Indicator Species and Scale of Observation

Abstract

Arthropods, because of their small size, diversity, and sensitivity to environmental variability, can be good indicators of habitat heterogeneity, ecosystem biodiversity, and environmental stress (Brown 1991; Hafernik 1992; Pearson & Cassola 1992; Kremen et al. 1993; Oliver 1993; Kremen 1994). Species richness among arthropod taxa can predict the richness of other taxa from an indicator (Pearson 1992), priority (Oliver 1993), or target (Kremen 1994) taxon. Erwin (1982) hypothesized that there were 30 million tropical arthropods, an estimate based partly on beetles comprising 40% of described taxa. May (1992) cited a lower estimate of six million, based on the percentage of British insects that were butterflies (0.3%). He pointed out, however, that relationships among taxa could vary from place to place, leading to great uncertainty in estimates of total richness. One factor in that uncertainty is that estimates of richness are invariably tied to areas of a particular size. Consequently, richness depends on scale of observation. If different taxa accumulate species at different rates as area increases, then correlations among taxa, or taxon percentages of total species richness, will vary with area size. Because a useful indicator has to be independent of sample size (Noss 1990), it is critical to measure the influence of sample size on indicator taxa. One way to test the influence of scale of observation is to collect samples in a hierarchical fashion so that smallscale samples can be aggregated into samples representing areas of increasing size. I sampled leaf litter invertebrates as part of the Missouri Ozark Forest Ecosystem Project (MOFEP), a large-scale, long-term study intended to examine the effects of timber harvesting on forest species. Litter samples were nested within plots nested within stands. By aggregating samples into plot data sets and plot data sets into stand data sets, the richness of eight arthropod taxa could be correlated across three scales of observation. By aggregating the stand data sets