River Hoogly, considered as an of import distributary of river Ganga, has been affected by indiscriminate discharging of contaminated and untreated sewage-sludge and industrial waste into waterways. A Water quality index has been developed utilizing five H2O quality parametric quantities like: dissolved O, biochemical O demand, pH, entire coliform and faecal coliforms at eight different Stationss along the river of Hoogly from 2002 to 2007. In this present survey, the WQI was calculated by both DELPHI and CCME procedure. Therefore these two methods reflect the quality of the H2O measured with regard to its pollution degree. The relationships among the Stationss are highlighted by bunch analysis to qualify the WQI. The survey represents a computer-simulated unreal nervous web ( ANN ) theoretical account for rating the relationship between the different parametric quantities of H2O quality sample collected at different Stationss along Hooghly river responsible for Water Quality measuring. Finally, both the H2O quality method ( CCME and DELPHI ) were statistically compared by the coefficient of finding ( R2 ) , root mean square mistake ( RMSE ) and absolute mean divergence ( AAD ) based on the proof informations set.
Keywords: ANN Model, CCME Method, Cluster Analysis, DELPHI Process, River Hoogly, Water Quality Index
The index of Water quality ( Abbasi, 1998 ; Coulston and Mrak, 1977 ) is a numerical index of physical, chemical, biological or radiological status of H2O beginnings and finding its quality before usage for assorted intents such as imbibing H2O, agricultural, aquatic life, recreational and industrial H2O etc. ( Carpenter et al. , 1998 ; Jarvie et al. , 1998 ; Sargaonkar and Deshpande, 2003 ) . The environment we live in is polluted greatly by assorted biological activities. If the equilibrium among the activities of different organisms- works, carnal and microorganism is earnestly disturbed, new equilibrium have to be attained sooner that may non be favourable to human or the works or animate being. Water pollution is the taint of H2O organic structures from chemical, particulate or bacterial affair that affects the H2O ‘s quality degree.
Sewage, agricultural and industrial waste discharges are the chief ground of H2O pollution. The River Hooghly, having the assorted domestic sewerage and industrial treated wastewaters and untreated wastewaters from the metropoliss on the Bankss is now the largely contaminated rivers of universe ; the H2O is even unhygienic for bathing and for imbibing intents. The contaminated H2O organic structures may hold unwanted coloring materials, smell, gustatory sensation, turbidness, organic affair contents, harmful chemical contents, toxic and heavy metals, pesticides, oily affairs, industrial waste merchandises, radiation, high Total Dissolved Solids ( TDS ) , acids, bases, domestic sewerage content, virus, bacteriums, Protozoa, rotifers, worms, etc. ( Carpenter et al. , 1998 ; Jarvie et al. , 1998 ) . The contaminated imbibing H2O may besides do human wellness hazard such as tumours, ulcers, tegument upsets ( Michael Hogan, 2010 ) .
However, all available H2O beginnings are non suited for all different intents. Water quality index ( WQI ) has been build to measure the suitableness of H2O for a assortment of utilizations. A H2O quality index is a individual figure quantitative look that provides overall H2O quality at a certain location and clip based on several H2O quality parametric quantities ( Bordalo and Wiebe, 2006 ; Boyacioglu, 2007 ; Brown et al. , 1970 ; Landwehr and Deininger, 1976 ; Pesce and Wunderlin, 2006 ; Sanchez et al. , 2007 ; Sargaonkar and Deshpande, 2003 ; Singh and Anandh, 1996 ) . Different techniques have been used in effort to alter complex H2O quality informations into simpler information that is easy to understand and available by the populace ( Bennetts et al. , 2006 ; Pulido-Leboeuf et al. , 2003 ) . This H2O quality index is a dimensionless values that ranking between 0 and 100 ( Parmar and Parmar, 2010 ) . A good H2O quality is represented by a higher index value of H2O quality ( Cude, 2001I? Pandey and Sundaram, 2002 ) . However, a H2O index based on some really of import parametric quantities that can provides a simple measuring of H2O quality. It gives the populace a general thought the possible jobs with the H2O in the part. There are several H2O quality parametric quantities to include in the index ( APHA, AWWA, WPCF, 1976 ; AWWA 1971 ; Hooda and Kaur, 1999 ; Kannel et al. , 2007 ; Karia and Christian, 2001 ; Metcalf and Eddy, 1992 ; Ramalho, 1983 ) . These parametric quantities are:
dissolved O ( DO )
biochemical O demand ( BOD )
2. Materials and methods
2.1. Analyze sites and collected informations
In order to find the H2O quality index, H2O samples were typically collected on a monthly footing across the river breadth of Hooghly at all the eight sites ( Berhampore, Palta, Srirampore, Howrah ( Shibpur ) , Garden Reach, Dakhshineswar, Uluberia, Dimond Harbour ) during the survey period ( 2002-2008 ) . Among 19 entire H2O quality parametric quantities, 5 of them were selected for analysis the H2O quality measuring. The five selected parametric quantities were pH, dissolved O, biochemical O demand, entire coliform, faecal coliform.
Several methods have been introduced in the yesteryear to bring forth a suited WQI method. Each method had some certain advantages and some disadvantages besides.
2.2. Calculation of CCME Water Quality Index ( CCME 2001 )
Each H2O quality index ( index ) was calculated utilizing methods developed by Canadian Council of Ministers of the Environment ( CCME, 2001 ) based on three different measurings of H2O quality: SCOPE ( F1 ) , FREQUENCY ( F2 ) , and AMPLITUDE ( F3 ) consequences as, ( Eq. 1 ) .
WQI ( index ) = ( 1 )
For each index, the rating graduated table followed the “ ranking ” graduated table is used five classs or degrees that corresponded to specific degrees of H2O quality which is shown in Table 1.
Table 1: Rating Scale used for the H2O quality index
Where, F1 ( range ) describes the extent of quality guideline non conformity over the clip period of involvement and it was calculated as
( 2 )
where the failed variables indicate the H2O quality variables with aims which are tested during the clip period for the index computation.
F2 ( frequence ) represents the per centum of single trials that do non transcend failed trials.
( 3 )
F3 ( amplitude ) represents the value by which the failed trial values do non run into their aims, and it was calculated in three stairss as:
When the trial value must non transcend the nonsubjective and the aim is termed an ‘excursion ‘ , so it expressed as follows:
( 4 )
For instances where trial value should non fall below the aim:
( 5 )
The corporate sum by which single trials are out of conformity is calculated by summing the jaunts of single trials from their aims and dividing by the entire figure of trials. This variable referred to as the normalized amount of jaunts or NSE calculated as
NSE = ( 6 )
F3 is so calculated by an asymptotic map that ranges the normalized amount of jaunts from aims ( NSE ) to give a scope between 0 and 100.
( 7 )
Once the CCME WQI value has been determined, H2O quality can be categorized by matching it to one of the undermentioned degree.
2.3. Calculation of H2O quality index with DELPHI procedure
The systematic technique was attempted to integrate the judgements of a big diverse system in H2O quality direction procedure ( Alexander, 1999 ; Saha et al. , 2007 ; Walski and Parker, 1974 ) . Two basic attacks are followed by the research workers: aggregate method and multiplicative method. An overall quality evaluation is derived by multiplying the concluding weights ( Wisconsin ) of each single parametric quantity with the corresponding quality evaluation ( qi ) , the amount of which gives the needed individual figure WQI. The quality evaluation is measured on a graduated table of 0 to 100 point ( i.e. , highest to lowest polluting ) .
Method 1: Aggregate Method ( Saha et al. , 2007 )
The WQI considered is of the signifier
( 8 )
where WQIa is the aggregate H2O quality index between 0 and 100, qi the quality of ith parametric quantity between 0 and 100, wi the weight of ith parametric quantity ( between 0 and 1 ) , and N is the entire figure of parametric quantities.
In this type of index, if any significantly relevant parametric quantity exceeds the allowable bound, the mean weighted indices does non see sufficient lowering of the H2O quality index. Table 2 is used to depict the high Indicator values corresponded to low degrees of taint ( i.e. , good H2O quality ) and low values indicated high degrees of taint ( i.e. , hapless H2O quality ) .
Table 2. Categorization of H2O quality based on WQI utilizing agrregative method
Good to Excellent
Good to Moderate
Bad to really Bad
Method 2: Multiplicative Method ( Saha et al. , 2007 )
Multiplicative signifier of index may be considered by
( 9 )
In this index, weights are calculated to the person parametric quantities based on a subjective sentiment. The categorization is shown in Table 3.
Table 3. Categorization of H2O quality based on WQI utilizing Multiplicative method
3. Consequences and Discussions
3.1. Bunch analysis
In order to avoid univariate statistical analysis job, multivariate analysis such as Cluster analysis is used in the survey to depict the correlativity amongst a big figure of meaningful informations without losing much information ( Jackson, 1991 ; Meglen, 1992 ) . Cluster analysis is a technique to sort groups of objects, or bunchs, in such a manner that the ensuing groups are similar to each other but distinguishable from other groups ( Helena et al. , 2000 ; Raghunath et al. , 2002 ; Simeonov et al. , 2003a ; Simeonova et al. , 2003b ; Simeonov et al. , 2004 ; Singh et al. , 2004 ; Vega et al. , 1998 ; Voncina et al. , 2002 ) . Cluster analysis can be performed on many different types of informations sets. Hierarchical bunch is a manner to look into grouping in informations, at the same time over a assortment of graduated tables, by making a bunch tree. The tree is non a individual set of bunchs, but instead a multilevel hierarchy, where bunchs at one degree are joined as bunchs at the following higher degree. Hierarchical agglomerate Cluster analysis was performed on the normalized informations set by agencies of the Ward ‘s method for sample categorization utilizing squared Euclidean distances as a step of close propinquity ( Einax et al. , 1997 ; Fovell, 1993 ) . Dendrogram has been developed utilizing the Matlab7 ( The Mathworks, Inc. ver. 7.0.1 ) with H2O quality index informations set of Hooghly river to happen out the similar sampling sites spread over the river stretch.
3.2. ANN patterning
In this present survey, different nervous web theoretical accounts and algorithm were tested and optimized to obtain the best theoretical account construction for the anticipation of H2O quality index of trying Stationss along the river Hooghly. Based on the rules of the feedforward backpropagation algorithm, the mold method has been developed ( Rumelhart et al. , 1986 ) . The ANN theoretical account was constructed on illustrations of deliberate datasets with known end products to analyse bing procedures. The ANN architecture typically comprises three types of nerve cell beds: an input bed ( independent variables ) , one or more figure of concealed beds and an end product bed ( dependent variables ) . The input bed, which merely connect one input value with its associated leaden values receives information from external beginnings and reassign this information to the concealed bed for processing ( Ozdemir et al. , 2011 ) . The net input for each nerve cell ( aj ) is the amount of all input values Xi ; each multiplied by its weight Wji, and added a bias term Zj which may be formulated as:
( 10 )
The end product value ( tj ) can be generated by treating all the information of concealed bed and net input neuron into the additive transportation map ( purelin ) of the nerve cell:
( 11 )
In this present survey, two types of transportation map have been applied: a tan-sigmoid transportation map ( tansig ) at hidden bed and a additive transportation map ( purelin ) at end product bed. The Levenberg-Marquardt back-propagation algorithm was used for web preparation. The inputs and end product parametric quantities to the ANN theoretical account were indistinguishable to the factors considered in bunch analysis attack, viz. pH, dissolved O, biochemical O demand, entire coliform, fecal coliform and Water quality index severally. All nervous web computations were implemented utilizing Neural Network Toolbox of MATLAB Version 7.0.1.
Cluster analysis was performed to place the spacial similarity for bunch of H2O quality index of trying sites under the monitoring web. It represented a dendrogram ( Fig. 1 and Fig. 2 ) by utilizing two different methods, grouping all the eight trying Stationss based on the H2O quality index of Hooghly River.
Fig. 1. Dendrogram based on agglomerate hierarchal constellating utilizing CCME method
Fig. 2. Dendrogram based on agglomerate hierarchal constellating utilizing DELPHI method
From the consequence it was observed that in CCME method, the bunch process generated two statistically important groups harmonizing to the norm calculated H2O quality index from the twelvemonth 2002-2008. Bunch 1 ( sites 2, 3, 4, 5, 7 ) and another bunch 2 ( site 6 is distantly related to sites 1 and 8 ) can be classified correspond to a comparatively moderate pollution, low pollution, really high pollution Stationss severally. From the DELPHI technique it was apparent that all eight Stationss on the river can be grouped into three major important bunchs with similar characteristic characteristics, bunch 1: Srirampore, Howrah ( Shibpur ) , Garden Reach, Dakhshineswar ( the scope of WQI is between 24 to 28.32 ) bunch 2: Palta and Uluberia ( WQI are 36.67 and 35.97 ) and cluster 3: Berhampore and Diamond Harbour ( WQI are 62.25 and 62.77 ) as presented in Fig. 2. It was clearly found that the 3rd major constellating group ( high significance of constellating ) was characterized by the highest Euclidian linkage distance than the other two constellating group. CA technique is utile in dependable categorization of H2O quality index in the whole part across the river basin and will do possible to plan a future spacial trying scheme in an optimum mode. Therefore, the figure of trying sites in the evaluating web will be reduced.
3.3. Development of ANN theoretical account
Machine larning techniques such as Artificial Neural Networks ( ANNs ) has increased late as a powerful tool in simulation of informations modeling and could be utile in ecological facets ( Moghaddam and Khajeh, 2011 ; Recknagel, 2001 ; Sinha et al. , 2012 ) . It is necessary to bring forth and optimise the ANNs for anticipation the best theoretical account constellation that gives lower mistake during developing with minimum calculating clip. In the present survey, an ANN based theoretical account was besides developed for depicting the Average calculated H2O quality index of all the eight trying Stationss along the river basin of Hooghly utilizing both CCME and DELPHI technique from the twelvemonth 2002-2008. All analyses were based on the deliberate information set. The preparation process is repeated until the mistakes become little plenty and the value of correlativity coefficient ( R ) between the theoretical account anticipation and experimental consequences is reached to 1. After the preparation, the ANN can be validated utilizing independent informations ( Tokar and Johnson, 1999 ) . The goodness of tantrum of the trained web was shown in Fig. 3 and Fig 4. Arrested development secret plan in Fig 4 has correlativity coefficient of 0.987 utilizing DELPHI method and correlativity coefficient of Fig 3 is 0.954 utilizing CCME technique.
Fig. 3. Arrested development secret plan on WQI ( Experimentally vs. Predicted ) utilizing CCME Method with five input variables, ten treating elements in concealed bed, and one end product variable
Fig. 4. Arrested development secret plan on WQI ( Experimentally vs. Predicted ) utilizing DELPHI Method with five input variables, ten treating elements in concealed bed, and one end product variable
The public presentation of the constructed DELPHI method and CCME method were besides statistically analysed by the root mean squared mistake ( RMSE ) , coefficient of finding ( R2 ) and absolute mean divergence ( AAD ) as follows ( Geyikci et al. , 2012 ) :
aˆ¦aˆ¦aˆ¦aˆ¦aˆ¦ ( 12 )
aˆ¦aˆ¦aˆ¦aˆ¦aˆ¦.. ( 13 )
aˆ¦aˆ¦aˆ¦aˆ¦aˆ¦aˆ¦.. ( 14 )
where N is the figure of informations points, is the predicted value from ANN consequences, is the existent WQI value calculated by CCME and DELPHI method, and the symbol ‘- ‘ is the norm of the related values. Table 4 represents the statistical comparing between CCME and DELPHI technique. In this present survey both CCME and DELPHI methods provided good findings of H2O quality of Hooghly river, yet the DELPHI method showed the clear high quality over CCME method for both informations adjustment by ANN theoretical account development and appraisal capablenesss.
Table 4. Comparison of DELPHI and CCME method for finding of WQI
AAD ( % )
Therefore it would be more rational and dependable to cipher the one-year H2O quality index utilizing five different parametric quantities such as pH, dissolved O, biochemical O demand, entire coliform, fecal coliform at 8 different sampling Stationss through a procedure of DELPHI.
From the present instance survey of H2O quality of assorted Stationss along the river Hooghly, it is noticed that H2O of Hooghly is slightly contaminated with the pollutants having from assorted industries and domestic beginnings. DELPHI and CCME both methods were applied to cipher the mean informations of every month of a twelvemonth. The Hierarchical bunch analysis and developed ANN theoretical account was applied to the Hooghly river basin to mensurate WQI, which has really hapless H2O quality. Overall H2O quality ranges shows from hapless to fringy quality depending on the river range and sample twelvemonth. Hierarchical bunch analysis grouped 8 trying Stationss into three major bunchs of similar features reflecting the H2O quality index calculated by DELPHI method. The WQI was formulated by both DELPHI and CCME technique and the root mean square mistake ( RMSE ) , coefficient of finding ( R2 ) and absolute mean divergence ( ADD ) were used together to compare the H2O quality public presentation of the CCME and DELPHI methods. The DELPHI method was found to hold higher predictive capableness than CCME method.