Use of creel data to understand factors influencing angler decisions to harvest fish

Objective Understanding the “who, what, when, where, and why” of angler harvest is a critical component of managing fisheries with harvest-based regulations. Previous research has suggested that these are complex questions deserving further examination at broader spatiotemporal scales. Here, we use a long-term creel survey data set to develop models for predicting likelihood of fish harvest using a suite of explanatory variables. Methods We used creel survey data to summarize 831,479 records of fish angled from Kansas waters between 1997 and 2024 (excluding 2020 and 2021) to develop models for predicting likelihood of angler harvest. Our approach included the development of a global binary classification model using a training data set (70% of records) that included harvest or release as the response variable and a suite of 20 explanatory variables, including fish species or taxa. The model was then applied to a testing data set (30% of records) to measure accuracy. This process was repeated for the 12 most occurring species/taxa in the creel database with all explanatory variables, excluding species/taxa. Variable importance from the global model and summarized variable importance from the species/taxa model were used to quantify factors associated with an angler’s decision to harvest a captured fish. Results The global classification model included 13 of 20 predictor variables for node split determinations, although only combinations of legal harvest status, fish length, and species/taxa were included in the final model. When applied to testing data, we observed accurate classification in 84.3% of records. Species/taxa models displayed considerable variability in included variables. Among the 12 models, 18 of 19 variables were included in node split determinations by the classification algorithms. Model accuracy varied from 67.6% for sunfishes Lepomis spp. to 96.4% for Smallmouth Bass Micropterus dolomieu. Summarized variable importance across all species/taxa models identified fish length and legal harvest status as the most consequential variables influencing fate of captured fish. Conclusions Legal harvest status, fish length, and species/taxa were important determinants of the likelihood of fish harvest in our study. However, some species/taxa (e.g., black basses Micropterus spp.) were infrequently harvested regardless of other associated variables. The interplay between examined factors highlights the importance of understanding social norms associated with fisheries and potential limitations of widespread harvest-based regulations.