PREEVENTS Track 2: Collaborative Research: Ocean Salinity as a predictor of US hydroclimate extremes

水的可用性是社会的基本必需品。作为最大的水分储存库和最终的水分来源，来自海洋的水维持了陆地降水，因此是理解陆地水循环变化的关键。洪水和干旱是水循环的极端现象，对社会造成巨大影响。近年来，西部的干旱导致了数十亿美元的农业损失和大面积的野火，而美国南部、中西部和东部的洪水也造成了类似的损失。它们是由海洋向陆地输出的水分过剩或不足造成的。海洋表面蒸发的水分是陆地降水的最终来源。因此，海洋水分供应的可用性调节了陆地上极端水文气候的严重程度。当水分离开海洋时，它会在海面盐度上留下印记。最近的研究提供了显著的新证据，表明盐度可以作为美国中西部、西南部和其他地区降水的熟练预测指标。盐度前兆明显优于基于温度的预测指标，特别是在强降水或异常干旱的年份。因此，海面盐度有很大的潜力为美国极端水文气候的季节性预报提供变革性的改进。该项目将根据这些对海洋盐度预测潜力的新见解，以及使用现场测量和卫星的不断扩大的盐度监测系统，为美国的干旱和洪水预警系统奠定科学基础。这将带来一系列社会效益：从野火和洪水中挽救生命和财产；更准确的季节性降雨预报提高了作物产量；通过更好地预测受干旱或洪水危机影响的不稳定地区，实现国家安全进步；更准确地预测能源需求和水资源短缺对发电厂的影响。一些本科生将有机会获得宝贵的研究经验，因此该项目将有助于培养下一代气候科学家。项目结果也将纳入通过麻省理工学院/世卫组织联合方案和杜克大学教授的研究生课程，并将向公众传播这些知识。将探讨产生极端降水和海面盐度之间新确定的关系的过程。将使用日降水数据和贝叶斯统计框架对美国的极端事件进行抽样。基于贝叶斯推断，将探索季前盐度前兆，并通过计算大气水分通量和地表盐度收支项确定水循环产生盐度特征的机制。此外，还将跟踪海洋对陆地的水汽通量，并评估通过大气中水汽供应和（或）能量再分配发展极端事件的过程。将开发和应用利用海面盐度前兆预测极端值的机器学习算法。该项目将使用新颖的方法，包括使用贝叶斯统计来确定极端降雨的最佳海面盐度和温度预测指标，分析海洋盐度预算以确定驱动大气变量，分析将水从海洋输送到陆地的大气环流，以及开发机器学习算法以提供极端干旱或洪水的最佳季节性预测。

Water availability is a fundamental necessity for society. As the largest moisture reservoir and ultimate moisture source, water from the oceans sustains terrestrial precipitation and is thus key to understanding variability in the water cycle on land. Floods and droughts represent extremes of the water cycle that have enormous consequences for society. In recent years Western drought has led to billions of dollars of agricultural losses and extensive wildfires, while floods produced similar losses in the South, Midwest and East of the US. They are caused by an excess or deficit of moisture exported from ocean to land. Moisture evaporating from the ocean surface is the ultimate source for terrestrial precipitation. Thus, the availability of the oceanic moisture supply modulates the severity of hydroclimate extremes on land. As moisture exits the ocean, it leaves a signature in sea surface salinity. Recent studies have provided remarkable new evidence that salinity can be utilized as a skillful predictor of precipitation in the US Midwest, Southwest and other regions. The salinity precursors significantly outperform temperature-based predictors, especially in the years with heavy precipitation or exceptional drought. Thus, sea surface salinity has great potential to provide a transformative improvement to seasonal forecasts of US hydroclimate extremes. This project will develop the scientific basis for a drought and flood early warning system for the US based on these new insights into the predictive potential of ocean salinity and the expanding salinity monitoring system that uses both in-situ measurements and satellites. This will lead to a number of societal benefits: lives saved and property preserved from wildfires and floods; improved crop yields resulting from more accurate seasonal rainfall forecasts; national security advances realized by better anticipation of destabilized regions affected by drought or flood crises; and more accurate forecasting of energy demand and the impact of water shortages on power plants. Several undergraduate students will have the opportunity to gain valuable research experience, and thus the project will help to train the next generation of climate scientists. Project findings will also be incorporated into graduate courses taught through the MIT/WHOI joint program and at Duke University, and the knowledge will be disseminated to the general public. The processes that produce the newly identified relationships between extreme precipitation and sea surface salinity will be explored. Daily precipitation data and a Bayesian statistical framework will be used to sample the extreme events in the US. Based on the Bayesian inference, the pre-season salinity precursors will be explored and mechanisms by which the water cycle generates the salinity signatures determined by calculating atmospheric moisture fluxes and the terms in the surface salinity budget. In addition, the oceanic moisture flux onto land will be tracked, and the processes assessed by which extremes develop through the moisture supply and/or energy redistribution in the atmospheric column. Machine-learning algorithms to predict extremes using the sea surface salinity precursors will be developed and applied. Novel approaches will be used in this project, including the use of Bayesian statistics to identify the optimal sea surface salinity and temperature predictors for rainfall extremes, analysis of the oceanic salinity budget to identify the driving atmospheric variables, analysis of the atmospheric circulations that transport water from ocean to land, and the development of machine learning algorithms to provide optimal seasonal predictions of extreme drought or floods.