Building a fish passage assessment protocol for New Zealand: implementation of Bayesian network models for estimating fish passage success

Migratory fish species are dependent on connected habitats to complete their life cycles. Instream structures such as culverts, weirs and dams can impede the movement of migratory species. Disruptions to migratory pathways impact ecosystem health by reducing the abundance and diversity of species present. A number of metrics are available for quantifying habitat fragmentation within river networks, but they are dependent on sufficient information being available on the location and severity of migration barriers. Characterising the likelihood of fish passage success at instream structures requires information on the characteristics of the structure and the capabilities of fishes. Biotelemetry and mark-recapture studies are the most effective approaches for quantifying passage success, but are impractical for broad-scale evaluation of multiple instream structures. Bayesian networks offer a flexible approach for deriving probabilistic models suitable for broad-scale rapid assessment of instream structures for barrier severity. We present a Bayesian network derived for evaluating the probability of fish passage success at culverts in New Zealand. A formal expert elicitation process was utilised to populate the prior probability distributions in the model. We present the results from over 350 culverts where the model has been applied. By taking advantage of expert knowledge, the model offers a practical and objective approach for rapidly quantifying the likelihood of fish passage success at multiple instream structures without the need for resource intensive tagging studies. The results are also consistent with requirements for developing environmental reporting metrics for stream connectivity and the model has been used in a new fish passage assessment protocol for New Zealand