Oil spills can disrupt marine and coastal ecosystem services leading to reduced employment opportunities and income. Ecosystem models can be used to estimate the effects of oil pollution; however, uncertainty in model predictions may influence damage assessment. We performed an uncertainty analysis for the Atlantis ecosystem model of the Gulf of Mexico (Atlantis-GOM), under a scenario simulating the effects of the Deepwater Horizon oil spill. Atlantis-GOM simulates major biophysical processes, including the effects of oil hydrocarbons on fish growth and mortality. We used all available fish stomach content data to inform parameter distribution for the Atlantis-GOM availability matrix, which represents predator total consumption potential and diet preference. We sampled the fish diet composition distribution and analyzed the variability of functional group biomass and catch predicted by Atlantis-GOM simulations to changes in the availability matrix. Resulting biomass and catch were then used to fit statistical emulators of the ecosystem model and predict biomass and catch given the complete diet parameter space. We used simulated and emulated data to assess changes in recovery time to oil spill effects. Uncertainty in diet composition had large effects on model outputs and may, therefore, influence damage assessment of oil exposure on economically important species.

漏油事件可能会破坏海洋和沿海生态系统服务，从而导致就业机会和收入减少。生态系统模型可用于估计石油污染的影响；但是，模型预测中的不确定性可能会影响损坏评估。在模拟“深水地平线”影响的情况下，我们对墨西哥湾的亚特兰蒂斯生态系统模型（Atlantis-GOM）进行了不确定性分析漏油事件。Atlantis-GOM模拟主要的生物物理过程，包括石油烃对鱼类生长和死亡率的影响。我们使用了所有可用的鱼胃含量数据来告知Atlantis-GOM可用性矩阵的参数分布，该矩阵代表了捕食者的总消费潜力和饮食偏好。我们对鱼的日粮组成分布进行了采样，并分析了Atlantis-GOM模拟预测的功能组生物量和捕获量随可用性矩阵的变化。然后将得到的生物量和捕获量用于拟合生态系统模型的统计仿真器，并在给定完整的饮食参数空间的情况下预测生物量和捕获量。我们使用模拟和仿真数据来评估恢复时间对溢油影响的变化。