ESTIMATES OF SOIL LOSSES IN WATERSHED UNDER TROPICAL OF ALTITUDE CLIMATE IN BRAZIL

model is a fast, simple, and inexpensive tool that contributes to the assessment of soil conservation in hydrographic subbasins.


INTRODUCTION
Soil is one of the most important and most complex natural resources, but current developments (climate change, soil erosion, and urbanization) increasingly threaten this valuable resource (Ayer et al., 2020;Spalevic et al., 2020;Chalise et al., 2019;Parsipour et al., 2019;Curovic et al., 2019.). Soils are essential for food production and various other ecosystem goods and services, including climate regulation and nutrient cycling (Greiner et al., 2017).
Soil degradation caused by erosion and rapid population increase is ranked among the most important environmental problems in the world Dimotta et al., 2017;Spalevic et al., 2016;Dimotta et al., 2016). Erosion is a key driver of land degradation, heavily affecting sustainable land management in various environments worldwide (Ouallali et al., 2019;Tavares et al., 2019;El Mouatassime et al., 2019;Nikolic et al., 2018;Spalevic, 2011). Water erosion is one of the main processes of tropical soil degradation and causes organic matter and nutrients losses, compromising the provision of soil ecosystem services (Olivetti et al., 2015;Bertol et al., 2007). According to Lal (2014), 1 billion hectares worldwide have already been affected by the erosion process, of which 70% are seriously committed to agricultural production. The worldwide annual rate of soil erosion from agricultural land ranges from 22 to 100 t ha −1 ; declines in productivity as much as 15-30% annually (Morgan, 2005).
Estimating soil loss and identifying hotspot areas support combating soil degradation (Girmay et al., 2020). Direct field measurements of soil erosion at permanent research or experimental stations using runoff plots with the known area, slope gradient slope length, and soil type could give reliable runoff and soil loss (Hurni et al., 2010) for experimental purposes, however, it is costly, laborintensive, and time-consuming (Alemayehu & Alamirew, 2012). Empiricalstatistical models were developed and improved to evaluate and quantify water erosion (Hazbavi et al., 2020;Amorim et al., 2010;Spalevic, 1999). These models include the Universal Soil Loss Equation -USLE (Wischmeier and Smith, 1978) and the Revised Universal Soil Loss Equation - RUSLE (Renard et al., 1997). RUSLE has a simple application and can be adapted to new geographical and edafoclimatic conditions. Moreover, combined this model with geographic information systems (GIS), it is possible to assess the spatial distribution of soil losses and identify areas most susceptible to erosion (Avanzi et al., 2013). Furthermore, soil losses can be compared with the soil loss tolerance (T) limits, which represent the maximum erosion rate that allows sustainable agricultural production (Wischmeier and Smith, 1978).
Coffee is the main agricultural crop in southern Minas Gerais, with economic and social prominence. Coffee plantations are concentrated in steepest areas, which are more vulnerable to water erosion. However, few studies evaluate the dynamics of the erosive process in these areas. Thus, the objective of this work was to evaluate the susceptibility to water erosion in steepest areas under predominant coffee cultivation using the RUSLE and compare the results to the T limit.

Study area
The research was carried out at the Ribeirão José Lúcio subbasin located in Conceição do Rio Verde Municipality, coordinated 473000 at 477000 m W and 7581000 at 7584000 m S, and the Ribeirão São Bento subbasin located in Cambuquira Municipality, coordinates 479000 to 484000 m W and 7570000 to 7575000 m S, Datum SIRGAS 2000, zone 23K UTM, both in southern Minas Gerais, at high altitudes (> 1000 m) in Serra da Mantiqueira, Brazil (Figure 1).
According to the Köppen system, the climate is classified as humid mesothermal, tropical of altitude subtype (Cwb), with an average temperature of 20° C and precipitation between 1,480 to 1,700 mm (Sparovek et al., 2007). In both municipalities, coffee production is the main economic activity (Figure 1). The elaboration of the land use map was made from the cartographic base and the crops mapped by Ipanema Agricola SA (Ipanema Coffees), and Landsat-8 Operational Land Imager (OLI) satellite images, bands 2, 3 and 4, corresponding to Orbit / Point 219/75 from Imaging Division (DIDGI) (INPE, 2019). The images were composting in ArcGIS 10.2 (ESRI, 2014), and the accuracy was verified in field surveys, with a 95% accuracy rate. Occupancy rates for land use classes are shown in Table 1. The altitudes range from 893 to 1,339 m and 849 to 1,096 m for the Ribeirão José Lúcio and Ribeirão São Bento sub-basins, respectively (Fig. 2). The digital elevation model (DEM), with spatial resolution of 12.5 m, was made from the contours extracted from the topographic map of Varginha (IBGE, 1979) and São Lourenço (IBGE, 1971), with the ArcGis 10.2 tool to Raster (ESRI, 2014).

Field sampling
Soil samples were collected based on land use and relief classes in 9 points at the Ribeirão José Lúcio subbasin and 18 points at the Ribeirão São Bento subbasin ( Figure 2). We collect three types of samples on the surface (0 to 20 cm) and subsurface (20 to 40 cm) soil layers: disturb, undisturbed by the clod method, and undisturbed with a cylindrical sampler (volume 92.53 cm³ and depth 5 cm).
The following analyses were performed: particle size distribution with and without NaOH (Bouyoucos et al., 1962 ;Blake et al., 1986); organic matter (MO) by oxidation with Na 2 Cr 2 O 7 2 mol L -1 + H 2 SO 4 5 mol L -1 ; pH with KCl and CaCl 2 -1: 2.5 ratio; sum of exchangeable bases (SB); soil density by the volumetric ring method; cationic exchange capacity at pH 7.0 (CEC-T) and effective cationic exchange capacity (CEC-t); aluminum saturation index (m), remaining phosphorus (P-rem), exchangeable Ca-Mg-Al with 1 mol L -1 KCl extractor, H + Al with SMP extractor; available phosphorus (P) by the colorimetric method using ascorbic acid; base saturation index (V%); flocculation index and water dispersed clay by the pipette method (Zhang, 1997); aggregate stability with weighted average diameter (MPD) and geometric mean diameter (DMG) calculation by wet sieving method and soil porosity with total pore volume calculation EMBRAPA (2011). The soil permeability variable was obtained in the field, from three replicates for each soil class with a Mini Disk Decagon Devices infiltrometer adjusted for the suction rate of 2 cm Zhang (1997). The moist color was visually classified according to the Munsell (2012) classification.

Revised Universal Soil Loss Equation
Soil loss rates at the study areas were calculated by the RUSLE (Equation 1) (Renard et al., 1997).

(Equation 1)
Where: A is a mean annual soil loss, Mg ha -1 year -1 ; R is the rainfall erosivity factor, MJ mm ha -1 h -1 year -1 ; K is the soil erodibility factor, Mg h MJ -1 mm -1 ; LS is the topographic factor expressing slope and ramp length (dimensionless); C is the factor for land use and management (dimensionless), and P is the factor for conservation practices (dimensionless) (Wischmeier and Smith, 1978).
The R factor was obtained from the rainfall erosivity map for the southern Minas Gerais state, with values ranging in the two areas from 5,145 to 7,776 MJ mm ha -1 h -1 year -1 with an average of 6,500 MJ mm ha -1 h -1 year -1 (Aquino et al., 2012). The K factor represents soil resistance to erosion. To Cambisol this parameter was calculated by the indirect method of Bouyoucos (1962) Table 3. Y = -3,89 x 10 -2 + 5,11 x 10 -3 X 14 -1,25 x 10 -2 X 15 + 5,41 x 10 -3 X 16 -7,27 x 10 -3 X 18 + 5,33 x 10 -2 X 33 + 3,21 x 10 -5 X 34 -5,66 x 10 -5 X 36 + 8,33 x 10 -4 X 2 -1,17 x 10 -2 X 4 + 1,53 x 10 -2 X 13 (Equation 3) Note: The description and parameter values of Equation 3 are described in Table  4. The values of the variables were obtained based on soil samples collected from the native forest. The LS topographic factor was estimated according to Moore and Burch (1986) in the ArcGIS 10.2 (ESRI, 2014) (Equation 4) from the DEM using the Raster Calculator tool. The model was efficient in determining LS, with higher factor values associated with steep slopes and more intense flows. The LS factor range from 0 to 238, with an average of 16.44 and 0 to 617, with an average of 7.28, for the Ribeirão São Bento and Ribeirão José Lúcio subbasins, respectively.
To determine the C and P factors, we consult the specialized literature. Areas with exposed soil present the highest C values, followed by eucalyptus cultivated down the hill, coffee, degraded pasture, facilities, indiscriminate floodplain soils, and native forest ( Table 5). The higher P factor was found in degraded pasture and exposed soils, while the lowest value found in the native forest was due to the dense vegetation cover (0.01). Coffee presents a P factor of 0.50 due to conservationist practices.  Roose (1977).

Validation
The validation of soil loss estimates was done by monitoring the annual sediment transport, according to Beskow et al. (2009). For this purpose, data of total solids in water and respective flow monitored from 1997 to 2010 by two hydrosedimentological stations operated by the Minas Gerais Institute for Water Resources Management (IGAM), located in the municipalities of Cambuquira (MG 473138 W and 7581539 S), and Conceição do Rio Verde (MG, 490706 W and 7572704 S). Afterwards, the annual sediment transport was calculated considering the flow of the sub-basins and the daily flow of data obtained from the National Water Agency (ANA).
The annual sediment transported was compare with the Sediment Delivery Ratio (SDR), which represents the eroded soil fraction that reaches the water bodies. The SDR value is determined according to Equation 5 Vanoni (1975).

Soil Loss Tolerance (T)
The T was calculated by Equation 6 (Bertol and Almeida, 2000).

(Equation 6)
Where T is the soil loss tolerance (Mg ha -1 year -1 ); h is the effective soil depth (cm), limited to 100 cm; r a is the ratio that expresses, mutually, the effect of the textural relationship between the horizons B and A and the clay content of the horizon A; m expresses the organic matter content in the 0 -20 cm soil depth; p is the soil permeability factor; Ds is the soil density (kg dm -3 ); and 1.000 is the constant that expresses the time period required to wear away a soil layer of 1,000 mm thickness.
Latosols and Cambisols of the study area present an effective soil depth (h) of 1000 mm and 800 mm, respectively. The other parameters were determined according to Bertol and Almeida (2000), using the soil analyses results. Both subbasins present a r a of 1 and an m and p of 0.7, with soil permeability classified as slow. Soil density of Latosols and Cambisols was, respectively, 1.23 kg dm -3 and 1.21 kg dm -3 for Ribeirão Jose Lucio and Ribeirão São Bento subbasins.

RESULTS AND DISCUSSION
The total soil loss of the Ribeirão São Bento subbasin was 1,032 Mg year -1 , while the Ribeirão José Lúcio subbasin presents a loss of 5,014 Mg year -1 . The sediment delivery ratio (SDR) was 0.045 and 0.38 indicating that 45% and 38% of eroded sediments in the respective Ribeirão São Bento and Ribeirão José Lúcio subbasins reach the water bodies. Thus, considering the SDR, the average soil loss estimated by RUSLE was 1.41 and 1.22 Mg year -1 ha -1 (Table 6).
According to Pandey (2007), errors smaller than 20% allow the validation of the water erosion models. Thus, the results generated by RUSLE illustrate the satisfactory efficiency of the method employed. Areas with exposed soil and steeper slopes have the highest rates of soil loss in both sub-basins (Figure 4). As expected, due to the greater fragility of Cambisols, the sediment generation rates in each class of land use showed that Cambisols are more susceptible to erosion compared to Latosols (Bertol and Almeida, 2000) (Table  7).
The soil loss rate estimated in the native forest was 0.01 Mg ha -1 year -1 , similar to Silva et al. (2007) that found soil loss rates range from 0.01 to 0.38 Mg ha -1 year -1 , in a native forest at the Rio Grande do Sul State. The low losses in native forests are due to natural conservation and the protection offered to the soil by the canopy of dense vegetation and litter.
Coffee areas presented average soil loss rates of 4.50 and 5.71 Mg ha -1 year -1 for the Ribeirão São Bento and Ribeirão José Lucio subbasins, respectively. The highest rates of erosion in coffee were found in young areas, when soil cover by the canopy of coffee trees is still low (Carvalho et al., 2007). The results obtained were lower than the values observed by Silva et al. (1999) (10.98 Mg ha -1 year) for Dystrophic Red-Yellow Latosol. These results are due to the conservation practices adopted in the coffee areas, with consequent lower P factor value (0.5).  (Table 7), due to the young age of the plants, which provides low canopy protection against erosion. The T limits determined for the Ribeirão São Bento and the Ribeirão José Lúcio subbasins were 8.3, 7.5, 7.1 and 6.7 Mg ha -1 year -1 , and 6.5, 8.5, 7.5 and 5.5 Mg ha -1 year -1 for the LVd1, LVd2, LVd3, and CX1, respectively. Ribeirão São Bento subbasin has 13.16% of the area with losses above T, while 7.9% of the Ribeirão José Lúcio subbasin area exceeded the T limits.
The T results obtained are below those found by Bertol and Almeida (2000) for Latosols from Santa Catarina State (10.62 to 12.50 Mg ha -1 year -1 ) and São Paulo State (9.60 to 15.00 Mg ha -1 year -1 ) according to Bertoni and Lombardi Neto (2012). This difference may be due to Bertol and Almeida (2000) method considering more attributes related to the soil formation factors in the T estimation.
Determining the T is quite difficult due to the difficulties in calculating soil formation rates. For this reason, soil properties, such as organic matter, water permeability in the soil and the textural relationship between horizons B and A, which indirectly reflect the rates of soil formation, are used to define T. Conceptually, every soil has a limit T, which is related to your training rate. Thus, T calculations are complementary to water erosion estimates and allow a more accurate assessment of soil degradation status Areas with soil losses above T should be prioritized in the adoption of conservation management practices, seeking to minimize water erosion, and ensuring the long-term sustainability of agricultural production. Better management practices such as terracing, level planting, and cover crops between the coffee lines could mitigate the erosion rates and decrease the runoff, consequently provide the conservation of watercourses, and improve the fertilizer use efficiency, which reduces the production costs (Bertoni and Lombardi Neto, 2012).
Considering the importance of coffee growing in high altitudes and steep areas in the south of Minas Gerais State, the results showed that the adoption of conservationist management practices provide low soil loss rates and contribute to the sustainability of coffee production. The studied subbasins presented distinct values of soil loss susceptibility but similar characteristics in the places most susceptible to erosion. The RUSLE model allowed the identification of areas with soil losses above the limits of T, especially in steep areas with coffee cultivation. Thus, it is an alternative tool for planning land use and management to promote sustainable agricultural systems.

CONCLUSIONS
Ribeirão São Bento and Ribeirão José Lúcio subbasins soil losses ranged from 0.01 to 28.45 Mg ha -1 year -1 , with an average of 1.41 and 1.22 Mg year -1 , respectively. The average soil loss in the coffee cultivation areas was 5.1 Mg ha -1 year -1 .
Revised Universal Soil Loss Equation modeling of water erosion showed higher losses rates in areas with steeper slopes and without conservation practices. The areas with soil loss above the tolerance limits should be a priority for the adoption of mitigation measures.
The RUSLE model is a fast, simple, and inexpensive tool that contributes to the assessment of soil conservation in hydrographic subbasins.