Water quality data analysis and modeling of the Langat river basin
Water quality data analysis and modeling of the Langat river basin
复制标题
冷岳河流域水质数据分析和建模
DOI：
--
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
H. Juahir
中科院分区：
文献类型：
--
作者：
H. Juahir
本文对水质时空格局的研究和人工神经网络预测模型的发展进行了研究。这些数据基于有关水质、水文和气象变量、土地利用变量和主要朗加特河沿线景观指标的次要数据。本研究结合环境计量学、地理信息系统和非参数趋势检验三种不同的工具，研究了河流水质和土地利用在季节和空间影响下的变化及其相互关系。新的景观指标使用斑块分析来表示兰加特河流域从上游到下游的土地利用变化造成的污染负荷的累积影响。最后，在上述分析结果的基础上进行了人工神经网络预测模型的建立。本研究使用了美国能源部1988-2002年7个监测站的23个水质数据。利用1974年、1981年、1984年、1988年、1990年、1991年、1995年、1996年、1997年和2001年的土地利用图，进行了基于农业、森林、城市、水体和其他五种土地利用类型以及LUSI的土地利用分析。对1公里、2公里和3公里缓冲区以及整个朗加特盆地地区进行了各种土地利用分析。应用层次凝聚聚类分析(HACA)、判别分析(DA)、主成分分析(PCA)、因子分析(FA)、多元线性回归和人工神经网络(ANN)等方法研究了主要水质变量的时空变化和污染源的来源。基于HACA的水质参数分析，形成了三个子流域群。这些集群分别被指定为高污染源(HPS)区、中等污染源(MPS)区和低污染源(LPS)区。因此，还使用DA确定了有助于区域集群化的重要水质参数。在时间上，基于季节(干湿)变化的DA方法成功地区分了10个水质变量，即温度(T)、溶解氧(DO)、pH、电导率(Cond)、盐度(SAL)、总固体(TS)、氯(Cl)、钾(K)、镁(Mg)和大肠杆菌。为了确定和定性描述河流污染的贡献者，使用了主成分分析和FA(varimax功能)。高污染源(HPS)区得到7个主成分(PC)，总方差为81%；中等污染源(MPS)和低污染源(LPS)区分别获得6个主成分，总方差分别为71%和79%。HPS和MPS的污染源来自人为来源(工业、城市废物和农业径流)。对于该区域，家庭和农业径流是已确定的主要污染源。在对兰加特流域土地利用和兰加特河水质参数进行分析的基础上，建立了用于建立人工神经网络预测模型的投入产出关系。基于不同的输入和输出参数，建立了五种不同类型的预测模型，用于预测(I)河流级别(两个模型)、(Ii)水质参数、(Iii)污染区域和(Iv)土地利用。用人工神经网络模型进行预测，取得了令人振奋的结果，预测精度可以接受。从这项研究中我们可以得出结论，在朗加特河流域数据上应用各种环境测量和统计方法，能够揭示这一大型复杂河流系统土地利用和地表水污染的时空变异性。基于数据的人工神经网络模型的开发也产生了有用的模型，这些模型可以作为决策者在规划更有效和更可持续的土地开发政策和水质监测方案时的决策工具。
This thesis concerns the investigation of spatial and temporal water quality pattern and the development of artificial neural network (ANN) prediction models. These are based on secondary data on water quality, hydrological and meteorological variables, land use variables and landscape metrics along the main Langat River. In this work three different tools, namely envirometrics, GIS and non-parametric test of trend were integrated to investigate the changes in river water quality and land use based on seasonal and spatial affects and their relationship with each other. The new landscape metrics were developed using patch analysis to represent the cumulative effects of pollution loading due to land use changes from upstream to downstream of the Langat River Basin. Finally the development of ANN prediction models was carried out based on the results obtained by the analyses mentioned above.23 water quality data collected from seven monitoring stations manned by DOE from 1988 to 2002 were used in this study. Land use analyses based on five land use types, namely, agriculture, forest, urban, waterbody and others as well as LUCI, were developed using land use maps of years 1974, 1981, 1984, 1988, 1990, 1991, 1995, 1996, 1997, and 2001. Various land use analyses were carried out for the 1, 2 and 3 km buffer areas as well as the whole Langat Basin area. Hierarchical agglomerative cluster analysis (HACA), discriminant analysis (DA), principal component analysis (PCA), factor analysis (FA), multiple linear regression and ANN were applied to study the spatial and temporal variations of the most significant water quality variables and to investigate the origin of pollution sources. Three sub basin clusters were formed based on water quality parameter analysis using HACA. These clusters are designated as high pollution source (HPS), moderate pollution source (MPS) and low pollution source (LPS) regions respectively. Significant water quality parameters that contribute to the clustering of the regions were also consequently determined using DA. Temporally, ten water quality variables, namely, temperature (T), dissolved oxygen (DO), pH, conductivity (Cond.), salinity (SAL), total solids (TS), chlorine (Cl), potassium (K), magnesium (Mg) and E.coli were successfully discriminated using DA based on seasonal (wet and dry) variations. In order to ascertain and qualitatively describe contributors to the pollution of the river, PCA and FA (varimax functionality) were used. Seven principal components (PCs) were obtained with 81% total variance for the high pollution source (HPS) region, while six PCs with 71% and 79% total variance were obtained for moderate pollution source (MPS) and low pollution source (LPS) regions respectively. The pollution sources for the HPS and MPS are of anthropogenic origins (industrial, municipal waste and agricultural runoff). For the LPS region, the domestic and agricultural runoffs are the identified main sources of pollution. Based on the analyses carried out on land use of the Langat Basin and water quality parameters of Langat River, input-output relationships were established for the development of ANN prediction models. Five different types of prediction models based on different input and output parameters were developed to predict (i) river class (two models), (ii) water quality parameters, (iii) pollution region, and (iv) land use. Encouraging prediction results were obtained using the ANN models with acceptable accuracies. From this study we can conclude that the application of the various envirometric and statistical methods on the Langat river basin data is able to reveal meaningful information on the temporal and spatial variability of land use and surface water pollution of this large and complex river system. Development of ANN models based on the data also yields useful models that can be employed as decision tools for policy makers in planning for more effective and sustainable land development policies and water quality monitoring programs.