IRD, Univ Brest, CNRS, Ifremer, LEMAR, Plouzané, France.
Wildlife Conservation Society, Gabon Program, Libreville, Gabon.
PLoS One. 2020 Jun 10;15(6):e0234091. doi: 10.1371/journal.pone.0234091. eCollection 2020.
In many developing countries, small-scale fisheries provide employment and important food security for local populations. To support resource management, the description of the spatiotemporal extent of fisheries is necessary, but often poorly understood due to the diffuse nature of effort, operated from numerous small wooden vessels. Here, in Gabon, Central Africa, we applied Hidden Markov Models to detect fishing patterns in seven different fisheries (with different gears) from GPS data. Models were compared to information collected by on-board observers (7 trips) and, at a larger scale, to a visual interpretation method (99 trips). Models utilizing different sampling resolutions of GPS acquisition were also tested. Model prediction accuracy was high with GPS data sampling rates up to three minutes apart. The minor loss of accuracy linked to model classification is largely compensated by the savings in time required for analysis, especially in a context of nations or organizations with limited resources. This method could be applied to larger datasets at a national or international scale to identify and more adequately manage fishing effort.
在许多发展中国家，小规模渔业为当地居民提供了就业机会和重要的粮食安全保障。为了支持资源管理，有必要描述渔业的时空范围，但由于努力的分散性质，以及从众多小型木船进行作业，这通常难以理解。在这里，我们在中非加蓬，利用隐马尔可夫模型从 GPS 数据中检测七种不同渔业（使用不同渔具）的捕捞模式。模型与船上观察员收集的信息（7 次航行）进行了比较，并在更大的范围内与目视解释方法（99 次航行）进行了比较。我们还测试了利用不同 GPS 采集分辨率的模型。当 GPS 数据采样率间隔不超过三分钟时，模型预测精度很高。与模型分类相关的较小的准确性损失在很大程度上被分析所需时间的节省所补偿，尤其是在资源有限的国家或组织的背景下。该方法可以应用于更大的国家或国际数据集，以识别和更有效地管理捕捞活动。
