*Article* **eNaBLe, an On-Line Tool to Evaluate Natural Background Levels in Groundwater Bodies**

**Daniele Parrone , Eleonora Frollini \* , Elisabetta Preziosi and Stefano Ghergo**

IRSA-CNR, Water Research Institute—National Research Council, Via Salaria km 29.300, PB 10, 00015 Monterotondo (Rome), Italy; parrone@irsa.cnr.it (D.P.); preziosi@irsa.cnr.it (E.P.); ghergo@irsa.cnr.it (S.G.) **\*** Correspondence: frollini@irsa.cnr.it

**Abstract:** Inorganic compounds in groundwater may derive from both natural processes and anthropogenic activities. The assessment of natural background levels (NBLs) is often useful to distinguish these sources. The approaches for the NBLs assessment can be classified as geochemical (e.g., the well-known pre-selection method) or statistical, the latter involving the application of statistical procedures to separate natural and anthropogenic populations. National Guidelines for the NBLs assessment in groundwater have been published in Italy (ISPRA 155/2017), based mainly on the pre-selection method. The Guidelines propose different assessment paths according to the sample size in spatial/temporal dimension and the type of the distribution of the pre-selected dataset, taking also into account the redox conditions of the groundwater body. The obtained NBLs are labelled with a different confidence level in function of number of total observations/monitoring sites, extension of groundwater body and aquifer type (confined or unconfined). To support the implementation of the Guidelines, the on-line tool evaluation of natural background levels (eNaBLe), written in PHP and using MySQL as DBMS (DataBase Management System), has been developed. The main goal of this paper is to describe the functioning of eNaBLe and test the tool on a case study in central Italy. We calculated the NBLs of As, F, Fe and Mn in the southern portion of the Mounts Vulsini groundwater body, within the volcanic province of Latium (Central Italy), also separating the reducing and oxidizing facies. Specific results aside, this study allowed to verify the functioning and possible improvements of the online tool and to identify some criticalities in the procedure NBLs assessment at the groundwater body scale

**Citation:** Parrone, D.; Frollini, E.; Preziosi, E.; Ghergo, S. eNaBLe, an On-Line Tool to Evaluate Natural Background Levels in Groundwater Bodies. *Water* **2021**, *13*, 74. https://doi.org/10.3390/w13010074

Received: 26 November 2020 Accepted: 27 December 2020 Published: 31 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Keywords:** natural background; groundwater body; conceptual model; preselection; nitrates; confidence level

## **1. Introduction**

The presence of inorganic potentially toxic elements in groundwater represents a significant problem in many parts of the world. They may derive from both natural processes and anthropogenic activities and natural background levels (NBLs) assessment is often used to distinguish these sources. The NBL has been defined in the Groundwater Directive (GWD) [1] as "the concentration of a substance or the value of an indicator in a groundwater body corresponding to no, or extremely limited, anthropogenic alterations, compared to unaltered conditions". The GWD requests the EU Member States to individuate appropriate threshold values (TVs) for various potentially harmful substances, taking into account NBLs when necessary, in order to evaluate the chemical status of groundwater bodies.

Currently, it is possible to distinguish a geochemical and a statistical approach for the NBLs assessment. The geochemical approach, originally called "pre-selection" (PS) within the BRIDGE (Background cRiteria for the Identification of Groundwater thrEsholds) project [2], requires the identification of groundwater with no or negligible human impact, using markers such as nitrates/ammonia in oxidizing/reducing environments, organic compounds and isotopes. Once the samples not affected by anthropogenic impact have been selected, a

representative value (usually a chosen percentile) of the natural background is derived. Many examples of PS method applications, including comparisons between groundwater bodies of different European countries [3,4], can be found in the literature [5–12].

The statistical approach involves the separation of uninfluenced and influenced populations by means of statistical procedures. For this purpose, numerous statistical techniques have been developed and tested. Some of these methods point to eliminate the outliers, assuming that the remaining data belong to the natural background. In some cases, the same methods can be used to separate different data populations. The "mean + 2σ" [13], the Median Absolute Deviation (MAD) [14], the Box and Whisker plot [15], the component separation [11,16] and other parametric or non-parametric techniques, including graphical methods as probability plots or quantile–quantile plots, have been largely used for this purpose [17–20]. Please refer to Preziosi et al. [21] for a review of the most common approaches for NBLs assessment, including many statistical techniques and the pre-selection methods.

In 2017, national Guidelines for the NBLs assessment in groundwater were published in Italy [22]. Starting from these Guidelines, Frollini et al. [23] applied the procedure to define the NBLs at the groundwater body (GWB) scale, while Parrone et al. [24] tested a multi-methodological approach on two case studies, also suggesting new criteria for the choice of the nitrate threshold to be used for the pre-selection of non-contaminated samples. An automated approach implementing both the component separation and the pre-selection methods was recently proposed by Chidichimo et al. [25].

The procedure is based mainly on the pre-selection method but different assessment paths are proposed according to the redox conditions of the GWB, the sample size in both the spatial and temporal dimension and the type of the distribution, normal or not, of the pre-selected dataset. The complexity of the schema is essential because of the great heterogeneity of the different GWB monitoring data. For this reason, moreover, the obtained NBLs are labelled with a different confidence level in function of: The number of total observations or the number of total monitoring sites (MSs), extension of groundwater body and aquifer type (confined or unconfined). For GWBs characterized by NBLs with low confidence levels, further monitoring activities are requested.

To provide support to operators involved in the use of the Italian Guidelines and to achieve a harmonization of procedures between the different structures involved, evaluation of natural background levels (eNaBLe), an on-line tool implementing the sequence of operations for NBL assessment, was developed at Water Research Institute – National Research Council (IRSA-CNR) [26]. The procedure implemented in the eNaBLe tool, written in PHP and using MySQL as DBMS, is organized into three logical blocks:


The main goal of this work is to test the eNaBLe tool on a case study in Central Italy. We calculated the NBLs of As, F, Fe and Mn in the southern portion of the Mounts Vulsini GWB, within the volcanic province of Latium (Central Italy), also operating a separation of data based on the different redox conditions found in the study area.

#### **2. Materials and Methods**

This section is dedicated to the detailed description of the tool and of all the steps that lead to the calculation of the natural background values. A description of the case study to which the Guidelines were applied is also included.

#### *2.1. The Italian Guidelines and the eNaBLe Tool*

eNaBLe operates in compliance with the provisions of the Italian Guidelines for NBLs assessment, the general scheme of which is shown in Figure 1.

*2.1. The Italian Guidelines and the eNaBLe Tool* 

eNaBLe operates in compliance with the provisions of the Italian Guidelines for

NBLs assessment, the general scheme of which is shown in Figure 1.

**Figure 1.** Italian Guidelines: Scheme of the procedure for the natural background levels (NBLs) assessment at the Groundwater Body (GWB) scale [23]. **Figure 1.** Italian Guidelines: Scheme of the procedure for the natural background levels (NBLs) assessment at the Groundwater Body (GWB) scale [23].

The general methodology adopted by the Guidelines is that of preselection which is accompanied by statistical evaluations in order to adapt the application of the procedure The general methodology adopted by the Guidelines is that of preselection which is accompanied by statistical evaluations in order to adapt the application of the procedure to the various groundwater bodies and allow the use of datasets of different numerical consistency. Four different evaluation paths (A, B, C, D) were therefore identified according to the spatial and temporal consistency of the available dataset. As a consequence of the great differences between the various GWB, both in terms of extension and number of monitoring networks, it was considered appropriate to associate also an index (confidence level) to the defined NBLs linked to the size of the statistical sample, to the extension and type of the aquifer (unconfined or confined) (Table 1). For more detailed explanation of the entire procedure, see [22].

**Table 1.** Confidence levels. The confidence level attributed to a NBL takes into account the characteristics of the dataset (case A, B, C, D), of the total number of observations and monitoring sites (MSs), also with reference to the extension of the GWB and the aquifer type (unconfined/confined). Confidence levels: H: High; M: Medium; L: Low; and LL: Very Low.


The complexity of the Guidelines framework advised the demand for an automated procedure which, while following the general formulation and the involved provisions, could facilitate the task of the operators. In 2018, a first automated approach was developed at IRSA-CNR. Later, the procedure was implemented as the eNaBLe on-line tool. The complete procedure, written in PHP [27] and using MySQL [28] as relational databases manager, is structured in modules, each one performing specific tasks on the input data, using specific configuration parameters. Some of these parameters are inherited from the Guidelines and cannot be changed; the others can be modified by the user. Each resulting dataset is then sent to the next module. The overall breakdown of eNable tool is shown in Figure 2. The modularity of eNaBLe has allowed us to easily integrate it into the Institute's Water Resources Database [29] which now is the source for the raw data for the NBL evaluation.

The user interface for the tool is divided into three logical phases which correspond to three different procedures and therefore to three distinct associated web pages: Selection of the analytical parameters for the evaluation of the NBLs and selection of the calculation options to be used; NBLs assessment and relative confidence levels; and output of results in graphical, tabular and report form. These procedures will now be described schematically.

2.1.1. Choice of the GWB, of the Parameters and the Calculation Options to Be Used for the NBLs Assessment.

Performing a query on the Water Resource Database [29], eNaBLe produces a table in which all the analytical parameters and their number are listed, proposing for the NBLs evaluation those parameters in which at least one exceedance of the relative TVs has been found. It is therefore possible to select the upper limits to be used in the validation, preselection and calculation procedures.

*Water* **2020**, *12*, x FOR PEER REVIEW 5 of 16

**Figure 2.** General scheme of the evaluation of natural background levels (eNaBLe) tool. **Figure 2.** General scheme of the evaluation of natural background levels (eNaBLe) tool.

> 2.1.1. Choice of the GWB, of the Parameters and the Calculation Options to Be Used for In particular, it is possible to define:


• The inclusion or not of those stations that, due to their overall characteristics, have

modified. It is also possible, by deselecting the checkbox, to completely disable the redox facies separation. be modified. It is also possible, by deselecting the checkbox, to completely disable the redox facies separation.

• Parameters to be used for the redox facies separation. The redox potential (ORP) or the dissolved oxygen concentration (DO) can be selected. The predefined limits proposed by the system for these two parameters (e.g., DO 3 mg/L or ORP 100 mV) can

• The limits to be applied for the preselection process to nitrates or ammonia in relation to the oxidizing or reducing facies. The system proposes values equal to 75% of the expected quality standards. • The limits to be applied for the preselection process to nitrates or ammonia in relation to the oxidizing or reducing facies. The system proposes values equal to 75% of the expected quality standards.

• A threshold for the control of the electrical balance, typically 5% [30] or 10% [2].

*Water* **2020**, *12*, x FOR PEER REVIEW 6 of 16

• The appropriate time interval for the data to be processed.


A screenshot of the configuration option selection window is shown in Figure 3. A screenshot of the configuration option selection window is shown in Figure 3.


**Figure 3.** Selection of the calculation parameters within the eNaBLe tool. **Figure 3.** Selection of the calculation parameters within the eNaBLe tool.

2.1.2. NBLs Assessment and Relative Confidence Levels

The general NBLs assessment procedure can be summarized in the following operational phases:

1. Query to the database to retrieve data of the selected GWB, relative to the selected parameters and to the parameters linked to the facies separation.

	- Case A: Sample size adequate to describe the temporal and spatial variability of the parameter for the dataset in question;
	- Case B: Sample size adequate to describe the spatial variability of the parameter for the dataset in question, but not the temporal variability;
	- Case C: Sample size adequate to describe the temporal variability of the parameter for the dataset in question, but not spatial variability;
	- Case D: Sample size inadequate to describe the spatial and temporal variability of the parameter for the dataset in question.

For sample size adequate to describe, from a purely statistical point of view, the spatial variability of the system in question, it means a minimum of 15 monitoring stations adequately distributed (N ≥ 15). For sample size adequate to describe, from a purely statistical point of view, the temporal variability of the system in question, it means a minimum of 8 observations (*n* ≥ 8) distributed regularly over at least 2 years for each station, over at least 80% of the monitoring stations. These requirements are those listed in the Guidelines and are not modifiable by the tool users.


the dimensional (area) and typological (unconfined or confined) characteristics of the GWB (Table 1). GWB (Table 1). 12. Assignment of a NBL at the groundwater body scale using the methodology selected in the configuration phase.

analogy with other GWBs, or portions of GWBs, characterized by similar conditions in terms of geochemical facies, hydrogeological context and anthropic pressures. 11. Determination of the confidence level to be associated to the calculated NBLs. It was considered appropriate to associate an index (confidence level) to the NBLs, as a function of the size of the statistical sample on which the calculation was based, of the dimensional (area) and typological (unconfined or confined) characteristics of the

12. Assignment of a NBL at the groundwater body scale using the methodology selected in the configuration phase. Steps 6 to 11 are performed by eNaBLe for the different redox facies and steps 7 to 12 for each of the analytical parameters selected.

*Water* **2020**, *12*, x FOR PEER REVIEW 8 of 16

Steps 6 to 11 are performed by eNaBLe for the different redox facies and steps 7 to 12 for each of the analytical parameters selected. 2.1.3. Output of Results

#### 2.1.3. Output of Results At the end of the calculation procedures, eNaBLe produces a summary with the con-

At the end of the calculation procedures, eNaBLe produces a summary with the configuration options and the results (TV, validated data, minimum and maximum representative values, calculation model and normality of distribution) and the calculated NBLs with relative confidence levels for the investigated parameters, differentiated by redox facies. Finally, the system produces a table which shows the NBLs relative to the entire GWB. If during the configuration phase the separation of redox facies has been deactivated, only the NBLs calculated for the entire GWB are produced. figuration options and the results (TV, validated data, minimum and maximum representative values, calculation model and normality of distribution) and the calculated NBLs with relative confidence levels for the investigated parameters, differentiated by redox facies. Finally, the system produces a table which shows the NBLs relative to the entire GWB. If during the configuration phase the separation of redox facies has been deactivated, only the NBLs calculated for the entire GWB are produced. By the appropriate links contained in the results page, files in CVS format, containing the intermediate and final datasets produced during the data processing, are also accessi-

By the appropriate links contained in the results page, files in CVS format, containing the intermediate and final datasets produced during the data processing, are also accessible. ble.

The tool will finally produce graphical reports of the selected parameters consisting of a table with the main statistical data (minimum and maximum value, mean, median, MAD, 95th percentile and normality of distribution), quantile-quantile plots and the georeferenced spatial distribution of the monitoring stations. A printable PDF files in which are summarized all the configuration parameters and the resulting NBLs, is also available. The tool will finally produce graphical reports of the selected parameters consisting of a table with the main statistical data (minimum and maximum value, mean, median, MAD, 95th percentile and normality of distribution), quantile-quantile plots and the georeferenced spatial distribution of the monitoring stations. A printable PDF files in which are summarized all the configuration parameters and the resulting NBLs, is also available.

#### *2.2. The Mounts Vulsini Groundwater Body 2.2. The Mounts Vulsini Groundwater Body*

The investigated area extends for about 60 km<sup>2</sup> . It is located on the southern flank of the Mounts Vulsini groundwater body, an unconfined aquifer hosted in the Pleistocene volcanites of the Vulsini volcanic district (Central Italy). Groundwater in the study area flows from N to SW and ESE (Figure 4). The investigated area extends for about 60 km2. It is located on the southern flank of the Mounts Vulsini groundwater body, an unconfined aquifer hosted in the Pleistocene volcanites of the Vulsini volcanic district (Central Italy). Groundwater in the study area flows from N to SW and ESE (Figure 4).

**Figure 4.** Groundwater level contour map of the southern Mounts Vulsini GWB and outcropping lithologies.

The main anthropogenic pressures are agriculture and animal husbandry. Urban areas and industrial sites including waste management facilities are also present in the area.

Groundwaters are mainly of the alkaline-earth bicarbonate and alkaline bicarbonate type; due to different natural phenomena (water–rock interaction, upwelling of geothermal fluids along fracture/fault systems and presence of mineral deposits), As and F are known to be widespread and co-present in the area [33–40], mainly in oxidizing conditions, with values often higher than the standards set for human consumption by WHO [41] and Directive 98/83/EC. Fe and Mn, on the other hand, are mainly linked to reducing conditions that are found locally, causing the reductive dissolution of their oxi-hydroxides and significant concentrations of the two elements in groundwater.

A total of 50 groundwater samples were collected from private wells in July 2017, 9 of which have been resampled twice more in January and July 2020 to increase the numerosity of the peculiar reducing facies existing in the area. Groundwater samples were collected following standardized sampling protocols [42,43]. Particular attention was paid to the inline measurement of physical–chemical parameters, whose correct determination is crucial for the definition of the aquifer conceptual model and in the setting of the calculation parameters included in the procedure for NBLs assessment (e.g., DO and ORP accurate measurements for the redox facies separation).

Physical–chemical parameters and chemical data were used to build the starting database for the calculation of NBLs, operated by using the eNaBLe tool.

#### **3. Results**

#### *3.1. Configuration and Preliminary Analysis on the Monitoring Stations*

The first stage of the procedure includes a series of operations to be applied to the 50 monitoring stations (MSs), all belonging to the Mounts Vulsini groundwater body. All analytical values below the limit of quantification should be replaced with a value equal to half the LOQ. However, none of the parameters of interest for this study (As, F, Fe, Mn) showed values lower than the LOQ. As regards the data validation, the threshold of the electrical balance (5%) did not result in the elimination of any sample (maximum error = 3.8%).

The next step concerns the redox facies separation, for which a DO threshold of 3.0 mg/L was used. This value led to the identification of an oxidizing facies (41 water points), with DO > 3.0 mg/L, largely dominant and widely present throughout the study area, and a reducing facies (9 water points), with DO < 3.0 mg/L, not spatially diffused but rather linked to a few isolated points (Figure 5).

The third and last step of this phase of the procedure is represented by the preselection of the MSs, operated using two different markers of anthropogenic contamination, according to the redox conditions: NO<sup>3</sup> − (<37.5 mg/L) for the oxidizing facies and NH<sup>4</sup> + (<0.375 mg/L) for the reducing facies. The selected concentration limits correspond to the 75% of the expected quality standards/TVs. As for the oxidizing facies, 15 samples were discarded, so the preselected dataset is composed of 26 MSs useful for the NBLs calculation (Figure 5). No points were discarded for the reducing facies, as all NH<sup>4</sup> <sup>+</sup> values are below the chosen threshold. The preselected dataset is therefore composed of the 9 initial stations.

Before continuing with the next parameter-specific phase, it was decided to separately evaluate the correlation between the elements of interest and the redox parameters (DO and ORP), in order to evaluate their sensitivity to the redox conditions. The Pearson parametric and Spearman non-parametric correlation indexes (Table 2) show a significant negative correlation with the redox parameters for Fe and Mn (mutually proportional), while As and F are well correlated with each other but not redox sensitive elements.

Consequently, in the NBLs assessment we have decided to proceed in a diversified way, keeping the two redox facies separate for Fe and Mn and instead defining a single value, relative to the entire preselected dataset (35 water points), for As and F, not affected by the redox conditions.

*Water* **2020**, *12*, x FOR PEER REVIEW 10 of 16

**Figure 5.** Redox facies subdivision based on dissolved oxygen concentration (DO) concentration and preselected MSs. **Figure 5.** Redox facies subdivision based on dissolved oxygen concentration (DO) concentration and preselected MSs.


**ORP 0.43** −0.39 **−0.45 −0.42** −0.21

Before continuing with the next parameter-specific phase, it was decided to separately evaluate the correlation between the elements of interest and the redox parameters **Table 2.** Pearson (*r*) and Spearman (*rs*) correlation coefficients for the redox parameters and the investigated elements. Statistically significant values in bold (α = 0.01).

#### **F** −0.21 **−**0.29 0.18 0.21 **0.51 Mn −0.66 −0.48** 0.16 **0.99** 0.13 *3.2. NBLs Calculation for As and F*

**Fe −0.53 −0.42** 0.28 **0.76** 0.11 **As −**0.04 **−**0.09 **0.59** −0.12 0.04 **DO ORP F Mn Fe As Spearman (***rs***)**  Consequently, in the NBLs assessment we have decided to proceed in a diversified way, keeping the two redox facies separate for Fe and Mn and instead defining a single value, relative to the entire preselected dataset (35 water points), for As and F, not affected For the 26 stations of the oxidizing facies, the available data relate to a single sampling (2017), for which only the spatial analysis was carried out. On the other hand, as regards the 9 water points in reducing facies, the dataset also includes the samples of January and July 2020. For the two parameters, a temporal processing was therefore carried out, defining a representative value for each station, given by the median of the values measured in the three campaigns. Consequently, the As and F dataset used for the spatial analysis consists of 35 data, of which 26 individual values of the oxidizing facies and 9 median values of the reducing facies.

by the redox conditions. *3.2. NBLs Calculation for As and F*  For the 26 stations of the oxidizing facies, the available data relate to a single sampling (2017), for which only the spatial analysis was carried out. On the other hand, as regards the 9 water points in reducing facies, the dataset also includes the samples of Jan-For both parameters, the presence of anomalous data was then analyzed using the Huber's non-parametric test, which identified two outliers for the F and one for the As. In the current state of knowledge, however, there are no valid scientific reasons to exclude these samples, which were therefore included in the final dataset for the definition of NBLs, consisting of 35 data. It is therefore the case B foreseen in the Guidelines, in which there is a significant spatial but not temporal dimension.

uary and July 2020. For the two parameters, a temporal processing was therefore carried out, defining a representative value for each station, given by the median of the values The Shapiro–Wilk test was then applied to verify the normality of data, showing that only F follows a Gaussian distribution. Consequently, according to the Guidelines, the NBL

measured in the three campaigns. Consequently, the As and F dataset used for the spatial

for F is given by the maximum of the statistical sample (3.56 mg/L), while for As it is equal to 95th percentile of the same statistical sample (20.5 µg/L).

Finally, considering the aquifer type (unconfined), its extension (60 km<sup>2</sup> ) and the number of total samples (35), for both parameters it was possible to associate a high confidence level to the defined NBLs.

#### *3.3. NBLs Calculation for Fe and Mn (Oxidizing Facies)*

As previously observed, for the 26 MSs in oxidizing facies, the available data belongs to a single sampling survey (2017), therefore only the spatial analysis was carried out. Huber's test found 2 outliers for Fe and 4 for Mn. However, also in this case there any scientific elements were identified to discard these data, hence the final dataset for the definition of NBLs remains the original one of 26 data. This is again the case B described in the Guidelines, in which there is a significant spatial but not a temporal dimension.

As expected, the subsequent Shapiro–Wilk test showed how the distributions of the two elements are far from normal, so in both cases the NBL should be set to the 95th percentile of the statistical sample (105.1 µg/L for Fe and 9.1 µg/L for Mn) and for both parameters a high confidence level was associated to the calculated NBL.

#### *3.4. NBLs Calculation for Fe and Mn (Reducing Facies)*

As regards the reducing facies, the dataset consists of only 9 water points, corresponding to the case D indicated in the Guidelines, the worst in terms of available information, in which neither a spatial nor a significant temporal dimension is reached. In these situations, the Guidelines suggest to estimate a provisional NBL, combining all available observations (in space and time). For Fe and Mn the total observations available are 26. Huber's test did not highlight possible outliers, so the 90th percentile of the observations for both parameters was calculated. The estimated NBLs are equal to 7364.7 µg/L for Fe and 805.0 µg/L for Mn. The associated confidence level is low and new monitoring observations are needed to improve the reliability of the dataset.

The calculated NBLs and the associated confidence levels are shown in Table 3. As indicated in the Guidelines, the NBL for each parameter was expressed with the same unit of measurement and rounded up with the same number of decimals as the relative limit set by Decree 152/2006 [44]. At present, as indicated by the Guidelines, for Fe and Mn the NBL for the groundwater body will correspond to that defined for the oxidizing facies, which has the highest confidence level.


**Table 3.** Calculated NBLs and associated confidence levels.

#### **4. Discussion**

The study on the natural background values for the southern Mounts Vulsini groundwater body confirmed the presence of high concentrations of As and F, which show numerous exceedances of the TVs set at the national level by Decree 30/2009 (1.5 mg/L and 10 µg/L) [45] and are naturally present in groundwater. The relatively low range of concentration, the existence of single data populations and the normal (for fluoride) or close to normal (for arsenic) distributions suggest that the presence of the two elements in groundwater can be largely attributed to the water–rock interaction processes within the volcanic aquifer. The statistical analysis does not highlight further overlapping phenomena and this also translates into the definition of a particularly reliable NBL for the investigated groundwater body. Fe and Mn show a wide range of concentrations, depending on the redox conditions found in the aquifer. In the most common oxidizing conditions, the

natural concentrations are generally low, while in the reducing conditions that can be found locally, the values are considerably higher and well above the limits set by Decree 152/2006 for the assessment of pollution while monitoring impacts on groundwater of e.g., industrial activities. The nature of these few peculiar points should be further investigated, in order to exclude any contribution of anthropogenic origin (currently not identifiable and quantifiable). Unlike the oxidizing facies, the few points belonging to the reducing facies do not show a particular spatial distribution within the GWB. This results in a geochemical population that can hardly be better characterized, due to the difficulty in finding other equally peculiar sampling points, which would make the dataset statistically more robust. The reliability of the NBL in this case can be improved mainly by monitoring these MSs, thus increasing the number of observations over time.

During the application of the Italian Guidelines for the NBLs assessment, some critical issues emerged, partly associated with the Guidelines themselves, others more specific of the online tool.

With regard to the first phase of the procedure, MSs specific, the software allows to perform the separation of the redox facies or alternatively to keep a single dataset. However, this choice cannot be applied in a different way depending on the parameters whose NBL is to be determined. Consequently, the procedure must be repeated twice, for the total dataset and for the separate facies.

Moreover, it is not clear in the Guidelines whether a temporal analysis of the preselection markers (nitrates/ammonia) should be done. Currently, eNaBLe calculates the median of the temporal data and if this is < the selected limit (e.g., 37.5 mg/L for nitrates or 0.375 mg/L for ammonia), the MS is considered useful for the calculation of NBLs; otherwise, it is discarded. However, this entails the risk of considering stations that have exceedances of these limits. In addition, since an analysis of temporal trends is not envisaged, stations that show ascending temporal trends of the markers, suggesting contamination in progress, could be included in the preselected dataset. In this regard, an official Guideline including different statistical methods for trend analysis, estimating concentration scenarios and identification of trend reversal, have been published in Italy in 2017 [46]. The use of these techniques, currently to be applied externally to the tool, could be helpful to integrate the temporal analysis of data, in particular for substances of clear anthropogenic origin such as nitrates. About this, Frollini et al. [47] have recently applied a slightly modified version of the Guideline to a groundwater body in Northern Italy featuring nitrate pollution, discussing its advantages and limitations.

Furthermore, the software does not currently allow to evaluate the correlation between chemical elements and redox parameters or even other chemical-physical parameters; it is therefore not possible to statistically evaluate whether they are redox-sensitive (operation currently executable only externally to the tool) and make a justified choice of the facies separation.

As regards the second phase of the procedure, parameter-specific, the software evaluates the presence of data outliers through application of the non-parametric Huber test (or alternatively, extrapolating them from a simple boxplot). This is clearly a simplification of the procedure, as non-parametric methods do not require any knowledge or assumptions about the form of data distribution. These are robust techniques, as they are applicable to any situation. However, the tool could also provide for the possibility of applying parametric tests (e.g., Rosner test), that are statistical procedures based on the assumption of normality of the data, more powerful and preferable to non-parametric tests, in particular when the data follow the hypothesized distribution. In the case study, for example, the Huber test identifies 2 outliers for F. However, the distribution of data is normal, both including and excluding the outliers, and the Rosner test [48], recommended for Gaussian distributions with more than 25 data, applied externally to the tool, does not detect anomalous values. Hence in the definition of the NBL we have included the outliers, since their exclusion is not supported by evident scientific reasons. However, the adoption of a parametric test, in this case, would have simplified the path of definition of the NBL.

Furthermore, again regarding the statistical study, it is limited only to the evaluation of the normality of the data and the presence of anomalous values, which can be correctly discarded or not. The evaluation of the existence of multiple populations, which would lead to a subdivision of the dataset before calculating the NBLs as required by the Guidelines, should be conducted aside. eNaBLe shows the distribution of the data at the end of the procedure, through Q-Q plot, but any subdivision of the dataset before the calculation is not allowed. In this regard, the possibility of implementing further partitions of the dataset, based on the study of the conceptual model of the GWB and conducted outside the tool, is currently being evaluated.

About the temporal analysis of data, currently the software allows to calculate a representative value for each monitoring station through the calculation of the median or, alternatively, through the elimination of any outliers and using the maximum value of the residual distribution. Therefore, at the moment the evaluation of temporal trends (step foreseen in the Guidelines), which could lead to the elimination of MSs that do not show outliers but simply increasing trends for certain parameters, suggesting a possible contamination in progress, should be performed aside. In the direction of improving the tool in this part, the implementation of a statistical test for the estimation of the slope of ascending trends (e.g., Mann–Kendall test) [49,50] is planned.

Finally, for the cases in which there is neither spatial nor temporal significance (Case D), without further specification by the Guidelines, eNaBLe evaluates the outliers putting together all the observations. Indeed, Case D refers to a GWB with a small number of MSs (to the limit of one only MS); therefore, if we reduce the multiple time observations to one representative value (median) for each station, we could not perform a robust statistical evaluation of the outliers.

Following the Guidelines, for each parameter only one NBL can be defined even in presence of multiple datasets, e.g., when different geochemical facies or redox conditions have been recognized and evaluated separately in the GWB. The Guidelines suggest choosing the highest value among those of the single datasets, only in case D will the NBL with the highest confidence level be selected. At present, therefore, the NBLs for Fe and Mn assigned to the entire GWB would be those, rather low, specific to the oxidizing facies. However, it is apparent that the high concentrations of these metals are to be associated with the peculiar reducing conditions that are found locally within the GWB. In order to assign them a NBL more representative of the conditions which promote their presence in solution, it is necessary to continue the temporal monitoring until reaching case C (significant temporal dimension), or to increase the number of MSs until reaching case B (significant spatial dimension). Only when the confidence levels are all the same, it will be possible to select the NBL given by the maximum value among the different datasets, therefore the one (clearly higher) associated with the reducing facies.

#### **5. Conclusions**

eNaBLe is a versatile and user-friendly instrument, meant to facilitate the assessment of the NBLs following the national Guidelines. It performs automatically the sequence of operations as indicated by the Guidelines, easily allowing a very quick calculation of NBLs at the groundwater body scale, even in the presence of a limited amount of data. By setting a few configuration parameters, an assessment is rapidly reached which appears sufficiently representative of the natural state of groundwater. However, it should be noted that these settings must be made in a reasoned way, starting from a thorough knowledge of the conceptual model of the aquifer. In particular, the redox facies separation assumes that redox conditions of groundwater are known. Its determination is commonly based on physical-chemical indicators such as DO or ORP whose measure is often critical. Further, this information is frequently missing or scarcely reliable in datasets of groundwater quality monitoring. The validation of the data and their organization into coherent datasets are also fundamental to obtain an evaluation that is as significant as possible.

Some critical issues have emerged in the statistical analysis path, even in part deriving from the reference Guidelines, but the software can be improved and refined in every its part starting from the indications and criticalities deriving from these first applications. The tool is still under testing for a thorough verification of all the possible variants in dataset structures and GWB conceptual models, also taking into account the suggestions of the stakeholders.

**Author Contributions:** Conceptualization, D.P., E.F., E.P. and S.G.; methodology, D.P., E.P. and S.G.; software, S.G.; validation, D.P.; formal analysis, D.P.; investigation, D.P., E.F., E.P. and S.G.; data curation, S.G.; writing—original draft preparation, D.P.; writing—review and editing, D.P., E.F., E.P. and S.G.; visualization, D.P. and E.F.; supervision, E.P. and S.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** The tool developing was not funded. Data used as case test were collected within a project funded by Regione Lazio (Area Ciclo Integrato Rifiuti), Italy (Det. G13172 18/10/2018) and Ecologia Viterbo s.r.l. (Contract 3269-2017 and 178/2020).

**Data Availability Statement:** Paper experimental data are not currently publicly available but can be supplied on request to the corresponding author.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


*Article*
