Influence of reach and watershed characteristics on fish distributions in small streams of eastern Montana

Abstract

The prairie biome is arguably the most endangered ecoregion in North America, and the chance of extinction for many fishes is greater than in other ecoregions. Fisheries scientists and managers must understand the ecology of fishes at multiple spatial scales for effective conservation. Statistical models designed to explain distributions of fishes using environmental characteristics measured at multiple scales have provided this critical information. The objectives of this study were to: i) identify the reach- (i.e., physicochemical and biotic) and watershed-scale characteristics that affect the distribution (i.e., presence or absence) of fishes in prairie streams of Montana; and ii) identify which scale best explains the distribution of fishes in Montana prairie streams: reach, watershed, or a combination of variables measured at either scale (i.e., combined models). Reach and watershed information from 120 sites sampled between 1999 and 2004 was used to model the presence or absence of 20 species using tree classifiers. Models were evaluated by strict criteria in this study as compared to similar studies of prairie fishes. In this study, variables used in models had to reduce a significant amount of the deviance in species distributions before they could be evaluated for their predictive ability; other studies have only examined whether reach- or watershed-scale variables could be used to predict the occurrence of species.Only six reach, four watershed, and seven combined models were found to be significant, suggesting that explaining and predicting the distribution of fishes in eastern Montana is difficult. Significant models did provide information that could be used to guide conservation efforts. Most of the reach- and watershed-scale variables used in these models described large-scale ecological gradients. Additionally, results indicated that significant models could predict individual species distributions with a high level of accuracy. Previous predictive modeling studies have indicated that GIS-derived watershed characteristics better explain fish distributions than reach-scale characteristics. However, this study did not find that watershed-scale variables were significantly better at explaining or predicting fish distributions in eastern Montana. Results from this study suggest further information is needed to understand how spatial scale affects the distribution of fishes in warmwater streams of Montana.