Effect of tillage on soil erosion before and after rill development

2.1 Study area

We used the soil and water conservation engineering laboratory in Yangling (34°17′56″N, 108°04′07″E) to carry out our rainfall simulation experiments. This station is on the southern margin of the Loess Plateau in China and has been used to study soil erosion for over 10 years. The soil in this area is classified as Eum-Orthic Anthrosol. Although the soil in this area was originally the cinnamon soil (Cinnamon series) that develops in the forest and bush, it has gradually formed a special agricultural soil through long-term cultivation and application of manure. The soil parent material is the alluvial loess of the quaternary sediments (Pan, 1961); soil texture is silt clay loam, total N and P are 0.91 and 0.50 g/kg, respectively, and the cation exchange capacity is 18.47 cmol/kg (Table 1).

Experiments were performed in three soil boxes, each made of steel plates and measuring 2 m length × 1 m width × 0.5 m depth. Each soil box had a V-shape outlet at its downslope end for collecting surface run-off and sediment samples from the box during a rainfall. The bottom of the box had four screened 1-cm diameter holes to drain infiltration water. The slope of the box was adjusted by using a screw jack.

2.2 Soil preparation in boxes

Before being placed in the boxes, the soils were sifted through a 10-mm sieve and were then air dried in the laboratory. The final soil moisture content was approximately 7%. Each box contained 40-cm deep air-dried soil (deposited in successive 5-cm-thick layers) with a bulk density of 1.30 g/cm³, similar to that of natural soil. After each box was filled, the soil surface was smoothed using a wooden plate.

Following completion of the 90-min rainfall simulation (described below) in the three treatment soil boxes, the top 20 cm was removed, and the remainder was allowed to dry for 3 days using fans. The top 20 cm of soil was then prepared using the same procedure for the first preparation as described above. This process ensured a similar moisture content in the soil boxes for each run. All experiments were done in triplicate.

2.3 Tillage preparation

Two agriculture tillage techniques, CT, which produces ridges and furrows, and RT, which produces depressions and mounds, were used to prepare the surface microtopography of the experimental boxes. One box was used for each tillage method. CT and RT are commonly used in China for agricultural production on sloping land. The control was a SS. Two types of farm tools, a chisel plow and hoe, were used by an experienced tillage operator to turn the soil in the soil boxes to simulate CT and RT treatments, respectively. For the CT, the ridges were 7–10 cm in height with a between-ridge distance of 25–30 cm. For the RT, the depth of depressions was 5–8 cm with a between-depression distance of 15–20 cm. The experimental slope for all treatments was 15°, which is the common gradient of agricultural land in this area.

2.4 Rainfall simulation and data collection

We used a side-sprinkle rainfall simulator as described by Zhao et al. (2014). The simulator was located 6 m above the ground, and eight evenly distributed calibration gauges showed that rainfall uniformity was greater than 90%, which is higher than that of the local natural rainfall as reported by G. Y. Wu et al. (2011).

Our experimental conditions were based on the natural conditions at the experiment station, which has a temperate semi-humid climate with a mean annual precipitation of 620 mm that is distributed unevenly throughout the year. Most rainstorms occur in summer, when sudden storms with rainfall intensities of 70–110 mm/hr are usual (data from Shaanxi Meteorology Bureau). In our experiments, we used a rainfall intensity of 90 mm/hr with total rainfall lasting 90 min for each surface treatment. In addition, we applied a 30-min prewetting rain at an intensity of 10 mm/hr to soil surfaces 24 hr before the rainfall to equalize soil moisture near the surface for all the rainfall runs. This provided a better comparison of the run-off and sediment generating processes from the different tillage methods.

During the rainfall experiments, the times for both the start of run-off and the start of the first rill formation were recorded. The combined surface run-off and sediment released at the outlet of the boxes was collected in 5-L buckets that were exchanged every 2 min with a new bucket and consecutively labelled. At rainfall termination, these buckets were weighed and then approximately 5 ml of saturated alum was added to each bucket to flocculate the solid fraction, which was placed in iron basins and oven dried at 110 °C for 8 hr. The dry sediments were weighed again, and run-off and sediment yields were calculated. The accuracy of the electronic scale used for the weighing was 0.01 g. Run-off and sediment yields were used to analyse run-off control benefit (RCB) and sediment control benefit (SCB) as a result of a particular tillage method on sloped land.

2.5 Statistical analysis

We defined the rainfall process as the overland flow erosion process before rill development on a slope and the rill flow erosion process after rill development. The RCB and SCB were calculated based on these two experimental processes. The RCB was calculated as

(1)

(2)

where RCB is the run-off control benefit of tillage (%); R₀(i) is the total run-off yield from an SS during rainfall event i (g); R_t(i) is the total run-off yield from a tilled surface during rainfall event i (g); and N is the number of rainfall events. If RCB > 0, the tillage is beneficial in controlling run-off; otherwise, there is no benefit, that is, tillage increases surface run-off.

The SCB was calculated as

(3)

(4)

where SCB is the sediment control benefit of tillage (%); S₀(i) is the total sediment from an SS during rainfall event i (g/m²); and S_t(i) is the total sediment from a tilled surface during rainfall event i (g/m²). If SCB > 0, the tillage is beneficial in controlling sediment loss; otherwise, there is no benefit, that is, tillage increases soil erosion.

We used MS Excel 10.0 for data arrangement and plots and DPS 7.05, a data processing system presented by Tang and Zhang (2013). Results are presented as the means and standard deviation of the three replicates. Significance at p < .05 of sediment and run-off between different tillage systems was determined using Fisher's least significant difference F test.

3.1 Effect of agricultural tillage techniques

We found different effects from the sloped boxes with different soil surface treatments (Table 2). The SS plots had the largest amount of accumulated run-off compared with the CT and RT treatments, which is in agreement with the findings of others that tilled rough surfaces lose less water than SSs (Cogo et al., 1984; García Morenoa et al., 2011; Helming et al., 1998). Surface run-off in the overland flow erosion process from the CT and RT treatments was significantly less than that of the SS treatment (p < .05), whereas the differences in the amount of surface run-off in the rill flow between the SS and CT treatments and between the SS and RT treatments were not significant (p > .05). In contrast, sediment loss among the SS, CT, and RT treatments varied. Although the RT treatment showed less sediment loss than the SS treatment in the overland flow, it sharply increased in the rill flow, and this increase led to greater accumulated sediment from the RT treatment compared with the SS treatment. As a result, the sum of sediment output by weight from the RT treatment for both the overland and rill flow erosion processes was significantly higher (p < .05) than that of the SS treatment. For the CT treatment, the sediment output was significantly less (p < .05) than that of the SS treatment for both the overland and rill flows.

To further quantify the effects of RT and CT on soil and water loss, the RCB and SCB of these treatments were calculated from Equations 2 and 4, respectively (Figure 1). Run-off control was found to be greater with overland flow than with rill flow (Figure 1a), suggesting that RT and CT have a stronger effect on run-off reduction before rill development. The RT and CT treatments reduced run-off by about 60% compared with the SS treatment during overland flow, whereas the effect of RT and CT treatments on run-off during rill flow was not significant. The average RCB was about 7% for the CT treatment and was below 0 for the RT treatment, suggesting that RT not only does not reduce surface run-off but can also promote it. Furthermore, when considering that the range of these values extends to either side of 0, RT and CT do not appear to have a significant ability to control run-off compared with an SS during the rill flow erosion process.

The positive SCB value for the CT for both overland and rill flows (Figure 1b) indicates that CT reduced soil loss by 30–60% compared with the SS. In contrast, the SCB of RT was above 0 for the overland flow but was about −25% for the rill flow, suggesting that RT cannot reduce soil loss during the rill flow erosion process. As a result, accumulated soil erosion yields from the RT treatment plot increased by approximately 16% compared with yields of the SS over all rainfall events. The SCB was significantly different (p < .05) between CT and RT for both the overland and rill flow erosion processes.

3.2 Role of tillage in the soil erosion process

Our findings showed that both agricultural tillage techniques reduced run-off and sediment generation during overland flow and that soil and water losses from these techniques were significantly less than from the SS treatment (p < .05). The CT treatment also reduced run-off and sediment during rill flow but to a different extent compared with that in overland flow. However, soil and water losses from the RT treatment were greater than those of the CT and SS treatments during the rill flow erosion process, perhaps due to the differences in surface microtopographies between the CT and RT treatments and their different effects on soil detachment and transportation. With CT, the surface microtopography of furrows and ridges was symmetric in character, with sediment from upslope blocked in furrows. With RT, the surface microtopography of depressions and mounds provided an anisotropic disposition. As a result, sediment from upslope was deposited in depressions, supporting the suggestions of Bochet, Poesen, and Rubio (2000) and Planchon and Mouche (2010) on how soil microtopography influences the ability to which sediment can be detached and transported. Another important factor that will influence the effect of surface microtopography on soil detachment and transportation is flow velocity. During the simulated rainfall, we attempted to measure flow velocity during the overland flow process using potassium permanganate but found that surface run-off moved slower than the diffusion of the potassium permanganate for both the RT and CT treatments. However, flow velocity was measurable during the rill flow process, and using this method, we recorded mean flow velocities of 0.19 and 0.13 m/s for the RT and CT treatments, respectively, based on five trials for each replicate (total of 15 trials for the three replicates for each treatment).

Tillage affects the soil characteristics in a number of ways. It changes the soil surface microtopography and increases soil surface variability (Römkens & Wang, 1986). The permeability of the topsoil increases with tillage, leading to a high infiltration capacity of soil and hence more water infiltration during rainfall (Dunne, Zhang, & Aubry, 1991; Manns, Berg, Bullock, & McNairn, 2014; Zhao et al., 2013). Steichen (1984) suggested that tilled surfaces have a higher infiltration capacity compared with untilled surfaces due to random roughness, and this has been supported by other researchers (Burwell & Larson, 1969; Freebairn, Gupta, Onstad, & Rawls, 1989; Magunda, Larson, Linden, & Nater, 1997). Another influence of tillage is that it increases surface storage and delays the initiation time of run-off. Zheng, Wu, He, Wang, and She (2007) demonstrated that the delay in surface run-off initiation ranged from 5 to 20 min and largely depended on tillage type and slope steepness. Using a rainfall event of 60 mm/hr intensity, they found that surface run-off initiation time followed the order CT > RT > SS, which is in agreement with our experimental results for both the overland and rill flow erosion processes (Figure 2). The wetting front of soil infiltration has been found to have the same spatial trend as the soil surface profile (Zhao et al., 2013). Zhao et al. (2013) also found that the accumulated infiltration of tilled surfaces is greater than 10% of that of an SS. Increasing SSR increases the depressional capacity of the soil, and this is another factor that delays the initiation of run-off on a slope.

The spatial distribution characteristics of surface microtopographies are known to differ among agricultural tillage techniques (Martin, Valeo, & Tait, 2008; Zhang et al., 2014). Ridges and furrows dominate the spatial geometry of the surface microtopography after tillage by CT, and these are relatively regular and show symmetry along the direction of the contour. With RT, mounds and depressions are the main spatial geometry and are randomly distributed over the whole surface (Zhao, Zhang, Liang, & Wu, 2011; Figure 3).

During run-off and erosion, surface microtopographies play different roles. On the basis of our observations, furrows and depressions have a storage impact, whereas ridges and mounds impact run-off generation. During rainfall, excess rainwater is temporally retained in furrows because ridges block the flow path of water until ridge failure occurs somewhere on the ridge downslope. The stored water begins to outflow from these pour points, and outflowing water contributes to overland run-off generation; at this point, parts of the ridges may be submerged by the outflowing water. Due to the continuous impact of run-off through the ridge and ridge failure, a pit appears on the ridge, and this develops into a short rill. In our study, the first ridge failure occurred approximately 20–35 min after rainfall began and also approximately 15 min after run-off was generated for all replicates with the CT treatment. Ridge failure occurred when the amount of rainwater exceeded the block capacity of the ridge. Denuding of the surface soil on the ridges was followed minutes later by pit formation. Down-cutting erosion occurred at these pits and proceeded until a rill developed that cut through the ridge between furrows. This caused more water and sediment to flow into the next furrow downslope. As neighbouring furrows are connected by a rill and the direction of rills is perpendicular to the contour, this provides a convenient means of draining water. Due to differences in the height of the ridges, the pour points of ridges across the surface are not usually along a straight line, and this characteristic means that the run-off and sediment in a rill do not travel between two neighbouring ridges. In turn, run-off and sediment travel a long or short distance in the furrow before the run-off water flows into the next downslope rill along the contour direction from the end of the last rill to the initial point of the next rill. During this period, the flow rate notably decreases as does the transportation capacity of the run-off. As a result, sediment from upslope is deposited, and a small alluvial fan forms. After a rainfall, many small alluvial fans can form at the low position of each rill. The process of soil erosion along a slope with ridge tillage can be impacted by microtopographic relief, and this microtopography can significantly reduce run-off and sediment production before rills form (Liu et al., 2014); however, once a rill has formed, the sediment detaching from a slope increases, and the subsequent rill erosion period can account for 87.2% of the total sediment yield measured for the whole rainfall–erosion and run-off–erosion processes. This is in agreement with the results obtained for our sloped plots. A rill network developed on the CT slope, and soil and water were lost from this channel (Figure 3); however, due to the rill network and alluvial fan formation, the CT-treated plot had low soil loss compared with the RT and SS treatments.

With RT, mounds and depressions play different roles in the run-off–erosion process compared with ridges and furrows. Mounds cannot be submerged by surface water, but they can divert surface run-off around their local summits and supply a continuous sediment load due to raindrop impact and surface run-off detachment. In contrast, depressions play a storage role, reducing surface run-off and sediment (Zhao et al., 2016). During rainfall, unconnected depressions are connected by outflowing water from other depressions and run-off from mounds; this leads to a continuous network over the whole surface. Once the flow network forms, the flow rate and transportation capacity of run-off increase, and soil erosion increases beyond what occurs with CT and SS. Eventually, the depressions fill in by sediment eroded from upslope, and their role in run-off and erosion processes disappears. Along with this, the mounds on RT slopes provide a greater source of sediment compared with SS under the same conditions. As a result, we found soil loss to be greater on the RT plot.

Another distinction between the two tillage methods is in the spatial pattern of rills. With RT, the surface microtopography has different spatial geometries compared with the microtopography with CT. This difference causes each rill to have distinctive features, including its length and direction. With RT, headward erosion begins at the outflow point of depressions and develops a consecutive rill in the short term with the direction of rills being from one depression to another. This pattern of rill favours sediment transport. However, with CT, the length of a rill is short, and the direction of the rills is perpendicular to the ridge; this characteristic of a rill inhibits sediment transport. Therefore, compared with RT, the rill pattern with CT provides greater soil and water conservation benefits.

In this study, we addressed the effect of two agricultural tillage methods on soil erosion between overland flow and rill flow processes, and the difference between them in run-off and sediment generation proved to be significant. Nevertheless, our experimental design used a single slope steepness and one level of rainfall intensity, and due to these limitations, our findings should not be generalized. Our study should now be extended to cover a range of slopes and rainfall intensities. In addition, the temporal variation of rills should be considered in future experiments.