A terrain-based paired-site sampling design to assess biodiversity losses from eastern hemlock decline

Biodiversity surveys are often hampered by the inability tocontrol extraneous sources of variability introduced intocomparisons of populations across a heterogenous landscape. If not specifically accounted for a priori, this noisecan weaken comparisons between sites, and can make itdifficult to draw inferences about specific ecologicalprocesses. We developed a terrain-based, paired-sitesampling design to analyze differences in aquaticbiodiversity between streams draining eastern hemlock (Tsuga canadensis) forests, and those draining mixedhardwood forests in Delaware Water Gap National RecreationArea (USA). The goal of this design was to minimize variancedue to terrain influences on stream communities, whilerepresenting the range of hemlock dominated streamenvironments present in the park. We used geographicinformation systems (GIS) and cluster analysis to define andpartition hemlock dominated streams into terrain types basedon topographic variables and stream order. We computedsimilarity of forest stands within terrain types and usedthis information to pair hemlock-dominated streams withhardwood counterparts prior to sampling. We evaluated theeffectiveness of the design through power analysis and foundthat power to detect differences in aquatic invertebratetaxa richness was highest when sites were paired and terraintype was included as a factor in the analysis. Precision ofthe estimated difference in mean richness was nearly doubledusing the terrain-based, paired site design in comparison toother evaluated designs. Use of this method allowed us tosample stream communities representative of park-wide forestconditions while effectively controlling for landscapevariability.