From structural diversity measures to ecosystem complexity: Experiments for deriving aggregated complexity indices

In times where biodiversity is globally under much pressure, effective monitoring of ecosystems is of great importance. As plants and particularly trees tend to shape the physical environment of ecosystems, indicators based on the structural complexity of plant communities are frequently used as surrogates for direct measures of biodiversity. A multitude of such quantitative diversity indicators exist and when considering multiple ecosystem services there is often the need to aggregate them in a single complexity index. We quantified the effects of four statistical techniques of aggregating contributing indices from three overarching tenets of α-diversity, i.e. location (or dispersion) diversity, size and species diversity. In addition we experimentally studied the influence of four different weights assigned to the contributing diversity measures. Overall the differences between the weights and aggregation methods used were comparatively small. Inverse correlation weights combined with arithmetic-geometric aggregation turned out to be the best choice for obtaining a clear complexity gradient for our study data from the boreal forest in Northern Sweden. In our analysis, it proved useful to rely on a small pool of global reference data with a strong structural gradient which served as contrasts and training data in addition to the data of our study plots. The application of random weights as statistical references was very helpful for understanding how weighting and index aggregation works. Our index-aggregation results suggested that our nine Swedish forest plots were at the lower end of global complexity and differed comparatively little in terms of forest structure.