Water quality assessment of Melen river using statistical and artificial intelligence methods

In this study, statistical analysis and artificial intelligence methods were employed to 26 physical and chemical pollution data obtained five monitoring stations on Big Melen River and its tributaries during the period 1995?2006 by State Hydraulic Works. Descriptive statistics such as mean, median, mode, variance, standard deviation, standard error were determined each data set. Water quality data were divided two part as high?low flow period and the periods were determined to investigate the high?low flow periods, rainy seasons and flow during 11 years. The PCA/FA and SOM-ANN was employed to evaluate the high?low flow periods correlations of water quality parameters, while the PCA and SOM techniques was used to extract the parameters that are most important in assessing high?low flow periods variations of river water quality. Factors/groups explained the pollution sources were identified as responsible for data structure at each data sets. So factors/groups are conditionally named mineral structure, soil structure and erosion, domestic, municipal and industrial effluents, agricultural activities (fertilizer, irrigation water), livestock wastes, waste disposal site and seasonal effects factors. PCA/FA and AOM were supported with MLR and ANN respectively, to determine the most important parameter in each factors/groups. APCS-MLR model were used for source apportionment and estimation of contributions from identified sources to the concentration of each parameter. APCS-MLR results was evaluation with fuzzy logic application to obtain comprehensible results for source apportionment and it was determined that which pollution sources affect the which parameters on which rate. The aim of this study is illustration the usefulness of multivariate statistical analysis and artificial intelligence for evaluation of complex data sets, in Melen River water quality assessment identification of factors/groups and pollution sources, for effective water quality management. It is thought that this study would suck advantage out of basin administrator, inspector and academic corporation with regard to evaluated and iterpreted especially continous and momentory data in basin monitoring studies.