ABSTRACT

Water monitoring programs are essential tools for planning and management, but their associated costs can present challenges, especially in the developing world where funding is tight. This paper addresses the use of data mining techniques in the periodic review of monitoring programs by analyzing the capacity of these techniques to reduce the number of monitored variables. This methodology was evaluated in two river basins in a southern Brazilian metropolis. The results demonstrated wide applicability that was not dependant on the level of anthropic influence within the basins. In addition, with complete databases and the use of automatic probes, periodic laboratory analyses may be used less frequently, saving human and financial resources and extending the effective water quality monitoring network. Of ten variables assessed, mathematical models were able to very accurately estimate nine and six in an urbanized basin and seven in a non-urbanized basin, demonstrating the effectiveness of this approach.