RII Track-2 FEC: Precise Regional Forecasting via Intelligent and Rapid Harnessing of National Scale Hydrometeorological Big Data

全球变暖已成为一个具有国家重要性的严峻问题，因为它导致极端天气和气候事件更加频繁，造成越来越大的经济损失，并对农业、交通、水资源管理、城市规划等众多部门产生不利的社会影响。为了更好地对天气和气候参数进行观测和数值模拟，以提高预报精度，本项目通过智能和快速利用全国范围的水文气象大数据进行精确的区域预报。它的目标是通过整合大量的大气数据集和收集到的地面数据来进行更精细的时空预测，包括基础研究和实验活动，从而改善目标地区的气象和水文预报。它的解决方案是创新的，利用实际收集的数据作为反馈，使预测模型在多个近期时间范围内生成更好的产品。通过利用(1)提出的一系列简单神经网络模型（称为模型集）和(2)研究团队成员开发或正在开发的多种加速方法，智能快速地利用大数据，可以获得更好的区域预测结果。基于模型的天气预报方案在空间和时间上的改进适用于全国所有地区，便于携带。路易斯安那州、阿拉巴马州和肯塔基州的五所大学以及美国地质调查局（U.S. Geological Survey）正在进行协同合作，从而在与加速数据分析、气象学和水文学相关的科学和工程领域广泛参与发现和创新。除了促进科学进步外，这个多学科项目还通过遏制全球变暖带来的潜在破坏来促进国家的繁荣和福祉。该项目还包括以下方面的全面努力：(1)通过合作和监督初级研究人员的职业发展，建立未来的领导力；(2)丰富重点学科的教育材料，加强学生研究，以促进劳动力发展；(3)积极招募和吸引代表性不足的参与者，以支持多样性。更好的天气和气候参数观测和数值模式提高了预报的准确性，能够抑制全球变暖导致的与极端天气和气候事件有关的灾害造成的经济损失和社会影响的上升趋势。这个多学科项目包括基础研究和实验活动，建立并扩展了团队成员在计算机科学与工程、气象学、水文学和电气与计算机工程等学科的早期工作。它解决了在全国范围内智能和快速利用水文大数据进行区域精确气象水文预报的技术挑战，其预期结果可能会推动神经网络智能大数据利用和通过各种方法快速处理数据的前沿。智能大数据利用来自适当的端到端简单神经网络模型（称为模型集）的结果，这些模型通过从近地观测（通过Mesonet站或水表）和基于计算物理大气方程的地理网格预测（通过高分辨率快速刷新的天气研究和预报模型）连续获得的大量数据集进行创造性地训练。加速大数据利用的各种方法正在开发中，并将在项目期间进行全面评估，包括(1)有效的计算机系统DRAM扩展，(2)执行弹性增强，以及(3)基于SC（随机计算）的加速器和具有理想调度策略的gpgpu的高计算密度支持模型训练和推理。通过智能和快速（PREFER）大数据利用，基于模型的精确区域预报解决方案在空间和时间上改善了天气预报，广泛应用于全国所有地区，易于携带，提供短期和精细的空间分辨率预报。它们的目标是解决气象预报应用中的重要问题（例如，强雷暴或热带系统的登陆）、加强洪水预警、调查回水湿地的蓄水能力以缓解河流洪水等。这项PREFER工作激发和滋养了路易斯安那州、阿拉巴马州和肯塔基州的五所大学以及美国地质调查局之间的跨学科研究和司法合作，有助于整合研究和教育，同时促进相互交织的项目活动的科学内容的发现和理解。它包含了以下方面的全面努力：(1)提升初级研究人员的职业发展，以建立未来的领导力；(2)丰富重点学科的教育材料，支持学生研究，以促进劳动力发展；(3)吸引代表性不足的群体积极参与；(4)广泛传播项目成果和软件工具。该奖项反映了美国国家科学基金会的法定使命，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。

Global warming has emerged as a stark problem of national importance, as it results in more frequent extreme weather and climate events that cause rising economic loss and adverse societal impacts on numerous sectors, such as agriculture, transportation, water resource management, urban planning, among others. For better observations and numerical models on weather and climate parameters to improve forecasting accuracy, this project addresses precise regional forecasting via intelligent and rapid harness on national scale hydrometeorological Big Data. It aims to improve meteorological and hydrologic forecasts at target regions of interest by integrating massive atmospheric data sets with gathered surface data for finer temporal and spatial predictions, containing both fundamental research and experimental activities. Its solution approach is innovative by leveraging the actual gathered data as feedback to make prediction models generate better products with multiple near-term time horizons. Better regional prediction results from harnessing Big Data intelligently and rapidly via utilizing (1) a collection of proposed simple neural network models (called modelets) and (2) multiple accelerating methodologies developed or under development by the research team members. The modelet-based solutions for improving weather prediction spatially and temporally are applicable to all regions in the nation, with easy portability. They are being undertaken synergistically by jurisdictional collaboration across five universities in Louisiana, Alabama, and Kentucky, plus U.S. Geological Survey, enabling broad engagement at the frontiers of discovery and innovation in science and engineering related to accelerating data analytics, meteorology, and hydrology. Besides promoting the progress of science, this multidisciplinary project advances the national prosperity and welfare by curbing potential disruption due to global warming. The project also includes comprehensive efforts for (1) building future leadership through collaboration and supervision of junior investigators for their career advances, (2) enriching educational materials on the focused disciplines and strengthening student research to boost workforce development, and (3) aggressively recruiting and engaging underrepresented participants to support diversity.Better observations and numerical models on weather and climate parameters improve forecasting accuracy, able to suppress the uptrend in economic loss and societal impacts caused by disasters pertinent to extreme weather and climate events, as a result of global warming. This multidisciplinary project involves both fundamental research and experimental activities, built upon and expanding earlier work of team members in the disciplines of computer science and engineering, meteorology, hydrology, and electrical & computer engineering. It deals with the technical challenges of intelligent and rapid harness on national scale hydrometeorological Big Data for precise meteorological and hydrological forecasting regionally, with anticipated outcomes likely to push the frontiers of intelligent bigdata harness by NNs (neural networks) and of speedy data processing through various methodologies. Intelligent bigdata harness results from proper end-to-end simple NN models (called modelets), which are trained inventively by huge datasets obtained continuously from both near-ground observations (via Mesonet stations or water gauges) and geo-gridded predictions based on computing physical atmospheric equations (via the Weather Research and Forecasting model with High-Resolution Rapid Refresh). Various methodologies for accelerating bigdata harness are under development and to be evaluated thoroughly during the project years, including (1) effective computer system DRAM expansion, (2) execution resilience enhancement, and (3) high-compute density support by SC (stochastic computing)-based accelerators and by GPGPUs with desirable scheduling policies for modelet training and inference. The modelet-based solutions for precise regional forecasting via intelligent and rapid (PREFER) bigdata harness improve weather prediction spatially and temporally for wide applications to all regions in the nation, with easy portability to offer short-term and fine spatial resolution prediction. They aim to address the important problems of meteorological forecast applications (e.g., landfalling of severe thunderstorms or tropical systems), flood warning alert enhancement, backwater wetland storage capacity investigation for river flood mitigation, among others. This PREFER work inspires and nourishes cross-disciplinary research and jurisdictional collaboration across five universities in Louisiana, Alabama, and Kentucky, plus U.S. Geological Survey, helping to integrate research and education while advancing discovery and understanding in the scientific contents of interwoven project activities. It contains comprehensive efforts for (1) lifting career development of junior investigators to build future leadership, (2) enriching educational materials on the focused disciplines and supporting student research to spur workforce development, (3) engaging active participation from underrepresented groups, and (4) disseminating project outcomes and software tools widely.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.