FOR 2630: Understanding the global freshwater system by combining geodetic and remote sensing information with modelling using a calibration/data assimilation approach (GlobalCDA)

当代全球水文(或地表)模型对年平均蒸散量或河流流域的低、中、高流量提供了相互矛盾的估计，导致对当前水资源可获得性或气候变化对淡水资源影响的估计大相径庭。拟议的研究组(RU)的中心目标是增进我们对全球淡水资源的了解，并更好地估计大陆水通量(径流、地下水补给、实际蒸散和可再生水资源)和蓄水量(雪、土壤、地下水和地表水以及冰川)。我们的假设是，只有将最先进的水文模型与多种新的和经过优化处理的大地测量和遥感数据结合在一起，采用基于集合的校准和数据同化(C/DA)方法，才能实现重大改进，这种方法允许根据建模目的进行灵活的参数(校准)和状态(数据同化)调整。这种办法还考虑到了模型结构和投入的不确定性，目前尚未实施。因此，我们为第一阶段制定了两个主要目标：(1)开发基于多观测集合的C/DA方法，以在全球范围的第二阶段将模型输出变量(基于量规的径流和GRACE和GRACE-FO总储水量异常的时间序列，以及关于积雪、地表水体和径流的范围和水位的遥感数据)的观测数据以最佳方式与水文模型相结合；(2)将该方法与全球水文模型Water GAP相结合，以改进对淡水通量和储水量的定量评估，包括它们对气候和人为强迫的不确定性，用于关键盆地和估计由于大陆水储存变化而引起的海洋质量变化。C/DA方法包括用于数据同化和参数校准的集合卡尔曼滤波法(EnCDA)和帕累托最优校准方法(POC)，该方法能够优化时间常数参数，但需要在第二阶段进行扩展，以考虑校准数据的不确定性和模型输出不确定性的估计。敏感性分析、C/DA提供的不确定信息以及针对独立数据的模型验证将使我们能够评估应用C/DA方法的附加值。

Contemporary global hydrological (or land surface) models provide conflicting estimates of e.g., mean annual evapotranspiration or low, mean, and high flows in river basins, resulting in strongly differing estimates of current water availability or of climate change impacts on freshwater resources. The central objective of the proposed Research Unit (RU) is to improve our understanding of global freshwater resources and to obtain better estimates of continental water fluxes (streamflow, groundwater recharge, actual evapotranspiration, and renewable water resources) and storages (in snow, soil, groundwater and surface water bodies as well as in glaciers). It is our hypothesis that a major improvement can only be achieved through combining state-of-the-art hydrological modelling and multiple new and optimally processed geodetic and remote sensing data in an ensemble-based calibration and data assimilation (C/DA) approach that allows a flexible parameter (calibration) and state (data assimilation) adjustment tailored to the modelling purpose. Such an approach, which also takes into account uncertainty due to model structure and input, has not been implemented yet. Therefore, in this RU we formulated two primary goals for the first phase: (1) develop a multi-observation ensemble-based C/DA methodology to combine, in the second phase at the global scale, observational data of model output variables (time series of gauge-based streamflow and GRACE and GRACE-FO total water storage anomalies as well as remotely-sensed data on snow cover, extent and water level of surface water bodies and streamflow) with hydrological models in an optimal manner, and (2) exploit this methodology with the global hydrological model WaterGAP to provide an improved quantitative assessment of freshwater fluxes and storages including their uncertainties in response to climate and anthropogenic forcing, for critical basins and for estimating ocean mass change due to continental water storage changes. The C/DA approach encompasses an Ensemble Kalman Filter approach (EnCDA) for both data assimilation and parameter calibration and a Pareto-optimal calibration approach (POC) that enables optimization of temporally constant parameters but needs to be expanded, in the second phase, to take into account the uncertainty of calibration data and estimation of model output uncertainty. Sensitivity analyses, uncertainty information provided by C/DA as well as model validation against independent data will allow evaluating the added value of applying the C/DA approach.