Identifying fishing grounds from vessel tracks: model-based inference for small scale fisheries

Recent technological developments facilitate the collection of location data from fishing vessels at an increasing rate. The development of low-cost electronic systems allows tracking of small-scale fishing vessels, a sector of fishing fleets typically characterised by many, relatively small vessels. The imminent production of large spatial datasets for this previously data-poor sector, creates a challenge in terms of data analysis. Several methods have been used to infer the spatial distribution of fishing activities from positional data. Here, we compare five approaches using either vessel speed, or speed and turning angle, to infer fishing activity in the Scottish inshore fleet. We assess the performance of each approach using observational records of true vessel activity. Although results are similar across methods, a trip-based Gaussian mixture model provides the best overall performance and highest computational efficiency for our use-case, allowing accurate estimation of the spat...

近年来技术的迭代发展，推动了渔船位置数据的采集速率持续攀升。低成本电子系统的研发，实现了小型捕捞渔船的追踪——该船队板块通常以数量众多、体量相对偏小的渔船为典型特征。此前数据匮乏的这一作业领域，即将迎来大规模空间数据集（spatial datasets）的产出，这为数据分析工作带来了新的挑战。学界已提出多种方法，可通过位置数据（positional data）推演捕捞活动的空间分布特征。本研究针对苏格兰近岸捕捞船队，对比了五种基于渔船航速、或航速与转向角（turning angle）的捕捞活动推演方法。我们以渔船真实作业的观测记录作为基准，评估各方法的性能表现。尽管各方法的整体结果较为相近，但基于航次的高斯混合模型（Gaussian Mixture Model）在本研究场景下展现出最优的综合性能与最高的计算效率，可实现对空间信息的精准估算。