Short-term effect of a crop-livestock-forestry system on soil, water and nutrient loss in the Cerrado- Amazon ecotone

Soil, water, and nutrient loss by water erosion are among the main factors leading to land degradation, decreasing soil productivity and the provision of ecosystem services. The Cerrado-Amazon ecotone in western Brazil has suffered rapid land-use cover changes with impacts on soil erosion and land degradation. Despite the importance of the region for Brazilian agriculture and environmental conservation, studies on soil, water, and nutrient loss are still scarce. We tested integrated crop-livestockforestry (ICLF) as a sustainable agriculture management system for the Cerrado-Amazon ecotone region. A field experiment was established in the north of Mato Grosso state to quantify total soil, water, carbon and nitrogen loss during the rainy season in 2012-2013 in plots of integrated crop-forestry (ICF), pasture (PAST), eucalyptus plantation (EUC), no-tillage crop succession (CS) and bare soil (BS). Total soil, water, carbon and nitrogen losses in BS were, on average, 96.7% higher than in ICF, EUC, PAST, and CS. ICF had significantly lower water loss than CS, EUC and PAST. Total loss of carbon (4.3 428.2 kg ha-1) and nitrogen (0.3 29.2 kg ha-1) differed significantly among treatments. The production systems with tree components (EUC and ICF) and PAST showed reduced soil and nutrients loss compared to CS. Our results demonstrated that ICLF can avoid soil quality loss and thus improve agriculture sustainability in the Cerrado-Amazon ecotone.


INTRODUCTION
Accelerated land-use and land-cover change has dramatically impacted soil erosion, leading to severe losses of soil, water and nutrients and corresponding loss of productivity and soil health of soils (Sobral et al. 2015;Dollinger and Jose 2018). Soil degradation due to erosion is a serious problem and needs to be addressed in the 21st century, notably in developing countries of the tropics and subtropics (Lal 2001) where erosion is often increased by factors, such as the removal of forest cover, intense rainfall, high erodibility of geologically old and weathered soils and inappropriate soil management (Wantzen and Mol 2013).
There is a growing concern on how to reduce the impact of agricultural production systems on natural environments while, at the same time, assuring food security (FAO 2009;2014). In this context, intensive but sustainable production systems such as integrated crop-livestock forestry (ICLF) can reduce environmental degradation while maintaining or increasing crop productivity (FAO 2010). The ICLF is a sustainable production strategy that integrates crops, livestock, and forestry in the same area through intercropping cultivation, cultivation in succession, and crop rotation, aiming at synergistic effects among the components of the agroecosystem (Balbino et al. 2011;. The system is based on diversified animal and plant production designed to optimize biological cycles and residue input. ICLF contributes to the recovery of degraded areas, as well as to the maintenance and restoration of vegetation cover and the adoption of good agricultural practices (GAP), including no-tillage soil management (FAO 2010;Balbino et al. 2011). The production strategy of ICLF also seeks to improve the human social condition in rural areas, through promotion and generation of jobs and incomes, conforming production units to environmental legislation and the appreciation of environmental services offered by agroecosystems (Balbino et al. 2011).
Despite the potential of ICLF for agricultural intensification and sustainability, studies on the quantification of the agricultural, social and environmental dimensions of ICLF systems are still scarce. Soil erosion and further effects on soil quality are expected under no-till and ICLF systems in the medium and long-term. However, soil and water loss may occur in short periods after ICLF establishment (Rieger et al. 2016;Zolin et al. 2016), which is more critical due to lower soil cover.
In the Brazilian Amazon, soil erosion could dramatically reduce the resilience of disturbed tropical forests, favoring the substitution by alternative vegetation forms that are persistently vulnerable to erosion (Flores et al. 2019). The north of Mato Grosso state, in the southern Brazilian Amazon, has experienced significant changes in land use and cover, which can directly impact soil and water conservation, as well as the hydrological regime of the region (Zaiatz et al. 2018;Oliveira et al. 2019). In this context, ICLF systems can play a critical role in coupling sustainable agriculture with minimizing negative impacts on the environment in the agricultural frontier of the southern Brazilian Amazon (the ecotone between the Amazon and the Cerrado savanna biomes). Furthermore, policies intended to provide preferential credit lines to finance investments into the lowcarbon economy are a major factor driving this region of Brazil to adopt sustainable agricultural practices such as ICLF (Carauta et al. 2018;Schaldach et al. 2018).
Although there are multiple studies evaluating the impact of agricultural practices (e.g. tillage and no-tillage systems, cover crops, soil management) on water erosion in Brazil (Beutler et al. 2003;Silva et al. 2005;Barbosa et al. 2010;Merten and Minella 2013), data on the ability of ICLF to reduce soil, water and nutrient loss relative to other systems, especially in regions of accelerated agricultural expansion, are still lacking. Also, little information exists on the impact of soil, water and nutrient loss in ICLF systems in the agricultural frontier of the southern Brazilian Amazon, which is one of the most important for soy and corn, as well as livestock production in Brazil.
Therefore we aimed to assess the suitability of ICLF for sustainable agriculture management in the Cerrado-Amazon ecotone. We established a field experiment to compare soil, water and nutrient loss in ICLF relative to other land uses. Our specific goals were: (a) to evaluate if ICLF is more efficient than other conservationist systems to reduce soil, water, carbon and nitrogen loss in the short-term; and (b) to determine the magnitude of the variability in carbon and nitrogen content in eroded sediment from ICLF-covered systems and bare soil.

Site description
The field experiment was conducted at the experimental farm (11°51'50"S, 55°37'39"W, 364 masl) of Embrapa Agrossilvipastoril, in the municipality of Sinop, in the state of Mato Grosso. The climate of the region is of Aw (tropical with dry winter) type, according to the Köppen classification, with average annual temperature of 24.7 °C and rainfall of 1974 mm, concentrated from October/November to March/April (Souza et al. 2013) (Figure 1). The soil of the experimental site is classified as dystrophic Red Yellow Latosol (Oxisol) (Viana et al. 2015), a Typic Dystrophic Hapludox (Soil Survey Staff 2014). The soil is developed on a gently rolling landscape, on sand feldspar sediments dominated by kaolinite and gibbsite and is characterized by a clayey texture with a moderate A (0.00-0.20 m) horizon and a deep B (0.20-1.00 m) horizon. Texture of the surface horizon is determined by 32% sand, 12% silt and 56% clay (Viana et al. 2015). VOL. 51 (2)

Experimental design
Soil, water and nutrient loss was monitored in five treatments: (1) integrated crop-forestry system (ICF), consisting of eucalyptus (Eucalyptus urograndis clone H13 (obtained from a commercial nursery with standard size for expedition ranging from 20-30 cm) as the tree component (three rows with 3.5 x 3 m spacing and 30 m between rows), and soybean followed by corn (Zea mays L.), intercropped with Urochloa brizantha cv. Marandu under a no-tillage system as the crop component (in the 30-m spaces between eucalyptus rows); (2) Urochloa brizantha cv. Marandu pasture (PAST); (3) eucalyptus (E. urograndis clone H13) plantation (EUC) with 3.5 x 3 m spacing ; (4) crop succession of soybean followed by corn, intercropped with U. brizantha in no-tillage system (CS); and (5) bare soil (BS). Single systems (PAST, EUC, and CS) were evaluated in 1-ha experimental plots, while for integrated systems (ICF and CS), 2-ha plots were used. For each production system, as well as for BS, one sub-plot for runoff measurement was used ( Figure 2). The fertilizers applied during the experiment are listed in Table 1.
Corn and soybean were sown at row spacing of 0.45 m, perpendicular to the slope. Pasture was also sown at row spacing of 0.45 m, with 4 kg ha -1 of viable seeds. Pasture was managed without grazing until December 2012, when it was harvested to simulate grazing. Eucalyptus seedlings were planted in furrows (with a furrower) at 3.5 x 3 m spacing. Weed control in the EUC treatment was performed in early March 2013 by mechanical mowing in the plant interrow. The eucalyptus trees had a mean height of 3.5 m in November 2012 and 6 m in April 2013.
In November 2011, after subsoiling to alleviate compaction (chisel plowing to 40 cm depth) and a light soil-harrow operation, we started the experiment, and in October 2012 we started monitoring soil, water, and nutrients loss. Measurements of runoff and soil erosion were made from sub-plots of 132 m 2 , based on the standard plot of the USLE model (Wischmeier and Smith 1978). One sub-plot with a slope of approximately 1.5% was established in each treatment ( Figure 2). The sub-plot was enclosed at the top and sides by 6 x 22 m galvanized sheets inserted in the soil to a depth of approximately 0.15 m. The length of the sub-plot was parallel to the slope of the terrain. Runoff-collection flumes were  ACTA AMAZONICA installed at the lower end, where the surface runoff flow was directed via PVC pipes into two 1,000-L collection tanks. A Geib divisor with 11 slots was installed between the tanks to conduct 1/11 of the excess runoff from the first into the second tank.

Data collection and laboratory analyses
Water volume and eroded sediment samples were collected to determine carbon (C) and nitrogen (N) content immediately after rainfall events that caused surface runoff during the rainy season, from November 2012 to April 2013 (30 events). In order to quantify soil and water loss, runoff and sediment samples were collected from the water tanks. Coarse sediments were held in a 50-L cotton bag installed in the first tank. After the runoff present in the tanks was homogenized, aliquots were collected and filtered with quantitative filter paper to determine the suspended sediments. The filtered material and coarse sediments were dried in an oven at 60°C for dry matter determination (Rieger et al. 2016).
Sediment samples used to determine C and N content were air dried, ground by a mill to pass through a 0.106 mm sieve and analyzed by CHNS analyzer (Vario Macro Elementar Analysensysteme, Hanau, Germany). We considered only sediment-associated C and N loss, thus, C and N loss in solution associated with surface runoff was not considered. As no inorganic C was present in the samples, total C corresponded to total organic C content.
Sediments from each runoff event were analyzed for C and N content by the product between the nutrient content and Figure 2. Location of the study area in Mato Grosso state, Brazil and an aerial image of the study area in the experimental farm of Embrapa Agrossilvipastoril, municipality of Sinop, showing the experimental plots of different production systems for the evaluation of short-term soil, water and nutrient loss. Below is a schematic drawing of the water and sediment collector system installed in each experimental plot (black rectangles on the aerial image) for the measurement of runoff and soil erosion. ICF -integrated crop-forestry system; PAST -pasture system; EUC -eucalyptus plantation; CS -no-tillage system with crop succession; BS -bare soil. Credit: Gabriel Rezende Faria (aerial image) and the authors (drawing).
the soil mass and then pooled to determine the monthly loss. Prior to the rainfall season, 20 soil samples were randomly collected from the 0 to 0.05 m layer in each experimental plot. The samples were pooled into one composite sample per plot to quantify the average content of C and N present in the surface soil layer. The pre-rainfall content and the content in the eroded sediment were used to calculate the Enrichment Ratio (ER) by dividing the average nutrient content from the eroded soil (average of the 30 events) by the average nutrient content in the pre-rainfall soil (Hernani et al. 1999;Silva et al. 2005). An ER >1 implies increase the and ER < 1 implies depletion of the nutrient content in the eroded sediment (Hernani et al. 1999).

Statistical analysis
Each rainfall event was considered a replication for soil, water, and nutrient loss in each treatment. The data series of water, soil, and nutrient loss for 30 rainfall events were tested for normality and homogeneity of variance with the Shapiro-Wilks and Levene tests, respectively. As the variables did not follow a normal distribution and were not homoscedastic, the data were compared among treatments with Mann-Whitney U-tests (for independent groups) at a significance level of 5%. The relationship of the overall means of soil loss with ER values per treatment, and of the monthly values of rainfall with C and N loss, was analyzed using simple linear regression.

C and N content in the eroded sediment
The C and N content in the eroded sediment ranged from 15.24 to 137 g kg -1 and 1.05 to 10.4 g kg -1 , respectively. BS, CS and ICF had similar median values for C and N content in the eroded sediments, and lower variability than PAST and EUC (Figure 3). Contents in PAST and EUC were more variable and had higher median values (Figure 3), resulting in a greater range of values of C and N in the sediments.

Monthly and accumulated C and N loss
C and N loss in BS increased significantly between November and February, the period of intense rainfall (Figures 4, 5) and showed a strongly positive relationship with monthly rainfall (r 2 = 0.94 and r 2 = 0.95, respectively). No relationship was found between monthly rainfall and C and N loss for CS, EUC and ICF treatments. Similarly, C and N loss in PAST also increased until February, while the other treatments did not follow the same trend, with highest values for C and N loss in CS in February, probably due to the soybean harvest (on 06 Feb 2013) and sowing of corn (on 19 Feb 2013).
Over the entire evaluation period (November 2012 to April 2013), total loss of C and N in BS (428 kg ha -1 C and  29 kg ha -1 N) was significantly higher than in ICF (4.8 kg ha -1 C and 0.35 kg ha -1 N), EUC (6.8 kg ha -1 C and 0.46 kg ha -1 N), PAST (4.3 kg ha -1 C and 0.27 kg ha -1 N) and CS (20.5 kg ha -1 C and 1.34 kg ha -1 N). There was no significant difference in C and N loss among ICF, EUC and PAST, however, losses were significantly higher in CS (Table 2).

Enrichment ratio
With the exception of EUC and PAST, eroded sediments were not enriched with C and N (Table 3). ER was higher in EUC and PAST. In EUC, ER of C and N was 1.82, which means that the C and N content was 82% higher in the eroded sediment than in the corresponding soil. In PAST, C content in the eroded sediment was only 2% (ER = 1.02) higher than in the soil, while N content was 20% (ER = 1.2) higher. ER values for the remaining treatments were lower than 1. No relationship was found between ER of C and N in sediments and soil loss. VOL. 51(2) 2021: 102 -112 ACTA AMAZONICA

Soil and water losses
Our results confirm the findings of other studies in Brazil on the importance of soil cover for soil conservation. A meta-analysis of the Brazilian experience with runoff and soil-erosion plot-scale studies under natural rainfall revealed similar values of soil loss, respectively, 0.1 and 0.3 Mg ha -1 year -1 in southern and central-western Brazil under pasture and grassland, concluding that conservation practices reduce erosion to ratios of those measured under natural vegetation (Anache et al. 2017).
An assessment of soil and nutrient loss and cost of water erosion in southern Brazil using a 22.1 x 3.5-m plot with 10% slope, resulted in soil loss of 85.3 Mg ha -1 on bare soil and 9.6 Mg ha -1 on conventionally tillaged soil, respectively 25 and 2.8 times higher than soil loss in a no-tillage system (Bertol et al. 2017).
In an area of natural savanna in the Brazilian Amazon, a three-month study found that, in addition to the natural vegetation, pasture with Urochloa brizantha and corn plantation were more efficient in containing erosion through soil, water, organic carbon and nutrient loss than cowpea-bean plantation and bare soil (Souza et al. 2019). In the latter study, average soil and water loss were, respectively, 8.9 and 3.9 times higher in bare soil than in the other treatments, showing that the harmful effects of erosion were attenuated by the soil cover.
However, a medium-term study in the same region of the northern Brazilian Amazon from 1988 to 1992, showed that conversion of primary forest to pasture can increase soil erosion 7.5 fold, which can impact the regional and global social economy (Barbosa and Fearnside 2000). As deforestation increases in the Amazon, unfavorable vegetation and soil conditions can decrease rainwater infiltration and increase surface runoff and sediment yield, which could ultimately cause land degradation (Zhu et al. 2018) and affect the hydrological budgets of large watersheds. In this regard, deforested areas converted to pasture lands, are more prone to deliver water with higher solute concentrations generated by erosion (Chaves et al. 2008). A recent and extensive literature review showed that soil erosion is the main resilience drain in disturbed tropical forests in the Brazilian Amazon (Flores et al. 2019).
Similar results were observed in other tropical regions. The suspended sediment yield from catchments under cropland on the forest frontier in southwestern Ethiopia was found to be four times larger than the yield from similar catchments under forest, with average values of 17 and 4 Mg ha -1 , respectively, which was attributed to higher soil erosion rates under cropland (Kassa et al. 2019). In Southwest China, the conversion of tropical rainforest to rubber monoculture increased sediment yield from 0.041 to 11.54 Mg ha −1 , a  Table 3. C and N content (%) in the eroded sediment (C sed and N sed ) and in the soil surface layer (C 0.05m and N 0.05m ) used to calculate the Enrichment Ratio for carbon (ER C ) and nitrogen (ER N ) in water-carried sediments from runoff sub-plots installed in experimental plots of different production systems in Sinop (Mato Grosso state, Brazil) from November 2012 to April 2013. PAST -pasture; ICF -integrated crop-forestry system; EUC -eucalyptus plantation; CS -no-tillage with crop succession; and BS -bare soil.  (Zhu et al. 2018).

C and N content in the eroded sediment
Similar values of C and N contents in the soil surface layer were observed in all treatments, which can be explained by the fact that measurements were made during the second year after the experiment was established, and significant managementinduced changes in total soil C and N content usually occur in the long term (Silva and Resck 1997).
The higher variability of C and N content observed in PAST and EUC sediments is likely related, on the one hand, to surface residue inputs (mowing of pasture and eucalyptus interrow during the collection period) and, on the other hand, to fertilization inputs in these treatments. In a study on nutrient loss by water erosion in agricultural soils of contrasting organic matter management, the eroded C was mainly derived from fresh organic residue added to the soil, while N loss was primarily associated with soil organic matter loss (Shi and Schulin 2018), corroborating the strong positive correlation between C and N loss observed in our study.
Soil organic C and its related components are preferentially transported by water erosion due to their low density of and that their content is higher in the soil surface layer (Strickland et al. 2012;Shi and Schulin 2018). Thus, high contents of C and N in the sediment are not directly related C and N content in the soil surface layer, but to the preferential transport of light organic components, since the content of C and N in the eroded sediment were negatively correlated with soil C and N content in the 0-0.05 m of the original source soils. Similar results were also reported by Owens et al. (2002), Girmay et al. (2009) andMartinez-Mena et al. (2008).
In EUC, the trees were 3.5 to 6 m high during the monitoring period, in a fast-growing stage, when they immobilize a significant amount of resources (water, carbon, and nutrients), and the system is in the open phase, when the biochemical cycling is prevailing (Laclau et al. 2010). Thus, it is likely that the C and N content in the EUC sediments were more related to soil management (subsoiling and furrower) and fertilization than to a direct effect of the trees at this stage.
Thus, the adoption of conservation-effective measures is essential to minimize the effects of rainfall erosivity, and to protect the stock of soil organic carbon (Silva et al. 2005;Keesstra et al. 2016). In Vietnam, total soil C losses of 0.001 and 0.008% were observed in the 0 -0.10-m layer after a single intense rainfall event, when the lowest C loss was observed for bare soil and the highest for a planted forest (Janeau et al. 2014). In a study in the Mediterranean region, agriculture was responsible for 70% of the soil loss by erosion, but for only 45% of the total eroded C, and 55% of the total eroded C was derived from soils under forest (Nadeu et al. 2014).
Therefore, the management systems that contribute to greater amount of residues on the soil surface are prone to losing some of the organic material by runoff. A proper tillage system can improve soil organic carbon accumulation and benefit land restoration .
In a study comparing the effect of antecedent soil water on sediment-bound carbon and nitrogen loss in the North-American Atlantic coastal plain, higher C and N loss was observed under conventional than under strip tillage (Strickland et al. 2012). In a semiarid Mediterranean region with only 386 mm average annual rainfall, Ruiz-Colmenero et al. (2013) reported C losses of 0.02 and 0.06 Mg ha -1 yr -1 were reported, higher under conventional tillage that under grassland management (Ruiz-Colmenero et al. 2013). In southeastern Brazil, where average annual rainfall is 1530 mm, Silva et al. (2005) reported C losses from 0.05 to 1.51 Mg ha -1 yr -1 (average 0.40 Mg ha -1 yr -1 ) in a Rhodic Hapludox soil without vegetation cover.

Monthly and accumulated C and N loss
A decrease in C and N loss in BS was observed during March 2013, despite high rainfall, which was related to soil-surface sealing due to raindrop impact and the lack of organic material on the soil surface (field observation by the authors). The lack of plant residue on the soil surface reduces the selective character of erosion (Silva et al. 2005), affecting C loss patterns in our plots, and, consequently, N loss, which was directly related to C loss. The strongly positive relationship between rainfall and C and N loss in BS during the period of intense rainfall (November-February) can be explained by the direct impact of the absence of vegetation cover, combined with soil surface sealing, on the conversion of rainfall into runoff (Zuazo and Pleguezuelo 2008).
The lower monthly soil loss in PAST, ICF and EUC as compared with CS, was also reported by Russelle (1996), indicating the benefits of well-managed pastures for soil conservation. In these treatments, the high correlation observed between rainfall and C loss from November to February may be related to the fact that eroded C was preferentially derived from the fresh organic residue present in the soil surface, as pointed out by Shi and Schulin (2018).
In general, soil and nutrient losses are higher during the early stages of crop establishment because of the bare soil, resulting in higher impact of highly erosive rains (Silva et al. 2005, Leite et al. 2009, Rieger et al. 2016, which is corroborated by our observation of highest values of C and N loss in February for CS. Moreover, our results indicate that even after a short-term period, the ICF system can have a positive synergetic effect on soil, water, and nutrient loss when compared to CS, as after two years, we observed a reduced water loss in ICF relative to PAST and EUC.

AMAZONICA
The cumulative losses and average contents of C and N in eroded sediments in our treatments indicated that the highest total losses did not necessarily correspond with the highest contents of C and N in the eroded sediment. This trend was especially evident in EUC, which had the highest content of C and N in the eroded sediment, but a total loss similar to ICF and PAST. Total C and N loss is related to the magnitude of the soil eroded (Schick et al. 2000;Santos et al. 2007), which is relevant to the total losses recorded in our study. In this regard, Suescún et al. (2017) also observed higher soil and nutrient losses under crop systems when compared to pastureland and oak forest.

Enrichment ratio
The ER of C and N for eroded sediment in EUC was probably related to the surface residue inputs (weed control by mechanical mowing in March 2013), which contributed to increasing SOC in the surface layer (easily transported by the water runoff) and the fertilization inputs, as previously mentioned. The ER of N for eroded sediment in PAST may be attributed to the transport of organic material as a consequence of mowing of pasture during the collection period (February 2013), contributing to a greater amount of residues on the soil surface, which was also observed by Shi and Schulin (2018).
Results similar to those observed for EUC and PAST were reported by Dedecek et al. (1986) and Schick et al. (2000). However, no ER was observed for BS, NT and ICF, as also reported by Eltz et al. (1984).
The ER for nutrients and organic C is partly attributed to selective erosion, as sediments of small diameter or low density (i.e., clay and organic matter) are most likely transported by the runoff, when compared to sediments that remain in the soil (i.e., silt and sand) (Langdale et al. 1985). This size-selective process may result in higher nutrient contents in sediments than in the source soil . Also, antecedent soil moisture may play an important role in this process, increasing runoff and sediment erosion in tilled soils through slaking of newly-formed, nonwater stable aggregates, which increases the proportion of high carbon silt + clay particles eroded during runoff (Strickland et al. 2012).

CONCLUSIONS
In this study we provided relevant insights on the short-term effect of a crop-livestock-forestry system on soil, water and nutrient losses in one of the most important and vulnerable agricultural frontiers in Brazil. We observed that both C and N loss was lower for production systems with tree components compared with a no-tillage crop succession system, indicating their higher capacity to reduce soil loss throughout the rainy season. Our results support the notion that ICLF systems can provide benefits for the Cerrado-Amazon ecotone sustainable agriculture management, with reduced soil, water, and nutrient loss. Moreover, we showed that, even under a short-term period of establishment, a positive synergetic effect for soil, water, and nutrient conservation is obtained through ICLF. Our results should encourage further studies to increase scientifically sound information on soil, water and nutrient loss under different production systems, to support policymakers to seek more sustainable approaches for this crucial agricultural and environmentally sensitive region of Brazil.