We can't identify 99 percent of the species that have gone extinct so trying to keep everything we do know about from going extinct is tilting at nature's windmill.

But if we played Noah and had a giant ark and wanted to engage in ecological Jenga (build a tower, then remove blocks without causing the tower to collapse), which species should we save first to prevent a collapse in ecological function?

An international research team has developed mathematical tools that can estimate which species are most influential in a food web and they found that long-lived, generalist top predators — such as otters— play the most influential roles within a food web.  

Long-lived, generalist top predators can highly influence ecosystems because they feed on different types of prey that occupy different parts of the food web. For example, otters feed on a wide variety of aquatic prey and can influence multiple species throughout the course of their relatively long lifespan. Removing otters from the ecosystem would cause long-term disruptions to all those species, a theory that the new models can now confirm for other species and ecosystems.

Understanding how the gain or loss of a single species affects a complex food web has been a difficult mathematical challenge, and the new findings provide fundamental insights into complex natural systems. The new study offers a rule of thumb to help other studies focus their research and data collection on species in order of their expected importance, and increase the efficiency of their research effort.

Lead author Helge Aufderheide of the Max Planck Institute and University of Bristol said, "The interactions in an ecosystem are so complex that one can often only guess about the roles that each species plays. Therefore, knowing how to find the key players makes all the difference for understanding where to focus studies." 

The new approach has non-ecological applications as well, they note. Even though the research team applied the computational tools on food webs, their approach also can be applied to other types of complex systems , such as electrical grids or online social networks, to identify influential components.