Collaborative Research: Framework: Improving the Understanding and Representation of Atmospheric Gravity Waves using High-Resolution Observations and Machine Learning

地球物理重力波是地球大气和海洋中普遍存在的一种现象，通过重力与层状或层状流体的相互作用而成为可能。当风在山上流动时，雷暴和其他强对流系统，以及冬季风暴加剧时，它们在大气中感到兴奋。重力波在大气动量和能量平衡中发挥着重要作用，它通过影响气候系统关键特征的变化，如急流和平流层极地涡旋，直接影响地面天气和气候。这些波对天气和气候预测提出了挑战：100米到100公里尺度的波既不能用传统观测系统进行系统测量，也不能在全球大气模型中得到适当的解析。因此，必须基于可直接模拟的已分解流动来表示或近似这些波。目前重力波的表示受到计算必要性和观测稀缺性的严重限制，导致短期天气和长期气候预测的不准确或不确定。该项目的目标是利用来自Loon高空气球的前所未有的观测，并使用专门的高分辨率计算机模拟和机器学习技术来开发准确的、数据信息丰富的重力波表示法。该项目的成果预计将产生更好的天气和气候模型，从而改善对极端天气的短期预测和对气候变化的长期预测，这具有重大的社会效益。此外，该项目将支持培训3名博士后、4名博士后和10名本科生暑期研究人员，从事大气动力学、气候建模和数据科学的交叉工作，从而为下一代科学家的跨学科职业生涯做好准备。该项目将带来两个关键进展。首先，它将开辟一个新的数据源，以限制重力波动量在大气中的传输。Loon LLC自2013年以来一直在发射超级压力气球，以提供全球互联网报道。从数千次飞行中可以获得非常高分辨率的位置、温度和压力观测(每隔60秒进行一次)。这提供了前所未有的高分辨率观测来源，以限制重力波源和传播。该项目将处理气球测量，并与新的高分辨率模拟相结合，建立一个公开可用的数据集，为观测受限的重力波源、传播和破裂评估开辟潜在的转换资源。第二个转变将是使用机器学习技术来开发计算上可行的重力波动量沉积表示法。目前基于物理的表示只考虑了波的垂直传播(即，它们是一维的)，而忽略了它们的水平传播。利用基于LOON测量和高分辨率模式的数据，将开发一维和三维数据驱动表示法，以更准确和有效地表示天气和气候模式中重力波的影响。这些新颖的表示将在理想化的大气模型中实现，以研究重力波在温带急流变化、准两年振荡(热带平流层风的缓慢变化)和冬季平流层极地涡旋中的作用，使人们能够更好地了解它们对大气温室气体浓度增加的反应。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。

Geophysical gravity waves are a ubiquitous phenomenon in Earth’s atmosphere and ocean, made possible by the interaction of gravity with a stratified, or layered fluid. They are excited in the atmosphere when winds flow over mountains, by thunderstorms and other strong convective systems, and when winter storms intensify. Gravity waves play an important role in the momentum and energy balance of the atmosphere, with direct impacts on surface weather and climate through their effect on the variability of key features of the climate system such as the jet streams and stratospheric polar vortices. These waves present a challenge to weather and climate prediction: waves on scales of 100 meters to 100 kilometers can neither be systematically measured with conventional observational systems, nor properly resolved in global atmospheric models. As a result, these waves must be represented, or approximated, based on the resolved flow that can be directly simulated. Current representations of gravity waves are severely limited by computational necessity and the scarcity of observations, leading to inaccuracies or uncertainties in short term weather and long term climate predictions. The objective of this project is to leverage unprecedented observations from Loon high altitude balloons and use specialized high resolution computer simulations and machine learning techniques to develop accurate, data-informed representation of gravity waves. The outcomes of this project are expected to result in better weather and climate models, thus improving short term forecasts of weather extremes and long term climate change projections, which have substantial societal benefits. Furthermore, the project will support the training of 3 Ph.D. students, 4 postdocs, and 10 undergraduate summer researchers to work at the intersection of atmospheric dynamics, climate modeling, and data science, thus preparing the next generation of scientists for interdisciplinary careers.The project will deliver two key advances. First, it will open up a new data source to constrain gravity wave momentum transport in the atmosphere. Loon LLC has been launching super pressure balloons since 2013 to provide global internet coverage. Very high resolution position, temperature, and pressure observations (taken every 60 seconds) are available from thousands of flights. This provides an unprecedented source of high resolution observations to constrain gravity wave sources and propagation. The project will process the balloon measurements and, in concert with novel high resolution simulations, establish a publicly available dataset to open up a potentially transformational resource for observationally constrained assessment of gravity wave sources, propagation, and breaking. The second transformation will be using machine learning techniques to develop computationally feasible representations of momentum deposition by gravity waves. Current physics-based representations only account for vertical propagation of the waves (i.e., they are one dimensional) and ignore their horizontal propagation. Using the data based on the Loon measurements and high resolution models, one and three dimensional data driven representations will be developed to more accurately and efficiently represent the effects of gravity waves in weather and climate models. These novel representations will be implemented in idealized atmospheric models to study the role of gravity waves in the variability of the extratropical jet streams, the Quasi Biennial Oscillation (a slow variation of the winds in the tropical stratosphere) and the polar vortex of the winter stratosphere, enabling better understanding their response to increased atmospheric greenhouse gas concentrations.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.