Data from: Fine-scale population dynamics in a marine fish species inferred from dynamic state-space models

Identifying the spatial scale of population structuring is critical for the conservation of natural populations and for drawing accurate ecological inferences. However, population studies often use spatially aggregated data to draw inferences about population trends and drivers, potentially masking ecologically relevant population sub-structure and dynamics.
The goals of this study were to investigate how population dynamics models with and without spatial structure affect inferences on population trends and the identification of intrinsic drivers of population dynamics (e.g. density dependence).
Specifically, we developed dynamic, age-structured, state-space models to test different hypotheses regarding the spatial structure of a population complex of coastal Atlantic cod (Gadus morhua). Data were from a 93-year survey of juvenile (age 0 and 1) cod sampled along >200 km of the Norwegian Skagerrak coast. We compared two models: one which assumes all sampled cod belong to one larger population, and a second which assumes that each fjord contains a unique population with locally determined dynamics. Using the best supported model, we then reconstructed the historical spatial and temporal dynamics of Skagerrak coastal cod.
Cross-validation showed that the spatially structured model with local dynamics had better predictive ability. Furthermore, posterior predictive checks showed that a model which assumes one homogeneous population failed to capture the spatial correlation pattern present in the survey data. The spatially structured model indicated that population trends differed markedly among fjords, as did estimates of population parameters including density-dependent survival. Recent biomass was estimated to be at a near-record low all along the coast, but the finer scale model indicated that the decline occurred at different times in different regions. Warm temperatures were associated with poor recruitment, but local changes in habitat and fishing pressure may have played a role in driving local dynamics.
More generally, we demonstrated how state-space models can be used to test evidence for population spatial structure based on survey time-series data. Our study shows the importance of considering spatially structured dynamics, as the inferences from such an approach can lead to a different ecological understanding of the drivers of population declines, and fundamentally different management actions to restore populations.

确定种群结构的空间尺度，对于自然种群保护以及精准开展生态学推断均具有关键意义。然而，种群研究常借助空间聚合数据推导种群趋势与驱动因子，这可能掩盖具有生态学相关性的种群亚结构与动态过程。
本研究旨在探究，有无空间结构的种群动态模型如何影响种群趋势推断，以及种群动态内在驱动因子（如密度依赖）的识别。
具体而言，我们构建了动态、年龄结构的状态空间模型（state-space models），以检验大西洋鳕（Gadus morhua）种群复合体空间结构的各类假说。研究数据源自一项为期93年的调查：沿挪威斯卡格拉克海岸超过200公里的海域，采样了0龄和1龄的幼龄大西洋鳕。我们对比了两类模型：其一假设所有采样鳕鱼隶属于同一大种群；其二则认为每个峡湾均拥有独立种群，其种群动态由本地因素决定。借助最优拟合模型，我们重构了斯卡格拉克海岸鳕鱼的历史空间与时间动态格局。
交叉验证结果显示，具备本地动态的空间结构化模型拥有更优的预测能力。此外，后验预测检验表明，假设种群为单一均质群体的模型，无法捕捉调查数据中存在的空间相关模式。空间结构化模型显示，不同峡湾的种群趋势差异显著，包括密度依赖存活率在内的各类种群参数估计结果亦呈现类似差异。沿整个海岸区域，近年生物量估计值均接近历史低位，但精细化尺度的模型显示，种群衰退在不同区域发生的时间各不相同。水温偏高与种群补充量不足存在显著关联，而栖息地的本地变化与捕捞压力或对本地种群动态产生了驱动作用。
从更广泛的视角来看，本研究展示了如何基于调查时间序列数据，利用状态空间模型检验种群空间结构的相关证据。本研究凸显了考虑空间结构化动态的重要性：采用此类方法得到的推断，可催生对种群衰退驱动因子的全新生态学认知，以及从根本上迥异的种群恢复管理措施。