The idea that biodiversity is linked to how well ecosystems function can be traced back to Charles Darwin. However, serious experimental testing of this concept didn't begin until the early 1990s. One of the most important studies was the BIODEPTH project, which tested plant-species biodiversity and ecosystem properties in test plots at eight grassland sites around Europe. This generated a vast amount of data, and Andy Hector, who worked on the project, has now figured out a way to pull it all together.

BIODEPTH spanned three years and tested how different numbers of plant species affect ecosystem functions, such as total hay production or how well the ecosystem retains certain nutrients. The data produced were standardized and each ecosystem function or process analysed, but only on an individual basis.

Hector, now an ecologist at the University of Zurich in Switzerland, wanted to know how important biodiversity's role is when all of these processes are put together. After much thought, he came up with a way to do this — by switching from conventional statistical theory that relies on probability to information theory. This approach allowed all possible combinations of species to be compared and a best set of species to be selected for each ecosystem process.

At this point, Hector recruited his colleague Robert Bagchi, now at the University of Oxford, UK. “He brought the maths to formalize our basic ideas of analysis and the programming skills to implement it,” says Hector. The two used the Akaike Information Criteria (AIC) to calculate the set of species that influenced each of seven ecosystem processes at each site. Simply put, the AIC balances explanatory power and simplicity to find the best model.

Hector and Bagchi then determined the overlap between pairs of ecosystem processes for all possible pair combinations. The average overlap varied from 0.2 to 0.5, meaning that in any given pair only one-fifth to one-half of the species were important to both ecosystem functions. They conclude that individual analyses of ecosystem functions underestimate the biodiversity required to keep multifunctional ecosystems healthy (see page 188).

“So far, all the work on ecosystem functions has broken them apart individually,” says Hector. “We are the first to do a serious, quantitative analysis of multiple functions of ecosystems.” He admits their analysis has limitations, but says it's reasonable to expect that the findings for these test plots will extend to real-world ecosystems, typically valued for their multiple functions.

Although 'the more biodiversity the better' seems obvious for ecosystem functioning, Hector says scientists have been working on these questions for only a decade and are still arguing about how to analyse the data. He hopes to apply his scientific findings to real-world problems, such as reforestation. In this example, the choice of which species or mix of species to plant would involve looking at factors including total overall production (for future logging) and carbon sequestration.

The study also refutes the old adage that biologists shy away from mathematical heavy-lifting. “The mathematics of ecology may not be as elegant as that of other disciplines, but it is often very, very complex,” says Hector. “You need people who can work with partial differential equations who can also climb up a 70-metre rainforest tree to collect pollen.”