Seasonal geochemical trends and pollution assessment of bottom sediments in the São Francisco hydrographic basin, Brazil: the Três Marias Reservoir

The Três Marias Reservoir is the ninth largest reservoir in Brazil, becoming crucial for national strategic development. However, many anthropic activities may affect the sediment quality, promoting the need for a proper environmental assessment. This research appraised the seasonal influences on the Três Marias Reservoir’s sediment geochemistry, elucidating possible anthropogenic impacts. The concentrations of Mg, Al, Ca, Cr, Fe, Co, Cu, Cd, Ti, Mn, Ni, Zn, Ba, and Pb were measured in 78 samples of bottom sediments regarding the two seasons of the area, a dry winter and rainy summer. The median ± 2 median absolute deviation (MAD) settled the geochemical background and environmental thresholds for the two seasons. The sediment quality guidelines CONAMA 344/12 highlight the possible adverse ecological effects of pollutants. The hierarchical clustering analysis, the geoaccumulation index, and the pollution load index delineated the polluted zones. The pollution load index ranges from 0.25 to 2.28 in the dry season and 0.56 to 2.11 in the rainy season, defining three affected zones in the reservoir. Forestry and agriculture are the probable pollution sources, reaching warning levels that should be considered in further environmental strategies.


Introduction
The evaluation and monitoring of potential toxic elements (PTEs) are fundamental in terms of environmental sustainability and human health safety (Srikanth et al. 2014). A geochemical assessment of PTEs in sediments is crucial to understand potential hazards to aquatic ecosystems once it can work as a geochemical reservoir capable of adsorbing and remobilizing pollutants. Therefore, the characterization of PTEs and trace elements in sediments can be a powerful tool to define possible anthropogenic stress in a watershed (Calmano et al. 1993;Peng et al. 2009). It is also important to highlight that trace element inputs to sediments can originate from geogenic sources (weathering, surface runoff, atmospheric deposition) and anthropic sources as agriculture and industrial and domestic discharges (Srikanth et al. 2014;Thin et al. 2020).
Sediment strata are also susceptible to register contamination coming from direct wastewater inputs in the watershed and coming from atmosphere pollution (Junior et al. 2014;Baird 2018). Various PTE can be linked with different sediment fractions through adsorption, coprecipitation, and complexation (Junior et al. 2014). Once PTEs are detected in the environment, it is crucial to assess the source of these elements and their bioavailability. For this reason, the geochemistry of sediments based on quantitative, spatial, and seasonal assessment is necessary to define trace elements' role in the environment (Bouezmarni and Wollast 2005).
A powerful tool to assess contaminants and pollutants in sediments is to define regional reference values and environmental thresholds for a specific area. For environmental proposes, settling a geochemical background range and environmental thresholds helps detect the local of potentially anthropogenic affected zones (Reimann and Garrett 2005;Gałuszka and Migaszewski 2011). The conception of geochemical background is mainly related to a specific chemical range that can be considered natural to an ecosystem, without non-natural/anthropogenic influences (Matschullat et al. 2000). However, the proper understanding of the geochemical background and the methodologies to calculate it are multiples and ambiguous and vary according to the context and application (Reimann and Garrett 2005;Rodrigues and Nalini Júnior 2009;Gałuszka and Migaszewski 2011). In addition, the environmental thresholds can be understood as concentrations of an element above which further investigation is required (Reimann and de Caritat 2016). In environmental risk assessment of the São Francisco basin, the geochemical background was adopted as synonymous of defining natural concentrations and environmental thresholds in which the ecosystem is probably under anthropogenic stress.
Nowadays, the São Francisco River can be considered the most important surface water resource regarding the domestic and industrial water supply in Brazil. In the Upper São Francisco basin context, the Três Marias Reservoir plays a fundamental role in national economic development. However, the reservoir's history has been marked by several anthropogenic impacts concomitantly with negligible environmental policies (e.g., CETEC 1983;Greenpeace 2002;Oliveira and Horn 2006;Theodoro et al. 2007;Junior et al. 2014;Bento 2020). Therefore, it is necessary to structure an environmental baseline to assess and monitor the anthropic evolution and their respective impacts on nature (Torres et al. 2019).
The construction of the reservoir in 1960 aimed to supply the region's energy shortage and control overflood issues in the São Francisco River (Prado and Pompeu 2014). A zinc mining industry was established in 1969 nearby the reservoir. At that time, the industrial discharges were disposed in the soil, lake, and tributaries, carrying enormous quantities of PTE to the water, air, soil, and sediments (Oliveira and Horn 2006;Junior et al. 2014). Concomitantly, the limestone, phosphates, and potash rock mining in the reservoir's tributaries are also a concern regarding PTE inputs to the sediments (Junior et al. 2014). In the 1990s, several studies indicate alarming concentrations of Cu, Cd, Zn, and Pb in the reservoir (Oliveira and Horn 2006). In 2005, it was noticed a large event of fish mortality in the reservoir (IGAM 2005;Junior et al. 2014). At that moment, there are few adjustments in the environmental licensing process which guided actions to mitigate these anthropic impacts. After changes in the regional environmental policies, agriculture (pesticides and fertilizers), fishing farms, and the small urban activities assume the protagonism as the main contamination source in the Três Marias Reservoir (Horn et al. 2014;Bento et al. 2019;Lima 2020). It is worth mentioning that in January 2019, the B1 iron ore tailing dam collapsed in Brumadinho. It was one of the worst mining-related disasters, with 270 human deaths and 12.106 m 3 tailings released to the environment (Parente et al. 2021). The magnitude and the extension of this environmental disaster are not fully dimensioned yet. There is the hypothesis that this accident might affect the Três Marias Reservoir ecosystem; however, the geochemical data presented in this research was collected right before this accident.
Nowadays, the Três Marias Reservoir is directly or indirectly responsible for the economic development of more than ten municipalities, highlighting Três Marias, Abaeté, and Morada Nova de Minas as the main urban centers in the region (Junior et al. 2014;Lima et al. 2021). The reservoir's surrounding area has been intensely used for agriculture, tourism, forestry, housing, fish farming, waterway traffic, industry, and pastures (Euclydes et al. 2001;Horn et al. 2014;Torres et al. 2019).
This research aims to better comprehend the behavior of Pb, Cr, Cu, Zn, Ba, and Ti and their respective environmental context of bottom sediments geochemistry regarding the following criteria: (1) Define a geochemical background based on the seasonal criteria.
(2) Use different environmental indexes to propose a temporal and spatial assessment to contaminated areas. (3) Elucidate possibilities of contamination/pollution sources regarding the anthropogenic activities and the geological context.

Study area
The Três Marias Reservoir is the ninth largest reservoir in Brazil, and it delimitates the segment between the Upper and Medium São Francisco River (Euclydes et al. 2001) (Fig. 1). The reservoir's maximum length is about 150 km and has an average depth of 16.8 m (Fonseca et al. 2012;Prado and Pompeu 2014). At maximum water level, the flood zone covers 1,050 km 2 reaching a maximum water volume of 21 × 10 9 m 3 (Prado and Pompeu 2014). The climate is classified as a typical tropical rainy system (Köppen, Aw), and the region experiences two different seasons: a dry winter from April to September and rainy summer from October to March (Godinho and Godinho 2003;Prado and Pompeu 2014;Torres et al. 2016). In this research, the hypothesis of a seasonal control in the PTE concentrations in sediments was assessed. The premise of seasonal effects in the geochemistry of the sediments is based on the following information: (1) The Três Marias Reservoir has two well-defined seasons with significant differences in temperature and rainfall.
(2) Bento (2020) conducted a hydrochemistry analysis in Três Marias watershed and indicated seasonal influences in PTE behavior. (3) Analysis conducted in fishes from Três Marias downstream also indicates PTE changes considering the seasonal effects (Gomes and Sato 2011). (4) The seasonal changes in runoff, precipitation, and oxygen availability in the deep layers might induce changes in the sediment geochemistry in Três Marias (Fonseca et al. 2012).
The regional geology describes two major stratigraphic elements that control the sediment loads in the Três Marias Reservoir: the Bambuí and the Mata da Corda Groups (Fonseca et al. 2011;Lima et al. 2021). The Bambuí Group is defined as marine paleoenvironment interlaying metapelites from Serra de Santa Helena Formation and Serra da Saudade Formation with carbonates from Sete Lagoas and Lagoa do Jacaré Formations. On the top, siliciclastic units from Três Marias Formation overlaps the metapelite-carbonate sequences (Campos and Dardenne 1997;Reis 2018). The alkaline mafic to ultramafic potassic rocks from the Mesozoic represent the Mata da Corda Formation in the area (Campos and Dardenne 1997;Fonseca et al. 2011).
Various land use and soil management were developed in the Três Marias Reservoir (Fig. 2). Since the 1990s, the natural vegetation, consisting in a transitional system between Savannah/Cerrado and Caatinga, has been replaced by the large production of housing, fishing, fruit farms, and forestry (Baggio and Horn 2010;Horn et al. 2014;Trindade et al. 2018). The evolution of small urban centers also promotes tourism and waterway traffic in the area. The Três Marias Reservoir and its main tributaries concentrate the largest number of fishers of Minas Gerais (Alvim and Peret 2004). The left shore of the reservoir, specifically in the Indaiá and Borrachudo Rivers, also has a high potential for unconventional hydrocarbon exploration that is still in development (Lima et al. 2020). Along the shores of the reservoir, the forestry is mostly marked by the anthropogenic forest of Pinus and Eucalyptus (Trindade et al. 2018).
The diversity of land uses and managements, the high potential for many economic activities, and a large number of municipalities directly or indirectly connected with the reservoir make the environmental assessment a real need regarding the principles of sustainable development.

Sampling and analytical procedures
In two campaigns, 40 samples were collected in the dry season (July/2018) and 38 in the rainy season (February/2019), regarding the two main seasons mentioned previously (Fig. 1). The sampling site distribution appraised all main tributaries, different land uses and soil management adjacent to the Três Marias Reservoir, and different lithologies observed in situ. The sampling procedures were executed by non-metallic tools, using a composite sampling methodology on the first 15 cm of the bottom sediments. The sediments were cooled correctly and transported to the laboratory for further analysis.
The Nucleus of Environmental Geochemistry Research (NGqA-CPMTC/UFMG), Belo Horizonte, Brazil, performed the sediments analytical procedures. The samples were oven-dried at 303.15 K for 36 h and then homogenized, pulverized, and sifted through a 0.063-mm sieve. Fine grain size particles were used for the chemical measuring. The chemical analyses were carried out following the EPA-3051A methodology, which settles the single extraction procedures using a microwave-assisted regarding the geological matrices (US, EPA 1998). The samples were digested using 10 ml of HNO 3 (65% v/v supplied by Merck) in a microwave MARS-CEM. The inductively coupled plasma optical emission spectrometry (ICP-OES) (Spectroflame-Spectro Analytical Instruments) measured the concentration of Mg, Al, Ca, Cr, Fe, Co, Cu, Cd, Ti, Mn, Ni, Zn, Ba, and Pb in the sediment samples. Quality assurance was given by the analysis of duplicates in 10% of the collected samples, and analytical blanks were performed for each 7-sample batch. The t test was applied to the duplicates, and it did not differ significantly regarding a p value of 0.02. The curves were performed against international and national laboratory standards.

Data analysis and geochemical background
Setting a range of concentrations expected of an ecosystem may be challenging, and it is crucial to settle a proper environmental assessment crossing different approaches. The median ± 2 median absolute deviation (± 2 MAD) was applied to assess the two well-defined seasons' background values in the Três Marias Reservoir. Choosing a ± 2 MAD method to define the threshold values creates highly conservative (low) levels for the background range, robust against outliers, and it may be more efficient for environmental proposes (Reimann and de Caritat 2016). The median values were taken as representative of the background concentrations range. Variables with a high amount of data below the detection limit were excluded from the background analysis. The censored data was adopted as 0.5 times the lower detection limit (Reimann et al. 2008;Hron et al. 2010).
The Shapiro-Wilk normality test was applied to the logtransformed data to appraise the behavior of each distribution. The Mann-Whitney test (Wilcoxon test) was used to assess possible statistical differences between seasonal concentrations regarding a significance level of 95% (p < 0.05).
In each sampling site, the PTE concentrations were compared with the Brazilian sediment quality guidelines CONAMA 454/2012. The Brazilian legislation adopts the threshold effect level (TEL) and probable effect level (PEL) following the Canadian Sediment Quality Guidelines for the Protection of Aquatic Life (CCME 2002). The PTE concentration was also compared with the upper continental crust (UCC) levels (Wedepohl 1995).

Environmental index assessment
It was possible to settle the contemporary context of the chemical contamination in Três Marias Reservoir crossing spatial data, statistical techniques, and two different environmental indexes. Table 1 detailed the procedures and classification of each applied environmental index. The PTEs were individually appraised in each sampling site using the geoaccumulation index (I geo ). The pollution load index (PLI) integrated multiple elements in each sample evaluation, giving an insight into the overall pollution content in the affected areas. The concentrations of Cr, Cu, Ba, Ti, Pb, and Zn were taken into account for the PLI calculus.
It is essential to highlight that the I geo and the PLI indexes were primarily developed to use in the < 2 μm sediment fraction (Förstner et al. 1990). However, the authors applied these indexes using geochemical concentrations in the 0.063-mm grain size fraction in this research. This choice was considered because the reference values applied in the indexes (B n values) were also obtained in the 0.063-mm grain size. In addition, the Brazilian sediment guidelines (CONAMA 454/2012) also work with 0.063-mm grain size to settle the TEL and PEL thresholds.
The I geo and PLI mapping were interpolated using inverse distance weighted (IDW) through the GIS platform using QuantumGIS (QGIS®). Those indexes' spatial assessment highlights the seasonal effects in the geochemistry of bottom sediments in the reservoir. The spatial approach also Table.1 Environmental indexes and their respective classification C n , the measured concentration at the sampling site. B n , the concentration adopted as reference values (background values) -in this research, the median value of each season was adopted as B n ( Uncontaminated to moderately contaminated 1 < I geo < 2 Moderately contaminated 2 < I geo < 3 Moderately to heavily contaminated 3 < I geo < 4 Heavily contaminated 4 < I geo < 5 Heavily to extremely contaminated I geo ≥ 5 Extremely contaminated Pollution load index (PLI) (Tomlinson et al. 1980) 0 < PLI ≤ 1 Baseline levels of pollutant present 1 < PLI ≤ 10 Polluted 10 < PLI ≤ 100 Highly polluted PLI > 100 Progressive deterioration of environment enhances the probable correlation between the chemical inputs.
The boxplot analysis applied in the I geo values using the Tukey inner fence (Tukey 1977) for outlier evaluation brought insights about the most polluted samples. The hierarchical clustering analysis (HCA) using Ward's method as a linkage procedure considering the Euclidean distance enforced the I geo patterns associated with contaminant inputs. The HCA was also performed for the I geo observation data (samples) using the same methods of linkage and distance, which give some information about the clustering structure along the study area.

Geochemical background and seasonal assessment
The descriptive statistics and geochemical background evaluation for the dry season are summarized in Table 2. The median concentration of the elements decreased in the dry season as follows: Fe > Al > Ca > Mg > Mn > Zn > Cu > Ba > Ti > Pb > Cr. Cd and Ni geochemical background was not appraised due to the large amount of censored data.
The background values (represented by the median concentration values of each element distribution) of Cu, Ti, and Pb are above the mean UCC levels, whereas the values of Fe, Al, Ca, Mg, Mn, Cr, and Zn are under the UCC threshold. The maximum concentration values of Cu, Zn, and Pb reach the TEL guideline, while the Cr's maximum values reach the PEL guideline.
The descriptive statistics and the geochemical background appraisal for the rainy season are summarized in Table 3. The median concentration of the elements decreased in the rainy season as follows: Fe > Al > Ca > Mg > Mn > Ba > Cr > Pb > Cu > Ti > Zn. Similar to the dry season, Cd and Ni geochemical background values were not settled due to the large amount of censored data. The background values (median) of Fe, Cu, Ti, and Pb are above the mean upper continental crust levels, while Al, Ca, Mg, Mn, Cr, and Zn are under the UCC threshold.
As detailed in Table 4, the Wilcoxon rank test denotes that Fe, Al, Ca, Cr, Ba, and Pb have statistical significant differences between those two seasons, which may imply that those elements are more susceptible to seasonal physical-chemical variabilities in the reservoir. On the other hand, Mg, Mn, Cu, Ti, and Zn show no statistically significant differences between the dry and rainy season. Cd and Ni were removed from the Mann-Whitney test due to the large amount of censored data.
It is possible to identify areas more susceptible to adverse environmental issues by comparing concentrations in each sample with the UCC, TEL, and PEL levels (Fig. 3). Seven samples reach the TEL level for Cr, whereas, in the rainy season, 11 samples and 2 samples reach the TEL and PEL threshold, respectively. Overall, the Pb concentrations increase in the rainy season. In both seasons, the Pb measured are above the TEL level in two different areas (samples  Figure 4 correlates the measured concentrations of Zn, Ti, and Cu, with the UCC, TEL, and PEL thresholds. The Zn values are under the environmental adverse effects levels, and only one sample has concentrations over the UCC standard reference. The Ti concentrations are all under UCC levels, and the Brazilian legislation does not settle sediment environmental guidelines for this element. The Cu ratios are above the TEL threshold in two different areas (samples 9-11, 39), implying a potential environmental hazard.
On the middle-left area of the reservoir, specifically in samples 9, 10, 13, and 14, it is observed that Cu and Pb concentrations are above the TEL levels, while the Cr concentration reaches the PEL levels. These concentrations are spatially correlated with the agriculture and forestry zones (Fig. 2). The peak of Ti in the same area (Fig. 4b) may imply in a lithophile sign (geogenic) input in those sediments. At first insight, the association between Ti-Cr-Pb may be a consequence of the geological context of Mata da Corda Formation (Theodoro et al. 2007;Fonseca et al. 2012;Lima et al. 2021). However, the higher concentrations of Cr and Pb, and considering the presence of Cu also in warning levels, might represent an additional anthropogenic input in the reservoir. These results are in agreement with the literature. These PTE (Cr and Pb) has strong correlation with the intensive use of agrochemicals nearby the reservoir (Junior et al. 2014).
The sampling sites 22 and 23 also have a similar environmental context of the samples 9-13. This area also shows concentrations over the TEL levels for Cr, Pb, and Cu, which means a certain probability of adverse effects in the biota. The geochemical provenance assessment proposed by Lima et al. (2021) grouped all those samples in the Mata da Corda Formation influencing area, which means that the potassic ultramafic rocks could naturally enhance the Cr and Pb concentration in those samples. The higher concentrations of Ti in those samples complement the hypothesis of a geogenic contribution of PTE. In spite of that, Fig. 2 also shows these samples spatially correlated with forestry and agriculture zones, which could imply an additional anthropogenic input in those places, explaining the high concentrations of Cu above the TEL thresholds. The surrounding land uses and management might also increase the surface runoff in those locations (Fonseca et al. 2012). Bento (2020) also indicates warning concentrations of PTE and organic pollutants in the water correlated to intensive agriculture and livestock practices. Therefore, this area may be affected by a geogenic input naturally enhanced with Cr, Pb, and Ti and, concomitantly, by an anthropogenic input explaining the higher Cr levels and a probable increment of Cu and Pb concentrations. In addition, an environmental assessment based on the geoaccumulation index-pollution load index may enhance this hypothesis.
Nearby the turbine uptake of Três Maria's reservoir, the dry season shows Cu concentration levels above the TEL threshold in sample 39. In the same sampling site, it is worth highlighting the Zn concentrations above the UCC levels and significantly higher than the overall concentrations of the reservoir. According to Bento (2020), this area is highly affected by urban activities and waterway traffic, also leading warning levels of Cu, Zn, Al, B, and NO 3 − in the water. Nonetheless, the rainy season does not have the same environmental-chemical behavior.
The Mann-Whitney test shows no statistical differences between seasons, enhancing the possibility of punctual contamination sources at this sampling site. The season's rainfall could also act on the reservoir as a purification (dilution) process, explaining this difference between seasons. The environmental index assessment was applied to  (Wedepohl 1995); TEL, threshold effect levels; PEL, probable effect levels; *the Brazilian legislation do not define TEL and PEL thresholds for Ba  (Wedepohl 1995); TEL, threshold effect levels; PEL, probable effect levels; *the Brazilian legislation do not define TEL and PEL thresholds for Ti, while the UCC limit is significantly above the scale of the graph

Geoaccumulation index assessment (I geo )
The I geo index was used as a reference for estimating the contamination content in both seasons in the reservoir (Table 5). The mean I geo evaluation for Cr, Cu, Ti, Zn, Ba, and Pb for both seasons indicates an overall unpolluted ecosystem (I geo ≤ 0). All I geo (Ba) and I geo (Pb) denoted an uncontaminated reservoir for both seasons. I geo (Cr) and I geo (Zn) values indicate unpolluted to moderately polluted environments in specific sites for both seasons. The I geo (Ti) reaches the unpolluted to moderately polluted system in the dry season and moderate contamination in the rainy season. The I geo (Cu) indicates an unpolluted to moderately polluted system in the dry season and unpolluted in the rainy season. Figure 5 elucidates the boxplot distribution of I geo index in each season. This analysis intends to highlight possible anomalies in the I geo distribution in the reservoir. I geo (Pb) and I geo (Ba) values do not show contamination in both seasons. I geo (Cu) values denote an unpolluted environment in the rainy season, while two samples highlighted as outliers in the dry season are classified as moderately contaminated. The statistical seasonal differences and the I geo assessment highlight the possibility of a mechanism that concentrates Cu in the sediments in the dry season or rainfall acts as a dilution factor in the rainy season. The positive outliers in the I geo (Zn) indicate a differential behavior of the Zn distribution in the reservoir in both seasons. The I geo (Cr) reaches the uncontaminated to moderately contaminated classification; however, no outliers appear in the distribution. The I geo (Ti) displays the highest I geo values in the reservoir, and one outlier in the dry season earns the moderately contaminated classification. Figure 6 shows a hierarchical clustering analysis regarding the I geo values in each season. In both seasons, two cluster groupings I geo (Pb)-I geo (Cr)-I geo (Ti) and I geo (Ba)-I geo (Cu)-I geo (Zn) were defined. This result may support the hypothesis of the geogenic inputs of Pb, Cr, and Ti, which are related to the lithotypes from Mata da Corda Group, as mentioned in the Geochemical background and seasonal assessment section. On the other hand, Zn, Cu, and Ba should have their provenance from both geogenic (Bambuí Group) (Fonseca et al. 2011;Lima et al. 2021) and anthropogenic sources. This interpretation is better understood in the I geo distribution maps in the Três Marias Reservoir. Figure 7 assesses the I geo (Cr), I geo (Ti), and I geo (Pb) distribution, while Fig. 8 assesses the I geo (Zn), I geo (Ba), and I geo (Cu) distribution maps. The I geo (Pb) assessment shows unpolluted to moderately polluted in a central south area of the reservoir in both seasons, with slight decreasing index values during the rainy season. The Pb concentrations are also above the TEL levels in these polluted areas. The I geo (Cr) distribution has significantly changed between the seasons, as exposed in Fig. 7a and b. It is worth noticing that the Wilcoxon test (Table 4) and the Cr concentration in each sample (Fig. 3a) also highlight the seasonal differences in the Cr distribution. The Cr-polluted areas on the east shore of the reservoir are mainly spatially correlated with agriculture/ forestry, as shown in the land use map (Fig. 2). The rainy season shows high Cr contamination signs in the reservoir's east margin. When those concentrations are compared with the Brazilian sediment guidelines (Fig. 3a and c), the Cr and Pb levels reach the environmental thresholds (TEL and PEL), standing an alarm about possible adverse concerns in those areas. The moderate contamination at the Indaiá River appears only in the rainy season, implying a punctual contamination source high influenced by seasonality. The I geo (Ti) distribution is quite similar in both seasons, showing its higher values in the reservoir's central zone.
The similarities between the Ti-Cr-Pb behaviors enhance a geogenic input hypothesis correlated to ultramafic rocks from the Mata da Corda Formation. The hierarchical clustering analysis in the I geo values (Fig. 6) reinforces this geochemical assemblage geogenic-related. Although, the higher I geo values of Cr and Pb near the agriculture zones highlight the possibility of anthropogenic inputs towards the sediment strata in the reservoir. Another concern is the Cr and Pb concentrations above the TEL and PEL thresholds, implying the need for proper environmental monitoring to avoid adverse biota impacts. Figure 8 shows analog distributions between the I geo (Zn)-I geo (Ba)-I geo (Cu) values. The I geo (Zn) distribution in both seasons is comparable, with a slight increase in the rainy season. The Wilcoxon test also shows no statistical seasonal differences in the Zn distribution. The I geo (Zn) mapping ( Fig. 8a and b) denotes moderately contamination spatially correlated with the forestry zones (Fig. 2). It is worth noticing that in any sample, the Zn concentrations surpass the environmental guidelines thresholds. The I geo (Ba) indicate unpolluted to moderately polluted at the reservoir's west shore in both seasons. The I geo (Cu) reaches a moderate level of contamination in the dry season and unpolluted to moderately polluted in the rainy season. The slight decrease of the I geo (Cu) in the rainy season may be an effect of the intense rainfall in the reservoir. However, the Wilcoxon test does not show statistical differences in Cu levels between seasons. The Cu-polluted areas are spatially correlated with forestry and agriculture (Fig. 2), and the Cu concentrations are above the TEL level in samples 9 to 12 (Fig. 4c).
It is worth noticing the spatial similarities between the I geo (Zn), I geo (Cu), and I geo (Ba) distribution also accordingly with the hierarchical cluster analysis. In the Três Marias Reservoir, the multi-element geochemical mapping associates Zn-Cu-Ba with the Bambuí Group (Lima et al. 2021). Therefore, the similarities between Zn-Cu-Ba may be geogenic-correlated. It is also important to notice that Cu, Cr, and Pb distribution are spatially correlated with the forestry and agriculture zones, both above the TEL levels and probably connected with anthropogenic inputs accordingly to the I geo classification.

Pollution load index (PLI) assessment and hierarchical cluster analysis (HCA) considering the observation data (samples)
The pollution load index (PLI) was applied to summarize the I geo assessment and delimitate probable susceptible areas in the Três Marias Reservoir. The PLI values in the dry season range from 0.25 to 2.28, whereas the PLI values in the rainy season range from 0.56 to 2.11. In both seasons, the PLI index (Table 1) classifies the reservoir predominantly as unpolluted to a baseline level of pollutant environment. However, in specific sites, the PLI indicates a sign of pollution. The PLI mapping (Fig. 9) provides an overall pollution The pollution load index sectorized the reservoir in different anthropic stressed areas. The middle south of the reservoir shows the higher levels of the PLI in both seasons. In those areas, the I geo index assessment shows pollution signs of Cr, Cu, Zn, and Pb. The west shore of this zone was influenced by forestry, whereas the east shore was affected by agriculture. In this zone, the Cr, Cu, and Pb reach the TEL threshold, indicating the possibility of adverse environmental effects on nature. The similar behavior of Cr-Pb-Ti and the hierarchical clustering analysis highlight a correlated geogenic input from the Mata da Corda Group that could increase their concentration (e.g., Fonseca et al. 2012;Lima et al. 2021). However, the I geo spatial assessment also shows signs of anthropogenic stress in that region, where the I geo (Zn) might be an indicator of anthropogenic input. Those elements may be associated with the use of pesticides in Fig. 7 Geoaccumulation classes were mapping for Cr, Ti, and Pb. I geo classes: 0, unpolluted; 1, unpolluted to moderately polluted; 2, moderately polluted. a I geo (Cr) in the dry season; b I geo (Cr) in the rainy season; c I geo (Ti) in the dry season; d I geo (Ti) in the rainy season; e I geo (Pb) in the dry season; f I geo (Pb) in the rainy season agriculture and forestry that could increase the PTE concentrations on sediments (Junior et al. 2014). Another hypothesis is that the marginal land use and management may intensify the runoff surface on the shores, raising the PTE concentration to warning levels (above the TEL threshold). In this case, the environmental indexes indicate a polluted zone, and the PTE ratios are higher than the sediment quality guidelines, implying a susceptible area for ecological issues.
The Indaiá River, on the western portion of the reservoir, shows higher PLI levels during the rainy season (samples 17 and 20). The I geo (Cr) assessment classified this zone as moderately polluted, and the Cr measured concentrations reaches the PEL reference value, indicating a probable adverse effect to the environment. Regardless, this evaluation does not apply to the dry season, enhancing the hypothesis of a pollution source highly affected by seasonality. Grassland   Fig. 8 Geoaccumulation classes were mapping for Zn, Ti, and Cu. I geo classes: 0, unpolluted; 1, unpolluted to moderately polluted; 2, moderately polluted. a I geo (Zn) in the dry season; b I geo (Zn) in the rainy season; c I geo (Ba) in the dry season; d I geo (Ba) in the rainy season; e I geo (Cu) in the dry season; f I geo (Cu) in the rainy season and pastures predominantly occupy the Indaiá River's downstream, which could not explain the Cr seasonal behavior in this area. The geological context also does not justify an intense Cr input in this site. The high concentration of Cr in this area could be connected with the role of agriculture and pisciculture that is strongly affected by the seasonality, once the Indaiá River has one of the largest Tilapia fish farming of Brazil. Bento (2020) also indicates warning levels of Cu, Mn, and Mg in the water in this area, which stand the potential risks in this place.
The PLI reaches values slightly higher than 1 in the extreme north (turbines uptake) of the Três Marias Reservoir for both seasons. This area has Zn concentrations higher than the UCC levels, and the Cu concentrations reach the TEL reference value. The I geo evaluation shows slight pollution of Zn and Cu in those sites. Bento (2020) also shows anomalies of Cu and Zn in the water explained by the heavy waterway traffic and small urban activities in this zone. However, right in the downstream of the Três Marias Dam, Junior et al. (2014) highlight higher levels of Cu and Zn in the sediments explained as a residual pollution associated with the Zinc mining in the 1960s. It might be a pollution caused by multifactorial inputs. It would be necessary a local assessment for better understanding the source of this contamination.
For a better comprehension of I geo value distribution in the study area, the HCA considering the observation data (samples) was applied using I geo values for Zn, Ba, Cu, Cr, Ti, and Pb in each season. Figure 10a and b show the cluster observation results, identifying three different groups for each season. Figure 10c and d identify the obtained clusters on the map, correlating the results with the PLI mapping from Fig. 9.
In both seasons, clusters 1 and 2 (the blue and green circles in the maps, respectively) grouped more samples in upstream and downstream, respectively. It may reflect a differential geochemistry signature in the sediment's structure between those two zones, either by geological features or anthropogenic activities in those regions (Theodoro et al. 2007;Fonseca et al. 2011;Lima et al. 2020Lima et al. , 2021. However, cluster 3 (yellow circles) separates samples with distinctive I geo associations, highlighting probable polluted areas. The spatial evaluation of cluster 3 has a strong correlation with high PLI values, enhancing the hypothesis of probably affected areas for both seasons.
Nowadays, several studies have been conducted in order to understand the bioavailability of these PTE and the environmental safety of nearby communities (e.g., Bento, 2020;GUAICUY, 2021). Other aspects as water quality, hydrochemistry, limnology of benthonic, and macroinvertebrates also should be analyzed to define a solid biogeochemical understanding of the Três Marias Reservoir. It is also crucial to keep a frequency of environmental monitoring studies to guarantee an environmentally sustainable development regarding the water resources of the São Francisco Basin.

Conclusion
The median ± 2 MAD defined a regional background based on the seasonal criteria. These new standard values allowed a proper environmental evaluation of the sediments' actual geochemical settings from the Três Marias Reservoir. Comparing the measured concentration in each sample with the Brazilian sediment quality guidelines brought insights into areas prone to adverse environmental issues. The hierarchical clustering analysis of the I geo classes denoted a geochemical association similar to the geogenic inputs from the reservoir's geological Fig. 9 Pollution load index mapping regarding the two main seasons of the Três Marias Reservoir. a PLI mapping on the dry season; b PLI mapping on the rainy season structures. However, the I geo classification indicates some polluted areas related to anthropogenic influences. The I geo index mapping allowed the spatial identification of probable anthropogenic stressed zones and different behaviors between two seasons. The PLI assessment summarized the role of the PTEs in the Três Marias Reservoir, indicating three more susceptible zones that should be better managed. The middle south of the reservoir shows signs of agriculture and forestry influences on the bottom sediments' quality. The I geo and PLI index assessment highlights the hypothesis about how the anthropic impacts could be affecting the sediments leading to warning concentrations (above the TEL and PEL standards), enhancing the need for an evaluation to comprehend the role of PTE on this site. The environmental index mapping also indicated a Cr contamination source highly associated with the seasonality behavior. A third probable affected zone is located in the extreme north of the reservoir regarding the Cu levels higher than the TEL reference value and the PLI slightly higher than 1. The HCA applied to the observation data (samples) was a multivariate technique based on the I geo values used to validate the PLI index interpretation.
These results provide an environmental assessment of a sensible area for Brazilian sustainable development on a national scale. The geochemical background regarding the seasonal criteria contributes as a powerful tool for further environmental assessment and management. The highlighted anthropic pressures using the environmental indexes furnished the most susceptible zones in the ecosystem. Further studies should be appraised to understand the real magnitude of the contamination issues. The results can enforce environmental policymaking and create a baseline for new strategies Fig. 10 Clustering observation analysis of dry and rainy seasons obtained in cluster 1 (blue), cluster 2 (green), and cluster 3 (yellow). The three clusters were spatially correlated with the pollution load index mapping for both campaigns. a Cluster observation results for the dry season; b cluster observation results for the rainy season; c cluster observation mapping over the PLI index mapping for the dry season; d cluster observation mapping over the PLI index mapping for the rainy season to improve the water resources quality in the Upper São Francisco River Basin.