Wouldn’t it be easier to find life on other worlds if we knew exactly where to look? Researchers have limited ability to collect samples on Mars or elsewhere, or access remote sensing instruments when looking for life beyond Earth. In a paper published in Nature Astronomy, an interdisciplinary study led by Kim Warren-Rhodes, Senior Research Scientist at the SETI Institute, mapped the sparse life found in salt domes, rocks and crystals in the Salar de Pajonales on the border between the Chilean Atacama and Desert and Altiplano is hidden. They then trained a machine learning model to recognize the patterns and rules associated with their distributions, so it could learn to predict and find the same distributions in data it wasn’t trained on. In this case, by combining statistical ecology with AI/ML, scientists were able to locate and detect biosignatures up to 87.5% of the time (vs. ≤10% with random search) and reduced the area needed for the search by up to 97% reduce.
“Our framework allows us to combine the power of statistical ecology with machine learning to discover and predict the patterns and rules by which nature survives and disperses across Earth’s harshest landscapes,” said Rhodes. “We hope that other astrobiological teams will adapt our approach to mapping other habitable environments and biosignatures. With these models, we can design custom road maps and algorithms to guide rovers to locations most likely to harbor past or present life — no matter how hidden or rare.”
Ultimately, similar algorithms and machine learning models could be automated for many different types of habitable environments and biosignatures aboard planetary robots, efficiently guiding mission planners to areas of any size with the highest probability of containing life.
Rhodes and the team at the NASA Astrobiology Institute (NAI) SETI Institute used the Salar de Pajonales as a Mars analogue. Pajonales is a high altitude (3,541 m), high U/V, hyperarid, dry salt lake bed considered inhospitable to many life forms but still habitable.
During the NAI project’s field campaigns, the team collected over 7,765 images and 1,154 samples and tested instruments to detect photosynthetic microbes living in the salt domes, rocks and alabaster crystals. These microbes secrete pigments that represent a possible biosignature on NASA’s ladder of life recognition.
At Pajonales, aerial drone images combined simulated orbital data (HiRISE) with soil sampling and 3D topographical mapping to extract spatial patterns. The results of the study confirm (statistically) that microbial life at the terrestrial analog site of Pajonales is not randomly distributed but is concentrated in patchy biological hotspots that are strongly associated with km- to cm-scale water availability.
Next, the team trained convolutional neural networks (CNNs) to identify macro-scale geological structures at Pajonales – some of which, such as patterned soil or polygonal networks, are also found on Mars – and micro-scale substrates (or “microhabitats”) , which most likely contain biosignatures.
Like the Perseverance team on Mars, the researchers tested how a UAV/drone could be effectively integrated with ground-based rovers, drillers and instruments (e.g. VISIR on “MastCam-Z” and Raman on “SuperCam” on Mars 2020 Perseverance Rover). .
The team’s next research goal at Pajonales is to test the CNN’s ability to predict the location and distribution of ancient stromatolite fossils and halite microbiomes using the same machine learning programs to see if similar rules and models can be applied to others that are similar but lightweight different natural systems apply. From there, entirely new ecosystems such as hot springs, permafrost and rocks in the Dry Valleys are explored and mapped. As more evidence accumulates, hypotheses about the convergence of livelihoods for survival in extreme environments will be iteratively tested, and biosignature probability plans for Earth’s major analogous ecosystems and biomes will be inventoried.
“While the high detection rate of biosignatures is a key finding of this study, it is no less important that datasets with vastly different resolutions from orbit to ground were successfully integrated and finally linked regional orbital data to microbial habitats,” said Nathalie A Cabrol, who PI of the SETI Institute’s NAI team. “In doing so, our team has shown a path that allows the transition from the scales and resolutions needed to characterize habitability to those that can help us find life.” In this strategy, drones were essential, but so was conducting microbial ecology field surveys requiring extended periods (up to weeks) of on-site (and on-site) mapping in small areas, a strategy that was critical to assessing local environmental patterns characterize that are favorable for life niches.
This study, led by the SETI Institute’s NAI team, has paved the way for machine learning to help scientists search for biosignatures in the Universe. Their paper, Orbit-to-Ground Framework to Decode and Predict Biosignature Patterns in Terrestrial Analogues, is the culmination of five years of the NASA-funded NAI project and a collaborative astrobiology research effort involving over 50 team members from 17 institutions.
The SETI NAI team’s project, Changing Planetary Environments and the Fingerprints of Life, was funded by the NASA Astrobiology Program (Mary Voytek, Director) under Grant No. Funded NNA15BB01A
The publication appeared in Nature Astronomy.
