1. Introduction
The strict EU legislation on groundwater has pushed analytical methods to new frontiers [
1,
2]. The main aim of this study was to develop a method for the single-shot detection of a broad range of organic compounds in relatively clean natural waters in the simplest and most cost-effective way possible. Organic compounds, identified as one of the emerging pollutant groups in groundwater, are now widely used in a range of human practices and activities. Their number is further increased by several million metabolites and degradation products of the parent compounds. Determining their concentration in groundwater is not practical or even feasible. In contrast to traditional organic pollutants, such as pesticides, aromatics, and halogenated solvents, recent research has focused on the detection of pharmaceuticals and other persistent chemicals in the environment [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18]. As a result, a broad range of organic compounds with very diverse physicochemical properties should be analysed using techniques that approach the limit of detection (LOD), not only to determine well-defined contamination clouds, but also to obtain a clear picture of pollution at very low concentrations, or even before the first relevant signs of pollution appear. The financial framework for such analytical techniques should remain reasonably cost-effective. Various analytical methods have been developed to measure organic contaminants in water. Hyphenated gas chromatography–mass spectrometry (GC-MS) for non-polar and moderately polar organic compounds, and liquid chromatography–mass spectrometry (LC-MS) techniques for polar compounds, are the most commonly used today [
1,
2]. Sample preparation is one of the crucial steps in obtaining reliable results below the ppt level with minimal secondary contamination.
The detection of contaminants in groundwater depends on the measurement approaches and techniques employed, of which two principal methods exist. The first is non-target screening (with no prior information), and the second is suspect or target screening (for suspect substances based on prior information) [
19]. While non-target screening is time-consuming, it provides important information on the compounds present in the sample. On the other hand, studies that use suspect screening are published more frequently [
3]. Studies of this type have identified low and variable concentrations as one of the several challenges involved in determining organic compounds in groundwater.
Interest in the principles of qualitative analysis has grown since 1997 [
20,
21], leading to an increase in the number of publications presenting detailed studies on screening methods [
22,
23,
24], their validation, and the uncertainties associated with estimation procedures [
25,
26,
27,
28]. Screening methods were first placed in a broader analytical context with the proposal by Valcárcel and Cárdenas [
29] on vanguard and rear-guard analytical strategies. Qualitative analyses are based on confirming the presence of chemical substances and providing analytical information from binary yes/no responses in a short time (e.g., detection of a chemical substance), while rear-guard analytical systems perform a complete analytical process for quantification purposes. Between these two approaches lies semi-quantitative analysis, discussed in a report by [
30], which deals with the validation of qualitative and semi-quantitative methods. Ultimately, its aim was to establish an acceptable balance between being “sufficient to detect problems” and not being so extensive that it would be too costly. This practical limitation is also the main reason for not complicating validation procedures for qualitative methods [
30].
However, groundwater monitoring programmes are still largely based on the collection of grab samples. This approach provides a snapshot of contamination at a particular point in time and may therefore not be truly representative of relevant environmental conditions over time. More than a decade ago, passive sampling was introduced as an attractive alternative to sampling natural waters [
3,
17,
31,
32,
33,
34,
35,
36,
37,
38,
39]. In 2012, ISO 5667-23 was published as the first standard for passive sampling of surface waters, followed by ASTM D7929-14 for passive sampling of groundwater in 2014. Compared to traditional sampling methods, passive sampling is less sensitive to random extreme variations in the concentration of organic pollutants in natural waters, and a wide range of contaminants can be detected simultaneously. A passive sampler can cover a long sampling period and integrate pollutant concentrations over time. Although compared to conventional monitoring, the use of passive samplers can significantly reduce analytical costs, a validation procedure that includes an assessment of the degree of sampling uncertainty remains a challenge [
15,
17,
40,
41,
42,
43,
44,
45,
46,
47,
48,
49].
The passive sampling procedure has been proven to be a powerful tool in preliminary observations of aquifers for periodic verification of an analyte list for quantitative monitoring, for early detection of various anthropogenic impacts on aquifers, and for other hydro-geological studies [
40]. In response to the lack of methods for simple and effective routine monitoring of microorganic substances, the aim of our research was to develop an analytical method using activated carbon fibre (ACF) passive sampling as a qualitative sampling method that complies with the ISO 5667-23:2011 standards for the detection of a wide range of organic compounds in groundwater by GC-MS. The positive properties of active carbon adsorption have encouraged researchers to use active carbon in virtually all areas of chemistry, mainly because of the simplicity of its design and operation, its selective tendency towards certain substances, and its complete elimination of pollutants, even from diluted solutions. This has led to an intensified search for durable, reliable, and selective alternatives for the protection and conservation of the environment [
50,
51]. This paper presents the development and application of this analytical method. Compared to other published passive sampling methods [
8,
31,
41,
49,
52,
53,
54,
55,
56,
57,
58,
59,
60,
61,
62,
63], the described procedure is one of the simplest and most efficient amongst those that allow for the detection of a wide range of organic compounds.
The development of new sampling and analytical methods or approaches is of great importance for hydrogeological studies. In groundwater, most pollutants are usually present at concentrations below the limit of quantification (LOQ), some of them also below the limit of detection (LOD) in the ppt and sub-ppt range. To date, researchers have investigated a wide range of emerging pollutants in groundwater, searched for possible sources of pollution, and studied the dynamics of pollution in aquifers and similar environments [
15,
64,
65]. Many of these studies are difficult to transfer to larger areas due to the cost of the analyses, as well as sampling and transport issues. The optimisation of precisely these aspects was the main goal of our research.
The applicability of the ACF method has been tested in various groundwater quality studies. This article presents the following three examples of its applicability: (a) detection of a wide range of organic compounds in groundwater at a regional scale, (b) the results of passive ACF sampling designed to analyse the possible source of contamination in the aquifer used for public water supply, and (c) comparison of grab and passive sampling results. In particular, we wanted to test whether this method could be used to detect organic compounds that had not been previously detected in the analysis of grab samples, i.e., emerging compounds.
The objectives of the present study were as follows: (a) to develop a non-target screening method that is simple and repeatable with the GC-MS analysis; (b) to simultaneously detect the presence of a wide spectrum of organic pollutants in groundwater; (c) to detect the presence of numerous organic pollutants, including those not yet detected by conventional analytical methods; (d) to apply the developed method in the field to investigate the origin and occurrence of organic pollutants in aquifers.
2. Materials and Methods
2.1. Chemicals and Reagents
Pure standards and standard solutions of analytes were purchased from Dr. Ehrenstorfer (Augsburg, Germany) and C/D/N Isotopes Inc. (Pointe-Claire, QC, Canada); MTBSTFA and t-BDMCS were obtained from Acros Organics (Geel, Belgium). Dichloromethane (DCM), HPLC special grade, acetone HPLC, and methanol (MeOH) were purchased from Rathburn Chemicals Ltd. (Walkerburn, UK). Ultra-pure water (UPW) used for AFC storage was purchased from EASYpure LF (Barnstead Thermolyne International, Dubuque, IA, USA).
All-glass syringes were obtained from Poulten & Graf (Wertheim-Reinhardshof, Germany), PTFE membrane filter, 0.2 µm from Sartorius AG (Göettingen, Germany), all-glass test tubes and weighing bottles from Lenz Laborglas (Wertheim, Germany); and helium (99.9%) and nitrogen gas (99.9%) from Messer Slovenija d.o.o. (Ruše, Slovenia). An SPE vacuum unit for evaporation of 12 samples was obtained from Grace/Alltech (Deerfield, MA, USA) and a Swinny filter holder (13 mm, stainless steel) from Merck (Darmstadt, Germany).
2.2. Sample Preparation
Zorflex
® FM10 active carbon fibres (ACFs) for passive sampling devices (
Figure 1), were purchased from Calgon Carbon Corporation (Pittsburgh, PA, USA). ACFs were used as a strong and reliable adsorbent [
50]. In one day, approximately 8 samples could be analysed in a GC-MS run consisting of samples, quality controls, and with a pure DCM analysis between each two samples.
2.2.1. ACF Purification and Transport to Sampling Site
Before installation, appropriate portions (3 cm2 each) of ACFs were heated for three hours at 300 °C in clean air from IQAir air purifiers. ACFs were transferred to the test tubes using tweezers and were not allowed to cool. Before cooling, a few drops of ultrapure water (UPW) were added to generate steam. The test tubes were then filled with UPW and sealed. The sealed test tubes were put in a flask with UPW and active carbon at the bottom of the flask and then transported to the installation point.
2.2.2. ACF Deployment
At the installation point, stainless steel meshes were equipped with ACFs just before installation (
Figure 1). All stainless steel materials used for ACF installation were purchased from MDM (Ljubljana, Slovenia). Stainless steel mesh was used for the passive sampler casing and was installed in an appropriately deep borehole on a stainless steel wire. Groundwater temperature, electrical conductivity, pH, and total organic carbon (TOC) were measured before and after installation of the passive sampler.
2.2.3. ACF Collection and Elution of Compounds
After exposure of passive sampling devices, the ACFs were immediately transferred out of their casings to weighing bottles filled with UPW using tweezers and transported to the laboratory in additional weighing bottles filled with UPW.
2.2.4. Elution of Compounds from ACFs
The UPW from the weighing bottles was removed and the ACFs were dried in an oven at 100 °C for one hour. The 3 mL of extraction solvent (DCM with 5% MeOH), along with internal standards caffeine-D9, phenol-D5, estrone-D4, and cholesterol-D4, was added to the weighing bottles for the elution of compounds, which was performed in an ultrasonic bath for 30 min. Approx. 2 mL of the remaining extraction solvent was collected in a syringe, filtered into a chromatographic vial, and concentrated to approx. 50 µl by nitrogen steam. As an alternative, derivatisation using a mixture of N-(t-butyldimethylsilyl)-N-methyltrifluoroacetamide (MTBASTFA) and 1% tert-butyldimethylsilyl chloride (t-BDMCS) was performed directly from the ACFs with the addition of 100 µl of a derivatisation reagent by reaction in an ultrasonic bath for 1 h, followed by the addition of 3 mL of extraction solvent and the elution of compounds in an ultrasonic bath for an additional 30 min. Two ACFs were installed at the same sample location in order to perform the derivatisation on one of them.
2.3. Analytical Method–Chromatographic Analysis
All samples were analysed using gas chromatography with mass spectrometry (GC-MS) (GC 6900/QP 5890, Agilent, Folsom, CA, USA) using a GERSTEL autosampler (Mülheim an der Ruhr, Germany) and a GC-MS Clarus 600 (PerkinElmer, Waltham, MA, USA), with a CombiPAL Autosampler (CTC analytics, Zwingen, Switzerland). Silanized injection liners with silanized glass wool were procured from SGE International Pty Ltd. (Ringwood, Australia), a DB-5ms ultra inert column, 30 m × 0.25 mm I.D., df 0.25 µm from Agilent (J&W Scientific (Folsom, CA, USA)); IQAir GCX air purifiers (INCEN AG, Staad, Switzerland); and an HP-5MS UI column, 30 m I.D. 0.25 mm, df 0.25 μm from Agilent.
The operational settings for Agilent GC-MS were as follows: Inlet at 270 °C; transfer line at 280 °C; ion source at 250 °C; quadrupole at 150 °C. Injection: 2 µl pulsed splitless (pressure, 300 kPa for 1 min). Carrier gas flow: He, 1.0 mL/min (constant flow). Total flow of 54 mL/min. Oven program: 35 °C (1 min) to 270 °C, with a total run time of 46.0 min. Total ion chromatogram (TIC) from 30 to 750 m/z.
Operational settings for the Perkin Elmer GC-MS were as follows: Inlet at 80 °C, after injection to 280 °C; transfer line at 280 °C; ion source at 250 °C; injection: 8 µl pulsed splitless (column flow 5 mL/min for 2 min); carrier gas flow: He, 1.0 mL/min (constant flow after 2 min); total flow 53 mL/min after 1.5 min; oven program: 35 °C (1 min) to 280° C, with a total run time of 50.0 min. TIC from 30 to 750 m/z.
Chemstation and Agilent Deconvolution Reporting Software (DRS) with retention time library 5989-5076EN (Agilent) and NIST 2008 spectral library and TurboMass software (PerkinElmer) and NIST 2008 spectral library were used for the interpretation of chromatograms. The GC-MS chromatograms were interpreted in two different ways. The first consisted of both a manual and an automatic evaluation of chromatograms to search for compounds. The results for the detected compounds were ranked on a scale from one (lowest) to five (highest) according to peak relative intensity compared to other peaks in the same GC-MS chromatogram. Only a few compounds with a maximal peak (5) area were reported. Others were ranked between two and four by the operator, according to their peak areas. Additional results from AMDIS deconvolution covered by commercial databases were also reported with rank-1 compounds.
The second evaluation of GC-MS chromatograms employed the integration of the most abundant mass fragments from the discovered compounds of interest, with the mass fragment m/z 203 of caffeine-D10 used as an “internal standard”. Ratios between the areas for compounds of interest and caffeine-D10 were calculated. Caffeine-D10 was chosen owing to its good mass spectrum definiteness and moderate polarity, which is similar to the average polarity values for the compounds of interest. The relationship allows for the comparison of different groundwater samples; consequently, the spatial distribution of the compounds can also be determined and shown.
2.4. Quality Control
The field and laboratory blank passive samplers were prepared, extracted, and analysed in parallel with each exposed sampler for quality control (QC) of the analytical method.
2.4.1. Sampling QC Procedure
Field blanks were tested by exposing the passive samplers to air at each installation point. The blanks were transported to the laboratory in the same way as the samples and were retained in the laboratory until the passive samplers were collected. Once the passive samplers were collected, the procedure was repeated using the blanks. The blanks were then analysed in the same way as the samples.
2.4.2. Analytical QC Procedure
Targeting retention times and signal-to-noise (S/N) calculations were subjected to the following quality controls: A first quality control solution (QC1) containing more than 100 semi-volatile compounds and a second quality control solution (QC2) of volatiles in DCM were injected using the same method. These compounds had similar physical and chemical properties as the compounds from QC4 (
Table 1). A 5 mg/L solution of endrine and p,p’-DDT was used as a third quality control solution (QC3) to test the degradation of sensitive compounds according to the EPA 525.2 method. Maintenance on the injection port was performed if the degradation exceeded 20% of the parent compounds.
Before the installation of passive samplers, the purity and adsorption capacity of each series were tested with two control solutions (QC4 solution and QC5 solution), whose compositions were selected according to their physio-chemical properties in order to represent a whole range of compounds of interest. In the QC4, 5 µl of the standard solution of analytes at approx. 100 mg/L from
Table 1 was added to 20 mL of UPW. A second control solution (QC5) was used for derivatisation, where 5 µl of a standard solution of analytes at approx. 100 mg/L from
Table 2 was added to 20 mL of UPW. The results for both control solutions are presented in
Table 1 and
Table 2. The TIC of both control solutions–QC4 and QC5–enabled all compounds from
Table 1 and
Table 2 to be identified, which confirms the criteria used for the successful analysis of passive samplers. The ACFs in both solutions were stored in a refrigerator overnight. The ACFs with the adsorbed compounds were rinsed with UPW and processed as regular samples, with or without the derivatisation step. One blank was processed in the same way as the samples for each sample site in each series. At derivatisation, a by-product estrone-TBDMS was also detected at t
R 38.4 min as described in the literature [
66] (see
Section 3.3).
In each series, an additional solvent blank (2 mL of DCM concentrated in the same way as the sample extracts) was analysed. In the chromatographic run, sample extracts, blanks, pure DCM, and control solutions with typical organic contaminants, respectively, were analysed. The results for the samples were substituted with those for the blanks.
2.5. Validation of Method
A cleaning procedure for ACFs was developed to preserve the original adsorption properties of ACFs after efficient cleaning. Adsorption properties were tested sensorially by shaking tetrahydrothiophene (a characteristically strong odour) and uranine (a characteristically strong colour) solutions with the cleaned original ACFs. The same adsorption properties of tetrahydrothiophene and uranine on ACFs were observed using this cleaning procedure.
We followed the validation guidelines for qualitative analytical methods [
30]. To this end, we performed repeatability experiments for extracts and blank ACFs. Positive identifications together with the observed intensities from passive samplers were compared with the results from the quantitative method used for grab samples [
67] in order to determine cut-off concentrations for the target group of compounds. Cut-off concentrations were determined by establishing false positive and false negative rates at a range of values above and below the expected cut-off concentration. The cut-off limit marks the point at which false negative rates for concentrations above the limit are low [
30]. Comparisons of the ratios of peak compound areas to caffeine-D10 were also possible with concentrations in grab samples collected at different times from the same research well for sampling points with relatively stable organic pollution.
2.6. Data Sets of Presented Examples
To test the usability and reliability of ACF passive samplers, we collected two sets of data to show three usability examples. The first refers to a set of data from various aquifers from the entire territory of Slovenia and the second refers to the pilot area of the Ljubljansko polje aquifer.
Slovenia: With an area of 20,273 km2, Slovenia ranks as a medium-sized European country. In Europe, Slovenia is among the countries with the most abundant groundwater resources. Its groundwater is located in aquifers with granular, fissured, and karstic porosity. The efficacy of the ACF method was tested on 470 groundwater samples from all over Slovenia in the period 2013–2019.
Ljubljansko polje aquifer: An aquifer of intergranular porosity consisting of the deposits of the Sava River. In some places, the deposits lie deeper than 100 m [
68,
69]. For the most part, the aquifer is open; in some areas less permeable layers appear. The hydraulic permeability of the aquifer layer is high, ranging from 1.2 × 10
−2 m/s in the central part of the field to 3.7 × 10
−3 m/s at the edge of the field [
70]. The experimentally determined groundwater flow velocity is estimated at 25 m/day, but it is also estimated that the groundwater velocity in the area ranges from a few metres to a few tens of metres per day. The groundwater generally flows parallel to the Sava River, from west to east. The aquifer is mainly recharged by two components: the Sava River and infiltrating precipitation [
71]. The catchment area is subjected to a variety of human activities that affect the quality of the groundwater.
The sampling design network used for testing the method covered the entire Ljubljansko polje aquifer area (
Figure 2). Passive sampling devices were installed in 13 observation wells in the saturated zone where contact with groundwater was constant. The exposure time of each passive sampler was 3 months. In case No. 2 (described in
Section 3.6.2), a total of 47 samples were collected across five campaigns during the period 2012–2017. The data for case No. 3 (
Section 3.6.3.) were obtained for both passive and point samples over the period from March to November 2015. Field parameters were also measured at the same time. During the sampling period, the temperature values varied between 7.00 and 14.04 °C, the pH electrical conductivity (at 20 °C) ranged from 326 to 838 μs/cm, the pH of water samples ranged from 7.2 to 7.8, and the total organic carbon (TOC) ranged from 0.20 to 0.61 mg/L during the exposure of passive samples in groundwater.