基于去耦合直接法的兰江流域地表水水质预测方法研究

针对影响因素众多、耦合机制复杂情况下的地表水污染物浓度预测问题，将河道污染物浓度的变化量表示为各种影响因子一阶偏导项和二阶偏导项的线性叠加。其中，一阶偏导项可描述影响因子变化与污染物浓度变化的直接关系，二阶偏导项可描述影响因子之间交互作用对污染物浓度变化的影响。在此基础上，提出了用以模拟地表水污染物浓度的去耦合直接法。采用2014—2016年兰江流域将军岩、低田、半潭、沈村、焦岩5个断面的水文和氨氮、高锰酸盐指数、总氮、总磷4项污染指标逐日实测数据，通过差分法求解了一阶和二阶偏导项，并采用2017—2019年实测数据对模型进行了验证和评价。结果表明：去耦合直接法能够有效预测地表水主要污染物浓度的变化方向和变化量，且模拟结果和实测结果的符合情况较好，4项污染指标模拟值的Nash-Sutcliffe系数为0.479～0.654,模拟值与实测值的偏差为0.070～0.352;汇流区面积增加后，影响因子不确性对污染物浓度的扰动减小，污染物浓度变化的规律性增强，去耦合直接法的模拟精度升高。与SWAT模型的对比分析结果显示，在污染成因不发生显著变化的情况下，去耦合直接法的模拟精度优于SWAT模型。

Prediction of Surface Water Quantity in Lanjiang River Basin Based on the Decoupled Direct Method

Aiming to pollutant prediction problems in surface water system affected by numerous factors and complicated coupling mechanisms,a decoupled direct method was proposed to predict pollutant concentrations,using a linear superposition of first- and second-order partial derivatives of influencing factors,which was expressed by the variation of the pollutant concentrations.The first-order partial derivatives could characterize the relationship between influential factor variation and pollutant concentration change,and second-order partial derivatives could represent the influence of the interaction among influencing factors on pollutant concentration change.Daily hydrological and water quantity (i.e.,ammonia nitrogen,permanganate index,total nitrogen and total phosphorus concentrations) data of five monitoring sections (Jiangjunyan,Ditian,Bantan,Shencun,and Jiaoyan) in Lanjiang River Basin from 2014 to 2016 were used to solve first- and second-order partial derivatives by difference method,and the data from 2017 to 2019 were used to validate and evaluate the model performances.The results suggested that,calibrated models could accurately predict the change direction and amount of the concentration of major pollutants in surface water,and the simulation results were in good agreement with the measured results.The Nash-Sutcliffe efficiency coefficients were 0.479 to 0.654,and the relative differences between the simulated and the measured concentrations were 0.070 to 0.352.The performances of proposed method simulating pollutant concentrations in large catchments were better than those in small catchments since most of the fluctuations in concentration caused by the uncertainty of influencing factors were balanced in the larger catchments.The accuracy of the decoupling direct method was better than that of the SWAT model under the conditions that the causes of pollution were not significantly changed.