【作者】 李家科；

【导师】 李怀恩；

【作者基本信息】 西安理工大学 ， 环境工程， 2009， 博士

【摘要】 随着点源污染控制力度的逐步加大,我国目前正处在污染构成快速转变时期,非点源污染负荷在水污染负荷中所占比重逐步上升,影响日益突出,因此,非点源污染负荷的定量化对流域水污染控制和水环境综合治理就显得越来越重要。非点源污染来源的复杂性、机理的模糊性和形成的潜伏性,决定了非点源污染负荷定量化研究具有较大的难度。本文在调查、监测的基础上,结合3S技术,以渭河流域为背景进行流域非点源污染负荷定量化研究。首先对渭河干流临潼水文站断面进行洪水期和非洪水期水质水量同步监测,明确非点源污染特征和其在总负荷中的比重；在此基础上,以渭河流域及其子流域陕西黑河流域为例,研究不同资料条件下流域非点源污染模拟和负荷预测方法。在有限资料条件下,针对目前非点源污染年负荷量估算统计模型中一般都是考虑单变量建立一元回归模型的情况,提出了几种考虑多个主要影响因素的非点源污染负荷预测方法。在资料允许的条件下,以GIS为平台,分别研究了分布式模型AnnAGNPS和SWAT在两种不同尺度流域——黑河流域、渭河流域的适用性和可靠性,同时探讨了黑河流域、渭河流域非点源污染物产出的时空分布特性,模拟了流域非点源最佳管理措施的效果。本研究的开展,不仅对全面认识渭河水污染的成因具有重要的科学意义,而且对渭河水环境的综合治理与水资源保护具有重要的现实意义。同时,考虑到非点源污染的普遍性,本文的研究方法与结果也可供其它类似地区和流域参考。取得的主要研究成果如下：1.2006年7-12月,对渭河临潼断面5次暴雨洪水过程和非洪水期3次24h平时水流进行了水质水量同步监测。结果表明：洪水期间,SS、浑水COD、浑水TP和浑水TN等指标监测值都远大于平时监测值,浑水中COD、TP、TN监测值高与表层土壤中天然腐殖质有关。上清液中总氮TN和总磷TP分别以水溶性氮和固态磷为主,浑水中COD、TP、TN与SS有很好的相关性。采用平均浓度法计算了各指标的非点源污染平均浓度,同时建立了各指标非点源污染负荷与地表径流量的相关方程；对2001-2007年临潼断面非点源负荷占总负荷的比重进行了计算,2006年CODS、TPS、TNS和无机氮非点源污染负荷相应的比例为38.17%、35.94%、33.32%和30.45%。2001-2007年CODS、TPS、TNS和无机氮非点源污染负荷占总负荷的比例多年平均分别为54.10%、51.72%、48.82%和45.53%。非点源污染在渭河水污染中占较大比重,其对渭河水质的影响不容忽视。2.将偏最小二乘回归理论引入到流域非点源污染负荷预测中,建立了非点源污染年负荷量预测模型,并与基于最小二乘的多元线性回归模型预测结果进行对比。采用渭河华县断面1976-1993年总氮非点源负荷及与其产生关系密切的径流、泥沙、降雨资料,前15a资料用作训练,后3a资料用作检验。计算分析表明：偏最小二乘回归分析实现了多元回归、主成分分析和典型相关分析的综合,能较好的处理变量之间的多重相关性问题,建模所需样本少,其计算结果合理,具有较好的推广应用价值。3.尝试将支持向量机技术应用于小样本渭河华县断面非点源总氮年污染负荷量预测。支持向量机(SVM)能在有限样本情况下,采用结构风险最小化准则,把学习问题转化为一个二次规划问题来获得全局最优解,从而克服了神经网络易陷于局部极小值的缺点。采用1976-1993年总氮非点源负荷及与其产生关系密切的径流、泥沙、降雨资料,前15a资料用作训练,后3a资料用作检验。经过与最小二乘支持向量机、BP神经网络和最小二乘回归方法预测结果比较,表明SVM方法预测精度要优于后两者,可用于有限资料条件下非点源负荷预测。4.尝试将自记忆原理引入非点源污染负荷的预测研究中,并对一般形式的自记忆模型进行改进,建立了改进的非点源污染负荷自记忆预测模型。采用渭河华县断面1976-1993年总氮非点源负荷及与其产生关系密切的径流、泥沙、降雨资料,前15a资料用作训练,后3a资料用作检验,经计算表明,本文所提出的改进的非点源污染负荷自记忆模型预测效果较好,该方法是一种可行的非点源污染负荷预测方法。5.以陕西黑河流域为研究区,借助GIS和相关资料提取参数,建立AnnAGNPS模型数据库。采用1991-1998年黑峪口断面月径流量、泥沙和无机氮、总磷监测数据率定和验证模型,分析AnnAGNPS模型在西北半干旱地区典型流域的适用性。结果表明,AnnAGNPS可用于该流域非点源污染的长期模拟。在此基础上,对模型主要参数进行了敏感性分析和非点源污染产出的时空分布特征分析。最后利用校准后的模型进行流域非点源污染管理措施模拟。为了达到同时削减径流、沉积物及总氮、总磷的输出量,在流域内主要应采取退耕还林的措施,治理非点源污染。6.以渭河华县断面以上流域为研究区(面积为10.65万km2),应用SWAT模型对该区域的径流、泥沙及氮污染负荷的产输出过程进行模拟计算。利用华县站1987-1988年的实测月径流、泥沙和污染物数据进行模型的调参计算,用1989-1990年的实测月径流、泥沙和污染物数据对模型进行验证,验证结果表明模型基本合理可行,可用于渭河流域非点源污染的模拟计算。利用率定好的模型对不同典型年进行了模拟计算,根据模拟结果对该区域的径流、泥沙和氮污染负荷的时空产输出特点进行了分析,为流域非点源污染控制防治规划提供科学依据。更多还原

【Abstract】 With the increasing control strength of point source pollution, water pollution constitution of our country is being at a rapid change stage, and the proportion of non-point source pollution (NSP) load compared to total pollution load is gradually rising and its influence is increasing. As a result,the quantification of NSP becomes increasingly important for water pollution control and water environment comprehensive management. The complexity of source, ambiguity of mechanism and latency of formation for NSP make it difficult to quantify it. On the basis of investigation and monitoring, combining with the technology of’3S’,this paper takes the Weihe river basin as an example to study the problem. Firstly, during flood and non-flood period in 2006, water quality and quantity at the Lin-tong section of the Weihe river were monitored, and the characterristic of NSP and its proportion compared to total pollution load are analyzed. Secondly, taking the weihe river basin and its subwatershed-the Heihe river watershed as the example, simulation and prediction of NSP for watershed under different data conditions is studied. Under the limited data condition, the statistical models of NSP annual load estimation are generally single-element regression models considering single variable at present. Pointing to this condition,several NSP load prediction methods considering multiple main relevant elements are proposed. Under enough data condition, the adaptability and reliability of AnnAGNPS model in the Heihe river watershed and SWAT model in the Weihe river basin are repectively studied by means of GIS. At the same time, the NSP output distribution of time and space in the heihe river watershed and the weihe river basin are analyzed, and the effects of best management practices (BMPs)are simulated. This research is not only important to comprehensively understand the cause of water pollution of the weihe river,but also important to water environment comprehensive management and water resources protection of it. Meanwhile, considering the universality of NSP in our country, the study methods and results can give references for other similar region and watershed.To sum up,main results of this study are listed as follows:1.From July to December in 2006, five storm events and three 24 hours normal flow events at the Lin-tong section were monitored. Raw water and clarified water quality of them was analyzed. The results indicates that: during the flood, the concentrations of SS, and the COD, TP as well as TN of raw water are higher than those in the non-flood period. The high values of the COD,TP and TN of raw water are related to natural humic substance in the surface layer soil. The total-N of clarified water is mainly made of dissolved nitrogen, and total-P is mainly made of particle phosphorus. TP and COD of raw water are closely related with SS. Mean concentration method was used to estimate NSP weighted mean concentration for each element. Water quality and quantity correlation method was used to establish correlation equation between NSP load and surface runoff for each element. The calculated NSP mean concentration and the established correlation equation are used to estimate NSP load of the Lin-tong section in the Weihe River from 1991 to 1999, the results of the two methods prove to be credible. The annual NSP load proportions compared to the total of the Lin-tong section from 2001 to 2007 are calculated. The NSP load proportion of COD, TP and TN of clarified water and inorganic nitrogen are repestively 38.17%,35.94%,33.32% and 30.45% in 2006,while which are repestively 54.10%,51.72%,48.82% and 45.53% from 2001 to 2007 on the whole.The effect of NSP to water quality is indispensable.2.The partial least square regressive method (PLS) is applied to build the model estimating the annual load of NSP, whose results are compared with the ones of least square method. The data series used in the study included NSP total-N load from 1976 to 1993 at the Hua-xian section, and environmental factor series such as flow, sediment, precipitation, etc.The first 15 years’data were used for training, and the last 3 years’data for testing. It is showed that PLS integrating the multi-regression, the principal component analysis and typical correlation analysis, can easily solve the multiple correlation problems in the multiple linear regressive analysis, and only need a few of data to build the model. The calculation results of PLS is credible, the estimated regressive parameters with PLS are robust. The method is feasible and practical.3. Support Vector Machine (SVM) method is applied in estimating annual NSP load at the Hua-xian section in the Wei River. The novel SVM method can transform the learning process into a secondary planning problem, and the global optimal solution can be obtained. This method avoids the potential shortcoming of artificial neural networks (ANN), which may be trapped in local optimums. The data series used in the study included NSP total-N load from 1976 to 1993, and environmental factor series such as flow, sediment, precipitation, etc. The first 15 years’ data were used for training, and the last 3 years’ data for testing. The SVM method performed better than that ANN and least square regression methods. This study indicated that the prediction of NSP could be improved by the SVM approach under limited data availability.4.Self-memory theory is applied to build the model estimating annual NSP load. Modified self-memory model of NSP loading is established based on improving common self-memory model. The data series of the Hua-xian section in the Wei River used in the study included NSP total-N load and environmental factor series such as flow, sediment and precipitation from 1976 to 1993. The first 15 years’data were used for training, and the last 3 years’data for testing. This study indicates that the modified self-memory method performed better, and the method can be used to predict NSP load.5.Taking shaanxi Heihe river watershed as study region, AnnAGNPS (Annualized Agricultural non-Point Source Model) foundation database is established using GIS(Geographic Information System) and relevant data to extract parameter.The model is calibrated and validated using observed stream flow, sediment load, inorganic nitrogen and total phosphorus month data from 1991 to 1998 in the Heiyukou section, which is used to judge the adaptability of the model in typical watershed in semi-arid region of Northeast of China. The results showed that the model can be used to long term simulation of the watershed NSP(non-point source pollution). Subsequently, Sensitivity of main parameters of the model is analyzed. Finally, the validated model is used to simulate the management measures. The results show that the reforestation project, which can simultaneously minify the amount of runoff, sediment, total nitrogen and total phosphorus should be adopted in Heihe watershed in order to control NSP.6.Weihe river basin above Hua-xian section is selected as the study area, with a drainage area of 10.65×104 km2. SWAT model is applied to simulate the producing and transporting process of flow, sediment and nitrogen non-point source pollution. Monthly flow, sediment and nitrogen pollutant data series from 1987 to 1988 are used to calibrate the correlative parameters of flow, sediment and nitrogen pollutant. The monthly flow, sediment and nitrogen pollutant data series from 1989 to 1990 are used to validate model. The results of calibration and validation show that the model is reasonable and available. The calibrated model is used to calculate the loadings of nitrogen non-point source pollutants of different typical years. Based on the simulated results, the spatial and temporal distribution of flow, sediment and non-point source pollution has been analyzed for the non-point source pollution control.更多还原