Collaborative Research: Adaptive Sampling and Ensemble-Based Data Assimilation

这个项目代表了毕晓普博士和麻省理工学院的伊曼纽尔博士之间的合作，伊曼纽尔博士是由ATM-9815114资助的。该项目将扩大首席调查员在大气自适应采样方面开展的工作。到目前为止，对大气层的常规现场观测要么是在固定时间的固定地点(例如，通过仪表式气象气球)进行的，要么是从船舶和商用飞机等有机会的平台进行的。自适应采样使用有关大气当前和(估计的)未来状态的信息，例如对分析不确定性和预测灵敏度分布的估计，以确定观测将导致预测最大改进的位置。然后，可编程的观测平台，如有人或无人驾驶的飞机，被引导到这些特殊位置进行观测。主要研究人员对数据所做的工作已经开发、改进和测试了几种适应性抽样策略。他们将继续这些研究，并延伸到飓风路径预测问题，这是我们今天面临的最重要的预测问题之一。这项研究的一个关键组成部分将是探讨数据同化及其在天气分析中以最佳方式利用这些新观测的作用。特别是，毕晓普博士和伊曼纽尔博士将研究同化过程中集合的使用，并将这种方法与当前的技术进行比较。在这项工作中，主要调查人员将采用模型的层次结构。

This project represents a collaboration between Dr. Bishop and Dr. Emanuel of MIT who is funded under ATM-9815114. The project will expand work undertaken by the principal investigators on adaptive sampling of the atmosphere. Heretofore, routine in-situ observations of the atmosphere have been made either from fixed locations at fixed times (e.g., by instrumented weather balloons) or from platforms of opportunity such as ships and commercial aircraft. Adaptive sampling uses information about the current and (estimated) future state of the atmosphere, such as estimates of the distribution of analysis uncertainty and forecast sensitivity, to determine locations where observations would lead to the greatest improvements in forecasts. Programmable observation platforms, such as manned or unmanned aircraft, are then directed to these special locations to make the observations. The principal investigators' work to data has developed, refined, and tested several adaptive sampling strategies. They will continue with these studies and extending to the problem of hurricane track prediction, one of the most important forecasting problems facing us today. A key component of this study will be the exploration of data assimilation and its role in optimally employing these new observations in the weather analyses. In particular, Drs. Bishop and Emanuel will be looking at the use of ensembles for the assimilation process and comparing this approach with current techniques. In the course of this work, the principal investigators will employ a hierarchy of models.