A sweeping machine learning study scanning over 86,000 high-res Mars images reveals that the mysterious streaks seen on Martian slopes are likely dry dust slides, not evidence of flowing water as previously hoped.
Long debated as possible indicators of present-day water (and thus habitability), these streaks—some recurring seasonally—now appear to be more aligned with wind activity and dust movement.
Doubt Cast on Martian Water Flows
A surprising new study from scientists at Brown University and the University of Bern is shaking up what we thought we knew about Mars. For years, mysterious dark streaks have been spotted crawling down the Red Planet’s cliffs and crater walls. Some experts believed these streaks were caused by liquid water, raising hopes that Mars might still have habitable environments today.
But the latest research tells a different story. Using machine learning and one of the most comprehensive datasets ever created for Martian slope features, the researchers found no signs of water. Instead, they discovered strong evidence that the streaks are likely the result of dry processes, such as shifting dust and wind activity.
“A big focus of Mars research is understanding modern-day processes on Mars — including the possibility of liquid water on the surface,” said Adomas Valantinas, a postdoctoral researcher at Brown who coauthored the research with Valentin Bickel, a researcher at Bern. “Our study reviewed these features but found no evidence of water. Our model favors dry formation processes.”
The research was published today (May 19) in Nature Communications.
across Mars’ dusty surface in Arabia Terra. Credit: NASA
Streaks Spotted Since the Viking Mission
These strange streaks were first captured in the 1970s by NASA’s Viking mission. They appear darker than the surrounding terrain and stretch for hundreds of meters down steep slopes. Some of them stick around for years, while others seem to appear and vanish with the seasons.
The shorter-lived versions are called recurring slope lineae, or RSLs. These tend to show up in the same places during the warmest times of the Martian year. That seasonal pattern led many scientists to believe water might be involved, possibly from melting underground ice, salty brines, or moisture pulled from the air.
If that were true, these streaks could represent rare, life-friendly oases on an otherwise barren world.
Not Everyone Bought the Water Hypothesis
Not all researchers were convinced by the water explanation. Some believed these streaks might just look like flowing liquid when viewed from space. In reality, they argued, the features could be caused by completely dry activity, like small rockslides or powerful gusts of Martian wind.
Hoping for new insights, Bickel and Valantinas turned to a machine learning algorithm to catalog as many slope streaks as they could. After training their algorithm on confirmed slope streak sightings, they used it to scan more than 86,000 high-resolution satellite images. The result was a first-of-its-kind global Martian map of slope streaks containing more than 500,000 streak features.
“Once we had this global map, we could compare it to databases and catalogs of other things like temperature, wind speed, hydration, rock slide activity, and other factors,” Bickel said. “Then we could look for correlations over hundreds of thousands of cases to better understand the conditions under which these features form.”
Wind, Dust, and the Real Culprits
This geostatistical analysis showed that slope streaks and RSLs are not generally associated with factors that suggest a liquid or frost origin, such as a specific slope orientation, high surface temperature fluctuations or high humidity. Instead, the study found that both features are more likely to form in places with above average wind speed and dust deposition, factors that point to a dry origin.
The researchers conclude that the streaks most likely form when layers of fine dust suddenly slide off steep slopes. The specific triggers may vary. Slope streaks appear more common near recent impact craters, where shockwaves might shake loose surface dust. RSLs, meanwhile, are more often found in places where dust devils or rockfalls are frequent.
Implications for Future Mars Missions
Taken together, the results cast new doubt on slope streaks and RSLs as habitable environments.
That has significant implications for future Mars exploration. While habitable environments might sound like good exploration targets, NASA would rather keep its distance. Any Earthly microbes that may have hitched a ride on a spacecraft could contaminate habitable Martian environments, complicating the search for Mars-based life. This study suggests that the contamination risk at slope streak sites isn’t much of a concern.
“That’s the advantage of this big data approach,” Valantinas said. “It helps us to rule out some hypotheses from orbit before we send spacecraft to explore.”
Reference: “Streaks on martian slopes are dry” by Valentin Tertius Bickel, and Adomas Valantinas, 19 May 2025, Nature Communications.
DOI: 10.1038/s41467-025-59395-w
Never miss a breakthrough: Join the SciTechDaily newsletter.
Follow us on Google and Google News.
28 Comments
What a deceptive little twat of a story. So, after “training it” on data that is admittedly a guess as to what it is, an AI identified it as something it was trained to say it was….Yeah, I wrote that correctly.
SO, it shouldn’t be hard to produce an image of the Earth doing the same thing, right? So where is it? I know proof is a weird thing to ask for buuuut we are making permanent conclusions soooo, where are the examples of this on Earth?
you are aware that mars is vastly different from earth? i mean, they are both orb shaped…. that’s about it. there’s been no definitive proof of water and there are many other things that exist in liquid form as well. being convinced that something exists without any proof is a shortcut to actual thought.
NASA is not going to reveal any truths they know…..on the Internet.
Like Mars is in Wyoming…
I agree with Glenn, if you don’t run your model AI against say the moon and earth, known information, it’s only possibility finding what it was asked to find. Wouldn’t be the first time bias was found in AI programming.
Carbon dioxide ice is common on Mars, and is sure to have mechanical surface properties that would facilitate enhanced flow downhill.
Dry ice on Earth has properties that are put to use. In ann evironment where CO2 ice is the ambient order, these streaks could be expected. Evaporates, re-freezes, and dust is re-covrred.
The moon doesn’t have wind or active geology and the Earth has too much water.
They are not compatible training tools and would taint the results.
The idea that they did not train the AI correctly is also evidence that you did not fully read or comprehend the article.
They stated that the algorithms were trained on tens of thousands of images (clearly with paired spectroscopic data) where they had already determined the lines to be dust.
They then used the trained algorithms to examine tens of thousands of additional images and the algorithm, examining patterns, determined that they were dust as well, not water.
I’m sure any anomalous image results were subject to much deeper human scrutiny as well.
People hear “AI” and they think Cortana from HALO: Fully capable of collecting and examining data in the wild and coming up with both a hypothesis and tests on the fly.
But real AI cannot do that.
Real AI is more like a Star Trek Tricorder: capable of collecting large volumes of data that it’s specifically told to look for and group according to detected patters.
That’s about it.
If the picture presented is representative then your AI is giving you a fairy tale. Obvious light colored sand slips ae seen in the lower left of the picture. They have a broad beginning as would be needed for the volume of sand needed to create the slide. Contrast that to the dark streaks which many have point like beginnings, are much longer, sinuous and begin at the top of a cone like structure. There is simply not enough sand possible, at the origen, to create a slide the length we see let alone sand of the right color.
Sand may be a componet of these streaks but a liquid carried it there!
Also as a note, explain a sand fountain without water ( note the mud cone like structureit it came from……maybe the “liquid” is from a CO2 (ultra cold gas flowing down hill coloring the sand Fe3+), geyser. Methane or ammonia geysers are also possible explanations. Check the spectrum of the streaks for elevated Fe 3+ or NaCl or both and you will get your answer.
I am no stranger to air photo interpretation or remote sensing and I agree with your analysis, especially the observation that the voluminous flows mostly seem to have a point source. It is difficult to reconcile the inferred volume of the “streaks” with the tiny sources unless one invokes a subsurface source of liquid. Also, liquids tend to darken soils and regoliths.
A problem that argues against it being some sort of ‘dust,’ is that silicates in general, indeed all oxides, tend to have a complex-refractive index with a small imaginary component (commonly called the ‘extinction coefficient’) that results in the appearance becoming lighter with finer particle sizes that approach the depth of penetration of light in a specimen. It is what is usually referred to as the ‘streak’ of a mineral on an unglazed porcelain plate.
The lede image (enlarged below, rotated 90 deg CW) shows an interesting feature for the left-most upper “streak.” There appear to be two dark, narrow bands of rock, like dikes or tilted strata, that cross the tip of the rock ridge that is the apparent source of the spreading “streaks.”
The image should be draped over a digital elevation model, as a stereo-pair, and examined in 3D with a stereoscope to be more certain about the relationships and down-slope directions. If it is dust, it could move up-slope under a strong wind. If it is liquid, the movement will always be down-slope.
What is needed is a little human intelligence, not artificial intelligence. Younger researchers are unnecessarily enamored with computers and haven’t learned to use the geological tools that were developed before they were born.
I took one two second glance at that photo and instantly concluded that it was the result of a minor landslide of the type that occurs frequently on sand dunes. I bet that the angle of the sun rays is wholly responsible for the dark smear effect. I believe you are clutching at misguided straws for some incomprensible reason.
David Harding, exactly what is your scientific or engineering discipline that allowed you to instantly know what is depicted in this photo? Or, are you a an avid skier, mountaineer, or exoplanetary geologist? As an engineering scientist, who was also an avid skier and mountaineer for much of my young adult life, I do not instantly see evidence of a landslide. The appearance is of flowing water or water ice. I believe you are clutching at misguided nonsense for some incomprehensible reason.
Your lack of programming experience is showing.
Your specious statement is as ridiculous as declaring “Dog trained to find people buried under rubble keeps finding people buried under rubble” to be a bad thing.
The Earth is not Mars. Earth has too much water, even in our dryest, harshest environments water plays it’s roles and our thicker atmosphere and higher gravity make Earth’s results useless for Mars research.
If Mars has polar ice caps then there most probably has been water there in the past if it was warmer. Why do scientists have to go to the complicated areas; can’t they drop the rover near the ice cap next time, ice is water when it’s warmer than 0’C Isn’t it.
The ice caps are composed of frozen carbon dioxide, not water ice. However water ice might be underneath the frozen CO2.
“… Not great news”
Data and its interpretation aren’t “great” or “not great” in this kind of scenario. It’s just analysis, not a crisis.
If Mars’ atmosphere doesn’t have enough pressure to support liquid water…why was anyone thinking it could be flowing water???
Because …there is still some gravity on Mars….and it is known there is still ice crystals in the more protected shadows , of the deeper asteroid strikes on Mars , as well as our moon….
There was nuclear war ! Cmon Nassa know that! On mars was like same on earth , there was people who start fighting like On earth – on one country press the button and another country press back ! At mommoent people destroid earth same like marss – and what Nassa , know that – they looking for new home because understand earth its finish ! And who gonna go on morss living ? Ofcorse They !! But no Nassa i promise that mission be filed , you guys stay back on earth – destroid earth and run away ? No chance , people just destroid everything, evrywhere, !
The above is why swipe typing is not a good idea.
Actually, swipe typing is more accurate for me. Illiteracy can’t be auto corrected though. As for Mars, wouldn’t there be rain clouds if water was present? Perhaps it’s too cold for evaporation though.
Big Dude…is that it? Swiping did that? Because I was left scratching my head…
Elon Stans who can’t wait to have particulate sorting cones like that are nonetheless using similar ML to try for nice lede images to follow on NASA’s and their own thesis. Give them a stochastic remote piezo band gap and they’ll terraform small arcs on that world?
Bollock.s bullshi..t rubbish!
Total garbage of an article and alleged research.
Agreed, DES57. Illiteracy appears rampant in the US, judging from the current state of their government. Far more likely than autocorrect, which whilst it may produce nonsensical sentences doesn’t typically butcher English.
I will have you know that not all Americans are gibbering dolts that will eat whatever they are spoon-feed. As for this article it is very clear that like most “scientists” today they found EXACTLY what they were looking for.
Where have you been hiding out ? Making nonsense comments , about Trump , as the world/Europe practically burns . As they all know Trump is doing things much better than the woke leaders in Europe…..As they falter , to even stay relevant , between China and US control of the larger world economy….
But there is hope for all , as we solve the Ai predicament, and help others with automation , and robots….As Europe is mighty shrinking in population , other than the Islam invasion , that is soon cutting out the Christian factions in all of Europe….Sweden now realizes , a lot 2 late it seems ? On taking on the soon battles in the streets , between any remaining Christians , and Islam followers…
We know nothing. Spectrum analysts shows there is water ice on the planet, but we can’t know, and neither can Ai, what is on the surface until it’s visited by a rover or a human.