CyberSEES: Type 1: Cyber-Enabled Ensemble Data Assimilation for Drought Monitoring, Forecasting and Recovery

有证据表明，在人口增长和其他因素的推动下，持续的土地开发加剧了干旱，这一点日益引起人们的关注。这相应地加剧了缺水，威胁到水资源的长期可持续性。由于美国是水足迹最大的国家之一，这个国家非常容易受到中度和严重干旱的影响。为了减轻干旱脆弱性，有效的干旱监测预报系统和干旱恢复时间评估对决策者至关重要。该项目将开发一个网络集成数据同化和陆地建模系统，以表征陆地表面状况，不仅可以评估农业干旱，还可以为概率干旱预测提供初始条件。这些估计数将为估计干旱恢复情况提供依据。该项目将成为大规模干旱研究和应用能力建设的样板，并将直接提高国家水资源管理者在干旱造成的水资源短缺期间采取适当和及时措施的能力。该研究依赖于大量的地球和环境观测数据，并结合先进的动力学和贝叶斯模型，以解释不确定性并提供可靠的干旱估计，目标是进一步提高淡水资源的可持续性。利用数据同化方法对地球系统科学中的不确定性进行量化的集成建模和概率估计一直是地球系统科学应用中的一个突出瓶颈。由于水文过程的多尺度性质和大多数水文系统的不确定性，以及认知和随机不确定性的存在，基于物理的动态和随机建模以概率表征干旱的发生并在季节尺度上进行预测，需要对系统进行适当的参数化。本研究提出了一个实时运行的现代集成数据同化系统，用于表征陆地表面状况，用于监测干旱，开发了一个有效数据同化的计算框架，并在计算框架中实现了一种针对计算可扩展性、功率和可靠性的方法。

There is growing concern with evidence that droughts have been intensified due to ongoing land development driven by population growth and other factors. This has correspondingly aggravated water scarcity, which threatens the long-term sustainability of water resources. Since the US is one of the largest in terms of water footprint, the country is very vulnerable to moderate and severe drought. To mitigate the drought vulnerability, an effective drought monitoring and forecasting system with assessment of drought recovery time is critical for decision makers. This project will develop a cyber-enabled ensemble data assimilation and terrestrial modeling system to characterize the land surface condition for not only assessing the agricultural drought but also providing the initial condition for probabilistic drought forecasting. These estimates will provide the basis for drought recovery estimation. The project will serve as a prototype to build capacity for large-scale drought studies and applications, and will directly enhance the ability of state water managers to take appropriate and timely measures during periods of water scarcity as a result of drought.The study relies on a variety of massive earth and environmental observational data in connection with advanced dynamical and Bayesian modeling, to account for uncertainties and provide reliable drought estimation with the goal to further freshwater resource sustainability. Ensemble modeling and probabilistic estimation to quantify the uncertainties in Earth system science by means of data assimilation has been a salient bottleneck in operationalization. Due to the multi-scale nature of hydrologic processes and under-determinedness of most hydrologic systems as well as the presence of epistemic and random uncertainties, dynamic physical-based and stochastic modeling to probabilistically characterize drought onset and predict it at seasonal scale requires proper parameterization of the system. This research proposes a modern ensemble data assimilation system operating in real time to characterize land surface conditions for monitoring drought, develops a computational framework for effective data assimilation, and implements an approach that targets computational scalability, power, and reliability in the computational framework.