*2.4. Statistical Analysis and Computing*

In this study, statistical analysis was conducted by using SPSS 25 for Pearson's correlation. Pearson's correlation coefficient (*r*) helps to quantify the significance of a relationship between two parameters and was widely used in groundwater quality assessment because it gives a quick correlation value. Its mathematical formula is expressed as follows [44]:

$$r\_{xy} = \frac{i = \sum\_{i=1}^{n} \left(\mathbf{x}\_i - \overline{\mathbf{x}}\right)(y\_i - \overline{y})}{\sqrt{\sum\_{i=1}^{n} \left(\mathbf{x}\_i - \overline{\mathbf{x}}\right)^2 \sum\_{i=1}^{n} \left(y\_i - \overline{y}\right)^2}} \tag{1}$$

where, *rxy* represents the correlation coefficient between the parameters *x* and *y*, *n* denotes the sample size, *x<sup>i</sup>* is the individual value of the parameter *x*, *x* is the mean value of the parameter *x*, *y<sup>i</sup>* stands for the individual value of the parameter *y*, and *y* denotes the mean value of the parameter *y*.

The values of correlation coefficient can be classified as very strong for *r* ≥ 0.80, strong for 0.60 ≤ *r* < 0.80, moderate for 0.40 ≤ *r* < 0.60, weak for 0.20 ≤ *r* < 0.40, and very weak for *r* < 0.20. In addition, the correlation coefficient is evaluated on the basis of *p* value. The correlation coefficient is statistically considered as highly significant when *p* < 0.01, marginally significant when *p* < 0.05, and not significant when *p* > 0.10 [44].

For various computing and plots, Microsoft Office 2016 (Excel and Word), Origin 2018, and Grapher 12 were used. Parameter analysis, Piper [45] and Durov [46] diagrams plots were executed using AqQA software. Finally, for mapping, ArcMap 10.3 software was used to locate samples and make a water quality distribution map. This map was obtained using Bayesian Kriging method, which is an automatic Geo-statistical interpolation package incorporated in ArcGIS software. The general Kriging equation can be described as follows [47]:

$$Z^\*\left(\mathbf{x}\_p\right) = \sum\_{i=1}^n \lambda\_i Z(\mathbf{x}\_i) \text{ with } \sum\_{i=1}^n \lambda\_i = 1 \tag{2}$$

where *λ<sup>i</sup>* is the Kriging weight; *Z\**(*xp*) estimates the unknown true value.
