Inference of Parameters for a Global Hydrological Model: Identifiability and Predictive Uncertainties of Climate‐Based Parameters
Inference of Parameters for a Global Hydrological Model: Identifiability and Predictive Uncertainties of Climate‐Based Parameters
复制标题
DOI：
10.1029/2021wr030660
复制
发表时间：
2022-01
影响因子：
5.4
通讯作者：
T. Yoshida;N. Hanasaki;K. Nishina;J. Boulange;M. Okada;P. Troch
中科院分区：
文献类型：
--
作者：
T. Yoshida;N. Hanasaki;K. Nishina;J. Boulange;M. Okada;P. Troch
全球水文模型（GHMs）的校准已经进行了20多年的尝试；然而，目前还没有一种有效的通用校准方法。我们提出了一个新的框架来校准温室效应因子，假设参数可以通过气候相似性进行区域化。我们将5000次随机生成参数的模拟结果汇总到11个Köppen气候类别中，并使用目标函数Nash-Sutcliffe Efficiency （NSE）对H08全球水文模型的4个敏感参数进行了校准。从100倍分裂抽样检验中，我们发现，当接受前5%的样本并为气候类别分配每个接受参数分布的中位数时，传递参数集的代表性和稳健性都得到了保证。基于气候参数的模拟在480个和234个站点（777个站点的61.7%和30.1%）分别获得满意（NSE > .0）和良好（NSE > . 0.0）的结果。存储容量（SD）和导电系数（CD）对气候类型敏感，并且在可接受样本中表现出良好的约束分布，而地下存储的衰退参数（γ和τ）对气候的解释能力很小或没有解释能力。所确定的气候类型参数与土壤形成和水汽传递效率的物理解释一致。所识别的参数值与物理基础的一致性表明确定了适当的参数，这确保了参数的鲁棒性，特别是当它们被转移到未测量的流域时。
Calibration of global hydrological models (GHMs) has been attempted for over two decades; however, an effective and generic calibration method has not been explored. We present a novel framework for calibrating GHMs assuming that parameters can be regionalized by climate similarities. We calibrated four sensitive parameters of the H08 global hydrological model by aggregating the results of 5,000 simulations with randomly generated parameters into 11 Köppen climate classes and using an objective function Nash–Sutcliffe Efficiency (NSE) with random sampling from the proposed parameter distribution. From a 100‐fold split‐sampling test, we found that both the representativeness and robustness of the transferred parameter sets were guaranteed when the upper 5% of the samples were accepted and assign the median of each accepted parameter distribution for the climate class. The simulation with the climate‐based parameters yielded satisfactory (NSE > 0.0) and good (NSE > 0.5) performances at 480 and 234 stations (61.7% and 30.1% of 777 stations), respectively. The storage capacity (SD) and the conductive coefficient (CD) were sensitive to the climate classes and exhibited well‐constrained distributions of the accepted samples, whereas the recession parameters for the subsurface storage (γ and τ) showed little or no explanatory power to climate. The identified parameters for climate classes exhibited consistency with the physical interpretation of soil formation and efficiencies in vapor transfer. The consistency of the identified parameter values with physical underpinnings indicates that the appropriate parameters were determined, which ensured the robustness of parameters, especially when they are transferred to ungauged watersheds.