New Techniques for Simulating and Analyzing Biogeochemical Tracers in a Seasonally Varying Global Ocean Models

摘要提案编号： 虽然反演方法和自动优化方法在全球海洋地球化学模式中的应用已经相当成功，但稳态方法未能捕捉到海洋基本的随时间变化的季节性。 将自动优化扩展到季节性变化模型的主要限制是巨大的计算成本。 在这项研究中，来自加州大学欧文分校的两名年轻的PI将采用最先进的牛顿-克雷洛夫方法，以大大减少旋转全球地球化学模型所需的计算时间。由此产生的快速求解器将使得能够更有效地探索参数空间，并且使得使用自动优化方法来调整不确定的模型参数变得可行。 他们将把求解器应用于复杂程度越来越高的全球海洋地球化学模型的层次结构。在最简单的模型中，生物吸收将通过将表面营养物恢复到其观测值来诊断参数化。在中等复杂性的模型中，生物吸收将具有明确的光、温度、磷酸盐和铁限制治疗的预后表示。最复杂的模型将是一个完整的生态系统模型，其中包括海洋生物地球化学循环所涉主要功能群的代表。这项研究的预期成果包括快速求解器，用于获得海洋生态地球化学模型的季节性变化周期稳定状态，可供社区使用，以大大降低与旋转不同气候状态或改变参数值的全球生态地球化学模型相关的计算成本。 此外，他们的努力将产生优化的地球化学模型参数，与气候和季节性变化的化学示踪剂数据相一致。在更广泛的影响方面，一名研究生和一名博士后研究员将获得海洋地球化学和先进的数值和优化方法方面的培训。本科生也将接触到科学研究作为NSF赞助的REU计划在UCI的一部分。 这两个PI将继续参与外展计划，为来自三个少数民族聚居、需求量大的南加州学区（圣安娜、科斯塔梅萨和康普顿）的小学教师提供地球科学培训。与NCAR CCSM海洋网格兼容的新的在线地球化学模型，快速Newton-Krylov求解器和参数优化程序将向社区公开。

ABSTRACTProposal No.: OCE-0623647Although application of inverse methods and automatic optimization methods to global ocean biogeochemistry models has been quite successful, the steady-state approach fails to capture the essential time-varying seasonality of the ocean. The main limitation of extending automatic optimization to seasonally varying models is the formidable computational costs. In this research, two young PIs from the University of California - Irvine will adapt a state-of-the-art Newton-Krylov method to greatly reduce the computational time needed to spin-up global biogeochemistry models. The resulting fast solver will make it possible to explore parameter space more efficiently and make it feasible to tune uncertain model parameters using automatic optimization methods. They will apply the solver to a hierarchy of global ocean-biogeochemistry models with an increasing level of complexity. In the simplest model, biological uptake will be parameterized diagnostically by restoring surface nutrients to their observed value. In models of intermediate complexity, biological uptake will have a prognostic representation with an explicit treatment of light, temperature, phosphate and iron limitations. The most complex model will be a full ecosystem model that includes representations of the major functional groups involved in ocean biogeochemical cycling. The anticipated outcomes of this research include the fast solver for obtaining the seasonally varying periodic steady states of ocean biogeochemistry models that can be used by the community to greatly reduce the computational costs associated with spinning up global biogeochemistry models for different climate states or with altered parameter values. In addition, their efforts will yield optimized biogeochemistry model parameters that are consistent with climatological and seasonally varying chemical tracer data. In terms of broader impacts, a graduate student and a postdoctoral researcher will obtain training in ocean biogeochemistry and advanced numerical and optimization methods. Undergraduate students will also be exposed to scientific research as part of the NSF-sponsored REU program at UCI. Both PIs will continue participating in outreach programs to provide training in the Earth sciences to elementary teachers from three heavily minority, high-need Southern California school districts (Santa Ana, Costa Mesa, and Compton). The new online biogeochemistry models compatible with the NCAR CCSM ocean-grid, the fast Newton-Krylov solver and parameter optimization routines will be made publicly available to the community.