Machine-learning predictions of annual lake-level recreational boat traffic in Minnesota​

Molly I. Tilsen , Aaron P. Muehler, Paul A. Venturelli, Nicholas B.D. Phelps, Grant L. Vagle , Amy C. Kinsley

Recreational boating provides human health and economic benefits, but can also cause harm through introductions of aquatic invasive species (AIS). Minnesota, USA, has a large number of lakes, an active boating community, and a watercraft inspection program that collects data regarding boat launches at select sites. However, there is currently no direct monitoring system that measures boating activity across the state, requiring the synthesis of boating surveys to estimate boat traffic. We addressed this challenge by predicting lake-level boating traffic for all Minnesota lakes ≥ 10 acres using an XGBoost machine learning algorithm that integrated observed boat launch data and standardized site information. We estimated an average annual launch frequency of 7.14 launches per registered boat, totaling 5.8 million annual launches during 2018 through 2023 boating seasons. During this period ~10% of Minnesota’s lakes were monitored. Our model suggests that an average of 7.08% of launched boats were inspected; lakes received an annual mean of 634 boats (SD = 932), and lakes infested with aquatic invasive species received, on average, 1,120 more visits than uninfested lakes. This framework supports ongoing assessment of recreational boating pressure and provides a foundation for understanding boating patterns, invasion risk, and environmental management across large lake networks.