Collaborative Research: Observed and Future Dynamically Downscaled Estimates of Precipitation Associated with Mesoscale Convective Systems

中尺度对流系统(MCS)是在比其所包含的风暴更大的范围内组织起来的雷暴的集合，其中单个雷暴协同作用来产生组织和维持该系统的大气运动。这些大型风暴系统产生了冰雹、洪水和龙卷风等极端天气，但在生长季节，它们也对美国大陆三分之二的东部地区的水资源做出了重要贡献。这个项目试图从总体上理解MCS的行为，包括MCS降水对CONUS总体水平衡的长期贡献，以及MCS活动在异常潮湿(洪水)或干燥(干旱)条件下年际变化的重要性。开展这项研究的一个关键工具是国际气象局(WSI)国家业务天气雷达(NOWrad)数据集，这是由国家气象局雷达站创造的20年来的纪录(目前为1996-2015年)，这些雷达站提供了对CONUS的连续近总覆盖。该项目的一个主要目标是开发和应用一种自动程序来检测和跟踪雷达数据中的MCS。该算法将MCS识别为雷达地图中沿系统长轴至少100公里的区域中超过阈值反射率值的连续或半连续特征。MCS在传播时分裂和合并的趋势使MCS跟踪变得复杂，该算法结合了一种识别合并和拆分的方法。另一个问题是，锋面气旋和登陆飓风中可能出现大片强降水区域，需要一种分类方案将这些区域与MCS区分开来。开发了一种机器学习技术来执行这种分类，使用专家判断来训练随机森林分类器(RFC)方案。通过一项调查征求专家的进一步判断，该调查邀请研究界参与跟踪和分类计划的制定和验证。然后利用MCS事件及其特征(强度、持续时间、结构等)来研究MCS的季节性、年际变率以及对CONUS降水(包括洪水和干旱)的贡献。进一步使用全球气候模式(GFDL-CM3)和区域对流允许模式(WRF-ARW，水平分辨率为4公里)来模拟当前和预测未来气候条件下CONUS上空的MCS。根据为NOWrad数据开发的跟踪和分类方案对模拟进行了分析，模式模拟允许检验MCS行为如何依赖于气候因素，如对流层湿度、土壤湿度、大气稳定性和大尺度大气环流。由于MCS降雨作为农业水资源的重要性以及与MCS活动相关的恶劣天气灾害，该工作具有更广泛的影响。为该项目制作的算法和数据集将通过一个在线门户网站与研究人员、业务气候学家和水文学家共享。此外，该项目支持和培训一名研究生，并为一名本科生提供暑期支助，从而为这一领域未来的科学工作人员提供支持。

A mesoscale convective system (MCS) is a collection of thunderstorms organized on a larger scale than the storms it contains, in which the individual thunderstorms act in concert to generate the atmospheric motion that organizes and sustains the system. These large storm systems produce extreme weather including hail, floods, and tornados, but they also make an important contribution to water resources over the eastern two thirds of the continental US (CONUS) during the growing season. This project seeks to understand MCS behavior in an aggregate sense, including the long-term contribution of MCS precipitation to the overall water balance of the CONUS and the importance of year-to-year variability in MCS activity for anomalously wet (flood) or dry (drought) conditions. A key tool for conducting the research is the Weather Services International (WSI) National Operational Weather radar (NOWrad) data set, a 20-year record (currently 1996-2015) created from the National Weather Service radar stations which provide continuous near-total coverage of the CONUS. A primary goal of the project is to develop and apply an automated procedure to detect and track MCSs in the radar data. The algorithm identifies MCSs as contiguous or semi-contiguous features in radar maps over an area of at least 100km along the system's major axis exceeding a threshold reflectivity value. MCS tracking is complicated by the the tendency of MCSs to split and merge as they propagate, and the algorithm incorporates a method for identifying mergers and splits. A further issue is that large regions of intense precipitation can occur in frontal cyclones and landfalling hurricanes, and a classification scheme is necessary to distinguish these regions from MCSs. A machine learning technique to perform this classification is developed using expert judgement to train a random forest classifier (RFC) scheme. Further expert judgement is solicited through a survey which invites the research community to participate in the development and validation of the tracking and classification schemes. The catalog of MCS events and their characteristics (intensity, duration, structure, etc) is then used to study MCS seasonality, interannual variability, and contribution to CONUS rainfall including floods and droughts.Further work uses a global climate model (GFDL-CM3) in combination with a regional convection permitting model (WRF-ARW at 4km horizontal resolution) to simulate MCSs over the CONUS under present-day and projected future climate conditions. The simulations are analyzed according to the tracking and classification schemes developed for the NOWrad data, and the model simulations allow examination of how MCS behavior depends on climatic factors such as tropospheric moisture, soil moisture, atmospheric stability, and large-scale atmospheric circulation.The work has broader impacts due to the importance of MCS rainfall as a water resource for agriculture and the severe weather hazards related to MCS activity. The algorithms and datasets produced for the project will be shared with researchers and operational climatologists and hydrologists through an online portal. In addition, the project supports and trains a graduate student and provides summer support for an undergraduate, thereby providing for the future scientific workforce in this area.