Collaborative Research: DC3--Integrating Probabilistic Forecasting and Optimization to Provide Decision Support to Meteorological Field Campaigns

在大型气象实验中部署复杂的现场数据收集平台(如特殊仪器的研究飞机)的努力取决于具体的天气条件、设备准备情况等，因此涉及后勤和智力挑战，以优化分配有限的资源，如为实现所述目标而资助的飞行小时数和可用的机组人员时间。这些挑战被数值天气预报(NWP)指导中固有的不确定性以及诸如多个优先的实验目标和地理上不同的研究区域等因素放大。这项研究最初将指导在深对流云和化学项目DC3期间计划在多个研究区域部署多架飞机。该研究小组之前的工作将概率预测方法与运筹学中的优化技术相结合，开发了一个自动天气驱动的决策支持系统。这一新的努力将在三个重要方面推动这项工作。首先，将开发一种新的统计后处理算法，将数值预报输出转换为现场实验决策者感兴趣的关键事件发生概率的校准估计。其次，将这些技术扩展到多目标决策模型。第三，将创建一个系统，允许用户(在这种情况下是飞行任务科学家)输入他们自己对选定变量的研究判断，以及从NWP模型收集的更客观的指导。这种独特的定制算法架构将结合专家判断和自动化的最佳属性，以优化快速发展的野外实验环境中的资源分配决策。该项目的智力优势集中在将现有的算法辅助决策支持的理论和实践扩展到受多个资源约束的多目标天气研究问题。这项工作将在金融和定量环境决策分析领域建立的方法之间架起桥梁。除了理想的跨学科学生教育之外，更广泛的影响将包括制定和应用严格的量化框架，以确定在大型气象实地活动期间实现具体科学成果所需的资源，并最终以尽可能最有效的手段提供实现多种研究目标所需的指导。

Efforts to deploy complex field data collection platforms (such as specially-instrumented research aircraft) in large meteorological experiments depend on specific weather conditions, equipment readiness etc. and thus involve both logistical and intellectual challenges to optimally allocate finite resources such as funded flight hours and available crew time to meet stated objectives. These challenges are magnified by uncertainties inherent in numerical weather prediction (NWP) guidance and factors such as multiple prioritized experimental objectives and geographically diverse study regions. This research will initially serve to guide deployment of multiple aircraft across multiple study regions as planned during DC3, the Deep Convective Clouds and Chemistry project. Previous work by this investigative team has integrated probabilistic forecasting methods with optimization techniques adapted from operations research to develop an automated weather-driven decision support system. This new effort will advance this work in three important ways. First, a novel statistical post-processing algorithm will be developed that converts NWP output into calibrated estimates of probability of occurrence for key events of interest to field experiment decision makers. Second, these techniques will be extended to a multi-objective decision model. Third, a system will be created that allows users (in this case, flight mission scientists) to enter their own studied judgments regarding selected variables in addition to more objective guidance gleaned from NWP models. This uniquely tailored algorithmic architecture will combine the best attributes of expert judgment and automation to optimize resource allocation decisions in rapidly-evolving field experiment setting.The intellectual merit of this project centers on extension of an existing body of theory and practice for algorithm-aided decision support to a multi-objective weather research problem subject to multiple resource constraints. This work will bridge methodologies developed in the arenas of finance and quantitative environmental decision analysis. Broader Impacts beyond desirable interdisciplinary student education will include development and application of a rigorous quantitative framework for determining those resources needed to achieve specified scientific outcomes during large meteorological field campaigns, and ultimately serve to provide guidance needed to achieve multiple research objectives by the most efficient means possible.