On Sept. 28, Alex Cox GR and earth sciences professor C. Brenhin Keller published a new model to computationally determine the factors that led to the extinction of dinosaurs. According to Cox, their study, which was published in “Science,” suggests that a volcanic eruption contributed to the Cretaceous-Paleogene event or K-Pg event, names for the dinosaur extinction incident.
According to Cox, their model features Bayesian inversion, a statistical inference method, to determine which gasses were present at the time of the K-Pg event. The results from the model were then compared to existing geological records found in fossil isotopes, Cox said. If the output of the model was drastically different from observed records, the model then randomly adjusted the carbon dioxide and sulfur dioxide levels in its input to produce an output closer to known data.
“It keeps [inverting the results] over and over again, each time recording solutions it’s particularly fond of because it does particularly well and rejecting some that don’t do as well,” Cox said.
Cox explained that approaching a “hotly debated” topic with nuance and mathematical reasoning was his motivation for the project.
“Our only hypothesis really was that gasses can reproduce environmental conditions associated with the K-Pg event,” Cox said. “We did not hypothesize initially that the volcanoes caused the extinction, or the asteroid caused the extinction; [instead], we think this model can reproduce these conditions, and, to that, [we can] apply physical and geological deductions.”
The foundation of the model is a predictive, computational algorithm called Markov chain Monte Carlo, Cox said. According to computer science professor Peter Winkler, MCMC is a “clever” and “powerful” method that can make “forward predictions.” For example, given information about the current weather, MCMC can predict the weather of the same location several days in advance. Cox’s and Keller’s uses of MCMC, however, are different, since it relied on present data to theorize a past event as opposed to theorize a future event, a method better known as inversion, Winkler added.
“You [currently] have a model that predicts what you have, but not what the scientific truth was,” he said. “You want to try to figure out what the scientific truth was, and you do that by guessing the scientific truth and then seeing whether it fits the data that you actually have.”
After analyzing output data from their model, Cox said that he and Keller found that the gasses required to recreate the environmental conditions across the Cretaceous boundary were “consistent” with those of a volcanic eruption.
“We do notice that carbon dioxide and sulfur dioxide causes these large-scale environmental changes, and a relatively small percent of that comes from asteroids — most of it comes from the volcanoes,” Cox said. “We’ve known that volcanoes have probably been an agent in causing this extinction for a long time, [and] we’re the first people to maybe work backwards from the geological record to [derive] potential outputs.”
To determine the existing levels of gasses in fossil isotopes, Keller said the study used the Long-term Ocean-Atmosphere-Sediment Carbon Cycle Reservoir, a model that computes carbon distribution between different environments. Fossilized foraminifera, a microscopic organism which scientists often survey in geological records, helped determine carbon levels, he added.
“We wanted to see that given these records, what inputs of noxious gasses and volatiles could have produced the observed trends,” Keller said. “If [the gasses] were the things that mediated the environmental damage, how much was emitted when, and does that line up with the [asteroid] impact, volcanism or both?”
According to Keller, the model ran 128 computers in parallel, simplifying a process that would normally take weeks.
“This parallelization turns out [to be] quite efficient,” Keller said. “It [gave] us an entirely new estimate of how many new gasses were emitted, and that helped us disentangle what was emitted based on where they were being emitted.”
One of the most popular theories behind the K-Pg event is the Alvarez hypothesis, which posited that an asteroid impact caused the death of the dinosaurs, Keller said. However, Cox and Keller’s analysis of gas levels at that time found that an extinction could have occurred without an impact, as volcanic activity affects CO2 and SO2 concentrations.
“This is not to say that the impact played no role … the impact might have caused a nuclear winter from throwing dust into the atmosphere. That’s something we wouldn’t see in our inversion, because we’re not inverting for dust [as an input],” Keller said.
