Incorporating egg-transporting pathways into conservation plans of spawning areas: An example of small yellow croaker (Larimichthys polyactis) in the East China Sea zone

Abstract

Backward-in-time Lagrangian model can identify potential spawning areas by reconstructing egg drift trajectories, contributing to accurately designing potential priority conservation plans for spawning areas. In this study, we apply this approach to investigate the small yellow croaker (Larimichthys polyactis) with commercial value in China. A two-step spatial random forest (RF) model is used to predict the occurrence probability and abundance of their eggs and describe the optimal ecological range of environmental factors. A priority protection index (BPPI) of the spawning areas is established by combining the sites with the optimal occurrence and abundance and integrating backward tracking pathways. The result indicates that the model with 1-2 day time lags of environmental variables shows the optimum explanatory power. Temperature and salinity are the most important factors affecting oogenesis and show a regime shift in the response curve. They reflect the physiological regulation of parental sexual maturation by the environment. In addition, egg abundance correlates more strongly with chlorophyll-a (Chl a) concentration and depth, suggesting that parents prefer environments with shallow water and high prey density for spawning activities. The egg retrieval shows that the potential spawning sources are distributed near the southeastern part of the oogenesis site, with a maximum egg dispersal distance of no more than 30 km. This finding confirms that the coastal regions of Jiangsu Province are an important spawning ground for the small yellow croaker, making a significant contribution to the productivity and resilience of the fish.

Evidence from interpretable machine learning to inform spatial management of Palau's tuna fisheries

“…There are several spatial regression type ML methods to explicitly account for spatial covariance (see Georganos & Kalogirou, 2022), including spatial RF with eigenvector spatial filtering (Liu, Jin, et al, 2022;Liu, Kounadi, et al, 2022;Reisinger et al, 2022)-an approach also used in a more conventional GLM(M)-based model of coral bleaching on the Great Barrier Reef (Hughes et al, 2021). We used the spatialRF R package (Benito, 2021) with eigenvector spatial filtering (Dray et al, 2006) and a matrix of the distance (in kilometers) between all sets to further explore species-specific spatial effects using the top 10 SHAP summary predictors determined previously using the species-specific ML models.…”