Evaluación de la representatividad de la calidad del agua en los puntos de monitoreo de la RCHB a partir de un análisis multivariado de cargas contaminantes

Abstract

The Water Quality Network of Bogotá is established in the 4 main rivers belonging to the city of Bogotá which are: Torca, Salitre, Fucha and Tunjuelo, this monitoring network has remained static for more than ten (10) years without taking into account that the quality and quantity of water in the city is highly dynamic in time and space, and that urban and industrial development processes have changed since this management tool came into operation in 2006. In addition, at the same time as these changes, the hydrology of the basins has been modified by the effects of the increase in impermeable areas due to the expansion of urban infrastructure.

Flow rates and concentrations of different determinants of water quality in the water quality network of Bogotá have made it possible to characterise the behaviour of the water resource in the urban basins of the City. However, evaluating these components independently to determine the relevance of a monitoring point limits the findings, since urban hydrology corresponds to a non-linear system that relates the quantity and quality of water from rain and wastewater collection systems, and the alterations to the system as a consequence of urban expansion.

Therefore, in the present investigation an analysis of the representativeness of the monitoring points was carried out through a multivariate evaluation of the polluting loads of the RCHB for the time period from 2006 to 2018. First, a spatial characterization was made on the influence area of the water quality network of Bogotá belongs, in order to know the hydrography and hydrology conditions, and also with the data provided by the District Secretariat of Environment was calculated of the polluting load integrating concentrations of quality variables (BOD5, COD, TSS, Total Nitrogen, Total Phosphorus, SAAM and Fats and Oils) and streamflows.

The analysis of the representativeness of the monitoring points was done through the application of different methods, in this way the detection of atypical data through the Mahalanobis distance algorithm was first evaluated, obtaining a maximum percentage of up to 22.77 % of atypical values in the monitoring points network. In accordance with the above, a new, more reliable database without outliers was obtained to be used in the selection of the most important quality and quantity variables that represent the dynamics of the water resource by river using the Random Forest algorithm.

Finally, the load data of the most important variables per river were evaluated with the grouping algorithm Expectation Maximization to determine the number of monitoring points that represent the dynamics of the polluting loads in the water quality network of Bogotá, thus obtaining an optimal number of monitoring points per River of 4 to Torca, 5 to Salitre, 6 to Fucha and 7 to Tunjuelo. The development of this project seeks to provide knowledge on the application of methodologies to optimize the monitoring activity, and thus improve the management processes, control and monitoring of water resources in the city of Bogotá.

Keywords: Water Quality Network, Mahalanobis distance, Random Forest, Expectation Maximization, pollutant load.