Hydrochemical Characteristics of Water Quality Around Nkalagu Area, Southern Benue Trough, Nigeria Using Multivariate Statistical Analysis

Hydrochemical characteristics of water quality around Nkalagu area has been studied and characterized using multivariate statistical analysis. Eighty water samples were collected in the area from spatially referenced boreholes, hand dug wells abandoned mines, catch pits and rivers located in and around the Nkalagu area and were analyzed for EC, pH, TDS, TH, Ca 2+ , Mg 2+ , Na + , K + , HCO 3- , Cl - , NO 3- , SO 42- and Fe 2+ according to EPA and APHA standards. Based on mean values, the order of abundance in ions is Cl - > HCO 3- > SO 42- > NO 3- , for anions and Na + > Ca 2+ > Mg 2 - > K + , for cations. TDS and Salinity hazard classifications characterized the water in the study area as soft to very hard with low to very high salinity hazard. Principal component analysis (PCA) reduced the hydrochemical data into two principal components which explain 78.553 %, of the total variance that characterize the water quality in relation to the source of its hydrochemistry. Cluster analysis (CA) grouped eighty water samples in the area into eight clusters of similar water quality characteristics related to water-rock interaction, agriculture and anthropogenic sources. Discriminant analysis (DA) showed that the discriminating parameters of water quality in the area are EC, TDS, TH, SO 4 , Cl, Mg, Ca, Na, and HCO 3 and this revealed that water quality in the area is controlled by both geogenic anthropogenic CA1-1 group, indicating that that water sampled in those stations have high pollution resulting from weathering of the host minerals. CA2-3 group has low positive loadings on PC1 and PC1. CA3-1 group has very high loadings on PC1. Water samples from these stations have high pollution resulting from weathering of the host minerals and mining activities. CA3-2 group has low positive loadings on PC1 and PC2.

wells (HDW), 13 from abandoned mine (AM), 3 from Catch pit (CP) and 4 from river (RV) in the month of March, 2019. The sampling locations were selected in order to cover residential, agricultural and industrial area so as to achieve a good sampling representation of the study area. Physiochemical parameters such as temperature (Temp), pH, total dissolve solid (TDS) and electrical conductivity (EC) were measured immediately in the field at each point once the sample was collected with the aid of field probes due to their transient characteristics and the remaining parameters were determined in the laboratory within 24 hours. Two set of samples were collected at each point. One set for anions test while the remaining set is for cations and were stabilized with two to three drops of diluted HCl. Samples were collected in pre-cleaned sterilized plastic bottles and stored in an ice box and the preservation and transportation of water samples were performed according to standard methods (APHA 2005). Chemical analyses were carried out at the chemical research laboratory, Abakaliki, Nigeria. The analytical methods used in the determination of the hydrochemical parameters are in accordance with the World Health Organization (WHO 2011) standards and in each of the samples, 14 parameters were tested for. Iron, Calcium, sodium, potassium and magnesium were determined by atomic absorption spectrophotometry. Chloride, nitrate and sulfate were determined by ion chromatography, bicarbonate and total hardness by Potentiometric titration. The accuracy of the chemical analysis was verified by calculating charge ratio between the sum of cations and sum of anions. Water samples result in the study area was classified and compared according to US Salinity Laboratory Staff (1954) based on EC, Davis & DeWiest (1996) based on TDS, Freeze & Cherry (1979) based on TDS and Sawyer & McCarty (1967) based on TH.
Three multivariate analysis techniques, namely, principal component analysis (PCA), cluster analysis (CA) and discriminant analysis (DA) were employed in characterizing the water quality in the study area. All the statistical analyses were carried out using Stagraphics Centurion XVI. The data were first of all standardized before they were used as input data in order to correct the effects of the varied range of measurements of the various parameters and differences in the units of measurements (Singh et al. 2004;Singh et al. 2009;Mohapatra et al. 2011). PCA, CA and DA were based on 13 physicochemical parameters as input variables in eighty water samples. Principal components with Eigenvalues ≥ 1.0 were considered significant (Kaiser 1958;Harman 1967). Principal component weight or factor loading ≥ 3.0 were considered significant for the physicochemical parameters and principal scores loadings ≥ 1.0 were considered significant on the water sampling location (Senthilkumar et al. 2008;Ayuba et al. 2013). Cluster analysis was based on Ward's method and squared Euclidean distance metric mode (Ward 1963;Güler et al. 2002). The discriminating factor used for the DA was the pollution loading class defined and identified by PCA and CA respectively. The data inputted were data collected in 80 different locations in the year 2019 after mining activities had resumed in the area.

Hydrochemical Characteristics
The result of the water analyses and the World Health Organization (WHO 2011) and Standard Organization of Nigeria (SON 2007) guideline limits is presented in Table 1. Table 1 shows the descriptive statistics (minimum, maximum, mean and standard deviation) overview of the chemistry of water in the study area generated from the analysis of the water samples collected in dry season. The range, mean and standard deviation values reveal considerable variations in the water samples with respect to their chemical composition. The pH values of water samples in the study area ranged from 5.25 to 8.25 (mean = 7.27). This reveals that the water in the study area is acidic to slightly alkaline in nature. EC is a measure of the total ionic components in water; the more solutes present in water, the higher the EC. The EC values in water samples ranged from 8.0 to 3996.0 µS/cm with a mean value of 1081.79 µS/cm. However, high values of EC were recorded in groundwater samples in the study area. The groundwater samples show very high EC values, especially in the dry season. This high values in the groundwater samples can be attributed to the high content of charged ions due to oxidation processes going on in the area. EC values revealed the high diversity in the geochemical processes that shape the chemistry of water in the area. The TDS values in water samples ranged from 75.0 to 1879.0 mg/l with a mean value of 545.96 mg/l. Presence of high level of TDS in water (> 1200 mg/l) can cause objectionable to consumers WHO (2011). The mean of water samples in the area exceeded the criteria of SON (2007) and WHO (2011). It was observed that some water samples show very high TDS values (> 120 mg/l). EC and TDS value was also observed to increase with depth in the groundwater samples. The high concentrations of TDS and EC in the groundwater samples might be attributed to the more pronounced water-rock interaction, such as the mineral dissolution and evaporation concentration functions of the host rock. These high TDS concentrations are due to the presence of high HCO3 -, SO4 2-, Cl -, Ca 2+ and Na + as showed in Table 1. According to Jaine et al. (2003), water that contains such high concentration of TDS could cause gastrointestinal irritation. High values of TDS also influence the taste, hardness, and corrosive property of the water (Haran 2002;WHO 2011). The hardness of water limits its use for domestic and agricultural activities. The TH values in water samples ranged from 32.44 to 467.78 mg/l with a mean value of 189.98 mg/l. Hardness in water in the study area is mostly due to the high TDS compared to Ca 2+ and Mg 2+ concentrations.
The calculated charge ratio between the sum of cations and sum of anions was ± 1.4 %, which is within the acceptable limits of < ± 5 % which confirms the reliability of the analytical results (Datta & Subramanian 1998;Singh & Hassin 2002). Na + and Ca 2+ dominate the observed cations concentration in water samples with mean values of 52.28 mg/l and 43.69 mg/l (Table 1) respectively. These ions represent 42.68 % and 35.67 % of the total major cations of water samples respectively while Mg 2+ represents 16.05 % and K + represents only 5.60 % of the total major cations of water samples (Table 2). Cland HCO3dominate the anions concentration with mean values of 103.01 mg/l and 100.42 mg/l respectively. These ions represent 34.59 % and 33.72 % of the total major anions respectively while SO4 2represents 26.25 % and NO3represents only 16.21 % of the total major anions of water samples.
The Fe values in dry season ranged from 0.01 to 5.52 ppm with a mean value of 0.77 ppm. The abundance of the major ions in the water samples in descending order is Cl -> HCO3 -> SO4 2-> NO3for anions and Na + > Ca 2+ > Mg 2 -> K + for cations. The standard deviations of the hydrochemical variables in general indicate that the water in the study area is heterogeneous and reveals the influence of complex contamination sources and geochemical processes. This variation could be attributed to differences in salinity and ionic composition. According to the US Salinity Laboratory (1954) classification 9 % of the water samples are classified as "Low class", 17 % as "Medium class" 55 % as "High class" and 19 % as "Very high class" as shown in Table 3. Consumption of such water could lead to gastro intestinal irritation. According to Davis & Dewiest (1966) water classification based on TDS (Table  4) classified 44 % of the water samples as "Desirable water" 34 % as "Permissible water" and 22 % as "Useful irrigation". Freeze & Cherry (1979) classification based on TDS (Table 5) also classified 78 % of the water samples as "Fresh water" and 22 % as "brackish water". Water classification based on TH value (Table 6) classified water in the study area as "Soft water type to very hard water type" according to Sawyer & McCarty (1967) water classification and will definitely require softening prior to domestic use. Hardness in water can give rise to the formation of scum (whitish scale) in pots, boiler rings, and irrigation equipment; it may also cause health problems to humans such as kidney failure (WHO 2011).   This hydrochemical facies indicates the dominance of alkali metals over alkaline earth metals (Na + K > Ca + Mg) and strong acidic anions over weak acidic anions (Cl + SO4 > HCO3). The origin of Na-Cl facies may be attributed to weathering of the lithographic units and dissolution of halite in water.
Ca-HCO3 facies (BH22, BH23, BH24, BH27, BH28, BH29, BH30, HDW12, AM13, AM8, AM9, AM10, AM11, AM12, CP1, CP2, CP3, RV1, RV2, RV3, and RV4 ) and this facies denotes the dominance of alkaline earth metal over alkali metals (Ca + Mg > Na + K) and weak anions over strong acidic anions (HCO3 > Cl + SO4). This suggested that carbonate weathering domination and rock-water interaction are the primary factors in increasing the major ion concentration in water. The origin of Ca-HCO3 facies could be traced to water recharge through precipitation. This facies type denotes primary (temporary) water hardness which relates to concentrations of calcium and magnesium in water and is usually expressed as an equivalent concentration of dissolved calcium carbonate (CaCO3). Primary hardness in water causes scale in water heaters, boilers, pipes, and turbines; it also consumes excessive quantities of soap during washing activities. Primary hardness in water can be removed by boiling (Freeze & Cherry 1979).
The predominance of the halite water type over the other water types denotes that the groundwater is seawater in nature and the variation in chemistry may be as a result of rock-water interactions and anthropogenic activities.

Correlation Analysis
Physiochemical parameters relationships The degree of linear association between any two water quality parameters is measured by the correlation coefficient (r) value. The correlation matrix of the physiochemical parameters in the study area is presented in Table 7. Parameters with correlation coefficient values that are significantly related at 0.01 and 0.05 levels are written with asterisks. The ionic pairs that are statistically related at 0.01 and 0.05 levels are thought to be released from the same sources or through same geochemical processes. The significant correlation between EC and the other hydrogeochemical parameters is highly positive with the exception of K + , HCO3 -, NO3 -, and Fe 2+ . The r value between EC and TDS is 0.983, which means TDS is highly positively correlated with EC and can be predicted from EC with 98 %. Additionally, the EC value of the water samples has high positive correlation with TH, Ca 2+ , Mg 2+ , Na + , Cland SO4 2with relative positive coefficient r values of 0.828, 0.804, 0.830, 0.889, 0.824 and 0.858 respectively. These positive correlations between EC and some of the major ions indicate that an increase in these ions concentrations would increase the EC value of the water in the area. The strong correlation of the major elements Ca 2+ , Mg 2+ , Na + , Cland SO4 2with EC is an indication of the contribution of these elements to the salinity or hardness of the water due to concentration of ions from evaporation of recharge water and water interaction with the geological formations. pH was found to be positively correlated (0.05 level) to K + and TH. This could be attributed to the anthropogenic influence on the water in the study area. The TDS values of the water samples show strong positive correlation with the major ions (Ca 2+ , Mg 2+ , Na + , K + , HCO3 -, Cl -, SO4 2and NO3 -,) that constitute it in the water solution. Table 4.17 also shows a strong positive correlation between TH and the cations Ca 2+ and Mg 2+ . This relationship is in line with fact that TH is determined based on these two cations. Ca 2+ shows highly positive correlation with Mg 2+ compared to Na + and strong correlation with Cland SO4 2compared to HCO3 -. This could be an indication of the source of Ca 2+ in the water (e.g. calcite, dolomite, gypsum and silicates) due to its strong association with Mg 2+ and suggest the type of water found in the study area. Mg 2+ positively correlated (0.01 level) to Na + , HCO3 -, Cland SO4 2-. Na + showed strong positive correlation with Cland SO4 2besides TDS compared to HCO3 -, which is an indication of the salinity found in some of the water samples. Clshowed strong positive correlation with SO4 2and NO3 -. This could be an indication of surface contamination due to agricultural activities in the study area.

Sources of ions and their controlling processes
PCA was employed in the determination of the various sources of the ions and processes controlling water chemistry and it was performed on 13 variables (pH, EC, TDS, TH, Ca 2+ , Mg 2+ , Na + , K + , Cl -, SO4 2-, HCO3 -NO3and Fe 2+ ) of the water samples in the study area. Table 8 shows the initial determined components, their Eigenvalues and the percent of variance contributed in each component. Only factors with Eigenvalues ≥ 1 were taken into consideration and this resulted into two principle components (PCs) that were sufficient in explaining 78.544 % of the variability in the original dataset from water samples. Absolute values of factor loadings of ≥ ± 3.5 were considered as strong correlation and rendered in bold and italic in Tables 8 to elucidate the relationship between the factors and the hydrochemical dataset. The two principle components shown in Table 8 are dominated by certain variables based on the prevailing hydrogeochemical processes and land use practices. EC, TDS, TH, Ca 2+ , Mg 2+ , Na + , SO4 2-, Cland HCO3have high positive loading factors on principal component (PC1) in water samples, explaining 15.521 % of the variation in the total dataset (Table 8). PC 2 has high positive loadings on pH, K + and Fe 2+ and negative loading on NO3 -, and SO4 2in water samples which explained 78.544 % of the variation in the total dataset (Table 8). As a result of the high associations and correlations between Ca 2+ +Mg 2+ and HCO3 -, Na + , and Cl -, PC 1 which explained the highest variance in the dataset may be defined as "hardness and salinity" factor. PC 2 which explained the least variance of the dataset has high negative loadings on NO3and SO4 2-. The occurrence of high loadings of SO4 2in PC 1 and PC 2 suggest multiple sources for the ions. In PC1, SO4 2has loading alongside with the major ions (Ca 2+ , Mg 2+ , Na + , Cland HCO3 -) and these ions are assumed to be released through various geogenic geochemical processes. SO4 2in PC 1 may have been released from geogenic processes also. The geogenic source of SO4 2may be derived from pyrite oxidation. The association of NO3with SO4 2in PC 2 suggests anthropogenic source of SO4 2in addition to its non-anthropogenic (geogenic) source. NO3is usually derived from anthropogenic sources and the association of this ion with SO4 2in PC 2 suggests an anthropogenic source for SO4 2-, in addition to those derived from oxidation of pyrite and related minerals. NO3may be released from domestic wastes and agricultural activities; likewise SO4 2may be released from domestic wastes as well as sulfate-rich fertilizers. PC 2 may be defined as "Anthropogenic" factor.

Discriminating physiochemical factors
DA was used to find out one or two functions (linear combinations) of observed data (discriminating functions) that best separate the water quality (high pollution loading and low pollution loading) of each of the water sampled in the area. One discriminating function (DF) which has the following qualities: eigenvalue >1.0, relative percentage explained > 70 %, and high canonical correlation > 0.5 (Mahmood et al. 2001) was extracted in the samples and was found to efficiently discriminate the water quality (Table 10). Wilk's lambda test showed that the extracted DF is a statistically significant DF because the P value is < 0.05 confidence (Table 4.9). The DF coefficient for water samples is presented in Table 11. From Table 11, EC, TDS, TH, SO4, Cl, Mg, Ca, Na, and HCO3 was found to best discriminate the water quality in the area. The DA shows that ions of both geogenic (Ca, Mg, SO4 and HCO3) and anthropogenic (Cl) origins best discriminate the water quality in the study area.

Conclusion
The study showed that the analysis of hydrochemical data using the multivariate statistical techniques such as (PCA, CA and DA) can give some information not available at first glance in the conventional hydrogeochemical analyses techniques. The classification of water types and dominant ions based on Piper diagram is: Na + -Cltype, Ca 2+ -HCO3type and Ca 2+ -Mg 2+ -Cl --HCO3type, with Na + -Clas the dominant water type. For all samples, the order of abundance in ions is Cl -> HCO3 -> SO4 2-> NO3 -, for anions and Na + > Ca 2+ > Mg 2 -> K + , for cations. PCA converted the thirteen parameters into two principle components (PCs), which explained 78.553 %, of the total variance. The first principle component (PC1) termed as "hardness and salinity" factor, explained 63.032 % of the total variance. The second principle component (PC2) can be termed as "anthropogenic" factor, which explained 15.521 % of the total variance. CA grouped 80 water samples in the area into eight clusters of similar water quality characteristics related to water-rock interaction, agriculture and anthropogenic sources. DA has shown that the principal physiochemical parameters which distinguish the water quality in the area are of geogenic and anthropogenic origins.
Hence, this study illustrates that multivariate statistical analysis is an excellent empirical tool for understanding complex water quality data sets and for understanding spatial variations, which are useful and effective for water quality management.