Aiding Rare Species Management Efforts through the Use of Mixed Models

Monitoring programs are frequently relied upon to support conservation decisions. Quantifying the population dynamics of rare and elusive species (e.g., threatened or endangered species), however, is a challenging process for both methodological and statistical reasons. Such species often occur in low numbers and may also be behaviorally cryptic. As a result, monitoring programs often produce data sets dominated by zero counts but also contain some observations of high numbers. Applying traditional statistical approaches to zero-inflated data typically requires data manipulation (e.g., making zero observations positive), and violations of statistical assumptions can still occur. Alternatively, mixed models can be applied to non-normal observations in order to assess potential relationships between abundance and landscape-level factors (e.g., those associated with surface coal mining, changing agricultural practices, and land development) and to partition variability occurring over time and space. Here, we fit a negative binomial mixed model to zero-inflated abundance data for blackside dace (Chrosomus cumberlandensis), in order to aid managers in identifying streams where future monitoring resources may be best allocated. Specifically, we hypothesized a negative relationship between dace abundance and water conductivity, while quantifying spatiotemporal variability associated with surveys from the Cumberland River system in Tennessee and Kentucky.