SURFACE RUNOFF IN TWO REHABILITATION PERIODS OF A BAUXITE MINE

The objective of this study was to monitor and compare runoff at 2.5 and 3.5 years of rehabilitation of a mined bauxite area with clonal planting of Eucalyptus. Seven plots were allocated for collecting runoff and precipitation was recorded with a pluviograph. The physical and environmental factors which interfere with runoff were characterized in both periods and subjected to principal component analysis (PCA) to show the most explanatory factors. The average monthly runoff percentage at 2.5 years was 0.25% (± 0.26%) of precipitation and equal (p-value<0.05) to the 3.5 years (0.15 ± 0.22%) of rehabilitation. Both values were lower than the Eucalyptus plantation area without mining (0.56%) in the same region and declivity. The high vegetation cover percentage by Eucalyptus crowns and Brachiaria sp. and the high infi ltration rates were the determining factors in PCA, and may explain the statistically equal runoff values between the rehabilitation periods. The combination of good edaphic conditions and the fast coverage of the plants resulted in runoff below 1% in a rehabilitated area after bauxite mining.


INTRODUCTION
Bauxite is among the minerals extracted in Brazil, coming from surface mines, which are also called open-pit mines (ABAL, 2018). Surface mining starts by removing vegetation and the upper soil horizons to expose the ore. After ore removal, the original waste and organic soil are replaced to reconstruct the topography of the area, and revegetation is carried out at the end (Vilas Boas et al., 2018).
The superfi cial mining process causes changes in the landscape such as topographical drawdown (Kuter, 2013), an increase in soil density (Barros et al., 2013), and the formation of cracks and fi ssures in the soil surface (Haigh and Sansom, 1999), thereby modifying hydrological processes with a reduction in runoff due to soil turning and the presence of species with high vegetation coverage (Merino-Martin et al., 2012).
The return to a similar or better functional ecosystem than that before the mining is the rehabilitation goal after surface mining (Reynolds and Reddy, 2012). A challenge in this process is controlling the water runoff (Rubio et al., 2013). Several techniques have been used to control it such as revegetation, which helps in recovering organic matter, restructuring the mined soil (Banning et al., 2008), and unpacking the soil with subsoiling, which can induce positive hydrological results in rehabilitated ecosystems when associated with revegetation (Sheoran et al., 2010). Thus, monitoring changes in soil and vegetation properties associated with hydrological processes (such as runoff ) in sequential rehabilitation periods is essential to assess and improve the implemented techniques (Shrestha and Lal, 2011).
Eucalyptus has shown high survival and rapid growth in mined areas (Schiavo et al., 2010), in addition to assisting in recovering organic matter from the mined soil (Banning et al., 2008). Planting Eucalyptus in rehabilitation is conditioned to the interest of the landowner in producing wood for economic purposes (Schiavo et al., 2010). It is known that runoff in areas without mining in Minas Gerais State is between 0.5 and 3.8% of rainfall in Eucalyptus plantations (Silva et al., 2011).
Runoff has been monitored in surface mines, mainly in China, the United States and Spain through fi eld experiments by installing plots in situ under simulated (Gomez-Gonzalez et al., 2016) and natural rain conditions (Merino-Martin et al., 2012), or by estimates using precipitation data (Liang et al., 2019;Taylor et al., 2009). However, only one study on the effi ciency of micro-damming in retaining runoff has been recorded in Brazil (Rubio et al., 2013). The lack of runoff monitoring in mining areas in Brazil may be related to the lack of requirements in legislation and procedures related to the environmental licensing process (Jeber and Profeta, 2018).
Therefore, the objective of this study was to monitor runoff in a rehabilitated area with clonal planting of Eucalyptus 2 years after interrupting a bauxite mining operation; to compare the runoff of 2.5 years with 3.5 years of rehabilitation; and to characterize the physical-environmental factors which interfere in runoff , seeking to discriminate those which can be improved.

Study area
The study was carried out in a rehabilitated bauxite mine (21° 3' 57.84" S, 42° 35' 42.38" W) in the municipality of Miraí, southeastern Minas Gerais, Brazil ( Figure 1A). The climate of the region is subtropical in altitude (Cwb) according to the Köppen classifi cation with two well-defi ned seasons, namely a rainy summer and a dry winter. Annual rainfall is 1,336 mm and average annual temperature is 19.3 °C (Alvares et al., 2013). The original vegetation type is Semi-deciduous Seasonal forest and a stratifi ed physiognomy of the Atlantic Forest which loses 25-50% of the canopy leaves in the dry period (Arruda et al., 2018). The predominant original soils are typical dystrophic Red-Yellow Latosol (Borges, 2013). The relief is rugged with hills and bauxite tops, forming part of the range of granulitic rocks called "Complexo Juiz de Fora" (Lopes and Adilson, 1990).
The mine of the rural property was leased by the mining company and the bauxite was extracted from July 2013 to July 2014. The ore removal was preceded by soil stripping with subsequent return to the topographic reconformation stage in August 2014. Revegetation was subsequently performed in December of the same year, following the operating standard of the mining company which is specifi c to the mined soils in the region (Lopes and Barros, 2017). The procedure starts with liming (4,000 kg ha -1 of dolomitic limestone) and phosphorus fertilization (2,000 kg ha -1 of reactive natural phosphate) to be carried out before planting. The clonal planting of Eucalyptus (hybrid AEC I144) at 2 x 3 m spacing was performed at a level with application of 300 grams of NPK (04-14-08) per seedling (Lopes and Barros, 2017). Brachiaria sp. was manually sown between the planting rows. The maintenance silvicultural treatments were replanting, covering fertilization and controlling ants (Lopes and Barros, 2017).

Surface runoff
A total of seven runoff collecting plots were installed in October 2016 in the experimental area inside the rehabilitated bauxite mine. Runoff was monitored from October 2016 to May 2018. The seven plots had dimensions of 10 x 6 m plus a triangular taper of 1 m in height from its base. The area and slope of the plots were: plot 1: 61.90 m 2 and 12.2°; plot 2: 61.96 m 2 and 12.7°; plot 3: 62.22 m 2 and 14.5°; plot 4: 62.04 m 2 and 12.9°; plot 5: 61.64 m 2 and 13.8°; plot 6: 61.73 m 2 and 15.1°; and plot 7: 61.46 m 2 and 15.4°. The plots were delimited by polyvinyl chloride (PVC) sheets and a triangular bottleneck was constructed with bricks at the lowest altitude (fl ow direction) for directing the water. A pipe was installed at the end of the triangular bottleneck to conduct the fl ow to a container with a capacity of 100 liters ( Figure 1B and 1D). The water volume drained in the plots was measured after rain events, with each collection consisting of one or more rains and then the collections for each month were added to obtain the monthly total. The runoff was calculated considering the runoff volume in relation to the contribution area of each plot (Equation 1):
The runoff was divided by the precipitation to obtain the runoff coeffi cient and multiplied by 100 to obtain the runoff in percentage (RP). Runoff was monitored in a calendar year (January to December 2017) and compared between 2.5 years (October 2016 to May 2017) and 3.5 years (October 2017 to May 2018) of rehabilitation, both monitored periods under natural rain ( Figure 1C).

Measurement of factors which interfere with surface runoff
Precipitation: The amount (P) and intensity (I) of precipitation were measured by a pluviograph (RainLog 2.0 RainLog 2.0 model from RainWise ® Inc) installed 50 m from the experimental area. The daily precipitation data (mm) were exported in an Excel spreadsheet and added to obtain the monthly total. The intensity data (mm.h -1 ) were used to obtain the maximum monthly intensities.
Vegetation cover: The plant stem circumference at the height of 1.30 m and the projection radius of the crowns were measured with a tape measure to calculate the diameter at breast height (DBH) and the crown area (Wink et al., 2012), and to characterize the Eucalyptus coverage in three plots previously selected at 2.5 (February 2017) and 3.5 years (February 2018) of rehabilitation.
Litter: Three litter samples at 2.5 (May 2017) and 3.5 years (May 2018) of rehabilitation were collected to check the litter water retention capacity (LWRC). A 0.25 m² square template was launched at three randomly selected points within a range 1.5 m away from the runoff collecting plots to collect the litter deposited on the soil surface. The saturated mass (after being submerged in water for 72 hours) and the dry mass (in an oven at 75 °C with forced air circulation) of the samples were measured to calculate the LWRC by Equation 2: Where: LWRC = litter water retention capacity (kg/kg); LSM = litter saturated mass (kg); LDM = litter dried mass (kg).
Physical properties of the soil: The texture, the overall density (Ds), the macro (ma) and microporosity (mi), the total porosity (TP) and the particle density (Pd) of the soil at 2.5 and 3.5 years of rehabilitation were analyzed to characterize their physical attributes (Teixeira et al., 2017).
Water infi ltration into the soil: The initial infi ltration rate (IIR), infi ltration capacity (IC) and the stable infi ltration rate (SIR) were measured in situ by the double ring infi ltrometer method using an IN2-W infi ltrometer from Turf-TecInternational © . The top layer of litter was removed to measure the water infi ltration rates into the soil. The tests were conducted until the infi ltration rate was constant in at least three consecutive measurements, reaching the stable infi ltration rate (SIR). Three trials were performed at 2.5 (May 2017) and 3.5 years (May 2018) of rehabilitation. Soil moisture at a depth of 0 to 20 cm was measured by the gravimetric method (Teixeira et al., 2017) on the respective test days and compared by the t-test (p≤0.05) before comparing the infi ltration rates. The infi ltration rate was calculated by the relation between the infi ltrated water column by time interval (Equation 3): Where: IR = Infi ltration rate (mm h-¹); h = height of the infi ltrated column of water (mm); t = time interval to infi ltrate the water column (hours).
The initial infi ltration rate (IIR) was the rate measured in the fi rst instant of the test. The infi ltration capacity (IC) was considered the maximum rate that the soil can absorb water after the infi ltration stabilized in a given time interval. The water slide height value was transformed into IC by Equation 4: Where: IC = infi ltration capacity (mm h -1 ); hAc = height of the accumulated infi ltrated column of water (mm); tAc = accumulated time interval (min).

Mechanical resistance of soil to penetration:
The mechanical resistance of the soil to penetration was measured using an automated SoloTrack PLG5300 penetrograph from Falker ® . Six observations were made to characterize the mechanical resistance of the soil to penetration at 2.5 (March 2017) and 3.5 years (March 2018) of rehabilitation. Soil moisture (0 to 20 cm) was measured by the gravimetric method (Teixeira et al., 2017) on the respective test days.

Statistical analysis
The data normality of the monthly runoff in percentage (RP), litter water retention capacity (LWRC), litter dry mass (LDM), physical characteristics of the soil, initial (IIR) and stable (SIR) infi ltration rate and infi ltration capacity (IC) was analyzed by the Shapiro-Wilk test and analyzes of variance were performed to verify the diff erences in these parameters between the 2.5 years with the 3.5 years of rehabilitation. The Scott-Knott test (p≤0.05) was applied to group similar soil mechanical resistance to penetration values in each rehabilitation period. Pearson linear correlation (r²) between runoff and precipitation was calculated to verify its infl uence on runoff (p≤0.05).
A principal component analysis (PCA) was applied to the dataset of the factors which interfere with surface runoff to check the interrelated variables and highlight the factors with the greatest infl uence. The factors used in the PCA were: precipitation (P), initial infi ltration rate (IIR), stable infi ltration rate (SIR), infi ltration capacity (IC), mechanical resistance of the soil to penetration (RSP), overall soil density (Ds), soil particle density (Dp), total porosity (TP), macroporosity (ma), microporosity (mi), vegetation cover area (VCA) and litter water retention capacity (LWRC).
The data were entered and processed in Microsoft Excel and the statistical analyses were performed using the R program (R Core Team, 2020).

Runoff and precipitation
The total precipitation in 2017 was 1,152 mm, generating a cumulative total of 3.3 mm of runoff in the area rehabilitated with Eucalyptus after bauxite mining, meaning that 0.29% of the rainfall was converted into runoff . The average monthly runoff percentage was 0.14%. The largest runoff occurred in November (2.1 mm), corresponding to 0.65% of the 317.4 mm of precipitation. The lowest precipitation (1.8 mm) and absence of the runoff were in September. The lowest runoff was between June and October, with an average of 0.01% of rainfall. These months were characterized by the lowest precipitation intensities of 0.63 mm h -1 (Figure 2A).
The average monthly runoff percentage at 2.5 years was 0.25% (± 0.26%) of precipitation and equal (p-value <0.05) to 3.5 years (0.15 ± 0.22 %) of rehabilitation. This runoff percentage varied between months and was higher in February (0.68%) and November (0.65%) at 2.5 and 3.5 years of rehabilitation, respectively ( Figure 2B). The runoff in the rainy season at 2.5 years of rehabilitation (October to May) was 5.0 mm (0.42%) for 1,193 mm of precipitation. Moreover, the surface runoff in the same months at 3.5 years was 2.5 mm (0.20%) for 1,292 mm precipitation. The maximum and minimum values of runoff occurred in diff erent months at 2.5 and 3.5 years of rehabilitation. The maximum  runoff (3.1 mm) at 2.5 years of rehabilitation was in December for a precipitation of 489 mm, while the minimum was in January with 30 mm of precipitation and 0.01 mm of runoff . The maximum runoff (2.1 mm) for 3.5 years of rehabilitation was in February, which constituted 0.64% of the rains in that month (317 mm). In contrast, the minimum runoff (0.001 mm) was in October, equivalent to 0.01% of the rains of 18 mm ( Figure 2C and 2D).

Factors which interfere with surface runoff
The vegetation cover area estimated by the projection area of the Eucalyptus crowns increased 27% from 2.5 years to 3.5 years of rehabilitation. The mean stem diameter increased from 10.94 cm (± 1.60 cm) to 12.09 cm (± 1.99 cm), and was equivalent to a basal area of 15.95 m² ha -1 and 19.55 m² ha -1 for 2.5 and 3.5 years of rehabilitation, respectively (Table 1).
The physical analysis of the soil showed that the clay content, soil density, particle density, microporosity and total porosity were equal between the rehabilitation periods. In contrast, the sand content increased and the macroporosity decreased at 3.5 years ( Table 1).

Correlation coeffi cients and principal component analysis
Positive correlations between runoff and precipitation and a reduction in this coeffi cient were observed from 2.5 (0.96; p-value 0.001) to 3.5 years (0.61; p-value 0.105).
The principal components analysis of the factors which interfered in the runoff at 2.5 years of age after bauxite mining explained 69.10% of the total variability in the fi rst two components, where 44.70% was explained by Component 1. Precipitation, rate and infi ltration capacity, macro and microporosity were the most explanatory factors in Component 1. Furthermore, soil density and vegetation cover had the highest eigenvectors in Component 2 (24.4%) ( Figure  4A). After 3.5 years of rehabilitation, the fi rst two components explained 75.90% of the total variability, while 39.40% of the variance was explained by Component 1 and 36.50% by Component 2. Precipitation, vegetation cover, soil density and total porosity, macro and micro porosity were the most explanatory variables in the 39.40% explanation of Component 1. Infi ltration rates and capacity were the most explanatory in Component 2 ( Figure 4B). grouping of resistances at 2.5 years (C) and 3.5 years (D) of rehabilitation. Similar letters indicate groupings of soil resistance between depths in the same treatment by the Scott-Knott test (p-value<0.05). Figura 3 -Resistência mecânica do solo à penetração das seis amostragens aos 2,5 anos (A) e aos 3,5 anos (B) de reabilitação; agrupamento das resistências aos 2,5 anos (C) e aos 3,5 anos (D) de reabilitação. Letras similares indicam agrupamentos da resistência do solo entre as profundidades no mesmo tratamento pelo teste de Scott Knott (p-valor <0,05).

DISCUSSION
The total annual rainfall was 408 mm less than expected in the climatological normal estimated from 1981 to 2010 for the region (1,560 mm) (INMET, 2018). The higher runoff in November can be explained by the greater amount and intensity of precipitation that occurred in that month, which reached 41.40 mm h -1 (Dourte et al., 2015). The accumulated runoff in 2017 and in the rainy season at 2.5 and 3.5 years of rehabilitation were lower than that found in Eucalyptus plantations with a slope of 17.6% in the Vale do Rio Doce region, Minas Gerais (1.2%) (Silva et al., 2011), and also for that reported in the rainy season in a Eucalyptus plantation with the same slope in a bauxite pre-mining area in southeastern Minas Gerais (0.56%) (Silveira, 2017). Surface runoff lower than that of non-mined areas may indicate that a greater amount of water is infi ltrating the mined soil, mainly due to the soil turning and high vegetation cover which traps the runoff , allowing greater water infi ltration into the soil (Merino-Martin et al., 2012).
The similarity of the runoff between 2.5 and 3.5 years of rehabilitation after bauxite mining can be explained by the rapid establishment of vegetation in the fi rst period (2.5 years) with more than 90% of the plots covered, which reduces runoff even in the early stages of rehabilitation (Carroll et al., 2004;Loch, 2000). Insignifi cant diff erences in runoff were also observed after vegetation was established in freshly mined plots in Central Queensland (Carroll et al., 2004).
The increase in the canopy projection area, diameter and basal area of Eucalyptus stem from 2.5 to 3.5 years of rehabilitation are associated with the growth dynamics of the plants which increases the occupied space over time (Wink et al., 2012). The diameter and basal area values are higher than those of Eucalyptus plantations via seeds in the region of this study (Silveira, 2017), and similar to the values of commercial plantations of the same age (Wink et al., 2012). The 27% increase in vegetation cover in one year is within expectations for the mining area, which could reach 60.26% in fi ve years, as recorded in China (Li et al., 2017). This high vegetation cover reduces runoff due to greater rainfall interception (Freitas et al., 2016;Zou et al., 2015) and decreased impact of raindrops on the soil (Armenise et al., 2018).
The greater accumulated litter stock after 3.5 years of rehabilitation can be explained by the larger size and growth of trees, which enhances the production and deposition of leaves in the soil (Mateus et al., 2013). The litter cover reduces the runoff speed, thereby facilitating rainwater infi ltration into the soil (Loch, 2000).
The statistically equal values of the initial infi ltration rate, infi ltration capacity and stable infi ltration rate between the rehabilitation periods (2.5 and 3.5 years, Table 1) can be explained by the similar humidity conditions and the physical properties of the soil, as reported in surface mines in Texas, United States, with infi ltration rates between 3 to 22 cm h -1 (Jarocki, 1994). Greater infi ltration occurs when the soil is dry and when texture and structure conditions have more empty spaces, constituting a characteristic which requires long periods for changes (Jarocki, 1994). The stable infi ltration rates at 2.5 and 3.5 years of rehabilitation are considered high when compared to natural soils in Eucalyptus plantations (78 and 165 mm h -1 ) and pasture (47 and 50 mm h -1 ) in Oxisols (Almeida et al., 2014;Silveira, 2017) and mined areas in Texas, United States (3 to 22 cm h -1 ) (Jarocki, 1994). The higher infi ltration rates in this study can be explained by the thick and porous topsoil layer which facilitates water infi ltration after mining (Huang et al., 2015) and by the presence of cracks and crevices resulting from the cross-subsoiling carried out during the soil preparation in rehabilitation (Haigh and Sansom, 1999).
The reduction in macroporosity and an increase in the sand content from 2.5 to 3.5 years of rehabilitation may be due to the reconstructed soil having a heterogeneous structure in the area generated by the disruption of soil horizons to remove ore (Li et al., 2014). The short period (one year) between sampling at 2.5 and 3.5 years of rehabilitation was not enough to modify the other physical soil characteristics (Table  1), as already reported in other surface mines (Chen et al., 2011;Ngugi et al., 2017).
The reconstructed soil reached the maximum penetrograph strength due to the presence of remaining saprolite concretions after mining (Silveira, 2017). These concretions result in a higher stony index from 20 cm deep due to the soil horizon restructuring after mining (Barros et al., 2013). The structural similarity of the soil explains the equal mechanical resistance values of the soil to penetration between 2.5 and 3.5 years of rehabilitation (Ngugi et al., 2017). The lower strengths up to 20 cm in depth are explained by the replacement of the topsoil layer removed before mining (Martín-Moreno et al., 2016).
The reduction in the correlation coeffi cient between runoff and precipitation at 3.5 years can be explained by the increase in vegetation cover. Vegetation intercepts rain drops, reducing its direct infl uence of precipitation (Freitas et al., 2016;Zou et al., 2015) and contributing to reduce runoff (Zhang et al., 2014).
The largest eigenvectors for infi ltration rate and capacity, macro and microporosity and soil density allow to infer that soil characteristics can be used to explain the equal runoff values (Santos et al., 2018). Except for macroporosity, these soil characteristics did not diff er in the interval of one year between the rehabilitation periods (2.5 and 3.5 years), demonstrating that they demand more time for modifi cation (Liu et al., 2017). Vegetation cover is another factor with a greater eigenvector in PCA and can explain the equal runoff values between the two rehabilitation periods due to its effi ciency in reducing runoff by interception after 2.5 years of rehabilitation (Vásquez-Méndez et al., 2010). Although the collections of factors which interfere with runoff were not carried out in the same month, it is believed that the diff erence of a few days between collections in the same period is not signifi cant (Liu et al., 2017), since there is a 12-month interval between collections for each period.

CONCLUSION
The runoff in the rehabilitated area after two years of bauxite mining showed the same pattern as unmined areas, with higher values in the rainiest months (November, December and February). The runoff was similar between 2.5 and 3.5 years of rehabilitation after bauxite mining, and less than the pre-mining runoff .
The physical soil characteristics showed insignifi cant changes between the two samplings, but the litter mass and its water retention capacity increased. Vegetation cover and soil characteristics, mainly infi ltration rates, were the determining factors for the runoff similarity in the two periods, indicating that the combination of adequate soil management and rapid plant cover provided runoff below 1% in a rehabilitated area after bauxite mining.

ACKNOWLEDGMENTS
We wish to thank to the "Conselho Nacional de Desenvolvimento Científi co e Tecnológico (CNPq)" for providing the fi rst author with a scholarship. The authors are also grateful to "Companhia Brasileira de Alumínio (CBA)" of Miraí, Minas Gerais, Brazil for providing the experimental area and the maintenance of the experimental plots.

AUTHOR CONTRIBUTIONS
Aline Gonçalves Spletozer: Conceived and designed the analysis, collected and analysis the data and wrote the paper. Lucas Jesus da Silveira: collected and analysis the data. Alexandre Simões Lorenzon and Aurora Yoshiko Sato: technical review. Herly Carlos Teixeira Dias: Conceived and designed the analysis and technical review.