Antarctic krill (Euphausia superba) is a keystone species in the Southern Ocean, and plays critical roles in the ecosystem through its spatial dynamics. To balance conservation and sustainable harvesting of this species, a revised management framework has been proposed, requiring precise krill distribution data. However, conventional species distribution models (SDMs) often fail to capture spatiotemporal autocorrelation, compromising their statistical integrity and predictive performance. This study focuses on providing the latest distribution information of Antarctic krill and examining how spatiotemporal structure affects the predictability of SDMs. We compared the performance of three SDMs fitted using the sdmTMB framework: a baseline model without random effects (non-spatial), a model incorporating spatial random effects (spatial), and a spatiotemporal model with spatiotemporal random effects (spatiotemporal), using 2013–2020 summer acoustic surveys near the South Shetland Islands. Cross-validated results revealed substantial performance disparities: the non-spatial model showed minimal explanatory power (R²=0.161), the spatial model demonstrated slight improvements (R²=0.165), while the spatiotemporal model achieved superior performance (R²=0.348). Salinity and year were identified as important predictors across all model formulations, while current velocity was a significant predictor in both the non-spatial and spatial models, but not in the spatiotemporal model. The estimated effects of these covariates changed when spatial and spatiotemporal random fields were incorporated. Model projections suggest a slight southward shift in krill distribution centroids. This study advocates for the integration of spatiotemporal structure into SDMs to support ecosystem-based management, offering insights for implementing spatially adaptive fisheries strategies in the rapidly changing Antarctic environment.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号