湖泊水动力模型外部输入条件不确定性和敏感性分析

以我国典型的大型浅水湖泊太湖为研究区域,采用国内外常用的环境流体动力学模型(EFDC),结合拉丁超立方取样(LHS)方法,研究湖泊水动力模型中4个重要的外部输入条件,即3个边界输入条件(出入湖流量、风速、风向)和1个初始输入条件(初始水位),对模型水动力模拟结果(水位、水龄以及流场)的影响与贡献.结果表明,初始水位的设定对模拟全湖水位和水龄产生决定性影响,不确定性的贡献率分别达到85.73%和66.125%,对垂向平均流速影响的贡献率只有3%;风速对表面流速模拟结果影响较大,贡献率达到58.70%,而对水位和水龄的贡献率分别为5.25%和3.00%.在垂向上,各层流速受外部输入条件不确定性的影响规律相似,贡献率排序为风速(55%~60%)>风向(10%~15%)>初始水位≈出入湖流量(1%~5%).因此在模拟大型浅水湖泊水动力过程时,可以根据不同的输出目标能够有针对性地提高外部输入条件的准确度,为提高模型精确度提供有效信息.

Uncertainty and sensitivity analysis of input conditions in large shallow lake hydrodynamic model

Uncertainty and sensitivity analysis of four important input conditions on the Environmental Hydrodynamic Fluid Code (EFDC) model results (i.e., water level, water age and currents) was investigated for a large shallow lake, Lake Taihu, China. The four input conditions included three boundary conditions (i.e., inflow/outflow, wind speed, wind direction) and an initial condition (i.e., initial water level). The Latin Hypercube sampling (LHS) as a global sensitivity method was used to estimate the uncertainty and sensitivity from the four input conditions. The results showed that uncertainties in the hydrodynamic process existed due to the uncertainties of model input conditions. Among the four input conditions, the initial water level was the most sensitive factor for the simulated water level and water age with the uncertainty contributions of 85.73% and 66.125% respectively, while it had barely 3% contributions to vertical averaged velocity. Wind speed played a significant role in the uncertainty of the velocity in the surface layer with a sensitivity coefficient of 58.70%, while it only had 5.25% and 3.00% contributions to the simulated water level and water age, respectively. Additionally, there was a similar impact of the four input conditions on the uncertainty of velocities in different layers. The four input conditions’ contributions to the velocities were as follows: wind speed (55%~60%) > wind direction (10%~15%) > initial water level ≈ inflow/outflow (1%~5%). Thus, the results provided reliable information for the model prediction of large shallow lakes like Lake Taihu. For different output targets, improving the precision of the input conditions with priority can efficiently enhance the precision of the hydrodynamic model.