A Novel Water Quality Index for Iraqi Surface Water

a


Introduction
When using WQIs to assess water quality monitoring data, results can be significantly interpreted, especially when pollutant concentrations are below the water quality criteria.In general, WQI can totally disregard the significance of sampling frequency in assessing water quality 1 .WQIs enable administrative decision-makers to evaluate the efficacy of regulatory programs and present information on water quality to the audience in an understandable and straightforward manner.They also assist professionals in separating monitoring data into a larger framework 2,3 .Indices were used for almost all monitoring programs, including environmental planning, water quality monitoring, assessment, treatment, and public awareness 4 .
The establishment of a scientific approach for selecting a numerical index for identifying chemical water contamination was encouraged, according to a panel of the president's science consultative

Abstract
The study aims to build a water quality index that fits the Iraqi aquatic systems and reflects the environmental reality of Iraqi water.The developed Iraqi Water Quality Index (IQWQI) includes physical and chemical components.To build the IQWQI, Delphi method was used to communicate with local and global experts in water quality indices for their opinion regarding the best and most important parameter we can use in building the index and the established weight of each parameter.From the data obtained in this study, 70% were used for building the model and 30% for evaluating the model.Multiple scenarios were applied to the model inputs to study the effects of increasing parameters.The model was built 4 by 4 until it reached 17 parameters for 10 sampling times.Obviously, with the increasing number of parameters, the value of the index will change.To minimize the effect of eclipse that arises in WQI and to solve the problem of overlapping quality and pollution, this study has created another index linked with IQWQI, which included both the quality and the degree of pollution.The second index is called the Environmental Risk Index (ERI), where only the variables that exceed the permissible environmental limits were included.Sensitivity Analysis was done to predicate IQWQI and to determine the most influential parameters in the IQWQI score; two types of models were chosen for the run of the sensitivity test, which are the Artificial Neural Network Regression (ANNR) and Backward Linear Regression (BLR).The results of IWOI and ERI for freshwater use during the dry season were very poor water quality with a high degree of risk.While in the wet season, both indices' values ranged from poor water quality to very poor water quality with a high degree of risk.Baghdad Science Journal committee on environmental pollution 5 .The panel stated that different chemical contaminants must be detected by the method used.Its outcome is essentially proportionate to the harmful consequences that water pollution has on people or aquatic life.The index enabled many changes in water quality that followed.In response to this claim, Horton released the first water quality indicator (WQI) the same year 6 .Since then, WQIs have developed into a common and useful tool for evaluating the water quality of various water bodies all over the world [7][8][9][10] .Following Horton, Brown et al. 11 developed a WQI with a structure that is comparable to Horton's index 6 .Still, with much rigidity in selecting parameters, the National Sanitation Foundation (NSF) provided funding for the research conducted by Brown et al. 11 .Because of this, Brown's index is sometimes known as NSFWQI.
A WQI 12 was developed in 2020 by a team of Iraqi experts to assess the suitability of rivers for drinking.Using the Delphi method, a survey of 44 water quality management experts asked them to select and rate only 10 from 27 water quality parameters.According to the panel's recommendation, only six parameters were chosen for the index: TDS, COD, DO, Total Hardness, TC and Cl, and based on their opinions, weights were given for each parameter.The subindex for each parameter was taken by the Due to the depletion of water supplies, expansion of agriculture, an increase in drainage, and high temperatures, the quality of the water is declining toward the middle and southern regions of Iraq.As a result, there are more salts and pollutants in this water, which is seen in the areas' drinking water quality 13 .In light of these factors, it is essential to regularly evaluate the river's water quality in order to determine its suitability for various uses and to detect pollution as soon as possible so that the appropriate authorities can take the necessary action 14 .
Because of the absence of water quality models that mimic the environmental reality of Iraqi water, this study aims to develop a water quality index that fits the Iraqi aquatic system consisting of physical and chemical factors.

Sites Description Climate
The Iraqi climate is arid to semi-arid, with dry, hot summer and cold winter.Moreover, it has low humidity and low precipitation 7 , and the mean annual rainfall is about 11.02 mm.Climate elements affect the hydrological characteristics of the river, as temperature affects the amount of evaporation.Temperature increases in summer, which leads to the evaporation of water and an increase in salinity in the surface water.The rise in water temperature affects aquatic organisms by, for example, decreasing oxygen, accelerating the organic dissolution of polluted organic materials, and increasing the toxicity of some chemical pollutants 15 .In addition, Intergovernmental Panel on Climate Change (IPCC) has identified Iraq as highly vulnerable to climate change 16 .According to the two main seasons (Wet and Dry) in Iraq are based on the relative humidity RH% Table 1, in which above 50 RH% is considered a wet season, while less than 50 RH% is considered a dry season 14 .

Study site
Five sites were chosen for conducting the study along the Tigris River within Baghdad City during 2020-2021, starting from Al-Muthana Bridge (north of Baghdad) and ending before the confluence between the Tigris and Diyala Rivers to the south of Baghdad City (Fig. 1), Table 2, represents the Global Positioning System for the sites.The first site (Al-Muthanna Bridge) is located at the entrance of the Tigris River into Baghdad city, this site represents the northern part of the Tigris River, a natural area influenced mainly by fisheries and agricultural activity and didn't have industrial activities.The second site (Al-Greaat Area) is located under a floating Bridge for pedestrian crossing, this site is about 7.99 km away from the first site, the area's nature is agricultural and rich, with palm groves and submerged plants on both edges and people visit this place to relax and go to restaurants, therefore, a lot of food scraps and plastic waste can be found near the river in this site.Site three (Al-Sarrafia Bridge) has a lot of human activity like restaurants, fisheries, residential buildings, etc, the distance between this site and the second site is about 7.52 km and it is located in the middle of Baghdad city.The fourth site (Al-Jadriyah Bridge) is predominantly urban with little agricultural activity on the campus of the University of Baghdad, the western part of this area has been converted into an artificial pool (Al-Jadriyah Lake for tourism), where water is pumped from the Tigris River into this lake, the distance between the third and fourth site is about 7.99 km.
The Fifth site (Al Za'franiya Area) is located southeast of Baghdad, before the mouth of the Diyala River, this site is influenced by many industrial activities which are located on the bank of the river, part of which belongs to the government sector and other parts to the private sector, like the vegetable oil plant under the Al-Dora Bridge and Al-Rasheed Power Station south of Baghdad (gas and thermal station) and various sources of water are brought to the river from these sectors, today this site is crowded with population due to urban development and an increase in municipal services.

Water Samples
Three water samples were taken from each site: one from each bank of the River and one from the middle.The average sampling time was between 7:00 AM to 6:30 PM.Each sample was collected from the subsurface (about 20-30 cm below the surface) in clean stopper-fitted polyethylene bottles.Before filling the bottles with the required sample, they were rinsed in river water several times.The samples were preserved in an ice-cool box until they were taken to the laboratory and subjected to physical and chemical analyses.Laboratory measurements were conducted 24 hours after sampling at the Environmental Research Center-University of Technology-Iraq.Field and laboratory measurements represented in Table 3 were carried out according to APHA 17 .

Iraqi WQI (IQWQI) Model Development Questionnaires
Delphi method was used to determine the final weight.Delphi technique can be defined as a communication method aimed at forming standards and guidelines and predicting trends 18 .A typical step was followed when using the Delphi method started with: A-In developing the initial Delphi questionnaire, 55 parameters were selected for 4 water usages (freshwater, aquatic life protection, agriculture, and raw drinking water) to prepare the questionnaire, including the parameters plus reasons and justifications for including them in the WQI.In the questionnaire, the respondents were asked to choose the most important parameters from their point of view and experience to evaluate the uses referred to above, giving a weight value for each parameter (from 1 to 5) (unconditional Integers), where the weight value (1) represents the least important and the weight value ( 5) is the most important (Supplement 1).B-Selecting the expert panel; 76 experts from academics and engineers with expertise in water quality management, starting with experts with Assistant Professor titles and above.C-Distributing the questionnaire; it was sent to the experts to collect the information and their opinions; the questionnaire will help identify the most appropriate parameters used to develop the Iraqi Water Quality Indices and assign a weight for each parameter.D-Collecting and analyzing the questionnaire, from the 76-expert panel, 32 responded, 4 refused to participate, and 40 did not respond.Eight respondents have been excluded from the 32 respondents due to a lack of information.

Parameters selection
Based on the purpose of the water uses, the parameters were chosen for the freshwater purposes for IQWQI, and the value of the standard for each parameter is shown in Table 3.

Weight Assignment
Parameter weighting help to assign relative importance to each parameter and illustrate interrelations between different parameters 22 .Based on the expert opinion, each parameter was assigned a weight (AW) from 1-5, and the main values of the weight were used.Then the temporary weight (tW) was calculated where a temporary weight of 5 was assigned to the parameter which gained the highest rating.All other temporary weights of the parameters were obtained by dividing the highest significance rating by the individual mean rating.Each temporary weight was then divided by the sum of all the temporary weights to arrive at the final weight, as shown in the following equation (Eq.1).

Sub-indices Formation and Aggregation of Functions
After assigning weights, index aggregation is performed to obtain the final index score.Aggregation occurs in sequential stages where the index aggregates sub-indices.The sub-index (SI) is determined for each parameter (Eq.2), and the quality rating is calculated as in Eqs. 3 and 4. The additive (arithmetic) method reached the final index (Eq.5).SIi= the sub-index of ith parameter; Qi= quality rating based on the concentration of ith parameter; Ci= is the observed value of the nth parameter; Si= is the standard value of the nth parameter; Cideal for DO= 14.6; Cideal for pH=7; Wi= final wight.=∑   /……….Eq. 5

Water Quality Rating
According to Tyagi et al. 23 , the best rating compatible with the Weighted Arithmetic Water Quality Index model is the NSFWQI model.It is given in Table 4, where WQI = 0 is the best value and WQI > 100 is unsuitable for use where the subindices  are not restricted to the range 0 -100.Consequently, it is possible that WQI > 100.Environmental Risk Index WQI, raises the problem of the eclipse, which is a term used to describe how the final WQI score hides the effects of the parameters that exceed the allowed levels and eventually masks the true nature of WQ, this situation occurs while applying the mathematical formula 24 , where lowly weighted subindices may be dominated by highly weighted subindices, or vice versa, putting the overall water quality rating in a questionable situation.Some researchers mentioned the eclipsing problem 22 .Ott 25 was the first author that pointed to eclipsing and described it as "poor environmental quality exists for at least one pollutant variable, but the overall index does not reflect this" the problems of eclipsing worsen as the number of parameters increase.Swamee and Tyagi 26 and Smith 27 referred to the eclipsing problem as "the index score hides the parameter responsible for limiting that water's suitability for the particular use and the degree by which it does this".The eclipsing can occur by one of the following; (i) inappropriate sub-indexing, (ii) parameter weightings that do not accurately reflect the relative importance of the parameters (iii) aggregation functions that are not appropriate 28 .
Example of eclipsing: in 4 virtual environmental parameters result, the observed value of one of them is beyond the permissible limit Table 5.The final index score might indicate good water quality, even though one of the parameters does not meet its permissible limit, so the parameter failure is hidden or eclipsed by the aggregation function.So, it was thought that there would be another index linked with the water quality index to be included in its calculation called Environmental Risk Index (ERI), where only the variables that exceed the permissible environmental limits are included.
The calculation depends on the concentration of each parameter that exceeded the permissible limit (Eqs.6.1, 6.2.).
Ci= is the observed value of the nth parameter Si= is the standard value of the nth parameter Based on the degree of contamination categories mentioned in 29 , the ERI was built with some modifications to be compatible with this study Table 6.The final calculation of the IQWQI was made by Microsoft Excel ver.19, where the fixed cell contains the component of the WQI; parameter, mean of respondents, temporary weights, final weight, observed value, standard value, and sub-index.All these cells are linked with the final WQI equation to generate the final score.

Canadian Council of Ministers of the Environment Water quality index (CCMEWQI)
The Canadian WQI was calculated in this study to be compared with the results of IQWQI.The CCMEWQI is a mathematical approach for evaluating surface water for various purposes following specific criteria 30 .The index is computed by summing the three factors according to Eq. 7. As indicated in Table 7.The index is based on three factors.

Statistical Analyses
Jeffrey's Amazing Statistics Program (JASP) for statistical analysis based on R programming language was used to conduct the sensitivity analysis for IQWQI for freshwater use (17 parameters).The dataset used in the calculations consisted of 108 values for each parameter.The data is split into 70% for training the network and 30% for testing.

IQWQI Calculation Parameters selection
70% of the data obtained from this study was used for developing the model.Water quality parameters are chosen based on the most concerned and available standards Table 3.The parameters set are selected based on Iraqi and international water maintenance standards.The first set for building the IQWQI was for freshwater use.This set contains 17 parameters Table 3.These parameters were used to identify the overall health of the Tigris River.

Weight assignment
The Delphi process obtained the parameter weight values.The parameter weight values are estimated based on the relative importance of the water quality parameter and/or the appropriate water quality guidelines 24 .The subindices were calculated for each parameter (for four water uses based on the expert panel drift from the Delphi method, first of all, it must take the average rating returned by respondents and then transform each parameter to temporary weights by dividing the parameter with the highest rate by the other parameters Table 8, first red box, and the parameter with the highest rating is given a full rating value which is 5, then, to determine the final weight for each parameter included in the model each temporary weight is divided by the sum of all temporary weights of parameters example 1, individual parameter concentrations is transformed to the same scale.Weighting aims to assign relative importance to each parameter and elucidate interrelations between different parameters.To ensure that the final wights are correct, the sum of all final wights must be 1, as reported by 2 , where the majority of WQI models applied unequal weighting techniques where the sum of all of the parameter weight values was equal to 1 Table 8, last red box.

IQWQI and ERI test
Sutadian 31 reported that CCME could work using four parameters for four sampled times.From this fact, Multiple scenarios were applied to the model inputs to see the effect of the increasing number of parameters.The model started to be built from 4 by 4 until it reached 17 parameters for 10 sampling times.With the increasing number of parameters, the index's value will change, which appears whenever the number of parameters and the sampling time increase, as proven in the cases below Table 9.
Scenario 1: 4 parameters (DO, BOD5, pH, CN -) by 4 sampled times, the result of the IQWQI was 30.3, and the ERI was 3.57, indicating a good water quality with a low degree of risk in the same time the IQWQI was compared with CCME to confirm that the new model was compatible with others models.The CCME result was 81.84 (good water quality) and was calculated for the same parameters used for IQWQI.The result of both indices came in the same category.Scenario 2: 5 by 5 (DO, BOD5, pH, CN -, PO4 3-), the resulting rank of both the indices IQWQI and ERI are still the same (a good water quality with a low degree of risk) with the change in the values 32.98 and 6.2, respectively.The CCMEWQI rank was in the good category.
Scenario 3: 6 by 6 (DO, BOD5, pH, CN -, PO4 3-, Cr + ) with the increase of parameters and the sample times, the value of water quality started to change, where IQWQI was 35.95 as shown in Table 9, the water quality is still within the good category, but the effect of eclipsing starts to rise which couldn't be indicated with IQWQI only, here the value of using the ERI appears as its value was 12.65 because the effect of the Cr + where its concentration was way beyond the limits where its concentration reach to 0.33 mg/l while the limits were 0.05 mg/l 18 , where this index focus on the effect of only the parameters exceeded the permissible limits, and it could be said that the water quality is good, but there is a medium degree of risk.The CCMEWQI rank has a fair category.It could be said it's compatible with IQWQI IQWQI was 85.23, and the value of ERI 78.21 where the turbidity exceeds the limit once.

Sensitivity Analysis for IQWQI
This study used the sensitivity analysis based on Artificial Neural Network Regression (ANNR) and Backward Linear Regression (BLR) to determine which water quality parameter most influences the score of IQWQI.Sensitivity analysis studies an output parameter's response concerning input parameter variations 32 .A model performance R 2 , RMSE and SSE were used for model performance evaluation for both ANNR and BLR and to make a comparison with them to see which will give the more accurate results, where these three criteria significantly affect the fitness and residual measurement of the ANNR and BLR models in WQI prediction.The dataset used as input data (108 values for each parameter) was subjected to Standardized to ensure a fair representation of parameters in the value of IQWQI.17 parameters (from WQI calculation) were selected as input and IQWQI as the output for ANNR-IQWQI and BLR-IQWQI models.
The first model includes all parameters and represents the input parameters called IQWQI-Ref, which serve as a reference model for ANNR and BLR.To assess the significance of the input parameters of IQWQI-Ref, the sensitivity analysis for each model was done by excluding one parameter from the 17 parameters.The ANNR performance model was evaluated using R 2 , RMSE and SSE, as shown in Table 10 and Table 11.The results show that the water quality index predicted with the ANNR model brings better and more reliable output (R 2 =0.957,RMSE =0.265) compared with the BLR-IQWQI (R 2 =0.901,RMSE = 0.504).ANNR consists of three layers, input layer, hidden layer and output layer.There are layers and nodes at each layer.Each node at the input and inner layers receives input values (parameters values) which are then processed and passed to the next layer.This process is conducted by weights representing the connection strength between two nodes.The model is shown in Fig. 2, where the input layer consists of 17 parameters and the hidden layer consists of 10 nodes.Output is the value of IQWQI predicted.Table 10, illustrates the sensitivity analysis result for IQWQI prediction by ANNR.The model was run 18 times, in each time, one parameter was excluded, ANNR-IQWQI-DO means the test calculated the IQWQI without the DO, and ANNR-IQWQI-BOD means the IQWQI was calculated without BOD, etc.By comparing the lowest R 2 and highest RMSE from Table 10, the most significant and influential parameters on IQWQI are Pb + , Ni + , Cr + , CN -, pH, PO4 3-, Zn + , DO, NO3 -, Al 3+ , and Fe 2+ .The residual error of the 18 models developed for IQWQI prediction is represented in.Fig. 3.

Result of IQWQI and ERI for this study
The results of IQWQI and ERI for the freshwater at different sites and seasons are represented in Figs. 4 and 5; 17 parameters were used in calculating the two indices.Table 2.During the dry season, all sites fall under the very poor water quality category with a high degree of risk (94.99 and 75.89 to 98.49 and 92.58, respectively).While in the wet season, the values of both indices were lower than in the dry season but still in the same categories except for site 2, where the IQWQI ranking was poor water quality but also with a high degree of risk (71.56-88.90 and 57.14-82.88,respectively).The parameters exceeding the Iraq rivers maintenance system in the dry and wet seasons are represented in Table 12.Pb + , SO4 2-and TDS concentrations were beyond the limits continuously, and Pb + concentration in this study was far beyond the limits.For this reason, the water quality falls into the very poor category.In general, according to IQWQI and ERI, the water quality of Tigris River in Baghdad city for different uses was ranked between good to unsuitable and never had an excellent ranking in any of the four water purposes.This situation is related to the increasing pollution in Tigris River due to discharging of effluent from various and uncontrolled sources such as industries, domestic waste, and agricultural activities, as confirmed by different researchers like the study of Fadhel 33 , which found increasing salinity content in Tigris River water in the Mosul city comparable with the past forty years, in addition to study of Al-Obaidy et al. 34 on Tigris river in Baghdad city where recorded high values for electrical conductivity reached to 1205.7 (µs.cm -1 ), and in the study of Nashaat et al., 35 on Tigris river south of Baghdad they notice an increase in the nutrient concentrations with decreasing dissolved oxygen.Abdul-Jabar and Thabi 36 applied Heavy Metal Quality Index on two sites on Tigris River in Bagdad City where they found that cadmium, lead, and chromium slightly affected to extremely affected the river's health.Also, Al-Obaidy et.al. 37 indicate serious contamination of Tigris River by heavy metals in both sediment and water.So, continuous river water quality monitoring is required to assess water quality for various uses.

Conclusion
Several water quality indices were used to assess the water situation in Iraq, and they showed a discrepancy in the WQI results due to the different variables used and the weights adopted in each index.The results of the new IQWQI showed high efficiency with the possibility of relying on a specific number of parameters that were chosen by water quality experts.Also, the index merges the quality and pollution indices, where IQWQI is linked with ERI to eliminate the eclipse effect in WQI.Finally, the proposed model allowed the Iraqi Water Quality Index (IQWQI) user to eliminate any parameter from the index only in case the final weight does not fall below 0.7.Sensitivity analysis using artificial neural network regression (ANNR), can produce a more reliable and accurate output of prediction of the IQWQI than backward linear regression (BLR).

Figure 1 .
Figure 1.Sampling sites across Tigris River, Baghdad City (google earth, 2022) (green boxes represent the sites and the orange boxes represent the distance between each two sites).

F3 1 . 1 .
When the test value must not exceed the objective.When the test value must not fall below the objective. = (∑ )/(  ) 3 = (/(0.01+ 0.01)) The comparison was made by removing one parameter each time from the calculation of IQWQI and comparing the results with the result of IQWQI, which includes all parameters (IQWQI-Ref.).High R 2 values and low RMSE and SSE values indicate non-influencing parameters in calculating water quality.In contrast, low R 2 and higher RMSE and SSE values indicate influencing factors in calculating water quality.

Figure 3 .
Figure 3. Residual error of the 18 models developed for IQWQI estimation for freshwater use based on sensitivity analysis.

Figure 4 .Figure 5 .
Figure 4.The result of IQWQI and ERI for the freshwater of Tigris River during the dry season

Table 8 . Weight Assignment for Studied Parameters
A linear scaling function was applied to convert parameter values to the sub-index (equation), where sub-index values were assigned based on the pollution condition 2 (example 2).It can be noticed from Table8that DO, Pb. pH and SO4 have a high value of sub-index, which correlates with the high the values of these parameters in the guidelines as previously explained and it must be kept in mind that the calculation of DO and pH differ from the rest where both of them must approach ideal values which are 14.6 for DO and 7 for pH.For a more detailed explanation of the calculation of IQWQI, see examples 1 and 2. The aggregate parameters collection process consolidates all parameters' quality scores obtained from subindices into a single water quality index score.A simple additive aggregation function was used to aggregate sub-indices.This final step is essential to produce a single unitless number representing overall water quality relative to the chosen guideline (Example 3).