Introduction

The contamination of the soil, air, and groundwater by liquids leaking from the underground storage tanks of gas stations can both impact the environment and have irreversible effects on human health (Hilpert et al. 2015; Moschini et al. 2005). These liquids may include fuel, toxic chemicals, and diluted effluents.

The underground storage tanks of gas stations represent a potential source of contamination, with the type and level of risk depending on the management of the stored fuel and the structural characteristics (e.g., type of soil, drainage) of the local landscape (Moschini et al. 2005).

When petroleum derivatives reach the soil, their components are partitioned into three phases: dissolved, liquid, and gaseous (Nadim et al. 1999). A small fraction of these components is dissolved in the aquifer, while a second fraction is retained in pure liquid form in the porous spaces within the soil as a residual saturation, and the third part evaporates and contaminates the atmosphere.

The spillage of hydrocarbon compounds may involve the contamination of the vadose and transition zones (the capillary fringe), which overlie the water table, by light non-aqueous phase liquids (LNAPLs). Below the water table, contamination is by dense non-aqueous phase liquids (DNAPLs). Hydrocarbons, such as gasoline, kerosene, and jet fuel, are common LNAPL contaminants (Jol 2009; Daniels et al. 1995; Domenico and Schwartz 1990).

The National Yearbook of Petroleum, Natural Gas, and Biofuel Statistics (ANP 2022) shows that a total of 42,401 gas stations were operating in Brazil at the end of 2021. Approximately 8.27% of these outlets are located in northern Brazil, of which, 1372 (3.24% of the total) operate in the state of Pará. Gas stations are commercial establishments that sell fuels derived from petroleum, and may thus be an important source of volatile organic compounds (VOCs) (Geraldino et al. 2021).

The presence of NAPLs in the subsurface can be inferred by the fact that the contaminated groundwater and soils will have lower electrical conductivity and lower relative permittivity than the surrounding, uncontaminated matrix (Atekwana et al. 2000). The relative permittivity of hydrocarbons ranges from approximately 2 F m−1 to 30 F m−1, compared with 80 F m−1 for water (Von Hippel 1961; Daniels et al. 1995), and the conductivity of hydrocarbons ranges from near 0 S m−1 to 0.02 S m−1 at frequencies of between 100 and 1000 MHz (Von Hippel 1961; Daniels et al. 1995).

Environmental pollution by the monoaromatic hydrocarbons benzene, toluene, ethylbenzene, and xylene (BTEX) is a widespread phenomenon. Human exposure to these compounds can lead to a number of health problems, including the induction of cancer, given that compounds such as benzene are classified by the World Health Organization (WHO) as potent carcinogens in humans. Benzene can damage the blood and the immunological system, and may cause cancers, such acute myeloid leukemia, and is also known to have genotoxic properties (Singh et al. 2017; Hilpert et al. 2015; Weelink et al. 2010; Badham and Winn 2007).

Hydrocarbon plumes may be delineated as areas of low resistivity if inorganic compounds are added to the contaminated groundwater for bioremediation. This will tend to increase the total amount of dissolved solids found in the subsurface (Benson et al. 1997; Asquith and Gibson 1982).

In addition to underground water, BTEXs may also contaminate surface water, the sediments of bodies of water, the soil, and the air (e.g., Bertolo et al., 2018; El-Naas et al., 2014; Falcó and Moya 2014; Junfeng et al., 2008; Nakamura and Daishima, 2005; CETESB, 2005; Ezquerro et al., 2004). Fernandes et al. (2014) identified toluene in the surface water of a small river basin and in a hydrographic microbasin. Bretón et al. (2021) evaluated the influence of seasonal and diurnal variations on the atmospheric concentrations of BTEXs. Allahabady et al. (2022) and Correa et al. (2012) evaluated the evaporative emissions from gas stations, and identified BTEX concentrations in the air of the study sites. Zhang et al. (2012) and Dutta et al. (2009) identified BTEXs in the air and reported the transfer of these compounds from the air to bodies of water by falling rain.

Different geophysical techniques (e.g., electrical resistivity tomography (ERT), induced polarization (IP), electromagnetic induction (EMI), magnetic, ground penetrating radar (GPR), and spontaneous potential, SP) have been applied increasingly to the investigation, description, and monitoring of areas contaminated by hydrocarbons (Biosca et al. 2020; Abbas et al. 2018; Subba Rao and Chandrashekhar 2014; Aal et al. 2001).

The GPR signal provides a high-resolution image of the contaminated zone, in particular in the vadose zone above the water table (Castro and Branco 2003). These authors detected the presence of LNAPLs in the vadose zone based on the presence of an area of strongly-reduced reflections, with enhanced reflections in the polluted transitional and saturated zones, and concluded that the saturated zone became more reflective due to the presence of hydrocarbons dissolved in water.

The present study investigated possible contamination by gas stations in the town of Bragança, in eastern Brazilian Amazonia. The data were collected in the vicinity of the town’s gas stations to identify possible plumes of contamination by hydrocarbon compounds in the subsurface. For this, the GPR was used in combination with gas chromatography–mass spectrometry to verify the possible contamination by BTEX in samples of groundwater collected near the gas stations. These methods were chosen for the present study due to their application in the diagnosis of contamination by hydrocarbons, rapid and non-destructive procedures, and their relatively low cost.

Materials and methods

The present study was based on the combination of an indirect investigation method, that is, ground-penetrating radar (GPR), and the direct analytical techniques of gas chromatography–mass spectrometry (GC–MS).

Study area

The study area is located in the town of Bragança in Pará state, northern Brazil (01°03′15″ S, 46°46′10″ W) which is located at an altitude of 19 m above sea level (Fig. 1). According to IBGE (2022) the municipality has an area of 2.124,734 km2, with an estimated population of 130,122 residents by 2021.

Fig. 1
figure 1

Study area and the location of the gas stations (red circles) surveyed in the town of Bragança (Pará, Brazil). Adapted and modified from IBGE (2015) and Google Earth Pro (2022)

The geographic region of Bragança, in the Brazilian state of Pará (IBGE 2017), in which the study area is located, was established on the local Neogene deposits. The basement of the coastal plain is composed of Neogenic–Paleogenic sediments of the Barreiras Group (Fig. 2), which forms the Coastal Plateau (Souza Filho and El Robrini 1996). This area is part of the Amazon coast, which extends from the mouth of the Oiapoque River, in the Brazilian state of Amapá, to the eastern extreme of Maranhão state. Bragança is located at the mouth of the Caeté River.

Fig. 2
figure 2

Source: Jorge (2017)

Geological map of northeastern Pará state (Brazil), showing the study area (Bragança, Brazil).

Geologically, the Bragança coastal plain is located within the Bragança-Viseu coastal basin, whose origin and evolution are closely linked to the formation of the equatorial Atlantic and the normal faults that permeate the current coastal zone (Souza-Filho and El-Robrini 1996). The geomorphology of this coastal plain has changed significantly in recent years, with the retraction of the mangroves from the coast, due primarily to the encroachment of the sand that covers the muddy substrates of the mangrove and asphyxiates the vegetation (Lara 2003).

The geometry and paleogeography of the region are associated with tectonic processes, which determined the thickness of the local Neogenic-Paleogenic and Quaternary deposits. The Neogene and Paleogene are represented in the study area by the Pirabas (Góes et al. 1990) and Barreiras formations (Rossetti et al. 2001), while the Quaternary is represented by sandy-clay sediments and Holocene deposits of the alluvial, estuarine, and coastal plains (Souza Filho et al. 2009) (Fig. 2).

The study area has a very humid, megathermic climate, with moderate water deficiency between July and December. Temperatures vary only discreetly over the course of the year, ranging from a minimum of 18 °C to a maximum of 33 °C, with an annual mean of 27 °C, and generally higher values between August and October. The region is dominated by the trade winds, primarily northeasterlies (with some shifts to the north and east), which are constant and moderate, with a maximum velocity of up to 7.5 m s−1 (Sousa Júnior et al. 2020; INMET 2020).

The climate of Pará state has two seasons, one rainy and the other, less rainy (or dry). The rainy season normally extends between January and May, while the dry season is from June to December (Magalhães et al. 2006). Precipitation in the rainy season ranges from 65 to 2744 mm, with a general mean of 1657 mm. In the dry season, precipitation ranges from 230 to 678 mm, with a mean of 487 mm (Moraes et al. 2005).

Ground penetrating radar (GPR)

We acquired GPR data to detect possible contamination plumes caused by hydrocarbons in the subsurface of the study area. Many previous studies have shown that GPR is an adequate geophysical tool for the detection of contaminated environments (e.g., Bacha et al. 2021; Aktürk and Doyuran 2015; Jiang et al. 2012; Castro and Branco 2003; Atekwana et al. 2000).

Six gas stations, identified as P01–P06 (Fig. 1), were surveyed in the present study. Ground Penetrating Radar uses electromagnetic (radio) waves of high frequency, that is 10–1000 MHz (Annan 1992), to survey underground structures. This geophysical tool is a non-invasive, versatile, and low-cost technique that can be used to survey an area rapidly, in both urban environments and areas of natural habitat. Ground Penetrating Radar has many geological applications, such as the high-resolution imaging of shallow soils and rock structures, the identification of buried conduits and cavities, the mapping of the water table, and archeology and forensic investigations. During the acquisition of data, a short pulse of the radar in the 10–1000 MHz frequency range is introduced into the subsurface. The velocity of the pulse is controlled by the dielectric constant and the conductivity of the subsurface (Kearey et al. 2002).

The dielectric constant is a dimensionless parameter often used in GPR applications, also known as the relative dielectric permittivity \(\left({\varepsilon }_{r}\right),\) defined by \({\varepsilon }_{r} = \varepsilon {{\varepsilon }_{0}}^{-1}\), where \({\varepsilon }_{0}\)=  dielectric permittivity in a vacuum \((8.854x{10}^{-12}F {m}^{-1})\). The dielectric constant expresses the amount of electrical energy stored when a substance is subjected to variable external electric fields. The finite pulse of the GPR can be considered to be composed of the superposition of monochromatic waves. As the source is finite, the amplitude of the resulting pulse is also affected by geometric scattering. This scattering causes only a decrease in the amplitude, given that it is a purely geometric factor that does not alter the frequency (Olhoeft 2000).

In a medium modeled by a perfect dielectric (constant phase velocity, attenuation coefficient equal to zero), then, the finite pulse captured by the receiving antenna should have the same form as the pulse emitted, and the frequency bandwidth received by the antenna should be equal to that emitted by the apparatus. Propagation effects cause the energy of the wave to decrease as it propagates. This loss of energy is manifested primarily in the attenuation, and is described by the absorption and dispersion, and by the geometric scattering of the wavefront. In other words, the attenuation occurs by geometric scattering and an increase in the conductivity of the soil (Annan 1996).

Absorption refers to the loss of the pulse amplitude that results primarily from the dissipation energy by its conversion to heat. Dispersion is reflected in the change in the shape of the pulse as it propagates. Geometric scattering accounts for the loss of energy as a function of the distance of the wavefront from the source, and is a factor that depends on the velocity distribution of the medium, although it does not vary with frequency (Neto and Medeiros 2006). The GPR data were collected at each gas station in both the rainy and thy seasons of 2017 and 2018. The data were acquired using a GSSI SIR-3000 system, with a 400 MHz antenna and time windows of 100, 150, and 200 ns (ns).

The system used for the acquisition of the electromagnetic signal was bi-static, with the common-offset (CO) configuration between the transmitting and receiving antennas. The GPR data were processed using the Reflexw software. We applied the following processing sequence (Delgado et al. 2022; Bacha et al. 2021; Sandmeier 2018; Rocha et al. 2015; Annan 1996): edition of the profile orientation, static correction of zero-time, gain-energy decay, temporal filtering (band-pass), dewow filter, running average, background removal, and time-to-depth conversion. The hyperbola overlap method was used to determine the propagation velocity, and the time-depth conversion was based on a velocity of 0.09 m ns−1.

Collection of water samples

Samples of groundwater were collected at each gas station and in the surrounding area in the rainy, rainy-dry transition, and dry seasons of 2017, 2018, 2021, and 2022. A total of 66 water samples were collected for the analysis of volatile organic compounds (BTEX). The samples were obtained from boreholes at the gas station sites and in adjacent residential areas. Informal interviews with local well diggers and the SIAGAS (2020) database confirmed that in general the water table of the study area is located at a depth of 6–10 m or less than 6 m.

All the collection points had a faucet installed. The faucet was first opened to drain for 2–3 min or long enough to eliminate the stagnant water in the pipe. The faucet was then disinfected by applying a sodium hypochlorite solution (100 mg L−1), with the excess solution being removed prior to the collection of the samples. A 100-mL sample of the water was collected from each point in a sterilized amber glass bottle, which was sealed and stored 4 °C, to avoid the volatilization of the hydrocarbons. These bottles were transported to the city of Belém, capital of the Brazilian state of Pará, for analysis at the Toxicology Laboratory of the Environment Sector (SAMAM) of the Evandro Chagas Institute/Brazilian Ministry of Health (IEC/SVS/MS).

The samples were collected and preserved in accordance with the methods established in the 21st edition of the Standard Methods for the Examination of Water and Wastewater, method SM-6010 B (APHA 2005), and the National Water Agency (ANA) of the Brazilian Ministry of the Environment (ANA 2011). The BTEX concentrations were determined by gas chromatography–mass spectrometry (GC–MS).

Gas Chromatography (GC) is a widely used technique for the identification of hydrocarbons and the quantification of mixtures of organic compounds (Pavón et al. 2007; Côcco et al. 2005). To analyze water samples containing BTEX, GC is normally combined with others techniques, such as Mass Spectrometry, or MS (Lee et al. 2007; Pavón et al. 2007; Jochmann et al. 2006; Nakamura and Daishima 2005).

The BTEX values identified in the present study were compared with the Maximum Value Allowed (MVA) in drinking water under Brazilian legislation. The MVA for benzene in water destined for human consumption is 5 μg L−1, while that of toluene is 30 μg L−1, ethylbenzene is 300 μg L−1, and xylene is 500 μg L−1 (BRASIL 2021).

Extraction of the BTEX from the water samples

The BTEX were extracted from the water samples using the automated Headspace (HS) method, which is based on the D 6040 method (Wallace and Stenerson 2008). Aliquots of 15 mL of the samples were placed in 20 mL bottles, sealed with an aluminum seal and teflon septum. The samples were then wrapped in aluminum blocks, placed in a Triplus RSH autosampler, and heated to 80 °C for 10 min.

The composition of the samples was quantified using a Trace 1300 gas chromatograph (Thermo Scientific) coupled to a TSQ 8000 mass spectrometer (Thermo Scientific), using a TG-5MS column (Thermo Scientific) composed of 5% phenyl and 95% dimethylsiloxane, with 30 m × 0.32 mm × 0.25 µm film. The oven temperature of the column was maintained at 40 °C for 1 minute, then raised to 70 °C, at 5 °C per minute, and finally to 70–220 °C at 30 °C per minute. The Detection Limits (DLs) considered here were 0.05 μg L−1 for benzene, ethylbenzene and xylene, and 0.1 μg L−1 for toluene.

The carrier gas was helium (99.999% pure) used at a flow rate of 1.0 ml min−1. The injector was operated at 280 °C in the split mode. The transfer line temperature was 250 °C, and the ion source was present at 230 °C. A 1000 µL aliquot of the sample contained in the headspace was injected into the chromatograph.

Results and discussion

Two gas stations (P02 and P06) presented radargrams which had a region of attenuated reflections or low amplitudes in the vadose zone. The GPR profiles of station P02 were obtained from transects of between 10 and 16 m in length, while at P06, the profiles were between 30 and 43 m long. The electromagnetic signal was generated at 2 m intervals along the length of each profile.

At gas station P02, hyperbolic electromagnetic signals were identified parallel to the location of the fuel pumps in the radargrams of profile L1 (Fig. 3), which correspond to the storage tank caps. Hyperbolic electromagnetic signals corresponding to the storage tank caps were also identified in the radargrams of profile L3 (Fig. 4), which is perpendicular to L1, together with responses that correspond to the pipes in the ground in the vicinity of the fuel pumps. Hyperbolic electromagnetic signals were also identified in the radargrams obtained from profile L4 at gas station P06 (Fig. 5), which corresponded to the locations of the storage tanks and pipes, as well as responses that correspond to the base of the support column of the gas station, as shown by Fuente et al. (2021), who also used GPR to evaluate the contamination of the subsurface of a gas station by hydrocarbons, and were able to identify pipes and tanks in the form of hyperbolic electromagnetic signals. Srigutomo et al. (2016) were also able to identify pipes in the subsurface of a petroleum refinery in Sumetera (Java, Indonesia) using GPR, which revealed hyperbolic signals similar to those recorded in the present study.

Fig. 3
figure 3

Radargrams obtained from profile L1 at gas station P02 in (A) April 2018 and (B) December 2018

Fig. 4
figure 4

Radargram obtained from profile L3 at gas station P02 in December 2017

Fig. 5
figure 5

Radargram obtained from profile L4 at gas station P06 in December 2017

The radargrams collected during the dry season also revealed the presence of low amplitude zones and hyperbolic electromagnetic signals, indicating the location of the underground tanks and their connecting pipes (Figs. 3, 6, and 7). These areas of low reflection are located in close proximity to the storage tanks. These features correspond to a hydrocarbon vapor phase in the vadose zone, or the region between the surface and the water table, in which the pores in the rock are not filled completely with water, as described by Castro and Branco (2003). A number of other studies have also applied GPR to investigate the contamination of the subsurface by leaked fuel, recording contaminant plumes at depths of up to approximately 4 m (e.g., Srigutomo et al. 2016; Subba Rao e Chandrashekhar 2014; Jiang et al. 2012).

Fig. 6
figure 6

Radargrams obtained from profile L2 at gas station P06 in December 2017

Fig. 7
figure 7

Radargrams obtained from profile L3 at gas station P06 in December 2017

The areas of low reflection identified in the present study can be attributed to the reduced plane wave reflection coefficients from the unsaturated sand filled partially by the vapor phase of the LNAPLs (Daniels et al. 1995). Gas station P02 had been in operation for 33 years at the time of the study, and P06, for 31 years, which implies that the age of the tanks may be a key factor here, given that corrosion and cracks tend to arise after 20 years of use (Blackman Jr 2001; Cole 1994).

Profile L2 (Fig. 6) was used to calculate the propagation velocity (0.081 m s−1), the electrical conductivity (50 S m−1), and the relative dielectric permittivity (13.8 F m−1) of the saturated clayey soil, at depths of below 2–2.4 m in the saturated zone (Table 1). The application of geophysical techniques at survey sites, including the examination of dielectric physical properties and electrical conductivity, has become increasingly important due to the presence of LNAPLs in the subsoil, which may alter its properties (Fuente et al. 2021; Subba Rao and Chandrashekhar 2014; Saucky 2000).

Table 1 Dielectric permittivity (ε), electrical conductivity (σ), and electromagnetic wave velocity (v) measured from the vadose and saturated zones

Profile L3 at station P06 (Fig. 7) was used to determine the physical properties of the organic contaminants in the low reflection zone, including the electromagnetic wave velocity (0.19 m s−1), the electrical conductivity (0.27 S m−1), and the dielectric permittivity (2.4 F m−1) of the dry, clayey soil (Table 1). It is also possible to conclude that the attenuation of the GPR signal near the surface is due to the low permittivity of the hydrocarbon product (gas vapor) relative to water. However, other factors may influence the leakage of hydrocarbons from underground storage tanks, such as accidental spills during the transportation and transfer of these substances, and the inadequate installation or maintenance of the storage tanks.

A number of studies have combined the use of geophysical tools with chemical and geochemical analyses to evaluate the contamination of the subsurface by hydrocarbons (e.g., Eze et al. 2021; Biosca et al. 2020; Abbas et al. 2018; Jiang et al. 2012). Atekwana and Atekwana (2010) recommend this approach to avoid ambiguities in the interpretation of the geophysical data.

In December 2017 (Table 2), toluene and xylene were detected in the samples from gas station P01. In sample A01, toluene was present at a concentration of 0.26 µg L−1 and xylene at 0.46 µg L−1. The same monoaromatic hydrocarbons were identified in sample A02 from the same gas station (P01), at concentrations of 0.13 µg L−1 (toluene) and 0.34 µg L−1 (xylene). All the other samples from this site presented values below the Detection Limit (< DL). These concentrations are much lower than those (MVA) defined by the Brazilian legislation for drinking water (BRASIL 2021). As gas station P01 has been in operation for 35 years, the age of the tanks may, once again, have provoked corrosion and cracks that would give rise to fuel leakage.

Table 2 Results of the analysis of BTEX in the water samples collected during the present study in Bragança, Pará, Brazil, in December 2017 (dry season)

None of the samples collected in May (rainy season) 2018 exceeded the DL for any of the BTEX in the CG–MS analysis. Given this, none of the samples can be considered to be contaminated by these compounds.

Similar values (< DL) were also recorded in most of the samples collected in the dry season month of December 2018 (Table 3), although benzene was recorded in sample A02 (0.558 µg L−1), from gas station P01, and toluene was found in sample A03 (10.390 µg L−1), from gas station P02. All the other samples were below the Detection Limit (< DL). While these values are much higher than those recorded in December 2017, also are still well below the MVAs defined by Brazilian legislation (BRASIL 2021).

Table 3 Results of the analysis of BTEX in the water samples collected during the present study in Bragança, Pará, Brazil, in December 2018 (dry season)

In October 2021 (dry season, 14 samples), the ethylbenzene concentrations of two samples (0.43 µg L−1 and 0.28 µg L−1; Table 4) were detectable. None of the other BTEX compounds were detected (< DL) in any of the samples, and ethylbenzene was not detected in any of the other samples.

Table 4 Results of the analysis of BTEX in the water samples collected during the present study in Bragança, Pará, Brazil, in October, 2021 (dry season)

None of the samples collected in January 2022 (rainy season, 14 samples) were above the Detection Limit (DL) for any of the BTEX compounds. In the rainy-dry transition period, i.e., July 2022 (14 samples), however, toluene was detected in six samples (Table 5), including samples A01 (0.151 µg L−1) from gas station P01, A02 (0.203 µg L−1) from a residence close to gas station P01, A03 (0.228 µg L−1) from gas station P03, A04 (0.342 µg L−1) from gas station P04, A05 (0.221 µg L−1) from gas station P05, and A14 (0.201 µg L−1) from gas station P07. Two samples contained xylene–A04 (0.868 µg L−1) from gas station P04 and A17 (0.421 µg L−1) from a residence near gas station P03.

Table 5 Results of the analysis of BTEX in the water samples collected during the present study in Bragança, Pará, Brazil, in July 2022 (rainy-dry transition)

There was no provision in the present study for the application of either physicochemical or biological measures to remove hydrocarbons, whether pure or dissolved in the underground water or substrate.

Conclusions

The GPR survey data identified low-amplitude reflections in the vadose zone of two gas stations in the Amazonian town of Bragança, in northern Brazil. Attenuation was recorded near the subsurface due to the low permittivity in the radargrams, which may have been caused by the reduced permittivity of the hydrocarbon products (gas vapor) in comparison with water.

The CG–MS also confirmed contamination by B, T, E, and X in water samples obtained from boreholes located within the study area. Toluene and xylene were the BTEX compounds found most frequently in the samples. The long-term data indicate that this contamination has persisted in both environments over a number of years.

Overall, some 19.7% of the 66 samples collected from the study area were contaminated, although none of these sample exceeded the threshold established by Brazilian legislation for drinking water. Even so, the presence of these contaminants in the water samples is preoccupying, and highlights the need for further monitoring, given the potential risks of the exposure to BTEXs of both gas station employees and local residents.

The combined GC–MS approach was successful and proved to be a potentially valuable tool for the diagnosis and monitoring of the presence of BTEXs in water samples obtained from boreholes near gas stations. Overall, the integration of the GPR geophysical tool with the chemical analyses (CG and MS) of groundwater samples proved effective for the complementary assessment of the contamination in the subsurface of the gas stations surveyed in the study area.