Fluoride levels in deep aquifers of Makurdi, North-central, Nigeria: an appraisal based on multivariate statistics and human health risk analysis

Fluoride enrichment of groundwater has been adjudged to be a global environmental challenge in the past decade as most humans depend on groundwater for their domestic needs. This study was conducted to investigate the ionic and fluoride concentrations in borehole water and its associated health risk potentials to residents of Makurdi town and its environs, Benue state, Nigeria. Multivariate statistical techniques were for the first time used to explain the mechanisms of fluoride occurrence in groundwater in the study area. An aggregate of sixty-three (63) groundwater samples were retrieved from boreholes in twenty-one (21) diverse points within the study area and assessed for its physico-chemical composition with emphasis on fluoride content and health risk potentials following standard field and laboratory procedures. It was observed that fluoride content in the sampled water exceeded the stipulated safe limit of 1.5 mg/L in about 33.33% of the total samples and ranged from 0.34 to 2.06 mg/L with an average of 1.26 ± 0.41 mg/L. Moderate affirmative relationships were observed to exist between F− and TDS, F− and EC, F− and Cl−, and F− and NO3− in the water samples indicative of a common source pollution. Principal component analysis (PCA) revealed that high fluoride content in the water samples was associated with the dissolutions from quartzite and shale into the underlying deep aquifers as well as from contributions from anthropogenic activities including fertilizer and pesticide uses. Fluoride risk assessment indicated that the hazard quotient (HQ) for ingestion of fluoride laden water exceeded the threshold value in 66.7, 71.4, 52.4, and 9.5% of the samples for infants, children, teenagers, and adults respectively. It was found that multivariate statistical procedures such as PCA and correlation analysis (CA) are capable of establishing the relationship among groundwater pollutants, while hierarchical cluster analysis (HCA) was found suitable for explaining the likely sources/processes of pollutant enrichment in the groundwater. It is recommended that the findings of this study would serve as a basis for policy makers and regulatory bodies towards ameliorating the menace of groundwater contamination within the study area.

1 3 fertilizer and pesticide uses. Fluoride risk assessment indicated that the hazard quotient (HQ) for ingestion of fluoride laden water exceeded the threshold value in 66.7, 71.4, 52.4, and 9.5% of the samples for infants, children, teenagers, and adults respectively. It was found that multivariate statistical procedures such as PCA and correlation analysis (CA) are capable of establishing the relationship among groundwater pollutants, while hierarchical cluster analysis (HCA) was found suitable for explaining the likely sources/processes of pollutant enrichment in the groundwater. It is recommended that the findings of this study would serve as a basis for policy makers and regulatory bod-

Introduction
Aside emerging contaminants such as pharmaceutical compound, plasticizers, pesticides, flame retardants, surfactants, personal care products, and endocrine disrupting compounds that have been recently detected in water resources (Dolatabadi & Ahmazadeh, 2020;Dolatabadi & Ahmazadeh, 2018), fluoride contamination in groundwater has also been identified and recognized to be among the major global environmental pollution concerns (Mahamud, 2012). In addition to nitrates and arsenic, the World Health Organization (WHO) has classified fluoride among the contaminants of drinking water that causes severe health challenges. The WHO has fixed 1.5 mg/L as the tolerable limit for safe drinking of water-containing fluoride (WHO, 2008). This presupposes that intake of fluoride at concentrations higher than the WHO allowable limit may cause several problems namely dental and skeletal fluorosis, low intelligent quotients in infants, reduced birth rates, neurological disorders, and thyroid gland injury (Dehghani et al., 2019). Also, epidemiological research has shown that potable water is the major gateway of fluoride into living organisms.
The elevated levels of fluoride pollution in groundwater have been recorded in various places worldwide. Globally, fluoride contamination in groundwater is prevalent especially in China, India, Kenya, Nigeria, South America (Andes and western Brazil), northwest Iran, Sri-Lanka, and Pakistan Dhanya & Shaji, 2017). Continentally, studies have reported elevated fluoride rates in African nations including Uganda, Tanzania, Sudan, Malawi, Algeria, Kenya, Nigeria, Ghana, Ethiopia, and South Africa (Malago et al., 2017).
Particularly in Nigeria, the most affected regions with fluoride contamination are the northern and south-western regions of the country with proportions above WHO discharge limit of 1.5 mg/L (Akpata et al., 2009). Recently, several researchers have reported high levels of fluoride pollution in groundwater resources in various locations in Nigeria, in which ninety percent of the residents depend on for domestic uses (Amadi et al., 2015;Bura et al., 2018;Dibal et al., 2016;Emenike et al., 2018a, b;Goyit et al., 2018;Gwaha, 2017;Malum et al., 2019;Okunola et al., 2016;Olasehinde et al., 2016).
Furthermore, studies have reported moderate to high prevalence rates of fluoride related diseases such as (skeletal and dental fluorosis) in various regions in Nigeria, thus substantiating fluoride contaminations of drinking water in such areas (Dirisu et al., 2016;Ephraim-Emmanuel et al., 2016;Fulata et al., 2017;Idon & Enabulele, 2018). The drinking water which is sourced from groundwater and consumed untreated in most parts of Nigeria has been identified as the main cause of fluoride linked diseases owing to neglected fluoride contamination of such water (Uriah et al., 2014).
Recently, scholars have imbibed the Human Health Risk Assessment (HHRA) methodology as well as the use of multivariate statistics to ascertain the relationship existing among the diverse nature and number of water pollutants with a view to identifying the actual source of pollution and the possible mitigation measures to be adopted. Some of these studies cut across areas such as Iran (Ashrafi et al., 2020;Pazand, 2016), Tunisia (Guissouma et al., 2017), Ghana (Ganyaglo et al., 2019;Salifu et al., 2012), India (Ahada & Suthar, 2019;Chabukdhara et al., 2017;Ghosh & Mondal, 2018), and Ethiopia (Haji et al., 2021) to mention a few.
In northern Nigeria, most studies have merely reported elevated fluoride levels in groundwater without a corresponding human health risk analysis and have failed to succinctly identify the likely sources for fluoride contamination of groundwater (Amadi et al., 2015;Bura et al., 2018;Dibal et al., 1 3 2016; Goyit et al., 2018;Gwaha, 2017;Malum et al., 2019;Okunola et al., 2016;Olasehinde et al., 2016). In southern Nigeria, Emenike et al. (2018a, b) and recently Egbueri and Mgbenu (2020) adopted similar approaches to reveal the fluoride levels and associated human health risk in Abeokuta town and Ojoto province respectively. Their findings provided very useful baseline data that could aid practical policy making in water supplies and management in the areas. Thus, it is evident that owing to the diverse nature of groundwater pollution across the globe, it is technically pertinent to ascertain the quality of groundwater in every urban or peri-urban environment in order to ascertain the risk associated with the consumption of such water. The adoption of the foregoing inference would assist in preserving the health of the residents of such environments.
In Makurdi metropolis, like many other urban and peri-urban centres in Nigeria, it is impracticable to access treated portable water as a result of poor provision of efficient water treatment systems and lack of stringent water quality standards. This has worsened the situation due to the population explosion in these areas (Chia et al., 2014;Emenike et al., 2016: Odjegba et al., 2014. The available surface water supplies in these areas are either insufficient or are poorly treated and may be contaminated with disease-causing microbial consortiums. Intrinsically, over 70% of residents in Nigerian urban and peri-urban communities have resulted to the use of borehole water for their domestic needs including those for drinking with the specious believe that groundwater is of higher quality (Adekunle et al., 2013). However, following the observed contamination of such groundwater with fluoride and its associated health risk as reported in other climes, it is highly desirable to comprehensively assess the quality of groundwater with regard to fluoride concentrations in Makurdi. This is highly essential in an area like Makurdi where groundwater is the major source of drinking water for the inhabitants. A comprehensive assessment study as intended in the current work will be beneficial to elucidate on the inherent risks to humans in relation to fluoride concentrations of groundwater and also facilitate in decision making in water supplies and regulations.
Consequently, this study for the first time attempts to extensively assess the quality of groundwater sourced from boreholes in 21 locations in Makurdi metropolis and its environs, North-central, Nigeria, during the peak of raining season (October, 2019) with emphasis on its fluoride content and associated human health risk potentials. Firstly, this study seeks to specifically reveal the spatial variations of fluoride and other groundwater quality factors in Makurdi metropolis. Secondly, multivariate statistical approach was employed to uncover the probable sources of fluoride contamination in the water samples as well as establish the interactions that exist among the water pollutants. Finally, the study seeks to adopt the standard risk assessment procedures as recommended by USEPA (2011) to establish the potential human health risk associated with the consumption of fluoride laden groundwater in the study area. The findings of the current study would be beneficial in safeguarding the lives of residents of Makurdi from the adverse consequences of fluoride contamination of groundwater.

Description of the study area
The current study was carried out by sampling groundwater from deep aquifers (boreholes) from 21 locations in Makurdi metropolis of Benue state-Nigeria. The map of Makurdi metropolis and the sampling points is as displayed in Fig. 1, while Table S1 (supplementary material) presents the coordinates of the sampling points.
Makurdi lies between latitude 7.733°N and longitude 8.5391°E of Greenwich meridian, which is within Southern Guinea Savanna Zone, characterized by distinct wet and dry seasons. The area has experienced a steady rise in urbanization activities in recent times which have also transformed most of its natural endowments.
The rainfall characteristics classified the patterns into dry season (low amount of rainfall) in November-April and wet season (high amount of rainfall) in May-October. The annual rainfall in Makurdi is estimated by the National Oceanic and Atmospheric Administration to be 1,237 mm (NOAA, 2016), while the period from November to January (hamartan weather) is relatively cool (Isikwue & Onyilo, 2010).

3
The landform is moderately undulating as revealed by the topographical map of the area. The total annual potential evapotranspiration (PET) is estimated at about 2602 mm, with mean annual relative humidity of about 40.7%.
The soil properties of this province are dominated by Makurdi sandstone which is part of the sedimentary basin in Nigeria in the same proportion (Isikwue & Onyilo, 2010) and slightly acidic (pH: 4.5-6). The sandstones in this zone are generally fine-to medium-grained, moderately sorted, micaceous, and feldspathic. In some parts, they are calcareous, micaceous, and shelly. Various types of cement like iron oxides, silica, dolomites, carbonates, and clay were shown to be present in the Makurdi sandstone (Akuh, 2014;Iorliam et al., 2013). In 2007, Makurdi population was estimated at 500,797 people (The World Gazetteer, 2020).

Sample collection and field/laboratory tests
Samples were collected in triplicates in the month of October 2019 from each of the 21 sampling points (P1-P21), giving a total of 63 samples. In each location, the samples were collected from taps connected to borehole supplies usually from an overhead tank. Prior to obtaining samples, the taps were opened and allowed to run onto the ground for about 10 min; then, samples were collected in freshly purchased 1L capacity polyethylene jars.
The freshly purchased polyethylene jars were initially washed with a mixture of distilled water and nitric acid (15%), rinsed again with distilled water alone and air dried prior to taking them for sampling campaign. At the sampling points, the jars were rinsed thrice with the borehole water before obtaining the samples. The collected samples were accurately labelled with the use of masking tapes and marker and immediately conveyed in an ice loaded cooler to the laboratory for tests, while the samples where refrigerated at 4 °C prior to the test to avoid sample deterioration.
Some of the water quality parameters including pH, electrical conductivity (EC), and total dissolved solids (TDS) were determined in the field immediately after the samples were collected. During the field tests, TDS was determined using a TDS meter (HM Digital TDS-4 Pocket Size), pH with a pH meter (Hanna Model HI98107) and EC with a conductivity meter (Suntex model SC-120).
The nitrate (NO − 3 ), bicarbonate (HCO − 3 ), chloride (Cl − ), carbonate (CO − 3 ), sulphate (SO 2− 4 ), sodium (Na + ), magnesium (Mg 2+ ), calcium (Ca 2+ ), and potassium (K + ) were measured following standard methods APHA (2005). The fluoride levels was measured potentiometrically using an ion selective electrode (Orion model 25,100) coupled to a multimeter (Xplorer GLX model PS-2002), and with the use of a Total Ionic Strength Adjustment buffer (TISAB I) in the ratio of 1:1 (ASTM Method D9214, 1996). Researchers have recently preferred such sensor-based techniques for quantification of water contaminants owing to their simplicity of use and accuracy of results Avazpour et al., 2020;Badakhshan et al., 2019). All laboratory tests were carried out in triplicate and mean values reported in order to ensure the reliability of the results.

Hydrogeochemical facies and water quality indices
The hydrogeochemical facies of the study location was evaluated using piper diagrams to understand the water type (composition) in terms of the ion (anions and cations) balances. This was performed using Aquachem software version 2014.
The Water Quality Index (WQI) is a recently adopted method for revealing the quality of water at a glance. Computation of WQI in the current study was in accordance with the method of Mahmud et al. (2020). The procedure involved three successive steps: The first step was 'assigning weight'. Each of the 13 parameters except carbonate was assigned weights (wi), according to its relative importance in the overall drinking water quality. The most significant parameters were given a weight of 5 and the least significant a weight of 3. In this study, the maximum weight of 5 was assigned for fluoride, TDS, pH, EC, chloride, and nitrate. The sulphate and potassium were assigned weights of 4, while the less harmful elements, such as sodium, calcium, magnesium, and bicarbonate, were assigned weights of 3 each (Mahmud et al., 2020). The carbonate was not assigned a weight as it is not regulated by the WHO and was excluded from the computation of WQI.
The second step was the calculation of relative weights for each water quality feature. The relative weight (Wi) was computed using Eq. 1 (Mahmud et al., 2020): where Wi is the relative weight, wi is the weight of each parameter, and n is the number of parameters.
The third step was the computation of the quality rating scale. The quality rating scale (qi) for each parameter was calculated using Eq. 2.
where qi is the quality rating; Ci is the concentration of each chemical parameter in each water sample in mg /L, except pH; and Si is the WHO standard for each chemical parameter and the water quality classification are as presented in Tables S2 and S3 in supplementary material. Finally, the Wi and qi was used to compute the SIi for each chemical parameter (Eq. 3), and then the WQI is calculated using Eq. 4.
where SIi is the sub-index of each parameter, qi is the rating based on concentration of each parameter, and n is the number of parameters. The computed WQI values for each location were categorized following the standard WQI classification scheme as depicted in Table S3, while the detailed computations for the WQI for each location and parameter are detailed in Table S4 ( (1)

Human health risk assessment for fluoride contamination
The human health risk assessment is used to account for the nature and probability of antagonistic wellbeing impacts on publics who may be vulnerable to undesirable substances in polluted ecological media, presently or in some other time. Therefore, in this study, the numerical health risk assessment of fluoride via intake of water was assessed in an urban population of Makurdi city, North-central, Nigeria, for the first time. In reflection of such interest, water samples collected from boreholes were taken from diverse locations within the city. The population was split into four age categories on the basis of physical and social variances in accordance with the method of Yousefi et al. (2018). Thus, the population was divided as follows: infants (< 2 years), children (2 to < 6 years), teenagers (6 to < 16 years), and adults (≥ 16 years). Exposure to fluoride was estimated using Eq. 5 (Radfard et al., 2018).
Estimated daily intake (EDI) of fluoride (mg/ kgB x/ /day) was estimated on the basis of the daily mean intake of portable water (C y ), amount of fluoride in drinking water (C j ), and body weight (B x ). The water intake statistics was appraised using questionnaire that were administered on the target populations (infants, children, teenagers, and adults). Data on the average body weight of the various populations were computed from the records department of an epidemiological centre in the study area where most residents are vaccinated from infancy to adulthood. The data collected for this computation was limited within the last 5 years (2015-2020) to reflect current realities.
The mean water intake levels in infants, children, teenagers, and adults were found to be 0.50, 1.0, 2.0, and 2.5 L day −1 respectively. Average body weights of the target groups were obtained to be 10, 18, 45, and 70 kg for infants, children, teenager, and adults respectively.
The non-cancer-causing risk of fluoride to human health was estimated in form of a hazard quotient (HQ) using Eq. 6 (Radfard et al., 2018).
The Oral Reference Dose (RfD) is an approximation of the day-to-day exposure of the human population to a substance that is not likely to be of any significant harmful risk effects in a lifetime. In this study, the RfD for fluoride was taken to be 0.06 mg/kg/day (IRIS, 2017;USEPA, 2011). The hazard quotient (which is described as the ratio of EDI to RfD) when obtained to be < 1 is assumed to be unlikely significant for exposed persons to experience any adverse health effects. However, if the hazard quotient is found to be > 1, it is an indication of a likely adverse health effect on the exposed population (Yousefi et al., 2018).

Statistical analysis of data
The results of laboratory test were firstly analyzed using descriptive statistics. The extent of violation of the WHO acceptable threshold for each water quality characteristics was also ascertained. Then the Pearson's pair wise correlation was used to elucidate on the relationship between the variables (water quality parameters) at the 0.05 level of significance.
The principal component analysis (PCA) was used to reduce the dimensionality of the water characteristics to facilitate rapid inference of results. PCA was done on the 13 water quality parameters using the Statistical Package for Social Sciences (SPSS) version 21 by the extraction of principal factors using the Varimax orthogonal rotation technique.
Hierarchical cluster analysis (HCA) which is a commonly used approach for grouping a set of interrelated data sets was then applied to the data (both for water quality features and locations). The Ward's method as described in Emenike et al. (2018a, b) was adopted for the HCA analysis in the present study using the SPSS statistical software version 21. The hydro-chemical facies of the water samples were elucidated via Piper, Schoeller, and Durov diagrams drawn using Aquachem (Aq*QA) version 2014 software.

Spatial variations in physico-chemical parameters of Makurdi groundwater
The characteristics of raw groundwater for all 21 locations are presented in Tables S5 and S6 in 1 3 supplementary material. The descriptive statistics is presented in Table 1. From Table 1, it was observed that majority of the samples were slightly acidic as the pH was noticed to range from 6.2 to 6.8 (mean = 6.52) across the 21 sampling points as can be seen in Fig. 2. Generally, almost 52.4% of the samples had pH levels that fell below the permissible limits for drinking water as stipulated by WHO (WHO, 2017), while the remaining 47.6% complied with the WHO standard. The slightly acidic nature of the water which was attributed to the extensive use of fertilizers and pesticides by famers in the water shed may negatively impact metabolism in humans (Chabukdhara et al., 2017).
Total dissolved solids (TDS) and electrical conductivity (EC) ranged from 89.80 to 1471.50 mg/L and 158.40 to 2487.40 µs/cm ( Fig. 2) with mean values of 708.02 mg/L and 1233.46 µs/cm, respectively. It was found that 28.57 and 38.10% of the samples violated the WHO limits of 1000 mg/L and 1500 µs/cm for TDS and EC, respectively. Although high concentrations of TDS and EC in drinking water may not be of any severe health implications, such elevated levels can affect the acceptability of such water for drinking purposes as a result of the high scaling effects of such water in household utilities and pipes (WHO, 2017).
Additionally, high TDS levels in water can cause gastrointestinal disorders in persons with kidney or renal problems (Chebet et al., 2020). The high levels of TDS and EC in the sampled water can be attributed to the extensive use of fertilizers and pesticides, deep infiltration of rainwater, dissolved sediments, and ion exchange (Egbueri & Mgbenu, 2020;Emenike et al., 2018a). It can also be specifically linked with the dissolution of bicarbonate, sulphate, sodium, calcium, and magnesium in the water from natural and anthropogenic processes as seen in the high levels of sulphate, sodium, and chloride ions in the water samples (Table 1). The high levels of these ions in the samples were also found to be responsible for groundwater salinity in the area (Chebet et al., 2020).
Similarly, chloride was found to range from 38.10 to 532.20 mg/L with average of 165.10 mg/L. Furthermore, about 28.57% of chloride concentration in the water samples exceeded the WHO limit of 200 mg/L. It should be noted that elevated chloride concentrations in water can induce a salty taste in the water, depending on the associated cation; the salty taste of the chloride anion could be triggered if the concentration falls between 200 and 300 mg/L for sodium, calcium, and potassium (WHO, 2017). The elevated concentrations of chloride in water could be connected with the release of industrial, commercial, and agricultural effluents and runoff (containing de-icing salts) into the environment as well as with the natural solubility of chloride bearing minerals (salts) in water. Similar observations were also reported in the study of Emenike et al. (2018a, b) and that of Asuma et al. (2020).
Carbonate and bicarbonate varied from 0.00 to 196.00 mg/L and 95.00 to 601.60 mg/L with average values of 18.70 mg/L and 331.56 mg/L, respectively. Although there are no strict regulations for carbonate content of drinking water, it was found that about 9.52% of the samples exceeded the WHO stipulated limit of 500 mg/L for bicarbonate in water. Activities relating to geological mineralization, irrigation, runoff, and seepage which mix up with groundwater during recharge may be responsible for the heightened levels of bicarbonate in the samples. Similar findings have been reported in previous studies (Akoteyon, 2013;Rasool et al., 2016).
Sulphate concentrations in the samples fell between 39.60 and 492.10 mg/L with an average of 283.88 mg/L, which was higher than the WHO acceptable limit (250 mg/L) for drinking water. In overview, about 71.4% violations were noticed from the 21 samples with respect to the WHO stipulated limits (250 mg/L) for sulphate ions in drinking water. Excessive levels of sulphate in water can lead to laxative effects in humans, while moderate levels could induce salty taste in drinking water depending on the associated cations. The foregoing is asserted to be responsible for the salty (saline) taste of water from deep aquifers within the study area as the excess sulphate interact with sodium cation to form sulphate salts. Furthermore, excessively high values of sulphate in drinking water may induce respiratory ailments in humans (Emenike et al., 2018b). The elevated sulphate concentration is also associated with Nitrate concentrations in the water sampled was observed to range from 0.58 to 16.39 mg/L with a mean concentration of 5.88 mg/L. Nitrate is a strictly regulated compound in drinking water due to its health implications. It causes the blue baby syndrome (methamoglobinemia) in infants and children when present in drinking water at levels higher than 50 mg/L (Sellami et al., 2019). Despite the excessive use of agro-chemicals and fertilizers by farmers in the study area, it was observed in this study that none of the samples exceeded the threshold value. Thus, the moderate/low nitrate levels in the groundwater may be linked to the removal of dissolved nitrate by filtration and adsorption activities of the soil profile during seepage and deep percolation (Emenike et al., 2018b).
Fluoride concentration varied from 0.23 to 2.03 mg/L with an average concentration of 1.26 mg/L. Most of the water samples complied with the minimum (0.5 mg/L) and maximum (1.5 mg/L) threshold values for fluoride in drinking water (WHO, 2017). Succinctly, about one-third (33.33%) of samples were found to exceed the maximum permissible limits, while only one of the samples had fluoride concentrations lower than the minimum acceptable limit (0.5 mg/L). High fluoride levels in water samples can be linked to the fluoride bearing rocks found in the geological formations, and myriad of anthropogenic activities in the study area. Fluoride concentration in water above 1.5 mg/L is reported to be responsible for dental fluorosis in humans (especially children), while excessive concentrations are reported to cause skeletal fluorosis in children and adults (WHO, 2004). It is usually referred to as the double edge sword as low intakes of fluoride in humans can also cause dental caries (Yousefi et al., 2018). Based on the reports from recent studies, it is presumed that excessive intake of fluoride can be carcinogenic in humans (WHO, 2004). Thus, fluoride removal from drinking water supplies is highly essential in the provision of potable drinking water.
Sodium concentrations in the samples ranged from 34.30 to 403.1 mg/L (with mean value of 211.94 ± 98.63 mg/L). There was an evident high variability in the sodium content of the water with about 57. 14% of the samples surpassing the WHO limits of 200 mg/L. The high sodium level in the water was attributed to aquifer chemistry and the mixing of wastewater effluent (generated through human activities) with the groundwater in the study locations (Tirkey et al., 2017).
Potassium, calcium, and magnesium concentrations in the water samples were found to range from 1.47 to 11.49 mg/L, 45.80 to 221.00 mg/L, and 28.60 to 237.70 mg/L, respectively. Potassium concentrations in the samples were found to fully comply (100%) with the WHO limits (50 mg/L). It should be noted that Mg + and Ca + concentrations in water are major pointers to its hardness. In the current study, about 47.6% of the samples surpassed the WHO most desirable limits of 75 mg/L for Ca + in drinking water, while 14.3% of the samples had Mg + concentrations above the WHO permissible limit (150 mg/L) and over 90% of the samples had values higher than the most desirable limits of 50 mg/L. It was inferred therefore that groundwater in most of study locations was generally hard and tasty. This was confirmed further by ingestion of such water samples. High levels of these cations in the water samples were traced to mineral dissolution in the basement and ion exchange activities (Magesh et al., 2017).

Water quality assessment
The quality of water samples was assessed through the Water Quality Index (WQI) and water type (hydrogeochemical facies) using piper diagram consisting of composite-double triangles and a diamond shape. Each of the triangles represents either the cation or anion composition of the water samples, while the diamond shape is a composite plot for both the cations and anions.
For the WQI, weights and relative weights were assigned to the water quality parameters as depicted in the supplementary material (Table S1). The data in Table S3 is the computations of the WQI and remarks for all the studied locations, and the results are presented in Fig. 3. It was observed that about 95% of the samples fell into class II (good water), see Table S2, while one sample (P14) had quality that fell into class I of the WQI classification, which suggest that the water is clean and safe. Furthermore, it was noticed that the major contributors to the overall WQI in the study locations were in the order of sulphate > EC > chloride > calcium > fluoride > TDS > pH > sodium. Magnesium and bicarbonate were observed to have moderate contributions to the water quality, while nitrate and potassium had the least contributions. The findings of this study are in tandem with that of Adimalla et al. (2018).
Although most of the samples were found to be good for drinking, the threshold values obtained for most locations (P1, P2, P4, P6, P8, P13) were observed to be very close to been ranked as class III water type ("poor water"). Thus, it is suggested that simple technologies such as filtration and adsorption could be deployed as treatment techniques for water in the affected locations to eliminate the major contributing pollutants such as sulphate, chloride, fluoride, calcium, and sodium, among others. In light of this, the affected water may be safer for human consumption.
Furthermore, the piper diagram ( Figure S1) shows that the Na + , Ca 2+ , and Mg 2+ are the major cations with Ca 2+ and Na + dominating, while K + was found to have a moderate contribution to the general description as indicated the in WQI computations. In the anions sections, SO 4 2− and Cl − were observed to dominate the facies, while contributions from HCO 3 were found to be moderate. The dominance of Ca 2+ and Na + in the water composition is an indicator of the possibility of ion exchange through rock weathering in the aquifer. The dominance of SO 4 2− and Cl − on the other hand reveal ion contributions as a result of silica weathering (Achary et al., 2016). Most notably, dissolution of trace gypsum and halite was found to be the major sources of Ca/SO 4 and Na/Cl in the studied groundwater samples.
Additionally, Aghazadeh et al. (2016) posited that when the ratio of calcium/magnesium ions in water is between 0.6 and a value higher than 2, it is a pointer to the possible dissolution of dolomite and silicate rock constituents in the aquifer. In the current study, more than 80% of the samples had calcium/magnesium ratios that fell between 0.6 and 3.4 confirming the interaction of silicate and dolomite materials in the aquifer. The overall characterization of the hydrogeochemical facies revealed the dominance of Na-SO 4 (42.8%), Na-Cl (14.30%), Mg-HCO 3 (14.30%), Mg-SO 4 (9.5%), Ca-HCO 3 (4.8%), Mg-Cl (4.8%), Ca-Cl (4.5%), and mixed water types (5%) in the studied locations (Fig. 4). The foregoing is indicative of the predominant salty taste of borehole water in the study area. Additional information on the hydro-chemical characterization of the water samples in form of the Durov and Schoeller diagrams can also be found in supplementary material ( Figures S2 and S3).

Multivariate statistical analysis and pollution source apportionment
In the current study, Pearson's correlation, principal component, and hierarchical cluster analyses were adopted to further corroborate the relationship and likely sources of contaminants in groundwater of the study area. These techniques have been utilized in several studies on the water quality assessment (Ali & Ali, 2018;Emenike et al., 2018b;Rasool et al., 2016;Ravikumar & Somashekar, 2017;Salifu et al., 2012). For correlation analysis (CA) when r values < 0.3, the association was taken as weak correlation; when it varied between 0.3 and 0.7, the relationship is regarded as being moderate; and when r was > 0.7, the relationship is considered as strong (Aravinthasamya et al., 2019;Emenike et al., 2018b;Salifu et al., 2012).

Correlation analysis
The results of correlation analysis (CA) are presented in Table 2. With regard to the relationship between fluoride and other ions in the water samples, it was found that fluoride was moderately and positively correlated with TDS, EC, Cl − , and NO 3 − . This suggests that these parameters have similar sources with fluoride, which occurs from rock weathering and geological interactions. For instance, Bempah (2014) reported similar relationship between fluoride, TDS, and EC in his study on arsenic contamination of groundwater in Ghana. Aravinthasamya et al. (2019) also reported such correlation of fluoride with TDS, EC, and Cl − in their separate study in India. Moderate positive correlation of fluoride and nitrate in the current work suggests that aside geogenic contributions, anthropogenic activities such as the use of chemical fertilizers are also responsible for fluoride enrichment in groundwater.   Similarly, F − was observed to be positively and weakly correlated with SO 4 2− , CO 3 − , pH, and all the cations except for Ca 2+ . This indicates that the sources of these contaminants in the water are probably from geological interactions. It was however observed that F − had a negative and weak correlation with HCO 3 − and Ca 2+ , signifying that they were likely from different sources. HCO 3 − and Ca 2+ were from calcite precipitation, while F − was contributed mainly from the dissolution of fluoride bearing minerals and the use of fertilizers and pesticides. Similar observations were reported by Emenike et al. (2018b) in addition to that of Adimalla & Venkatayogi (2017).
Also, strong positive correlations were observed for EC-TDS, Na + -SO 4 , Ca 2+ -Na + , and Ca 2+ -K + (Table 2), thus indicating source similarities of the parameters. No negatively strong or moderate correlations were observed; however, most of the parameter (pH-Cl − , pH-CO 3 − , pH-HCO 3 − , NO 3 − -Mg 2+ among others) exhibited a weak negative relationship suggesting that these parameters were introduced in groundwater samples from different routes (Table 2).
Comparatively, fluoride concentrations in the current study were observed to be higher than the mean values reported from other studies (Emenike et al., 2018b;Salifu et al., 2012;Shen and Schafer, 2015;Ravikumar & Somashekar, 2017). It is important to note that fluoride content in groundwater tends to increase as the TDS and EC values increase (Rafique et al., 2009). This trend was also observed in the current study. Although fluoride was found to be weakly correlated with pH in this study, the findings support the assertion that fluoride solubility is pH dependent (fluoride solubility in water is favoured when the pH ranges from 6 to 9).
The pH of the samples were found to range between 6.2 and 6.83 in the current study with an average of 6.52 which is likely responsible for the high fluoride values recorded (Saxena & Ahmed, 2001). Furthermore, the presence of fluoride in the water samples could be linked to the dissolutions from quartzite and shale (which are the major geological sediments of the study locations) into the underlying aquifer (Eneji, 2011).

Principal component analysis
In this study, PCA revealed four principal components (PC) with a cumulative variance of 71.53% (rotation sum of squared loadings) as illustrated in Table 3 and the scree plot ( Figure S4). It was observed from Figure S4 that about 13 factors were extracted in total but only four of them were seen to have Eigen values > 1 and they were considered as the principal components (PC). PC1 had strong and positive factor loadings for sulphate, sodium, calcium, potassium, and magnesium with an explicated variance of 24.7% (Table 3). This also indicates the contribution of rock weathering and ion exchange activities in the hydro-geo-chemistry of the studied aquifers, which also implies that the cations originate from similar geogenic sources (dolomite, shale and gypsum).
PC2 was observed to have strong positive loadings for pH, TDS, EC, and fluoride with a total explainable variance of 19.6%. This indicate that the occurrence of these contaminants in water owing to similar natural and/or anthropogenic activities. High TDS, EC, and fluoride levels are controlled to a large extent by high evaporation of groundwater, especially in a flat terrain characterized by slow groundwater flows (Gao et al., 2020). The foregoing relationship also implies the role of pH in the dissolution and mobility of anions (especially fluoride) and cations in water, as the solubility of fluoride in water is favoured in the pH range of 6-9 as earlier stated (Saxena & Ahmed, 2001). Since both TDS and EC values are major indicators of ion concentrations in water, in this study, it was observed that sodium and sulphate ions were the major contributors to the TDS and EC values of groundwater and as such the strong positive loading of fluoride, TDS, and EC in PC2 is a further confirmation that fluoride enrichment of groundwater is chiefly influenced by the dissolution of basement rocks. PC2 also confirms the positive correlation of TDS and EC with fluoride in water and further substantiate the assertion that fluoride concentrations in water tends to increase with upsurge in TDS and EC (Rafique et al., 2009). The component plot in rotated space illustrates the factor loadings for the principal components as shown in Figure S5. PC3 and PC4 were found to have negligible contributions to the water chemistry. PC3 had strong positive factor loadings for chloride, carbonate, fluoride, and magnesium with an explainable variance of 13.74%. The dominance of carbonate in PC3 indicates its contribution in the dissolution and distribution of fluoride and chloride in groundwater. Nitrate and fluoride had high positive factor loadings in PC4 with an explainable variance of 13.47% suggesting anthropogenic contributions. PC4 represent the minor anthropogenic (use of fertilizers and pesticides) contributions of this parameters to the groundwater quality.
It was also observed that fluoride and magnesium exhibited a quasi-independent behaviour in PC1, PC2, PC3, and PC 4, which implies that they are introduced into the water majorly from natural (geogenic) and on a minor note through anthropogenic sources (Hosseini et al., 2019).

Hierarchical cluster analysis
The HCA were performed in this study using the Ward's method by selecting the squared Euclidean distance as the interval and the hydro-chemical results of all samples were statistically analyzed. The data used was normalized before the analysis. The HCA was performed in the Variable-and Case-modes, such that the Variable-mode provided the relationships between water quality parameters, while the Casemode revealed the relationships between the sample locations (Bempah, 2014;Sharaf & Subyani, 2011).
The HCA was performed for the set of water samples from 21 locations with three replications each (giving rise to 63 cases) and 13 water quality parameters (variables). The result of HCA in form of a dendrogram for the Variable-mode is shown in Figure S6, while that for the Case-mode is depicted in Figure S7. It was observed that the variables were grouped into two clusters. The parameters were observed to be grouped on account of their concentrations in the water samples. Cluster 1 was further divided into 3 sub-clusters. The pH, K + , NO 3 − , and F − formed the first sub-cluster in cluster 1, which indicated that both natural and anthropogenic activities were responsible for the occurrence of the parameters in groundwater. The second sub-cluster in cluster 1 comprised of CO 3 − , Ca 2+ , Mg 2+ , and Cl − which shows that these ions are present in groundwater as a result of anthropogenic contributions. SO 4 2− , Na + , and HCO 3 − were grouped in the third sub-cluster in cluster 1, indicating that the sources of these ions were from rock weathering and dissolutions (geogenic activities). TDS and EC formed the second and most important cluster which suggest the dominance of natural processes such as evaporation which also promoted the groundwater salinity (Aravinthasamya et al., 2019). Despite TDS and EC having relatively higher linkage as compared to the other variable, they were found to be closely related owing to their strong positive correlations with most of the studied parameters (Bempah, 2014) and were proven to be the major indicators of groundwater quality in the area.
Furthermore, Na + , SO 4 2 , and HCO 3 − were observed to be closely connected which indicated the impacts of rock weathering and ion exchange activities on the water chemistry as earlier shown in the correlation analysis, PCA and the hydrogeochemical facies of the studied locations. The pH, Cl − , CO 3 − , NO 3 − , F − , K + , Ca 2+ , and Mg 2+ were also found to be closely clustered suggesting source similarities and the involvement of pH in the mobility and speciation of the ions in groundwater. The close linkage distance between NO 3 − and F − further confirms that fluoride is contributed to the groundwater from the use of fertilizers.
On the other hand, it was observed that the sample locations were grouped into four clusters and two groups on the basis of the similarities in the water chemistry of the locations. Locations P1, P2, P3, P4, P5, P6, P8, P10, P12, and P18 were grouped into clusters 1 and 2 (saline water), while locations P7, P9, P11, P13, P14, P15, P16, P17, P19, P20, and P21 were grouped into clusters 3 and 4 (non-saline water). The implications of the groupings may be linked to the close proximity of the sample locations as earlier stated in the coordinate locations. In general, there was no significant distinctions in the water quality of the studied locations as the samples were observed to have similar characteristics in terms of its overall quality except for salinity levels. It was further observed that the depths of the boreholes were well correlated with the salinity levels. This was because most of the boreholes grouped as "saline" were deeper (30-40 m), while the non-saline ones were observed to be shallower (15-25 m). The foregoing was also an indication that groundwater salinity in the area was caused by dissolution of basemen rock minerals.

Human health risk analysis for fluoride contamination of groundwater
Despite the seemingly good quality of the groundwater for drinking purposes as demonstrated in the results of the WQI analysis, it is essential to evaluate the potential health risk of the exposed populations to fluoride contamination. This is apparent as some fractions of the samples were observed to contain fluoride concentrations that exceeded the allowable limits for drinking water in order to prevent fluorosis in humans.
The results of the estimated daily fluoride intake for each location is presented in Figure S8, while that for the hazard quotient per location for every human category is shown in Fig. 4. Figure S8 represents the exposure levels of the populations (infants, children, teenagers, and adults) to fluoride risk, while Fig. 4 depicts the hazard quotient (an indicator of actual risk) for risk of fluorosis in the vulnerable population. Details of the computations for the risk assessment are available as supplementary material (Table S7).
From Figure S8, it was observed that the EDI Values were considerably high in most of the locations especially in infants and children. A moderate level of EDI was observed in most locations for teenagers, while adults in all locations except one (P13) were found to have EDI levels for fluoride that were lower than the RfD value of 0.06 mg/kg/day. Generally, EDI values ranged from 0.023 to 0.100, 0.025 to 0.110, 0.020 to 0.088, and 0.016 to 0.071 mg/kg/day for infants, children, teenagers, and adults, respectively. Among the studied cases locations, P14 and P13 had the lowest and highest EDI values respectively, which follow after the mean fluoride concentrations of these locations.
From the computations of the hazard quotient (HQ) (Fig. 4), it was noticed that for infants, about 66.7% of the sample locations had HQ values in excess of the set limit (1). This implies that this population is vulnerable to high risk of fluoride related disorders due to ingestion of fluoride laden groundwater. Similarly, with respect to children, teenagers, and adults, it was observed that about 71.4, 52.4, and 9.5% of the sample locations were at elevated risk of developing fluoride related diseases from the consumption of high-fluoride content water.
It is evident that children had the highest potential risks for developing fluorosis even higher than that for infants, since the ratio of the daily water intake rates and body weight for the children is considerably higher than that for the infants. Moreover, this is expected as the children undergo more metabolic activities as a result of mobility and play which demands for more water consumption as compared to the infants. As for the teenagers and adults, the risk as earlier stated could be moderate and low respectively. This could be attributable to the ratio of the daily water intake rates and body weight, coupled with the associated metabolisms. As expected, teenagers are more prone to high metabolic rates and thus higher water intake rates irrespective of their moderate body weight.
Despite the moderate and low risk levels for teenagers and adults respectively when considering the entire study location, it is pertinent to note that locations P13 and P19 had very high risk levels for all age categories, and therefore, this study recommend that groundwater from these locations should be subjected to affordable defluoridation systems prior to consumption. Several studies have reported analogous risk levels for fluoride in drinking water in many parts of the world (Emenike et al., 2018b;Karunanidhi et al., 2019;Nakazawa et al., 2020;Radfard et al., 2018;Yousefi et al., 2018).

Conclusions
This study successfully assessed groundwater in Makurdi metropolis with emphasis on its fluoride content and associated human health risk. The findings from this study shows that significant variation in the physico-chemical parameters across the sampling points considered exists. It was observed that fluoride content of the water exceeded the WHO limit of 1.5 mg/L in 33.3% of the various sampling locations. The multivariate statistical data shows that the source of groundwater contamination was largely from geogenic activities majorly from the dissolution of quartzite, dolomite, and shale into the deep aquifers of the study area with a minimal contribution from anthropogenic sources. Human health risk assessment for fluoride contamination of the groundwater revealed that 66.7, 71.4, 52.4, and 9.5% of the population in the sampling locations for infants, children, teenagers, and adults respectively are at high risk of developing fluoride related disorders from the consumption of fluoride laden water. Moreover, multivariate statistics was successful and effective in describing the relationships that exist among the diverse nature of pollutants in groundwater and was able to establish the source of pollutants in the studied groundwater samples. Water quality index analysis revealed that aside the elevated fluoride levels in some of the samples, the water was largely suitable for domestic use.
The results of this study (first of its kind) will serve as baseline data and provide information on the quality status of groundwater obtained from boreholes in Makurdi town and its environs with specific emphasis on fluoride pollution levels and its associated adverse health effects. Thus, this study proposes that groundwater sources in the study area at the household scales should be subjected to affordable and efficient water treatment technologies such as batch adsorption or cloth filtration prior to consumption for the containment of fluoride and related contaminants.

Limitation of the study
Although the temporal variations of groundwater quality was not conducted in the current study, the researchers ensured the validity of their findings by designing the sampling protocols to coincide with the peak rainfall period (October) in the study area. This ensured that the worst scenario was represented in the current work. Thus, the findings in the current work can serve as a baseline data for further related works considering seasonal/temporal changes in groundwater quality in the study area.
Secondly, water quality is a broad subject area that involves a large number of parameters. Thus, this study focused mainly on the parameters that are most crucial to drinking water quality with reference to the WHO standards. The other quality parameters such as heavy metals and bacteria are as well important.