LTREB Renewal: Ecological Dynamics in an Experimentally-Tractable Natural Ecosystem

生态学的一个核心目标是记录环境是否以及如何变化，确定这些变化的原因，并预测这些变化对生态系统的影响。这是一项挑战，因为生态系统的生物体和非生物部分之间有着复杂的联系网络。 数学模型是跟踪这些生态相互作用并预测它们将如何应对环境变化的重要工具。然而，模型需要与来自自然的数据联系起来。开发环境变化预测模型的两个主要挑战是：（1）收集关于一组完整的物种和环境因素之间的相互作用如何随时间变化的足够数据，以及（2）通过实验严格测试模型预测。这项研究将结合联合收割机的四分之一个世纪的一系列数据，100多个物种和相关的环境变量在岩石海岸线的塔图什岛在华盛顿州，与长期的实地实验，模仿灭绝的一个关键物种，加州贻贝。长期数据将应用于几种不同的建模方法，随后将通过长期现场实验对这些模型的预测进行测试。该研究将确定最有前途的生态预测建模方法，并将其提供给对物种灭绝和全球变化等环境影响后果感兴趣的生态系统管理者和政策制定者。综合数据系列也将提供给其他科学家，作为进一步研究的平台。 该项目还将让本科生参与实地研究，数据管理，数学建模，以及与公众，管理人员和政策制定者的沟通。此外，由于理解物种相互作用网络的挑战与处理复杂网络的其他科学学科共享，因此项目结果将在其他学科中具有普遍价值。研究人员将以两种方式对海岸线上的植物和动物复制永久性地块进行年度调查：1）记录5年期间2 600个固定点下的物种身份，并生成物种之间的年度转移概率，2）在永久性60 x 60厘米普查地块中生成丰度估计。 15个实验区将通过选择性地移除出现的加州贻贝个体来维持，而不干扰所有其他物种。 将使用潜水数据记录器和陆基气象站每30分钟收集一次环境数据。将监测水的化学性质，包括关键的营养物质。 这些数据将以几种方式进行分析，包括1）参数化基于过渡的模型（马尔可夫链模型，空间显式细胞自动机）与环境的依赖性，2）参数化多物种种群动态模型，从情节计数，3）应用多空间交叉收敛映射和测试是否准确地检测关键物种已知有很强的因果关系的影响，从独立的实验，4）应用神经网络模型并测试它们对物种灭绝后果的预测，5）测试物种丰度随时间的变化与其对生态系统的重要性之间是否存在关系，如独立实验所评估的那样。由丰富的长期数据集支持的社区建模项目具有很强的潜力，可以促进我们对社区动态及其对环境变化的响应机制的理解。

A central goal for ecology is to document if and how the environment is changing, to determine the causes of these changes, and to predict what the consequences of these changes will be to ecological systems. This is a challenge because of the complex network of connections among the living organisms and the non-living parts of ecosystems. Mathematical models are essential tools to keep track of these ecological interactions and to predict how they will respond to environmental changes. However, models need to be linked to data from nature. Two major challenges in developing predictive models of environmental change are 1) collecting sufficient data on how interactions among a complete set of species and environmental factors change over time, and (2) rigorously testing model predictions with experiments. This study will combine a quarter-century long series of data on 100+ species and relevant environmental variables in the rocky shoreline of Tatoosh Island in Washington state, with a long-term field experiment that mimics the extinction of a key species, the California mussel. The long term data will be applied to several different modeling approaches and predictions from these models will subsequently be tested with the long-term field experiment. The research will identify the most promising modeling approaches for making ecological prediction, and make them available to ecosystem managers and policy makers interested in the consequences of environmental impacts such as species extinction and global change. The comprehensive data series also will be made available to other scientists to be used as a platform for additional studies. This project will also engage undergraduate students in field research, data management, mathematical modeling, and in communicating with the public, managers, and policy makers. Furthermore, because the challenge of understanding networks of species interactions is shared with other scientific disciplines that deal with complex networks, project results will be of general value in other disciplines. The researcher will conduct annual surveys of replicated permanent plots for plants and animals on the shoreline in two ways: 1) by documenting the species identities under 2,600 fixed points over a 5-year period and generating annual transition probabilities among species, and 2) by generating abundance estimates in permanent 60 x 60 cm census plots. Fifteen experimental plots will be maintained by selectively removing individuals of Mytilus californianus when they appear, leaving all other species undisturbed. Environmental data will be collected every 30 minutes using a submersible data logger and a land-based weather station. Water chemistry, including critical nutrients, will be monitored. These data will be analyzed in several ways, including 1) parameterizing transition-based models (Markov chain models, spatially-explicit cellular automata) with environmental dependencies, 2) parameterizing multi-species population dynamic models from plot counts, 3) applying multi-spatial cross-convergent mapping and testing whether it accurately detects key species known to have strong causal effects from independent experiments, 4) applying neural network models and testing their predictions about the consequences of species extinction, and 5) testing whether there is a relationship between the variability of a species' abundance through time and its importance to the ecosystem as assessed by independent experiments. The community modeling projects enabled by the rich long term data sets have a strong potential to advance our understanding of mechanisms underlying community dynamics and their response to environmental change.