Predicting marine species distribution and identifying priority areas for conservation – comparing and coupling food web and Bayesian hierarchical modelling approaches

海洋生态系统受到许多人为压力因素的影响，资源的过度开发是主要威胁之一。减轻人为压力和保护海洋生物多样性和生态系统功能的一个重要工具是实施海洋保护区。了解生物多样性、目标物种和基本栖息地的地理分布，是成功确定和优先确定合适保护区的关键。随着不同的保护规划算法的发展，各种各样的统计建模方法用于理解和预测物种分布。然而，最终的预测总是伴随着不同程度的不确定性；尽管关于不确定性水平的知识有助于在面临高风险时做出决策，并允许调整管理计划，但在保护规划工具中很少考虑不确定性。此外，大多数相关的物种分布模型和保护区选择算法只考虑了相关的环境因素，而生态过程和人类活动也可以驱动物种分布。因此，为了更好地预测物种分布和加强保护区的确定，关键的一步包括发展明确说明生态过程和人类活动的方法，以及量化与所产生的保护计划有关的不确定性。贝叶斯层次物种分布（B-HSD）模型是一种新颖的技术，它允许在建模过程中结合空间随机效应项、变量的空间相关性和参数的不确定性，从而更加真实和准确地估计不确定性。生态过程和人类活动可以在诸如生态空间生境容量（E-HFC）模型等机制模型中明确考虑。E-HFC可以结合食物网动态和渔业影响，从累积的物理、海洋学和环境影响中驱动物种的觅食能力。此外，生态空间中的“目标函数”允许识别有效的保护区。然而，与大多数生态系统模型和保护规划工具一样，E-HFC模型没有考虑与物种分布相关的不确定性。因此，本项目的目的是探讨B-HSD模型与E-HFC模型在物种分布预测和优先保护区域确定方面的互补性和适用性。为此，我使用了热带海湾系统Chwaka湾（坦桑尼亚）及其最近开发的Ecopath模型作为案例研究。因此，我试图进一步发展E-HFC模型的适用性，并为物种分布预测的混合方法的发展做出贡献。

Marine ecosystems are impacted by numerous anthropogenic stressors, with overexploitation of resources being among the central threats. A prominent tool to mitigate anthropoogenic pressures and to protect marine biodiversity and ecosystem functioning is the implementation of marine protected areas. For a successful identification and prioritization of suitable conservation areas it is crucial to understand the geographic distribution of biodiversity, target species and essential habitats. A wide variety of statistical modelling approaches for understanding and predicting species distribution have been developed along with different conservation planning algorithms. However, the resulting predictions are always associated with varying degrees of uncertainty; and although the knowledge on the level of uncertainty facilitates decision-making in the face of high risks involved and allows for the adaption of management plans, uncertainty is rarely accounted for in conservation planning tools. Furthermore, most correlative species distribution models and reserve selection algorithms only account for associated environmental factors, while ecological processes and human activities can also drive species distribution. A crucial step towards better predicting the distribution of species and towards an enhanced identification of conservation areas, therefore, encompasses the development of approaches that explicitly account for ecological processes and human activities and that quantify uncertainties associated with resulting conservation plans. Bayesian hierarchical species distribution (B-HSD) modelling is a novel technique that allows the incorporation of spatial random-effect terms, spatial correlation of the variables and the uncertainty of the parameters in the modelling process, resulting in a more realistic and accurate estimation of uncertainty. Ecological processes and human activities can be explicitly considered in mechanistic modelling such as the Ecospace habitat capacity (E-HFC) model. E-HFC allows to spatially drive foraging capacity of species from cumulative physical, oceanographic and environmental effects in conjunction with food web dynamics and fisheries impacts. Furthermore, the ‘objective function’ in Ecospace allows for the identification of effective conservation areas. However, as most ecosystem models and conservation planning tools the E-HFC model does not incorporate the uncertainty related to species distribution. The objective of the project is, thus, to explore the complementarity and applicability of B-HSD models and the E-HFC model for the prediction of species distribution and the identification of priority areas for conservation. For this purpose I use the tropical bay system Chwaka Bay (Tanzania) and its recently developed Ecopath model as case study. With this I attempt to further develop the E-HFC model applicability and to contribute to the development of hybrid approaches for the prediction of species distribution.