模型参数与边界条件对滇池水质变化的全局敏感性分析

Global sensitivity analysis of parameters and boundary conditions to water quality model in Lake Dianchi

借助湖泊系统模拟可定量地追踪滇池水质的变化过程,而模型输入水质参数和边界条件对机理模型的结果有显著的影响.然而,目前研究主要关注参数或边界条件对模拟结果的敏感性影响,缺乏兼顾模型内部方程参数与外部输入边界条件对结果的影响,并对比分析其相对大小.以滇池为例,通过对其构建的IWIND水质模型中的90个水体参数、83个底泥参数、6个气象边界条件、51个河流输入边界条件进行联合采样,计算叶绿素a、总氮、总磷、氨氮以及溶解氧5种指标的对数纳什系数(现状为基线),并进行Sobol全局敏感性分析,获得滇池的敏感性参数和边界条件.结果表明:通过比较两大类输入数据的全局敏感性,气象边界条件敏感性最大,其次是水体参数、底泥参数,河流输入边界条件敏感性最小.其中气象敏感边界条件是风速、气压、气温、相对湿度、辐射且风速敏感性最大,水体敏感水质参数主要是藻类代谢及生长过程、碳循环碳矿化过程、氧循环复氧过程与温度相关的参数(KTG2d、TRc;KTMNL;KTR),惰性颗粒态有机物沉降速率(WSrp).底泥敏感参数主要是第二层泥沙浓度(rM2)、底泥中有机碳降解过程中温度相关系数(ThKC1)、惰性颗粒态有机碳转化为底泥不同活性的有机物比例(FCRP2).新宝象河、盘龙江、海河敏感性与其流量和负荷呈正相关,但不同河流对滇池不同区域的水质影响有一定差异.本研究识别出的敏感参数和边界条件有助于提升模型的精确度,对水质模型的模拟研究有重要价值.

Changes in lake water quality can be traced quantitatively through numerical simulation. The modeling parameters and boundary conditions have significant influence on the mechanism model. However,current studies mainly focus on the sensitivity effects of parameters or boundary conditions on simulation result,lacking the coupled impacts and further comparative analysis. We took Lake Dianchi as an example,where we calculated the logarithmic Nash coefficient of the time series of chlorophyll A,total nitrogen,total phosphorus,ammonia nitrogen and dissolved oxygen using the IWIND model for a sample of 90 waterbody parameters,83 sediment parameters,6 meteorological boundary conditions,51 river input boundary conditions. The sensitivity indicators of the model parameters and boundary conditions were obtained by Sobol global sensitivity analysis. The results showed that the meteorological boundary conditions were the most sensitive,followed by water quality parameters and sediment parameters,and river input of boundary conditions. The wind speed,air pressure,temperature,relative humidity and radiation were the top most sensitive meteorological indicators,with wind speed being the most sensitive. The temperature-related parameters of algae metabolism and growth process,carbon mineralization process in the carbon cycle,the reoxygentation process in the oxygen cycle(KTG2d,TRc;KTR;KTMNL)and the sedimentation rate of inert particulate organic matter(WSrp)were the most sensitive water quality parameters. The sediment concentration of the second layer(rM2),the temperature correlation coefficient during the degradation of organic carbon(ThKC1) and the proportion of inert particulate organic matter converting to organic matter of different activities(FCRP2) were the most sensitive sediment parameters. The sensitivity of Xinbaoxiang river,Panlongjiang River and Haihe River were positively correlated with their discharge and load,but the water quality impacts of different rivers on different areas of Dianchi Lake had certain differences. The sensitive parameters and boundary conditions identified in this study are helpful to improve the accuracy of the model and have great value in the simulation of water quality model.