The legacy of mining revealed by environmental DNA: long-term ecological structuring of marine benthic communities after the Fundão dam collapse

TL;DR

This study uses environmental DNA to show how a 2015 mining disaster in Brazil continues to affect marine life more than three years later.

Contribution

The study reveals long-term ecological impacts of the Fundão dam collapse using eDNA metabarcoding to track benthic community changes.

Findings

The Front region near the river mouth showed a community dominated by metal-tolerant species like diatoms and protists.

The North and South regions had higher biodiversity and included benthic invertebrates like nematodes and sea cucumbers.

Species richness and diversity varied significantly across regions, indicating persistent ecological disruption from mining waste.

Abstract

Coastal marine ecosystems are key components of biodiversity and ecosystem functioning but have been increasingly degraded by human activities. One of the most severe environmental disasters in Brazil occurred in November 2015, when the Fundão tailings dam collapsed in Mariana (Minas Gerais), releasing approximately 40 million m3 of iron ore waste into the Rio Doce basin and adjacent coastal environments. To evaluate the long-term biological consequences of this event, we analyzed the taxonomic composition and diversity of marine communities using environmental DNA (eDNA) metabarcoding from sediment cores collected in 2018 across three coastal sectors—Front (mouth of the Doce River), North, and South. A total of 761,517 reads generated 11,061 unique amplicon sequence variants (ASVs) assigned to 148 taxa revealing significant spatial variation in taxonomic (species-level) composition and diversity indices (PERMANOVA, pseudo-F = 16.55; p = 0.047). The South region exhibited the highest species richness (q₀ = 103 taxa), followed by the North (97) and Front (70). Cluster and SIMPER analyses indicated two distinct biological assemblages: (1) the Front region, dominated by diatoms (Mediophyceae, Bacillariophyceae) and protists tolerant to metal enrichment, and (2) the North–South regions, characterized by higher evenness and presence of benthic invertebrates such as Holothuroidea and nematodes (Desmodorida). Species abundance distribution (SAD) models differed among areas, reflecting ecological gradients associated with the dispersal and chronic accumulation of mining residues. These results demonstrate a persistent imbalance in marine communities near the Doce River mouth, suggesting that the legacy of historical contamination and the Fundão dam failure continues to shape benthic biodiversity patterns more than three years after the disaster.

The online version contains supplementary material available at 10.1007/s10661-026-15092-9.

Figures

Map of the Rio Doce coastal region. The five sampling stations—F1 (Front), N1–N2 (North), and S1–S2 (South)—are displayed near the river mouth

Analysis for patterns of co-occurrence and similarity among taxa. a CLUSTER analysis: The CLUSTER analysis with the SIMPROF test identified that the three-sampling area were divided into two groups (colored cyan and coral) that were statistically distinct. The gray dashed lines indicate the levels of similarity at which the clusters were grouped by the SIMPROF test. b UpSet Plots: The UpSet plots were used to verify the relationships between sets of taxa and identify patterns of co-occurrence among the groups formed in the CLUSTER analysis. Colors represent different areas: coral for front, bl

Relative read abundance derived from eDNA metabarcoding (a), species richness (Hill number *q*₀), diversity of rare taxa (*q*₁), and dominance index (*q*₂) in marine benthic communities from areas adjacent to the Doce River mouth (Front), North, and South regions. Abundance values represent relative ASV read proportions, not direct organism counts. Different letters indicate statistically significant differences among regions

Partitioning of beta diversity among benthic communities from the Front, North, and South sectors based on the Baselga ([2010](#CR10)) framework. Lower‐triangular heatmaps show pairwise Sørensen dissimilarity (βSOR) and its turnover (βSIM) and nestedness (βSNE) components calculated from presence–absence data. All panels share a common color scale, highlighting the dominance of species turnover in driving spatial differentiation among sectors

Abundance distribution models of species sampled at the mouth of the Rio Doce and adjacent North and South areas. The values for the Akaike criteria are presented in the supplementary material (Supplementary Table 5)

Chord diagram illustrating dissimilarities between different sampling areas according to the SIMPER analysis results. The diagram shows taxa contributing to 80% of the dissimilarities between sampling points: a North, South, and Front; b North and South; c South and Front; and d North and Front. Connections between groups reflect the relative contributions of each taxon to differences in community composition

Full text

Introduction

Coastal and marine ecosystems support high levels of biodiversity and provide essential ecological services, including nutrient cycling, carbon sequestration, and fisheries productivity (Cruz & Manuel, 2021; Czachur et al., 2021; Ochieng et al., 2023). However, these ecosystems are increasingly exposed to anthropogenic stressors that reduce environmental quality, disrupt community structure, and impair ecosystem functioning (Halpern et al., 2008; Milon & Álvarez, 2019).

In Brazil, one of the most severe examples of large-scale environmental degradation was the collapse of the Fundão tailings dam, located in Mariana (Minas Gerais), on 5 November 2015. The disaster released approximately 40 million cubic meters of iron ore tailings into the Rio Doce watershed, which were transported for over 600 km downstream until reaching the Atlantic Ocean (Gomes et al., 2021; Mendes et al., 2020). This event caused extensive contamination of sediments and aquatic biota, altering the estuarine and coastal dynamics of the region and generating long-lasting environmental impacts (Merçon et al., 2021; Bonecker et al., 2019, 2022; Condinni et al., 2022).

Once discharged into the coastal zone, hydrodynamic and oceanographic processes promoted the wide dispersion of fine tailings, which remained in suspension for extended periods before settling on the inner continental shelf (Aguiar et al., 2023; Lemos et al., 2022; Quaresma et al., 2015, 2020). Deposition was particularly pronounced near the river mouth, where reduced current velocities favor the accumulation of fine sediments and the formation of nearshore mud belts (Bastos et al., 2015; Quaresma et al., 2015). These deposits are enriched with metals such as iron (Fe), manganese (Mn), arsenic (As), cadmium (Cd), copper (Cu), zinc (Zn), nickel (Ni), and chromium (Cr), which can persist in sediments and exert long-term effects on benthic ecosystems (Cagnin et al., 2017; Gabriel et al., 2020; Nascimento et al., 2023). Such contamination alters sediment composition, modifies habitat suitability, and drives shifts in benthic community structure, reinforcing the relevance of evaluating biological responses to both historical mining activities and the Fundão dam failure in coastal marine environments (Bernardino et al., 2019; Bonecker et al., 2019, 2022; Condinni et al., 2022).

Previous studies conducted in the Doce River basin and adjacent coastal environments have documented severe biological consequences in systems affected by the Fundão dam collapse, including metal bioaccumulation, oxidative stress, genetic damage, and simplification of community structure in both freshwater and marine environments (Barrilli et al., 2024; De Biasi et al., 2023; Lines et al., 2023). These findings indicate that the impacts of the disaster extend beyond acute effects, interacting with the legacy of two centuries of mining-related contamination that already characterized the Doce River basin (Cagnin et al., 2017; Gabriel et al., 2020; Nascimento et al., 2023).

Traditional environmental assessments based on chemical and physical analyses or the monitoring of visible fauna remain essential tools for evaluating environmental status (Cremonez et al., 2014; Cross et al., 2021; Smith, 1991). However, they often fail to capture subtle or long-term ecological changes, as they rely on spatially and temporally limited sampling (Banyal et al., 2019; Dias et al., 2019; Rosenberg et al., 2018). To overcome these limitations, molecular methods such as environmental DNA (eDNA) metabarcoding have become powerful alternatives, enabling high-sensitivity detection of taxa from DNA fragments preserved in sediments (Clarke et al., 2021; Taberlet et al., 2018; Thomsen & Willerslev, 2015). Sediments act as natural archives of biodiversity, retaining genetic material that integrates both current and historical community composition (Capo et al., 2020).

In this study, sedimentary eDNA was used to assess spatial variation in the taxonomic structure of marine benthic communities across three regions adjacent to the Doce River mouth—Front, North, and South—sampled in 2018, 3 years after the Fundão dam collapse. Previous studies have shown that the dispersal of mining tailings and local hydrodynamics strongly influence contamination gradients and benthic assemblages in this region (Aguiar et al., 2020, 2023; Quaresma et al., 2015, 2020), and that areas historically exposed to metal contamination tend to be dominated by pollution-tolerant or opportunistic taxa (Giongo et al., 2020; Ngole-Jeme & Ndava, 2023). Based on this framework, this study tests the hypothesis that marine benthic community composition differs among the Front, North, and South regions as a function of long-term exposure to mining-derived sediments.

By integrating molecular diversity metrics and ecological modeling, this work provides new insights into the long-term effects of mining legacies on benthic biodiversity and the resilience of impacted marine ecosystems in the Doce River coastal region.

Materials and methods

The study was conducted on marine sediments collected from the continental shelf adjacent to the mouth of the Doce River, on the eastern Brazilian coast. This region presents contrasting sedimentary and hydrodynamic characteristics between its northern and southern sectors, influenced by river discharge, coastal currents, and storm-induced resuspension (Bastos et al., 2015; Quaresma et al., 2015).

Sampling was designed to capture spatial variability along this coastal gradient and was carried out in November 2018.

Sample collection

In November 2018, 3 years after the disaster, five sediment cores were collected using a gravity corer: one directly in front of the river mouth (F1, 22-m depth), two to the north (N1, 14 m; N2, 12 m), and two to the south (S1, 20 m; S2, 16 m). Each core was sliced every 2 cm throughout its length to examine vertical stratification. Although surface (0–2 cm) and deep (28–30 cm) sediment layers were initially selected to represent recent and older depositional periods, respectively, these layers were combined for subsequent analyses (Fig. 1). This approach was adopted because sedimentary eDNA integrates biological signals over extended temporal scales due to DNA preservation, bioturbation, and resuspension processes, and separating layers would not provide sufficient replication for robust temporal comparisons.Fig. 1. Map of the Rio Doce coastal region. The five sampling stations—F1 (Front), N1–N2 (North), and S1–S2 (South)—are displayed near the river mouth

Immediately after collection, samples were frozen, transported to the laboratory, and preserved in 100% ethanol at − 20 °C until analysis.

eDNA extraction and amplification

Environmental DNA (eDNA) was extracted using the FastDNA™ SPIN Kit (MP Biomedicals). A filter sample obtained by filtering water from an aquarium with known species was used as a positive control. Amplification was performed using primers specific for the 18S gene, as described by Pochon et al. (2013). The first round of amplification was performed using the Platinum™ SuperFi II PCR Master Mix (ThermoFisher) protocol with an annealing temperature of 55 °C and 30 reaction cycles. A negative control was included for each PCR reaction batch. After the first round of amplification, a second round was performed to attach Illumina adapters and indices to all samples that produced visible bands on an agarose gel. Samples were purified before each round of amplification.

Environmental DNA (eDNA) was extracted from approximately 0.25 g of each sediment slice (0–2 cm and 28–30 cm) using the FastDNA™ SPIN Kit for Soil (MP Biomedicals, USA), following the manufacturer’s protocol. All procedures were performed in a clean area under UV-sterilized laminar flow hoods, using sterile equipment and filter tips to minimize contamination. Extraction blanks were included for each batch, and a positive control (aquarium seawater containing known marine taxa) was used to verify amplification success.

The 18S rRNA gene was selected as a universal marker for eukaryotic diversity, enabling detection across a wide range of benthic and planktonic taxa (Taberlet et al., 2018; Thomsen & Willerslev, 2015). Because the 18S rRNA gene provides limited taxonomic resolution for metazoans compared to mitochondrial markers, interpretations involving multicellular taxa were restricted to higher taxonomic levels and community-level patterns rather than species-level inference. Amplification targeted the V9 hypervariable region using primers 1380 F (5’-CCCTGCCHTTTGTACACAC-3’) and 1510R (5’-CCTTCYGCAGGTTCACCTAC-3’) (Linda et al., 2009). PCR reactions were conducted in triplicate (25 µL total volume) containing 1 × PCR buffer, 1.5 mM MgCl₂, 0.2 mM dNTPs, 0.4 µM of each primer, 1 U of Taq DNA polymerase, and 2 µL of template DNA. Cycling conditions were initial denaturation at 94 °C for 5 min, followed by 35 cycles of 94 °C for 30 s, 57 °C for 45 s, 72 °C for 45 s, and a final extension at 72 °C for 10 min.

PCR products were visualized on 1.5% agarose gels, pooled, and purified using the QIAquick PCR Purification Kit (QIAGEN, Germany). Libraries were prepared using the Illumina Nextera XT kit and sequenced on a MiSeq platform (2 × 300 bp, Illumina Inc.).

Bioinformatic

Raw reads were inspected using FastQC v0.11.9 (Andrews, 2010) and trimmed for quality and adapters using Cutadapt v3.5 (Martin, 2011), removing sequences shorter than 250 bp or with an average Phred score below Q30. Denoising, merging of paired-end reads, and chimera removal were performed with the DADA2 algorithm implemented in QIIME2 v2023.9.1 (Bolyen et al., 2019).

Unique Amplicon Sequence Variants (ASVs) were inferred, and only ASVs occurring in at least two samples and representing > 0.005% of total reads were retained to minimize sequencing artifacts. Taxonomic assignment was conducted using a Naïve Bayes classifier trained on the SILVA v138.1 reference database (Quast et al., 2013), filtered for the 18S rRNA V9 region. Although alternative curated databases such as PR2 are widely used for eukaryotic metabarcoding, all downstream analyses were deliberately conducted at the family level to minimize potential database-related misclassifications and to ensure robust ecological interpretation.

Following quality control, a total of 761,517 high-quality reads were obtained, generating 11,061 ASVs across 148 taxa. Sequence data were rarefied to the lowest read depth (32,500 reads per sample) before downstream analyses to standardize sampling effort. All scripts were implemented in QIIME2 and R v3.4.3 for transparency and reproducibility.

Data analysis

All analyses were performed at the family level to facilitate comparisons between approaches, as only approximately 40% of the taxa were identified at the species level. All analyses used* R* version 3.4.3 (R Core Team, 2017).

Hierarchical clustering (CLUSTER analysis), using the unweighted pair-group method with arithmetic means, was applied to identify species that tend to co-occur in samples and group samples with similar taxonomic compositions. To assess the significance of the clusters, the Similarity Profile (SIMPROF) test, which employs the Bray–Curtis distance, was used to measure differences in sample composition based on species abundance. In this study, abundance was inferred from the relative read abundance of amplicon sequence variants (ASVs) obtained through eDNA metabarcoding, rather than from direct organism counts or microscopic observations. This technique performs a permutation test of the null hypothesis that a specified set of samples with no predefined clusters does not differ in its multivariate structure. In this study case, 10,000 permutations were applied to calculate the average similarity profile, 999 simulated profiles were generated, and a significance level of 5% was adopted. Additionally, to analyze the intersections and overlaps of taxonomic data between the three areas, UpSet plots were generated using the UpSetR package in R.

Diversity was quantified using Hill numbers (Hill, 1973), a unified family of diversity indices that vary in their sensitivity to species relative abundances (Jost, 2006). We calculated q₀ (species richness), q₁ (equivalent to the exponential of Shannon entropy), and q₂ (equivalent to the inverse of Simpson’s dominance index), using the entropart package in R (Marcon & Hérault, 2015). These indices allowed the assessment of richness, evenness, and dominance patterns across the studied regions.

Spatial variation in community composition among the Front, North, and South sectors was assessed using beta diversity metrics based on presence–absence data aggregated by region, minimizing biases associated with read abundance and sequencing effort. A taxon was considered present in a given sector when detected in at least one sampling station within that area. Total beta diversity was calculated using the Sørensen dissimilarity index (βSOR) and partitioned into turnover (βSIM) and nestedness (βSNE) components following Baselga (2010). Pairwise dissimilarities were computed using the betapart package in R (Baselga & Orme, 2012), and the relative contribution of each component was summarized.

To test for differences in species composition and abundance among areas, we performed a permutational multivariate analysis of variance (PERMANOVA; Anderson, 2006) using the adonis2 function implemented in the vegan package. The model included “Area” as a fixed factor, and the response variables were species composition, total abundance, and diversity indices (q0, q1, and q2), analyzed separately. Differences in each variable were assessed using metabarcoding data, where the relative reads of each species represented abundance. Significance was assessed with 999 permutations for each variable.

To identify the families that contributed the most to the differences in taxonomic composition between sites, a SIMPER (Similarity Percentage Analysis) was performed using the vegan package in R (Oksanen et al., 2022). The analysis was based on a dissimilarity matrix generated using the Bray–Curtis index, which is appropriate for abundance data, with a cut-off threshold of 80% for cumulative species contribution. Additionally, Chord diagrams implemented in the circle package in R were used to visualize the taxonomic flow between areas.

Species Abundance Distribution (SAD) models were applied to characterize the sampled environments and infer the ecological processes shaping community structure. These models were adjusted by comparing observed data with the following theoretical distributions: Broken-stick, Preemption, Lognormal, Zipf, and Zipf-Mandelbrot (Magurran, 2004; McGill et al., 2007). The best-fitting model was identified using maximum likelihood estimation and evaluated through Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Deviance Information (model residuals), with lower criterion values indicating a better fit. By enabling the assessment of the balance between rare and dominant species, SAD models help detect responses to environmental disturbances or changes in community composition (Matthews & Whittaker, 2015; Barrilli et al., 2024,2025). Comparing theoretical and observed distributions provides insights into the mechanisms driving community dynamics, allowing predictions of ecological trajectories and evaluations of ecosystem resilience (Cheng et al., 2012). To explicitly link the proposed hypotheses to the analytical framework, we adopted an integrative approach combining multivariate, diversity-based, and distributional analyses. This framework allowed us to evaluate spatial differentiation among coastal sectors, identify dominance patterns associated with disturbance, and assess ecological organization along a gradient of mining influence.

Results

A total of 761,517 high-quality reads were obtained, resulting in 11,061 unique ASVs assigned to 148 taxa across all regions. Each sediment core contributed two slices (0–2 cm and 28–30 cm), analyzed together to represent integrated benthic communities from each site. (Supplementary Table 1).

Multivariate PERMANOVA based on Bray–Curtis dissimilarity confirmed significant differences in community composition among regions (pseudo-F = 16.55, p = 0.047; R^2^ = 0.32; Supplementary Table 2). Hierarchical clustering and SIMPROF tests identified three distinct assemblages corresponding to the North, South, and Front (central) regions (Fig. 2a). Despite limited replication at the Front station, its samples formed a clearly separated cluster, associated with finer sediments and higher concentrations of metals (Fe, Mn, As, Cu, Zn). UpSet plots (Fig. 2b) showed that 42% of ASVs were shared between the North and South regions, while only 18% occurred in all three areas, confirming spatial segregation and low overlap in community composition.Fig. 2. Analysis for patterns of co-occurrence and similarity among taxa. a CLUSTER analysis: The CLUSTER analysis with the SIMPROF test identified that the three-sampling area were divided into two groups (colored cyan and coral) that were statistically distinct. The gray dashed lines indicate the levels of similarity at which the clusters were grouped by the SIMPROF test. b UpSet Plots: The UpSet plots were used to verify the relationships between sets of taxa and identify patterns of co-occurrence among the groups formed in the CLUSTER analysis. Colors represent different areas: coral for front, blue for south, and cyan for north

Relative organism abundance, inferred from ASV read counts, did not differ among the three sampled areas (Front, North, and South), as shown in Fig. 2a. In contrast, diversity metrics exhibited clear spatial variation among coastal sectors (Fig. 3b–d; Supplementary Table 2). Diversity based on rare species (q₀) was highest in the South (103 ± 8 taxa), followed by the North (97 ± 10 taxa) and the Front region (70 ± 6 taxa).Fig. 3. Relative read abundance derived from eDNA metabarcoding (a), species richness (Hill number q₀), diversity of rare taxa (q₁), and dominance index (q₂) in marine benthic communities from areas adjacent to the Doce River mouth (Front), North, and South regions. Abundance values represent relative ASV read proportions, not direct organism counts. Different letters indicate statistically significant differences among regions

Although differences in diversity based on rare species were evident, diversity based on common (q₁; Fig. 3c) and dominant species (q₂, Fig. 3d) was more comparable among regions. Even so, the composite samples revealed a consistent diversity gradient along the coast, with the South sector exhibiting higher diversity weighted toward dominant species (q₂ = 0.72 ± 0.05), whereas the Front sector was characterized by stronger dominance of a few tolerant taxa (q₂ = 0.54 ± 0.04). These spatial patterns were statistically supported by PERMANOVA, which detected significant differences among regions for diversity based on rare species (q₀; pseudo-F = 69.02, p = 0.027), but not for diversity based on common (q₁; p = 0.729) or dominant species (q₂; p = 0.889).

Beta diversity analyses revealed pronounced compositional differentiation among the Front, North, and South sectors (Fig. 4). Sørensen dissimilarity (βSOR) was high between the Front and North (0.49) and between the Front and South (0.50), whereas dissimilarity between the North and South sectors was considerably lower (βSOR = 0.18), indicating greater taxonomic similarity between these two regions. Partitioning of βSOR demonstrated a clear predominance of species turnover (βSIM), which accounted for most of the observed dissimilarity across all pairwise comparisons, particularly between the Front sector and the other two regions (βSIM = 0.38–0.39). In contrast, nestedness contributed only marginally to total beta diversity (βSNE = 0.03–0.12), indicating that spatial differences in community composition are primarily driven by species replacement rather than by ordered species loss. (Supplementary Table 3).Fig. 4. Partitioning of beta diversity among benthic communities from the Front, North, and South sectors based on the Baselga (2010) framework. Lower‐triangular heatmaps show pairwise Sørensen dissimilarity (βSOR) and its turnover (βSIM) and nestedness (βSNE) components calculated from presence–absence data. All panels share a common color scale, highlighting the dominance of species turnover in driving spatial differentiation among sectors

The Relative Abundance Distribution (RAD) models revealed contrasting community structures among the studied areas (Fig. 5). The Preemption model best described the Front region (AIC = 1159, Supplementary Table 4), indicating a strong dominance pattern typical of disturbed or metal-enriched sediments. In contrast, power-law models provided the best fit for the other areas, with the Zipf model describing the North assemblage (AIC = 3054) and the Zipf–Mandelbrot model best fitting the South assemblage (AIC = 1665).Fig. 5. Abundance distribution models of species sampled at the mouth of the Rio Doce and adjacent North and South areas. The values for the Akaike criteria are presented in the supplementary material (Supplementary Table 5)

SIMPER analysis indicated that a small set of taxa accounted for over 80% of total dissimilarity among regions (Supplementary Table 5). The main contributors were Mediophyceae (15.3%), Bacillariophyceae (9.7%), Pseudoperkinsidae (7.6%), and Holothuroidea (5.7%). The Chord diagram (Fig. 6) illustrates the strength of these associations and shows distinct compositional transitions between North–South (balanced communities) and South–Front (microbialdominance). The Front region was dominated by Cyclopoida, Pseudoperkinsidae, Mediophyceae, and Ulvophyceae. The North exhibited a more balanced composition with Phyllodocida, Desmodorida, and Prasinophytae, whereas the South was characterized by Holothuroidea, Spionida, Desmoscolecida, and Pleosporales.Fig. 6. Chord diagram illustrating dissimilarities between different sampling areas according to the SIMPER analysis results. The diagram shows taxa contributing to 80% of the dissimilarities between sampling points: a North, South, and Front; b North and South; c South and Front; and d North and Front. Connections between groups reflect the relative contributions of each taxon to differences in community composition

Discussion

The collapse of the Fundão tailings dam in 2015 caused one of the most severe mining-related environmental disasters in South America, profoundly altering the physical and biological integrity of the Doce River system and adjacent marine habitats. Using sedimentary eDNA, our study reveals clear spatial differentiation in benthic community structure among the Front, North, and South sectors, consistent with persistent contamination legacies and the heterogeneous dispersion of mining-derived sediments along the coastal gradient (Aguiar et al., 2023; Lemos et al., 2022). The integration of surface and deeper sediment layers provides a composite signal of benthic communities, supporting our focus on long-term spatial structuring rather than short-term or depth-specific temporal dynamics. Although quantitative analyses were standardized at the family level, occasional references to other taxonomic ranks were used solely to support ecological interpretation and link detected taxa to well-established functional traits.

The marked differentiation among the Front, North, and South sectors reflects the operation of environmental filtering under chronic contamination, a process widely recognized as a primary driver of benthic community organization in impacted coastal systems (Bernardino et al., 2019; Elliott & Quintino, 2007; Ngole-Jeme & Ndava, 2023). The Front region, directly at the river mouth, exhibited the lowest diversity based on rare species and the highest diversity based on dominant species, consistent with ecologically stressed environments subject to irregular sediment deposition (Zhang et al., 2021). In contrast, the South region, despite being exposed to multiple anthropogenic pressures including historical mining activities, riverine transport of contaminated sediments, coastal urbanization, and port-related activities (Coelho et al., 2023; Ferreira, 2022; Macêdo et al., 2023), exhibited the highest diversity based on rare species as well as the highest diversity based on common species, indicating a comparatively more diverse assemblage with signs of partial resilience. The North region displayed intermediate diversity patterns, consistent with reduced fluvial influence and comparatively milder contamination gradients (Smeti et al., 2019). Also, the higher diversity observed in the South sector is likely favored by greater sediment stability, reduced influence of the riverine plume, and more consistent hydrodynamic conditions, which together promote habitat heterogeneity. These conditions contrast with the Front region, where frequent resuspension events and fine sediment deposition limit the persistence of sensitive organisms.

Beta diversity analyses helped elucidate the ecological processes structuring benthic assemblages along the coastal gradient. The high Sørensen dissimilarity between the Front sector and both the North and South sectors indicates strong compositional segregation at the river mouth, where local environmental conditions differ markedly from adjacent areas. In this region, recurrent sediment resuspension, fine particle deposition, and elevated contaminant loads create unstable and stressful habitats that restrict the establishment of sensitive taxa and favor organisms capable of tolerating physical disturbance and chemical stress (Barrilli et al., 2024; Ellis et al., 2017). The predominance of species turnover over nestedness suggests that spatial differences in community composition arise mainly through taxonomic replacement rather than simple species loss (Baselga, 2010; Barrilli et al., 2026), with sensitive taxa being progressively excluded and replaced by tolerant or opportunistic groups adapted to unstable substrates. In contrast, the lower beta diversity observed between the North and South sectors likely reflects more comparable environmental conditions, including greater sediment stability and reduced direct influence of the riverine plume, which likely promote higher ecological connectivity and allow a broader range of taxa to persist.

Species abundance distribution models further support this interpretation by revealing differences in community organization consistent with varying levels of environmental constraint across the study area. The dominance of Preemption-type abundance distribution in the Front region indicates resource monopolization by few tolerant taxa (Tatsumi et al., 2021), while Zipf and Mandelbrot models in the North and South reflect, respectively, community equilibrium and high species turnover under fluctuating environmental conditions (Barrilli et al., 2021, 2024; Gray, 2002). Together, beta diversity patterns and abundance distribution models reveal a gradient of resistance and resilience across the Doce coastal system: the Front remains impoverished and dominated by opportunists, the South harbors resistant but functionally limited assemblages, and the North retains intermediate diversity with partial recovery potential. These patterns align with eDNA-based evidence of persistent ecological imbalance across trophic levels (De Biasi et al., 2023).

This spatial heterogeneity is reflected in marked differences in taxonomic composition among coastal sectors, with specific groups contributing disproportionately to regional dissimilarities. In the most impacted areas, particularly the Front sector, assemblages were dominated by taxa associated with diatom-rich communities, including Mediophyceae and Bacillariophyceae, as well as ichthyosporeans such as Pseudoperkinsidae. These groups are known to tolerate elevated metal concentrations and low-oxygen conditions, and their prevalence is consistent with environments subjected to chronic contamination and physicochemical stress (Hancock et al., 2021; Ngole-Jeme & Ndava, 2023). The frequent occurrence of copepod taxa affiliated with Cyclopoida and ciliates within Heterotrichea, widely recognized as indicators of organic and metal enrichment in disturbed sediments (Dahms et al., 2009; Yang et al., 2009), further reinforces this pattern. The presence of Ulvophyceae and Zoantharia in the Front sector, opportunistic taxa that commonly proliferate under nutrient enrichment and hypoxic conditions (Hardouin et al., 2016; Holzinger et al., 2015), likely reflects eutrophic pulses and altered trophic pathways following tailings deposition.

These biological patterns are closely linked to the long-term contamination history of the Doce River basin. Elevated concentrations of Fe, Mn, Cr, Ni, Cu, Zn, and As have been documented in the basin well before the Fundão dam collapse, reflecting more than two centuries of mining activity (Bernardino et al., 2019; Cagnin et al., 2017; Gabriel et al., 2020). The dam failure subsequently delivered additional pulses of fine sediments enriched in Fe, Mn, and As to the estuarine and coastal environments (Kananizadeh et al., 2024; Nascimento et al., 2023; Segura et al., 2016), contributing to the formation of persistent metal-enriched depositional zones on the inner continental shelf. Oceanographic processes, including southward sediment transport and resuspension during storm events, likely facilitated the redistribution of these materials, helping to explain the spatial extension of ecological stress and the altered community structure observed, particularly in the southern sector (Aguiar et al., 2023; Quaresma et al., 2020).

Taken together, these results indicate community restructuring driven primarily by environmental filtering, whereby tolerant organisms persist under chronic stress while sensitive species are progressively constrained or excluded (Whitfield, 2021; Wilson & Chiarucci, 2000). Although some taxa may exhibit resilience through physiological tolerance or behavioral plasticity, the persistence of contamination and sediment instability suggests that full functional recovery remains uncertain (Hernández-Andreu et al., 2024; Valdivia et al., 2021). In this context, the observed ecological patterns are consistent with the biological traits of dominant and sensitive taxa along the contamination gradient. Diatoms and ichthyosporeans, characterized by high reproductive rates, short generation times, and physiological mechanisms that confer tolerance to metal enrichment and hypoxia, are able to persist under chronic environmental stress (Hancock et al., 2021; Ngole-Jeme & Ndava, 2023). In contrast, benthic invertebrates typically associated with more stable substrates, including several nematode and echinoderm taxa, tend to be more sensitive to sediment contamination and oxygen limitation, resulting in reduced occurrence in highly impacted areas. Together, these trait-based differences support an interpretation of long-term community restructuring driven by environmental filtering rather than complete ecological recovery.

Conclusions

This study provides the first eDNA-based evidence of the long-term ecological consequences of the Fundão dam collapse on marine benthic communities along the Doce River coast. Although direct chemical measurements were not included in this study, the consistent spatial differentiation of eukaryotic assemblages likely reflects the legacy of mining-derived sediments, with community composition serving as a biological proxy for the presence and persistence of contaminants in the system.

Our results indicate that the Front region functions as a depositional sink, dominated by metal-tolerant diatoms, protists, and opportunistic taxa, whereas the South and North sectors host more diverse assemblages with signs of partial functional recovery. The dominance of the Preemption abundance distribution in the Front region, contrasted with the prevalence of power-law-based distributions (Zipf and Zipf–Mandelbrot) in the North and South, highlights distinct ecological processes associated with environmental stress, community reorganization, and increasing structural complexity.

These findings support the hypothesis that environmental filtering plays a central role in shaping benthic biodiversity in mining-impacted areas, favoring tolerant taxa and reducing community evenness shortly after disturbance. While our data reflect conditions observed three years after the disaster, they suggest that recovery trajectories remain constrained by the persistence of fine, contaminated sediments, as also reported in previous studies of the Doce River system.

Finally, the integration of eDNA metabarcoding, multivariate analyses, and species abundance distribution modeling proved to be a robust approach for detecting subtle and spatially structured impacts of mining activities. We recommend the adoption of long-term genetic biomonitoring frameworks, integrated with chemical and sedimentological assessments, to better evaluate ecosystem recovery and inform restoration strategies in the Doce River estuary and other mining-affected coastal ecosystems worldwide.

Supplementary Information

Below is the link to the electronic supplementary material.ESM 1(DOCX 50.5 KB)