Testing Groundwater Quality in Jouamaa Hakama Region (North of Morocco) Using Water Quality Indices (WQIs) and Fuzzy Logic Method: An Exploratory Study

Abstract

Groundwater is one of the important determinants of human health in many regions of Morocco. Indeed, it is one of the government’s top concerns. However, slow and wrong decisions are hindering the advancement of the quality of groundwater in Morocco. The traditional monitoring methods are costly, time-consuming, and human-resource-intensive, especially in such a developing country. An exploratory study was conducted in the Jouamaa Hakama region in Morocco to test and compare groundwater quality using in situ measurements, water quality indices (WQIs), and a fuzzy logic (FL) method. The methodology followed in this study involves using and comparing four WQIs and FL based on in situ measurements at nine different wells along the Ouljat Echatt River downstream of the Chrafate wastewater treatment station. Twelve physical–chemical and bacteriological parameters: potential of hydrogen (pH), temperature (T°), turbidity (TURB), chemical oxygen demand (COD), biochemical oxygen demand in five days (BOD₅), suspended matter (SM), phosphate (PO₄³⁻), nitrite $(N{O}_2^{-}$), nitrate $(N{O}_3^{-}$), ammonium $(N{H}_4^{+}$), dissolved oxygen (DO), electrical conductivity (EC), and fecal coliforms (FC) were measured in 2016 and 2017. The results show that all wells are of poor quality according to FL and WQIs; the Weighted Geometric WQI and Oregon WQI indicate that the groundwater is below the acceptable standard for human consumption, while the Weighted Arithmetic WQI and Logarithmic WQI indicate that the majority of wells are of good quality. These tested indices represent an excellent tool to support decision making and can be highly helpful in monitoring groundwater quality in vulnerable areas such as the Jouamaa Hakama region in the north of Morocco.

1. Introduction

Among the several sources of water used, groundwater is the most valuable natural source. It is utilized for irrigation, industry, and domestic water supplies. Due to population growth, urbanization, and industrialization, the demand for drinking water has steadily increased [1]. However, promoting sustainable development and the advancement of both socioeconomic and environmental goals is crucially connected to the quality of groundwater [2]. In most nations across the world, determining the quality of water is critical and plays a significant role in its sustainable development. Numerous potential sources of contamination make groundwater pollution a challenging issue [3], such as anthropogenic activity, agricultural and industrial practices [4], or contamination from dirty rivers [2], and some waters are now unsafe to consume. Drinking-water contamination is a source of illness and is frequently linked to the spread of ailments such as cholera, dysentery, poliomyelitis, and diarrhea [5]. According to the World Health Organization (WHO), polluted waters are responsible for around 80% of illnesses in humans. Therefore, groundwater and surface water monitoring and management have become a necessary process that should be integrated into political perspectives for sustainable development [6]. The groundwater in Morocco has always been a priority as part of a strategic decision to develop the water sector [7]. It is a significant component of the nation’s hydraulic legacy [8]. A substantial reduction in precipitation at the start of this century has been observed, which is concerning for the decades to come [2]. Like many other nations in the Mediterranean region, the groundwater in Morocco is contaminated.

The management and protection of water supplies depend greatly on the evaluation of water quality [9]. For the country to thrive sustainably, groundwater quality monitoring is essential. The definition of groundwater quality includes the thermal, physical, chemical, and bacteriological characteristics of the water body [10]. Indeed, a decision index determines sustainable consumption and water resources quality [11]. Due to the wide usage of groundwater (irrigation and human consumption), the monitored parameters vary and are usually linked to consumer safety, which is the purpose of this study, as well as understanding the nature of pollutant sources [12].

Groundwater quality indicators for human consumption include pH, water T°, TURB, COD, BOD₅, SM, PO₄³⁻, $N{O}_2^{-}$, $N{O}_3^{-}$, $N{H}_4^{+}$, DO, EC, and FC [13,14]. These parameters are traditionally identified by field sampling, in situ measurements, and laboratory analysis and compared with their standard values [15,16]. This approach is costly, time-consuming, and human-resource-intensive. In addition, many WQIs have been developed using these in situ parameters for numerical outcome scores [17,18,19].

WQIs have the benefit of providing information regarding the status of water resources to the general public, water managers, and decision-makers [20,21]. This makes it possible to evaluate the success or failure of any management plan for enhancing the quality of the water [22]. To make groundwater quality information easily understandable to the general public, decision-makers, and non-specialists, the WQI converts a collection of chosen water quality criteria into a single dimensionless number [23].

WQIs were first introduced by Horton in 1965 [24]. The aim of a WQI is to combine all of the parameters into a unique number that can be used to categorize water based on in situ parameters. Numerous works and studies on WQIs and water quality measures have used WQIs for water quality [25]. The steps required to generate WQIs were described by Horton [24] as follows: (1) collect measurements for individual water quality indicators, (2) convert the measurements into “sub-index” values to represent them on a common scale, and (3) combine the various subindex values into a single overall value [26]. The five forms of WQI aggregation functions are (a) the arithmetic aggregation function, (b) the multiplicative aggregation function, (b) the geometric mean, (c) the harmonic mean, and (d) the minimum operator. Various scholars have attempted to build a water quality index based on these functions [26].

Later, Brown et al. [27] created a new WQI that was based on the judgment of a panel of 142 water quality experts, who determined the weighting of each variable and established five categories for water quality: red (very bad), orange (bad), yellow (average), green (good), and blue (excellent). The original index that Brown et al. suggested was in an arithmetic form. Later, Brown et al. (1973) [28] believed that a geometric aggregate was preferable to an arithmetic aggregate because it is more sensitive when a single variable exceeds the norm.

In the same period, 1971, Prati et al. [29] proposed a different index based on water quality criteria, while Deininger and Landwehr [30] proposed their own WQI that is conceptually related to Brown et al. [27]. Three specific-purpose water quality indices were presented by Nemerow [31], and when combined, they yielded a general water quality index. In their strategy, they combined the maximum value and the average value of each variable. A variable’s impact is lessened by the approach when it significantly exceeds the permitted limits [32]. In Alabama (USA), Dinius (1972) [33] presented another WQI based on the Horton index [32]. This WQI establishes a decreasing scale from 100 to 0, with 100 representing the ideal water quality. Dinius [34], in 1987, developed another WQI using the subindex method introduced by Dalkey with some changes [35].

In 1989, House and Newsome [36] demonstrated how a WQI might be used to assess “good” and “bad” water by consolidating data on a variety of biological and physical–chemical characteristics into a single, straightforward index. Afterwards, new indices appeared, such as the global pollution index (Sargaonkar and Deshpande) [37]; the River WQI; and the Scatter Score Index, which is used to assess water quality around mine sites in the United States and detect changes in water quality over time and space [32,38]. Meanwhile, in Canada, the Canadian Council of Ministers of the Environment (CCME) has introduced a WQI, which is a non-linear index [32].

Another WQI was created by Said et al. [39] using logarithmic aggregation, with which they intended to preserve the precision of the index while reducing the number of variables that needed to be added and changing the aggregation method.

Introduced by Icaga in 2007 [40], the most recent technique for computing a WQI is based on FL and was motivated by Silvert’s [41] work in 2000 on environmental assessment. FL is a form of multivalued logic that expresses the partial truth between being false and true. It takes any real number between 0 and 1, unlike Boolean logic [40], where the truth values of variables can only be the integer values 0 or 1. Zadeh (1965) [42] introduced fuzzy logic as yet another practical and affordable technique for determining the quality of water. It has the potential to be an excellent tool since it simulates complex systems under ambiguous and imprecise situations using qualitative and quantitative models.

In general, WQIs have been used to measure water quality in a variety of water bodies around the world [18,43,44,45,46] using different concepts [47,48,49]. The selection of quality indicators, the application of weights, the construction of subindices, and the determination of the aggregate quality value are the foundations of the methodology used to calculate WQIs [50]. Similar to this, numerous works and studies on WQIs and water quality measures have been conducted in Morocco [51,52,53,54,55].

WQIs can be seen as water quality models that depict a complicated reality simply by selecting variables and defining procedures for weighting and aggregating variables [32]. Different ways of aggregating variables have been used. These are mainly: the weighted arithmetic mean (WAWQI), the weighted geometric mean (WGWQI), the weighted and unweighted harmonic quadratic mean (WUHQMWQI), in particular, the Oregon Water Quality Index (OWQI), and, more recently, aggregation using logarithmic (Logarithmic WQI) and fuzzy logic (FL)-based functions. The role of WQIs is to simplify a set of parameters, while the role of FL is to analyze these parameters [56].

The objective of this exploratory study was to use the most well-known WQIs and FL to assess and compare groundwater quality in the Jouamaa region in the north of Morocco. It also aimed at confirming the superiority of the FL method in accurately determining water quality from the used parameters. Moreover, it is intended to support decision-makers and managers to periodically assess and monitor water quality for appropriate management strategies to minimize the level of water contamination.

2. Materials and Methods

2.1. The Study Area

The Jouamaa Hakama region is located in the north of Morocco, 17 km away from the city of Tangier. The study area is characterized by the Tangier Automotive City (TAC) industrial zone, the Chrafate wastewater treatment station, the Ouljat Echatt river, 9 wells, and the Ibn Batouta dam (Figure 1). The Ouljat Echatt river receives the treated TAC discharges from the Chrafate wastewater treatment station [57,58], and all 9 wells are distributed along this river (Figure 1 and Figure 2) [59]. The wells are between 5 and 10 m deep and used for the population’s daily needs (Figure 2). The water from these wells is drawn by a pump into well n° 3 (P3) (see Figure 1), while for the other wells, the water is drawn manually by buckets.

2.2. Water Quality Analysis

Eighteen samples were extracted from the nine wells along the Ouljat Echatt river for physical, chemical, and bacteriological analyses.

Table 1. The analyzed parameters during this study and the reasons for the parameters’ selection with references.

Type of Parameter	Parameter	Reason for Choice	Reference
Physical–chemical	BOD5, COD, DO, pH, and T° EC TURB $N{O}_2^{-}$, $N{O}_3^{-}$, $N{H}_4^{+}$, $P{O}_4^{3-}$, and SM	An increase in T° leads to a pH decrease. All parameters have an impact on DO and can lead to algal growth. Pure water is not a good conductor of electric current. - Associated with cancer, birth defects, and other health effects (methemoglobinemia). - $N{H}_4^{+}$ is not toxic to the human body but can easily be converted to $N{O}_2^{-}$/$N{O}_3^{-}$. - Excessive amounts are the main cause of water eutrophication. Can concentrate and transmit water pollutants and impede their removal by water treatment.	[60,61,62,63,64,65].
Bacteriologic	FC	Potential transmission of enteric pathogens such as Salmonella, Shigella, Vibrio cholerae, and E. coli. Potential transmission of enteric pathogens such as Salmonella, Shigella, Vibrio cholerae, and E. coli.

shows the selected parameters and the reasons for their choice in this study. Before sampling, each bottle was rinsed with distilled water, and sterilized bottles were used for bacteriological and other analyses. Each sample was tested in situ for pH, T°, DO, EC, and TURB. A volume of 1.5 L was taken using flasks at the well for other physical–chemical and bacteriological parameters (Table 1). The samples were identified using sampling sheets (date, geographical coordinates, time of sampling, number, etc.). The sample bottles were transported to the laboratory in a cooler at low T° (4 °C). For bacteriological analyses, the samples were analyzed within a period of less than 24 h after collection. The physical–chemical parameters were analyzed in the “Laboratory of Physical Chemistry of Materials, Natural Substances and Environment, Chemistry Department, Sciences, and Technology Faculty, Abdelmalek Essaâdi University, Tangier, Morocco”, while the bacteriological analysis was conducted in the laboratory “National Office for the Sanitary Safety of Food Products ONSSA” Tangier, Morocco.

Table 2. The parameters measured, the analytical methods and Moroccan standards used during the analysis, and the interpretation of the results.

Parameters	Analytical Methods	Units	Moroccan Standards [66]	References
pH	pH meter in situ	---	6.5-8.5	NM 03.7.009
T°	Thermometer	°C	30
EC	Conductimeter	µS/cm	1300	NM 03.7.011
BOD5	BOD5 m	mg/L	3	NF T90-103
COD	COD meter	mg/L	30	NF T90-101
DO	Oximeter	mg/L	7 (90%)
SM	Gravimetric	mg/L	50	NF T90-105-2
TURB	Turbidimetry	NFU	5	NF T90-033
$N{O}_2^{-}$	Spectrophotometry	mg/L	0.5	NF T 90-040
$N{O}_3^{-}$	Spectrophotometry	mg/L	0.5
$N{H}_4^{+}$	Spectrophotometry	mg/L	0.5
$P{O}_4^{3-}$	Spectrophotometry	mg/L	0.2–0.5
FC	Water quality—Research and enumeration of Escherichia Coli and coliform bacteria; Part 1: Membrane filtration method	MPN/100 ml	≤20	NM 03.7.003

summarizes all of the parameters measured, the analytical methods and Moroccan standards used during the analysis, and the interpretation of the results.

2.3. The WQI Approach

The process of creating a WQI consists of condensing a number of different indicators into a single, concise value that identifies the water quality of a specific source and is understandable by a broad audience, including non-experts such as the general public or decision-makers [51].

There are water quality indices based on indicators and aggregation methods currently used around the world, such as the US National Sanitation Foundation’s Water Quality Index (NSFWQI); the Canadian Council of Ministers of the Environment Water Quality Index (CCMEWQI); the British Columbia Water Quality Index (BCWQI); and the OWQI.

2.3.1. Chosen WQIs and Data Collection

Five WQIs based on five different equations were used in this study for the following reasons:

• Each of the chosen equations belongs to an aggregation category of the WQI.
• The 5 equations can be compared to assess the quality of groundwater belonging to the study area.
• The chosen equations will help us to model and project a relevant method for the future monitoring and surveillance of ecosystems downstream of discharges of treated wastewater from industrial areas.
• Modeling from WQIs will provide decision-makers with a reliable and easy tool for controlling and monitoring water resources downstream of industrial areas.
• For the sampling and selection of samples, we targeted 9 wells on either side of the river downstream from the Chrafate wastewater treatment station to the dam. The choice of wells is for the purpose of monitoring whether there is groundwater degradation and whether there is longitudinal and transverse propagation of pollution.

The results of the samples were compared to Moroccan standards [66] according to Table 2, and the WQIs were calculated based on Moroccan standards for drinking water.

For the chosen equations,

Table 3. Chosen WQIs.

Aggregation Methods	Equations
WAWQI	WQI = $\frac{{{\displaystyle \sum }}_{i=1}^n\left(Wi x Qi\right)}{{{\displaystyle \sum }}_{i=1}^{n}Wi}$
WGWQI	WQI = ∏qiwi
OWQI	$OWQI=\sqrt{\frac{n}{{{\displaystyle \sum }}_{i=1}^n\frac{1}{S{I}_i^2}}}$
Logarithmic aggregation	NewWQI = log ($\frac{{\left(DO\right)}^{1.5}}{{\left(3.5\right)}^{TP} \ast {\left(TURB\right)}^{0.15}\ast {\left(15\right)}^{\frac{FC}{10000}}+0.14 \ast {\left(SC\right)}^{0.5}}$)
Fuzzy logic	Fuzzy WQI= $\frac{\int \mu \mathrm{Ã}\left(x\right).xdx }{\int \mu \mathrm{Ã}\left(x\right)dx}$

below summarizes the equations used to compute each WQI.

1.

Weighted Arithmetic Average: The First NSFWQI in 1971 [67]

Brown et al.’s (1965) initial index proposal included an arithmetic form. Choosing the appropriate variables to include in the WQI was one of the difficulties with the Horton approach [67]. With the support of the National Sanitation Foundation, Brown et al. created a different version of the WQI; thus, the name of this index is NSFWQI (Equations (1)–(5)). Brown’s method, which is based on Dalkey’s method (the Delphi technique developed by the Rand Corporation) [35], was carried out by carefully choosing the variables, creating a standard scale, and allocating weights to the variables (

Table 4. Weights assigned to parameters [32].

Parameters	pH	T	COD	BOD5	DO	EC	SM	TURB	$\mathit{N}{\mathit{O}}_2^{-}$	$\mathit{N}{\mathit{O}}_3^{-}$	$\mathit{N}{\mathit{H}}_4^{+}$	$\mathit{P}{\mathit{O}}_4^{3-}$	FC
Weight (wi) *	4	2	3	3	5	4	2	2	4	5	3	4	5

* Each criterion is assigned a weight (Wi) ranging from 1 (lowest influence on water quality) to 5 (greatest impact) based on the potential impact on health and the environment.

). A group of 142 specialists were asked to rate the total water quality using the chosen variables on a scale from 1 (highest) to 5 (lowest).

The relative weight (Wi) of each parameter was calculated by Equation (1):

$$Wi=\frac{wi}{{\sum }_{i=1}^{n}wi}$$

(1)

where wi represents the weight of the parameter (i), and (n) is the total number of parameters.

The quality rating was calculated by Equation (2):

$$Qi=\frac{Ci}{Si}\ast 100$$

(2)

where Qi is the quality rating, Ci is the concentration, and Si is the standard.

For the pH, the rating scale was determined by Equation (3):

$$Qi=\frac{\left(CpH-8.5\right)}{\left(6.5-8.5\right)}$$

(3)

The subindex (SI) for each parameter was calculated by Equation (4):

SI = Wi ∗ Qi

(4)

where Wi is the relative weight, and Qi is the scale rating of the parameter.

DO, FC, pH, BOD, $N{O}_3^{-}$, PO₄³, T°, TURB, and SM are the variables included in the calculations. Each variable received a rating that was calculated using the arithmetic mean, and ratings were then changed into temporary weights. The final weight was then calculated by dividing each temporary weight by the total number of temporary weights [68,69,70,71,72]. Brown et al. [28] thought about using a color spectrum to show the range of water quality in each area, with dark red signifying very poor water quality (WQI = 0–10), a narrow strip of yellow showing average quality (50 WQI), and dark blue signifying excellent water quality (WQI = 90–100) (

Table 5. Classification of water quality [72,73].

WQI Value	Class
Excellent water	<50
Good water	50–100
Poor water	100–200
Very poor water	200–300
Unsuitable for drinking	>300

).

$$WQI=\frac{{{\displaystyle \sum }}_{i=1}^n\left(Wi\ast Qi\right)}{{{\displaystyle \sum }}_{i=1}^{n}Wi }$$

(5)

where Wi is the unit weight of the nth water quality parameter; Qi is the quality rating of the ith water quality parameter.

2.

Weighted Geometric Average: In 1973, the NSFWQI was updated for the second time. Brown et al. proposed another formula for calculating the WQI after discovering that the multiplicative formula was better suited to the judgments of experts. Researchers used the same scale of classification of water quality [23,74,75] and employed the geometric mean form of weighting in the WQI (Equation 6 and

Table 6. Descriptor Words and WQI Value Ranges [4].

Rating	Range
Excellent	91–100
Good	71–90
Medium	51–70
Bad	26–50
Very bad	0–25

). Later, other researchers [76,77,78] also used a weighted geometric mean for aggregation [32].

WQI = ∏qi^wi

(6)

where qi is the quality class for the nth variable; Wi is the relative weight for the nth variable (∑Wi = 1). Brown et al. kept the same scale for the classification of water quality [32].

3.

Weighted and Unweighted Harmonic Square Average: Canadian Council of Ministers of the Environment (CCEM) 2001, Dojido et al. 1994 [48]

In order to calculate the WQI, Dojido et al. (1994) [48] introduced the harmonic mean. Individual indicator weights are not used in this average. Researchers discovered that it reduces eclipsing while accounting for the influence of other indicators and is more responsive to the more degraded indicators than the arithmetic form [79]. The British Columbia Water Quality Index and the CCEM Water Quality Index are further indicators based on harmonic means. A harmonic-mean-based methodology has been used in several WQIs [21,80,81]. Another often used WQI that is available to the general public is the OWQI, which was created by the Oregon Department of Environmental Quality (ODEQ) in the late 1970s and revised multiple times since then [81]. Due to the vast amounts of information and resources required for the calculation and presentation of the findings, the original OWQI was, however, abandoned in 1983. The OWQI was updated in 1994 by improving the original subindices (adding T° and a total phosphorus subindex) and thereby improving the aggregation process in response to advances in computer technology, improved displaying tools, and data visualization, which allowed a better understanding of water quality. The NSFWQI, which used the Delphi approach for variable selection, served as a basis for the original OWQI [74].

The OWQI takes into account measurements of the eight factors that affect water quality, including T°, DO, BOD, pH, $N{O}_3^{-}$, $P{O}_4^{3-}$, and FC (Equation (7)). The categories of impairment used to classify the water quality factors include oxygen depletion, eutrophication (the potential for excessive biological development), dissolved chemicals, and health risks [82].

$$OWQI=\sqrt{\frac{n}{{{\displaystyle \sum }}_{i=1}^n\frac{1}{S{I}_i^2}}}$$

(7)

$$OWQI=SQRT (\frac{8}{\frac{1}{SIT^2}+\frac{1}{SIDO^2}+\frac{1}{SIBOD^2}+\frac{1}{SIpH^2}+\frac{1}{SITS^2}+\frac{1}{SIN^2}+\frac{1}{SIP^2}+\frac{1}{SIFC^2} })$$

4.

Logarithmic Aggregation: New Water Quality Index Proposed by Said et al. 2004 [39]

A new WQI (Equation (8) and

Table 7. WQI characterization of water quality [32].

Index Value Ranges	Classification
Highest purity	3
Marginal purity	<2
Poor quality	<1

) was created by Said et al. [39] utilizing logarithmic aggregation in an effort to alter the aggregation method while retaining the precision of the index and cutting down on the number of variables that must be added. After testing their index with a random database, they demonstrated that it produced findings that were comparable to those of the NSFWQI and WEP-WQI [32]. Said et al.’s proposal in 2004 [39] was to divide the WQI calculation into two parts. Variables related to water quality were ranked in the first one based on their importance. To give DO the highest weight, followed by fecal coliform and total phosphorus, and to maintain the index in a straightforward equation, numerous versions were tried in the second stage. Finally, the logarithm was employed to produce small numbers that the general public, stakeholders, and management decision-makers could use with ease. The following factors were chosen for the new WQI based on their amount of influence on water quality: DO, $P{O}_4^{3-}$, FC, TURB, and specific conductivity.

$$\mathrm{New}WQI=\log (\frac{{\left(DO\right)}^{1.5}}{{\left(3.5\right)}^{TP} \ast {\left(TURB\right)}^{0.15}\ast {\left(15\right)}^{\frac{FC}{10000}}+0.14\ast {\left(SC\right)}^{0.5}})$$

(8)

where DO is the dissolved oxygen (% oxygen saturation); TURB is the turbidity (Nephelometric turbidity units (NTU)); TP is the total phosphates (mg/L); FC is the fecal coliform bacteria (counts/100 mL); and SC is the specific conductivity (MS/cm at 25 °C);

The index was designed to range from 0 to 3. The maximum or ideal value of this index is 3.

In very good water, the value of this index will be 3. From 3 to 2, the water is acceptable, and less than 2 is marginal and needs remediation; if one or two variables have deteriorated, the value of this index will be less than 2.

If most of the variables have deteriorated, the index is less than 1, which means that the water quality is poor [32].

5.

Fuzzy Logic: Icaga 2007 (Equation (9)) [40]

Icaga 2007 [40] developed the FL approach after being influenced by Silvert’s [41] work on the evaluation of the environment. According to Zadeh [42], FL is a type of multivalued logic that expresses a partial truth that can either be true or false. Contrary to Boolean logic [83], where variables’ truth values can only be the integer values 0 or 1, it accepts any real number between 0 and 1 [32]. Some people view the idea of “acceptability” as being ambiguous [41]. In order to translate quantitative data into something more palatable, fuzzy membership functions must first be defined to represent the level of acceptability (Figure 3).

$$\mathrm{Fuzzy} WQI=\frac{\int \mu \mathrm{Ã}\left(x\right).xdx }{\int \mu \mathrm{Ã}\left(x\right)dx}$$

(9)

According to Icaga, six steps are needed to develop an FL index for a conventional classification containing four classes, as follows [40]:

• Determination of the quality classes for the measured variables;
• Arrangement of the variables according to their classes into the four groups;
• Application of membership functions;
• Rule bases;
• Use of the fuzzy algorithm: In the fuzzy algorithm, the Mamdani [84] approach is used. Fuzzy inferences of the groups are determined using grades of membership functions of the variables;
• Defuzzification of the inferences using centroid methods, which calculate the center of gravity of the output function.

In order to develop systems more in line with the spirit of human thinking, Zadeh [42] introduced FL, which has since grown to be one of the most popular methods in the field of artificial intelligence. It is considered suitable for creating environmental indices. It has been demonstrated that this technique can handle complicated systems in ambiguous and imprecise situations [85].

It is an effective way to communicate evaluation results in a way that the general public, decision-makers, and non-experts in general may understand [41,85,86]. To characterize water quality, fuzzy-logic-based water quality indices are being developed [19,87,88,89,90]. The intricacy of the environmental circumstances has an impact on the outcome; this issue is connected to the “principle of incompatibility” [91].

2.3.2. FL Formalism

Based on fuzzy set theory, Zadeh developed FL in 1965 [42,92], which enables the consideration of any ambiguity or uncertainty. The idea of membership functions is a fundamental one in FL [51].

2.3.3. Fuzzification

To activate rules that are defined in terms of linguistic variables, net input values are mapped to those variables using membership functions. Additionally, the fuzzifier uses membership functions to assess the degree to which input values belong to each fuzzy set.

2.3.4. Evaluation of the Rule

Combining the fuzzy inputs yields a calculation of the rule’s strength. Multiple connective antecedents are combined using the fuzzy intersection operation. The strength value of the rule’s antecedent has a correlation with the rule’s consequent, and the most frequent way to entail a rule is to cut the consequent membership function at the level of the antecedent truth. Clipping is the name of this technique [51].

2.3.5. Aggregation of Rule Outputs

Using the fuzzy union of each individual rule contribution, the outputs of all rules are then combined into a single fuzzy distribution [51].

2.3.6. Defuzzification

The aggregated output fuzzy set is mapped to a precise number using several methods, including (1) the “centroid”, which is the most popular defuzzification method and involves calculating the center of gravity of the aggregate fuzzy set, (2) the “maximum”, (3) the “mean of the maxima”, (4) the “height”, and (5) the “modified height” [51].

2.3.7. Membership Functions

A membership function, which can take trapezoidal, triangular, Gaussian, etc., forms, is used to represent a fuzzy subset of an input. Each point in a fuzzy set is assigned a membership score between 0 and 1 by the membership function. To define membership functions, several approaches can be utilized, including expert knowledge [93], genetic algorithms [94], fuzzy clustering, and neural networks [95].

2.3.8. Fuzzy Set Operations

The relationships among the fuzzy subsets are defined by using fuzzy set operators, which can be used in developing different systems based on fuzzy logic, of which three basic operators are: 1. Intersection (AND), 2. Union (OR), and 3. Negation (NOT).

2.4. Using Python Programming Language

The Python programming language is a sophisticated, object-oriented scripting programming language with dynamic semantics and an efficient and intuitive coding style [96,97,98]. It is a viable and attractive analytical tool for environmental research [99]. In this work, the SciKit-fuzzy [100] library was used to model the FL WQI (Figure 4). Physical–chemical and bacterial parameters were the predictor variables, while the WQI and FL models were the dependent variables. The generation of rules is a crucial step. It was achieved based on expertise and knowledge of the area under study.

2.5. Rules for Fuzzy Logic

The procedure performed within FL is described in

Table 8. Linguistic terms and fuzzy sets for the fuzzy index.

Parameters	Units	Very low			Low			Medium			High			Very high
		a = b	c	d	a	b	c	a	b	c	a	b	c	a	b	c
T°	°C	0	5	10	5	10	15	10	15	20	15	20	25	20	25	30
pH	---	0	3	5	3	5	7	5	7	9	7	9	11	9	11	14
EC	µS/cm	0	500	600	500	600	700	600	700	800	700	800	900	800	900	1500
SM	mg/L	0	25	50	50	100	1000	100	1000	2000	1000	2000	3000	2000	3000	6000
DO	mg/L	0	2	4	2	4	6	4	6	8	6	8	10	8	10	12
BOD5	mg/L	0	1	2	1	2	3	2	3	4	3	4	5	4	5	6
COD	mg/L	0	10	20	10	20	30	20	30	40	30	40	50	40	50	60
$N{O}_3^{-}$	mg/L	0	25	50	25	50	75	50	75	100	75	100	125	100	125	150
$N{H}_4^{+}$	mg/L	0	1	2	1	2	3	2	3	4	3	4	5	4	5	6
$P{O}_4^{3-}$	mg/L	0	1	2	1	2	3	2	3	4	3	4	5	4	5	6
TURB	NTU	0	1	2	1	2	3	2	3	4	3	4	5	4	5	6
FC	MPN/100 mL	0	10	20	10	20	100	20	100	1000	100	1000	2000	1000	2000	20,000

and Figure 4. We suggest that the levels of the thirteen parameters are sufficient to assess the quality of groundwater by means of an aggregate index called Fuzzy Water Quality (FGWQ). We chose “very low”, “low”, “medium”, “high”, and “very high” fuzzy sets for the inputs and “excellent”, “acceptable”, and “poor’’ fuzzy sets for the output. Trapezoidal membership functions were defined for the fuzzy sets in the extremities, while triangular ones were used for the in-between fuzzy sets (Figure 4 and Figure 5).

The robustness of the system depends on the number and quality of the rules [19]. In this study, we built the rules below:

#Poor

If T is “very high” or pH is “very low” or EC is “very high” or SM is “very high” or DO is “very low” or BOD5 is “very high” or COD is “very high” or NO3 is “high” or NO3 is “very high” or NH4 is “ high” or NH4 is “very high” or PO4 is “ high” or PO4 is “very high» or TURB is “very high” or FC is “very high” then Quality is “Poor”.

#Acceptable

If (T is “very high” or T is “low” or T is “medium” or T is “ high”) and (pH is “low” or pH is “medium” or pH is “ high” or pH is “very high”) and (EC is “very low” or EC is “ low” or EC is “medium” or EC is “ high”) and (SM is “very low” or SM is “low” or MES is “medium” or SM is “ high”) and (DO is “low” or DO is “medium” or DO is “high” or DO is “very high”) and (BOD5is “very low” or DBO5is “low” or BOD5 is “medium” or BOD5 is “ high”) and (COD is “very low” or COD is “low” or COD is “medium” or COD is “high”) and (NO3 is “very low” or NO3 is “low” or NO3 is “medium” and NH4 is “very low” or NH4 is “low” or NH4 is “medium”) and (PO4 is “very low” or PO4 is “low” or PO4 is “medium” and TURB is “very low” or TURB is “low” or TURB is “medium” or TURB is “ high”) and (FC is “very low” or FC is “low” or FC is “medium” or FC is “high”) then Quality is “Acceptable”.

#Excellent

If T is “medium” and pH is “medium” and EC is “high” and SM is “very low” and DO is “ high” and BOD5 is “very low” and COD is “low” and NO3 is “very low” and NH4 is “very low” and PO4 is “very low” and TURB is “ high” and FC is “very low” then Quality is “Excellent”.

3. Results and Discussion

3.1. Physical–Chemical and Bacteriological Parameter Assessment

Table 9. Values of water quality parameters in 2016 and 2017 for all 9 wells.

Parameters	Range	Year: 2016Mean	Standard Deviation	Range	Year: 2017Mean	Standard Deviation	Moroccan Standards [66]
T (°C)	18.8–20.9	20.36	0.68	19.9–20.7	20.34	0.27	30
pH	7.17–8.05	7.60	0.29	6.82–7.82	7.21	0.35	6.5–8.5
EC (µS/cm)	980–3100	1544.44	804	820–3250	1568	874.29	1300
SM (mg/L)	2–130	21.11	41.49	3–13	7.78	4.02	50
DO (mg/L)	2.08–7.56	5.52	1.83	1.59–6.40	4.20	1.46	7 (90%)
BOD5 (mg/L)	2–29	7.55	8.62	7–20	12.44	4.67	3
COD (mg/L)	1.05–32	10.94	9.75	17.90–42.90	26.13	8.22	30
$N{O}_3^{-}$ (mg/L)	0.13–26.6	5.05	8.56	0.13–12.42	3.26	4.20	50
$N{H}_4^{+}$ (mg/L)	0.38–0.57	0.47	0.06	0.33–0.55	0.45	0.08	0.5
$P{O}_4^{3-}$(mg/L)	0.07–0.99	0.37	0.28	0.41–3.48	1.29	1.15	0.4
TURB (NTU)	1.16–17.7	4.98	5.35	1.25–20	5.16	5.82	5
FC (MPN/100 mL)	83–1191	585.22	435.64	83–1190	585	435.64	20

shows the results of the physical–chemical, mineral, and bacteriological parameters of nine samples. The groundwater T° was 18.8–20.9 °C in 2016 and 19.9-20.7 °C in 2017, with average T° values of 20.36 and 20.34, respectively. The majority of the samples in the study showed nearly neutral properties, lower than the maximum value recommended by Moroccan standards, with a pH range between 7.17 and 8.05.

The value of EC varied between 820 and 3250 µS/cm in 2016 and 2017, with a remarkable maximum value at sites P2, P3, P4, and P6, exceeding the criteria of the Moroccan standard set at 1300 µs/cm (Table 9). Except for sites P1, P2, and P4, all site points had DO concentrations between 1.05 mg/L and 7.56mg/L, complying with Moroccan standards (Table 9). The concentrations of BOD5 show that the values varied from 2 to 29 mg/L and were relatively high compared to the value of 3 mg/L set by the Moroccan standard (Table 9). Given that the COD concentrations increased from 2016 to 2017, with a range of 1.05–42.90 mg/L, it is also reported that there were levels higher than those recommended by Moroccan norms. TURB is a measure of the lack of clarity of the water. The TURB values of the samples were between 1.25 and 20 NTU, and the admissible limit of TURB is 5 NTU according to Moroccan standards (Table 9).

High concentrations of $N{O}_3^{-}$ and $N{H}_4^{+}$ were observed and ranged from 0.13–26.6 mg/L ($N{O}_3^{-}$) to 0.33–0.57 mg/L ($N{H}_4^{+}$) in 2016 and 2017, respectively, with a pH that was somewhat neutral (Table 9). $N{H}_4^{+}$ is formed through the fixation of nitrogen from the atmosphere by soil microbes, while nitrifying bacteria convert nitrogen to $N{O}_3^{-}$ under aerobic conditions [101]. The reason for the presence of water pollution in these wells during the rainy season is the transport of soluble substances to the water table, so most of the sources of nitrogen pollution come from the soil, septic tanks, and animal manure.

Normal concentrations of $P{O}_4^{3-}$ recorded at the nine site points ranged from 0.07 mg/L to 3.48 mg/L, which is less than the 5 mg/L recommended by Moroccan regulations (Table 9). Hence, this confirms the absence of eutrophication [100]. FC showed high values in 2016 and 2017, with an average of 585–585.22 MPN/100 mL, which is more than the 20 MPN/100 mL recommended by Moroccan regulations (Table 9). This indicates recent fecal contamination [102] in the wells, which are used in this region for drinking water. It also means that there is a greater risk that pathogens are present [100]. Based on this result, the source of this fecal pollution could be directly related to TAC and other agriculture activities [58]. It is crucial to note the contact between TAC discharge, the Charafate wastewater treatment plant discharge, the Ouljat Echatt river, and the nine wells [58]. Indeed, organic and mineral pollution negatively impacted the groundwater’s chemical and physical properties [58].

Indeed, mineral pollution negatively impacts the groundwater’s chemical and physical properties. $N{O}_3^{-}$ is one of the most common mineral pollution sources in groundwater across the world [103]. Jalali et al. (2011) [104] conducted research on groundwater $N{O}_3^{-}$ contamination in western Iran and discovered $N{O}_3^{-}$ leaching from agricultural soils, which can raise $N{O}_3^{-}$ concentrations in groundwater. Chen et al. (2016) [105] discovered a greater concentration of $N{O}_3^{-}$ in northwest China groundwater and discovered that an increased fertilizer rate, intensive irrigation, and higher permeability might be the primary reasons for elevated $N{O}_3^{-}$ in northwest China groundwater. Furthermore, Burow et al. (2010) [106] discovered that the global rise in nitrogen fertilizer usage over the last few decades had resulted in higher $N{O}_3^{-}$ concentrations in groundwater.

Groundwater pollution can have an influence on human health, environmental quality, and socioeconomic development. Many studies, for example, have found that excessive levels of organic pollutants pose a health risk to human populations [107]. This is especially important for newborns and children, who are more vulnerable to the impacts of these pollutants than adults [108,109]. For example, “blue baby syndrome,” also known as newborn methemoglobinemia, is caused by high $N{O}_3^{-}$ concentrations in the drinking water used to create baby formula. Groundwater pollution can also have an impact on human health through its impacts on the food production chain.

3.2. Evaluation of the Groundwater Quality Using Different WQIs

Based on prior work on WQIs, this study proposes a methodology for evaluating the performance of the groundwater quality utilizing data from all well-water samples at the Jouamaa Hakama site. To that end, this study compared four water quality indices and FL. Table 9 displays the values of the water quality metrics used in this investigation for all wells in 2016 and 2017.

According to the FL WQI, WGWQI and OWQI findings, all or the majority of the wells are of poor quality, indicating that the groundwater is below the permitted limit for human consumption based on Moroccan regulations. However, the WAWQI indicates that the majority of the wells are of good quality, and the Logarithmic WQI indicates that the quality is of the greatest purity in eight cases, intermediate quality in six cases, and poor quality in the remaining five cases (

Table 10. Comparison between results of four WQIs and FL for nine wells in 2016 and 2017.

Year	POINTS	WAWQI	WAWQI Quality	WGWQI	WGWQI Quality	OWQI	OWQI Quality	Logarithmic WQI	Logarithmic Quality	Fuzzy WQI	Fuzzy WQI Quality
2016	P 1	66.41	Good water	19.24	Very bad	0.069	Very poor	2.62	Highest purity	22.59	Poor
2016	P 2	92.95	Good water	23.55	Very bad	0.096	Very poor	1.73	Marginal quality	22.59	Poor
2016	P 3	68.36	Good water	29.65	Bad	0.138	Very poor	5.66	Highest purity	22.59	Poor
2016	P 4	69.75	Good water	33.65	Bad	0.220	Very poor	6.07	Highest purity	22,59	Poor
2016	P 5	54.22	Good water	25.30	Bad	0.103	Very poor	2.52	Highest purity	22.59	Poor
2016	P 6	70.38	Good water	21.69	Very bad	0.166	Very poor	1.71	Marginal quality	22.69	Poor
2016	P 7	79.16	Good water	31.67	Bad	0.106	Very poor	2.65	Highest purity	22.59	Poor
2016	P 8	174.57	Poor water	28.35	Bad	0.136	Very poor	0.45	Poor	22.68	Poor
2016	P 9	43.89	Excellent water	0.00	Very bad	0.000	Very poor	1.84	Marginal quality	22.59	Poor
2017	P 1	109.90	Poor water	30.26	Bad	0.215	Very poor	1.25	Marginal quality	22.59	Poor
2017	P 2	106.44	Poor water	37.67	Bad	0.179	Very poor	0.83	Poor	22.59	Poor
2017	P 3	143.47	Poor water	60.14	Medium quality	0.261	Very poor	0.78	Poor	22.59	Poor
2017	P 4	79.39	Good water	27.62	Bad	0.213	Very poor	3.07	Highest purity	22.59	Poor
2017	P 5	88.60	Good water	29.21	Bad	0.113	Very poor	2.13	Highest purity	23.47	Poor
2017	P 6	72.33	Good water	21.17	Very bad	0.134	Very poor	1.52	Marginal quality	22.79	Poor
2017	P 7	86.00	Good water	38.05	Bad	0.217	Very poor	1.98	Marginal quality	22.59	Poor
2017	P 8	108.78	Poor water	32.69	Bad	0.127	Very poor	0.03	Poor	22.59	Poor
2017	P 9	92.18	Good water	30.62	Bad	0.101	Very poor	0.04	Poor	22.59	Poor

and Figure 6, Figure 7 and Figure 8).

According to Table 10, the WAWQI in 2016 ranged between 43 and 174, indicating good quality in the majority of wells. In 2017, the same index ranged from 72 to 143, indicating acceptable quality in just 55% of the wells and low quality in the remaining wells (indicates degradation in P1, P2, P3, and P9). The Logarithmic WQI shows similar results to the WAWQI, with the highest purity in the majority of the wells, except for P8 in 2016 and P2, P3, P8, and P9 in 2017.

Considering the WGWQI, we discovered that water quality ranges from bad to extremely poor in the majority of the cases, with the exception of P3 in 2017. The OWQI classified all samples as very poor quality between 2016 and 2017, with a maximum value of 0.26 in P3. Finally, FL WQI rates all samples as low quality, with a score of 22.59 in the majority of cases.

There is a significant disparity between the findings obtained by the various WQIs and FL. This can be explained by either the differences in the aggregation technique of the input parameters and the problems arising from them or the fact that many parameters were left out of certain WQIs.

In the case of the WAWQI, which is based on the additive method, the true overall quality is not reflected since lower values of one or some subindices are overridden by higher values of other subindices, or vice versa [110]. For example, the water quality in well P9 in 2016 was “Excellent” according to the WAWQI, while the FC concentration in the same location was estimated to be 864 (MPN/100 mL), indicating fecal contamination by far exceeding the Moroccan standard of 20 (MPN/100 mL), hence making the water unsuitable for drinking.

In the case of the Logarithmic WQI, the index considers only five parameters: DO, turbidity, TP, EC, and FC. The quality results from the latter index can be misleading, especially in the case of the ideal values of these five parameters, where the index will classify the point as being of the highest purity, but the other parameters (not included in the equation) by far exceed the standard limits.

However, the WGWQI, which is based on the geometric mean, and the OWQI, which is based on the harmonic mean, can be considered accurate in the context of this study, since they coincide with the results of the parameter assessment and account for the contribution of parameter values indicating low quality. This is explained by the fact that, from their aggregation functions, they are more sensitive to the more degraded input parameters, which alleviates the eclipsing effect [111].

The FL WQI gives equal consideration to both its measured value and the groundwater quality standard in its evaluation, while the former only considers the water quality standard. The FL WQI operates in a way such that all parameters’ quality requirements should be met to achieve the quality objective. The FL WQI is more effective in this context because it collects expert knowledge using the Mamdani inference system, which has proven to simulate an expert’s knowledge in a way that compares to human thinking [85,112]. Furthermore, the effectiveness of the FL WQI has been validated in previous work, such as [113,114], and the results are confirmed through the parameter assessment in Section 3.1 and are close to the results obtained in [58], in which researchers assessed water quality in the same study area by leveraging GIS and analyzing physical–chemical indicators with respect to Moroccan standards. The latter study concluded that water in almost all samples had been contaminated due to urban and agricultural activities.

4. Conclusions

The results of the physical–chemical, mineral, and bacteriological analyses in Jouamaa Hakama groundwater showed high concentrations of TURB, COD, BOD5, SM, $P{O}_4^{3-}$-, $N{O}_2^{-}$, $N{O}_3^{-}, N{H}_4^{+}$, DO, EC, and FC in nine wells. These exceed the Moroccan standards, except for pH, T°,$and N{O}_3^{-}$, which were found to be normal in all wells. This showcases the vulnerability of the water quality of the wells where the Ouljat Echatt river runs with treated TAC discharges from the Chrafate wastewater treatment station. Additionally, these results confirm the pollution of the groundwater in the Hakama region and the need for more attention and monitoring of water resources in the north of Morocco.

This exploratory study of using WQIs and FL for testing the groundwater quality determination performance in order to support decision-makers in the north of Morocco showed the poor quality of groundwater wells. The findings indicate that the groundwater is below the human consumption norm based on Moroccan standards. In detail, the comparison between the WQIs showed that the arithmetic WQI indicates that the majority of the wells are of good quality, and the Logarithmic WQI indicates that the quality is of the greatest purity in seven cases, intermediate quality in six cases, and poor quality in the remaining five cases. The Weighted Geometric WQI and Oregon WQI indicate that the groundwater is below the acceptable standard for human consumption. FL indicates poor quality at all sampling points.

The comparison of the data produced from the different WQI computations and FL reveals that there is a significant difference between the results obtained by the WAWQI and Logarithmic WQI with the FL model. On the other hand, similar water quality was indicated by the FL WQI, WGWQI, and OWQI. While the WAWQI and log WQI indicate high water quality in the majority of the samples examined, these WQIs are deemed unreliable in this study since they do not reflect the reality of the area. Generally, FL showed more accurate results and may be used to confirm other WQI outcomes. This study implies that FL might be a potential alternative to WQIs and standard groundwater quality models, particularly in unpredictable situations.

Additionally, the study presents the first step in the implementation of new models based on WQIs and FL that can assist decision-makers on wells’ closure. The relationship between these methods provides an estimated determination of potential pollution in groundwater. Indeed, the future development of the method based on this exploratory study can provide low-cost and spatial information for Jouamaa Hakama groundwater monitoring.

Hence, this study demonstrates the potential of WQIs and FL—using water quality indicators—as a useful tool for organic and mineral groundwater quality and, in particular, the use of these data for developing prediction models, which could be a cost- and time-effective technique compared to classical methods used in the north of Morocco.

This is a preliminary result for monitoring groundwater quality using WQIs. To improve this tool and the Jouamaa Hakama region’s groundwater quality, some objectives following this study are as follows:

-

Didactical field sampling: two times per month to obtain more data.

-

Inclusion of more bacteriological parameters in WQIs, such as total coliforms and pathogens, hydrochemical analysis, and heavy-metal parameters, to enhance water quality assessment, because most nitrate, $N{O}_2^{-}$, and FC pollutants come from sewage water and industrial areas.

-

Control of the effectiveness of Charafate wastewater treatment plants, such as physical and chemical (jar test) analyses and bacteriological analysis, to confirm the source of pollution in this area.

-

Control of TAC wastewater quality and its compliance with Moroccan environmental regulations.

-

More studies on the techniques and methods of applying suitable treatment for liquid effluents from agriculture, industrial, and urban areas.

-

Establishment of the deep monitoring of the domestic water sanitation network in urban riparian areas to maintain its balance.

-

The adoption of machine-learning techniques for the prediction of important pollution indicators in our study area.

-

The development of a real-time model for groundwater monitoring based on the prediction of pollution indicators for supporting decision-makers in the north of Morocco.