Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Can machine learning predict citizen-reported angler behavior?

Julia S. Schmid, Sean Simmons, Mark A. Lewis, Mark S. Poesch, Pouria Ramazi

TL;DR

This study addresses predicting citizen-reported angler behavior using machine-learning models trained on auxiliary environmental, socioeconomic, and fisheries-management data across freshwater bodies in Canada. It defines two targets, the total fishing duration $D_{w,t}$ and the mean catch rate $C_{w,t}$, and evaluates predictions across multiple temporal resolutions and spatial extents using nine algorithms. The strongest performance occurs at the water-body level with monthly aggregation, yielding about 88% accuracy for $C_{w,t}$ and 87% for $D_{w,t}$, while regional and provincial daily predictions cluster near 60% accuracy; predictions at larger extents and finer resolutions are less accurate. Additional temporal features provide only modest gains, and a simple baseline often underperforms relative to feature-based models. Overall, the work demonstrates the feasibility of large-scale spatiotemporal prediction with auxiliary data for citizen-reported angler behavior, while highlighting data-bias limitations and guiding future directions toward feature selection, data expansion, and applications to conventional survey data or angler profiling.

Abstract

Prediction of angler behaviors, such as catch rates and angler pressure, is essential to maintaining fish populations and ensuring angler satisfaction. Angler behavior can partly be tracked by online platforms and mobile phone applications that provide fishing activities reported by recreational anglers. Moreover, angler behavior is known to be driven by local site attributes. Here, the prediction of citizen-reported angler behavior was investigated by machine-learning methods using auxiliary data on the environment, socioeconomics, fisheries management objectives, and events at a freshwater body. The goal was to determine whether auxiliary data alone could predict the reported behavior. Different spatial and temporal extents and temporal resolutions were considered. Accuracy scores averaged 88% for monthly predictions at single water bodies and 86% for spatial predictions on a day in a specific region across Canada. At other resolutions and scales, the models only achieved low prediction accuracy of around 60%. The study represents a first attempt at predicting angler behavior in time and space at a large scale and establishes a foundation for potential future expansions in various directions.

Can machine learning predict citizen-reported angler behavior?

TL;DR

This study addresses predicting citizen-reported angler behavior using machine-learning models trained on auxiliary environmental, socioeconomic, and fisheries-management data across freshwater bodies in Canada. It defines two targets, the total fishing duration and the mean catch rate , and evaluates predictions across multiple temporal resolutions and spatial extents using nine algorithms. The strongest performance occurs at the water-body level with monthly aggregation, yielding about 88% accuracy for and 87% for , while regional and provincial daily predictions cluster near 60% accuracy; predictions at larger extents and finer resolutions are less accurate. Additional temporal features provide only modest gains, and a simple baseline often underperforms relative to feature-based models. Overall, the work demonstrates the feasibility of large-scale spatiotemporal prediction with auxiliary data for citizen-reported angler behavior, while highlighting data-bias limitations and guiding future directions toward feature selection, data expansion, and applications to conventional survey data or angler profiling.

Abstract

Prediction of angler behaviors, such as catch rates and angler pressure, is essential to maintaining fish populations and ensuring angler satisfaction. Angler behavior can partly be tracked by online platforms and mobile phone applications that provide fishing activities reported by recreational anglers. Moreover, angler behavior is known to be driven by local site attributes. Here, the prediction of citizen-reported angler behavior was investigated by machine-learning methods using auxiliary data on the environment, socioeconomics, fisheries management objectives, and events at a freshwater body. The goal was to determine whether auxiliary data alone could predict the reported behavior. Different spatial and temporal extents and temporal resolutions were considered. Accuracy scores averaged 88% for monthly predictions at single water bodies and 86% for spatial predictions on a day in a specific region across Canada. At other resolutions and scales, the models only achieved low prediction accuracy of around 60%. The study represents a first attempt at predicting angler behavior in time and space at a large scale and establishes a foundation for potential future expansions in various directions.
Paper Structure (31 sections, 6 equations, 10 figures, 2 tables)

This paper contains 31 sections, 6 equations, 10 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Study area and data
  4. Citizen-reported angler behavior
  5. Features on the environment, socioeconomics, and fisheries management
  6. Model prediction tasks
  7. Target variables
  8. Different temporal resolutions
  9. Different spatial extents
  10. Different temporal extents
  11. Model training and testing
  12. Evaluation of predictions
  13. Modified prediction tasks
  14. Predictions with additional features
  15. Predictions without features
  16. ...and 16 more sections

Figures (10)

  • Figure 1: Prediction performance over time (and space) for the mean catch rate (left) and total fishing durations (right). Accuracy scores on test sets of data subsets for (A) the 20 water bodies with the most samples, (B) the 20 regions with the most samples, and (C) the three considered provinces (AB, BC and ON), water body type subsets (lakes, rivers) and the entire data set. Colors show different temporal aggregation levels of reported fishing trips. Marks show minimum, maximum and mean accuracy scores. N: Mean number of data samples for model training and testing, F: Mean number of features in the models. See Fig. \ref{['fig:BasicAnalysis_TempRes']} for more details.
  • Figure 2: Prediction performance over space for the mean catch rate (left) and total fishing duration (right). Accuracy scores on test sets of data subsets for (A) the 20 days, (B) the 20 weeks, and (C) the 20 months with most samples in a specific area. Bodies show different spatial extents of reported fishing trips (single regions, single provinces and entire study area (three provinces). Marks show minimum, maximum and mean accuracy scores. N: Mean number of data samples for model training and testing, F: Mean number of features in the models. See Fig. \ref{['fig:BasicAnalysis_SpatRes']} for more details.
  • Figure S1: Number of samples per water body (over open-water seasons of five years) and per day (over the three provinces). On average, a water body has 7 samples (days with reports) and a day has 41 samples (water bodies).
  • Figure S2: Distribution of the mean catch rate and total fishing duration in samples across the three provinces (entire data set).
  • Figure S3: Inter-annual curve of the mean catch rate (left) and total fishing duration (right) in samples across the three provinces (entire data set). The line shows the mean at the specific (A) date, (B) week number and (C) month over 5 years (2018-2022) and all water bodies.
  • ...and 5 more figures