Modelagem matemática e análise estatística multivariada aplicadas à avaliação da qualidade da água da sub-bacia hidrográfica do alto Paraopeba

Abstract

The discharge of urban and industrial effluents into the waters without adequate treatment, as well as the transport of sediments, mainly from agricultural areas, have altered the availability of water in quality and quantity. Therefore, it is necessary to monitor, evaluate, and diagnose the quality of water, identifying sources of pollution and trying to anticipate possible impacts that may occur, and thus provide subsidies for the effective management of water resources. In this context, mathematical modeling of water quality and multivariate statistical analysis are tools that can contribute to the evaluation of water quality. The former makes it possible to simulate the current and future scenarios of water bodies according to the behavior of the simulated variables. Statistical analysis can help in understanding the correlation of data and the dynamics of water quality variables. The main objective of this work was to evaluate the water quality of the upper Paraopeba sub-basin to support the management of water resources. For the mathematical modeling the parameters dissolved oxygen (DO), biochemical oxygen demand (BOD), nitrogen variations (organic, ammoniacal, nitrite and nitrate), phosphorus variations (organic, inorganic and total) and Escherichia coli were modeled using the QUAL-UFMG model. Two scenarios of intervention in the basin were simulated, considering the reduction of the pollutant load discharged with the implementation of sewage treatment plants (STPs) in the municipalities of the sub-basin with a horizon for the year 2035. For the multivariate statistical analysis the principal components and cluster techniques were used applied to nineteen variables in eight water quality monitoring points distributed throughout the basin. In the mathematical modeling, there was less adherence of the model in the simulation of the Maranhão River. In the current scenario, the variables presented higher values for the RMEQ index (Root Mean Square Error), which presents the similarity between observed and modeled data, and the lower this value, the more similar are the modeled and observed data. The RMEQ was 1.42 and 1.51 for BOD and DO respectively, and the modeled stretches remained within the release standards of CONAMA 357/2005 by 21% for BOD and 39.8% for DO. In the scenario simulation for the Maranhão River, there was an improvement in relation to BOD, which started to meet the recommended limits in 39% of the stretch. For the Paraopeba River, the RMEQ index showed results with lower values, the highest for BOD, with 0.70. The variables total P and E. coli were not considered in the simulation. The variables total P and E. coli presented the lowest percentages of compliance with the legislation, with 44.5% and 19.8% of the modeled stretch, respectively. In the simulation of future scenarios, there was an improvement in this result with 73.6% for total P and 24.9% for E. coli. The principal components analysis showed five components that together explained 78.46% of the variation in the data. The cluster analysis grouped the eight monitoring stations analyzed into three groups with similarities between the points, and the third group formed by the two stations located in the sub-basin of the Maranhão River showed the greatest dissimilarity with the others.