Online Uncertainty Quantification for Novel Atmospheric Measurements

在数值天气模式中，当前对大气状态的观测作为初始条件被同化到模式中。为了使观测尽可能有用，需要确定测量中潜在不确定度的范围。对于全新的测量类型，例如来自卫星的遥感，很难独立地验证观测结果并确定不确定性特征。该项目中的这项工作将评估各种数值技术，以确定这些新观测的不确定性。该项目的主要影响将是天气预报，这些信息有可能被用于其他各种领域。一名研究生将参与该项目，确保培训下一代数据同化专家。本项目将讨论大气观测的不确定度量化(UQ)问题。更具体地说，该项目将以“新的”测量为目标，在这种测量中，新的观测无法与独立的观测进行验证。研究小组将对当前的方法进行审查，并开发新的方法来推进在线观察UQ的实践。该项目的第一步将是利用双尺度洛伦兹(L96)模型开展实验，并随后评估各种现有的不确定度量化战略。基于核密度估计(KDE)的新理论发展也将得到测试和成熟。然后，研究团队将把分析扩展到一般流通模型(GCM)用例。该项目的结果将是：1)对建议用于地球科学的领先观测UQ技术所做的假设进行详尽的评估，以及2)用于非高斯误差估计的新UQ技术。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

In numerical weather modeling, current observations of the state of the atmosphere are assimilated as an initial condition into the model. For the observations to be as useful as possible, the range of potential uncertainty in the measurements needs to be determined. For completely new types of measurements, such as remote sensing from satellites, it is difficult to independently validate the observations and determine the uncertainty characteristics. This work in this project will assess a variety of numerical techniques to determine the uncertainty of these new observations. The main impact of the project will be on weather forecasting, with the potential for the information to be used in various other fields. A graduate student will be involved in the project, ensuring the training of the next generation of data assimilation experts. This project will address the topic of uncertainty quantification (UQ) for atmospheric observations. More specifically, the project will target “novel” measurements, where new observations are unable to be validated against independent observations. The research team will perform an examination of current methodology and develop new methods for advancing the practice of online observation UQ. The first step of the project will be the development of experiments using the two-scale Lorenz (L96) model and the subsequent evaluation of various existing strategies for uncertainty quantification. A new theoretical development based on Kernel density estimates (KDE) will also be tested and matured. The research team will then expand the analysis to general circulation model (GCM) use cases. The result of the project will be: 1) An exhaustive evaluation of assumptions made by leading observation UQ techniques suggested for geoscience, and 2) A new UQ technique for non-Gaussian error estimation.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.