LabEx COTE, Integrative and Theoretical Ecology, University of Bordeaux, Bordeaux, France.
Institute of Mathematics of Bordeaux, CNRS, Talence, France.
J Anim Ecol. 2020 Jul;89(7):1628-1644. doi: 10.1111/1365-2656.13227. Epub 2020 May 18.
Dynamic food web models describe how species abundances change over time as a function of trophic and life-history parameters. They are essential to predicting the response of ecosystems to perturbations. However, they are notoriously difficult to parameterize, so that most models rely heavily either on allometric scaling of parameters or inverse estimation of biomass flows. The allometric approach makes species of comparable body mass have near-identical parameters which can generate extinctions within a trophic level. The biomass flow approach is more precise, but is restricted to steady-states, which is not appropriate for time-varying environments. Adequately parameterizing large food webs of temperate and arctic environments requires dealing both with many species of similar sizes and a strongly seasonal environment. Inspired by the rich empirical knowledge on the vertebrate food web of the Białowieża forest, we parameterize a bipartite food web model comprising 21 predators and 124 prey species. Our model is a non-autonomous coupled ordinary differential equations system that allows for seasonality in life-history and predation parameters. Birth and death rates, seasonal descriptions of diet for each species, food requirements and biomass information are combined into a seasonal parameterization of a dynamic food web model. Food web seasonality is implemented with time-varying intrinsic growth rate and interaction parameters, while predation is modelled with both type I and type II functional responses. All our model variants allow for >80% persistence in spite of massive apparent competition, and a quantitative match to observed (seasonal) biomasses. We also identify trade-offs between maximizing persistence, reproducing observed biomasses, and ensuring model robustness to sampling errors. Although multi-annual cycles are expected with type II functional responses, they are here prevented by a strong predator self-regulation. We discuss these results and possible improvements on the model. We provide a general workflow to parameterize dynamic food web models in seasonal environments, based on a real case study. This may help to better predict how biodiverse food webs respond to changing environments.
动态食物网模型描述了物种丰度如何随时间变化，作为营养和生活史参数的函数。它们对于预测生态系统对干扰的反应至关重要。然而，它们的参数化非常困难，因此大多数模型要么严重依赖参数的比例缩放，要么依赖生物量流的反演估计。比例缩放方法使具有相似体型的物种具有几乎相同的参数，这可能导致营养级内的灭绝。生物量流方法更精确，但仅限于稳态，不适用于时变环境。在温带和北极环境中充分参数化大型食物网需要同时处理许多体型相似的物种和强烈季节性的环境。受白俄罗斯森林脊椎动物食物网丰富的经验知识的启发，我们参数化了一个由 21 种捕食者和 124 种猎物组成的二分食物网模型。我们的模型是一个非自治的耦合常微分方程组系统，允许在生活史和捕食参数中存在季节性。出生率和死亡率、每个物种的饮食季节性描述、食物需求和生物量信息被组合成一个动态食物网模型的季节性参数化。食物网的季节性通过时变的内在增长率和相互作用参数来实现，而捕食则通过 I 型和 II 型功能反应来建模。尽管存在明显的竞争，但所有模型变体都允许>80%的持久性，并且与观察到的（季节性）生物量定量匹配。我们还确定了在最大化持久性、再现观察到的生物量和确保模型对采样误差的稳健性之间的权衡。尽管 II 型功能反应预计会出现多年周期，但这里强烈的捕食者自我调节阻止了这种情况。我们讨论了这些结果和模型的可能改进。我们提供了一种基于实际案例研究的季节性环境下参数化动态食物网模型的一般工作流程。这可能有助于更好地预测生物多样性丰富的食物网如何应对不断变化的环境。
