Predictive power of food web models based on body size decreases with trophic complexity

Food web models parameterized using body size show promise to predict trophic Interaction Strengths (IS) and abundance dynamics. However, this remains to be rigorously tested in food webs beyond simple trophic modules, where indirect and intraguild interactions could be important and driven by traits other than body size. We systematically varied predator body size, guild composition and richness in microcosm insect webs and compared experimental outcomes with predictions of IS from models with allometrically scaled parameters. Body size was a strong predictor of IS in simple modules (r2=0.92), but with increasing complexity the predictive power decreased, with model IS being consistently overestimated. We quantify the strength of observed trophic interaction modifications, partition this into density-mediated vs. behaviour-mediated indirect effects and show that model shortcomings in predicting IS is related to the size of behaviour-mediated effects. Our findings encourage development ...采用体型参数化的食物网模型在预测营养交互强度（Trophic Interaction Strength, IS）与种群丰度动态方面展现出应用潜力。然而，此类模型尚未在简单营养模块之外的食物网中得到严格检验——在这类食物网中，间接相互作用与集团内相互作用可能发挥重要作用，且其驱动因素并非仅为体型这一性状。我们在微宇宙（microcosm）昆虫食物网中系统性调控捕食者体型、功能群组成与物种丰富度，并将实验结果与采用异速缩放参数的模型所预测的IS值进行对比。在简单营养模块中，体型对IS具有极强的预测能力（r²=0.92），但随着系统复杂度提升，模型的预测能力逐渐下降，且模型对IS的预测值始终偏高。我们量化了观测到的营养交互修饰强度，将其划分为密度介导间接效应与行为介导间接效应两类，并证实模型在预测IS时的缺陷与行为介导效应的强度存在相关性。本研究结果为……的发展提供了重要启示。