Application of multivariate statistical analysis to incorporate physico-chemical surface water quality in low and high flow hydrology

Abstract

Multivariate statistical methods, such as principal components analysis (PCA), discriminant analysis (DA) and general linear models (GLM) were applied to incorporate physico-chemical surface water quality in low and high flow hydrology in Northern Iran, based on analysis of the 7-day low flow index and existing water quality data. In view of this, 7-day low flows were calculated for 15 water years (1991–2006) at 15 monitoring stations. Eleven water quality parameters were extracted during the low flows from the water quality data and compared to water quality during high flows. Significant differences in water quality were noted for some monitoring stations and the pattern and magnitude of the statistically significant responses (t test, p < 0.05) varied among sites. PCA, was applied to the data sets of the two low and high flow periods, and resulted in three effective factors explaining 77.8 and 67.4 % of the total variance in surface water quality data sets of the two periods, respectively. The main factors obtained from PCA indicated that the parameters influencing surface water quality are mainly related to natural, point and non-point source pollution in the study area. DA provided an important data reduction as it used only three parameters, i.e. magnesium (Mg2+), calcium (Ca2+) and bicarbonate (HCO3−) affording 60 % correct assignations, to discriminate between the two low and high stream flow periods. General regression models revealed that surface water quality parameters were explained by low and high flow and specific discharge. The results of this study can be useful for water managers for effective surface water quality management under climate change.