Key points are not available for this paper at this time.
ABSTRACT Stream fragmentation is a leading threat to freshwater fish diversity and is often caused by the installation of impassable culverts at road–stream crossings. In the United States alone, millions of culverts contribute to this fragmentation. Therefore, estimating culvert barrier impact on aquatic organism passage and prioritizing barriers for removal are imperative. These prioritization efforts require inventory databases of culvert barriers, but conducting field surveys of all culverts is unrealistic. Prioritizing areas for inventory field surveys can be improved with statistical models quantifying relationships between remotely sensed watershed characteristics and culvert attributes. However, models linking culvert attributes to landscape predictors have not been compared across ecoregions. In this study, we investigated regional variation in the relationships between estimates of culvert attributes and landscape variables across the southeastern US. We used Bayesian hierarchical models to quantify the effects of stream gradient, watershed area, and land use on culvert outlet drop and openness within and among 15 southeastern US ecoregions. Overall, we found that the magnitudes and ranges of most predictor variable coefficient estimates depended on ecoregion identity. This regional variation was most apparent in the relative importance of predictor variables, with truly contradicting relationships being rare. These results imply that quantifying this regional variation is necessary for a fuller understanding of the relationships between watershed characteristics and culvert attributes, which itself is necessary for prioritizing areas for direct barrier measurements. Further investigating this variation will allow improved efficiency in building culvert barrier inventories, ultimately improving barrier removal efforts.
Sliger et al. (Tue,) studied this question.