Analysis and classification of bottled waters in the Maghreb Arab region Open Access

The consumption of bottled natural mineral, spring and table waters has increased considerably in recent years in many regions in the world and, particularly, in the Maghreb countries (North Africa) namely Algeria, Tunisia, Morocco, Libya and Mauritania (Miguel 2006; Lourenço et al. 2010; Felipe-Sotelo et al. 2015; Horowitz et al. 2018; Aslani et al. 2021). This prodigious consumption evolution is attributed to several factors, in particular, customer concern about increased water contamination, bad taste and odors, and bottled water was more closely controlled, safer and sometimes even therapeutic, social norms, affect, image on water consumption and public health campaign (Saleh et al. 2001; Lourenço et al. 2010; Huang & Liu. 2017; Etale et al. 2018; Totaro et al. 2018; Geerts et al. 2020). Doria (2006) summarized factors leading consumers' preferences in drinking bottled waters in USA, Canada and France (Table 1).

Mineral water comes from a source tapped at one or more bore holes or springs, and originates from a geologically and physically protected underground water source. It can be differentiated by its particular content of mineral salts, trace elements or other constituents from other forms of water intended for human use. These mineral waters may have therapeutic properties favorable to human health (Lourenço et al. 2010; Jora 2015; USFDA 2022).

Spring water, as their name suggests, comes often from a specific underground source which gets to the land surface to emerge and is safeguarded from any possible pollution (Jora 2015). Spring water is bottled and delivered as it is at the source. Unlike mineral water, their mineral salt composition is not fixed. The only treatments they can undergo are basic physical treatments: decantation, filtration and aeration, if they naturally contain dissolved carbon dioxide (Lourenço et al. 2010; Jora 2015; USFDA 2022).

The table water is usually considered as tap, well or lake waters, bottled, after treatment, and sometimes adding minerals. Their composition is often variable, frequently gasified and finally dechlorinated (Jora 2015; USFDA 2022).

Several studies have been performed to evaluate the composition and quality of bottled waters (Cüneyt 2007; Grošelj et al. 2010; Liidia & Valter 2010; Eggenkamp & Marques 2013; Totaro et al. 2018; Aslani et al. 2021), contrary to the Maghreb countries where we can only cite a few articles dealing with bottled waters in Algeria (Hazzab 2011a, 2011b; Sekiou & Kellil 2014; Labadi & Hammache 2016; Kerdoun et al. 2021), in Tunisia (Gharbi et al. 2010; Sghaier & Ben abdallah 2018) and in Morocco (Gharibi et al. 2018).

In Algeria, Hazzab (2011a, 2011b) discussed, in particular, the evolution in legislation regarding the exploitation, production and marketing of bottled water and Sekiou & Kellil (2014) exposed the classification of spring bottled water. In the latter, the results showed that spring bottled waters can be divided into two distinct major groups. Labadi & Hammache (2016) presented a comparison between mineral and spring waters produced in Algeria, whereas Kerdoun et al. (2021) analysed fluoride in bottled waters in southern Algeria which indicated a maximum value found is 1.65 mg/l.

In Tunisia, Gharbi et al. (2010) investigated the isotopes ^234and238U in 10 bottled waters and showed that the levels of these elements vary between 0.16 and 2.02 mSv/a, and do not exceed the maximum dose recommended by the WHO.

In the same context, Sghaier & Ben abdallah (2018) studied the statistical classification of 20 Tunisian bottled waters and concluded that the 20 brands can be grouped in 10 distinct classes with similar chemical characteristics.

While in Morroco, Gharibi et al. (2018), through the water quality index, showed that among the 17 bottled waters studied (mineral water, spring water and table water), only carbonated mineral waters were considered unacceptable for human consumption.

In any case, to our knowledge, there is no global work that encompasses all bottled waters in all the Maghreb countries.

Many constituents are very important and their presence in bottled water could have a serious negative impact on the population health, only particular parameters that define the characteristics of the water are requested by the Maghreb countries legislations (NM 1991; Jora 2011, 2014; NT 2013). Maghreb countries' regulations state that physicochemical parameters like, dry residue, anions, cations and some trace elements must be surveyed at the source. In addition, the mineral water must be free of parasites and pathogenic microorganisms and must contain no fecal contaminants (Jora 2006, 2015; NT 2007a).

Using different empirical, graphical and multivariate techniques, the study discusses the water quality and classification according to the Maghreb, the European (EU), and the World Health Organization (WHO) legislations.

In the literature (McQueen 1967; Rajesh et al. 2002; Marc-Henri 2004; Cüneyt 2007; Gaetano et al. 2007; Lourenço et al. 2010; Tanaskovic et al. 2012; Eggenkamp & Marques 2013; Alfaifi 2019; Taşan et al. 2022), several methods for characterizing and classifying groundwater, in particular packaged waters, have been provided.

The advantage of these techniques, such as principal component analysis (PCA), Factor analysis (FA), Cluster (CA), is their ability to analyze large and complicated data. Such methods, sometimes, create new variables in the comparison and interpretation of data (Gaetano et al. 2007).

The use of different multivariate statistical techniques, the hierarchical cluster analysis (HCA), K-Means clustering (KMC), principal component analysis (PCA), along with graphical piper and empirical Stuyfzand classifications, facilitates the interpretation of complex data to better understand the bottled waters quality and their classification. This combination facilitates the determination of the hydrochemical processes affecting groundwater chemistry (Alfaifi 2019).

The starting assumption is that the bottled waters are indistinguishable from each other in terms of their characteristic constituents as used by most consumers throughout the Maghreb region. So the present study aims to report the first quality assessment of bottled waters produced and sold in these countries, comparison of characteristic constituents with Maghreb, EEC and WHO legislations, and finally, water classification using empirical, graphical and multivariate statistical techniques. For the purpose of the study, the constituents' data sets of bottled waters are reported by means of their labeled physico-chemical compositions.

The ionic charge balance error considered was within 10% (Cüneyt et al. 2002; Yekdeli et al. 2010; Kloppmann et al. 2011). In this case, consequently, 11% of the water brands data were rejected.

In general, electrical conductivity and the quantity of dissolved solids are essential parameters for characterizing the differences between mineral waters (Moreno et al. 2022) but are not included in the commercial label details (Van der 2003), so they are computed for this purpose (Simler 2014).

The chemical studies of mineral, spring and table waters are intended to classify their facies and the sources of their key chemical components; they are carried out using characteristic reports and are regarded as indices of water quality (Olive 1996). As mineralization increases, the Piper plot is especially suited to studying the evolution of water facies, or to compare groups of samples with each other and to indicate the dominant cation and anion types (Piper 1944).

It is composed of two triangles, allowing the representation of the cationic and the anionic facies, and a rhombus that gives the global facies (Rasouli & Pouya. 2012). Concentrated point clouds reflect the fusion of cationic and anionic elements for the various samples (Olive 1996). The Stuyfzand classification, subdivides water into four levels according to the most relevant chemical characteristics (Stuyfzand 1986). The ‘main type’ is evaluated according to the chloride content, the ‘type’ is evaluated with respect to hardness index and the ‘subtype’ classification is determined based on the cations and the dominant anions. Finally, the ‘class’ is evaluated with respect to the sum of Na⁺, K⁺ and Mg²⁺ in meq/l (Stuyfzand 1989).

Multivariable statistical techniques can be used to test water quality data and decide if samples can be put into distinct groups that may essentially be in the hydrochemical context.

Clustering, as an approach of finding subgroups within observations, is used widely in applications like water quality classification where we try and find some structure in the data.

It is a question of grouping individuals iteratively, beginning from the bottom (the two closest individuals) and progressively constructing a tree or a ‘dendrogram’ grouping all individuals at the root into a single class (Boral et al. 2020). Hierarchical classification is used because in a class of the next partition, each class of a given partition is included.

This involves understanding how the distance between an individual and a group or the distance between two groups at each stage should be calculated or grouped. Therefore, the use of this approach needs to make an additional choice: how to decide the distance between two groups and what the aggregation criterion is to hold (Bouroche & Saporta 2006). According to the different aggregation criteria, multiple classifications will be made; the classification outcome is presented in a dendrogram (Sekiou & Kellil 2014). HCA reveals group structure when within group variation is sufficiently less than between-group variation (Worley & Powers 2013).

A clustering algorithm that divides observations into K clusters is K-means. It can be easily used in classification, as we can dictate the number of clusters, where we divide data into clusters that can be equal to or greater than the number of groups.

Based on their simplicity of implementation and fast execution (Davidson 2002; Kaushik & Mathur 2014), KMC is often used to classify water into various hydrochemical groups; this approach varies from the HCA because at the beginning of the study, the number of clusters is preselected. The KMC method will produce exactly different clusters with the greatest possible distinction. Clustering begins with k random clusters, and then transfers objects between clusters to minimize and maximize variability within and between clusters respectively.

Unlike HCA, it is not possible to present outputs of KMC in a dendrogram for a simple visual evaluation of the results; instead, the results are presented in large table that shows members of clusters and their distances from respective cluster centers (Cüneyt et al. 2002).

PCA's goal is to return to a small-size space while distorting the reality as little as possible. Therefore, the most appropriate summary of the initial data must be collected. It is the variance-covariance matrix (or that of the correlations) that will make it possible to produce this relevant summary, essentially by analyzing the dispersion of the data considered.

From this matrix, we will extract, by a suitable mathematical process, the factors we are looking for, in small numbers. This will let us achieve the desired graphics in this small dimensions space, deforming the overall configuration of individuals and initial variables as little as possible. The interpretation of these graphs would allow the structure of the analyzed data to be understood (Bouroche & Saporta 2006), using the Kaiser criterion corresponding to eigenvalues >1 to determine the number of factors to be retained (Kaiser 1960). PCA can provide the function of classification and clustering (Abdi & Williams 2010).

The strength of statistical techniques makes it better to graphical methods to grouping water samples. However, the empirical and graphical methods do, successfully, help to obtain more chemistry information about groups. By combining the two techniques, we can gain additional information that neither technique by itself can offer (Sekiou & Kellil 2014).

The descriptive statistical characteristics of position and dispersion are summarized in Table 3. It appears that, for major cations of bottled waters, the calcium contents are between 8.02 mg/l (observed in Bahia water, Morocco) and 413 mg/l (observed in Ben Haroun water, Algeria).

Table 3

Descriptive statistics of the physicochemical parameters of bottled waters in the Maghreb region

Parameter .	Mineral water .				Spring water .				Table water .
	M .	Min-max .	SD .	VC .	M .	Min-max .	SD .	VC .	M .	Min-max .	SD .	VC .
pH	7.14	6.57–7.85	0.28	3.9	7.38	6.92–8.03	0.27	3.6	6.78	6.50–7.11	0.32	4.7
EC	956.36	275–4.476	907.26	94	662.08	256–1.265	214.51	32	492.83	358.0–656.0	126.50	26
Ca2+	89.51	12.02–413	67.31	75	77.49	17.63–134.38	28.53	37	22.73	8.02–40.00	12.97	57
Mg2+	29.34	3.16–80.25	21.28	73	19.78	2.64–44	12.91	65	15.23	7.29–27.00	7.88	52
Na+	75	5.50–680	135.43	180	32.11	1.16–125.0	30.63	95	46.62	36.50–52.00	6.37	14
K+	4.02	0.10–26	5.40	134	1.99	0.21–8	1.78	89	1.84	1.00–3.00	0.94	51
	355.27	85–1.809	319.00	89	271.25	122–402.6	69.59	25	50.26	24–85.12	22.86	45
Cl−	100.11	10–1.038	176.73	176	43.74	4.00–131	35.86	82	90.78	78.22–140.0	24.44	26
	74.94	3–514	101.64	136	51.04	3.7–167.88	49.27	97	44.54	10.49–79.75	32.14	72
	7.12	0.02–34	7.92	111	11.78	0.20–46.5	8.68	74	3.48	0.20–9.18	3.13	90
	0.00	0–0.02	0.01	–	0.01	0.00–0.01	0.01	–	–	–	–	–
F−	0.64	0.3–4.8	0.30	47	0.50	0.19–1	0.27	54	0.66	0.65–0.67	0.02	3

Parameter .	Mineral water .				Spring water .				Table water .
	M .	Min-max .	SD .	VC .	M .	Min-max .	SD .	VC .	M .	Min-max .	SD .	VC .
pH	7.14	6.57–7.85	0.28	3.9	7.38	6.92–8.03	0.27	3.6	6.78	6.50–7.11	0.32	4.7
EC	956.36	275–4.476	907.26	94	662.08	256–1.265	214.51	32	492.83	358.0–656.0	126.50	26
Ca2+	89.51	12.02–413	67.31	75	77.49	17.63–134.38	28.53	37	22.73	8.02–40.00	12.97	57
Mg2+	29.34	3.16–80.25	21.28	73	19.78	2.64–44	12.91	65	15.23	7.29–27.00	7.88	52
Na+	75	5.50–680	135.43	180	32.11	1.16–125.0	30.63	95	46.62	36.50–52.00	6.37	14
K+	4.02	0.10–26	5.40	134	1.99	0.21–8	1.78	89	1.84	1.00–3.00	0.94	51
	355.27	85–1.809	319.00	89	271.25	122–402.6	69.59	25	50.26	24–85.12	22.86	45
Cl−	100.11	10–1.038	176.73	176	43.74	4.00–131	35.86	82	90.78	78.22–140.0	24.44	26
	74.94	3–514	101.64	136	51.04	3.7–167.88	49.27	97	44.54	10.49–79.75	32.14	72
	7.12	0.02–34	7.92	111	11.78	0.20–46.5	8.68	74	3.48	0.20–9.18	3.13	90
	0.00	0–0.02	0.01	–	0.01	0.00–0.01	0.01	–	–	–	–	–
F−	0.64	0.3–4.8	0.30	47	0.50	0.19–1	0.27	54	0.66	0.65–0.67	0.02	3

EC [μS/cm]. electrical conductivity at 20 °C; Ca²⁺[mg/l]. calcium ion; Mg²⁺[mg/l]. magnesium ion; Na⁺[mg/l]. sodium ion; K⁺[mg/l]. potassium ion. [mg/l]. bicarbonate ion; Cl⁻[mg/l]. chloride ion; [mg/l]. sulphate ion; [mg/l]. nitrate ion; [mg/l]. nitrite ion; F[mg/l]. fluoride ion.

Min-max. minimum-maximum values; SD. standard deviation; VC. variation coefficient.

View Large

Calcium is primarily present in carbonate rocks in water, but is also found in some igneous rock minerals (Potelon & Zysman 1998; Parizot 2008; Rodier et al. 2009). The most possible source of magnesium in water is the dissolution of carbonate rocks, magnesium minerals and soil cation exchange capacity (Potelon & Zysman 1998, Parizot 2008; Rodier et al. 2009). Magnesium concentrations vary from 2.64 mg/l (observed in Besbassa Water, Algeria) to 80.25 mg/l (observed in Oktor water, Tunisia). Sodium is subjected to the same processes of adsorption/desorption as calcium and magnesium, so its dissolution is of comparable complexity, ranging from 1.16 mg/l of sodium (observed in Mileza water, Algeria) to 680 mg/l of sodium (observed also in Ben Haroun water, Algeria).

Mainly found in igneous rocks and clays, potassium is an element rarely detected in groundwater at levels greater than 10 mg/l (Potelon & Zysman 1998, Parizot 2008; Rodier et al. 2009). The potassium concentrations observed in the studied waters are between 0.1 mg/l (at Sidi El Kebir water, Algeria) and 26 mg/l (observed in Oulmes water, Morocco). Whereas, for major anions, it appears that chlorides range from 6.2 mg/l (observed in Mileza water, Algeria) to 1,038 mg/l (observed in Oktor water, Tunisia). Chlorides in groundwater can have multiple origins, particularly water-igneous interactions (Potelon & Zysman 1998, Parizot 2008; Rodier et al. 2009). Bicarbonate contents range from 24 mg/l (observed in Ciel water, Morocco) to 1,809 mg/l (observed in Ben Haroun water, Algeria). In general, bicarbonates derive from the dissolution of carbonate minerals and the action of CO₂ in meteoric waters and soils (Potelon & Zysman 1998, Parizot 2008; Rodier et al. 2009). Finally sulfates range from 3 mg/l (observed in Salsabil water, Algeria) to 514 mg/l (Ben Haroun, Algeria). The presence of sulfates in groundwater can have two origins: the oxidation of minerals rich in sulfur, the levels measured are then from a few mg/l to a few tens of mg/l, or the leaching of evaporitic formations, the contents are then of the order of one hundred or thousands of mg/l (Potelon & Zysman 1998, Parizot 2008; Rodier et al. 2009).

These waters have nitrate contents ranging from 0.02 mg/l (observed in Sidi Driss water, Algeria) to 46.5 mg/l (observed in Arwa water, Algeria) and very low nitrite levels (the maximum is less than 0.021 mg/l). Bottled mineral, spring and table waters have fluoride contents ranging from 0.19 mg/l (observed in Melina water, Tunisia) to 4.80 mg/l (observed in Garci water, Tunisia).

Generally, the mineral water originates from groundwater; its mineralization is mostly determined by total dissolved solids (TDS) and electrical conductivity (EC). The chemical composition of the mineral water depends heavily on local geological and climatological conditions. The large dispersions in TDS and EC values of mineral waters reveal differences in the mineralogy of the aquifers nature (Sekiou & Kellil 2014). According to Cüneyt et al. (2002), similar water samples often present similar hydrological history, similar recharge areas and similar infiltration and flow paths in terms of climate, mineralogy and residence time. In any way, the water chemistry highly depends on the availability of mineralizing agents, such as temperature, CO₂ concentration, redox conditions and the type of adsorption complex (Van der 2003).

The high values of the statistical parameters of dispersion (standard deviation, coefficient of variation and extent) of the chemical components of bottled mineral waters show the great variation in the quality of these waters compared to spring and table waters (Table 3).

The contents of the various chemical elements of bottled mineral waters are lower than the limit values of the Algerian mineral water standards (Jora 2000, 2004, 2006, 2015), the Tunisian mineral water standards (NT 2007a), the European directives of the mineral waters (EEC 2003) and the WHO standards related to the mineral waters (Table 4). In the same way, the spring and table waters of the Maghreb countries comply with the Algerian standards of spring waters (Jora 2000, 2004, 2006, 2015) and those of Tunisian packaged waters (NT 2007b).

Table 4

Drinking standards for bottled waters quality

Parameter .	mineral water (limit values) .				Spring water (limit values) .		Drinking water (limit values) .
	Algerian standards JORA (2015)) .	Tunisian standards NT 09.33 (2007a) .	EU Directive EEC (2003)  .	WHO standards .	Algerian standards JORA (2015)  .	Tunisian standards NT 09.83 (2007b) .	Algerian standards JORA (2014)) .	Tunisian standards NT 09.14 (2013) .	Moroccan standards .	WHO standards 2008 Guideline values .	EU Directive EEC (1998)  .
pH					6.5–8.5		≥6.5 et ≤9	>6.5 et <8.5	6.5–8.5		6.5–9.5
EC					2.800		2.800	>300 et <2.500	2.700		2.500gv
Ca2+					200		200	200
Mg2+					150		–	100
Na+					200		200	200		200a	200
K+					20		12	–
					-		500
Cl−					500		500	500	750	250a	250gv
					400		400	500	400	500a	250gv
	50	50	50	50	50	50	50	45	50	50	50
	0.1	0.1	0.1	3	0.1	0.02	0.2	0.2 b 3 c	0.5	3	0.5
F−	5	1.5	5	1.5	2	2	1.5	1.5	1.5	1.5	1.5

Parameter .	mineral water (limit values) .				Spring water (limit values) .		Drinking water (limit values) .
	Algerian standards JORA (2015)) .	Tunisian standards NT 09.33 (2007a) .	EU Directive EEC (2003)  .	WHO standards .	Algerian standards JORA (2015)  .	Tunisian standards NT 09.83 (2007b) .	Algerian standards JORA (2014)) .	Tunisian standards NT 09.14 (2013) .	Moroccan standards .	WHO standards 2008 Guideline values .	EU Directive EEC (1998)  .
pH					6.5–8.5		≥6.5 et ≤9	>6.5 et <8.5	6.5–8.5		6.5–9.5
EC					2.800		2.800	>300 et <2.500	2.700		2.500gv
Ca2+					200		200	200
Mg2+					150		–	100
Na+					200		200	200		200a	200
K+					20		12	–
					-		500
Cl−					500		500	500	750	250a	250gv
					400		400	500	400	500a	250gv
	50	50	50	50	50	50	50	45	50	50	50
	0.1	0.1	0.1	3	0.1	0.02	0.2	0.2 b 3 c	0.5	3	0.5
F−	5	1.5	5	1.5	2	2	1.5	1.5	1.5	1.5	1.5

EC [μS/cm], electrical conductivity; Ca²⁺[mg/l], calcium ion; Mg²⁺[mg/l], magnesium ion; Na⁺[mg/l], sodium ion; K⁺[mg/l], potassium ion; [mg/l], bicarbonate ion ; Cl⁻[mg/l], chloride ion; [mg/l], sulfate ion; [mg/l], nitrate ion; [mg/l], nitrite ion; F[mg/l], fluoride ion.

a, limit for water intended for infant consumption; gv, Guideline value; b, long-term exposure; c, Short- term exposure.

View Large

Most bottled waters are loaded with bicarbonates, calcium and magnesium (Table 5). It appears, particularly, that Ca-HCO₃ facies is dominant in Algerian and Tunisian mineral waters by 70.37% and 72.72% respectively, while the majority of Moroccan mineral waters are characterized by two facies: Ca-HCO₃ and Na-HCO₃.

It appears also that all Algerian spring waters, 83.33% of Moroccan spring waters and 71.42% of Tunisian spring waters are classified under the Ca-HCO₃ type, while 83.33% of both Tunisian and Moroccan table waters are under the Na-Cl type.

According to the European Economic Community (EEC 2009), the classification of mineral waters, in terms of TDS, shows that 52% of the bottled mineral water fall in the range of ‘moderately mineralized’ (TDS > 500 mg/l), whereas 48% as ‘oligomineral’ waters (TDS < 500 mg/l).

As regard to nutritional benefits, the Maghreb waters are well known by some characteristics, Table 6 illustrates these cases, it appears that 9.5% of these bottled waters are considered containing magnesium, 6.7% containing chloride, 6.7% containing calcium 5.5% containing sodium 4% of containing sulfate, and particularly 32.87% of waters are recommended for a low-sodium diet.

Table 6

Distribution of water brands in the Maghreb countries with respect to their major characteristics

Water Characteristics .	Algeria .		Tunisia .		Morocco .
	Water brand .	Water type .	Water brand .	Water type .	Water brand .	Water type .
Rich in mineral salts (TDS greater than 1,500 mg/l)	Ben Haroun Mouzaia	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing sulfate (⁠ > 200 mg/l)	Ben Haroun N'gaous	mineral	Oktor	mineral	–	–
Containing chloride (Cl− > 200 mg/l)	Ben Haroun	mineral	Oktor Garci	mineral	Oulmes Sidi Harazem	mineral
Containing bicarbonate (⁠ > 600 mg/l)	Ben Haroun Mouzaia	mineral	Garci	mineral	Oulmes	mineral
Containing calcium (Ca2+ > 150 mg/l)	Ben Haroun Bourached	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing magnesium (Mg2+ > 50 mg/l)	Mouzaia N'gaous Ben Haroun Mensourah	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing sodium (Na+ > 200 mg/l)	Ben Haroun	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing fluoride (F > 1 mg/l)	–	–	Garci Sabrine	mineral	–	–
Suitable for a low-sodium diet (Na+ < 20 mg/l)	14 watersa	10 mineral 4 spring	5 watersb	3 mineral 2 spring	6 watersc	1 mineral 5 spring

Water Characteristics .	Algeria .		Tunisia .		Morocco .
	Water brand .	Water type .	Water brand .	Water type .	Water brand .	Water type .
Rich in mineral salts (TDS greater than 1,500 mg/l)	Ben Haroun Mouzaia	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing sulfate (⁠ > 200 mg/l)	Ben Haroun N'gaous	mineral	Oktor	mineral	–	–
Containing chloride (Cl− > 200 mg/l)	Ben Haroun	mineral	Oktor Garci	mineral	Oulmes Sidi Harazem	mineral
Containing bicarbonate (⁠ > 600 mg/l)	Ben Haroun Mouzaia	mineral	Garci	mineral	Oulmes	mineral
Containing calcium (Ca2+ > 150 mg/l)	Ben Haroun Bourached	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing magnesium (Mg2+ > 50 mg/l)	Mouzaia N'gaous Ben Haroun Mensourah	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing sodium (Na+ > 200 mg/l)	Ben Haroun	mineral	Oktor Garci	mineral	Oulmes	mineral
Containing fluoride (F > 1 mg/l)	–	–	Garci Sabrine	mineral	–	–
Suitable for a low-sodium diet (Na+ < 20 mg/l)	14 watersa	10 mineral 4 spring	5 watersb	3 mineral 2 spring	6 watersc	1 mineral 5 spring

^aSalsabil, Ifri, Chifa, Hammemet, Batna, Youkous, Ain Souda, Milok, Sidi Driss, Lalla Khadija, Besbassa, Mileza Manbaa ELghozlane Nestlé.

^bSafia Ain Mizeb, Fourat, Hayet, Palma, Dima.

^cAin Saiss, Cristaline, Chaouen, Ain Soltane, Ain Ifrane, Rif.

View Large

It is very important to notice that the entire waters characteristics are mineral waters, with the exception of waters suitable for a low-sodium diet which 44% are spring. Also five mineral waters are considered as rich in mineral salts and no waters correspond to a very low mineral content (TDS < 50 mg/l).

Referring to the legal limits, it appears that among the characteristic waters, the Ben Haroun Algerian mineral water brand exceeds the limit values of Algerian, Moroccan, WHO and EU standards of water intended for human use (N.M. 1991; EEC 1998; Jora 2011, 2014) in term of EC, calcium, sodium and sulfate; Oktor Tunisian mineral water brand exceeds these limit values in terms of EC, calcium, sodium, potassium and chloride; Garci Tunisian mineral water brand exceeds also these limit values in terms of EC, sodium, potassium and fluoride; and finally Oulmes Moroccan mineral water brand exceeds the limit value of standards in terms of sodium. The multiple exceeding of limit values of water standards calls into question the abundant and long-term consumption of these waters.

Stuyfzand classification shows that 66 bottled waters brands among 74 studied are considered oligohaline to fresh, more precisely 34 mineral waters, 26 spring waters and 6 table waters. The same classification shows that 87.84% of bottled waters brands have an alkalinity ranging from moderately low to moderately high, more precisely 35 mineral waters, 26 spring waters and 2 table waters (Table 7).

Empirical and graphical methods provide useful and quickly accessible chemistry details about bottled waters (degree of mineralization, hardness, alkalinity, salt richness, facies, etc.) and allow for modest water classification.

These techniques have a number of disadvantages; in particular: they are monovariable or at most based on a few variables, as is the case with Suyfzand's method, which does not allow explanatory graphic illustration.

The piper approach offers an explanatory visual example but uses a small number of parameters as well (major ions).These analytical and graphical methods provide modest details to differentiate between distinct groups.

Multivariate statistical techniques offer a great tool for analyzing physico-chemical data of bottled waters and determine if these waters can be categorized into separate classes. The PCA reduces the number of dimensions present in data,

The result indicates that the key variables are comprised in two principal components (corresponding to eigenvalues >1).

This hierarchical classification analysis confirms the extreme quality of Ben Haroun, Oktor, Garci, Oulmes and Mouzaia waters (Group 1 in Figure 4). In Figure 3, the point cloud reveals a continuous variation of the chemical properties of bottled waters: no coincidence between groups, we have delimited the groups as determined by the HCA.

The KMC method is used to validate the composition of each group resulting from the HCA classification, with the number of clusters selected being equal to four clusters. It appears that, KMC analysis placed more than 98% of the samples within the same groups; The KMC results are largely consistent with those of the PCA and HCA.

The application of variance analysis shows that the inter-group differences are, at the thresholds of 5%, significantly higher than the intra-group differences. It comes that groups established are distinct and significantly different (Table 8).

Table 8

Synthesis of ANOVA analysis

Parameter .	SS model .	df between groups .	SS Erreur .	df within groups .	F .	p .
EC	79696230	3	5739391	69	319.37	0.0000
Ca2+	612815	3	86059	69	163.78	0.0000
Mg2+	65358	3	4416	69	340.40	0.0000
Na+	873258	3	161272	69	124.54	0.0000
K+	1696	3	344	69	113.401	0.0000
	7543844	3	2200072	69	78.86	0.0000
Cl−	1339232	3	462462	69	66.61	0.000000
	473591	3	321866	69	33.84	0.000000

Parameter .	SS model .	df between groups .	SS Erreur .	df within groups .	F .	p .
EC	79696230	3	5739391	69	319.37	0.0000
Ca2+	612815	3	86059	69	163.78	0.0000
Mg2+	65358	3	4416	69	340.40	0.0000
Na+	873258	3	161272	69	124.54	0.0000
K+	1696	3	344	69	113.401	0.0000
	7543844	3	2200072	69	78.86	0.0000
Cl−	1339232	3	462462	69	66.61	0.000000
	473591	3	321866	69	33.84	0.000000

EC [μS/cm], electrical conductivity; Ca²⁺ [mg/l], calcium ion; Mg²⁺ [mg/l], magnesium ion; Na⁺ [mg/l], sodium ion; K⁺ [mg/l], potassium ion; [mg/l], bicarbonate ion ; Cl⁻[mg/l], chloride ion; [mg/l], sulfate ion; SS, sum of squares; df, degrees of freedom; F, F test statistic; P, p-value.

View Large

Multivariate statistical techniques are very efficient at grouping bottled waters by physical and chemical similarities, but as noted by Cüneyt et al. (2002), they are not immediately useful for identifying trends and chemical interpretation.

The drawback of multivariate statistical techniques is that they do not provide adequate information on the chemistry of the identified groups; on the other hand, the graphical and empirical methods provide useful information on the existence and chemical facies of groups.

Therefore, the combination of multivariable statistical methods and graphical and analytical techniques provides for this reason a solid means of classifying the bottled waters of the Maghreb countries, while at the same time retaining the ease of graphical representation and chemical analysis of water groups.

This article provides the first survey on the physico-chemical composition of bottled natural mineral, spring and table waters of the Maghreb countries. In addition to water quality analysis, this paper discusses the empirical, graphical and multivariate classification of these waters namely: Stuyfzand, classification, Piper classification, PCA classification, HCA classification, KMC classification and ANOVA analysis.

Referring to the Algerian, Tunisian, the European Union (EU) and the World Health Organization (WHO) standards, our survey showed that the concentrations of the different chemical parameters are lower than the limit values of the standards of mineral, spring and conditioned waters.

It appears from this study that Ca-HCO₃ facies is the dominant facies of Algerian and Tunisian mineral waters, while the majority of Moroccan mineral waters are characterized by Ca + Na-HCO₃ facies. All Algerian spring waters and the majority of Tunisian and Moroccan ones are Ca-HCO₃ facies, while the majority of table waters are Na-Cl facies. The bottled waters of the Maghreb countries are marked by some key characteristics: sulphated, chlorinated, bicarbonated, containing calcium, sodium and fluoride and referring to the diet with low sodium.

Empirical and graphical techniques provide useful and rapid information on bottled water chemistry, but these techniques are based on monovariable or bivariable analysis or, in the best cases, on the use of major ions. The classification of Stuyfzand does not group the data into chemical classes and little detail is given by the Piper diagram to distinguish between the separate groups.

Multivariable statistical techniques, unlike these classifications, are not limited by the number of parameters and are very successful in grouping bottled waters into separate groups, but do not include any water chemistry information.

For this reason, the combination of the two methods makes it possible to provide a classification methodology that allows the preservation of the advantages of each technique, thus enabling the bottled waters of the Maghreb countries to be classified in a multivariate manner, while at the same time maintaining the ease of graphic presentation and the interpretation of separate groups of waters.

The classification result shows essentially that bottled waters of the Maghreb countries can be divided into four main distinct groups. The first group includes five characteristic mineral waters rich in salts: Ben Haroun, Oktor, Garci, Oulmes and Mouzaia, with Na + K-Cl facies. The waters of the second group are marked by two facies: Ca + Mg-HCO₃ and Ca + Mg-Cl+ SO₄, while the waters of the third and the fourth groups are essentially marked by Ca + Mg-HCO₃ facies.

With respect to recommended daily amount for consumption, because the mineral content varies so widely between different kinds of mineral water, there are only guidelines for how much nutrients (calcium, magnesium, …) human should get in mineral water.

The authors are thankful to the anonymous reviewers for their careful reading and precious comments.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

The authors are not affiliated with or involved with any organization or entity with any financial interest or nonfinancial interest in the subject matter or materials discussed in this paper.

The authors declare there is no conflict.

All relevant data are included in the paper or its Supplementary Information.