Developing a predictive model to identify Sea Lamprey parasitism on Lake Trout using biologgers

Sea Lamprey Petromyzon marinus remain problematic for Lake Trout Salvelinus namaycush restoration in the Laurentian Great Lakes. Fisheries assessments would benefit from knowledge of spatial–temporal patterns of Sea Lamprey parasitism on Lake Trout; however, such patterns are challenging to estimate from wounding rates on caught Lake Trout. Electronic tags have been used to identify distinct fish behaviors (e.g., foraging or spawning) using measurements of acceleration or heart rate. We hypothesized that Sea Lamprey attachment would elicit changes in the heart rate and swimming behavior of Lake Trout. Here, we determined whether tagging devices could record these changes and whether we could accurately predict lamprey attachment on Lake Trout using these recordings.

Adult Lake Trout (n = 34) were implanted with acceleration and heart rate tags and then were subjected to Sea Lamprey parasitism within a laboratory setting. Approximately 70 different acceleration and heart rate metrics were collected and tried as predictors of lamprey attachment. The top variables were used to train random forest models and then tried on test data sets. The accuracy of these models was then validated using a jackknife approach.

Metrics related to body orientation and heart rate were identified as the best predictors of Sea Lamprey attachment. The best models predicted lamprey attachments with high accuracy; however, individual‐level jackknife tests resulted in less accurate cross‐individual prediction and regularly predicted false negatives. These findings may be related to individual variance in the Lake Trout response to attachment, but there was evidence that the shifting of tags after implantation impacted predictive performance, which could be remedied with adjustments during implantation.

Our study highlights the potential to use tagging devices for quantifying Sea Lamprey attachments on Lake Trout in the wild. Further development appears necessary; however, once improved, these predictive models have the potential to generate field‐based estimates of Sea Lamprey attack rates on Lake Trout.