Ecologists drifted long ago from the simplistic model of the food chain to food webs, intricate, multi-tendril interactions between species that paint a more accurate picture of an ecosystem’s network. But, as with most sciences, as the models become more complex, so too does the analysis required to answer questions about the role each animal plays in an ecosystem. In a chain, if you remove one piece, the whole network falls apart. But how do you rank the importance of organisms in a system that looks like the tangle of wires behind an entertainment center?
Stefano Allesina, a brand new assistant professor in the University of Chicago Department of Ecology & Evolution (like, really brand new, as in moved into town last week) found the answer to this question in a brand name rapidly taking over our lives: Google. Specifically, he got a hunch that the algorithm Google uses to operate its search engine could be turned into a tool for detecting what species are most integral to an ecosystem’s health.
“One of the main problems in conservation is to forecast what’s going to happen if the species we are looking at is going down or going toward extinction,” Allesina said. “This single extinction can cascade in the loss of other species that are apparently unrelated, because all things are interdependent and it’s a very complex machinery. Or you could take away one piece and maybe the whole thing will reshape itself.”
So Allesina, and his collaborator Mercedes Pascual from the University of Michigan, constructed a computer model, published earlier this month in PLoS Computational Biology, to find vulnerabilities in an ecosystem. As Allesina describes it, they tried to help the cause of conservation by looking for the best way to destroy an ecosystem.
“How can we damage the network in the fastest possible way? How can we take away the most important species first so we can make the whole system collapse? It’s the best solution, but it’s actually not very good for the environment,” Allesina laughed.
Extracting such a solution from a typically tangled food web is a mammoth task without a shortcut – “It’s like finding a needle in a haystack that’s as big as the universe,” Allesina said. So the researchers turned to Google’s PageRank algorithm, which has a similar purpose: sniffing out the most relevant material from a vast morass of information. PageRank, the process that sends you to this site when you search “university chicago medical blog,” determines what web pages are most “important” by the number of other web pages, and the importance of those outgoing websites, sending a link that direction. So the New York Times comes up first when you search “new york newspaper,” while LoHud.com, aggregating the newspapers of New York’s Lower Hudson Valley, doesn’t show up until page 51 of the search results.
Allesina adapted the PageRank algorithm to biological systems where the “links” are not who sends web traffic to another site, but who eats whom. As the paper describes it, “a web page is important if important pages point to it, species are important if they point to important species.” This method is more accurate than simply counting the number of links between species, and designating the species with the most links as the most important. If that were the case, a Facebook model would work to describe an ecosystem, where the most “popular” species are the most important. But as Allesina says, the “friendships” in a food web are not equal to both sides.
“If you’re a sheep and I’m a wolf, we have very different kind of relationship,” Allesina said. “It’s one-way, at least in terms of energy.”
So the most important species in a given ecosystem are not those with the most links to other species, but those with the most links to species which are themselves important. For example, Allesina said, consider a species (let’s call it the Googlefish) that serves as a food web link between the deep part of a lake and the surface area. Even if the Googlefish only eats one or two species from the deep zone, and one or two species from the surface zone selectively eat it, the Googlefish’s extinction would be traumatic to the lake’s ecosystem. Suddenly, the surface zone predators would have no substitute prey, and they may slide toward extinction themselves. Meanwhile, the Googlefish’s prey, would thrive, having lost a predator that selectively feasted on them, and this surge could also throw off the ecosystem.
“A species is important if it supports on top of it some important species like a pillar,” Allesina said. “There are some that are critical, and if you knock them out, the whole building collapses.”
Ecologists possess other tools to locate these critical species, but they are much slower and more difficult to program, Allesina said. In the paper, the Google algorithm performed just as well as these older algorithms and far outshone simple “link-counting” models in 12 ecosystems used as test conditions. Allesina thinks that his model could eventually be used as an “alarm bell” for ecosystems, identifying key species where limited conservation funds should be directed. He’s also looking at ways that the same algorithm could be adapted to seek out vulnerable pieces of a terrorist organization or the metabolic pathways of certain cancers. But while Allesina said his model would be “dangerous” to use in its early, simple form, he hoped that it would demonstrate the need to consider a more holistic view of ecosystems in conservation efforts.
“You can’t conserve species in isolation; they’re not in deep space, they’re all interconnected,” Allesina said. “You have move to from a single species type of idea for conservation to multiple species and trying to conserve the nework rather than the single species.”