Comparison of linear and nonlinear modeling approaches to develop an abundance index based on voyage and market data for a data-limited fishery
Frontiers in Marine Science(2024)
Abstract
IntroductionWorldwide coastal fish resources face severe threats from fisheries overexploitation. However, the evaluation of abundance trends in most coastal fisheries is constrained by limited data. This study took blackmouth croaker (Atrobucca nibe), a stock depleted by coastal trawl fishery in southwestern Taiwan, as an example to showcase the development of a relative abundance index from data-limited fishery (only landing data were available).MethodsThis study employed unique data sourcing from voyage data recorders (VDRs) to estimate fishing effort (in combination with landing data to estimate the catch per unit effort, CPUE) that demonstrated the potential application in global data-limited fisheries and assessed alternative approaches for predictors of fishery-targeting practices to condition effort for producing more accurate metrics of relative abundance. The nominal CPUE was standardized using three statistical models: generalized linear model, generalized additive model (GAM), and vector-autoregressive spatiotemporal models (VASTs) with two treatments of each of the four effects: environmental (sea temperature, salinity, density of mixing layer, seafloor temperature, and chlorophyll), vessel, spatial, and targeting effects. A total of 15 models were designed and compared for these effects, and their explanatory power (EP) was evaluated using cross-validation R2 and other metrics.Results and discussionResults indicated that the targeting effect exerted the most significant influence on standardization and was suggested to be addressed through the principal component analysis (PCA) approach. Both vessel and spatial effects demonstrated considerable influence, whereas the environmental effect exhibited a limited impact, possibly due to the small fishing area in this study. Regarding models’ EP, given the nonlinear nature of the PCA algorithm and environmental data, the study highlighted the superiority of the GAM over linear-based models. However, incorporating nonlinear features in VAST (M15) makes it the most effective model in terms of predictive power in this study. Concerning the stock status, despite variations in relative CPUE trends among major models, a general declining trend since 2015 signals the potential decline of the blackmouth stock and urges fishery managers to consider further design of management measures.
MoreTranslated text
Key words
data-limited fisheries,fishing effort estimation,CPUE standardization,VAST,GAM (generalized additive models),spatiotemporal effects
AI Read Science
Must-Reading Tree
Example
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined