Soil loss and runoff in southern Brazil in conservation systems: a long-term experiment

. Brazil has high soil loss rates due to its tropical and subtropical climate characteristics. In this sense, soil use and management practices may minimize such losses. In this study, experimental plots were monitored for eight years in southern Brazil during natural rainfall events. Treatments were as follows: I) bare soil under conventional tillage, II) barley-oat/ soybeans succession under conventional tillage, III) oat-lupine-vetch/ maize succession under conventional tillage, IV) barley-oat/ soybeans succession under no-tillage, and V) oat-lupine-vetch/ maize succession under no-tillage. Data on rainfall amount and erosivity indices (EI 30 , EI 20 , and EI 10 ) were subjected to regression analyses, evaluating the effects of climatic variables on soil and water losses. We could also analyze crop species, conventional and no-tillage systems, and winter and summer seasons. The highest soil and water losses occurred in treatments under conventional tillage. In addition, soybeans treatments increased soil and water losses. Maize under no-tillage was the most efficient practice, as it had soil losses (3.4 times) and water losses (1.5 times) lower than soybeans under conventional tillage. Rainfall amounts and EI 30 were better correlated with soil and water losses in conventional tillage treatments. However, EI 10 could better explain the soil losses from the conservationist treatment.


Introduction
Erosion is considered a major global threat to soil and water degradation (Borrelli et al., 2020), especially in tropical regions, as it causes soil particles to detach and be transported more easily (Miranda, Scarpinella, Silva, & Mauad, 2015).Research on water erosion has been often carried out in delimited plots, whose surface runoff and soil losses are monitored after each rainfall event (Strohmeier, Laaha, Holzmann, & Klik, 2016;Anache, Wendland, Oliveira, Flanagan, & Nearing, 2017;Yang et al., 2019).In Brazil, the first experimental studies at a plot scale were conducted in the 1940s.However, since the peak of studies on soil erosion in the 2000s, field experiments have decreased by roughly 86% (Anache et al., 2017), and about 50% of experimental studies have only two or fewer years of monitored data.Between 1990 and2001, research in this field focused primarily on the USA and Europe, then expanding to China and Australia, and finally reaching South America and eastern and southern South Africa (Zhuang, Du, Zhang, Du, & Li, 2015).
Experimentation for several years is crucial due to climatic phenomena that have frequently occurred in southern Brazil over the last years (Grimm, Almeida, Benetti, & Leite, 2020;Nóia Júnior, Fraisse, Karrei, Cerbaro, & Perondi, 2020).In recent years (15 to 20 years), the frequency of extreme events has increased, with alternating periods of drought and flood.The phenomena known as El Nino and La Nina have intensified in recent years.Between 2009 and 2010, a large volume of rainfall was observed, generating heavy flooding and reductions in crop yields (Theisen, Vernetti Jr., & Silva, 2009).In the 2011-2012 crop season, according to data from Empresa de Assistência Técnica e Extensão Rural [EMATER] (2012), crop yields in Rio Grande do Sul (RS) were severely affected by the lack of water in the soil (Cunha et al., 2011).And more recently, in the 2021/2022 crop season, RS again experienced a severe drought, which was similar to that in the 2004-2005 crop season, which was considered the worst drought in RS according to meteorological records.

Study site
The experiment was conducted for eight years (1977 to 1985), monitoring one of the research centers of the current Department of Agricultural Diagnosis and Research (DDPA), formerly known as the State Foundation for Agricultural Research (FEPAGRO/ Floresta), part of the Department of Agriculture, Livestock, and Irrigation (SEAPI).It is located in the district of Boca do Monte, in the municipality of Santa Maria (Figure 1).Santa Maria is in the physiographic region of the central depression, in Rio Grande do Sul State, southern Brazil (29º30'45" S, 54º48'15" W, and 95-135 m altitude; Figure 1).According to Köppen's classification, the region has a Cfa humid subtropical climate without droughts.The average annual rainfall was 1578 mm, according to daily rainfall data provided by the National Institute of Meteorology (INMET), considering the period 1962-2012.The soil was classified as a Ultisol with low natural fertility, low organic matter content, and a sandy-clayey-loamy texture (IUSS Working Group WRB, 2015).

Experimental design and treatment characterization
The experiment was carried out with five treatments and two repetitions each, totaling ten experimental plots (22.0 × 3.5 m), on a 9% slope terrain.The plots were delimited by galvanized sheets to a depth of about 10 cm on the sides and upper end, with a drainage collection system at the lower end, which consisted of a gutter and two collection tanks.Coarser sediments were collected in the first tank (reservoir 1), while thinner ones were in the second tank (reservoir 2).
The treatments comprised: i) bare soil under conventional tillage.The soil was tilled twice a year, as plots were prepared with annual crops.Weeding was performed throughout the year using hoes or manually removing spontaneous vegetation to keep the soil bare.The standard plot, bare soil under conventional tillage, was referred to as SB; ii) oat-lupine-vetch/ maize succession under conventional tillage.In the first two years, oat-maize succession under conventional tillage with one plowing and two harrowings towards the slope.After five years, vetch (Vicia sativa) replaced white lupine as a winter crop.Crop substitutions during the experiment occurred due to phytosanitary issues.Oat-lupine-vetch/ maize succession under conventional tillage was referred to as Maize CT; iii) barley-oat/ soybeans succession under conventional tillage.The crops were harvested manually with a sickle, and straw was spread over the plot before tilling.After two years (1980)(1981), oat (Avena sativa) was replaced with barley as a winter crop due to phytosanitary issues.Barley-oat/ soybeans succession under conventional tillage was referred to as Soybeans CT; iv) oat-lupine-vetch/ maize succession under no-tillage.This treatment followed the same cropping protocol as in treatment 2, and differed regarding minimum tillage at sowing, characterizing direct sowing.Oat-lupine-vetch/ maize succession under no-tillage was referred to as Maize NT; and v) barley-oat/ soybeans succession under no-tillage.It followed the same cropping protocol as in treatment 3 and differed regarding minimum tillage at sowing, also characterizing direct sowing.Barley-oat/ soybeans succession under no-tillage was referred to as Soybeans NT.

Crop treatments
The soil was fertilized in October 1977 as recommended by the Soil Analysis Laboratory of the Department of Agriculture (Eltz, Cassol, Guerra, & Abrão, 1984).It was tilled towards the slope and harrowed to incorporate lime and corrective fertilizers.Another tilling was carried out with two more harrowings towards the slope, followed by sowing maize and soybeans under a conventional tillage system.
In all treatments, weed was controlled using applications of paraquat and glyphosate herbicides before crop sowing or whenever needed, along with manual weeding.The plots and freshly cut oat were harvested manually.For oat, two cuts were made per cycle, with the green mass being cut and removed from plots to simulate grazing.Grains from the remaining crops were crushed, and straw returned to the soil after being manually chopped using a machete, being incorporated in conventional tillage treatments, and merely distributed superficially in the no-tillage treatments.

Rainfall monitoring
Rainfall data were obtained using a pluviograph that was located at plots to calculate the amount, intensity, and rainfall erosivity indices (EI 30 , EI 20 , and EI 10 ), according to criteria proposed by Wischmeier (1959) and later adapted for regions with tropical climate by Cabeda (1976).

Runoff and sediment monitoring
Surface drainage collection systems were installed at the lower ends of the plots and consisted of gutters connected by PVC pipes to tank 1 (300-500 L capacity).When the collected volume reached the tank capacity, the surplus finest material was transferred to tank 2 (400-600 L capacity).Tank 2 was connected to tank 1 by a Geib-type flow divisor, which allowed the passage of 1/7, 1/9, or 1/11 of the total runoff, discarding the remaining runoff.Runoff volume was measured by reading the water level in the tanks.Sediment samplings were taken to quantify soil losses in each natural rainfall and erosive runoff event.The samples were taken to the laboratory for weighing, drying, and final quantification of solid mass (g).Subsequently, the remaining material in tanks 1 and 2 (composed of suspended sediments) was homogenized again to sample the suspended sediments.These samples were collected using 300-mL vials, placing two for each treatment per tank (1 and 2).
The vials were weighed, and Kal (SO 4 ) 2 was added to each one to precipitate sediments and determine suspended sediment volumes.After 48h, the supernatant liquid was removed from the vials using a siphon with a plastic hose, leaving about 1.0 cm of water covering the sediment.The vials were then taken to a ventilated air-forced oven at 55 -60ºC until reaching a constant mass.
Some soil and water samples were collected in more than one rainfall event.Thus, rainfall and rainfall erosivity data were aggregated.A total of 276 individual and aggregate rainfall events were analyzed, of which 139 refer to the winter period (June to November) and 137 to the summer period (December to May).

Erosivity indices
Based on the rainfall intensity, we could calculate the kinetic energy (KE) of each individual and erosive rainfall event according to Equation (1) of Foster, McCool, Renard and Moldenhauer (1981) for each rain segment.The erosivity indices (EI 30 , EI 20 , and EI 10 ) were calculated based on the method proposed by Wischmeier and Smith, (1958).EI 30 was used to calculate the maximum rainfall intensity for 30 consecutive minutes.Likewise, EI 20 and EI 10 were calculated using the maximum rainfall intensity in 20 and 10 minutes, respectively (Equation 2) (Wischmeier & Smith, 1958).KE = 11.87 + 8.73LOG 10 I (1) where in: KE is the kinetic energy expressed in J m -2 mm -1 ; I is the rain intensity of each segment in mm h -1 ; EI 30 , EI 20 , and EI 10 are expressed in MJ mm ha -1 h -1 ; and I 30 , I 20 , and I 10 are the maximum intensities at 30, 20, and 10 min., respectively (expressed in mm h -1 ).

Water determination
Where in: Q runoff is the runoff volume for each rainfall event (mm), and rainfall is the rainfall amount (mm).

Data analyses
Data between the years 1978-1986 were used.The first year (1977) was excluded since it was when all experimental plots were tilled.Averages were grouped into periods referring to each agricultural year, between tillage and sowing of winter/ spring crops (oat, barley, and vetch) and summer/ autumn crops (soybeans and maize), and tillage for the next crops.The data were separated into winter/ spring periods (called winter periods) and summer/autumn periods (called summer periods).
With the 8-year series of data, simple regression analyses were performed between (I) soil losses (t ha -1 ) and rainfall amount (mm), (II) soil losses and 30-min.rain erosivity index (EI 30 , MJ mm ha -1 h -1 ), (III) water losses (mm) and rainfall amount (mm), and (IV) water losses (mm) and 30-min.rain erosivity index (EI 30 , MJ mm ha -1 h -1 ).These analyses were performed per (agricultural) season and divided into summer and winter periods due to their climatic differences.Subsequently, simple regression analyses were also carried out between (V) soil losses (t ha -1 ) and rainfall erosivity indices (EI 30 , EI 20 , and EI 10 , MJ mm ha -1 h -1 ), and (VI) water losses (mm) and rain erosivity indices (EI 30 , EI 20 , and EI 10 , MJ mm ha -1 h -1 ).The analyses (V) and (VI) were also performed per (agricultural) season, but disregarding winter and summer differences, distinguishing the treatment referred to as standard plot (BS) from those where maize and soybeans were included as summer crops.This way, we could effectively analyze the effects of both crops on the results and of conventional and no-tillage systems on the relationship between the different indices and soil and water losses.

Rainfall analysis between 1978 and 1986
The annual average rainfall was 1,673.1 mm (Table 1), which is close to the historical average for Santa Maria (1,722.0mm).Erosivity indices (EI 30 , EI 20 , and EI 10 ) are also detailed in Table 1, reflecting how total energy and peak intensity interact in a particular rainfall event.In total, 839 rainfall events occurred in the 8-year series, considering erosive and non-erosive rainfall, 421 in the winter, and 418 in the summer.Erosive rainfall totaled 370, with 168 in the winter and 202 in the summer.Unsurprisingly, when comparing the three indices, EI 10 generated the highest values per period (winter and summer).Within the eight years, the R factor was 8,486.5 MJ mm ha -1 h -1 (Table 1), oscillating between 8,000.0 and 10,000.0MJ mm ha -1 h -1 year -1 for the 1980-2013 period (Almagro, Oliveira, Nearing, & Hagemann, 2017).Figure 3 describes the monthly averages for rainfall and erosivity indices.The months with the highest rainfall amounts may unnecessarily have the highest yearly erosivity.July had the highest volume of rainfall (190 mm), while October, February, and April showed the highest erosivity indices (EI 30 , EI 20 , and E 10 , respectively) (Figure 2).In Rio Grande do Sul State, the most significant rainfall erosivity was found in periods of low soil cover.In October (the month with the highest EI 30 ), the soil tillage and sowing of the main commercial crops, such as maize and soybean, occurs (Panizzi, 2013).

Effects of soil management and cover on runoff
Runoff decreases, on average, only 1.5 times from CT to NT treatments (Table 2).This result reveals that NT treatments partially controlled runoff because the efficiency in reducing runoff was lower than in reducing soil loss, which was reduced by 3.4 and 2.8 times for maize and soybeans, respectively, from CT to NT treatments (Table 2).Table 2 shows the runoff coefficients -C (%), which were obtained as the ratio between runoff and rainfall.The highest C values were found in CT treatments, with the highest one under soybeans (25.6%) (Table 2).When rainfall intensity increased, the effects of soil management decreased.Besides, other soil features, including bulk density, had a more substantial influence.In a simulated rainfall trial with high intensity (120 mm h -1 ), Lemos, Cassol, and Barros (2020) obtained a C value of 77% for maize under NT, which was about 1.71 times higher than that for CT.

Effects of rainfall amount and EI 30 on runoff
Table 3 displays the regression analyses between runoff and rainfall amount (PPT) and EI 30 .The analysis for the summer period was best explained by BS by PPTR (R² = 0.72).The summer PPT also revealed a good relationship with maize and soybeans under CT (Table 3), unlike NT, where winter rainfall amount was most related to runoff for maize and soybeans (Table 3).This result might be due to soil management and rainfall characteristics during winter.In other words, voluminous and prolonged rains, characteristic of the winter period in Rio Grande do Sul State, added to low evapotranspiration and the presence of soil cover may have kept soil moisture closer to saturation.Thus, rains of lower intensity, but that can be voluminous, generate surface runoff in conservation systems.When runoff was related to EI 30 , the same results were found for CT concerning rainfall amount (Table 3).However, the results are different for NT, as erosive rainfall occurring in the summer best explains its runoff (Table 3).The coefficients of determination improved for rainfall amount and EI 30 when runoff was related to EI 30 , both in NT and CT systems.
EI 30 best correlated with runoff in all CT and NT treatments (Table 3).In Jiangxi (China), Chen, Liang, Zhang, and Zhang (2020) monitored 66 rainfall events in experimental plots and reported that soil and water losses increased significantly as I 30 increased in treatments with bare-soil, leveled, slope-direction, and mulched crops.
Nevertheless, the results were different when the winter and summer periods were evaluated separately (Table 3).EI 10 was the best to explain runoff in NT treatments in the summer, matching the soil loss results (Table 5, Figure 4).When comparing the KE effects of rainfall with I 30 on the hydrological variables peak flow and runoff depth during 81 rainfall events, KE explained these variables in a small rural catchment (Ramon, Minella, Merten, Barros, & Canale, 2017).

Effects of soil management and cover on soil losses
Table 4 lists the soil loss averages in agricultural years for each treatment.Bare soil (BS) had the highest soil losses every year, with an annual average of 330.8 t ha -1 , which was already expected due to tilling in this treatment and the absence of protective cover crops.In conventional tillage (CT) treatments, soil losses were also more significant than in no-tillage (NT) treatments.On average, soil losses decreased by 3.4 times from Maize CT to Maize NT succession treatments (Table 4).As for soybeans succession, losses decreased by 2.8 times from CT to NT (Table 4).These results demonstrate the beneficial effects of conservation systems in controlling soil erosion (Didoné et al., 2017;Le Gall et al., 2017) and corroborate the literature regarding their efficiency for soil losses in subtropical climate regions (Bertol et al., 2008).
The highest averages of soil losses in NT occurred in the agricultural years 1983-1984, with 107.4 and 103.5 t ha -1 for maize and soybeans, respectively (Table 4), five years after experiment installation (considering 1977 as year zero).This fact can be explained by a liming operation incorporated with one plowing and two harrowings in all experimental plots, according to recommendations so far known, including limestone incorporation every five years (Eltz et al., 1984).However, the effects of conservation systems could be evaluated, since soil losses were much lower in the 1982-1983 agricultural years than in the following year, even with more significant rainfall volumes and erosivity among all study periods (Tables 1 and 4).
In addition to the difference in soil losses, when comparing CT and NT, it is noteworthy that soil loss was higher in treatments where soybeans were introduced as summer crops instead of maize (Table 4).This outcome can be attributed to the different contributions and quality of residues added to the soil both for winter and summer crops.Soybeans have a decomposition rate that exceeds that of grasses, which is characteristic of leguminous species known for their low plant C/ N ratios (Rossi, Pereira, Giácomo, Betta, & Polidoro, 2013;Acosta, Amado, Silva, Santi, & Weber, 2014).This feature helps quickly reduce soil cover and protection by soybeans residues, baring the soil and leaving it more susceptible to erosion.Notwithstanding, maize crop residues are more persistent because of their higher carbon, cellulose, and hemicellulose contents, which is typical of species of the Poaceae family (Redin et al., 2014), reducing decomposition rates (Watthier, Peralta Antonio, Gomes, Rocha, & Santos, 2020).

Effects of rainfall amount and EI 30 on soil losses
Table 5 demonstrates the soil losses in the different treatments and their relationship with rainfall amounts (PPT) and EI 30 .PPT and EI 30 could explain soil losses in BS.By isolating the individual effects of each period (winter and summer), both rainfall depth and EI 30 evidenced a more substantial relationship in the summer, with r 2 values of 0.81 (BS x PPT) and 0.75 (BS x EI 30 ), respectively (Table 5).BS, unlike the others, had no vegetation cover at any time of the year.This way, no physical barriers protected the soil against raindrop impacts, breaking particles down and starting the erosion process faster (Carvalho, Eduardo, Almeida, Santos, & Sobrinho, 2015;Panachuki, Bertol, Sobrinho, Oliveira, & Rodrigues, 2011).Soil losses in CT, both under soybeans and maize crops, were better explained by PPT in the winter; however, it could not explain soil losses in NT, which had a low r 2 , even when winter, summer, and agricultural seasons were isolated (Table 5).When relating soil loss to EI 30 , it better explained the dependent variable (soil loss) in CT than did PPT, which explains the soil losses under NT (Table 5), mainly when isolating winter and summer periods.In other words, erosive rainfall occurring in the summer, for both soybeans and maize, could better explain these losses (Table 5).Notably, in treatments under CT, soil losses were better described by EI 30 in the winter for maize crops and in the summer for soybeans (Table 5).Ramon et al. (2017) reported that I 30 is limited in explaining sediment yields, while its concentration (suspended sediments concentration) shows a better relationship with this variable (I 30 ) in a small rural catchment in RS, Brazil.Nevertheless, the rainfall amount variable better -elucidated sediment yield (Ramon et al., 2017).
EI 30 is globally recognized and used to relate rainfall energy with soil losses (Alewell, Borrelli, Meusburger, & Panagos, 2019).Conversely, this study revealed that depending on soil use and management, EI 10 and EI 20 are better suited to explain soil losses, as they are more related to the dependent and explanatory variables (Table 5, Figure 3).EI 30 was the best to explain soil losses in BS, as it was developed based on these plot conditions (Table 5, Figure 4A).Between 2015 and 2018, the maximum rainfall intensity in 30 min (I 30 ) was considered the best indicator of soil losses in bare soil and runoff depth (Chen et al., 2020).Apart from BS, CT treatments with vegetation cover (i.e., maize and soybeans -CT) also showed the best relationship between soil losses and EI 30 (Figure 4B and C,respectively).
When the results of NT with maize (Figure 4B, Table 5) and soybeans (Figure 4C, Table 5) were observed, EI 10 was the best explanation for soil losses.These findings corroborate (Stocking & Elwell, 1973), wherein, for plots with vegetation, EI 5 and EI 15 were better correlated with soil losses than EI 30 .These results were also found when analyzing the periods (Table 5).We observed the same behavior: EI 10 during the winter better explained soil losses regardless of the crop for NT.However, the results were different for the summer period, when EI 20 was the best to explain NT treatments, in addition to the two treatments under BS and Maize CT (Table 5).Evaluating the effects of different rainfall intensities (10, 20, 30, 40, 50, 60, and 90 min.)on soil losses in experimental plots with an average 18 -23% slope and soil covered by spontaneous vegetation, the maximum intensity in 20 min.was better related to soil losses for 12 erosive rainfall events (Mohamadi & Kavian, 2015).

Conclusion
Conservation systems based on soil cover (disregarding crop rotation) and no-tillage have been efficient in controlling soil losses over the years.This result is mainly based on studies in experimental plots with natural and simulated rainfall.Nonetheless, the same efficiency was not observed in controlling runoff, as demonstrated by our findings.In other words, soil loss decreases, on average, by 3.4 times from oat-lupinevetch/ maize succession under conventional tillage to the same succession under no-tillage.In the barleyoat/ soybeans succession, losses decreased by 2.8 times from conventional to no-tillage.Runoff decreases, on average, only by 1.5 times from conventional to no-tillage.When rainfall amount and erosivity index (EI 30 ) were related to soil losses, both explained them for bare soil and treatments under conventional tillage, both for summer and winter rainfall events.However, plots under no-tillage were better correlated with winter rainfall, both in energy (EI 30 ) and volume (mm).Moreover, different soil management and cover types were observed by evaluating rainfall energy through erosivity indices (EI 30 , EI 20 , and EI 10 ) under annual soil losses.That is, EI 30, as expected, is the variable that best describes soil losses for bare soil and conventional tillage.However, EI 10 best explains soil losses in no-tillage systems.For runoff, less rainfall intensity is required for starting, showing that EI 30 was the best related to an annual runoff under both conventional and no-tillage systems.

Figure 1 .
Figure 1.Location of the study area in the city of Santa Maria, Rio Grande do Sul State, Brazil.

Figure 3 .
Figure 3. Influence of erosivity indices on runoff in the bare soil plot (A) and NT and CT treatments, evaluated with maize (B) and soybeans (C).

Figure 4 .
Figure 4. Influence of erosivity indices on soil loss in BS plot (A) and NT and CT systems, evaluated with maize (B) and soybeans (C).

Table 3 .
Equation fits from runoff regression analysis.

Table 4 .
Annual soil losses in each treatment in t ha -1 .

Table 5 .
Equation fits from regression analysis Soil loss.