Six counties from Middle Tennessee namely Canon, Coffee, Davidson, Rutherford, Warren, and Williamson were chosen for the collection of the tap water samples (Figure 1 and Figure 2). These counties were chosen because they represent Urban, Urbanizing and Rural settings in Middle Tennessee. According to the US Census Bureau [16], one county is considered urban, two as urbanizing, and three as rural. The Davidson county (population 648,295) is an urban county, while Rutherford (population 274,454) and Williamson (population 192,911) counties are urbanizing. On the other hand, the rural counties are Canon (population 3811), Coffee (population 53,222), and Warren (population

39,839). The tap water samples were collected from three locations in each county. Tap water was collected randomly from six counties in Middle Tennessee while 12 different brands of bottled water were purchased from various supermarkets. Most of the selected sampling locations were publicly accessible places such as supermarkets, gas stations, and food courts or restaurants.

Tap water samples were collected randomly from three different locations within each county. The tap water samples were collected between May 2012 and July 2013. Bottled water was also purchased during the same period. A total of 18 tap water samples were collected from the selected six counties (Table 1). Pre-cleaned sterilized plastic bottles were used to collect the water samples. The water samples were transported to the laboratory. Elemental constituents were determined using ICP-OES.

All the chemicals used in the analysis of the water samples were reagent grade and were purchased from Fisher Scientific (Suwanee GA, USA) and Hach (Loveland, CO, USA). Distilled water was used as feed water to produce ultra-pure deionized water by the Nano-pure Infinity UV/UF Deionizer (Barnstead Thermolyne, IA, USA; resistivity: 18.3 m Ω/cm at room temperature and TOC: < µg/L). A mixed standard (100 mg/L containing a suite of elements) was purchased from SPEX Certiprep, Inc. (Metuchen, NJ, USA). Other metal standards (1000-mg/L each) for the calibration of the ICP were also purchased from Fisher Scientific. The method for the analysis of elements in samples was validated with Standard Reference Material (SRM 1643e Trace elements in natural water). A Varian Vista-Pro ICP-OES (Varian Inc., Walnut Creek, CA 94598, USA) was used for the measurement.

The ICP-OES was calibrated using calibration standard solutions made from the stock standard solutions. Additionally, an internal standard (scandium) was used as part of quality assurance. SRM 1643e was used to check the accuracy and precision of the analytical method. The Standard Reference Material (SRM 1643e): trace elements in natural water, was purchased from the National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA.

Tap water samples were collected randomly from three different locations within each county. A total of 18 tap water samples were collected from the selected six counties (Table 1). While microbial work was not done in this study, pre-cleaned sterilized plastic bottles were used to collect the water samples. The water samples were shipped to the laboratory for analysis. Elemental constituents were determined using ICP-OES. The collected water samples did not contain any particulates; hence the water samples were not filtered prior to analysis. The Operation condition of the Inductively Coupled Plasma is illustrated in Table 2.

Milli-Q water (resistivity of 18.2 MΩ/cm) was used for sample dilutions, rinses and preparation of diluted standards. All glassware and polyethylene containers in contact with sample digests were washed with metal-free soap, rinsed multiple times, soaked in 50% nitric acid for 24 hours and finally rinsed with deionized water. The trace metal grade concentrated nitric acid was used as a reagent and purchased from Fisher Scientific (St. Louis, Missouri, USA). The ICP tune and calibration solutions were from SPEX Certiprep, Inc. (Metuchen, NJ, USA). Standard reference material (SRM 1640: trace elements in natural water) (Table 3) was purchased from the National Institute of Standards and Testing (NIST), Gaithersburg, MD 20899, USA. The SRM samples were used in recovery and method validation experiments (Table 3).

The percentage trace element recoveries (Table 3) from the standard reference materials were close to the provided certified values by the manufacturers. Recoveries ranged from 70% - 120%. The measured values of SRM 1643 used as an ICP check solution were also close to the certified values reported by NIST (data not reported).

The elements determined by ICP-OES were Aluminum (Al), Antimony (Sb), Arsenic (As), Boron (B), Barium (Ba), Beryllium (Be), Bismuth (Bi), Calcium (Ca), Cadmium (Cd), Cobalt (Co), Chromium (Cr), Copper (Cu), Gallium (Ga), Indium (In), Iron (Fe), Lead (Pb), Lithium (Li), Magnesium (Mg), Manganese (Mn), Mercury (Hg), Molybdenum (Mo), Nickel (Ni), Phosphorus (P), Potassium (K), Rubidium (Rb), Selenium (Se), Sodium (Na), Silicon (Si), Silver (Ag), Strontium (Sr), Sulfur (S), Thallium (Tl), Tin (Sn), Titanium (Ti), Uranium (U), Vanadium (V), and Zinc (Zn).

The physicochemical parameters such as pH, specific conductivity, Total Dissolved Solids (TDS) were determined in tap water and compared with the US Environmental Protection Agency Maximum Contaminant Level (MCL). These parameters were found to be below their respective MCL limit (Table 4).

Statistical analysis showed that the selected urban areas had a significantly higher percentage of heavy metals than those of rural areas (Figure 3). Additionally, heavy metals (Arsenic, Cadmium, Cobalt, Lead, Mercury, and Silver) concentrations were highest in urbanized areas relative to rural settings. However, the values were below the Maximum Contaminant Levels (MCLs).

No differences were detected in heavy metal concentrations among the six counties (χ² = 4.07, df = 5, P = 0.5395). However, when counties were compared for radioactivity (presence of uranium U and rubidium Rb), significant difference were observed (χ² = 25.47; df = 5; P = 0.0001). Canon county had significantly more radioactive elements than the other counties (Figure 4). The reason could be attributed to the underlying bedrock and natural radioactive elements.

From the Principal component analysis of all the elements, it was found that the first 7 components explain 90% of the variations among the variables (Table 5; Figure 5). Also Component 1 and 2 cumulatively accounted for 53% of the variability in the dataset (Table 5).

Cluster analysis or clustering is the task of grouping a set of objects in such a way that the objects in the same group (called a cluster) are more similar (in some sense or another) to each other than to those in other groups (clusters). It is the main task of exploratory data mining, and a common technique for data analysis, used in many fields. Cluster analysis (agglomeration, nearest neighbor

*The first seven components explain 90% of the variations among the variable’s elements. The rest 31 of the components has been truncated due to redundancy.

method, squared Euclidean distance and z-score transformation) produced dendrogram plot (Figure 6) from all the major inorganic concentrations (except Fl- Cl- Ni, Co, Fe, Sn) in drinking water types. The hierarchical agglomerative

procedure resulted in two major clusters. The dendrogram plot suggests that tap water in these counties was very similar to the bottled water types.

It is noteworthy that most of the toxic metals were below the EPA recommended limit. However, more epidemiological studies are required to see the effect of several of the elements.

The data described herein, tend to indicate that there is minimal or no significant difference between bottled water and tap water. However, some chemical constituents like calcium and magnesium tend to be higher in tap water collected from rural counties than in the urbanizing counties. Similarly, the same elements tend to be higher in natural spring bottled water, indicating that the occurrence of these elements might be due to the geology of the drinking water. In middle Tennessee for example, there is an abundance of limestone rocks. These rocks are prevalent in rural counties of middle Tennessee. Due to the industrialization and urban sprawl in the urban counties, there are less rocks. As a result, the concentration of calcium and magnesium tend to be lower in tap water. Also, the concentrations of fluoride were relatively higher in tap water. Fluoride was below the detection limits in bottled water; indicating that manufacturers of the bottled water brands seldom add this element (fluoride) to bottled water. Certainly, there is a dire need to conduct more epidemiological studies to see the effect of several of these elements in drinking water sources.