Landscape-scale controls on trace metals partitioning and mobility in tropical soils affected by legacy lead smelting

TL;DR

This study examines how lead and other metals from a former smelter in Brazil are distributed and behave in different soil types.

Contribution

The study reveals landscape-specific metal mobility patterns and stabilization mechanisms in contaminated tropical soils.

Findings

Pb, Zn, and Cd concentrations decrease with distance from the smelter and vary by soil type.

Pb stabilization differs between hillslope (carbonates) and floodplain (Fe oxyhydroxides) soils.

Floodplain soils act as temporary metal storage rather than permanent sinks.

Abstract

The long-term legacy of a decommissioned lead smelter in Santo Amaro, Bahia (Brazil), has produced one of the most metal-contaminated urban areas worldwide. This study investigates the spatial distribution, geochemical partitioning, and potential mobility of trace metals across contrasting landscape units, namely hillslope and floodplain soils. A total of 120 soil samples were analyzed using portable X-ray fluorescence (XRF), scanning electron microscopy coupled with energy-dispersive spectroscopy (SEM–EDS), and the BCR sequential extraction procedure. XRF results revealed extremely high concentrations of Pb (up to 24,962 mg kg−1), Zn (up to 8572 mg kg−1), and Cd (up to 454 mg kg−1), with strong spatial heterogeneity related to distance from the former smelter and landscape position. Cadmium and Pb were predominantly associated with labile and reducible fractions, indicating high chemical lability (mobility factor) and sensitivity to environmental conditions. SEM–EDS observations revealed contrasting stabilization mechanisms of Pb, associated with carbonate phases in hillslope soils and co-precipitated with Fe oxyhydroxides in floodplain soils. In contrast, Cr, Ni, and Cu were mainly associated with the residual fraction, indicating dominant lithogenic control and limited mobility. Geostatistical modeling still showed strong spatial dependence for Pb, Zn, and Cd with decreasing concentrations away from the smelter. The results demonstrate that floodplain soils do not act as permanent sinks but as transitional environments, where metals mobilized from hillslopes are temporarily retained and subsequently transferred toward downstream compartments.

The online version contains supplementary material available at 10.1007/s10653-026-02994-3.

Figures

Localization of the study area and sampling points, and topography map. See details and coordinates in Table S1

Distribution map of metallic elements in the area most impacted by a former lead smelter in Santo Amaro (soils with [Pb] > 100 mg kg^−1^)

Relative distribution of metals among BCR fractions in hillslope (HS) and floodplain (FP) soils, Santo Amaro. F1 = exchangeable fraction; F2 = oxidizable fraction; F3 = reducible fraction; F4 = residual fraction

Mobility factor for Cd, Cr, Cu, Ni, Pb, and Zn in in the area most impacted by a former lead smelter in Santo Amaro (soils with [Pb] > 100 mg kg^−1^), in hillslope (HS) and floodplain (FP) soils

Principal Component Analysis between labile fractions F1 and F2, and physical and chemical soil attributes (COT—total organic carbon). (1) Hillslope soils; (2) Floodplain soils

Scanning electron microscopy and EDS microanalyses: EDS layered image of sample 24, hillslope soil (A) and of sample 44, floodplain soil (B); Pb mapping of sample 24, hillslope soil (C) and of sample 44, floodplain soil (D). EDS layered image of individual mineral, sample 44, floodplain soil, with Fe mapping (E) and Pb mapping (F). Biplots graphical of Ca, Fe, and Pb spot geochemical analysis in both samples (G e H)—see spectra in Figures S4, S5 and S6 (Supplementary material)

Full text

Introduction

One of the primary concerns following the closure of mining and metallurgical operations is the long-term management of the waste they generate (Tayebi-Khorami et al., 2019). Although some of this material can be repurposed or remined, the majority poses serious environmental risks due to accumulating inadequate storage (Ceniceros-Gómez et al., 2018; Ettler et al., 2005; Falagán et al., 2017; Helser & Cappuyns, 2021; Karaca et al., 2017). In the state of Bahia, northeastern Brazil, widespread soil contamination has been documented in the municipality of Santo Amaro (Carvalho et al., 1989; Carvalho et al., 1984; de Andrade Lima & Bernardez, 2011; de Andrade & Moraes, 2013; dos Santos & dos Anjos, 2022; Magna et al., 2013; Paoliello & De Capitani, 2007; Silvany-Neto et al., 1989). This pollution originated from the operations of Plumbum Mineração e Metalúrgica Ltda., a multinational lead extraction company that operated from the 1960s until 1993 and produced approximately 490,000 tons of slag containing Pb, Cd, Zn, Cu, and As. Recent studies highlight the persistence of contamination, largely due to the stability and low mobility of these metals under natural conditions. Thirty years after the smelter’s closure, residues remain present in soils (de Andrade & Moraes 2013; dos Santos & dos Anjos, 2022; dos Santos et al., 2017a, 2017b), riverbeds and marine sediments (Bomfim et al., 2018; da Silva et al., 2017; Gloaguen et al., 2021; Motta et al., 2018; Ramos et al., 2023), and even in human and animal tissues (de Souza Guerra et al., 2015; Magna et al., 2013; Niemeyer et al., 2015; Paoliello & De Capitani, 2007). Concerns persist regarding the effectiveness of implemented remediation efforts (Anjos & Sánchez, 2023; Batista et al., 2017), as well as the adequacy of public policy responses (dos Santos & dos Anjos, 2022).

Despite numerous investigations, few studies have addressed the distribution of toxic metals in this environment from a geological, geomorphological, and pedological perspective. The study area is located within a sedimentary basin formed by the infilling of a Cretaceous rift, and is dominated by shallow marine green shales rich in organic matter, interspersed with carbonate lenses. The Pb background concentration in these shales and the derived Vertisols is 22.3 g kg^−1^, but only 4.9 g kg^−1^ in Arenosols developed from sandstones (Gloaguen & Passe, 2017). The most contaminated zone around the smelter—where Pb concentrations range from 100 to over 10,000 mg kg^−1^—is primarily composed of very clayey Vertisols. Extensive areas of Vertisols are relatively rare in Brazil, making Santo Amaro a unique case in Latin America. Some examples of metal-contaminated Vertisols have been reported elsewhere, including the Ranipet industrial zone in Tamil Nadu, India (Veluprabakaran & Kavitha, 2023) the Challawa industrial area in Kano, Nigeria (Akan et al., 2009), and the Sukinda chromite mining region in Odisha, India, where Vertisols are affected by Cr, Ni, and Pb pollution (Bhup et al., 2011). Comparable findings have also been documented in the Hunter Valley region of New South Wales, Australia (Lamb et al., 2009), and in the Zimapán mining district in Hidalgo, Mexico (Armienta et al., 2016), where legacy mining and smelting activities have contributed to persistent soil contamination.

In such areas, the remobilization of metals is a critical concern. The combination of a humid tropical climate, sedimentary lithologies, and horst–graben tectonics, as observed in the Recôncavo Basin in Santo Amaro, creates a rugged terrain characterized by hills, valleys, and incised canyons. This geomorphological setting exacerbates the dispersion of metals adsorbed onto clay particles, particularly during tropical rainy seasons. Therefore, analyzing the distribution of metals across distinct landscape units is essential to understand their transport and depositional dynamics. As reported by Anand et al. (2003), the proximity to the smelter does not necessarily correlate with soil metal concentrations or disruption of microbial communities; instead, factors such as phosphorus content or organic matter may play more significant roles. Landscape features influence the biogeochemical cycling of trace metals: for instance, riparian and wetland zones in valley bottoms can function both as sources and sinks for metals due to their high organic matter content and redox variability (Jaja et al., 2022). Moreover, vegetation structures such as copses and hedgerows have been shown to significantly affect metal accumulation in soils, likely due to edge effects (e.g., advection and inflow)—with higher concentrations in woodlands compared to grasslands, moorlands, or open fields (Fritsch et al., 2010).

X-ray fluorescence (XRF) spectrometry has become a widely adopted tool for rapid field assessment of contaminated areas (Gloaguen et al., 2024; Lemière, 2018; Potts & West, 2008; Qu et al., 2022; Schneider et al., 2016; Silva et al., 2018; US Environmental Protection Agency 2007). This technique allows for non-destructive elemental analysis with reduced cost and minimal waste generation. However, total metal concentrations alone are insufficient to fully evaluate environmental risk. Understanding the potential mobility and bioavailability of metals is essential for interpreting their environmental behavior beyond traditional static pollution indices (Fernández-Ondoño et al., 2017; Helser & Cappuyns, 2021; Król et al., 2020; Vijver et al., 2004). Assessing the geochemical forms and distributions of elements across soil fractions is fundamental to predicting mobility, bioavailability, and risk, especially under variable environmental conditions and as influenced by soil physicochemical properties (Cipullo et al., 2018; Fazle Bari et al., 2020; Gabarrón et al., 2019; Vollprecht et al., 2020). Studies of metal availability in soils and sediments impacted by smelter slags typically apply sequential extraction protocols (Kumkrong et al., 2021; Mbodji et al., 2022; Swęd et al. 2022). The BCR method has been extensively applied to assess metal bioavailability in a range of environmental matrices, including mining residues, sediments, soils, sewage sludge, and related compartments (Arain et al., 2008; Cotta et al., 2023; ELTurk et al., 2018; Li et al., 2018; Roig et al., 2013; Swęd et al. 2022; Wang et al., 2018; Yin et al., 2014). These techniques promote a more accurate panorama of geochemical metal dynamics: metals associated with carbonates, Fe and Mn oxides, or organic matter exhibit contrasting sensitivities to pH, redox fluctuations, and hydrological dynamics, particularly in tropical and subtropical environments. For example, carbonate-bound metals are sensitive to pH variations, whereas metals associated with Fe and Mn oxides are particularly mobile in reductive environment (Rodgers et al., 2015). Mineralogy also plays a considerable role, with fine-textured clayey soils richer in metal oxides due to higher sorption capacity, favoring metal retention (dos Santos et al., 2017a, 2017b; Gloaguen & Passe, 2017; Gomes et al., 2016). Organic matter can further influence metal behavior by forming stable complexes, with interactions modulating by soil pH, ionic strength, and microbial activity (Li et al., 2022; Luko-Sulato et al., 2024; Nierop et al., 2002). Furthermore, geomorphology and hydrological processes often result in spatial decoupling between contamination sources and observed metal concentrations, promoting progressive redistribution and homogenization of contamination across watersheds (Anand et al., 2003; Derakhshan-Babaei et al., 2022; Jaja et al., 2022). Consequently, an integrated approach combining geochemical partitioning, spatial analysis, and landscape context is essential for understanding the long-term mobility and persistence of metals in soils impacted by mining and smelting activities.

The primary aim of this study is to investigate the spatial distribution and potential mobility of metals in soils from two contrasting landscape units, hillslopes and floodplains, affected by legacy lead smelting.

Material and methods

Study area

The study area is located in the municipality of Santo Amaro, Bahia, Brazil, which covers an area of 492 km^2^ and has a population of approximately 57,800 inhabitants. It is situated near the estuary of the Subaé River, which flows into Todos os Santos Bay (Fig. 1). Spatial delineation of the study site was based on a previously developed geochemical soil map (Fig. S1). The selected area spans 15.22 km^2^ and corresponds to the most contaminated portion of the municipality, where lead concentrations exceed 100 mg kg^−1^ (Nero, 2020). According to the Köppen–Geiger climate classification, the region has a humid or superhumid tropical climate (Af), with no distinct dry season. Annual precipitation averages 1713 mm, peaking in May and reaching its lowest levels in September. The average annual temperature is 24.7 °C. Elevation ranges from 290 m inland to sea level along the southeastern margin, where coastal plains and extensive mangrove ecosystems are found bordering Todos os Santos Bay. Geologically, the region is composed of Cretaceous sedimentary rocks, primarily green shales of the Candeias Formation (Santo Amaro Group), with some sandstone outcrops of the Sergi Formation (Upper Jurassic), and marine and continental coastal deposits from the Holocene (Dalton de Souza et al., 2003).Fig. 1. Localization of the study area and sampling points, and topography map. See details and coordinates in Table S1

The soils are predominantly Vertisols derived from Cretaceous shales (Aliança Formation). Field classification identified Hydromorphic and Haplic Vertisols, Haplic and Thionic Gleysols, Haplic Cambisols, and Fluvic Neosols (Fig. 1). Based on geomorphological position and soil type, the area was divided into two landscape units:

Hillslopes: These areas are dominated by Haplic Vertisols (VX), covering 9.84 km^2^ (64.7% of the study area). These soils pose management challenges due to their shrink–swell behavior: they become hard and blocky when dry, and extremely plastic and sticky when wet. They occur on undulating terrain (3–8% slope) at elevations of 60–90 m (summits and shoulders), and at lower altitudes (30–60 m) on steeper slopes (> 45%). These sites exhibit low topographic wetness index values, poor drainage, and higher surface runoff, making them susceptible to erosion.

Floodplains: These areas occupy valley bottoms with slopes < 3% and cover 35.3% of the study area. Sedimentation and depositional patterns are influenced by drainage channels, and soil texture varies significantly over short distances (Fig. 1). The dominant soils are Hydromorphic Vertisols (99% of the area), with minor occurrences of Thionic Gleysols, Haplic Cambisols, and Fluvic Neosols. These sites show higher wetness index values, intense water saturation, and fluctuating redox conditions. Under such conditions, we hypothesize a substantial increase in pollutant transfer risk.

Sampling was conducted at summit or shoulder positions on hillslopes, and at depositional environments in floodplains. From a mineralogical perspective, although quantitative determinations of free iron oxides (DCB- or oxalate-extractable Fe) were not available, the soils are expected to contain significant amounts of pedogenic Fe oxides and hydroxides, such as goethite and hematite, inherited from the weathering of Fe-rich shales. In floodplain soils, periodic water saturation and redox fluctuations favor the dissolution, redistribution, and reprecipitation of amorphous and poorly crystalline Fe phases, whereas in hillslope soils, Fe oxides tend to be more stable and better crystallized. These contrasting mineralogical environments are critical for understanding differences in metal retention and mobility between landscape units.

Soil sampling

Soil sampling was conducted during the dry season, in 2022 and 2023, with temperatures from 24 to 28 °C. During the rainy season the floodplains in bottom valley are inundated for months (swamp and river—anoxic conditions). Sampling sites were distributed across hillslope and floodplain environments, both upstream and downstream of the former lead smelter. Predominant wind directions in the region are from southeast to northwest, which historically favored atmospheric dispersion along this axis, as indicated by the spatial pattern of soils contaminated with Pb concentrations > 100 ppm (Fig. 1). A total of 120 soil samples were collected from 60 locations, representing both topsoil (0–5 cm) and subsurface (5–20 cm) layers (Fig. 1 and Table S1). Sampling was conducted using a stainless-steel hand auger, following the guidelines of the Brazilian Society of Soil Science (Teixeira et al., 2017). Sample locations were selected using a stratified random sampling strategy based on Conditioned Latin Hypercube Sampling (cLHS), designed to maximize environmental variability in the covariates used for predictive modeling (Minasny & McBratney, 2006). This approach has been widely applied in recent years (Adamchuk et al., 2011; Brungard & Boettinger, 2010; de Menezes et al., 2013; Levi & Rasmussen, 2014; Van Wijnen et al., 2012). The environmental covariates considered included elevation, slope, topographic wetness index (derived from a 30 × 30 m digital elevation model), lead concentrations above 100 mg kg^−1^, access roads (OpenStreetMap), and soil classes (classified in the field according to the second categorical level following Santos et al., 2018). The cLHS library in RStudio was employed for the analysis (R Development Core Team, 2022). Some of the variables are presented in Fig. S1.

Physical and chemical analysis

Analytical procedures for soil classification, total organic carbon (TOC), pH, and particle size distribution were conducted according to the methods described in the EMBRAPA Brazilian Soil Analysis Methods Manual (Teixeira et al., 2017), at the Trace Metal Laboratory, Universidade Federal do Recôncavo da Bahia (UFRB). Soil samples were air-dried, gently disaggregated, and entirely passed through a 2 mm sieve. The particle size distribution was analyzed by pipette method. Chemical dispersion was performed using sodium hexametaphosphate, as recommended in EMBRAPA Manual for calcareous soils and soils dominated by high-activity 2:1 clay minerals, such as Vertisols, Alfisols, Aridisols and Molisols. Sand (> 53 µm) was measured by sieving and drying, and the proportions of clay (< 2 µm) and silt (2–53 µm) were calculated based on sedimentation time and Stokes’ law. Soil pH was measured in a 1:2.5 soil-to-water suspension, and TOC was quantified through wet oxidation using 0.2 M potassium dichromate in sulfuric acid, followed by titration with 0.2 M ammonium ferrous sulfate.

X-ray fluorescence analyses were performed using a Bruker S1 Titan 600 portable spectrometer (Billerica, MA, USA). For this purpose, 2 g of quartered air-dried soil from each sample were homogenized and ground in an agate mortar and passed through a 250 µm sieve, following USEPA Method 6200 for X-ray fluorescence analysis. (US Environmental Protection Agency 2007). Prior to XRF analysis, all soil samples were standardized to ensure analytical comparability. Samples were analyzed under identical measurement conditions. A constant sample mass was used for each analysis, and measurements were performed on loose, homogenized material to minimize matrix effects related to particle-size variability and surface roughness. Calibration was performed using certified reference materials (San Joaquim, Soil Montana 1, CS-M2, and IPT-32), yielding analytical precision below 5.1% and calibration coefficients (R^2^) greater than 0.99.

Although portable XRF provides rapid and non-destructive elemental analysis, its detection limits for Cd are higher than for other trace metals such as Pb and Zn. As a result, Cd concentrations in some samples, particularly those collected at greater distances from the former smelter or in less impacted areas, fell below the instrumental detection limit. These values were treated as non-detects and excluded from geostatistical modeling where appropriate.

Additionally, two representative samples were analyzed via scanning electron microscopy (SEM) using a Zeiss EVO LS15 system under high vacuum conditions at the UFRB SEM Laboratory. SEM–EDS analyses were conducted on two representative samples selected to illustrate dominant metal stabilization mechanisms in contrasting landscape units (hillslope and floodplain soils). These analyses were intended to provide qualitative micro-scale support to the interpretations derived from bulk geochemical data. To enhance conductivity and minimize charging effects, a thin carbon coating was applied using a Quorum Q150T ES carbon coater. Imaging was carried out at 15–25 kV and 500–850 pA. Both secondary electron (SE1) and backscattered electron (BSD) detectors were used at magnifications ranging from 200 × to 10,000 × . Elemental mapping and chemical microanalyses were performed in seven selected areas using an Oxford Instruments X-Max EDS detector.

Sequential extraction was not performed on all samples, but on a representative subset corresponding to approximately 30% of the dataset. Samples were selected to encompass the main soil classes identified in the study area, both landscape units (hillslope and floodplain), and a wide range of Pb concentrations. This selection strategy ensured that the sequential extraction results captured the dominant pedological, geomorphological, and geochemical variability of the contaminated area. The samples were freeze-dried (LIOTOP model L101) and sieved to 250 µm prior to extraction. The four-step sequential extraction followed the European Community Bureau of Reference (BCR) protocol. All extractions were carried out using approximately 1.0 g of soil in 50 mL polyethylene centrifuge tubes. The exchangeable and acid-soluble fraction (F1) was extracted by adding 40 mL of 0.11 mol L^−1^ acetic acid. The suspensions were agitated on an orbital shaker at 60 rpm for 16 h at room temperature, followed by centrifugation at 3000 rpm for 20 min. The supernatants were filtered and stored in polyethylene bottles for subsequent analysis. The reducible fraction (F2), targeting metals associated with carbonates and Fe and Mn oxides, was extracted by adding 40 mL of 0.5 mol L^−1^ hydroxylamine hydrochloride (NH_2_OH·HCl), adjusted to pH 1.5 with 1:1 HNO_3_, to the solid residue from F1. The samples were then agitated, centrifuged, and filtered following the same procedures applied for F1. The oxidizable fraction (F3), representing metals bound to sulfides and organic matter, was extracted by adding 10 mL of hydrogen peroxide (H_2_O_2_, 8.8 mol L^−1^) to the residue from F2. The mixture was allowed to react at room temperature for 1 h with occasional agitation and subsequently heated in a water bath at 85 °C until the volume was reduced to approximately 3 mL. Additional 10 mL H_2_O_2_ was then added until near-complete evaporation of the reagents. After cooling, 50 mL of 1.0 mol L^−1^ ammonium acetate (NH_4_OAc), adjusted to pH 2 with 1:1 HNO₃, was added, followed by agitation, centrifugation, and filtration. The residual fraction (F4), corresponding to metals bound to primary and secondary mineral lattices, was determined by digesting the solid residue from F3 after drying at 65 °C following the USEPA Method 3050B. The solid residues obtained after each fraction (except F4) were washed with 20 mL of deionized water, agitated for 15 min, centrifuged, and the supernatants discarded prior to the next extraction step. Metal concentrations in the supernatant were determined by inductively coupled plasma optical emission spectrometry (ICP-OES, Agilent 720 Series, USA) at the Lepetro Laboratory, Federal University of Bahia. Technical specifications of the equipment are presented in Table S2.

Data treatment

Geostatistical analyses of the XRF data were conducted using RStudio. Data normalization involved logarithmic and Box-Cox transformations to reduce the influence of outliers. Semivariograms for each element were developed using ordinary kriging (OK). Variogram models were manually adjusted by visually comparing and modifying parameters such as sill, range, and nugget to maximize the coefficient of determination (R^2^) and minimize the root mean square error (RMSE). RMSE values close to 1 were prioritized through iterative testing. Co represents the nugget effect, which accounts for microscale variability and analytical uncertainty, whereas C1 corresponds to the structured variance associated with spatial autocorrelation. The ratio Co/(Co + C1) was used to assess the degree of spatial dependence of each element, with lower values indicating stronger spatial structure. Among the models tested, exponential, Gaussian, and spherical functions yielded the best RMSE values. The selected theoretical models and their associated error estimates are summarized in Table S3. Resulting geochemical maps were generated using ArcGIS 10.8 (ESRI Inc., USA).

To assess metal availability, the results of the BCR sequential extraction were analyzed using both descriptive and multivariate statistical approaches in RStudio. Due to the non-normal distribution of the data (Shapiro–Wilk test, p < 0.05), non-parametric methods were employed. Statistical analyses were conducted in RStudio. The Mann–Whitney U-test (p > 0.05) was used to identify significant differences between sampling depths and topographic positions. Correlation matrix analysis was applied to examine relationships among soil properties, geochemical fractions, and metal concentrations. Principal component analysis (PCA) was performed using the VARIMAX rotation method. Its applicability was confirmed by Bartlett’s test of sphericity and the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy, with KMO values greater than 0.50 indicating suitability for factor analysis.

The results of the sequential extraction were further interpreted using the Mobility Factor (MF), which considers the sum of the exchangeable and reducible fractions (F1 + F2) as the most mobile and environmentally available forms of metals (Kabala & Singh, 2001; Osakwe, 2010). Metal mobility generally follows the order F1 > F2 > F3 > F4. The MF was calculated using Eq. 1 (Iwegbue et al., 2018; Vaněk et al., 2005):

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$MF=\frac{\mathrm{F}1+\mathrm{F}2}{\mathrm{F}1+\mathrm{F}2+\mathrm{F}3+\mathrm{F}4}*100$$\end{document}

where Fn represents the concentration of each fraction obtained from the BCR sequential extraction. Higher MF values indicate greater metal mobility and, consequently, higher potential bioavailability (Gope et al., 2017).

Results and Discussion

Soil particles size

Particle-size distribution reflects the strong dominance of fine fractions typical of Vertisols. Clay content exceeded 35% in all samples, with mean values of 447 ± 120 g kg^−1^ in hillslope soils and 481 ± 98 g kg^−1^ in floodplain soils. Silt contents were also high (348 ± 109 g kg⁻^1^ and 338 ± 100 g kg^−1^, respectively), whereas sand contents were comparatively lower and more variable. Coefficients of variation for clay and silt fractions indicate moderate spatial heterogeneity, particularly in floodplain environments where depositional processes lead to short-range textural variability. The predominance of clay and silt fractions plays a key role in controlling trace metal retention through adsorption onto mineral surfaces and association with Fe oxides and organic matter.

Total concentration and geochemical maps

As expected, soils showed high levels of Pb, the primary pollutant, and Cd, an associated contaminant, throughout the study area. Maximum concentrations reached 24,962 mg kg⁻^1^ for Pb and 454 mg kg⁻^1^ for Cd (Table 1). The coefficients of variation (CV) followed the descending order: Zn > Pb > Ni > Cd > Cr > Cu, suggesting that these metals are strongly influenced by external (especially anthropogenic) factors, consistent with findings in polluted environments. The high variability of Zn, Pb, and Cd concentrations (Table 1) reflects the intense impact of smelting activities. The asymmetric data distribution and presence of outliers further highlight the strong impact of historical smelting activities on soil metal concentrations near the former lead smelter in Santo Amaro.Table 1. Mean total concentrations of Cd, Cr, Cu, Ni, Pb and Zn (coefficients of variation into parentheses in %) in soils within the area most impacted by smelter activities (Pb > 100 mg kg^−1^), in the city of Santo Amaro, Bahia, Brazil0–5 cm (n = 60)5–20 cm (n = 60)Regional geochemical backgroundGeometric meanMaxGeometric meanMaxCd10.7 (223)454.37.7 (334)1376.4 < 0.2Cr86.6 (27)149.079.5 (30)138.0108.8Cu116.4 (30)266.4114.7 (39)407.1534.2Ni9.5 (107)80.014.4 (92)129.044.9Pb471.9 (271)24,961.7441.2 (275)24,325.111.3Zn211.7 (278)8.572.4203.6 (289)8.581.730.1***Values significantly different between soil layer, within the same line (Mann–Whitney test, p < 0.05)**For Vertisol. Gloaguen and Passe (2017)

Cu, Cr and Ni displayed lower coefficients of variation, suggesting less influence from smelter pollution, and more association with parent material, in particular for Cr and Ni. Cu was identified as a minor component in the lead ore, which may explain locally elevated concentrations, despite the overall lithogenic control observed for this element. A statistically significant difference between topsoil and subsurface soil was observed only for Ni, with higher concentrations at depth.

A simple Principal Component Analysis (PCA) was used only with metals as an integrative framework to confirm controlling tendencies and to distinguish between anthropogenic and lithological influences across the study area (Chen et al., 2022; Kolawole et al., 2023). The first principal component (PC1, Fig. S3), indicating the historical lead smelting activities explains the largest proportion of variance and is strongly loaded by Pb, Zn, and Cd. In contrast, Cr loads primarily on PC2, reflecting lithological control related to parent material composition.

Geostatistical modeling was used to estimate metal concentrations in unsampled locations (Zhen et al., 2019). The spatial dependence of each element was assessed using the Co/(Co + C1) ratio, which reflects the proportion of structured spatial variability. Ratios below 0.25 indicate weak spatial correlation, while values above 0.75 suggest strong spatial dependence. Except for Cd, which often fell below the detection limit and could not be modeled, the spatial correlation coefficients (R^2^) were high for Ni, Pb, and Zn, and moderate for Cr and Cu (Table S3). The geochemical maps (Fig. 2) indicate that all studied metals show elevated concentrations near the former smelter site. Two main spatial patterns are evident: (1) a distribution pattern aligned with the former ore transportation route, particularly the railway that crosses the area from the center-east to the northwest, which is especially prominent for Zn and Pb, and coincides with SE-NW wind directions; and (2) a geological control linked to the Candeias Formation, composed of metal-rich shales from the Santo Amaro Group, which contributes to the elevated natural trace metals background levels (Gloaguen & Passe, 2017).Fig. 2. Distribution map of metallic elements in the area most impacted by a former lead smelter in Santo Amaro (soils with [Pb] > 100 mg kg^−1^)

In summary, Pb, Zn, and Cd concentrations show a general decrease with increasing distance from the former smelter site (COBRAC), although this trend is not strictly linear. Spatial patterns are strongly influenced by prevailing wind directions, which favored the atmospheric dispersion of metal-bearing particulates, as well as by topography and hydrological connectivity, which promote the downslope transport and accumulation of contaminated fine sediments in floodplain environments. In contrast, Ni and Cr display more homogeneous spatial distributions, largely independent of distance from the smelter, indicating predominant lithogenic control related to parent material composition. Copper exhibits an intermediate behavior, reflecting mixed lithological control and secondary anthropogenic inputs associated with smelting activities. Together, these results highlight the combined roles of emission sources, landscape position, and geological background in shaping trace metal distribution across the study area.

Geochemical behavior of metals

Recent studies have demonstrated that a robust interpretation of metal contamination in mining- and smelting-affected environments requires the combined use of spatial analysis, multivariate statistics, and geochemical partitioning approaches. For instance, PCA has been successfully applied to distinguish anthropogenic metal inputs from lithogenic backgrounds, while sequential extraction and mineralogical analyses provide mechanistic insight into metal retention and potential mobility (Mufalo et al., 2024; Soro et al., 2023; Wunn et al., 2025).

In this study, the results of the sequential extraction are presented as absolute values and percentages for each fraction in Table 2 and Fig. 3. In general, soils from floodplain areas exhibited lower concentrations of Zn, Pb, Cd, and Cu than those from hillslopes and summits. However, high standard deviations obscured significant differences. Ni and Cr concentrations did not vary substantially between the two landscape units, reflecting their strong association with parent material. Soils are moderately acidic (Table 2), a condition that may enhance the solubility of trace metals and thus increase their potential mobility (Ali Sungur & Ozcan, 2019; Paula et al., 2022). The low variability of the pH (SD = 0.5 or 0.6) results from the homogeneity distribution of the Vertisols on Cretaceous shales in the studied area. Total organic carbon (TOC) ranged from 0.96 to 8.34% in hill soils and from 1.77 to 5.92% in floodplain soils, with mean values of 3.77% and 3.76%, respectively, indicating no systematic accumulation of organic matter in floodplain environments relative to upland areas. Particle-size distribution reflects the dominance of fine fractions (clay and silt), consistent with the prevalence of Vertisols, which by definition contain more than 35% clay.Table 2BCR concentrations of metals (mg kg^−1^), pH, total organic carbon (TOC in %), and particle size distribution (in g kg^−1^) in hillslope and floodplain soils, Santo Amaro. F1 = exchangeable fraction; F2 = oxidizable fraction; F3 = reducible fraction; F4 = residual fractionHillslope (n = 24)Hillslope (%)Floodplain (%)Floodplain (n = 10)pH6.3 ± 0.5––6.2 ± 0.6TOC3.8 ± 1.7––3.8 ± 1.4Sand205.5 ± 153.2––181.1 ± 178.2Silt347.6 ± 109.3––338.0 ± 100.2Clay446.9 ± 120.4––480.9 ± 98.0CdF110.5 ± 16.532.330.98.0 ± 12.8F210.5 ± 11.742.843.68.8 ± 12.7F31.4 ± 1.112.713.01.5 ± 1.2F41.1 ± 0.312.312.51.1 ± 0.4Total23.719.6CrF11.0 ± 0.04.24.01.0 ± 0.0F21.0 ± 0.04.24.01.0 ± 0.0F36.1 ± 3.024.815.44.0 ± 1.9F419.9 ± 8.466.776.520.2 ± 5.6Total28.126.2CuF13.8 ± 8.14.02.71.3 ± 0.5F24.6 ± 11.54.82.41.2 ± 0.7F321.1 ± 31.625.917.713.3 ± 19.9F439.1 ± 23.665.477.341.4 ± 24.1Total68.657.2NiF11.2 ± 0.510.66.91.0 ± 0.0F22.8 ± 3.017.97.91.1 ± 0.2F31.7 ± 1.112.517.32.0 ± 2.2F49.1 ± 4.859.067.810.8 ± 4.3Total14.814.9PbF1532.2 ± 1452.35.14.7158.2 ± 240.2F21840.6 ± 2969.054.059.91258.7 ± 1694.4F3739.0 ± 1398.023.418.3566.6 ± 959.2F4275.1 ± 411.117.617.1363.6 ± 612.3Total3386.92347.1ZnF1173.8 ± 409.015.911.543.0 ± 63.1F2190.9 ± 352.327.120.364.9 ± 96.0F3117.3 ± 248.714.612.642.3 ± 62.0F4188.3 ± 334.442.455.6115.9 ± 132.2Total670.3100266.0266.0*Statistically different within the same line, between HS and FP (Mann–Whitney test-U, p < 0.05)Fig. 3. Relative distribution of metals among BCR fractions in hillslope (HS) and floodplain (FP) soils, Santo Amaro. F1 = exchangeable fraction; F2 = oxidizable fraction; F3 = reducible fraction; F4 = residual fraction

Lead, the primary pollutant, was predominantly associated with the reducible (F2) and oxidizable (F3) fractions, both of which are considered moderately mobile. Metals bound to amorphous Fe and Mn oxides and hydroxides can be readily leached under reducing conditions, increasing their environmental availability (Islam et al., 2023; Tong et al., 2020). The dominance of Pb in F2 indicates its strong adsorption to Fe–Mn oxides, which play a critical role in controlling early-stage Pb mobility. This behavior has been similarly reported in other urban and industrial soils (Davidson et al., 2006; Famuyiwa et al., 2022; Umoren et al., 2007; Wu et al., 2008). Anthropogenic Pb accumulation in F2 can be destabilized under reducing conditions, leading to its release (Botsou et al., 2016; Iwegbue et al., 2018; Kumari & Paul, 2017; Sungur et al., 2020). The mobility of Fe_2_O_3_ in wetland (gley) horizons, along with the immobilization of Al and Si, is controlled by hydromorphic conditions and organic matter content, affecting Pb dynamics in floodplain soils.

This geochemical behavior is consistent with recent studies emphasizing that Pb mobility and persistence in contaminated environments are strongly controlled by mineralogical associations and landscape position rather than by total concentrations alone. Soro et al. (2023) demonstrated that Pb retained by Fe oxyhydroxides represents a potentially unstable pool in environments subject to redox oscillations, particularly in hydromorphic soils, where periodic reduction can trigger Pb release into the soil solution. At the landscape scale, Mufalo et al. (2024) highlighted that floodplain and low-energy depositional environments often act as secondary accumulation zones for Pb, where metals initially deposited by atmospheric emissions are subsequently redistributed and stabilized by Fe-rich phases. This mechanism explains the persistence of elevated Pb levels in floodplain soils even decades after the cessation of smelting activities.

Zinc displayed different behavior in hillslope soils, with relatively equivalent distribution across all four geochemical fractions, and concentrations exceeding natural background in all fractions (30.1 mg kg^−1^, Table 1). Its substantial proportion found in the residual fraction indicates partial immobilization in these soils (Shaheen & Rinklebe, 2014). In the floodplain, Zn concentrations in the labile fractions (F1, F2, and F3) decreased significantly, suggesting increased mobility in response to hydromorphic processes. This contrasting behavior of Zn between hillslope and floodplain soils is consistent with recent studies highlighting the sensitivity of Zn to changes in redox conditions and sedimentary dynamics. Soro et al. (2023) showed that, although Zn may be partially stabilized within residual mineral phases in well-drained soils, it remains more responsive than Pb to geochemical perturbations, particularly under fluctuating redox and pH conditions. Wunn et al. (2025) demonstrated that Zn typically clusters with anthropogenic metals such as Pb and Cd in PCA, but exhibits broader fractionation patterns reflecting its higher geochemical reactivity.

Cadmium was predominantly found in the labile fractions (F1 and F2; Table 2, Figs. 3 and 4) in both hillslope and floodplain soils. As one of the most toxic trace elements, Cd poses serious risks to biodiversity, soil microbial communities, and human health (Demková et al., 2017). Its prevalence in the exchangeable and reducible fractions is particularly concerning due to its high leaching potential (Jalali & Hemati, 2013; Li et al., 2022). The elevated Cd concentrations suggest that significant active contamination remains, even thirty years after smelter deactivation, with persistent geochemical mobility across the landscape, supported by the high mobility factor values (Fig. 4).Fig. 4. Mobility factor for Cd, Cr, Cu, Ni, Pb, and Zn in in the area most impacted by a former lead smelter in Santo Amaro (soils with [Pb] > 100 mg kg^−1^), in hillslope (HS) and floodplain (FP) soils

The trace metals less influenced by smelting activities (Cr, Ni and Cu) were predominantly associated with the residual fraction (F4—> 59%), indicating a dominant association with stable mineral phases resistant to weathering. These elements showed no clear relationship with anthropogenic contamination and are primarily controlled by lithological sources. Cr and Ni are commonly enriched in mafic and ultramafic rocks, and their presence in this region likely reflects the geochemistry of the Recôncavo and Tucano sedimentary basins, which include fluvial sediments from northwestern Archean greenstone mafic belts during Rift Late Cretaceous filling (dos Santos et al., 2017a, 2017b; Silva et al., 2001). Soils developed from these formations are naturally enriched in Cr and Ni (dos Santos et al., 2017a, 2017b; Gloaguen & Passe, 2017), contributing to elevated regional background values.

The Mobility Factor (MF), calculated from the relative contribution of labile geochemical fractions, should be interpreted as a context-dependent indicator of potential mobility. Cd and Pb are clearly the most labile metals, followed by zinc. In floodplain soils, which are frequently affected by periodic water saturation and redox oscillations, the exchangeable and reducible fractions (F1 + F2) provide a meaningful estimate of potential metal mobilization, as metals associated with amorphous and poorly crystalline Fe and Mn oxides may be released under reducing conditions. In contrast, in well-drained hillslope soils, metals associated with the oxidizable fraction (F3), including those bound to organic matter and sulfides, may also contribute to potential mobility under changing redox conditions. The absence of a clear distinction in the mobility factors of Cd, Pb, and Zn between these two geochemically distinct environments suggests that runoff-driven transport of these metals is still active between hillslopes and floodplains, contributing to a homogenized geochemical distribution pattern.

It should also be noted that the hydroxylamine hydrochloride used in the reducible step of the BCR protocol is a relatively weak extractant and does not fully dissolve well-crystallized iron oxides. As a result, a fraction of Fe-associated metals may remain in the residual fraction, even though they can still be mobilized under prolonged reducing conditions. This limitation of the sequential extraction scheme helps reconcile the fractionation results with the associations observed in SEM–EDS analyses, highlighting the complementary nature of bulk geochemical and microanalytical approaches in the next section.

These findings confirm that the primary pollutants Cd, Pb, and Zn, are mobile metals in soil samples from both landscape units. This is further supported by the high mobility factor (MF) values observed in both environments. The continued dynamic input of metals from hillslopes and their release in hydromorphic floodplain soils indicate ongoing transfer to nearby aquatic systems, including estuarine mangroves located less than one kilometer from the contaminated zone.

Factors influencing the mobility of the metals

Numerous studies have demonstrated that the mobility of trace metals is strongly influenced by the physicochemical properties of soils and mining residues (Fernandes et al., 2018; Shaheen & Rinklebe, 2014; Souza et al., 2018; Teixeira et al., 2021). To better understand these relationships, a principal component analysis (PCA) was conducted using the concentrations of labile fractions (F1 + F2) and soil properties across both landscape units floodplains (FV) and hillslopes (EC) (Fig. 5).Fig. 5. Principal Component Analysis between labile fractions F1 and F2, and physical and chemical soil attributes (COT—total organic carbon). (1) Hillslope soils; (2) Floodplain soils

In hillslope soils, the first two principal components explained 86.91% of the total variance: 55.17% by PC1 and 31.74% by PC2. The variable “distance to the smelter” was positioned within the main cluster of variables, indicating its stronger role in metal mobility compared to floodplains. Variables such pH, clay, and TOC emerge as the most influential factors in explaining the mobility of trace metals.

In the floodplain, the first principal component accounted for 60.71% of the variance, while the second explained 25.54% (total: 86.25%). The main cluster included also pH, clay, total organic carbon (TOC), and labile metal fractions, but also silt, indicating that fine-textured soils with high organic matter content are more effective at retaining metals. Sand content was clearly separated from the primary variables, highlighting its weaker contribution to metal retention. The variable "distance to the smelter" also appeared somewhat isolated, implying that, although proximity to the pollution source influences metal concentrations, its role in determining mobility is less important than soil texture and organic matter in floodplain environment.

Soil texture and organic matter play a fundamental role in modulating trace metal retention and potential mobility in the study area. The dominance of fine fractions (clay and silt) in Vertisols enhances the adsorption of metals through high surface area and reactive mineral sites, particularly in soils developed from Fe-rich shales. Although total organic carbon (TOC) contents were comparable between hillslope and floodplain soils, the interaction between organic matter and fine mineral fractions contributes to the stabilization of metals in the solid phase.

Spatial patterns of metal concentrations are not solely controlled by distance from the former smelter, but are strongly influenced by topography and sediment redistribution processes. Downslope transport of fine, contaminated particles promotes the accumulation of metals in floodplain environments, even at locations more distant from the emission source. Conversely, well-drained hillslope soils closer to the smelter may exhibit lower concentrations of labile metals due to limited sediment retention and more stable geochemical conditions. This interaction between texture, organic matter, and landscape position explains the heterogeneous distribution of metals observed across the study area and highlights the importance of considering geomorphological controls in assessments of soil contamination and metal mobility.

Insights from scanning electron microscopy and microanalyses

Scanning electron microscopy coupled with energy-dispersive spectroscopy (SEM–EDS) was employed as a complementary microanalytical approach to support the interpretation of bulk geochemical data. The SEM–EDS observations provide localized evidence of element associations within individual particles or coatings, but they do not represent the dominant mineralogical phases at the soil scale. Consequently, interpretations of element partitioning and potential mobility in this study are primarily based on the results of the BCR sequential extraction, which integrates information from the entire soil matrix. SEM–EDS results are used to corroborate these interpretations by illustrating plausible associations between metals and specific mineral components.

SEM and elemental mapping analyses revealed distinct spatial associations between Pb, Ca, and Fe in hillslope and floodplain soils. In hillslope soils, the presence of carbonate lenses within the underlying rock formations contributed to the development of partially carbonate-rich Vertisols, supporting a clear association between Ca and Pb (Fig. 6D). In these soils, Pb appears diffusely distributed and possibly stabilized by carbonate phases derived from the parent material, suggesting the likely presence of secondary Pb carbonate phases, such as cerussite. The higher proportion of Pb in the acid-soluble fraction (F1, Table 2) indicates a greater association with carbonate-related phases, which is consistent with this spatial association between Pb and Ca observed in EDS elemental maps.Fig. 6. Scanning electron microscopy and EDS microanalyses: EDS layered image of sample 24, hillslope soil (A) and of sample 44, floodplain soil (B); Pb mapping of sample 24, hillslope soil (C) and of sample 44, floodplain soil (D). EDS layered image of individual mineral, sample 44, floodplain soil, with Fe mapping (E) and Pb mapping (F). Biplots graphical of Ca, Fe, and Pb spot geochemical analysis in both samples (G e H)—see spectra in Figures S4, S5 and S6 (Supplementary material)

In contrast, floodplain soils exhibit different Pb behavior. The association between Pb and Fe (Fig. 6C and E) in floodplain soils suggests co-precipitation or strong sorption of Pb onto Fe oxides under fluctuating redox conditions. The BCR results (Table 2), characterized by a marked shift of Pb toward the reducible and oxidizable fractions (F2–F3) in floodplain soils, corroborate the EDS findings. Diffuse erosion on hillslopes can promote the downslope transport of contaminated particles into alluvial environments, where hydromorphic conditions favor the dissolution and reprecipitation of Fe oxyhydroxides typical of hydromorphic Vertisols and Gleysols in the region.

While Pb stabilization in hillslope soils may be associated with carbonate phases, its association with Fe oxides dominates in floodplain environments. This contrast highlights distinct Pb retention mechanisms across landscape units. Pb carbonates are generally more stable under oxic, well-drained conditions, whereas Pb associated with Fe oxides and hydroxides may become less stable under redox cycling (Mufalo et al., 2024; Soro et al., 2023). Such conditions may enhance Pb release into the soil solution, increasing its mobility and potential co-transport with fine Fe-rich particles to downstream fluvial and estuarine systems. Consequently, caution is warranted when assessing the long-term stability and mobility of Pb in these soils, as hydromorphic conditions in floodplain areas may intensify environmental risks. Overall, the results highlight the critical roles of soil mineralogy and hydrology in determining Pb stabilization and mobility in contaminated landscapes.

Conclusion

This study demonstrates that trace metal contamination in soils affected by legacy lead smelting in Santo Amaro is governed by a combination of spatial, pedological, and geomorphological controls. Spatial distribution patterns reveal strong heterogeneity, with elevated Pb and Zn concentrations not strictly decreasing with distance from the former smelter, but strongly influenced by landscape position, sediment redistribution, and hydrological connectivity. These elements are largely concentrated in the exchangeable and reducible fractions, indicating a higher environmental mobility and potential bioavailability, particularly in acidic conditions.

The high proportion of Pb and Cd in most labile fractions (F1 and F2) is of particular concern. The predominance of these fractions indicates a greater potential for Pb and Cd mobilization under changing environmental conditions. Besides, the geochemical behavior of metals differs markedly between hillslope and floodplain environments. In well-drained hillslope soils, Pb is predominantly stabilized by carbonate phases derived from parent material, whereas in floodplain soils, hydromorphic conditions and redox fluctuations promote stronger associations between Pb and Fe oxides, enhancing its potential mobility. Less mobile elements such as Cr and Ni are largely controlled by lithology and remain predominantly associated with residual mineral phases across both landscape units, while Cu exhibits an intermediate behavior reflecting both lithological inheritance and secondary anthropogenic inputs.

Identifying the allocation of metals among geochemical fractions is critical for environmental assessment and management, as total concentrations alone do not adequately reflect potential mobility or environmental concern. The integration of sequential extraction data with spatial analysis provides a more robust framework for distinguishing between lithogenic and anthropogenic contributions and for identifying areas where metals are more susceptible to redistribution, with potential transfer into nearby water systems, including rivers and estuaries located less than one kilometer from the former contamination source. Under hydromorphic conditions, the redistribution of metals from soils to aquatic environments may persist, highlighting the long-term environmental concern for downstream ecosystems, even more than 30 years after the deactivation of the smelter.

Supplementary Information

Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 2504 kb)