Wagner, T., S.R. Midway, T. Vidal, B.J. Irwin, and J.R. Jackson. 2016. Detecting unusual temporal patterns in fisheries time series data. Transactions of the American Fisheries Society 145:786-794.
Abstract
Fishery-independent surveys are an important source of information for making inferences about the temporal dynamics of ecologically and recreationally important fisheries. For example, time-series of catch per effort data are often examined for the presence of long-term trends. However, it is also of interest to know if certain sampled locations are exhibiting temporal patterns that deviate from the overall pattern exhibited across all sampled locations. These ‘unusual’ sites may be the result of site-specific abiotic (e.g., habitat) or biotic (e.g., the presence of an invasive species) factors that cause them to respond differently to natural or anthropogenic drivers of population dynamics or to management actions. We present a Bayesian model selection approach that allows for detection of ‘unique’ sites—locations that display temporal patterns with documentable inconsistencies from the overall global average temporal pattern. We apply this modeling approach to long-term gillnet data for yellow perch in Oneida Lake, but the approach is also relevant to shorter time-series data. We used this approach to identify six unique temporal patterns among 15 sampled locations, and describe the magnitude of difference between these patterns and the lake-wide average. Detection of unique sites may be informative to management decisions related to prioritization of rehabilitation or restoration efforts, stocking decisions, determination of fishable areas, and to further understanding changes to ecosystem dynamics.