life.
Zooming in further, Warren-Rhodes and her colleagues examined what they call “microhabitats,” which are tiny environments on a scale of just a few inches. It turned out that bacteria absolutely love to live in deposits of alabaster: a very porous, powdery form of gypsum. Small patches of alabaster are “almost universally inhabited,” write Warren-Rhodes and her colleagues. And microstructures — those little ridges that divide up patterned ground — are all teeming with photosynthetic microbes.
Meanwhile, microbes apparently snub sand and flat, bare ground. Crystals were hit or miss; microbes preferred to live in gypsum crystals if they were in domes, or if they had a loose outer layer of alabaster.
Alabaster tends to form at the base of ridges and domes, where humidity is a little higher, while crystals with alabaster layers tend to form near open cracks in domes, where there’s also a little more humidity. And alabaster’s inner structure is full of tiny pores, perfect for holding onto water as long as possible — and for transporting brine upward from underground.
What Does It All Mean?
Warren-Rhodes and her colleagues used all of that data to train a convolutional neural network to predict the chances of finding life in a particular spot, from a scale of miles to a scale of inches. And when they tested it on another area of the lakebed, the network predicted correctly between 70 percent and 90 percent of the time.
“Whereas a random search yielded a 9.2 percent probability of detecting biosignatures, the targeted search [...] guided by machine learning models, delivered up to an ~87.5 percent chance of locating biosignatures in the first sample,” write Warren-Rhodes and her colleagues.
Mission planners for NASA’s Mars rovers already target the areas considered most likely to hold traces of ancient alien life — they’re not out there just taking random samples and hoping for the best. In particular, rover teams look for places where water probably flowed or stood during the planet’s warmer, wetter past.
But Warren-Rhodes and her colleagues argue that the chances of success are much better if rovers target the right microhabitats — like a colony of photosynthetic bacteria thriving in a three-inch-wide patch of alabaster but not in the sand a few inches away. To do that, however, researchers need an intimate knowledge of each specific environment, on a scale of just a few inches. A dry lakebed like Pajonales has different microhabitats than an ancient river delta like Jezero Crater, for example.
Warren-Rhodes and her colleagues’ ultimate goal is a database of information – and trained neural networks — for many different analogs: environments here on Earth that share features with places on Mars.
“Such a library could assist future Mars mission scientists in the selection of facies, mineral assemblages, and structures with the highest chance of containing biosignatures,” Warren-Rhodes and her colleagues write.
That microhabitat library could also be useful in the search for life on other worlds, seemingly even more alien than Mars: icy moons like Titan, Enceladus, and Europa.
