Compensation: an alternative method for analyzing diversity-productivity experiments


  • Peter B. Adler,

  • John B. Bradford

P. B. Adler and J. B. Bradford, Graduate Degree Program in Ecology and Dept of Rangeland Ecosystem Science, Natural Resources, Colorado State Univ., Fort Collins, CO 80523, USA (


Although recent experimental results demonstrate a positive effect of diversity on primary productivity, the interpretation of these experiments has been controversial, creating a need for new methods of analysis. The methods developed in response to this need all use the production of individual species grown in monocultures to calculate the expected production of each species mixture, then analyze departures from these expectations as a function of species richness. We propose an alternative method that treats the same assembly experiments as species removals, and calculates the expected production of each mixture based on the production of individual species when grown together in the full community (the experimental mixture containing all species in the pool). Using the observed production of the full community, and the observed and expected productions of less diverse mixtures, we calculate an index of compensation that measures the degree of functional recovery following species loss. To explore whether losses of dominant versus subordinate species have different ecosystem effects, we suggest a multiple regression approach that tests the influence of both species richness and expected production on compensation. If compensation varies with species richness or expected production consistently in many experimental systems, then we may be able to predict the ecosystem effect of different types of extinctions.
While existing monoculture approaches more directly test hypotheses about complementary resource use, the compensation approach offers two advantages: 1) it is more appropriate for testing how extinctions will affect ecosystem function, and 2) it may provide an important link between assembly experiments in artificial communities and removal experiments in natural systems.