A mix of data- and computation-intensive XSEDE resources are enabling researchers to scale up their climate, vegetation and agent-based human behavior models to tackle fundamental questions of how Homo sapiens came to dominate the planet, according to archeologist Colin Wren of Arizona State University (ASU). The international collaboration, led by Curtis Marean of ASU, will next use XSEDE machines to combine those models for a full simulation of the emergence of “modern” human behaviors, Wren said at the XSEDE15 conference in St. Louis on July 29.

“Fundamentally we don’t know exactly why our species was able to outcompete other hominids, and that’s one of the things driving this research,” Wren said after his presentation. “Part of the behavioral development in our species is hypersociality—the ability to cooperate with people who are not related to you in a systematic way, including defending resources. We’re trying to understand whether these characteristics … gave us a kind of special recipe that allowed us to take over the rest of the planet.”

Of particular interest to the researchers is the period between 100,000 and 200,000 years ago, when humans began to display signs of advanced cognition, social learning and cooperation far beyond the apparent abilities of other hominids. This period, which began with a glacial event, was clearly a turning point in our evolution, Wren noted.

Genetic evidence from modern humans suggests a major population bottleneck around this time; all of us may descend from 15,000 or fewer survivors. Archeological evidence suggests little or no human habitation except in the Cape Floral Region of South Africa at roughly this time as well. More intriguing, this is also when archeologists begin to see evidence of utilization of more difficult to gather but nutritionally valuable foods such as shellfish, artistic use of pigments and other advanced behaviors.

With help from XSEDE campus champion and project collaborator Eric Shook of Kent State University, the researchers harnessed the Pittsburgh Supercomputing Center’s memory-intensive Blacklight system and the Texas Advanced Computing Center’s computation-intensive Stampede to simulate the African climate in this period. The models suggest that when the rest of Africa was too arid for human habitation, the Cape Floral Region would have supported predictable, high quality and dense human food sources such as tuber plants and shellfish.

It’s the quality, quantity and seasonality of food available in the Cape Floral Region that Marean and his colleagues suspect may have spurred modern human behaviors.

“With resources that are as dense and predictable as these coastal resources, and the cognitive ability to exploit them, suddenly there is value in defending that resource,” Wren said. “You have to be able to predict when they’re going to be there, and make sure nobody else is going to exploit them.”

It may have been this imperative that provided a survival benefit for humans to cooperate in larger and ultimately unrelated groups. This partly genetic ability, which meant survival through glacial cycles, conveyed an overwhelming advantage over other hominids once modern humans began to spread out of Africa around 50,000 years ago.

“One possibility we are computing is whether we outcompeted other hominids directly or indirectly,” Wren said. “With the Neanderthals, there was some interbreeding; but we’re the only lineage that’s left.”

Currently the investigators are up-scaling their earlier models for climate, vegetation and human behavior to HPC platforms—Stampede for climate simulations; Blacklight for climate simulations, vegetative models and agent-based simulations; and San Diego Supercomputer Center’s Comet system for agent-based simulations. The ultimate goal will be to combine the simulations to test the hypothesis, along with further archeological evidence, that climate change led to population collapse, exploitation of new food sources, resource guarding, hypersociality and finally replacement of other hominids. Marean charts out the hypothesis in detail in Scientific American this month.

“We’ve used a number of different XSEDE systems for different purposes,” Wren said. “I see Comet as being particularly useful for the agent-based models.” Many of the new systems announced at XSEDE15 “are intended for the ‘long tail’ of scientific HPC use,” he added, “and I think archeology is going to be able to make a lot of use of them.”