Abstract Flood is one of the most catastrophic natural disasters in the United States, particularly in the Southeast states where hurricanes and tropical storms are most prevalent, causing billions of dollars in damage annually and significant losses of life and property. The Weather Research and Forecasting Hydrological model (WRF-Hydro) is a community-based hydrologic model designed to improve the skill of hydrometeorological forecasts, such as river discharge, through simulating hydrologic prognostic (e.g., soil moisture) and diagnostic (e.g., energy fluxes) variables. These quantities are potentially biased or erroneous due to the uncertainties involved in all layers of hydrologic predictions. In this study, we use an ensemble based Data Assimilation (DA) approach to explore the benefit of independently and jointly assimilating remotely sensed SMAP (Soil Moisture Active Passive) soil moisture (at different spatial resolutions) and USGS streamflow observations to improve the accuracy and reliability of WRF-Hydro model predictions while accounting for uncertainties. This study is conducted over a large region near to Houston, Texas where heavy rainfall from hurricane Harvey caused flooding in 2017. Before implementing DA, we first calibrated the WRF-Hydro model parameters using four United States Geological Survey (USGS) stream gauges installed within the watershed. In this step, we identified the most dominant model parameters, which were used later in the development of joint state-parameter DA. The findings of this study showed that the multivariate assimilation of soil moisture and streamflow observations results in improved prediction of streamflow as compared to univariate assimilation configurations and regardless of the watershed's streamflow regime. The results also revealed that, during the normal streamflow condition, assimilation of downscaled SMAP soil moisture at 1 km spatial resolution, would improve the accuracy of streamflow simulation more than the assimilation of coarse resolution products (i.e., the native SMAP at 36 km spatial resolution and its interpolated version at 9 km spatial resolution). However, during the period of hurricane Harvey, the soil moisture observations at different resolutions showed a similar impact on improving the streamflow prediction.

中文翻译：

---

用于增强社区 WRF-Hydro 模型预测的多变量遥感和原位数据同化

摘要 洪水是美国最具灾难性的自然灾害之一，尤其是在飓风和热带风暴最盛行的东南部各州，每年造成数十亿美元的损失和重大的生命财产损失。天气研究和预报水文模型 (WRF-Hydro) 是一种基于社区的水文模型，旨在通过模拟水文预测（例如土壤湿度）和诊断（例如能量通量）来提高水文气象预报的技能，例如河流流量) 变量。由于水文预测的所有层面都存在不确定性，这些数量可能存在偏差或错误。在这项研究中，我们使用基于集合的数据同化 (DA) 方法来探索独立和联合同化遥感 SMAP（土壤水分主动被动）土壤水分（在不同空间分辨率下）和 USGS 流量观测的好处，以提高 WRF 的准确性和可靠性-考虑不确定性的同时进行水力模型预测。这项研究是在德克萨斯州休斯顿附近的一个大区域进行的，那里的飓风哈维在 2017 年导致洪水泛滥。在实施 DA 之前，我们首先使用安装在其中的四个美国地质调查局 (USGS) 流量测量仪校准了 WRF-Hydro 模型参数。分水岭。在这一步中，我们确定了最主要的模型参数，这些参数后来用于联合状态参数 DA 的开发。这项研究的结果表明，与单变量同化配置相比，无论流域的水流状况如何，土壤水分和水流观测的多元同化都可以改进水流预测。结果还表明，在正常水流条件下，在 1 km 空间分辨率下同化缩小的 SMAP 土壤水分比粗分辨率产品（即 36 km 空间分辨率下的原生 SMAP）同化更能提高水流模拟的精度。及其在 9 公里空间分辨率下的插值版本）。然而，在飓风哈维期间，不同分辨率下的土壤水分观测对改进流量预测显示出类似的影响。