The sample locations and types are reported in Table 1 and mapped in Figure 1. The geological characteristics of the sampling sites and the descriptions of the samples collected are detailed in the following section.

All materials used for the recovery of major cation and trace metal analysis were prepared in a clean room and decontaminated using a 24 h nitric acid (1.5 M HNO₃) wash at 45°C, followed by a 24 h wash with deionized ultrapure water (18 MΩ-cm) at 45°C.

In order to compare the enrichment or depletion of REE as a whole series, they are generally normalized to a standard set of values. For this study we normalized REE concentrations in soil and water samples to Taylor and McLennan (1989)’s Upper Continental Crust (UCC) values. Elements were further classified into three groups, the light REE (LREE) which range from elements La to Nd, middle REE (MREE) Sm to Tb, and heavy REE (HREE) Dy to Lu. For soil, sediment, and aqueous samples the Ce anomaly (Ce*) and Eu anomaly (Eu*) were calculated using a linear extrapolation of UCC-normalized REE, Eqs. 1, 2 respectively: C e * = 2 C e U C C L a U C C + Pr U C C (1) E u * = 2 E u U C C S m U C C + G d U C C (2)

For comparisons of plant samples, the bioconcentration factor (BCF) and translocation factor (TF) were calculated using Eqs. 3, 4, accordingly, where [REE_AG] is the concentration of REE in the above-ground tissue (mg kg⁻¹) [REE_BG] is the concentration of REE in the below-ground tissue (mg kg⁻¹), and [REE_soil] is the concentration of REE in the soil or aquatic sediment collected from the plant rhizosphere (mg kg⁻¹). The resulting BCF and TF values are dimensionless ratios. The BCF is reflective of the plant’s ability to accumulate an element from the soil/aquatic sediment and TF is reflective of the ability for the plant to translocate elements from the rooting system to aerial tissue. The Ce* and Eu* for plant samples were determined using BCF values calculated from Eqs 1, 2, instead of UCC-normalized values: B C F = R E E A G R E E S o i l (3) T F = R E E A G R E E B G (4)

For statistical treatment, standard diagnostic tests were performed to determine whether data needed transformation. If a transformation was required, the box-cox method for transformation was used to determine the best fit. Data passing diagnostic tests were then tested with an Analysis of Variance (ANOVA), for statistical significance the p-value was limited to p < 0.05. The number, n, of all datasets analyzed is reported in Figure 1. Missing data are handled as random, allowing the analysis of imbalanced datasets, particularly soils and aquatic sediments. Post-hoc testing was completed using Tukey Honestly Significantly Difference (Tukey HSD).

In situ physicochemical water quality parameters are displayed in Supplementary Table S1. Waters collected from outside of São Domingos and Lousal were circumneutral, pH 7.1 and pH 6.7, accordingly. Similar Eh values were observed for water collected from outside of São Domingos and Lousal, +397 mV and +360 mV, respectively. Of the water samples collected from inside of the mining sites, the open pit at Quinta do Bispo had the highest pH, 5.1. The sample collected at Quinta do Bispo still had an oxidizing Eh of +524 mV. Samples collected from the São Domingos River Valley display a pH ranging from 2.4 to 2.9 and were strongly oxidizing with an Eh ranging from +691 mV to +723 mV. Similarly, samples collected from the Lousal treatment system had a pH ranging from 2.3 to 3.0 with an Eh ranging from +615 mV to +808 mV.

The EC for the samples collected from the mining areas tended to be higher than the EC observed outside of the mining areas. The range of DOC for the samples collected ranges from 0.45 mg·L⁻¹–9.71 mg·L⁻¹. Supplementary Table S1 presents the concentrations of anions, measured in aqueous samples. The dominant inorganic anion observed in the mining sites was sulfate, in contrast to the dominant anion outside of the mining area which was chloride. The presence of sulfate is anticipated as sulfuric acid is generated through AMD and used during ISL.

Upper Continental Crust (UCC) normalized aqueous REE patterns are displayed in Figure 2 while REE concentrations are tabulated in Table 2. The aqueous samples collected from outside of the mining area contained less REE than the samples collected from inside of the mining areas. The samples collected outside of the mining area exhibited depletion of LREE relative to both the MREE and HREE. The depletion of LREE was punctuated by a negative Ce anomaly (Ce*) for samples collected outside the mining area. Samples collected from the Lousal passive treatment system had the highest REE concentrations with 628.84 μg·L⁻¹ associated with the leachate treatment tank. The sample collected from the sixth treatment pool showed a marked decrease in dissolved REE with 238.68 μg·L⁻¹. Aqueous samples collected from the AMD areas also contained higher concentrations of REE than in the open pit at Quinta do Bispo, perhaps indicating that AMD is a more potent mobilizer of REE than ISL. There were some similarities between the aqueous REE patterns, all mines showed a MREE enrichment as compared to the LREE and HREE. In addition to the MREE concavity, samples collected from São Domingos and Quinta do Bispo had a skewed pattern, with MREE > HREE > LREE. The open pit sample from Quinta do Bispo also exhibited a strong positive Eu anomaly (Eu*), not shared by the other sites.

A summary of major cation concentrations for water and soils/aquatic sediments reported in Supplementary Tables S2, S3, respectively. Table 3 displays the REE concentration of each individual soil. When evaluating soil REE concentrations there were three effects considered: 1) Prevailing geochemical context, 2) the mining context, and 3) rhizosphere effects.

When considering the responses of plant REE concentration, BCF, and TF there were three main effects investigated: 1) Prevailing geochemical context, 2) the mining context, and 3) the plant type.

Though none of the studied mines targeted the extraction of REE ores, REE can be distributed throughout silicate, carbonate, oxide, phosphate, and oxyhydroxide (Migaszewski and Gałuszka, 2015) minerals, which are susceptible to dissolution under low pH conditions or changes in reduction/oxidation conditions (Cao et al., 2001; Mihajlovic et al., 2017). The MREE enrichment in aqueous samples is a typical signature of AMD (Perez-Lopez et al., 2010). The pH range and Eh observed for both São Domingos and Lousal are also consistent with the geochemical context of AMD (Dold, 2014). The pH observed in Quinta do Bispo was relatively high when compared with the pH of São Domingos and Lousal. This is most likely due to the fact that sulfate mineral oxidation or AMD is negligible in the surrounding geology (Carvalho et al., 2005) and the ISL methods employed at Quinta do Bispo ended in 1999 (European Commission Directorate-General for Energy, 2012). The saturation of cations and anions in waste streams, represented by the EC measured in aqueous samples, favors the precipitation of metals (Nieto et al., 2007), and is therefore also a parameter used to evaluate the precipitation/dissolution dynamics at these sites. Low pH, high EC, and elevated sulfate concentrations were indicative of the geogenic and anthropogenic dissolution of primary minerals inside of the three legacy mine sites. These geochemical changes were induced by AMD processes or ISL practices.

In addition to dissolution, processed REE were susceptible to sorption and desorption processes, which were evaluated in this study based on the CEC_Metson and OM content of soils and aquatic sediments. A soil with higher CEC_Metson and OM content has a high metal loading capacity, whereas a soil with a low CEC_Metson and OM content has less potential metal-sorbing surfaces, thus the metal holding capacity of soils collected from São Domingos is higher and the metal holding capacity of aquatic sediments collected from Lousal is lower. The sediments collected from Lousal had lower concentrations of REE indicative of the lower metal holding capacity of these sediments relative to the other sampled sites. As a result, Lousal also displayed the highest aqueous REE concentrations.

All of the samples collected from inside of the mining sites are considered impaired simply on the basis of having low pH (pH < 6.5) (Baker et al., 1990). As Baker et al. (1990) reported, biological diversity is lost starting at pH < 6.5, and an interruption of nutrient cycling at pH < 5.0. While pH directly impacts the ecology of a system, the increased metal solubility and bioavailability caused by acidification (Baker et al., 1990; Cao et al., 2001; Batista et al., 2013) can also induce toxicity (Baker et al., 1990; Plante, 2007; Batista et al., 2013).

Inorganic anions typically associated with REE complexation include: SO₄ ²⁻, PO₄ ³⁻, Cl⁻, CO₃ ²⁻, and F⁻ (Millero, 1992; Vaziri Hassas et al., 2021). As stated in the results, the mining areas were dominated by sulfate. A study by Zhimang et al. (2000) implicates sulfates in the bioavailability of REE to plants, mainly showing that REE accumulation in below-ground tissue is limited by sulfate. Secondary sulfate salts such as jarosite and alunogen have been shown to be temporary stabilizers of polyvalent cations (including REE) in AMD (Ferreira et al., 2021). Subsequent solubilization of metals associated with jarosite, alunogen, and other secondary sulfate minerals has been implicated as a cause for delayed revegetation (Ferreira et al., 2021). This demonstrates that plants growth in mining environments is dependent on the prevailing geochemistry, as primary and secondary minerals constitute a source or sink for potentially toxic metals.

Aside from sulfates being a sink for REE, sulfates have also been implicated in the selective partitioning of REE in soils and waters (Ferreira da Silva et al., 2009; Welch et al., 2009; Grawunder et al., 2014; Soyol-Erdene et al., 2018). REE have even been suggested as tracers of mining pollution due to their unique MREE-enriched signature in mining wastes (Merten et al., 2005; Perez-Lopez et al., 2010). This signature is presumably due to sulfate and sulfate intermediates showing higher stability in complexation with MREE (Perez-Lopez et al., 2010) and precipitation/sorption processes occluding LREE (Liu et al., 2019) and HREE (Byrne and Li, 1995) from solution, paired with dissolution/desorption of MREE-enriched Fe and S minerals (Welch et al., 2009; Perez-Lopez et al., 2010; Grawunder et al., 2014). The occurrence of a correlation between REE and Fe, as seen in the water samples collected in this study, may indicate that Fe-mineral dissolution contributes to the soluble REE in these systems (Cao et al., 2001; Merten, 2005; Prudêncio et al., 2015; Prudêncio et al., 2017; Riley and Dutrizac, 2017; Liu et al., 2022). Many geochemical factors may contribute to the observed high REE concentrations and unique MREE enrichment in AMD surface waters, a detailed study on REE speciation is still needed.

The DOC concentrations anticipated in AMD waters should be low as acidic pH promotes an increase of organic aggregation, inducing a decrease of their colloidal stability (Johnson and Hallberg, 2003; Pédrot et al., 2009). Though limited amounts of DOC and OM are associated with AMD systems, no broad generalizations can be made about the fractionation of REE by organics, which is specific to the type of organic matter. Organic matter forms a significant sink for REE, both as a sorbing surface and as a bridging ligand (Byrne and Li, 1995; Pourret et al., 2007). Organic ligands make a more stable complex with HREE, causing HREE to deplete from solution into complexes with sediment surface ligands (Byrne and Li, 1995). More complex organics, such as humic substances, have been shown to produce a fractionation of MREE enrichment over both LREE and HREE, specifically in natural systems characterized by a high metal loading like the sites studied (Pourret et al., 2007; Pédrot et al., 2008).

The oxidation state of REE is generally trivalent, with the exception of Eu²⁺ and Ce⁴⁺. In this study, a negative Ce* was associated with samples collected from outside the mining areas. Geochemically, the waters collected from outside of the mining area are less oxidizing and have a higher pH. Therefore, the speciation of metals may vary between waters collected from inside of the mining area versus those collected from outside of the mining area. The negative Ce* observed in surface waters collected from outside of the mining areas is indicative of preferential Ce⁴⁺ sorption to the solid phase, which is typical of circumneutral pH waters under oxidizing conditions (Bau, 1999).

The positive Eu* in liquors and ores associated with the open-pit can be attributed to the hydrothermal deposit where the Eu²⁺ oxidations state is preferentially incorporated into high-temperature hydrothermal seams over the other REE, leading to enriched Eu concentrations (Sverjensky, 1984; Danielson et al., 1992). The proceeding ISL processes increased mobilization of all REE contained in the Eu-enriched ore, accordingly producing an enrichment of Eu in the remaining liquor and analogous waste rock. This mechanism explains the positive Eu* seen in the open-pit water sample and the Salix sp. rhizosphere soil, which was collected from the waste rock pile. All other soil samples collected from Quinta do Bispo had a negative Eu*. Consequently, the conclusion can be drawn that the soils collected from the surrounding material were depleted in Eu as it was partitioned into the hydrothermal seam. Thus, in the soil samples collected from the surrounding areas, there is a negative Eu* corresponding to this partition.

In order for a plant type to be considered a REE hyperaccumulator, the REE concentration in the above ground-tissue must exceed 1,000 μg·g⁻¹ (Liu et al., 2018). None of the plants collected in this study met the hyperaccumulator definition. Liu et al. (2018) acknowledges the importance of plant biomass, as high biomass can accrue significant amounts of REE. The concentrations of REE observed in the collected plants do not show any particular elevation or depletion when compared with other studies of plant REE (Wyttenbach et al., 1998; Liang et al., 2008; Migaszewski and Gałuszka, 2015). The P. aquilinum plant species did accumulate more REE than any other plant type: though this figure represents 6–53 times the amount accumulated by the other plant samples in the present study, the pteridophyte sample is not remarkably high. Pteridophytes are generally known to accumulate much higher concentrations of REE over other plant types (Tyler, 2004). P. aquilinum also tended to concentrate certain REE at a higher level in the above-ground tissue than in the below-ground tissue. Wang et al. (1997) made a similar observation in Dicranopteris dichotoma, another pteridophyte, which accumulated more REE in the above-ground tissue than in the below-ground tissue.

As mentioned in the materials and methods section, the BCF is reflective of biological occlusion or assimilation of particular REE from the surrounding media. The closer the BCF value is to 1, the closer the plant concentrations reflect the soil concentrations. In this study, any uptake of REE by the plant was insignificant in comparison with the soil pool. This was proven by a comparison of bulk and rhizosphere soils which showed no depletion or enrichment dependent on plant growth, and further supported by the BCF which for most plants was much less than 1.

The BCF can also be looked at in terms of slope and anomaly, indicating the effects of soil speciation and plant species selectivity towards the assimilation of particular REE into the above-ground tissue. Ultimately this study showed that for certain plant types (P. aquilinum), above-ground tissues were enriched in LREE relative to the surrounding soil. This finding is congruent with previous research, which demonstrated that LREE tend to be enriched in the above-ground tissue of plants compared to the MREE and HREE (Wang et al., 1997; Fu et al., 1998; Liang et al., 2008; Gonzalez et al., 2014). This may have to do with the plant’s affinity to uptake LREE (Brioschi et al., 2012) and porewater chemistry restricting available HREE (Millero, 1992; Byrne and Li, 1995; Fu et al., 1998; Tyler, 2004; Brioschi et al., 2012). However, this pattern was not maintained across all plant types. In fact, T. latifolia and Salix sp. showed an MREE-enrichment compared with LREE and HREE, supported by Wyttenbach et al. (1998).

Salix sp. and P. aquilinum had positive Eu* in their BCF signatures. Positive Eu* in plant tissue has been observed in other studies (Durães et al., 2014; Krzcuik and Gałuszka, 2019; Krzciuk and Gałuszka, 2020). There are a couple of to-date unconfirmed hypotheses on why plants may have Eu*. Potentially Eu³⁺ is transformed to Eu²⁺ and is taken up by plants under the redox conditions of the rhizosphere. This may be a result of diurnal patterns, which dictate oxygen supply to the roots (Krzcuik and Gałuszka, 2020), or it may be related to the acquisition of other essential nutrients; such as Fe, which is reduced at the root dermis (Marschner and Rӧ;mheld, 1994; Wyttenbach et al., 1998). In both scenarios, the supply of Eu and the redox conditions are a vital control on the bioavailability of Eu. Many authors (Wang et al., 2012; Brioschi et al., 2013; Durães et al., 2014; Thomas et al., 2014) have suggested that the origin of positive Eu* in plants is the result of Eu²⁺ substitution for an essential plant nutrient such as Ca²⁺ due to the similarity in ionic radius (Shannon, 1976). All of the above-ground tissue REE concentrations positively correlate with above-ground tissue Ca²⁺ concentrations for this study.

The investigation of the TF can be used to make conclusions about the transfer and relative accumulation of REE in and to different parts of the plant. In this study, the REE transfer from below-ground tissue to above-ground tissue did not relate to the geochemistry or mining context. It was only reflective of plant type differences. The subsequent enrichment in the LREE seen in the C. ladanifer and P. aquilinum plants is then an artifact of the preferential translocation of LREE into the above-ground tissue. Fu et al. (1998) posited that preferential uptake and translocation of REE in a pteridophyte was related to plant Si processes and that in situ recrystallization of Si would allow for substitution of MREE and HREE occluding the larger ionic radius of LREE, leaving the LREE more mobile and able to reach the distal and aerial tissues. Evidence supporting this may be seen in the strong positive correlation between all of the REE and Si (Supplementary Table S6) among all samples collected in this study.

Combining the information presented on the BCF enrichment of MREE with the lack of a TF MREE enrichment, it can be concluded that the MREE fractionation into the above-ground plant tissue, observed in T. latifolia and Salix sp., originates from the preferential uptake of MREE in the below-ground tissue. While the total bioavailable pool of metals can include the soil (water-soluble, ligand-exchangeable, and organic or carbonate bound metals), the dissolved metals in water are typically thought of as readily bioavailable (Adriano, 2001; Di Bonito et al., 2008). These dissolved concentrations could explain the elevated MREE in plants, as the enrichment of the MREE in plant tissue mirrors the enrichment of MREE in surface waters impacted by mining.

The information presented may show that the MREE enrichment seen in the roots is an artifact of the soluble REE pool in these mine sites, whereas the LREE enrichment and Eu* in the above ground plant tissue demonstrates plant-specific fractionation. Thus, this study provides evidence that the REE pattern displayed by each plant may result from an interaction between the plant type and growth medium. This conclusion is contrary to one posited by Laul and Weimer (1982). Laul and Weimer (1982) studied a selection of biological samples and concluded that the REE patterns closely resembled their growth medium when normalized to chrondrites, however the REE patterns for plants were never normalized to the soil media to confirm enrichment or depletion, which would have obviated any preferential uptake of specific REE. Wang et al. (1997) similarly found that the concentration of REE in most plants collected from sites with REE-rich ores had similar patterns to the soils they were collected from, but also notes an accumulation of LREE over HREE. Pisciotta et al. (2017) made a similar conclusion that the plant tissue reflected the geochemistry of the collection site, however this conclusion was based on one plant type in one geochemical context. Another study by Krzcuik and Gałuszka (2020) showed that plants generally had a similar pattern for uptake regardless of the land usage, proving quite the opposite of earlier studies by Laul and Weimer (1982) and Pisciotta (2017), since this would mean that regardless of geochemical background, plants had a certain physiological signature for REE-uptake. In the context of Krzcuik and Gałuszka (2020) it should be noted that researchers did find some differences based on collection of samples inside or outside of AMD areas. A study conducted by Barbera et al. (2021) actually showed that a planted system when spiked with REE did not necessarily lead to increased uptake of REE into the above-ground tissue. Barbera et al. (2021) instead shows increased REE concentration in the below ground tissue with limited transfer of REE into the above-ground tissue. Wyttenbach et al. (1998) similarly refuted the idea that the growth media was completely responsible for REE fractionation in above-ground tissues. Thus, there is evidence to support claims from each of these previous studies and create a more cohesive theory that both geochemical background and plant type contribute to REE patterns in plant tissues. The results from the present study seem to indicate that some plant types may be reflective of the area geochemistry, whereas others may show increased fractionation of soil REE, which demonstrates the importance of capturing many plant species in differing contexts.