Remotely sensed ensemble of the water cycle

We present a new REmotely Sensed ENsemble of the water cycle (REESEN). The REESEN approach generates a large number of realizations of the remotely sensed water budget and enforces closure for each realization. The REESEN approach is applied to 24 large river basins from Oct. 2002- Dec. 2014. Three water balance closure algorithms are evaluated, ranging from simple redistribution of residuals to more complex Kalman-filtering and multiple-collocation approaches, to understand the impact of algorithm choice on the resulting water budget partitioning. Therefore, three REESEN ensembles are generated at each basin, one for each closure technique. Compared with a published climate data record, the ensemble shows strong agreement for precipitation, evapotranspiration and changes in storage (R2: 0.91-0.95), with less agreement for streamflow (R2: 0.42-0.47), which may be indicative of LSM biases in the climate data record. Water balance residual errors resulting from combinations of raw products vary significantly (p&lt;0.001) with latitude, with a tendency for positive biases for low- and mid-latitude basins, and negative biases elsewhere. Overall, residual errors are equivalent to 15% of total precipitation when averaged across all data products and basins. Enforcing water balance closure reduces uncertainty relative to raw retrievals for 88%-100% of all timesteps. This observation-based dataset is distinct from modeled estimates and therefore has the potential to preserve important information of anthropogenic effects on the water balance. For example, the ensemble shows higher evapotranspiration and greater storage depletion than the model-based climate data record during months of heavy irrigation over the Sacramento-San Joaquin basin system. MATLAB files were created for REESEN ensembles used in Abolafia-Rosenzweig et al. (2020). There is one file for each basin system and each closure algorithm: proportional redistribution (PR), multiple collocation (MCL), constrained Kalman filter (CKF). The file descriptions are described in the README. Shapefiles used in the study (Fig. 1 of Abolafia-Rosenzweig et al., 2020) are in corresponding folders to each basin.本研究提出一种全新的遥感水循环集合数据集（Remote Sensed Ensemble of the Water Cycle, REESEN）。REESEN方法可生成大量遥感水文收支的实现样本，并对每个样本强制实现水量闭合。该方法被应用于2002年10月至2014年12月期间的24个大型流域。为探究算法选择对最终水文收支分配结果的影响，本研究评估了三种水量闭合算法，从简单的残差重分配方法，到更为复杂的卡尔曼滤波与多重配准方法。因此，每个流域均生成三套REESEN集合，分别对应一种水量闭合算法。与已发布的气候数据记录相比，该集合在降水量、蒸散发以及储水量变化方面表现出极强的一致性（决定系数R²：0.91~0.95），而在径流量方面一致性较弱（R²：0.42~0.47），这可能暗示了气候数据记录中地表模式（Land Surface Model, LSM）存在偏差。由原始数据集组合得到的水量平衡残差误差随纬度呈现显著差异（p<0.001）：低纬度和中纬度流域多表现为正偏差，其余流域则多为负偏差。总体而言，当对所有数据集和流域取平均时，残差误差相当于总降水量的15%。相较于原始反演结果，强制实现水量闭合可在88%~100%的所有时间步长上降低数据集的不确定性。本观测驱动数据集不同于模型模拟结果，因此有望保留人类活动对水量平衡影响的重要信息。例如，在萨克拉门托-圣华金流域系统的高强度灌溉月份，该集合的蒸散发量高于基于模型的气候数据记录，且储水量消耗也更为显著。本研究为Abolafia-Rosenzweig等人（2020）中使用的REESEN集合生成了MATLAB格式文件，每套流域系统与每种水量闭合算法对应一个文件：比例重分配（Proportional Redistribution, PR）、多重配准（Multiple Collocation, MCL）以及约束卡尔曼滤波（Constrained Kalman Filter, CKF）。文件的详细说明详见README文档。本研究中使用的形状文件（对应Abolafia-Rosenzweig等人2020年论文中的图1）存放在各流域对应的文件夹中。