Sustainable soil management requires simple, intuitive, and repeatable indicators for assessing changes in soil function over a range of temporal and spatial scales.  Soil microarthropods are closely associated with decomposition and nutrient cycles and may be particularly responsive indicators for soil management practices.  However, identification of appropriate bioindicator species for many systems has been severely limited by a lack of information on species taxonomy, distribution, and functional role.  Here, we evaluate Indicator Species Analysis (ISA; Dufrene and Legendre, 1997) as an objective method for assessing the indicator potential of different species without regard to their ecological role or expected management response.  Studies were conducted within the context of a meadow-forest gradient in the central Appalachian hill-lands a mosaic of small pasturelands (<200 acres) within a forested matrix.  Soils in meadow, forested, and edge plots were sampled monthly from April 2004 to April 2005 and extracted for soil microarthropods.  We then used ISA to (1) identify significant indicator species from soils under forest, meadow, and edge vegetation covers, (2) compare seasonal trends and vegetation classifications derived from the entire dataset and the reduced indicator species matrix using a combination of ordination, clustering, and multivariate hypothesis testing approaches, and (3) test the predictive power of bioindicator species to classify microarthropod communities from an independent set of samples.   

Results/Conclusions

Restricting ordination and site classification to significant indicator morphotaxa reduced the dimensionality of the community data matrix by 69% while only slightly decreasing the efficiency of unsupervised classification (from 87.2 to 84.4%); the percentage of total variability explained by first three PCA axes increased following ISA.  When these same indicator species were used to classify an independent set of samples, the percentage of total variability explained by the first three PCA axes increased from 64.2% to 77.1%; cluster analysis of the test dataset correctly classified 47 out of 50 plots by cover type (94% accuracy).  However, restriction of analysis to indicator species alone reduced detection of differences between sampling dates relative to the complete dataset.  Although care needs to be taken to ensure that the dataset used for indicator selection is fully representative of underlying temporal and spatial variability, ISA appears to overcome many of the limitations associated with parametric and multivariate approaches for identifying indicator species and has the potential to greatly reduce the taxonomic expertise and labor costs associated with sorting and identification of soil microarthropods.