In the search for life on exoplanets, finding nothing is something too

What if humanity’s search for life on other planets returns no hits? A team of researchers led by Dr. Daniel Angerhausen, a Physicist in Professor Sascha Quanz’s Exoplanets and Habitability Group at ETH Zurich and a SETI Institute affiliate, tackled this question by considering what could be learned about life in the universe if future surveys detect no signs of life on other planets.

The study, published in The Astronomical Journal and carried out within the framework of the Swiss National Center of Competence in Research, PlanetS, relies on a Bayesian statistical analysis to establish the minimum number of exoplanets that should be observed to obtain meaningful answers about the frequency of potentially inhabited worlds.

Accounting for uncertainty

The study concludes that if scientists were to examine 40 to 80 exoplanets and find a “perfect” no-detection outcome, they could confidently conclude that fewer than 10 to 20% of similar planets harbor life. In the Milky Way, this 10% would correspond to about 10 billion potentially inhabited planets.

This type of finding would enable researchers to put a meaningful upper limit on the prevalence of life in the universe, an estimate that has, so far, remained out of reach.

There is, however, a relevant catch in that ‘perfect’ null result: Every observation comes with a certain level of uncertainty, so it’s important to understand how this affects the robustness of the conclusions that may be drawn from the data.

Uncertainties in individual exoplanet observations take different forms: Interpretation uncertainty is linked to false negatives, which may correspond to missing a biosignature and mislabeling a world as uninhabited, whereas so-called sample uncertainty introduces biases in the observed samples. For example, if unrepresentative planets are included even though they fail to have certain agreed-upon requirements for the presence of life.

Asking the right questions

“It’s not just about how many planets we observe—it’s about asking the right questions and how confident we can be in seeing or not seeing what we’re searching for,” says Angerhausen. “If we’re not careful and are overconfident in our abilities to identify life, even a large survey could lead to misleading results.”

Such considerations are highly relevant to upcoming missions such as the international Large Interferometer for Exoplanets (LIFE) mission led by ETH Zurich. The goal of LIFE is to probe dozens of exoplanets similar in mass, radius, and temperature to Earth by studying their atmospheres for signs of water, oxygen, and even more complex biosignatures.

According to Angerhausen and collaborators, the good news is that the planned number of observations will be large enough to draw significant conclusions about the prevalence of life in Earth’s galactic neighborhood.

Still, the study stresses that even advanced instruments require careful accounting and quantification of uncertainties and biases to ensure that outcomes are statistically meaningful.

To address sample uncertainty, for instance, the authors point out that specific and measurable questions such as, “Which fraction of rocky planets in a solar system’s habitable zone show clear signs of water vapor, oxygen, and methane?” are preferable to the far more ambiguous, “How many planets have life?”

The influence of previous knowledge

Angerhausen and colleagues also studied how assumed previous knowledge—called a prior in Bayesian statistics—about given observation variables will affect the results of future surveys. For this purpose, they compared the outcomes of the Bayesian framework with those given by a different method, known as the Frequentist approach, which does not feature priors.

For the kind of sample size targeted by missions like LIFE, the influence of chosen priors on the results of the Bayesian analysis is found to be limited and, in this scenario, the two frameworks yield comparable results.

“In applied science, Bayesian and Frequentist statistics are sometimes interpreted as two competing schools of thought. As a statistician, I like to treat them as alternative and complementary ways to understand the world and interpret probabilities,” says co-author Emily Garvin, who’s currently a Ph.D. student in Quanz’s group.

Garvin focused on the Frequentist analysis that helped to corroborate the team’s results and to verify their approach and assumptions.

“Slight variations in a survey’s scientific goals may require different statistical methods to provide a reliable and precise answer,” notes Garvin.

“We wanted to show how distinct approaches provide a complementary understanding of the same dataset, and in this way present a roadmap for adopting different frameworks.”

This work shows why it’s so important to formulate the right research questions, to choose the appropriate methodology and to implement careful sampling designs for a reliable statistical interpretation of a study’s outcome.

“A single positive detection would change everything,” says Angerhausen, “but even if we don’t find life, we’ll be able to quantify how rare—or common—planets with detectable biosignatures really might be.”