A groundbreaking machine learning study analyzing over 86,000 high-resolution images of Mars has revealed that the enigmatic streaks observed on the planet's slopes are likely the result of dry dust slides rather than flowing water, which many had hoped to find. This research challenges long-held beliefs that these streaks could indicate the presence of liquid water and potentially habitable environments on Mars.
In a surprising turn of events, scientists from Brown University and the University of Bern have released a study that fundamentally alters our understanding of Mars. For decades, dark streaks have been noted on the cliffs and crater walls of the Red Planet. Initially, some experts posited that these streaks were caused by liquid water, igniting hopes for discovering habitable conditions on Mars. However, the latest findings present a different picture.
Utilizing advanced machine learning techniques along with one of the most extensive datasets of Martian slope features ever compiled, the researchers found no evidence supporting the existence of water in these streaks. Instead, they demonstrated strong indications that these features are primarily the result of dry processes, including dust movement and wind activity. Adomas Valantinas, a postdoctoral researcher at Brown, remarked, “Our model favors dry formation processes.”
The mysterious streaks were first documented in the 1970s by NASA’s Viking mission. These streaks, which appear darker than the surrounding terrain, can stretch for hundreds of meters down steep slopes. Some remain visible for years, while others exhibit seasonal patterns, leading scientists to classify them as recurring slope lineae (RSLs). These RSLs tend to appear in the warmest periods of the Martian year, sparking theories about the involvement of water, possibly from melting underground ice or brines.
Despite overwhelming interest in the water hypothesis, not all researchers were convinced. Some experts argued that the streaks could merely resemble flowing liquid when viewed from space, suggesting they might result from dry processes such as rockslides or strong winds. In a bid to clarify these findings, Bickel and Valantinas employed a sophisticated machine learning algorithm to catalog slope streaks. After training their algorithm with confirmed sightings, they analyzed over 86,000 high-resolution satellite images, culminating in a comprehensive global map of slope streaks with more than 500,000 documented features.
The geostatistical analysis yielded significant insights into the formation of slope streaks and RSLs. The data indicated that these features are not typically associated with conditions favourable to liquid water or frost, such as specific slope orientations, temperature fluctuations, or high humidity. Instead, the study revealed that these streaks are more prevalent in areas with elevated wind speeds and dust deposition, pointing to a predominantly dry origin. The researchers concluded that the streaks likely form when fine dust layers abruptly slide off steep slopes, with triggers varying from recent impact craters to frequent dust devils.
These findings cast doubt on the potential for slope streaks and RSLs to represent habitable environments on Mars, significantly impacting future exploration missions. While habitable sites are typically viewed as ideal exploration targets, the risk of contamination from Earth-based microbes poses a substantial concern. This study suggests that the likelihood of contamination at slope streak sites is minimal, alleviating some of the worries associated with exploring these areas.
Valantinas noted, “That’s the advantage of this big data approach. It helps us to rule out some hypotheses from orbit before we send spacecraft to explore.” The implications of this research could shift the focus of Martian exploration, directing efforts toward areas with more promising conditions for discovering life.
For further reading, refer to the study titled “Streaks on Martian slopes are dry” by Valentin Tertius Bickel and Adomas Valantinas, published on May 19, 2025, in Nature Communications. DOI: 10.1038/s41467-025-59395-w.
