Connectivity and conditional models of access and abundance of species in stream networks

Barriers to passage of aquatic organisms at stream road crossings are a major cause of habitat fragmentation in stream networks. Accordingly, large investments have been made to restore passage at these crossings, but often without estimation of population-level benefits. Here, we describe a broad-scale approach to quantifying the effectiveness of passage restoration in terms interpretable at population levels, namely numbers of fish and length of stream gained through restoration, by sampling abundance in a study design that accounts for variable biogeographic species pools, variable stream and barrier configurations, and variable probabilities of capture and detectability for multiple species. We modified an existing zero-inflated negative-binomial model to estimate the probability of site access, abundance conditional on access, and capture probability of individual fish. Therein, we modeled probability of access as a function of gradient, stream road-crossing type, and downstream access by fish simultaneously with a predictive model for abundance at sites accessible to fish. Results indicated that replacement of barriers with new crossing designs intended to allow for greater movement was associated with dramatically higher probability of access for all fishes, including migratory Pacific salmon, trout, sculpin, and lamprey. Conversely, existing non-replaced crossings negatively impacted fish distributions. Assuming no downstream constraints on access, we estimated the potential length of stream restored by the program ranged between 7.33 (lamprey) and 15.28 km (small coastal cutthroat and rainbow trout). These contributions represented a fraction of the total length available upstream (187 km) of replaced crossings. When limited ranges of species were considered, the estimated contributions of culvert replacement were reduced (1.65-km range, for longnose dace to 12.31 km for small coastal cutthroat and rainbow trout). Numbers of fish contributed ranged from less than 500 (longnose dace) to greater than 100 000 (sculpin). Although our framework can address the question of effectiveness in a broad array of stream and crossing configurations, much stronger inferences would be possible if future restoration efforts were designed to address the limitations we encountered in this study, particularly the lack of available information on crossings and species presence prior to restoration, and nonrandom selection of crossings to be replaced.