Multivariate Analysis for selection of monitoring parameters in water reservoir of Juiz de Fora, Minas Gerais

Abstract The need to control land use and monitor water quality for public supply requires the evaluation of many parameters, which are often difficult to analyze and interpret due to cost and the large volume of data. In this context, this work selected the most relevant parameters for monitoring the Dr. João Penido Dam, the main supply source of Juiz de Fora, MG. The evaluation was conducted based upon a series of thirteen parameters of the catchment provided monthly by the Municipal Sanitation Company between 1998 and 2012. The data were statistically processed using Factor Analysis and Principal Component Analysis and parameters with highest variances were extracted. Component 1 (FV1) reflected the influence of dissolved solids associated with runoff processes in the watershed and with the erosion of margins. FV2, as well as FV3, are strongly associated to the potential discharge of sewage in the watershed area. Using multivariate analysis, we removed five parameters that showed no variance: hardness, consumed oxygen, pH, alkalinity and manganese.摘要 为管控土地利用并监测公共供水水质，需对多项参数开展评估，但受成本约束与数据体量庞大的影响，此类参数往往难以开展分析与解读。在此背景下，本研究选取了巴西米纳斯吉拉斯州茹伊斯迪福拉市主要供水水源——若昂佩尼杜博士大坝（Dr. João Penido Dam）水质监测所需的核心参数。本次评估基于1998年至2012年间由该市市政卫生公司（Municipal Sanitation Company）按月采集并提供的流域13项监测参数开展。研究采用因子分析（Factor Analysis）与主成分分析（Principal Component Analysis）对数据集开展统计处理，并提取方差贡献占比最高的参数。主成分1（FV1）反映了与流域径流过程及岸坡侵蚀相关的溶解态固体污染物影响；主成分2（FV2）与主成分3（FV3）则均与流域范围内潜在的污水排放行为强相关。本研究通过多变量分析剔除了5项无方差贡献的参数：硬度、耗氧量、pH、碱度与锰。