Date of Award

Fall 2006

Project Type


Program or Major

Natural Resources and Environmental Studies

Degree Name

Doctor of Philosophy

First Advisor

Robert T Eckert


Reference conditions play a vital role in many challenges facing both conservation and natural resources management. This dissertation sought to establish minimally-impacted reference conditions for stream biota and habitat in New Hampshire and explore alternative statistical methodologies to predict reference conditions for biological and habitat assessments. The fish, stream-dwelling salamander, macroinvertebrate and periphyton assemblages as well as the co-occurring physical habitat and riparian conditions of 76 minimally-impacted first to fourth order streams in New Hampshire were estimated using USEPA Environmental Monitoring and Assessment Program protocols over a four year period. Several statistical approaches and data standardizations for classifying multi-taxonomic assemblages were investigated for the strength of the classification they produced; log transformed abundances classified using TWINSPAN produced the best classification as measured using specific criteria. Seven natural biotic community types primarily arranged along the longitudinal stream profile were classified. Geographic classifications based on ecoregions and watersheds poorly explained organism distributions and abundances. Organism distributions were primarily associated with substrate characteristics, elevation, latitude, and the proportion of mesohabitat types (e.g. pool, riffle, etc.).

A new approach to constructing a biological assessment index that is based on the Bray-Curtis percent similarity between the observed and predicted communities was developed to allow taxa density information into the multivariate predictive assessments. Separate linear regression models to predict the densities of each taxon resulted in the most accurate predictions of expected community structure. Multivariate predictive models that included classification steps were not in general less accurate than approaches based on continuous prediction of taxon densities such as nearest-neighbor or ordination-based analyses. Including abundance information into the predictive models did not increase relative prediction error compared to an AusRivAS-style assessment index based solely on predicted taxon occurrences. Habitat prediction followed similar results. Inter-annual variation in three streams sampled every year of the study was highest in the vertebrates and lowest in the macroinvertebrates. In contrast, vertebrate assemblages were more resistant to a summer spate than the macroinvertebrates. Greater sampling intensity in the field and laboratory are probably the only remaining avenues for increasing assessment accuracies and reducing unexplained variation in reference conditions.