Multivariate statistical techniques, such as cluster analysis (CA) and principal component analysis/factor analysis (PCA/FA), were used to investigate the temporal and spatial variations and to interpret large and complex water quality data sets collected from the Kaduna River. Kaduna River is the main tributary of Niger River in Nigeria and represents the common situation of most natural rivers including spatial patterns of pollutants. The water samples were collected monthly for 5 years (2008-2012) from eight sampling stations located along the river. In all samples, 17 parameters of water quality were determined: total dissolved solids (TDS), pH, Thard, dissolved oxygen (DO), 5-day biochemical oxygen demand (BOD5), chemical oxygen demand (COD), NH4-N, Cl, SO4, Ca, Mg, total coliform (TColi), turbidity, electrical conductivity (EC), HCO3 (-), NO3 (-), and temperature (T). Hierarchical CA grouped 12 months into two seasons (dry and wet seasons) and classified eight sampling stations into two groups (low- and high-pollution regions) based on seasonal differences and different levels of pollution, respectively. PCA/FA for each group formed by CA helped to identify spatiotemporal dynamics of water quality in Kaduna River. CA illustrated that water quality progressively deteriorated from headwater to downstream areas. The results of PCA/FA determined that 78.7 % of the total variance in low pollution region was explained by five factor, that is, natural and organic, mineral, microbial, organic, and nutrient, and 87.6 % of total variance in high pollution region was explained by six factors, that is, microbial, organic, mineral, natural, nutrient, and organic. Varifactors obtained from FA indicated that the parameters responsible for water quality variations are resulted from agricultural runoff, natural pollution, domestic, municipal, and industrial wastewater. Mann-Whitney U test results revealed that TDS, pH, DO, T, EC, TColi, turbidity, total hardness (THard), Mg, Ca, NO3 (-), COD, and BOD were identified as significant variables affecting temporal variation in river water, and TDS, EC, and TColi were identified as significant variables affecting spatial variation. In addition, box-whisker plots facilitated and supported multivariate analysis results. This study illustrates the usefulness of multivariate statistical techniques for classification and processing of large and complex data sets of water quality parameters, identification of latent pollution factors/sources and their spatial-temporal variations, and determination of the corresponding significant parameters in river water quality.
