Collaborative Research: Estimating Ecosystem Model Uncertainties in Pan-Regional Syntheses and Climate Change Impacts on Coastal Domains of the North Pacific Ocean

将开发一系列贝叶斯分层模型(BHM)，以合成北太平洋边界各重点区域的沿海生态系统动态和对气候变化的反应。BHM是一种统一的概率建模方法，它在存在多平台观测的情况下更新关于过程模型和参数的不确定分布知识。对由此产生的“后验”分布的汇总测量提供了对中心趋势和不确定性的现实的定量估计。研究人员将在了解区域海洋学的北太平洋生态系统模型(Nemuro)之后开发我们的过程模型分布。因此，这项研究的一个重要成果将是定量了解和比较北太平洋各地区Nemuro状态变量和参数的相对不确定性。BHM发展计划分三步走，将在三年的时间表上处理泛区域综合、气候变化影响和生态系统管理工具概念。最初的BHM发展将是一个可重新定位的、依赖于时间的一维(垂直)模式，旨在总结不同制度(陆架、坡度、上升流位置、边界流延伸等)的生态系统动力学。在感兴趣的沿海地区内。将充分利用北太平洋沿海地区多学科观测方案和确定性模型实施的数据和见解。除了强调实地观测外，BHM方法还将把确定性模式输出(如区域海洋模式系统或ROMS)作为数据，提供对北太平洋沿海海洋生态系统了解状况的严格和完整的综合。调查人员将专注于美国GLOBEC计划部分地区(即加州海流系统CCS和阿拉斯加湾沿岸CGOA)和西太平洋(WPAC)来自北太平洋海洋科学组织(PICES)的数据和模型。一维BHM还将用于气候规模计算，以记录和比较北太平洋沿海海洋生态系统内部和整个气候变化的影响，并量化这些比较中的不确定性。最终的BHM实施将是三维的，考虑到中尺度海洋动力对沿海生态系统区域的影响，并展示了BHM方法潜在的生态系统管理优势。这项研究的智力优势来自于概率建模方法(即BHM)的新扩展，该方法综合了主要海洋盆地东西边界沿海地区的不同观测和确定性模型模拟。BHM在生物海洋学中的应用代表着一种变革性的研究步骤，并引入了一种新的范式。本文提出的研究结合了确定性模型和概率模型的优点，以获得现代低营养层海洋生态系统模型状态变量和参数的不确定性估计。这项研究的一个更广泛的影响将是对博士后和研究生(统计学和海洋学)进行关于海洋建模方法的这种新的协同作用的培训。随着生态系统管理者和科学家学会利用概率分布中的状态和参数信息，生态系统模型的不确定部分可以被用于更密集的观测和/或更复杂的参数化。

A sequence of Bayesian Hierarchical Models (BHM) will be developed to synthesize coastal ecosystem dynamics and responses to climate change across focus regions bounding the North Pacific Ocean. BHM is a unified probabilistic modeling approach that updates uncertain distributional knowledge about process models and parameters in the presence of multi-platform observations. Summary measures of the resulting "posterior" distributions provide realistic quantitative estimates of central tendencies and uncertainties. The investigators will develop our process model distributions after the North Pacific Ecosystem Model for Understanding Regional Oceanography (NEMURO). So, a significant outcome of the research will be quantitative understanding and comparisons of the relative uncertainties of NEMURO state variables and parameters, region-by-region across the North Pacific. A three-step BHM development plan will address pan-regional syntheses, climate change impacts, and ecosystem management tool concepts, over a three-year schedule. The initial BHM development will be a relocatable, time-dependent, one-dimensional (vertical) model intended to summarize ecosystem dynamics for different regimes (shelf, slope, upwelling loci, boundary current extensions, etc.) within the coastal regions of interest. Data and insights from multi-disciplinary observational programs and deterministic model implementations in coastal regions of the North Pacific will be fully exploited. In addition to emphasizing field observations, the BHM methodology will incorporate deterministic model output (e.g. the Regional Ocean Modeling System or ROMS) as data, providing a rigorous and complete synthesis of the state of understanding for coastal ocean ecosystems of the North Pacific. The investigators will focus on data and models in the Eastern Pacific from parts of the US GLOBEC program (i.e. California Current System, CCS; and Coastal Gulf of Alaska, CGOA) and in the Western Pacific (WPAC) from the North Pacific Marine Science Organization (PICES). The 1D BHM will also be implemented in climate-scale calculations to document and compare climate change impacts within and across North Pacific coastal ocean ecosystems, and to quantify uncertainties in these comparisons. The ultimate BHM implementation will be in three dimensions, accounting for mesoscale ocean dynamical impacts on the coastal ecosystem regions, and demonstrating potential ecosystem management advantages of the BHM approach.The intellectual merit of this research derives from a novel extension of probabilistic modeling methods (i.e. BHM) to synthesize disparate observations and deterministic model simulations from coastal regions on eastern and western boundaries of a major ocean basin. Application of BHM in Biological Oceanography represents a transformative research step and introduces a new paradigm. The research proposed here combines the strengths of deterministic and probabilistic models to obtain uncertainty estimates for state variables and parameters of a modern lower-trophic level ocean ecosystem model. A broader impact of the research will be the training of postdoctoral and graduate students (in statistics and oceanography) in this new synergy of ocean modeling approaches. As ecosystem managers and scientists learn to utilize state and parameter information in probability distributions, uncertain parts of the ecosystem model can be targeted for more intensive observations and/or more sophisticated parameterizations.