Water quality analysis using a variable consistency dominance-based rough set approach

Abstract

Analysis and evaluation of water quality and its dynamics are of prime importance for water resources and environmental monitoring. Diverse methods such as multivariate statistics, time series analysis, and neural networks have been used for modeling and analysis of water quality indicators. Although these methods are useful to explore the main body of knowledge related to the water pollution problem, they are less effective for considering inherent uncertainties and vagueness in water pollution data. In this study, a variable consistency dominance-based rough set approach (VC-DRSA) was used to explore the underlying knowledge related to data for total dissolved solids (TDSs) in the Latyan Watershed, north of Tehran, Iran. Environmental parameters for the period of 2002–2007, including precipitation, river water temperature, runoff measured at 22 monitoring sites, and two products of the MODIS sensor (16-day NDVI and land surface temperature) were the explanatory variables. VC-DRSA was used in data mining analysis to explore the most effective and reliable rules for relating TDS data to the explanatory variables. Rule validation results show that the extracted rules were very effective and straightforward for examining the important relationships between the environmental parameters and TDS data. Application of the moving average filter in the TDS data led to decreased noise and a considerable reduction in the width of the boundary region between the lower and upper approximations.