Study area and sampling

The study was carried out in a disposal site for oil well drilling and prospecting wastes. The area is located in Santa Maria do Oeste, Paraná State - Brazil (24º 56' 20" S, 51º 51' 46" W). The local soil was previously characterized according to ^{Donagemma et al. (2011}) and classified by ^{Santos et al. (2013}) as typical dystrophic Red Latosol, moderate A, and wavy relief. Table 1 describes the soil physical and chemical features.

Waste from oil well drilling and prospecting was deposited in the studied soil over 20 years ago. It is composed of fragmented rock mixed with well drilling fluid (removed during well drilling). In the area, there is also deposition of sediments from upstream soil erosion by surface runoff.

A geostatistical model was used to measure spatial variability of barium (Ba) and lead (Pb) in the area. To this end, a 15-m soil sampling grid was created, with 76 sampling points. Each point was georeferenced at each grid vertex using a topographic global positioning system (GPS) with post-processing differential correction (DGPS - Trimble Pro XT model). The UTM coordinate system was used (fuse 22 and horizontal datum SAD 69). To further improve the accuracy, a differential correction was used, with sub-meter accuracy (average location error equal to 0.6 m). At each point, soil samples were collected from the following depth layers: 0.0-0.3, 0.3-0.6, 0.6-0.9, and 0.9-1.2 m. Each sampling trench was dug using a backhoe loader, and soil samples were collected by hand.

Chemical analyses

All samples were analyzed for pseudo-total levels of Ba and Pb by EPA 3051 (^{USEPA, 1994}). Thirty of the samples with the highest pseudo-total concentrations were selected for analysis by NBR 10004 (^{ABNT, 2004a}). Leaching and solubilization tests were carried out following NBR 10005 (^{ABNT, 2004b}) and NBR 10006 (^{ABNT, 2004c}) standards, respectively.

For leaching assay, a 100-g sample was sieved through a 0.0095-m mesh sieve and added with 11.4-mL extractant solution of glacial acetic acid in 2-L deionized water (pH = 4.93). This solution was stirred at 30 rpm and 25 ° C for 18 h, using a flask of inert material (borosilicate glass). Then it was filtered through 0.6- to 0.8-µm fiberglass filter to obtain the leachate for analysis. Additionally, a 250-g sample was sieved through a 0.0095-m mesh sieve and added with deionized water in a 1,500-mL flask, then stirred at low speed for 5 min. The flask was covered by a PVC film and left to rest for 7 days at 25 ºC. After, the solution was filtered through a 0.45-µm membrane, and solubilized extract was analyzed.

Concentrations of Ba and Pb in soil extracts were determined by plasma emission spectrometry (ICP-OES) (Perkin Elmer, OPTIMA 3000 model). The method detection limit (DL) was calculated as the blank mean plus three times the standard deviation of the blank of all analyses (10 replications). The DLs for Ba and Pb were 0.036 and 0.25 mg kg^-1, respectively. The QLs (quantification limits) for Ba and Pb were 0.36 and 0.5 mg kg^-1, respectively. The methods were validated by certified reference material: NIST SRM 2709a (San Joaquin Soil) with Ba content of 979±28 mg kg^-1 (95 % of recovery) and Pb content of 17.3 mg kg^-1 (92% recovery). The results of certified reference materials were within the confidence limits stated by NIST.

Barium and lead spatial variability

Ordinary kriging geostatistical interpolation was used to create Ba and Pb concentration isoline maps. After spatial dependence was identified and measured, maps were produced with precision and without trends. Isovalue maps were then generated using the following procedures: 1) descriptive statistical analysis, 2) spatial dependence determination (variography), 3) semi-variogram modeling and validation, and 5) interpolation. All the steps were performed by GEOESTAT software (^{VIEIRA et al., 1983}).

The models fitted to the semivariograms were validated by nugget effect (C₀), range (a), sill (C₀+C₁), spatial dependence index, C0C0+C1 and R² values, which are often used to evaluate interpolation performance.

Descriptive analysis of soil levels of barium and lead

The overall sample mean for Ba level was 732.15 mg kg^-1 (Table 2), the maximum was 10,557.00 mg kg^-1 and the minimum was below 0.036 mg kg ^-1 (undetected). As standard values have not been established yet for Paraná State, we used the reference values in resolution n° 420 of the Brazilian Council for the Environment - CONAMA (^{BRASIL, 2009}). At all depth layers, Ba contents were higher than the limit in the most restrictive scenario (Agricultural/AP-Max), which is 300 mg kg^-1. The Ba contents were about 30 times higher than the limit for agricultural areas. Barium concentrations were even higher than the limit for industrial areas, which is 750 mg kg^-1. According to ^{Magalhães et al. (2014b}) and ^{Freitas et al. (2015}), Ba concentration in areas contaminated by oil well drilling residues is increased due to waste disposal near the wells. It is worth mentioning that Ba has no direct relationship with the genesis of local rocks or respective mineralization, as seen in local geochemical surveys (^{LICHT; BITTENCOURT, 2013}). In the soils of northern Paraná State, natural levels of Ba fluctuate within a range between 38.5 to 63 mg kg^-1 (^{PAVELHAO, 2015}). In the coastal plain of the state, the levels are even lower (< 0.17 mg kg^-1) (^{BUSCHLE, 2013}). That said, the abnormal levels found in this study may result from oil well drilling waste incorporation.

The overall sample mean for the Pb level was 22.58 mg kg^-1, the maximum was 602.49 mg kg ^-1, and the minimum was below 0.25 mg kg ^-1 (undetected). The maximum values at the depth layers of 0.0-0.3, 0.3-0.6, 0.6-0.9 and 0.9-1.2 m were 602.49, 314.41, 218.59, and 165.65 mg kg ^-1, respectively, while the minimum value was below 0.25 mg kg ^-1 (undetected) at all four depth layers (Table 2). In one of the sampling points, the Pb contents were also higher than the limit in the most restrictive scenario (Agricultural/AP-Max), which is 180 mg kg^-1. Moreover, in the three upper layers, the maximum values reached were higher than the reference values established by CONAMA (^{BRASIL, 2009}). In the 0.0-0.3 m depth interval, the maximum content was about three times higher than the references. Therefore, the area is not only contaminated by Ba but also by Pb up to a depth of 0.9 m. When considering maximum allowable Pb contents for farmland of other countries, the values found in the study area are also above the standards of the European Union (50-300 mg kg^-1) and the United States (150 mg kg^-1) (^{MCGRATH et al., 1994}).

Spatial variability in depth of barium and lead

For Ba, the Gaussian model showed determination coefficient values (R²) equal to or above 0.80, at the four depth layers (Table 3). However, for Pb, the 0.0-0.3 and 0.6-0.9 m depth layers showed a more erratic behavior when using the Gaussian model, with R² values much lower than those observed for Ba at these depths. At the 0.3-0.6 m and 0.9-1.2 m depth layers, the spherical model presented better fit, with higher R² values.

Comparing each parameter for Ba in the different depth layers, a total agreement was observed, that is, according to depth and metal type the parameters had a good fit. This is because Co (nugget effect) varied from 0.10 to 0.19 (Table 3), R² was greater than 0.8, and range above 50 m. ^{Amaral Sobrinho et al. (2018)} found similar Gaussian model fit results to Ba distribution data, which were also collected in waste disposal areas for oil well drilling and prospecting. Accordingly, a 15 x 15-m grid was suitable for the range values found. However, for Pb, the nugget effect (Co) and R² values with the best fit were found only at the 0.9-1.2 m depth layer; yet, the range values were suitable with a good autocorrelation for all depth layers. These values of range highlight that the Pb had a spatial variability pattern other than random (^{GUEDES et al., 2012}). The mean R² values at the depths 0-0.3, 0.3-0.6, and 0.6-0.9 m do not necessarily denote that data interpolation was lower at these layers. This is because R² refers only to model fit for all semivariance points. What matters, however, for geostatistical interpolation is the model fitting to the semivariance points within the range limit.

Spatial dependence of Ba and Pb contents was identified by means of geostatistical treatment; hence, geostatistical interpolators could be used to generate spatial variability maps for the four soil layers (0.0-0.3, 0.3-0.6, 0.6-0.9, and 0.9-1.2 m) (Figure 1). The green areas correspond to classes 1 and 2 (^{BRASIL, 2009}), which have concentrations up to the limit value for prevention, for Ba (0 and 150 mg kg^-1) and for Pb (between 0 and 72 mg kg^-1). The blue area comprises class 3, in which concentrations were above the prevention limits and below investigation limits, for Ba (between 150-300 mg kg^-1) and for Pb (73-180 mg kg^-1). Shades from pink to black were classified as class 4. In the case of Ba, other hues were added due to a large range of values. Also for Ba, the sum of the percentages of areas classified as class 3 and class 4 reaches more than 50% of the total at all depth layers. For Pb, classes 1 and 2 represent over 90% of the area, and most of the contamination is within the first 0.6 m depth and around the waste disposal dike. This indicates that more than 50% of the area has Ba concentrations that disturb soil main functions and pose direct or indirect risks to human health (^{BRASIL, 2009}).

Figure 1 Spatial variability of barium (A, B, C, and D) and lead (E, F, G, and H) at the depth layers of 0-0.3, 0.3-0.6, 0.6-0.9 and 0.9-1.2 m, respectively. 

Leaching and solubilization tests

Thirty samples with higher pseudo-total concentrations of Ba and Pb were selected. All these samples had concentrations of both elements below the limit for leaching test according to NBR-10005 (^{ABNT, 2004b}), 70 mg L^-1 for Ba and 1.00 mg L^-1 for Pb. Likewise, the contents of all samples were below the limit for solubilization test according to NBR-10006 (^{ABNT, 2004c}), 0.7 mg L^-1 for Ba and 0.01 mg L^-1 for Pb (Table 4).

The low solubility of barite-derived Ba has already been reported in studies on soils with varying organic matter, iron, calcium, and sulfur contents (^{CAPPUYNS, 2017}) and soil from A horizon of a Red-Yellow Latosol (^{MAGALHÃES. et al., 2011}; ^{LIMA et al., 2012}; ^{MAGALHÃES et al., 2014a}). This low solubility is due to the high pH and redox potential (Eh), resulting in its slow release to the soil. Barium tends to remain in poorly soluble forms (^{NOGUEIROL; ALLEONI, 2013}) with sulfate ion, which makes it more associated with sediment contamination than with water (^{DAVIDSON et al., 2005}). Barium soluble fraction only increases under extreme reduction conditions (Eh<-250 mV), hence increasing its mobility and bioavailability (^{LIMA et al., 2012}; ^{MAGALHÃES et al., 2014a}; ^{MAGALHÃES et al., 2014b}; ^{SAMPAIO JUNIOR et al., 2015}).

Pseudo-total concentrations of Ba and Pb to reach the maximum limit for leaching and solubilization assays were estimated using data from Table 4. This way, regression curves could be plotted (Figure 2), finding R² values equal to 0.85, 0.95, and 0.89 for Ba solute, Ba leachate, and Pb leachate, respectively.

Figure 2 Regression analysis for contents of barium (leached and solubilized) (A) and lead (leached) (B) as a function of total contents in the soil. 

The regression curve equation shows that Ba pseudo-total concentrations above 61,000 mg kg^-1 are able to reach the upper limit for leaching (70 mg L^-1). This is 6 times the value of the most concentrated sample (10,557.00 mg kg^-1), for both leaching and solubilization analysis. On the other hand, for Pb, the concentrations should be above 1,071.33 mg kg^-1 to reach the limit for leaching (1 mg L^-1). This is about 1.8 times the value of the most concentrated sample (602.50 mg kg^-1), according to the maximum limits defined in NBR-10004 (^{ABNT, 2004a}).