PREEVENTS Track 2: Collaborative Research: Flash droughts: process, prediction, and the central role of vegetation in their evolution

干旱通常被认为是一种逐渐蔓延的灾难，随着时间的推移慢慢出现。相比之下，“闪电干旱”在短短几周内急剧加剧。近年来，美国发生了一些此类事件，对农业和经济造成了重大和意外的损害。目前的预报系统对突发性干旱的反映不足，妨碍了抗旱准备工作。该项目的动机是需要推进对闪电干旱的理解，以提高我们预测它们的能力。为了做到这一点，我们将重点关注植物在突发性干旱的发展中发挥的关键作用。新的卫星技术和实地测量方法使人们有可能在植物的水分胁迫肉眼可见之前几周就检测到这种胁迫。当植物压力迅速增加时，就有很高的发生暴旱的风险。利用这种理解，我们将产生覆盖整个美国本土的闪电干旱定义和检测系统。然后，我们将根据天气和植被相互作用导致干旱的方式对闪电干旱进行分类。这些相互作用对于不同的地区或土地使用可能会有很大的不同，因此确定类别是改善预测的重要一步。使用这些类别，我们将应用最近开发的统计方法，将联合收割机植物胁迫观测与天气预报相结合，提前两周到三个月预测闪电干旱风险。在这些时间尺度上的预测可以为种植决策和救灾工作提供信息。最后，将利用先进的天气模式，选择破坏性极强的突发干旱进行详细研究，以了解土地管理和气候如何促成特别严重的事件，该项目将通过针对以下三个已知特征，推进对突发干旱的了解和预测：（1）植被和土壤湿度的观测可以在重要的提前期提供突发干旱风险的早期迹象;（2）蒸发需求是突发干旱发生的主要驱动力，并且它适合于熟练的亚季节到季节（S2 S）预报;（3）植被通过土壤水分和湍流热通量在突发干旱发展中起着核心作用。为了利用这些功能进行预测，我们提出了一个新的框架，定义闪电干旱的基础上，植被压力的快速增加是定义闪电干旱特征的核心。该框架利用先进的卫星和地面观测。我们将根据气象，水文和生态因素对美国大陆的历史性突发干旱事件进行分类，使我们能够区分具有不同过程和可预测性特征的不同类型的事件。这种分类将支持概率统计和机器学习预测模型，这些模型结合了最近开发的观测数据集和全球S2 S预测系统的联合收割机信息。对干旱等级和可预测性的分析将反过来用来选择案例，进行详细的动态模拟研究，以隔离植被的作用及其对可预测性的贡献。最后，项目期间建立的模拟基础设施将用于研究气候和土地覆盖对突发干旱的敏感性，有助于预测未来的突发干旱风险和评估土地管理备选方案。总之，这些活动将为突发干旱预测带来新的工具，有助于干旱的动态模拟，并将这些极端事件的理解和预测置于气候趋势和陆地碳平衡的更广泛背景下。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Drought is often thought of as a creeping disaster; one that emerges slowly over time. In contrast, "flash droughts" intensify dramatically in just a few weeks. A number of these events have struck the United States in recent years, leading to significant and unexpected damage to agriculture and the economy. Flash droughts are poorly represented in current forecast systems, hindering drought preparedness. This project is motivated by the need to advance understanding of flash droughts in order to improve our ability to predict them. To do this, we will focus on the critical role that plants play in the development of a flash drought. New satellite technologies and field measurement methods make it possible to detect water stress in plants weeks before that stress can be seen by eye. When plant stress increases rapidly there is a high risk of flash drought. Using this understanding, we will produce flash drought definitions and detection systems that cover the entire contiguous United States. We will then categorize flash droughts according to the ways in which weather and vegetation interact to cause the drought. These interactions can be very different for different regions or land uses, so identifying categories is an important step for improving prediction. Using these categories, we will apply recently developed statistical methods to combine plant stress observations with weather forecasts to predict flash drought risk from two weeks to three months in advance. Predictions at these time scales can inform planting decisions and relief efforts. Finally, highly damaging flash droughts will be selected for detailed study using advanced weather models, in order to understand how land management and climate contribute to particularly severe events.This project will advance flash drought understanding and forecasting by targeting three known characteristics: (1) observations of vegetation and soil moisture can provide early indications of flash drought risk at significant lead times; (2) evaporative demand is a leading driver of flash drought onset, and it is amenable to skillful subseasonal-to-seasonal (S2S) forecasts; (3) vegetation plays a central role in flash drought development via soil moisture and turbulent heat fluxes. To leverage these features for prediction, we propose a new framework for defining flash droughts based on the understanding that a rapid increase in vegetation stress is the core defining flash drought characteristic. This framework makes use of advanced satellite and ground observations. We will classify historic flash drought events across the Contiguous United States on the basis of meteorological, hydrological, and ecological factors, allowing us to distinguish different types of event that have distinct processes and predictability characteristics. This classification will support probabilistic statistical and machine learning forecast models that combine information from recently developed observation datasets and global S2S forecasting systems. Analysis of drought classes and predictability will, in turn, be used to select cases for detailed dynamically-based simulation studies that isolate the role of vegetation and its contribution to predictability. Finally, the simulation infrastructure established during the project will be used to examine climate and land cover sensitivities of flash droughts, contributing to projections of future flash drought risk and assessment of land management options. Taken together, these activities will bring new tools to flash drought prediction, contribute to dynamically-based simulation of drought, and place both understanding and prediction of these extreme events into the broader context of climate trends and the terrestrial carbon balance.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.