Scholarlink

<p>Species distribution models (SDMs) are commonly used in predicting fish species distribution pattern and its change. However, the data collected from fishery-independent surveys may have deviations from the natural situation, leading to bias in ecological studies. This study aimed to explore the impact of sampling designs of fishery-independent surveys on the prediction of SDMs for marine fish species. The sampling designs included simple random sampling (SRS), stratified random sampling (StRS), systematic sampling (SYS) and stratified systematic sampling (StSS). Generalized linear model (GLM), generalized additive model (GAM), random forest (RF), classification and regression tree (CART), and support vector machine (SVM) were developed based on the simulated data from different sampling designs to predict the spatial distribution of target fish species. The predictive abilities of all SDMs enhanced and the differences between the prediction performances of SDMs decreased with increasing sample size. SDMs using the data from SYS always generated more accurate predictions than those from other sampling designs, particularly in the case of GAM and GLM when the sample sizes were small. The predictive ability of GAM was unstable with small sample size, but was getting similar with those of RF and SVM with increasing sample size. The predictive abilities of SDMs were influenced by the data collected from different sampling designs in fishery-independent surveys. Thus, sampling designs from which data are collected should be considered when choosing proper SDMs to investigate the spatial distribution of fish species.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction Performances of Species Distribution Models for Marine Fish Species Vary with Sampling Design and Model Algorithm

  • 期刊论文

  • 发表日期：

    2026年5月22日

查看全文
  • Yingdong Li,
  • Chongliang Zhang,
  • Yihong Ma,
  • Ying Xue,
  • Yupeng Ji,
  • Yiping Ren,
  • Binduo Xu

摘要

Species distribution models (SDMs) are commonly used in predicting fish species distribution pattern and its change. However, the data collected from fishery-independent surveys may have deviations from the natural situation, leading to bias in ecological studies. This study aimed to explore the impact of sampling designs of fishery-independent surveys on the prediction of SDMs for marine fish species. The sampling designs included simple random sampling (SRS), stratified random sampling (StRS), systematic sampling (SYS) and stratified systematic sampling (StSS). Generalized linear model (GLM), generalized additive model (GAM), random forest (RF), classification and regression tree (CART), and support vector machine (SVM) were developed based on the simulated data from different sampling designs to predict the spatial distribution of target fish species. The predictive abilities of all SDMs enhanced and the differences between the prediction performances of SDMs decreased with increasing sample size. SDMs using the data from SYS always generated more accurate predictions than those from other sampling designs, particularly in the case of GAM and GLM when the sample sizes were small. The predictive ability of GAM was unstable with small sample size, but was getting similar with those of RF and SVM with increasing sample size. The predictive abilities of SDMs were influenced by the data collected from different sampling designs in fishery-independent surveys. Thus, sampling designs from which data are collected should be considered when choosing proper SDMs to investigate the spatial distribution of fish species.