Biodiversity assessment is a necessary component of much research in freshwater and benthic ecology, yet there is little evaluation of how well single habitat biodiversity assessments reflect overall biodiversity. In June/July 2005 a pilot study was conducted in two tributary watersheds of the Saint John River (New Brunswick, Canada) to determine sampling intensity required to assess biodiversity patterns in the Odonata. Sampling followed national protocols and focused on in-channel and off-channel habitats at a variety of sites in each watershed. A total of 34 species were collected across 40 sites in both watersheds. Although they do not saturate, rarefaction curves indicate greater diversity in off-channel vs. in-channel habitats. Cluster analysis (TWINSPAN) revealed a general separation between off-channel and in-channel sites based on species composition. Canonical Correspondence Analysis suggests that channel width and macrophytes are important factors controlling community composition of in-channel habitats, and pH and macrophytes are important factors for off-channel habitats. Results are discussed in the context of current freshwater biodiversity assessment programs, and we offer suggestions for refining our current sampling approach. Additionally we discuss the role of biodiversity assessment in assessing the effects of large-scale land-use patterns on benthic invertebrate diversity patterns, including community similarity.