Fish recruitment is complex and difficult to predict. Data-driven approaches show promise for predicting recruitment and understanding its drivers. We used a random forest model to infer relationships between year-class strength and 17 variables describing potential recruitment drivers across 30+ years of walleye (Sander vitreus) data from Minnesota’s nine largest inland lakes. Our model explained 20% of the variation in year-class strength overall, with predictive performance varying among lakes (–8% to 37% explained variance). Of the variables analyzed, degree-days during the first year of life and first winter severity were the most important for predicting recruitment, with relatively weak year classes predicted to occur with cold first growing seasons and severe first winters. Other thermal variables were secondarily important predictors of year-class strength. Predicted year-class strength was positively related to stock size and stocking and negatively related to the presence of invasive species; however, these variables were less important than thermal variables. Our results indicate that thermal conditions in early life can have a substantial impact on walleye recruitment and highlight the potential for differing recruitment drivers and dynamics among lakes.
|Original language||English (US)|
|Number of pages||13|
|Journal||Canadian Journal of Fisheries and Aquatic Sciences|
|State||Published - 2020|
Bibliographical noteFunding Information:
We thank the members of the Red Lake Department of Natural Resources, Dave Staples, Tony Kennedy, Ben Vondra, Carl Pederson, Brett Nelson, Eric Jensen, Matthew Hennen, Gerry Albert, Tyler Ahrenstorff, and Doug Schultz for their cooperation and assistance with data compilation and analysis and for providing useful insights. We also thank the members of the Hansen Lab and Greg Sass for providing helpful comments on the manuscript.
© 2020, Canadian Science Publishing. All rights reserved.