RESUMO
Biodiversity surveys are often hampered by the inability to control extraneous sources of variability introduced into comparisons of populations across a heterogenous landscape. If not specifically accounted for a priori, this noise can weaken comparisons between sites, and can make it difficult to draw inferences about specific ecological processes. We developed a terrain-based, paired-site sampling design to analyze differences in aquatic biodiversity between streams draining eastern hemlock (Tsuga canadensis) forests, and those draining mixed hardwood forests in Delaware Water Gap National Recreation Area (USA). The goal of this design was to minimize variance due to terrain influences on stream communities, while representing the range of hemlock dominated stream environments present in the park. We used geographic information systems (GIS) and cluster analysis to define and partition hemlock dominated streams into terrain types based on topographic variables and stream order. We computed similarity of forest stands within terrain types and used this information to pair hemlock-dominated streams with hardwood counterparts prior to sampling. We evaluated the effectiveness of the design through power analysis and found that power to detect differences in aquatic invertebrate taxa richness was highest when sites were paired and terrain type was included as a factor in the analysis. Precision of the estimated difference in mean richness was nearly doubled using the terrain-based, paired site design in comparison to other evaluated designs. Use of this method allowed us to sample stream communities representative of park-wide forest conditions while effectively controlling for landscape variability.