Studies on thewater quality of rivers are extremely important, particularly because riversare one of the main sources of water supply for potable, agricultural, andindustrial usage.
Unfortunately, river pollution has become one of the mostimportant environmental problems (Fanet al., 2009, Tsihrintzis, 2013, Parsaie and Haghiabi, 2017). Monitoring the parameters of waterquality of rivers allows better management of water quality. In addition, thiswill lead to an improvement in the public healthlevel; therefore, continuous surveying of the water quality of rivers is ofhigh importance (Ishaket al.
, 2012). Water quality indices areparameters that are related to the biological, physical, and chemicalproperties of water. Usually, water quality is determined by measuring itsbiochemical oxygen demand (BOD), chemical oxygen demand (COD) and declined dissolvedoxygen (DO) level. The dissolved oxygen level is a measure of the health of theaquatic system. A certain minimum level of DO in water is required for theaquatic life to survive (Basant et al., 2010). The sources of DO in a water body include re-aeration from the atmosphere, photosynthetic oxygen productionand DO loading. The sinks include oxidation of carbonaceous and nitrogenousmaterial, sediment oxygen demand, andrespiration by aquatic plants (Kuo et al.
, 2007). The chemical oxygen demand is used as a measure of the oxygenequivalent of the organic matter content of a sample that is susceptible tooxidation by a strong chemical oxidant. The COD is used to measure the totalquantity of oxygen-consuming substancesin the complete chemical breakdown of organic substances in water. It is animportant parameter in measuring quality and determining what organic load ispresent in the water (Verma and Singh, 2013). The Biochemical oxygen demand is an approximate measure of theamount of biochemical degradable organic matter present in a water sample.
Itis defined by the amount of oxygen required for the aerobic microorganismspresent in the sample to oxidize the organic matter to a stable organic form (Chapman, 1996). Excessive BOD loads damagethe quality of river water. It causes low DO (dissolved oxygen) concentrationand unsuitable living conditions forflora and fauna in the river. At the sametime, BOD–DO relationships include an exchangewith the river bed and nitrification and denitrification(Radwan et al., 2003). Nutrients and light in the phytoplankton growth, the relationshipbetween DO and phytoplankton concentrations and ammonia affect the BODdegradation (Lopes et al.
, 2005). Dissolved oxygen levels, water temperature, water flow,chlorophyll a and nutrient levels (ammonia, nitrite, nitrate) are among themost critical factors for biochemical oxygen demand (BOD) in the rivers. Theoxygen consumption from degradation of organic material is normally measured asBOD and COD, so there is an important relationshipbetween them. Performing the test for BOD requires significant time andcommitment for preparation and analysis.
This process requires 5 days, withdata collection and evaluation occurring on the last day. A test is used tomeasure the amount of oxygen consumed by these organisms during a specifiedperiod of time (usually 5 days at 20 C). The difference in initial DO readings(prior to incubation) and final DO readings (after 5 days of incubation) isused to determine the initial BOD concentration of the sample. This is referredto as a BOD5 measurement (Dogan et al., 2009). Several water quality models such as traditional mechanisticapproaches have been developed in order to manage the best practices forconserving the quality of water. Most of these models need several differentinput data which are not easily accessible and make it a very expensive and time-consuming process (Suen and Eheart, 2003).
In recentyears, several studies have beenconducted on water quality forecast models (Kurunç et al., 2005, Li, 2006,Goyal et al., 2013).
However, since a large number of factors affecting the water quality have acomplicated non-linear relation with the variables; traditional data processingmethods are no longer good enough for solving the problem (Nasr et al., 2012, Mokarram, 2015). Onthe other hand, the artificial neural networks (ANNs) capable of imitating thebasic characteristics of the human brain such as self- adaptability, self-organization, and error tolerant and havebeen widely adopted for model identification, analysis and forecast, systemrecognition and design optimization (Diamantopoulou et al., 2005, Niu etal., 2006, Sarkar and Pandey, 2015).Unlikemany statistically based water quality models, which assume a linearrelationship between response and prediction variables and their normaldistribution, ANNs are able to map the non-linear relationships that arecharacteristics of aquatic eco-systems (Lek and Guégan, 1999).
During last about two decades, ANNs have undergone an explosivedevelopment in application in almost all the areas of research (Lerner et al., 1994, Kung and Taur,1995, Raman and Chandramouli, 1996, Chu and Bose, 1998, Li, 2006, Ciampi andLechevallier, 2007, Messikh et al., 2007, Hanbay et al.
, 2008, Dürrenmatt andGujer, 2012, Abyaneh, 2014).The ANN approach has several advantages over traditional phenomenological orsemi-empirical models since they requireknown input data set without any assumptions (Gardner and Dorling, 1998). The ANN develops a mapping of the input and output variables,which can subsequently be used to predict desired output as a function ofsuitable inputs (Friedman and Kandel, 1999). A multi-layer neural network can approximate any smooth,measurable function between input and output vectors by selecting a suitableset of connecting weights and transfer functions (Gardner and Dorling, 1998).ANN models have been widely applied to the water quality problems (Rogers and Dowla, 1994, Wen andLee, 1998, Lek and Guégan, 1999, Bowers and Shedrow, 2000, Cancelliere et al.,2002, Kuo et al., 2004).
Recently, by developing soft computingtechniques in most areas of water engineering as a powerful tool for modeling(Azamathulla et al., 2008) researchers have attempted touse artificial intelligence for modeling water qualityproblems (Nooriet al., 2010, Abyaneh, 2014, Emamgholizadeh et al., 2014, Noori et al., 2015,Dehdar-Behbahani and Parsaie, 2016, Parsaie and Haghiabi, 2017).
The presentstudy uses Mg, Ca, Na, Cl, SO4, HCO3, CO3, TDS, EC, K and SAR as the input andconsidered DO, BOD and COD as the output to investigate the quality of Karunwater in the vicinity of Ahvaz city. Thenovelties of this study in comparison with the previous ones are a) application of newer data, b)sensitivity analysis of parameters by Gamma test before simulation by the neural network and c) application of more inputdata for construction of neural network architecture.