Collaborative Research: Process-Based Statistical Interpolation Methods for Improved Analysis of WATERS Test-bed Observations and Water Quality Models

切萨皮克湾是一个很好的例子，说明了复杂的水动力学、生物地球化学和来自一个大流域的不同输入是如何导致人类活动对一个关键的环境、经济和社会资源的影响的不确定性的。对这类系统进行更好的科学理解和工程管理需要仔细综合的方法，最大限度地利用所有现有的观测和建模工具，不仅是为了更好地预测未来的影响，而且也是为了更好地了解过去和当前的观测。在此背景下，以及在规划和设计采样方案的背景下，开发现有观测数据的四维（即空间和时间）插值新方法是环境观测站的一个至关重要的需求。这项研究将通过利用切萨皮克湾几十年来建立的丰富资源基础，以及最近建立的切萨皮克湾环境观测站（CBEO）的原型，帮助满足这一需求，该观测站已成为美国国家科学基金会支持的WATERS网络的潜在节点。目前提出的研究目标是开发、测试和应用更好的统计模型来插值水质观测，从而更有效地利用目前可用的水动力和水质模型中捕获的过程理解。更具体地说，这项工作将产生观测数据统计内插的新方法，方法是使用基于过程的“影响度量”（而不是距离）来定义内插（即克里格）所需的相关结构。需要测试的其他影响指标包括旅行时间、水龄和示踪剂比例，所有这些都是通过运行完善和校准的切萨皮克湾水动力和水质模型生成的。基于模型的理解也将用于探索在不同时间间隔和历史环境条件下获得的水质参数之间可能的相互关联。经过开发和全面评估后，新的插值方法将用于探索：(1)历史数据记录上的缺氧发展；(2)确定性模型预测与观测到的水质时空趋势之间持续不一致的原因。新开发的基于过程的插值方法有望克服在流动水体中使用克里格时通常遇到的许多困难。利用统计模型和基于过程的模型对综合观测数据集进行综合分析，将最大限度地利用每种方法的优势，分别包括不确定性估计和预测能力。将这些方法应用于海湾缺氧的紧迫科学问题将显示出它们的优点。总的来说，这项工作将进一步评估和展示环境观测站在改变我们对当前和历史数据的使用和理解方面的力量。更好的分析和理解缺氧的工具的产生将对切萨皮克湾的管理产生深远的影响。目前，使用插值工具来量化海湾水域不符合水质标准的程度，并使用过程模型来预测管理活动（如TMDL开发）的影响。对这两种工具的改进和两者的综合使用将有助于更好地了解和预测水质退化，从而有助于制定最有效的管理办法。该项目的所有人员都与环保署的切萨皮克湾项目合作，因此能够将这些改进的工具带给海湾管理人员。概念性的方法也应该证明在任何地方都同样有价值，只要开发良好的基于过程的仿真模型是可用的。研究结果将通过国内和国际科学会议、同行评议期刊上的出版物以及通过WATERS网络上的CBEO节点（由圣地亚哥超级计算机中心维护）提供新方法来传播。这项研究是跨学科的，跨两所大学合作，包括研究生和本科生。对K-12教育的影响将通过与马里兰大学正在进行的教育项目的合作来实现，该项目使用互动式教育模块来教授中学生有关地表水“死区”（缺氧）的问题。

0854329 / 0853765 Ball / DiToro The Chesapeake Bay is a prime example of how complex hydrodynamics, biogeochemistry, and varying inputs from a large watershed can lead to uncertainty about the impacts of human activities on a crucial environmental, economic, and social resource. Better scientific understanding and engineering management of such systems requires carefully integrated approaches that make maximum use of all available observations and modeling tools, not only for better predictions of future impacts, but also for better understanding of past and current observations. In this context, and also in the context of planning and designing sampling programs, the development of new methods for 4D (i.e., space and time) interpolation of existing observational data is a critically important need for environmental observatories. This research will help meet this need by taking advantage of a rich resource base that has been established over many decades of Chesapeake Bay research and most recently through a prototypical Chesapeake Bay Environmental Observatory (CBEO) that has been established as a potential node for the NSF-supported WATERS Network. Objectives of the currently proposed research are to develop, test, and apply better statistical models for the interpolation of water quality observations that make more effective use of the process understanding captured in currently available hydrodynamic and water quality models. More specifically, the work will generate new approaches for statistical interpolation of observations by using process-based "metrics of influence" (as opposed to distance) for defining the correlation structure that informs interpolation (i.e., kriging). The alternative metrics of influence to be tested include travel time, water age, and tracer proportion, all generated through runs of well-established and calibrated Chesapeake Bay hydrodynamic and water quality models. Model-based understanding will also be used to explore possible cross correlations among water quality parameters, as obtained over different time intervals and historical environmental conditions. After their development and thorough evaluation, the new interpolation methods will be applied toward exploring: (1) hypoxia development over a historical data record, and (2) causes for continuing inconsistencies between deterministic model predictions and observed temporal and spatial trends in water quality.The newly developed process-based interpolation methods are expected to overcome many of the difficulties commonly encountered in using kriging in flowing water bodies. The integrated analysis of comprehensive observational data sets with both statistical and process-based models will take maximum advantage of the strengths of each approach, which include uncertainty estimation and predictive ability, respectively. The application of these methods to pressing science questions on Bay hypoxia will demonstrate their merit. Overall, the work will further evaluate and demonstrate the power of environmental observatories to transform our use and understanding of current and historical data.The generation of better tools for analyzing and understanding hypoxia will have far reaching impacts on the management of the Chesapeake Bay. Currently, interpolation tools are used to quantify the extent of Bay waters not meeting water quality criteria, and process models are used to predict impacts of management activities, such as TMDL development. Improvements to both types of tools and integrated use of the two will allow better understanding and prediction of water quality degradation and thus help target the most effective management options. All of the personnel on this project have worked collaboratively with EPA's Chesapeake Bay Program and are thus able to bring these improved tools to Bay managers. The conceptual approach should also prove to be equally valuable at any location where well-developed process-based simulation models are available. The findings will be disseminated through national and international scientific meetings, through publications in peer reviewed journals, and by making the new methods available through the CBEO node on the WATERS network (as maintained through the San Diego Supercomputer Center). This research is interdisciplinary and collaborative across two universities, including both graduate and undergraduate students. Impact on K-12 education will be achieved through collaborations that assist an on-going educational program at the University of Maryland which uses interactive educational modules to teach middle-school students about the issues surrounding "dead zones" (hypoxia) in surface waters.