Long-term soil, vegetation, and surface temperature dynamics in mining tailings–affected landscapes revealed by multitemporal remote sensing

ABSTRACT

Tailings dam failures cause severe and persistent environmental damage, yet their long-term ecological impacts remain poorly quantified. On November 5, 2015, the collapse of the Fundão Dam in Mariana, Minas Gerais, Brazil, released over 34 million cubic meters of iron ore tailings, contaminating approximately 650 km of the Rio Doce River system and affecting thousands of hectares of surrounding land. This study evaluates the multi-decadal effects of this disaster on vegetation cover, soil exposure, and surface temperature using a Landsat 5 and 8 time series spanning 1989–2022. We integrated four complementary remote sensing (RS) indicators: the Normalized Difference Vegetation Index (NDVI) to assess vegetation health, Bare Soil Frequency (BSF) to quantify persistent soil exposure, Synthetic Soil Image (SySI) to characterize surface composition, and Land Surface Temperature (LST) to detect thermal anomalies. Analyses across four regions of interest revealed that between 1989 and 2015, forest cover declined by 6.14 %, while mining areas expanded by 7.7 %. Immediately after the disaster, NDVI dropped sharply, BSF and SySI indicated widespread soil exposure, and LST increased by approximately 5 °C. The estimated organic carbon storage potential declined by 635,500 Mg CO₂-eq before the disaster and by an additional 27,700 Mg CO₂-eq immediately afterward. By 2020, vegetation had partially recovered; however, elevated soil exposure and surface temperatures persisted, indicating ongoing ecological stress. These results demonstrate the effectiveness of multi-indicator RS analyses in detecting and monitoring the long-term impacts of environmental disasters. This framework provides a transferable tool for long-term monitoring of mining-related disasters in tropical landscapes.

Keywords
environmental hazard; organic carbon stock; reflectance; Mariana Dam; Fundão Dam

INTRODUCTION

Mining contributes significantly to global economic growth, yet inadequate management can impose substantial environmental risks (Buch et al., 2021). Among these, tailings dam failures have repeatedly caused severe and long-lasting harm to ecosystems and human communities. Tailings dams store fine-grained waste from ore processing, typically in large volumes behind earthen embankments, which makes them susceptible to structural collapse. In Brazil, at least ten catastrophic tailings dam failures have occurred in the past 50 years, most in Minas Gerais, a state that contains more than 60 % of the mineral reserves of Brazil (Tonietto and Silva, 2011; Rotta et al., 2020). Such failures not only transform landscapes but also degrade water quality, threaten biodiversity, and pose serious risks to human health (Armstrong et al., 2019).

On November 5, 2015, the Fundão tailings dam in Mariana, Minas Gerais, failed, releasing more than 34 million cubic meters of iron ore waste into the environment (Ibama, 2015; Silva et al., 2015; Freitas et al., 2016; Governo de Minas Gerais, 2016). The collapse destroyed the villages of Bento Rodrigues and Paracatu de Baixo, resulting in 17 fatalities, and contaminated approximately 650 km of the Rio Doce River system (Armstrong et al., 2019; Ibama, 2015). Tailings deposition covered an estimated 15 km², burying native vegetation, altering river morphology, and introducing toxic elements such as arsenic, cadmium, copper, and lead into soils and sediments (Carvalho et al., 2017; Duarte et al., 2021; Ferreira et al., 2021). These contaminants have created persistent ecological and public health risks, with measurable impacts still evident years after the disaster.

Mining-related disturbances degrade watershed health by altering soil properties, accelerating erosion, fragmenting habitats, and reducing biodiversity (Carvalho et al., 2017; Rotta et al., 2020; Ferreira et al., 2021). The resulting vegetation loss weakens ecosystem resilience and reduces carbon sequestration capacity, thereby hindering efforts to mitigate climate change (Lal, 2005; Mendes et al., 2024). In particular, deforestation and increased soil exposure from mining activities decrease organic carbon storage in both biomass and soils, affecting local and regional climate regulation. Assessing these impacts over extended timescales supports the development of effective restoration strategies and informs sustainable land-use planning in mining-affected regions.

Remote sensing (RS) and geographic information systems (GIS) are indispensable tools for assessing environmental change, especially in data-scarce or physically inaccessible regions (Demattê et al., 2025). Recent studies have diversified the use of remote sensing (RS) techniques to monitor mining impacts, ranging from the analysis of spectral signatures of heavy metals in rare earth areas (Li et al., 2026) to the use of radar sensors (SAR) and soil moisture data to identify dam failure mechanisms, such as those observed in Brumadinho (Thompson et al., 2020; Rotta et al., 2020). Research conducted in global hotspots has employed high-resolution satellite constellations, such as PlanetScope, for the near-real-time detection of Land Use and Land Cover (LULC) changes (Pacheco et al., 2025), as well as predictive models such as CA-Markov to project future degradation in mining landscapes (Obodai et al., 2024). However, most RS-based evaluations of mining impacts concentrate on short-term effects or single environmental indicators, such as vegetation indices (Padmanaban et al., 2017; Rudorff et al., 2018). In the specific context of the Fundão dam collapse, the literature has predominantly focused on short-term vegetation responses and water turbidity (Padmanaban et al., 2017; Rudorff et al., 2018; Rotta et al., 2020). This narrow focus constrains the ability to capture cumulative and interacting processes, thereby limiting understanding of the mechanisms that drive long-term ecological recovery or degradation following tailings dam failures.

This article differs from conventional approaches by its multi-decadal temporal depth (1989–2022) and by the integration of a robust framework that combines four complementary indicators: the Normalized Difference Vegetation Index (NDVI) to assess vegetation vigor (Rouse et al., 1973; Huete et al., 2002), land surface temperature (LST) to evaluate thermal stress (Li et al., 2013), bare soil frequency (BSF) to quantify persistent soil exposure (Demattê et al., 2018, 2025), and Synthetic Soil Images (SySI) to characterize soil properties and lithology (Demattê et al., 2018). Unlike studies relying on isolated metrics, the use of the Geospatial Soil Sensing System (GEOS3) algorithm enables the generation of SySI and the quantification of BSF, thereby distinguishing natural soils from the persistent dispersion of mine tailings. This data represents the proportion of valid observations in which a given pixel is classified as exposed soil over the analyzed time period (Demattê et al., 2018). Furthermore, the study advances by correlating thermal anomalies (LST) with losses in vegetation vigor, revealing ecological feedback mechanisms and enabling an unprecedented quantification of organic carbon deficits, thereby adding a climatic dimension to disaster monitoring.

We hypothesized that tailings deposition from the Mariana disaster caused sustained declines in vegetation health, increased soil exposure, and elevated surface temperatures, with the strongest impacts occurring in the most affected areas. We aimed to (i) quantify pre- and post-disaster changes in vegetation, soil exposure, and surface temperature; (ii) assess the persistence and magnitude of these changes over three decades; and (iii) demonstrate the applicability of long-term, multi-indicator RS analysis for ecological monitoring, restoration planning, organic carbon accounting, and land-use management in mining-affected landscapes. To achieve these goals, we analyzed the long-term environmental legacy of the Fundão dam collapse using an integrated remote sensing approach. Landsat 5 and 8 imagery (1989–2022) was examined across four regions of interest (ROIs), each representing a distinct level of disturbance.

MATERIALS AND METHODS

Study area characterization

This study took place in the municipality of Mariana, Minas Gerais, southeastern Brazil, situated between 43° 11' 51.71" W and 43° 32' 26.50″ W and 20° 15′ 48.81″ S and 20° 16′ 58.76″ S (Figure 1). Mariana lies in the central portion of the state, within the Iron Quadrangle (Quadrilátero Ferrífero), a globally significant mining region with extensive iron ore deposits. The local geology comprises Archean formations, including the Belo Horizonte Complex (gneisses and migmatites) and the Rio das Velhas Supergroup (schists and phyllites), as well as Paleoproterozoic units from the Minas Group. The Cauê Formation (itabirite and hematite) and the Gandarela Formation (dolostones) underpin the economic importance of mineral extraction in the region (Dorr, 1969).

Figure 1
Localization map of the study area with conceptual workflow showing the four regions of interest from the mining company to the stream, with a 1 km buffer.

The soils reflect this complex geological history. Ferralsols and Cambisols dominate the region, exhibiting high contents of iron and aluminum oxides due to prolonged weathering of iron-rich lithologies. These soils typically have low natural fertility and high acidity, requiring careful management for agricultural purposes. According to the Köppen climate classification system, Mariana exhibits Cwa and Cwb climates, characterized by hot, wet summers and mild, dry winters, with mean annual temperatures ranging from 19.5 to 21.8 °C and annual precipitation between 1,300 and 1,600 mm (Alvares et al., 2013). The Atlantic Forest is the predominant biome, interspersed with Cerrado (Brazilian Savanna) and Campo Rupestre (rocky field) vegetation (Alves and Kolbek, 1994; Vasconcelos et al., 2020).

The ROI encompassed the area most severely impacted by the Fundão dam collapse, with a focus on the Bento Rodrigues subdistrict. To delineate the spatial extent of the disaster, we generated a 1 km buffer around the Doce River channel and dam boundaries, encompassing approximately 40,000 ha. This buffer ensured the full inclusion of impacted areas and provided a consistent framework for spatial analysis. Four ROIs split the study field to capture gradients of disturbance intensity. The ROI-1 (23,643.05 ha) includes the mining infrastructure and dam site; ROI 2 (6211.28 ha), immediately downstream, experienced the highest direct impact and received the largest volume of tailings; ROI 3 (5507.87 ha); and ROI 4 (5001.25 ha), located farther downstream, exhibit progressively lower disturbance levels. Observations from field surveys and satellite imagery indicate that areas closer to the dam, particularly along the Gualaxo do Norte River, show more severe environmental degradation. This spatial stratification enables assessment of both localized impacts near the dam and broader downstream effects. Applying identical analytical procedures to each ROI allows quantification of spatial heterogeneity in degradation patterns and ecosystem recovery following the disaster.

Digital image processing

To ensure cross-sensor consistency between Landsat 5 TM and Landsat 8 OLI imagery (LANDSAT/LT05/C02/T1_L2 and LANDSAT/LC08/C02/T1_L2, respectively), a harmonized preprocessing workflow was applied to the entire dataset using the Google Earth Engine (GEE) platform (Gorelick et al., 2017). The analysis relied on Collection 2 Level-2 surface reflectance products, which are radiometrically and atmospherically corrected by the USGS using sensor-specific algorithms (LEDAPS for Landsat 5 and LaSRC for Landsat 8), thereby reducing inter-sensor radiometric discrepancies. Then, atmospheric correction was performed using the standard USGS algorithms embedded in Landsat Collection 2 Level-2 products (LEDAPS for Landsat 5 and LaSRC for Landsat 8).

Cross-sensor consistency was further ensured by selecting spectrally equivalent bands to align TM and OLI wavelength ranges and by applying uniform quality assessment masks. Temporal standardization was achieved by restricting the analysis to the dry season (April–September), minimizing phenological and illumination-related variability across years (Demattê et al., 2018). All preprocessing steps were applied consistently across the time series, ensuring that observed variations in vegetation, soil exposure, and surface temperature primarily reflect environmental changes rather than sensor-related artifacts.

Key years represented distinct stages in the historical trajectory of the disaster: 1989 (pre-mining baseline), 2015 (pre-disaster conditions), 2016 (immediate post-disaster impact), and 2020 (early recovery phase). This temporal framework enables assessment of both the long-term mining effects and the ecological legacy of the Fundão dam collapse. Additionally, annual time-series analyses of NDVI, BSF, SySI, and LST detail the dynamics of these factors over time. This approach captures potential non-linear changes between these milestones.

A standardized image preprocessing workflow was applied to the entire dataset to ensure both temporal and spatial comparability. The workflow included radiometric calibration, atmospheric correction, and cloud and shadow masking using quality assessment (QA) bands, as well as spectral band harmonization between Landsat 5 TM and Landsat 8 OLI sensors. Harmonization minimized spectral discrepancies arising from sensor differences, enabling consistent calculation of spectral indices. To support reproducibility, all GEE scripts, parameter thresholds, and image filtering criteria are available upon request. This level of transparency ensures that other researchers can replicate the processing pipeline and apply it to similar post-disaster monitoring scenarios.

Vegetation analysis

The spectral index NDVI quantified vegetation dynamics, as it is a well-established spectral indicator of vegetation vigor and canopy density (Rouse et al., 1973; Huete et al., 2002). The NDVI leverages the differential reflectance of healthy vegetation, which strongly absorbs visible red light for photosynthesis while reflecting near-infrared radiation. We calculated NDVI for all selected years and for the complete annual time series using equation 1.

Eq. 1 $NDVI=\frac{NIR-RED}{NIR+RED}$

In equation 1, NIR represents reflectance in the near-infrared band, and RED represents reflectance in the red spectral band. We classified NDVI values on a continuous scale from −1 to +1, in which values near -1 indicate non-vegetated surfaces (e.g., water, bare soil, built-up areas) and values near +1 indicate dense, healthy vegetation. We extracted NDVI from Landsat 5 TM and Landsat 8 OLI imagery following radiometric calibration, atmospheric correction, and cloud/shadow masking. To ensure spectral consistency between sensors, we harmonized the Landsat 5 and Landsat 8 bands according to USGS guidelines for cross-sensor time-series analysis, which minimized the risk of sensor-induced biases in multi-decadal comparisons.

Temporal changes in NDVI, alongside BSF, SySI, and LST, revealed concurrent patterns of vegetation decline, increased soil exposure, and surface heating. We analyzed NDVI spatially across the four ROI and temporally for the selected years (1989, 2015, 2016, 2020) to capture both localized and broad-scale vegetation dynamics. To evaluate interannual variability, we generated boxplots to compare NDVI distributions before and after the Fundão dam collapse. These visualizations highlighted median shifts, variability ranges, and outliers, which enabled us to differentiate between short-term shock effects and longer-term recovery trends.

It is essential to note that factors other than vegetation health, such as soil background reflectance, atmospheric conditions, and mixed-pixel effects in heterogeneous landscapes, can influence NDVI. We integrated NDVI with BSF and SySI to reduce the risk of misclassifying bare soil or sparse vegetation as a healthy canopy. Additionally, we employed an annual time-series analysis to minimize the influence of anomalous years, which allowed us to highlight persistent directional trends rather than relying solely on discrete temporal snapshots. This approach enabled us to detect vegetation loss and recovery trajectories, providing a reliable baseline for assessing the ecological legacy of the Fundão dam collapse.

Land use and land cover information were obtained from the MapBiomas Project, a nationally validated, spatially explicit land-use classification system developed for Brazil using Landsat satellite imagery and cloud-based processing in Google Earth Engine. MapBiomas provides annual LULC maps at 30 m spatial resolution and uses a hierarchical classification framework fully compatible with international standards set by the Food and Agriculture Organization (FAO) and the Brazilian Institute of Geography and Statistics (IBGE).

The MapBiomas classification system is structured into three hierarchical levels. At Level 1, six major land cover and land use classes are defined: forest, non-forest natural formation, agriculture, non-vegetated area, water bodies, and non-observed areas. Level 2 refines these categories into 12 more detailed classes, including mining, pasture, agriculture, urban infrastructure, planted forests, and other anthropogenic and natural land-use types relevant to environmental assessments. Level 3 provides further subdivision exclusively for the forest class, distinguishing natural forest formations into forest, savanna, and mangrove ecosystems (Souza Junior et al., 2020).

Bare surface analysis

Assessing BSF is critical for monitoring ecosystem health, particularly in areas affected by mining waste. Exposed soil can indicate contamination from mining residues, such as iron ore. It also often reflects degradation due to erosion and vegetation loss (Novais et al., 2025). We use BSF to provide a quantitative measure of the impacts from mining and other intensive land uses (Demattê et al., 2018). By analyzing bare soil areas, we can assess the effect of mining residues on soil and vegetation dynamics. Continuous monitoring of soil exposure also helps us understand how pollution influences soil cover and overall ecosystem health, especially in the context of climate change and extreme events.

Synthetic soil image

The SySI is a composite reflectance product that aggregates bare surface data from multi-temporal Landsat imagery. We generate it using GEOS3, a data-mining algorithm designed to detect and compile bare soil pixels across time-series satellite archives. Originally developed with Landsat 5 TM collections from 1984 to 2011, GEOS3 identifies bare soil fragments at a 30-meter spatial resolution and integrates them into a spatially continuous surface. This process enables characterizing soil and land-surface properties (Demattê et al., 2018).

SySI comprises six spectral bands: Blue (band 1), Green (band 2), Red (band 3), Near Infrared (NIR, band 4), Shortwave Infrared 1 (SWIR1, band 5), and Shortwave Infrared 2 (SWIR2, band 6). It serves as a robust proxy for lithological, pedological, and ecological variability at biome scales. A rule-based classification isolates bare soil pixels using spectral indices and quality assessment masks. These masks exclude clouds, shadows, water bodies, and vegetated areas, both photosynthetic and non-photosynthetic.

We have validated SySI across diverse soils and land uses, and we consider it a powerful tool for environmental monitoring, degradation assessment, and land-use planning. Its scalability from regional to global applications (Demattê et al., 2025) makes it particularly useful. However, densely forested areas or regions with limited historical imagery can reduce its accuracy. Image quality and seasonal variability also affect detection precision, which requires caution when extrapolating results to dissimilar contexts.

Bare soil frequency

GEOS3 algorithm (Demattê et al., 2018, 2025) systematically registers bare soil occurrences from multi-temporal Landsat imagery, which we derived from BSF (Equation 2). This proxy quantifies the proportion of time each pixel was classified as bare soil over a specific period. For bare soil, we established thresholds of NDVI between -0.25 and 0.25, representing bare soil or urban areas, while negative values correspond to water bodies or non-vegetated surfaces, Normalized Burn Ratio 2 (NBR2) (Equation 3) between -0.30 and 0.10, and the Visible-to-Near-Shortwave Infrared Tendency Index (VNSIR) (Equation 4) <0.9. These spectral constraints filtered out vegetated, burned, and non-soil surfaces, ensuring accurate bare soil detection.

Eq. 2 $BSF=\frac{Numberofbaresoilobservation}{Totalnumberofvalidobservations}\times 100$

Eq. 3 $NBR2=\frac{(SWIR1-SWIR2)}{(SWIR1+SWIR2)}$

Eq. 4 $VNSIR=1-\left[\right(2\times RED-GREEN-BLUE)+3(SWIR2-NIR\left)\right]$

in which: Blue, Green, Red, NIR, SWIR1, and SWIR2 are the harmonized spectral bands from Landsat 4 TM, 5 TM, 7 ETM+, and 8 OLI, respectively.

Land Surface Temperature analysis

We derived Land Surface Temperature (LST) from the thermal bands of Landsat 5 TM and Landsat 8 TIRS imagery after radiometric calibration and atmospheric correction. Land Surface Temperature added a thermal dimension to our analysis, enabling us to detect heat anomalies associated with vegetation loss and bare-soil expansion (Li et al., 2013). We focused our analysis on the dry season (April–October) to reduce the influence of seasonal moisture variability.

Pixels with LST values overlaid pixels classified as bare soil in BSF and SySI served to examine the relationship between soil exposure and surface heating. This approach allowed us to assess whether areas of persistent bare soil consistently exhibited higher surface temperatures, which could further inhibit vegetation regeneration. By integrating BSF, SySI, and LST, we developed a multi-dimensional perspective on soil exposure, composition, and thermal stress. This method enhances our ability to detect persistent degradation patterns, attribute them to mining-related disturbances, and evaluate ecological recovery implications.

Data analysis

We analyzed the spatial and temporal patterns of environmental degradation by integrating NDVI, BSF, SySI, and LST (Table 1) using datasets from each of the four ROI. Our analytical framework aimed to identify both immediate and long-term changes associated with the Fundão dam collapse while distinguishing them from background variability.

Table 1
Summary of remote sensing-based techniques utilized

Index	What does it do?	Reference
NDVI	Quantifies vegetation greenness and serves as a proxy for vegetation vigor.	Rouse et al. (1973)
SySI (GEOS3)	Provides a representation of BS and can also be related to lithological and pedological variability, carbon-rich pools, and biome changes	Demattê et al. (2018)
BSF	Evaluates in percentage the bare surface frequency	Demattê et al. (2018; 2025)
LST	Provides an estimated temperature value in the land surface	Li et al. (2013)

• Source: the authors.

First, we developed a multi-year time series for NDVI and LST covering 1989 to 2020. To identify long-term trends and abrupt post-disaster changes, we extracted annual mean values for each ROI and visualized them with line graphs. Boxplots comparing pre- and post-disaster distributions complemented these temporal analyses. This allowed us to assess shifts in medians, variability, and extreme values. The visual supports provide a robust basis for distinguishing short-term anomalies from persistent changes in vegetation cover and surface temperature.

Second, Pearson correlation coefficient analyses identified possible statistical relationships among BSF, NDVI, and LST. This strategy helps to quantify the strength and direction of associations between soil exposure, vegetation health, and thermal anomalies. Positive correlations between BSF and LST indicate that increased soil exposure is associated with elevated surface temperature. In contrast, negative correlations between NDVI and both BSF and LST demonstrate that vegetation loss is linked to higher soil exposure and thermal stress.

Third, we conducted spatial overlay analyses to identify co-occurrence patterns among the indicators. We mapped areas where low NDVI coincided with high BSF and elevated LST to delineate degradation hotspots, which could serve as priority sites for restoration. We compared these spatial patterns across the four ROI to assess how proximity to the dam and river connectivity influenced the severity of environmental degradation.

We maintained consistency across all ROI and years to ensure comparability throughout the analysis. We performed all geospatial and statistical operations in GEE and R (version 2024.12.1+563) using standardized, reproducible workflows. By integrating temporal trend analysis, statistical correlations, and spatial pattern detection, we developed a comprehensive, reproducible approach to evaluate the persistence, magnitude, and spatial variability of environmental impacts following large-scale mining disasters.

It is important to note that this framework is exploratory and does not establish causality. For future assessments, we recommend incorporating multivariate regression models and spatially explicit approaches, such as geographically weighted regression (GWR) or time-series decomposition. These techniques can account for confounding factors, such as topography, rainfall, and non-mining land use, thereby enhancing causal interpretation of observed patterns.

RESULTS AND DISCUSSION

Variation in land use and land cover over time

Long-term RS analysis revealed substantial shifts in LULC within the study area, reflecting both the gradual influence of mining activities and the acute disturbance caused by the Fundão dam collapse. The multi-decadal dataset captured gradual changes in forest and non-forest classes, as well as abrupt alterations associated with tailings deposition (Figures 2a, 2b and 2c). Between 1989 and 2015, native forest cover declined from 60.3 to 54.2 %, corresponding to a loss of approximately 2,468 ha. This trend continued through 2020, resulting in a net reduction of 6.48 % (~2,615 ha) relative to the 1989 baseline, as also observed by Omachi et al. (2018). These findings are consistent with documented deforestation patterns in the Iron Quadrangle driven by the expansion of mining activities and infrastructure development (Rotta et al., 2020; Duarte et al., 2021). Over the same period, mining areas increased by 7.72 % (~3,115 ha), confirming intensified mineral extraction in the region (Ferreira et al., 2021).

Figure 2
Land use and land cover (LULC) dynamics within the study area were derived from the MapBiomas classification system. (a) Spatial distribution of LULC classes for the reference years 1989 and 2020, representing baseline landscape conditions before large-scale mining activities and the cumulative effects observed after more than three decades of disturbance, including the Fundão dam collapse; (b) Annual temporal evolution of major LULC classes from 1985 to 2022, depicting long-term trends in forest cover, mining areas, and other land-use classes throughout the pre-mining, post-disaster, and recovery phases; and (c) detailed annual trajectory of LULC classes indicating periods of abrupt change associated with the expansion of mining activities and the dam failure, as well as subsequent stabilization and recovery processes.

Other LULC classes also declined, including pasture (−1.41 %; ~570 ha), grassland (−0.57 %; ~233 ha), and planted forests (−0.50 %; ~201 ha). These reductions reflect both direct land conversion for mining operations and indirect effects from tailings deposition in disaster-affected areas. Such land transformations alter habitat structure, reduce ecological connectivity, and disrupt local ecosystem services (Mendes et al., 2024).

The 2015 Fundão dam collapse represented a sharp inflection point in LULC dynamics. Post-disaster imagery from 2016 revealed extensive deforestation, increased soil exposure, and substantial geomorphological changes along the Rio Doce Basin. These patterns reflect both the immediate destruction caused by the tailings wave and the subsequent environmental stress resulting from altered soil conditions and contaminant deposition (Carvalho et al., 2017; Duarte et al., 2021).

Annual time-series data (Figure 2c) show an abrupt decline in vegetation cover immediately after 2015, followed by partial regrowth in subsequent years. Recovery remains spatially heterogeneous, with persistent degradation in areas of high tailings accumulation, where unfavorable soil and hydrological conditions limit natural recolonization (Duarte et al., 2021; Mendes et al., 2024). Overall, the LULC assessment highlights a dual-impact scenario: long-term deforestation driven by the expansion of mining activities, compounded by sudden, large-scale habitat loss from the Fundão dam collapse. Together, these processes have caused substantial structural changes in the landscape, reduced ecological resilience, and increased the complexity of restoration planning in the region (Mendes et al., 2024).

Detection of tailings spread across the affected areas by orbital reflectance imagery

The combined use of BSF and SySI allowed detection and quantification of the spatial distribution of tailings following the Fundão dam collapse (Figures 3a, 3b, 3c, 3d, 3e and 3f). This integration provided both spatial and compositional information, enabling differentiation between tailings deposits and naturally exposed soils. Since the mining implementation in 1989 (Figure 3a) to the post-disaster analysis for 2016 and beyond, the SySI has revealed a distinct spectral signature corresponding to deposited tailings (Figures 3b, 3c and 3d).

Figure 3
Spatial and temporal dynamics of Bare Soil Frequency (BSF) within the study area. (a, b, c and d) Spatial distribution of exposed soil for the reference years 1989, 2015, 2016, and 2020, illustrating progressive soil exposure associated with mine installation, the immediate impact of the Fundão dam collapse, and subsequent partial recovery. The selected years represent distinct ecological states rather than evenly spaced time intervals. For each year, the total exposed soil area (ha) and its relative contribution (%) to the study area are indicated; (e) Complete annual BSF time series (1985–2022) expressed as total exposed soil area, illustrating key phases of landscape transformation, including pre-mining conditions, mine installation, expansion, peak soil exposure, and post-collapse regeneration; and (f) annual bare soil frequency (%) for the study area, showing long-term trends in persistent soil exposure, with dashed lines marking the onset of mining activities and the dam collapse.

Figures 3e and 3f show the dynamics of bare soil in the study area. High BSF values, together with SySI-derived reflectance characteristics, delineated extensive sediment coverage along the Gualaxo do Norte River and adjacent floodplains. The highest concentrations occurred in ROI 2, immediately downstream of the dam, where BSF exceeded 80 % in several contiguous areas. The ROI 1, encompassing the dam infrastructure, also exhibited high BSF values, reflecting both pre-existing mining surfaces and newly deposited tailings. The ROI 3 and 4 had progressively lower BSF values, although evidence of tailings dispersion persisted along the creek.

SySI outputs also confirmed alterations in soil spectral properties across all ROI, particularly in areas with high BSF (Figures 3a, 3b, 3c, and 3d). These changes reflect modifications in surface mineralogy and particle-size distribution, consistent with the deposition of fine-grained tailings enriched in metals such as arsenic, cadmium, and copper (Carvalho et al., 2017). Such compositional shifts have important implications for post-disaster recovery, as they can affect contaminant mobility, water–soil interactions, and vegetation establishment in affected areas (Duarte et al., 2021). Annual SySI composites are shown in figure 3e, indicating a peak in bare soil in 2016, which is also observed in BSF (Figure 3f).

Land surface temperature outcomes

Figure 4 presents the spatial and temporal dynamics of vegetation vigor and surface temperature associated with the Fundão dam collapse. Figures 4a and 4b show the spatial distribution of the NDVI in 2015 (pre-collapse) and 2016 (immediate post-collapse), respectively, highlighting abrupt vegetation loss in areas directly affected by tailings deposition. Figure 4c summarizes these changes through NDVI boxplots for 2015 and 2016, evidencing a marked reduction in vegetation vigor after the disaster, with the dashed line indicating the bare soil threshold. Subfigure (d) presents boxplots of LST for the same years, revealing a pronounced increase in surface temperatures following the collapse. To place these short-term impacts in a broader context, figure 4e illustrates the annual NDVI time series (1985–2022) for the entire study area, with annotations marking the onset of mining activities in 1989 and the dam failure in 2016. Finally, figure 4f depicts the annual LST time series (1985–2022) across the four ROI, demonstrating progressive warming trends that intensified after the expansion of mining activities and were further exacerbated by the dam collapse.

Figure 4
Spatial and temporal dynamics of vegetation vigor and surface temperature associated with the Fundão dam collapse. (a–b) Spatial distribution of the Normalized Difference Vegetation Index (NDVI) in 2015 (pre-collapse) and 2016 (immediate post-collapse), indicating abrupt vegetation loss in areas affected by tailings deposition; (c) Boxplots of NDVI values for 2015 and 2016, illustrating the sharp reduction in vegetation vigor following the disaster; the dashed line indicates the bare soil threshold; (d) Boxplots of Land Surface Temperature (LST) for 2015 and 2016, showing a marked increase in surface thermal conditions after the collapse; (e) Annual NDVI time series (1985–2022) for the study area, indicating long-term vegetation dynamics, with annotations marking the onset of mining activities (1989) and the dam failure (2016); and (f) Annual LST time series (1985–2022) across the four ROI, evidencing progressive warming trends intensified after the expansion of mining activities and the dam collapse.

A focused post-collapse analysis, by overlaying BSF maps with LST data, revealed a strong association between persistent soil exposure and elevated surface temperatures, particularly in ROI 1 and 2 (Figures 5a and 5b). These elevated temperatures likely result from the low albedo and reduced moisture retention of tailings deposits, which increase heat absorption and exacerbate thermal stress on adjacent vegetation (Li et al., 2013). Such conditions create microclimatic barriers to vegetation regrowth, contributing to the persistence of degraded zones in the years following the disaster. Overall, integrating BSF, SySI, and LST provided a robust and reproducible framework for mapping tailings dispersion and characterizing its environmental footprint. This methodology identified the most severely affected areas and highlighted priority targets for remediation, where persistent bare surfaces and altered thermal regimes may delay ecological recovery (Duarte et al., 2021).

Figure 5
Temporal evolution of Land Surface Temperature (LST) for selected post-collapse years, evidencing thermal anomalies associated with soil exposure and vegetation loss. (a) mean LST image under SySI, highlighting the four points in the study area (the purple areas indicate soils, in a composition 543 RGB Landsat); and (b) line chart with mean LST value for each point. See the full annual LST time series from 1985 to 2022 in figure 4f.

The role of remote sensing in environmental monitoring

This study highlights the pivotal role of RS in long-term environmental monitoring, particularly in post-disaster landscapes (Novais et al., 2025). By providing high-frequency, spatially explicit observations, this approach enables a detailed assessment of land degradation and ecological recovery following industrial accidents. In the Mariana disaster, satellite-based analyses supported documentation of the spatial and temporal extent of environmental damage, tracking recovery dynamics, and informing mitigation and rehabilitation strategies.

The RS techniques are effective for identifying, delineating, and monitoring pollution sources and affected areas, including abandoned lands, land-use changes, and alterations in water bodies (Miller et al., 2023). These capabilities support evaluation of environmental and socioeconomic risks, reduce disaster impacts, and reveal industry-wide LULC patterns (Buch et al., 2021). Earth observation data also provide a crucial basis for comparing mining impacts across commodities, geographies, and operational contexts (Werner et al., 2019; Chaddad et al., 2022; Nasser et al., 2024). Nevertheless, studies that integrate multiple RS-derived indicators remain scarce in assessments of areas affected by mining tailings.

Our integration of NDVI, BSF, LST, and SySI provided a robust framework for detecting and monitoring ecological impacts. The NDVI served as a sensitive indicator of vegetation health and vigor, detecting stress from canopy loss or alteration. The BSF quantified the extent and persistence of bare soil, revealing areas degraded by vegetation removal and substrate exposure. The LST captured localized thermal changes that exacerbate environmental stress and influence ecological processes, while SySI offered complementary information on soil characteristics, enhancing understanding of pedological alterations.

Our correlation analysis (Table 2) revealed meaningful relationships among the metrics. The BSF correlated moderately and positively with LST, which indicates that increased soil exposure is associated with higher surface temperatures, likely due to enhanced solar absorption. Conversely, BSF showed a negative correlation with NDVI, which confirms that areas with frequent soil exposure tend to have reduced vegetation vigor. Elevated LST values also correlated with lower NDVI, suggesting that thermal stress inhibits vegetation growth, as observed by Yue et al. (2007). Our findings align with previous research. Palma et al. (2024) reported that combining NDVI with proximal sensors, such as temperature and CO₂ measurements, improves our understanding of soil dynamics and ecological impacts in mining-affected areas. Similarly, Silva Junior et al. (2018) and Rudorff et al. (2018) documented sharp post-disaster declines in vegetation indices within the Rio Doce Basin.

Table 2
Correlation matrix analysis of Bare Soil Frequency (BSF), Land Surface Temperature (LST), and Normalized Difference Vegetation Index (NDVI). Correlation coefficients were calculated using annual mean values (1989–2022)

	LST	BSF	NDVI
LST	1.00	0.37	-0.10
BSF	0.37	1.00	-0.56
NDVI	-0.10	-0.56	1.00

• Source: the authors.

Integrating vegetation, thermal, and soil indicators provides a holistic perspective on post-disaster environmental processes, enhancing monitoring capabilities in ecologically complex regions (Mello et al., 2025). This multi-indicator RS framework allows spatial identification of impacted areas, temporal tracking of degradation and recovery, and evaluation of mitigation efforts, positioning it as a strategic tool for the sustainable management of mining-affected landscapes.

Impact of the mining tailings dam collapse

The collapse of mining tailings dams produces profound and multifaceted socioeconomic and environmental impacts (Santos et al., 2019). Economically, they destroy infrastructure, reduce land value, and disrupt local livelihoods such as agriculture and fisheries, while remediation and cleanup require substantial financial investment. Socially, these disasters displace communities, degrade living conditions, and increase health risks, particularly respiratory and dermatological diseases associated with contaminated water and soil.

The environmental consequences are severe and persistent. Carvalho et al. (2017) reported that copper concentrations in the surface waters of the Doce River following the disaster were 86 times the maximum permissible limit established by Brazilian regulations. Tailings deposition altered river hydromorphology, increased turbidity, and impaired aquatic biodiversity and ecosystem services (Rudorff et al., 2018). The long-term patterns documented in this study confirm that mining tailings can cause persistent ecological degradation, often undetectable immediately after dam failures but revealed through sustained RS monitoring.

Our findings align with Rotta et al. (2020), who investigated the 2019 Brumadinho collapse and identified seepage erosion and internal liquefaction as key failure mechanisms. Using satellite-derived soil moisture and SAR data, they observed progressive saturation, vertical subsidence of up to 0.30 m, and persistent surface water accumulation that weakened the dam before failure. While Rotta et al. (2020) focused on early-warning indicators and short-term impacts, our analysis highlights the cumulative legacy of tailings deposition over three decades, particularly the long-term degradation of vegetation health, increased soil exposure, elevated surface temperatures, and substantial depletion of organic carbon stocks. Together, these complementary perspectives underscore the value of integrating predictive and retrospective satellite-based approaches for comprehensive disaster assessment and recovery planning.

Thompson et al. (2020) examined the short-term effects of the Brumadinho collapse using biogeochemical, microbiological, and ecotoxicological data collected from the Paraopeba River one week and four months post-disaster. They reported water quality levels up to 30 times higher than permitted by Brazilian standards (Conama, 2005). In comparison, the Mariana disaster had a greater spatial extent and downstream reach, producing broader ecological consequences for the Doce River Basin (Ibama, 2015; Santos et al., 2019).

Contamination studies further underscore the severity of these impacts. Duarte et al. (2021) reported elevated arsenic (17 mg kg^-1) and cadmium contents in the affected region, posing significant ecological risks. Using X-ray fluorescence, Ferreira et al. (2021) found total concentrations of potentially toxic elements, including chromium and lead, were substantially higher in soils impacted by the collapse than in adjacent areas. These findings reinforce the value of RS-based indicators for detecting and monitoring persistent post-disaster contamination, supporting earlier observations by Ye (2022) in mining-impacted landscapes. Collectively, these results demonstrate that mining tailings spills cause widespread destruction, loss of life, and long-lasting environmental contamination. Effective mitigation requires stringent regulatory enforcement, investment in safer mining technologies, and comprehensive recovery plans that integrate environmental rehabilitation with long-term monitoring.

Influence of mining on forest degradation

Between 1989 and 2020, the conversion of 2,615 ha of Atlantic Forest to mining areas resulted in a substantial reduction in carbon sequestration capacity. Omachi et al. (2018) observed a massive loss of Atlantic Forest following the collapse of the Fundão Dam tailing dam in Mariana, Brazil. Tropical forests, as noted by Lal (2005), are critical components of the global carbon cycle, storing approximately 123 Mg ha^-¹ of C in soils and 121 Mg ha^-¹ of C in biomass, for a total of about 243 Mg ha^-¹ of C. Based on these estimates, our analysis indicates a lost sequestration potential of roughly 635,500 Mg CO₂-eq (Figure 6a), highlighting the severe ecological cost of forest removal for the expansion of mining activities. This loss represents both a setback for climate change mitigation and a long-term depletion of vital natural carbon reservoirs.

Figure 6
Estimated storage carbon capacity by soil and forest. (a) Organic Carbon deficit immediately after the disaster. (b) Comparison of Carbon Deficit and Projected Organic Carbon Stock during all mining activity.

Researchers estimated that organic carbon stock could have reached 794,300 Mg CO₂-eq, a 25 % increase, had deforestation been avoided, emphasizing the compounded effects of the expansion of mining activities and the Fundão dam collapse. Preserved forest cover could have supported an estimated 800.000 Mg CO₂-eq of carbon storage, contributing significantly to local and regional climate regulation (Ibama, 2015). The disaster further intensified CO₂ emissions, degraded ecosystem services, and disrupted ecological balance, reducing the resilience of both natural systems and human communities. These cascading effects underscore the urgency of implementing sustainable land management strategies to preserve and restore carbon-rich habitats.

According to our estimates, this vegetation had sequestered approximately 27,700 Mg CO₂-eq from organic carbon before being buried under tailings. The deposition abruptly halted its role as a carbon sink, disrupting the local carbon cycle and eliminating a critical buffer against atmospheric CO₂ accumulation. Such sudden losses, compounded by longer-term deforestation trends, illustrate the scale at which mining and associated disasters can compromise carbon storage, exacerbate climate impacts, and threaten biodiversity.

Limitations and opportunities

This study primarily relies on RS data and secondary sources to estimate ecological variables, including organic carbon stocks, vegetation health, and soil exposure. Certain methodological constraints highlight opportunities for future research. For example, the absence of ground-based validation, largely due to the historical scope of the analysis, limits the ability to verify satellite-derived proxies in situ. We assumed fixed per-hectare carbon stock values and generalized vegetation regrowth rates to maintain consistency across the multi-decadal series. These assumptions simplify the forest’s structural heterogeneity and ecological complexity.

Organic carbon loss calculations were derived from per-hectare coefficients reported for tropical forests (Lal, 2005; Mendes et al., 2024). While biome-appropriate for large-scale assessments, these values do not fully capture spatial variability in forest structure, degradation intensity, or biomass density over time (Omachi et al., 2018). Consequently, the organic carbon loss estimates presented here (e.g., Figure 6) should be interpreted as approximate averages rather than exact measurements. Based on reported variability in tropical and subtropical forests, we estimate an uncertainty of approximately ±15 %, reflecting differences in canopy composition, successional stage, and disturbance history. This uncertainty does not alter the conclusion that tailings deposition caused substantial carbon depletion but underscores the importance of integrating region-specific allometric equations and field measurements in future post-disaster carbon accounting.

The use of correlation analysis to examine relationships among BSF, LST, and NDVI provided useful exploratory insights but does not imply causation. Environmental drivers such as topography, precipitation, and non-mining land use changes may also influence these patterns (Rudorff et al., 2018; Silva Junior et al., 2018; Palma et al., 2024). Advanced methods, including multivariate regression, geographically weighted regression (GWR), and time-series decomposition (not applied in this study), are promising tools for improving causal attribution.

Although these relationships are consistent with mining impacts, other environmental factors such as topography, precipitation, and unrelated LULC changes may also influence the observed trends. Our current exploratory approach does not fully isolate mining-specific effects. Future studies should incorporate multivariate regression, time-series decomposition, or spatial modeling (e.g., geographically weighted regression) to control for confounding variables, thereby strengthening causal attribution. Including covariates such as rainfall, slope, and elevation could improve discrimination between mining-induced degradation and natural variability.

Limitations related to spatial resolution and atmospheric correction must also be considered. The 30 m resolution of Landsat imagery may not capture fine-scale disturbances, such as narrow riparian zones or small regeneration patches. Additionally, atmospheric correction using DOS1, while computationally efficient and consistent across the multi-decadal Landsat archive, does not fully account for aerosol variability or bidirectional reflectance effects. Although previous studies and limited scene-based comparisons suggest that DOS1 performs reliably for detecting relative changes, minor biases or dampening of subtle trends cannot be ruled out. Advanced algorithms such as LaSRC, which integrate MODIS aerosol retrievals and radiative transfer modeling, may further reduce atmospheric artifacts over long time series (Vermote et al., 2016). Comparative analyses using LaSRC for Landsat 8 and LEDAPS for Landsat 5 could help evaluate the sensitivity of LST, NDVI, BSF, and SySI to the choice of atmospheric correction. Finally, the inclusion of ground-based validation, currently absent, would enhance the accuracy of interpretations and improve methodological transparency.

Despite these limitations, the RS-based framework provides robust long-term insights into the spatial and temporal dynamics of forest degradation and carbon loss associated with both the expansion of mining activities and the Fundão Dam collapse. Future research should prioritize integrating high-resolution imagery, ground-truth ecological data, and advanced spatial-statistical models to refine indicator sensitivity and improve uncertainty quantification. Expanding the framework to include additional biophysical parameters, such as soil moisture, evapotranspiration, and surface albedo, would further enhance its diagnostic capacity. By combining assessments with predictive modeling, environmental agencies can improve early-warning systems, optimize restoration planning, and strengthen climate adaptation strategies in mining-impacted regions.

CONCLUSIONS

This study confirms that RS time series are essential for diagnosing and monitoring the long-term environmental impacts of industrial disasters. Using Normalized Difference Vegetation Index (NDVI), Land Surface Temperature (LST), Bare Soil Frequency (BSF), and Synthetic Soil Image (SySI) indices, we detected severe and persistent changes in soil and vegetation dynamics following the 2015 Fundão Dam collapse in Mariana, Brazil. Between 1989 and 2015, forest cover in the study area declined from 60.34 to 54.22 %, corresponding to a loss of approximately 2,468 ha, while mining areas expanded by 3,115 ha. This land-use transformation resulted in an estimated loss of 635,500 Mg CO₂-eq of potential carbon sequestration. Our observations demonstrate that an additional 27,700 Mg CO₂-eq of carbon may have been lost due to deforestation and sediment burial in the immediate aftermath of the collapse.

Although NDVI trends after 2016 suggest gradual vegetation recovery, the continued presence of exposed soils, as indicated by BSF and SySI, reflects lasting ecological stress. The spatial footprint of tailings and elevated surface temperatures highlights the persistent challenges to environmental recovery in mining-impacted landscapes. These findings demonstrate the capacity of long-term RS to characterize enduring environmental degradation and underscore its strategic value for restoration planning and organic carbon stock assessment.

The methodological framework presented here is robust, scalable, and suitable for integration into institutional ecological monitoring programs. Nonetheless, several limitations must be acknowledged, including the lack of ground-truth validation, the relatively coarse spatial resolution of Landsat sensors, and reliance on optical imagery, which is susceptible to cloud cover. Future research should address these constraints by incorporating higher-resolution datasets and systematic in situ measurements to calibrate and validate RS-derived indices, thereby improving the accuracy of estimates of soil and vegetation degradation.

Overall, this study delivers a transferable model for assessing environmental degradation in disaster-affected landscapes and provides clear evidence that long-term RS can reliably diagnose persistent ecosystem alterations, even when constrained by the limitations inherent to available data.

ACKNOWLEDGMENTS

We are thankful for the technical support provided by the members of the Geotechnologies in Soil Science group (GeoCiS, https://esalqgeocis.wixsite.com/geocis), which is greatly appreciated. We also extend our gratitude to Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) for the scholarship awarded to the second author (project 2024/06285–1).

How to cite:

Novais JJM, Mello DC, Veloso GV, Melo BMD, Santos UJ, Rosas JTF, Braga AC, Bartsch BA, Sousa GPB, Amorim MTA, Makino FY, Rosin NA, Rodríguez-Albarracín HS, Palma HP, Francelino MR, Falcioni R, Nanni MR, Couto EG, Demattê JAM. Long-term soil, vegetation, and surface temperature dynamics in mining tailings–affected landscapes revealed by multitemporal remote sensing. Rev Bras Cienc Solo. 2026;50:e0250204. https://doi.org/10.36783/18069657rbcs20250204

DECLARATION OF GENERATIVE AI AND AI-ASSISTED TECHNOLOGIES IN THE MANUSCRIPT PREPARATION PROCESS

The authors declare that, during the preparation of this study, they employed AI-based methods to revise the grammar and structure of the English text, enhancing clarity, coherence, and cohesion. After using these tools, the authors reviewed and edited the content as necessary and assumed full responsibility for the accuracy and integrity of the publication.

FUNDING

The authors thank the University of São Paulo for providing financial support through project numbers 2022/14935-0 and 2021/05129-8. Gratitude is also extended to the National Council for Scientific and Technological Development (CNPq) and the Coordination for the Improvement of Higher Education Personnel (CAPES) for funding this research.

DATA AVAILABILITY

The archives generated in this study are available on request directly from the first author, the corresponding author, or by emailing the Geotechnologies in Soil Science ([e-mail]).

REFERENCES

• Alvares CA, Stape JL, Sentelhas PC, Gonçalvez JD, Sparovek G. Köppen’s climate classification map for Brazil. Meteorol Z. 2013;22:711-28. https://doi.org/10.1127/0941-2948/2013/0507
  » https://doi.org/10.1127/0941-2948/2013/0507
• Alves RJV, Kolbek J. Plant species endemism in savanna vegetation on table mountains (Campo Rupestre) in Brazil. Vegetatio. 1994;113:125-39. https://doi.org/10.1007/BF00044230
  » https://doi.org/10.1007/BF00044230
• Armstrong M, Petter R, Petter C. Why have so many tailings dams failed in recent years? Resour Policy. 2019;63:101412. https://doi.org/10.1016/j.resourpol.2019.101412
  » https://doi.org/10.1016/j.resourpol.2019.101412
• Buch AC, Niemeyer JC, Marques ED, Silva-Filho EV. Ecological risk assessment of trace metals in soils affected by mine tailings. J Hazard Mater. 2021;403:123852. https://doi.org/10.1016/j.jhazmat.2020.123852
  » https://doi.org/10.1016/j.jhazmat.2020.123852
• Carvalho MS, Moreira RM, Ribeiro KD, Almeida AM. Concentração de metais no rio Doce em Mariana, Minas Gerais, Brasil. Acta Bras. 2017;1:37-41. https://doi.org/10.22571/Actabra13201758
  » https://doi.org/10.22571/Actabra13201758
• Chaddad F, Mello FAO, Tayebi M, Safanelli JL, Campos LR, Amorim MTA, Sousa GPB, Ferreira TO, Ruiz F, Perlatti F, Greschuk LT, Rosin NA, Rosas JTF, Demattê JAM. Impact of mining-induced deforestation on soil surface temperature and carbon stocks: A case study using remote sensing in the Amazon rainforest. J South Am Earth Sci. 2022;119:103983. https://doi.org/10.1016/j.jsames.2022.103983
  » https://doi.org/10.1016/j.jsames.2022.103983
• Conselho Nacional do Meio Ambiente - Conama. Resolução 357, de 17 de março de 2005: Dispõe sobre a classificação dos corpos de água e diretrizes ambientais para o seu enquadramento, bem como estabelece as condições e padrões de lançamento de efluentes, e dá outras providências.. Brasília, DF: Conama; 2005 [cited 2025 Nov 20]. Available from: https://www.gov.br/ibama/pt-br/assuntos/fiscalizacao-e-protecao-ambiental/emergencias-ambientais/manchasdeoleo/arquivos/2023/res_conama_357_2005_classificacao_corpos_agua_rtfcda_altrd_res_393_2007_397_2008_410_2009_430_2011.pdf
  » https://www.gov.br/ibama/pt-br/assuntos/fiscalizacao-e-protecao-ambiental/emergencias-ambientais/manchasdeoleo/arquivos/2023/res_conama_357_2005_classificacao_corpos_agua_rtfcda_altrd_res_393_2007_397_2008_410_2009_430_2011.pdf
• Demattê JAM, Fongaro CT, Rizzo R, Safanelli JL. Geospatial Soil Sensing System (GEOS3): A powerful data mining procedure to retrieve soil spectral reflectance from satellite images. Remote Sens Environ. 2018;212:161-75. https://doi.org/10.1016/j.rse.2018.04.047
  » https://doi.org/10.1016/j.rse.2018.04.047
• Demattê JAM, Rizzo R, Rosin NA, Poppiel RR, Novais JJM, Amorim MTA, Rodriguez-Albarracín HS, Rosas JTF, Bartsch BDA, Vogel LG, Minasny B, Grunwald S, Ge Y, Ben-Dor E, Gholizadeh A, Gomez C, Chabrillat S, Francos N, Fiantis D, Belal A, Tsakiridis N, Kalopesa E, Naimi S, Ayoubi S, Tziolas N, Das BS, Zalidis G, Francelino MR, Mello DC, Hafshejani NA, Peng Y, Ma Y, Coblinski JA, Wadoux AMJC, Savin I, Malone BP, Karyotis K, Milewski R, Vaudour E, Wang C, Salama ESM, Shepherd KD. A global soil spectral grid based on space sensing. Sci Total Environ. 2025;968:178791. https://doi.org/10.1016/j.scitotenv.2025.178791
  » https://doi.org/10.1016/j.scitotenv.2025.178791
• Dorr JVN. Physiographic, stratigraphic, and structural development of the Quadrilátero Ferrífero, Minas Gerais, Brazil. Washington, DC: U.S. Geological Survey; 1969. (Professional Paper 641-A). https://doi.org/10.3133/pp641A
  » https://doi.org/10.3133/pp641A
• Duarte EB, Neves MA, Oliveira FB, Martins ME, Oliveira CHR, Burak DL, Orlando MTD, Rangel CVGT. Trace metals in Rio Doce sediments before and after the collapse of the Fundão iron ore tailing dam, Southeastern Brazil. Chemosphere. 2021;262:127879. https://doi.org/10.1016/j.chemosphere.2020.127879
  » https://doi.org/10.1016/j.chemosphere.2020.127879
• Ferreira GWD, Ribeiro BT, Weindorf DC, Curi N, Guilherme LRG. Assessment of iron-rich tailings via portable X-ray fluorescence spectrometry: The Mariana dam disaster, southeast Brazil. Environ Monit Assess. 2021;193:203. https://doi.org/10.1007/s10661-021-08982-7
  » https://doi.org/10.1007/s10661-021-08982-7
• Freitas CM, Silva MA, Menezes FC. O desastre na barragem de mineração da Samarco – fratura exposta dos limites do Brasil na redução de risco de desastres. Cienc Cult. 2016;68:25-30. https://doi.org/10.21800/2317-66602016000300010
  » https://doi.org/10.21800/2317-66602016000300010
• Gorelick N, Hancher M, Dixon M, Ilyushchenko S, Thau D, Moore R. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sens Environ. 2017;202:18-27. https://doi.org/10.1016/j.rse.2017.06.031
  » https://doi.org/10.1016/j.rse.2017.06.031
• Governo de Minas Gerais. Relatório: Avaliação dos efeitos e desdobramentos do rompimento da Barragem de Fundão em Mariana-MG. Belo Horizonte: Secretaria de Estado de Desenvolvimento Regional, Política Urbana e Gestão Metropolitana; 2016. Available from: https://www.agenciaminas.mg.gov.br/ckeditor_assets/attachments/770/relatorio_final_ft_03_02_2016_15h5min.pdf
  » https://www.agenciaminas.mg.gov.br/ckeditor_assets/attachments/770/relatorio_final_ft_03_02_2016_15h5min.pdf
• Huete A, Didan K, Miura T, Rodriguez EP, Gao X, Ferreira LG. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens Environ. 2002;83:195-213. https://doi.org/10.1016/S0034-4257(02)00096-2
  » https://doi.org/10.1016/S0034-4257(02)00096-2
• Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis - Ibama. Laudo técnico preliminar: Impactos ambientais decorrentes do desastre envolvendo o rompimento da barragem de Fundão, em Mariana, Minas Gerais. Brasília, DF: Ibama; 2015. Available from: https://www.ibama.gov.br/phocadownload/barragemdefundao/laudos/laudo_tecnico_preliminar_Ibama.pdf
  » https://www.ibama.gov.br/phocadownload/barragemdefundao/laudos/laudo_tecnico_preliminar_Ibama.pdf
• Lal R. Forest soils and carbon sequestration. For Ecol Manag. 2005;220:242-58. https://doi.org/10.1016/j.foreco.2005.08.015
  » https://doi.org/10.1016/j.foreco.2005.08.015
• Li Z-L, Tang B-H, Wu H, Ren H, Yan G, Wan Z, Trigo IF, Sobrino JA. Satellite-derived land surface temperature: Current status and perspectives. Remote Sens Environ. 2013;131:14-37. https://doi.org/10.1016/j.rse.2012.12.008
  » https://doi.org/10.1016/j.rse.2012.12.008
• Li S, Dan Z, Yin T. Mining-induced land use change and ecological restoration: lessons from three developing economies. Phys Chem Earth Pt A/B/C. 2026;142:104267. https://doi.org/10.1016/j.pce.2025.104267
  » https://doi.org/10.1016/j.pce.2025.104267
• Mello DC, Francelino MR, Moquedace CM, Baldi CGO, Silva LV, Siqueira RG, Veloso GV, Fernandes-Filho EI, Thomazini A, Demattê JAM, Ferreira TO, Gomes LC, Senra EO, Schaefer CEGR. Global warming may turn ice-free areas of Maritime and Peninsular Antarctica into potential soil organic carbon sinks. Commun Earth Environ. 2025;6:143. https://doi.org/10.1038/s43247-024-01937-z
  » https://doi.org/10.1038/s43247-024-01937-z
• Mendes RG, Valle Junior RF, Feitosa THS, Silva MMAPM, Fernandes LFS, Pacheco FAL, Pissarra TCT, Lana RMQ, Melo MC, Valera CA. Carbon footprints of tailings dams’ disasters: A study in the Brumadinho region (Brazil). Sci Total Environ. 2024;949:175026. https://doi.org/10.1016/j.scitotenv.2024.175026
  » https://doi.org/10.1016/j.scitotenv.2024.175026
• Miller ME, Ghisolfi RD, Barroso GF. Remote sensing monitoring of mining tailings in the fluvial-estuarine-coastal ocean continuum of the Lower Doce River Valley (Brazil). Environ Monit Assess. 2023;195:542. https://doi.org/10.1007/s10661-023-11123-x
  » https://doi.org/10.1007/s10661-023-11123-x
• Nasser FC, Mello DC, Francelino MR, Krause MB, Moura Soares H, Demattê JA. Mapping deactivated mine areas in the Amazon forest impacted by seasonal flooding: Assessing soil-hydrological processes and quality dynamics by remote sensing techniques. Remote Sens Appl Soc Environ. 2024;34:101148. https://doi.org/10.1016/j.rsase.2024.101148
  » https://doi.org/10.1016/j.rsase.2024.101148
• Novais JJM, Melo BMD, Neves Junior AF, Lima RHC, Souza RE, Melo VF, Amaral EF, Tziolas N, Demattê JAM. Online analysis of Amazon's soils through reflectance spectroscopy and cloud computing can support policies and the sustainable development. J Environ Manag. 2025;375:124155. https://doi.org/10.1016/j.jenvman.2025.124155
  » https://doi.org/10.1016/j.jenvman.2025.124155
• Obodai J, Bhagwat S, Mohan G. Gold mining's environmental footprints, drivers, and future predictions in Ghana. Remote Sens Appl Soc Environ. 2024;33:101103. https://doi.org/10.1016/j.rsase.2023.101103
  » https://doi.org/10.1016/j.rsase.2023.101103
• Omachi CY, Siani SMO, Chagas FM, Mascagni ML, Cordeiro M, Garcia GD, Thompson CC, Siegle E, Thompson FL. Atlantic Forest loss caused by the world's largest tailing dam collapse (Fundão Dam, Mariana, Brazil). Remote Sens Appl Soc Environ. 2018;12:30-4. https://doi.org/10.1016/j.rsase.2018.08.003
  » https://doi.org/10.1016/j.rsase.2018.08.003
• Pacheco AP, Nascimento JAS, Ruiz-Armenteros AM, Silva Junior UJ, Silva Junior JA, Oliveira LMM, Santos SM, Reis Filho FD, Galdino CAPM. Land cover transformations in mining-influenced areas using PlanetScope imagery, spectral indices, and machine learning: A case study in the hinterlands de Pernambuco, Brazil. Land. 2025;14:325. https://doi.org/10.3390/land14020325
  » https://doi.org/10.3390/land14020325
• Padmanaban R, Bhowmik AK, Cabral P. A remote sensing approach to environmental monitoring in a reclaimed mine area. ISPRS Int J Geo-Inf. 2017;6:401. https://doi.org/10.3390/ijgi6120401
  » https://doi.org/10.3390/ijgi6120401
• Palma HP, Mello DC, Francelino MR, Schmitz D, Veloso GV, Santos APM, Krum DN, Thomazini A, Fernandes-Filho EI, Schaefer CEGR, Oliveira TS, Demattê JAM. Application of sensing techniques for quantifying CO₂ flux and dynamics in environments affected by the Fundão dam collapse, Mariana, Brazil. J South Am Earth Sci. 2024;146:105099. https://doi.org/10.1016/j.jsames.2024.105099
  » https://doi.org/10.1016/j.jsames.2024.105099
• Rotta LHS, Alcântara E, Park E, Negri RG, Lin YN, Bernardo N, Mendes TSG, Souza Filho CR. The 2019 Brumadinho tailings dam collapse: Possible cause and impacts of the worst human and environmental disaster in Brazil. Int J Appl Earth Obs Geoinf. 2020;90:102119. https://doi.org/10.1016/j.jag.2020.102119
  » https://doi.org/10.1016/j.jag.2020.102119
• Rouse JW, Haas RH, Schell JA, Deering DW. Monitoring vegetation systems in the Great Plains with ERTS. Proc 3rd Earth Resour Technol Satellite Symp. 1973;SP-351:309-17. Available from: https://ntrs.nasa.gov/api/citations/19740022614/downloads/19740022614.pdf
  » https://ntrs.nasa.gov/api/citations/19740022614/downloads/19740022614.pdf
• Rudorff N, Rudorff CM, Kampel M, Ortiz G. Remote sensing monitoring of the impact of a major mining wastewater disaster on the turbidity of the Doce River plume off the eastern Brazilian coast. ISPRS J Photogramm Remote Sens. 2018;145:349-61. https://doi.org/10.1016/j.isprsjprs.2018.02.013
  » https://doi.org/10.1016/j.isprsjprs.2018.02.013
• Santos OSH, Avellar FC, Alves M, Trindade RC, Menezes MB, Ferreira MC, França GS, Cordeiro J, Sobreira FG, Yoshida IM, Moura PM, Baptista MB, Scotti MR. Understanding the environmental impact of a mine dam rupture in Brazil: Prospects for remediation. J Environ Qual. 2019;48:439-49. https://doi.org/10.2134/jeq2018.04.0168
  » https://doi.org/10.2134/jeq2018.04.0168
• Silva D, Ferreira M, Scotti MR. O maior desastre ambiental brasileiro: De Mariana (MG) a Regência (ES). Arq Mus Hist Nat Jard Bot. 2015;24:85-7. Available in: https://periodicos.ufmg.br/index.php/mhnjb/article/view/6264
  » https://periodicos.ufmg.br/index.php/mhnjb/article/view/6264
• Silva Junior CA, Coutinho AD, Oliveira-Júnior JF, Teodoro PE, Lima M, Shakir M, Gois G, Johann JA. Analysis of the impact on vegetation caused by abrupt deforestation via orbital sensor in the environmental disaster of Mariana, Brazil. Land Use Policy. 2018;76:10-20. https://doi.org/10.1016/j.landusepol.2018.04.019
  » https://doi.org/10.1016/j.landusepol.2018.04.019
• Souza Junior CM, Shimbo JZ, Rosa MR, Parente LL, Alencar AA, Rudorff BFT, Hasenack H, Matsumoto M, Ferreira LG, Souza-Filho PWM, Oliveira SW, Rocha WF, Fonseca AV, Marques CB, Diniz CG, Costa D, Monteiro D, Rosa ER, Vélez-Martin E, et al. Reconstructing three decades of land use and land cover changes in Brazilian biomes with Landsat archive and Earth Engine. Remote Sens. 2020;12:2735. https://doi.org/10.3390/rs12172735
  » https://doi.org/10.3390/rs12172735
• Thompson F, Oliveira BC, Cordeiro CM, Masi BP, Rangel TP, Paz P, Freitas T, Lopes G, Silva BS, Cabral AS, Soares M, Lacerda D, Vergilio CS, Lopes-Ferreira M, Lima C, Thompson C, Rezende CE. Severe impacts of the Brumadinho dam failure (Minas Gerais, Brazil) on the water quality of the Paraopeba River. Sci Total Environ. 2020;705:135914. https://doi.org/10.1016/j.scitotenv.2019.135914
  » https://doi.org/10.1016/j.scitotenv.2019.135914
• Tonietto A, Silva JJMC. Valoração de danos nos casos de mineração de ferro no Brasil. Rev Bras Criminalística. 2011;1:31-8. https://doi.org/10.15260/rbc.v1i1.15
  » https://doi.org/10.15260/rbc.v1i1.15
• Vasconcelos TNC, Alcantara S, Andrino CO, Forest F, Reginato M, Simon MF, Pirani JR. Fast diversification through a mosaic of evolutionary histories characterizes the endemic flora of ancient Neotropical mountains. Proc R Soc B. 2020;287:20192933. https://doi.org/10.1098/rspb.2019.2933
  » https://doi.org/10.1098/rspb.2019.2933
• Vermote E, Justice C, Claverie M, Franch B. Preliminary analysis of the performance of the Landsat 8/OLI land surface reflectance product. Remote Sens Environ. 2016;185:46-56. https://doi.org/10.1016/j.rse.2016.04.008
  » https://doi.org/10.1016/j.rse.2016.04.008
• Werner TT, Bebbington A, Gregory G. Assessing impacts of mining: Recent contributions from GIS and remote sensing. Extr Ind Soc. 2019;6:993-1012. https://doi.org/10.1016/j.exis.2019.06.011
  » https://doi.org/10.1016/j.exis.2019.06.011
• Ye P. Remote sensing approaches for meteorological disaster monitoring: recent achievements and new challenges. 2022;19:3701. https://doi.org/10.3390/ijerph19063701
  » https://doi.org/10.3390/ijerph19063701

*

Corresponding author: e-mail: [e-mail]

Editor:

José Miguel Reichert https://orcid.org/0000-0001-9943-2898 and Waldir de Carvalho Junior https://orcid.org/0000-0002-8619-0989

DECLARATION OF COMPETING INTEREST

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Publication Dates

• Publication in this collection
  31 Aug 2026
• Date of issue
  2026

History

• Received
  24 Nov 2025
• Accepted
  05 Feb 2026