Soil & Tillage Research 147 (2015) 50–59

Influence of ridge height, row grade, and field slope on soil erosion incontour ridging systems under seepage conditions

Q.J. Liu a,*, J. An a, L.Z. Wang a, Y.Z. Wua, H.Y. Zhang a,b

a Shandong Provincial Key Laboratory of Soil Conservation and Environmental Protection, Linyi University, Linyi, Shandong 276000, PR Chinab State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Institute of Soil and Water Conservation, Chinese Academy of Sciences,Yangling, Shaanxi 712100, PR China

A R T I C L E I N F O

Article history:Received 24 December 2013Received in revised form 17 October 2014Accepted 27 November 2014

Keywords:Contour ridgeSeepageSoil erosionRainfall simulation

A B S T R A C T

Seepage in contour ridge systems is a common phenomenon that can exacerbate soil erosion, however,the characteristics of soil erosion under seepage conditions in contour ridge systems are not clear. Theobjective of this study was to analyze the soil erosion process under seepage conditions and quantify theeffects and interactions between the ridge height, row grade, and field slope on runoff and sedimentyield. Twenty-three treatments for these three factors were arranged by an orthogonal rotatable centralcomposite design. A new type of experimental plot for simultaneously changing the row grade and fieldslope and creating seepage conditions was used to imitate the microtopographic relief of contour ridgesystems. In each run, seepage samples from the row sideslope were collected every 2 min for 60 min, andthen artificial rainfall simulation was performed for 30 min during which runoff samples were collectedevery 1 min. The results showed that four soil erosion sub-processes were observed, including interrillerosion, headward erosion, contour failure, and rill erosion. Second-order polynomial regression modelspredicted the sediment yield (R2 = 0.74) better than the runoff (R2 = 0.56). Interactions between thesefactors did not significantly affect the runoff or sediment yield even at p < 0.1. The row grade and fieldslope exerted a greater effect on the sediment yield than on the runoff, whereas the ridge heightinfluenced the runoff more with an increasing positive effect. The effect of these three factors onsediment yield revealed a convex curve with an increasing factor value. The field slope exhibited a greaterincreasing effect before the maximum sediment yield occurred and a greater decreasing effect after thatthan the other two factors did. The maximum runoff and sediment yield occurred at similar row grades(7.5� and 7.1�, respectively) and field slopes (10.9� and 10.8�, respectively). However, the minimum runoffoccurred at a ridge height of 6.7 cm, and the maximum sediment yield at a ridge height of 12 cm. Thefindings have important implications for assessing and modeling soil erosion in contour ridge systems.

ã 2014 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Soil & Tillage Research

journal homepage: www.else vie r .com/locate /s t i l l

1. Introduction

Seepage, defined as the reemergence of soil water at the surface,has been highlighted recently for producing areas susceptible toerosion on hillslopes (Chu-Agor et al., 2008; Huang and Laften,1996) and stream banks (Fox et al., 2007b; Karmaker and Dutta,2013), which are considered the dominant source of river sedimentin many areas (Fang et al., 2012; Fox and Wilson, 2010; Shi et al.,2013). Under seepage conditions, water-saturated soil looses itsmatric suction, which effectively reduces the stress of surface soilparticles (Vandamme and Zou, 2013; Al-Madhhachi et al., 2014). Inaddition, the exfiltration gradient can work against gravitational

* Corresponding author. Tel.: +86 13562990782; fax: +86 539 8766700.E-mail address: Liuqianjin@lyu.edu.cn (Q.J. Liu).

http://dx.doi.org/10.1016/j.still.2014.11.0080167-1987/ã 2014 Elsevier B.V. All rights reserved.

forces, further decreasing the effective stress (Huang and Laften,1996). The seepage flow itself possesses erosive power, and fineparticles can be eroded through the voids between coarse grains,significantly increasing hydraulic conductivity and decreasing soilstrength (Ke and Takahashi, 2012). Nouwakpo et al. (2010)observed that the average erodibility under seepage regimeswas 5.64 times larger than that under a drainage regime, and thecritical shear stress decreased dramatically as the hydraulicgradient increased from negative to positive (Nouwakpo et al.,2010). The effect of the hydraulic gradient on soil strength has alsobeen confirmed by Ke and Takahashi (2012).

Recent laboratory studies and field observations have demon-strated that the decrease in soil stress and increase in erodibilitycaused by seepage can exacerbate soil erosion (Fox et al., 2007b;Huang and Laften, 1996; Nouwakpo and Huang, 2012; Zheng et al.,2000). Huang and Laften (1996) observed that at a 5% slope, the

Table 1Code values determined by the orthogonal rotatable central composite design andcorresponding factor values of row grade, field slope and ridge height.

Code values Factor values

Row grade (�) Field slope (�) Ridge height (cm)

1.68 10.0 15.0 16.01 8.4 13.0 14.40 6.0 10.0 12.0

�1 3.6 7.0 9.6�1.68 2.0 5.0 8.0

Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59 51

sediment concentration was six times higher for a surface under20 cm seepage pressure compared with that of a surface drainedfor seven days, and the seepage could accelerate headcutdevelopment. Zheng et al. (2000) observed that sediment deliverywas three to six times greater in artesian seepage conditions thanunder drainage conditions with run-on runoff feeding. The erosionrate was 2.1 times higher in seepage conditions than in drainageconditions under high rainfall and run-on intensity (6.8 � 10�4

m�3 s�1), and the channel erosion rates were doubled underseepage conditions (Nouwakpo and Huang, 2012). Fox et al.(2007b) identified a maximum seepage of 1.0 L min�1 andsediment concentration of 100 g L�1 at a stream bank site in theGoodwin Creek watershed and observed at least three bankcollapses due to the seepage-erosion-initiated undercutting of thebank. Wilson et al. (2007) and Midgley et al. (2013) studiedseepage in situ and confirmed that seepage erosion was animportant factor in streambank failure. Field observations revealedthat seepage can induce the development of rills, leading to moresoil loss than interrill erosion (Valentin et al., 2005), and thisphenomenon was also verified by laboratory studies by Huang andLaften (1996).

Due to a lack of experimental observations and adequateresearch tools, the erosion process under seepage conditions haslargely been overlooked in prediction models, such as in the WaterErosion Prediction Project (WEPP) (Fox and Wilson, 2010;Nouwakpo and Huang, 2012; Valentin et al., 2005). Recent studiesmainly focused on two regimes related to seepage. One regime isthe downslope of the slope land, which is modeled by adjusting thelevel of supply water feeding to the bottom. Seepage and drainageconditions can be created when the water level is higher or lower,respectively, than the soil surface in laboratory conditions(Gabbard et al., 1998; Huang and Laften, 1996; Zheng et al.,2000). Based on the submerged jet test (jet erosion test, or JET)used in situ as well as in the laboratory (Hanson and Simon, 2001;Hanson and Hunt, 2007; Al-Madhhachi et al., 2013), a mini JET witha device to impose seepage forces was developed by Al-Madhhachiet al. (2014) and was considered as an in situ mechanistic approachto investigating soil erosion under seepage conditions. The otherregime is soil banks, which can be created by supplying water fromthe upper side of the soil matrix with a controlled water head (Chu-Agor et al., 2008; Fox et al., 2007a). To study seepage erosion underfield conditions, Midgley et al. (2013) created an innovative trenchinjection system that can provide a constant head on a near-streambank groundwater system that has been successfully usedto research seepage-induced streambank erosion and instability.However, another practical regime – the contour ridge system –

has not been considered until now. Contour ridging can increasethe infiltration of retained rainwater in furrows and result inseepage.

Contour ridging is an effective agricultural practice for soilconservation and crop promotion (Barton et al., 2004; Shi et al.,2004; Stevens et al., 2009). Soil erosion in contour ridge systemshas garnered increasing attention because variations in field slopeand microtopographic relief can produce ineffective erosioncontrol (Hatfield et al., 1998; Liu et al., 2014; USDA-ARS, 2008).The effects of ridge height, row grade, and field slope on soilerosion before contour failure have been considered to the greatestextent in the Revised Universal Soil Loss Equation, Version 2(RUSLE2) (USDA-ARS, 2008; Hessel et al., 2003). The ridge height,with a positive effect on water infiltration that results in less runoffand sediment yield, is used to compute the effect of contouring onerosion in the RUSLE2 model. The row grade along the furrows canform some depression areas where runoff accumulates. When theponded rainwater exceeds the storage within a contour row,overflow occurs and may result in severe ephemeral gully erosion(Flanagan and Livingston, 1995; USDA-ARS, 2008). The

conservation function of the contour ridge as the field slopeincreases is described as a concave curve, increasing from no soilconservation capacity to the greatest conservation benefit andthen decreasing to no benefit again (USDA-ARS, 2008).

To analyze the effects and interactions between the row grade,field slope, ridge height and width, and rainfall intensity on soilerosion induced by contouring failure, Liu et al. (2014) conducted32 rainfall simulation experiments arranged in an orthogonalarray. The results revealed that the interaction between the fieldslope and rainfall intensity had a significant effect on the runoff,and the ridge height was the most important factor for sedimentyield. However, this result and the research findings used in thedevelopment of the Universal Soil Loss Equation (USLE) (Wisch-meier and Smith, 1978), RUSLE (Renard et al., 1997), and WEPP(Flanagan and Livingston, 1995) did not consider seepageconditions. Recently, Al-Madhhachi et al. (2014) incorporatedseepage forces into the Wilson model (Wilson, 1993) based on JETtechniques and developed a new detachment model i.e., themodified Wilson model, for predicting the influence of seepage onsoil detachment. Because of the seepage creation method, in whichwater was supplied to the bottom of the soil matrix by an attacheddevice, more studies on the application of this modified model forthe estimation of soil erodibility concerned with seepage incontour ridge systems are necessary. Thus, a better understandingof soil erosion under seepage conditions and its influencing factorsin contour ridge systems will advance our knowledge of soilerosion and potentially improve soil erosion modeling andconservation practices. Therefore, this study was undertaken toexamine soil erosion under seepage conditions. The specificobjectives were to: (i) analyze the soil erosion process underseepage conditions and (ii) quantify the effects and interactionsbetween the ridge height, row grade, and field slope on runoff andsediment yield.

2. Methods and materials

2.1. Experimental design

An orthogonal rotatable central composite design was used toinvestigate the effects of three influential factors: the row grade,field slope, and ridge height. With this method, the number oftreatments required to estimate all of the terms of a second-orderpolynomial equation can be considerably reduced compared withthe full factorial design. Most importantly, the response modelcoefficients could be uncorrelated and estimated as a function ofonly the distance from the center and not the direction (St-Pierreand Weiss, 2009). Therefore, this method is widely used in variousfields (Domínguez et al., 2010; Hadjmohammadi and Sharifi, 2012;Hou et al., 2009; Tang and Feng, 2006; Zhou et al., 2007). Based onthe results of field investigations and previous studies, theminimum and maximum values of these factors at the codevalues of 1.68 and �1.68 were determined, and then the values atthe other code values (i.e., 1, 0, and �1) could be calculated, asshown in Table 1. Here, the code values were determined by the

Table 2The orthogonal rotary combination design for the three influencing factors at five code values and measured values of seepage discharge, runoff and sediment yield for eachtreatment.

Treatment No. Code values Experiment results

Rowgrade

Field slope Ridge height Seepage dischargea Runoff Sediment yield

(x1) (x2) (x3) (L min�1) (L) (kg)

1 1 1 1 0.98 34.37 1.192 1 1 �1 0.69 34.80 0.613 1 �1 1 0.87 35.68 0.334 1 �1 �1 0.50 27.01 0.405 �1 1 1 0.99 50.62 2.096 �1 1 �1 0.70 24.63 0.147 �1 �1 1 0.97 37.37 0.148 �1 �1 �1 0.52 32.92 0.099 1.68 0 0 0.80 39.27 3.83

10 �1.68 0 0 1.21 42.22 0.1711 0 1.68 0 0.54 33.71 2.0912 0 �1.68 0 0.74 34.23 0.1713 0 0 1.68 1.26 46.14 0.7714 0 0 �1.68 0.43 34.56 2.0315 0 0 0 0.51 40.02 2.2016 0 0 0 0.63 34.13 3.4817 0 0 0 0.63 40.02 2.7718 0 0 0 0.35 35.06 3.8919 0 0 0 0.83 35.57 3.1920 0 0 0 0.59 40.03 3.1721 0 0 0 0.62 28.74 2.0822 0 0 0 0.80 29.17 2.8423 0 0 0 0.43 35.57 3.08

a Means the seepage volume per min in the final 2 min of this experiment duration (60 min).

52 Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59

orthogonal rotatable central composite design. Twenty-threetreatments were arranged with different combinations of thesefactors (Table 2). In this experiment, due to the inherent repetitionof the factors in orthogonal treatments 1–8 and rotatable centraltreatments 15–23, treatments 9–14 were replicated twice, and atotal of 37 runs were performed.

2.2. Experiment plots

A new type of experimental plot designed by Liu et al. (2014)was used in this study (Fig.1). This plot could change the row gradeand field slope simultaneously. Its soil box consisted of twostainless steel cassettes (80 cm wide, 160 cm long and 50 cm deep)hinged together along an adjacent boundary. A row grade rangingfrom 0� to 15� could be created by rotating the screw (a), while afield slope from 0� to 20� could be obtained by adjusting the screw(b). To create the seepage conditions, water was pumped to thefurrow through two pipes (c) (with an inner diameter of 2 cm). Onthe wall of the pipes, eight holes with a diameter of 1 mm were

Fig. 1. Rainfall simulation plot for adjusting row grade and field slope simultaneously by

flow rate of 3 ml min�1 to the furrows where the water level was controlled by pipes (a, screw for row grade adjustment; b, screw for field slope adjustment; c, pipes for supplythe ridge (f); e, pipes for collecting redundant water; g, outlet for collecting seepage a

drilled. These holes and the ended outlets of the pipes were bandedby gauze. The two pipes could supply water with a discharge rangefrom 0 to 20 L min�1, and the required water supply discharge(3 L min�1) in this study could be obtained by adjusting the powerof the water pump. To avoid overflow, two pipes (d) were fixed inthe middle of the box bottom to let the redundant water flow outthrough the tubes (e) under the box. The water levels in thefurrows were kept approximately 1 cm lower than the concavespot at the ridge (f). The seepage and runoff were collected fromthe outlet (g) fixed in the middle of the downside of the box (Fig.1).

The sandy brown soil collected from the plough layer was usedin this experiment. The textural information for the soil is listed inTable 3. After being air dried, the soil was passed through a10.0 mm sieve. The bottom layer of the box was packed with soil ata bulk density of 1.6 g cm�3 in four 5 cm layers. Then, the ridge wasbuilt at a bulk density of 1.2 g cm�3. Based on a previous fieldinvestigation, the projection length of the lower row sideslope onthe box bottom flat was designed to be two times as long as that ofthe adjacent upper sideslope. Before packing, the bound lines of

rotating screw (a) and (b) and creating seepage conditions by supplying water with ad).ing water; d, pipes for adjusting the water level 1 cm lower than the concave spot atnd runoff.

Table 3Textural characteristics, organic C contentand bulk density of the plow layer and sole of the soil used in all experiments.

Gravel(%)

Sanda

(%)Silta

(%)Claya

(%)Organic C(g kg�1)

Bulk density of the plow layer(g cm�3)

Bulk density of the plow sole(g cm�3)

22.2 71.2 28.1 0.7 13.3 1.2 1.6

a The soil component classifications are based on the USDA soil classification system.

Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59 53

each layer and the ridges were drawn on the inside walls of the soilbox. The weight of the soil for each layer could be calculated basedon the determined bulk density, layer volume and soil watercontent measured in real-time. When the soil was compacted tothe layer bound lines, the soil surface was scratched by a woodboard to merge the soil matrix with the upper layer.

2.3. Seepage experiments

Seepage experiments to create stable seepage conditions bysupplying water to furrows were conducted before rainfallsimulation experiments. About 12 h before supplying water tothe furrow, a 60 min pre-rain at an intensity of 20 mm h�1 wasapplied to settle the soil surface and consolidate the looseaggregates to form a uniformly sealed, wet surface. The stableseepage conditions created here were used to reduce the potentialinfluence of the dramatic seepage change on soil erosion in thefollowing rainfall simulations, because the change of seepagedischarge process, presenting an increasing trend with an “S”shape, has been proved to have an obvious effect on soil erosion inour preliminary experiments. We also detected that it tookapproximately 60 min after the continuous seepage flow occurredthrough the outlet (g) for the seepage flow to stabilize under aconstant water supply rate of 3 L min�1. Therefore, the seepageexperiments were performed for 60 min beginning from whencontinuous seepage flow appeared. The stable seepage discharge,as a quantified index of seepage gradient, could influence soilerosion greatly, but it was not considered as an independent factorin this study, because seepage discharge is determined by theexperimental factors (i.e., ridge height, row grade and field slope)with other conditions (e.g., the soil properties, soil packing andwater supply) being equal. During the water supply process, theridge can sink and thus leading to a lower ridge height; therefore,the pipes (d) used to adjust the furrow water level should bechanged manually. When the continuous seepage flow occurred,the seepage flow was collected manually at the outlet (g) every2 min using pre-weighed plastic buckets. Therefore, a total of30 samples of seepage discharge were collected in each run.

2.4. Rainfall simulation experiments

In our preliminary experiment, the Horton flow runoffgenerated on the row sideslope in the first 1 or 2 min aftersimulated rainfall began at a rainfall intensity of 39 � 0.3 mm h�1.Because of the runoff, it is difficult to measure the seepagedischarge, which could occur until soil water is saturated given nopipeflow or preferential flow occurs. Therefore, rainfall simulationswere performed after the seepage experiments. Following theseepage process, two seepage conditions could be created for therainfall simulation programs. One condition maintained theaccumulated water in the furrows at the current level, with thewater supplying and excess water draining continued. Under thiscondition, the stored water in the furrow could barely overflow theridge, and interrill erosion on the row sideslope was therefore themain erosion type in the ridge system. The other condition haltedthe water supply and drainage, allowing the rainwater toaccumulate and flow over the ridge; in this condition, contourfailure might occur at a rainfall intensity of approximately

40 mm h�1, which could mimic the real circumstance better thankeeping the water level unchanged. Under the latter condition,because of the supplementation of rainwater and runoff, seepagedischarge was not dramatically reduced by closing the watersupply. Additionally, under this condition, more soil could beeroded when overflow or contour failure occurred, which was themain focus of this study. Therefore, in this study, once the seepageprocess was finished, the water supply was stopped, and the pipes(d) were closed. A rainfall simulation at an intensity of 39 � 0.4mm h�1 was performed for 30 min. The equipment used here was atrough rainfall simulator with fixed Veejet 80100 nozzles (Xie et al.,2008; Zhang et al., 2007) with a homogeneity coefficient greaterthan 0.89, which was available in the Shandong Provincial KeyLaboratory of Soil Conservation and Environmental Protection, PRChina. For each rainfall event, all runoff mixed with sediment wascollected with previously weighed plastic buckets at 1 minintervals. The collected samples were weighed immediately anddried in forced-air ovens at 105 �C for 12 h. The dried samples wereweighed again to calculate the water mass and volume bysubtracting the weight from the mixed sample weight. The totalrunoff or sediment yield in each rainfall event was obtained bysumming up the volumes or weights of all the individual samples.

2.5. Data treatment

The change in the row grade caused the section area of the plotsto change accordingly. Compared to the plot with a lower rowgrade, the section area was smaller for the higher row grade plot,which resulted in less runoff and sediment yield. Assuming thedifference area was located at the upper board of the ridge, wherethe interrill erosion occurred, Liu et al. (2014) generated Eq. (1) forthe sediment yield and runoff. The coefficient of the formula was sosmall (5E-9) that the runoff generated from the difference betweenthe plots with the highest and lowest row grade (approximately500 mL for treatment Nos. 9 and 10) could not harbor aconsiderable amount of sediment. Therefore, the sediment yieldvalues used here were not calibrated.

S ¼ 5E � 09Q2:6411n ¼ 16R2 ¼ 0:75 (1)

where S is the sediment yield per min (g min�1), and Q is the runoffper min (mL min�1).

The estimation of the regression coefficients, analysis of factoreffects and interactions were implemented using the DPS dataprocessing system (Tang and Feng, 2006).

3. Results and discussion

3.1. Runoff and sediment yield process

Twenty-three treatments, arranged in an orthogonal rotatablecentral composite design (St-Pierre and Weiss, 2009; Tang andFeng, 2006), were performed. The seepage discharge, runoff andsediment yields for each treatment are listed in Table 2. Soilerosion occurred under different seepage discharge ranging from0.35 to 1.26 L min�1. Compared with the runoff, the sediment yieldvaried more considerably, with the maximum value of 3.89 for No.18 being approximately 40 times larger than the minimum value of0.09 for No. 8. Different erosion processes were observed in the two

Fig. 2. Runoff and sediment yield process during artificial simulation within 30 minfor (a) treatment No. 8 and (b) treatment No. 18.P1, interrill erosion period; P2, headward erosion period; P3, contour failure period;P4, rill erosion period.

54 Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59

events, which could represent the two erosion mechanismsrepresentative of all of the experiments. The erosion process forNo. 8 included two sub-erosion periods (Fig. 2a), interrill erosionand a headward erosion period. In the first period, soil erosion wasmainly caused by raindrop detachment and raindrop-induced flowand transport, similar to the observations of Kinnell (2000). Mostof the sediment concentration in this period was less than2.0 g min�1. During the headward erosion period, soil erosion wasdominated by little water steepfall headwarding and transportunder overflow conditions. In this period, the overflow of thestored rainwater in the furrow can denude the soil matrix when thecritical shear stress of the soil is lower than that of the overflow,after which small waterfalls can appear, which can lead to highersoil erosion than interrill erosion (Kinnell, 2000; Meyer andHarmon, 1987). It was observed that when the eroded soilaccumulated at the bottom of the waterfall and hindered sedimenttransport, less sediment could be transported out. When themounded soil was pushed down by the surface flow, moresediment was observed. This hindering and pushing impact onsediment led to fluctuations in the sediment yield curve (Fig. 2).The water flow was weakly influenced by the mounded soil on the

Table 4The regression coefficient significance test for the second-order polynomial regression

Factor Runoff

Regression coefficient Standard regression coefficient t-test p Value

RG �1.37 �0.18 1.04 0.32

FS 0.77 0.10 0.59 0.57

RH 4.26 0.57 3.24 0.01

RG*RG 1.10 0.16 0.90 0.38

FS*FS �1.30 �0.19 1.06 0.31

RH*RH 0.96 0.14 0.79 0.44

RG*FS 0.19 0.02 0.11 0.91

RG*RH �2.78 �0.29 1.61 0.13

FS*RH 1.56 0.16 0.90 0.38

RG: row grade; FS: field slope; RH: ridge height. *indicates an interaction.

saturated soil surface; therefore, little dramatic fluctuations wererecorded.

When the headward erosion process continued, a rill wouldform in the row sideslope along which the stored water in thefurrow rushed down, resulting in ridge collapse. As a consequenceof ridge collapse, a mass of sediment yield in the poured wateroccurred in the following period, which was termed as contourfailure period here. This phenomenon was observed in treatmentNo. 18 (Fig. 2b). It was obvious that when contour failure occurred,a considerable quantity of soil was eroded. Contour failure candecrease the soil conservation capacity of contour ridge systemand cause ephemeral gully erosion (Hessel et al., 2003; USDA-ARS,2008). Induced by contour failure, the rill erosion in the ephemeralgully was the dominant erosion type, with rill bank failureoccurring occasionally due to geotechnical instability (Al-Madh-hachi et al., 2014; Midgley et al., 2013). In this period, the sedimentyield was much higher than that in the interrill and headwatererosion period.

However, contour failure and subsequent rill erosion did notoccur in all the treatments. In the treatment where the sedimentyields were less than 1 kg (Table 3), the ridge was not washed outcompletely. After erosion by concentrated flows the remainingridge heights differed among the treatments, which indicated thatthe soil conservation abilities of the contour ridges were destroyedto different degrees. Therefore, in the case of overflow occurrence,the ridge height decayed more rapidly than it did under interrillerosion. This decay in ridge height indicates that extra attentionshould be paid to ridge height modeling, for which only interrillerosion and precipitation were considered in RUSLE2 (USDA-ARS,2008).

In the interrill erosion period, both the runoff and sedimentyield displayed a slightly increasing trend in the first few minutesand then began to decrease slightly. The increase may have beencaused by the addition of rainfall-generated runoff from thesaturated (middle part) or near saturated (upper side) soil surfaceof the row sideslope. The decreases of the runoff and sedimentyield may be caused by seepage reduction where the runoffgenerated by rainfall alone tended to increase. This decreases couldbe explained by the following two phenomena observed during theexperiments. Notably, the water level in the furrow continuouslyincreased until overflow occurred, which indicated that rainfallcould result in a lower seepage rate than that in water suppliedconditions. The effect of rainfall on seepage could be attributed tothe clogging of rainfall-detached soil particles (Fox and Wilson,2010) or the compaction imposed by rainfall and water infiltrationon the soil matrix (Fohrer et al., 1999). The other observationindicated that when the duration of the interrill erosion percentwas long (e.g., treatment No. 1), the water level first became lowerafter the supplied water closed and then increased to a high leveluntil it overflowed. The decreasing water level in this processsuggests that a lower waterhead or gradient could reduce the seep

model of runoff and sediment yield.

Sediment yield

s Regression coefficient Standard regression coefficient t-test p values

455.86 0.27 1.95 0.07460.70 0.27 1.97 0.0728.45 0.02 0.12 0.90

�508.56 �0.32 2.35 0.03�815.74 �0.51 3.77 0.00�720.26 �0.45 3.32 0.01�116.38 �0.05 0.38 0.71�185.88 �0.08 0.61 0.55318.13 0.14 1.04 0.31

30

35

40

45

50

Run

off

(L)

(a)

Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59 55

(Nouwakpo et al., 2010; Zheng et al., 2000). Although the reductionof the seepage in the rainfall simulation period could be deducedqualitatively, it was not sufficient to reveal the soil erosionmechanism under seepage conditions. Therefore, measuring theseepage discharge during precipitation in situ will be an importantfocus of future studies.

3.2. Significance of the effects and interactions of the influencingfactors

The second-order polynomial regression models for runoff andsediment yield were established with the code values of row grade(x1), field slope (x2), and ridge height (x3) as independent variables.The significance of the effects and interactions of these influencingfactors are listed in Table 4. Because of the orthogonal design, theregression coefficients of the variables were uncorrelated and canbe used to assess the importance of these factors. The “+” and “�”

symbols indicated that the effect was positive or negative,respectively (Jiang et al., 2011; Tang and Feng, 2006).

Model (2) indicated that only 56% of the runoff (YQ) variancecould be explained by regression. The high p value (0.15) indicatedthat the regression was not significant. In addition, only the ridgeheight exhibited a positive significant effect (p < 0.01) on the runoff(Table 4). The row grade negatively affected the runoff and the fieldslope exerted a positive effect on it, albeit not significantly. Theregression model (3) for sediment yield (YS) exhibited a highefficiency coefficient (R2 = 0.74) and reflected the actual circum-stance significantly with a low p value (0.01). Table 4, whichdisplayed the significance of the regression coefficients, illustratedthat the row grade, field slope and quadratic terms significantlyaffected the sediment yield at p < 0.1. The ridge height did notexhibit a significant effect, whereas its quadratic term exhibited asignificant and negative effect.

YQ ¼ 35:45 � 1:37x1 þ 0:77x2 þ 4:26x3 þ 1:10x21 � 1:30x22 þ 0:96x23

þ 0:19x1x2 � 2:78x1x3 þ 1:56x2x3ðR2 ¼ 0:56 p ¼ 0:15Þ (2)

YS ¼ 2:98 þ 0:46x1 þ 0:42x2 þ 0:03x3 � 0:51x21 � 0:82x22 � 0:72x23� 0:12x1x2 � 0:19x1x3 þ 0:32x2x3ðR2 ¼ 0:74 p ¼ 0:01Þ (3)

The interactions between these three factors had no significanteffects, even at p < 0.05 (Table 4). Some interactions between the

Fig. 3. Effect of field slope on soil erosion in contour ridge systems at watershed andplot scale with different ridge height (higher rate of the erosion with contour to theerosion without contouring indicating the lower soil conservation benefit) (USDA-ARS, 2008).

above-mentioned influencing factors have garnered attention inthe RUSLE2 model (Fig. 3) (USDA-ARS, 2008). Fig. 3 illustrates thatthe ridge height had diverse effects on the benefit of soilconservation at different field slopes. Additionally, Liu et al.(2014) observed that the ridge height and field slope exertedsignificant (p < 0.01) positive effects on the runoff and sedimentyield caused by contour failure. The runoff during the rill erosionperiod was significantly affected by the interactions between thefield slope and ridge height and between the ridge height and rowgrade (Liu et al., 2013). However, under seepage conditions, nosignificant interactions between these factors (i.e., ridge height,row grade, and field slope) on runoff and sediment yield weredetected. Therefore, the interactions could be neglected and theimpact of these factors could be analyzed individually, indicatingthat soil conservation measures concerned with these factorscould be adopted independently in such circumstances.

3.3. Monofactor and border effect of the influencing factors

The monofactor effect models (Tang and Feng, 2006) wereobtained by fixing the other two factors at zero in model (2).Models (4)–(6) were the sub-models of runoff with only the rowgrade, field slope, and ridge height considered, where the runoffswere represented as YQRG

,YQFSand YQRH

, respectively. In the sameway, sub-models (7)–(9) for sediment yield were also derived frommodel (3), and the sediment yields were represented by YSRG , andYSRH , respectively. Based on models (4)–(6) and (7)–(9),Fig. 4 wasplotted to interpret the effect of these factors on runoff andsediment yield. These three factors all exhibited a nonlinearinfluence on the runoff (Fig. 4a). The runoff continuously increasedas the ridge height increased. For the row grade, the runoffdecreased to a minimal value and then increased. The field slopeexhibited diverse effects, increasing initially then decreasing to aminimum value. All sediment curves exhibited a convex shape,

20

25

-1.68 -1.0 0 1. 0 1.68Code values

RG FS RH

-0.50.00.51.01.52.02.53.03.5

-1.68 -1. 0 0 1.0 1.6 8

Sedi

men

t yi

led

(kg)

Code value s

RG FS RH

(b)

Fig. 4. Monofactor effect curves of row grade (RG), field slope (FS) and ridge height(RH) for (a) runoff and (b) sediment yield.RG: row grade; FS: field slope; RH: ridge height. Code values were determined bythe orthogonal rotatable central composite design.

56 Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59

meaning that the influences of the row grade, field slope, and ridgeheight exhibited similar patterns: with increasing code value, thesediment yield increased to a peak value and then declined.

YQRG¼ 35:45 � 1:37x1 þ 1:10x21 (4)

YQFS¼ 35:45 þ 0:77x2 � 1:30x22 (5)

YQRH¼ 35:45 þ 4:26x3 þ 0:96x23 (6)

YSRG ¼ 2:98 þ 0:46x1 � 0:51x12 (7)

YSFS ¼ 2:98 þ 0:46x2 � 0:82x22 (8)

YSRH ¼ 2:98 þ 0:03x3 � 0:72x23 (9)

To detect the rate of change of the influential factor effect on therunoff and sediment in the range of the code values, monofactorborder effect equations (Tang and Feng, 2006) were obtained bytaking the derivatives of models (4)–(6) and (7)–(9), respectively.The monofactor border effect equations for the runoff comprisedmodels (10)–(12), and those for the sediment comprised models(13)–(15). Fig. 5 illustrates the monofactor border effect on therunoff and sediment yield. In Fig. 5, the curves above the X-axis,with a positive Y-value, indicate that the factor effect is positive,and vice versa. The intersection of the X-axis and the curve is thedividing point of the effect, which can be either positive ornegative. At this point, extreme runoff or sediment yield occurs.The gradient of the curve indicates the degree of the increase or

-10.0-8.0-6.0-4.0-2.00.02.04.06.08.010.0

-1.68 -1. 0 0 1.0 1.68

Run

off

(L)

Code values

RG FS RH

(a)

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

-1.68 -1. 0 0 1.0 1.68

Sedi

men

t yi

eld

(kg)

Code values

RG FS RH

(b)

Fig. 5. Monofactor border effect curves of row grade (RG), field slope (FS) and ridgeheight (RH) for (a) runoff and (b) sediment yield.RG: row grade; FS: field slope; RH: ridge height. Code values were determined bythe orthogonal rotatable central composite design.

decrease of the effect. Fig. 5a shows that the runoff was positivelyinfluenced by the ridge height, with an increasing effect indicatedby the above X-axis curve with a positive gradient. The curves forthe row grade and field slope factor intersected with the X-axis,which indicated that a converse effect occurred as the factorincreased. For the sediment yield, all border effect curves exhibiteda descending trend with increasing values of row grade, field slope,and ridge height (Fig. 5b). The border effect curve for the field slopedisplayed a larger gradient, which indicated that before theintersection point with the X-axis, the field slope exerted a greaterincreasing effect on the sediment yield compared to the other twofactors, whereas after the point field slope exhibited a greaterdecreasing effect.

dYQRG

dx1¼ 2:2:0x1 � 1:37 (10)

dYQFS

dx2¼ �2:60x2 þ 0:77 (11)

dYQRH

dx3¼ 1:92x3 þ 4:26 (12)

dYSRG

dx1¼ �1:01x1 þ 0:46 (13)

dYSFS

dx2¼ �1:64x2 þ 0:46 (14)

dYSRH

dx3¼ �1:44x3 þ 0:03 (15)

By setting the right part of these monofactor border effectequations to zero, the monofactor code values and the corre-sponding values at which extreme runoff and sediment yieldoccurred could be obtained (Table 5). It was clear that themaximum runoff and sediment yield occurred at similar rowgrades (7.5� and 7.1�, respectively) and field slopes (10.9� and 10.8�,respectively). However, the extreme runoff and sediment yieldoccurred at quite different ridge heights. The minimum runoffoccurred at a ridge height of 6.7 cm, but the maximum sedimentyield occurred at a ridge height of 12 cm.

3.4. Interpretations of the factor effect on runoff and sediment yield

The row grade is considered an important factor in soil erosionin contour ridge systems in the soil erosion models of USLE

Table 5Monofactor code values and factor values for row grade, field slope and ridge heightat which the extreme runoff and sediment yield occurred.

Factors Runoff Sediment yield

RG Code value 0.62 0.46Factor value (�) 7.5 7.1

FS Code value 0.30 0.28Factor value (�) 10.9 10.8

RH Code value �2.22 0.02Factor value (cm) 6.7 12.0

RG: row grade; FS: field slope; RH: ridge height; “–” symbol means minimum valueoccurred.

Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59 57

(Wischmeier and Smith, 1978) and RUSLE (Renard et al., 1997).Although the absolute row grade, which is the decrease inelevation over the distance along the furrows (rise/run), givesmuch credit for soil loss precision, it was only used in special casesin the RUSLE (USDA-ARS, 2008). In RUSLE2 (USDA-ARS, 2008), therelative row grade was defined as the ratio of the row grade to theaverage steepness of the overland flow path and was considered acontouring subfactor (pc). Indeed, it was assumed that contouringrapidly decreases in effectiveness as the relative row gradeincreases (Fig. 6). In fact, the relative row grade is measured asthe ratio of the ridge–furrow orientation to the overland flow path,and only five specialized classes are used to represent contouring,regardless of any warranted further precision (USDA-ARS, 2008).The influence of the row grade in contour ridge systems in theRUSLE2 model is considered in the context of rill–interrill erosion,and the ephemeral gully erosion caused by contour failure isneglected. Recently, only a few studies have investigated the role ofrow grade on soil erosion. Using a new type of soil box in which therow grade and field slope are adjusted simultaneously, Liu et al.(2013) observed that the row grade significantly and positivelyaffected the sediment yield per time in the rill erosion periodbefore contour failure, whereas it did not significantly influencethe runoff or sediment yield induced by contour failure (Liu et al.,2014). Unlike the previous results (Liu et al., 2013, 2014; USDA-ARS,2008), in this study, the influence of row grade on soil erosion wasassessed under seepage conditions, and contour failure wasconsidered. The results indicated that the row grade exerted agreater effect on the sediment than on the runoff (Table 4). Itseffect on the runoff and sediment can be expressed as a quadraticcurve (Fig. 4) with the minimum and maximum values occurringapproximately at the code value of 0.5. This result indicated thatthe row grade had opposite effects on the runoff and sedimentyield. Comparing the results in Fig. 4b and Fig. 6 (USDA-ARS, 2008),it can be observed that the row grade exerted an observablydifferent effect on the sediment load, likely because thisexperiment was conducted under seepage conditions. Contourfailure was also considered. However, when increasing the rowgrade to 7.1�, the row grade positively affected the sedimentregardless of whether the soil was under drainage or seepageconditions. The similar row grades for maximum runoff andsediment yield (Table 5) indicated that adjusting the row gradewould affect the runoff and sediment yield in similar ways.

Fig. 6. Effect of relative row grade on the contouring subfactor (pc) inRUSLE2 erosion model (USDA-ARS, 2008).

The field slope has been investigated much more thoroughly forthe RUSLE2 model than in other studies and models (e.g., WEPP orUSLE). The effect of field slope on soil erosion was considered as aconvex curve (Fig. 3) (Liu et al., 2014; USDA-ARS, 2008). The soilconservation benefit transitioned from no effect to the greatestbenefit and then decreased to no effect with increasing slope, andthe greatest benefit occurred when the slope ranged from 3% to 7%due to the influence of ridge height (USDA-ARS, 2008). Further-more, the conservation benefit was maximized (i.e., minimum soilerosion occurred) at a slope of approximately 7.0%, while beyondthis slope, the field slope positively affected soil erosion. Liu et al.(2014) also observed that the field slope positively affected interrilland rill erosion on the row sideslopes, and contour failure occurredon the slope from 8.7% to 17.6%. In this study, the field slope exerteda similarly effect on the runoff and sediment yield (Fig. 4), with themaximum values occurring at 10.9� and 10.8� (Table 5), respec-tively. In addition, the field slope exerted a greater influence on thesediment yield than it did on the runoff (Table 4). The results inFig. 3 (USDA-ARS, 2008) and Fig. 4 suggest that the effect of fieldslope under seepage conditions may be similar to that underdrainage conditions when no contour failure has occurred. Thefield slope positively affected the sediment yield within fieldslopes of 5–10�. However, the field slope played a different rolewhen contour failure was considered: the sediment yield in thisstudy declined when the field slope increased continuously(Figs. 4 and 5). On higher slope land, the contour ridge was stilluseful for controlling soil erosion before the elevation of the ridgetop became lower than the adjacent furrow on the upper side, asstated in the RUSLE2 user guide (USDA-ARS, 2008). The field slopehad a similar effect on the runoff and the sediment yield with themaximum values occurring near 10.9� and 10.8�, respectively,indicating that changes in the field slope would have similarimpacts on runoff and sediment yield.

The ridge height has two effects in the RUSLE2 model (USDA-ARS, 2008). When the ridge is oriented up and down the hillslope,increasing the ridge height increases the row sideslope steepnessand area, which results in higher soil erosion. When the ridge isused as contour tillage, a higher ridge height can reduce soilerosion for more water infiltration. In addition, the effect of theridge height is influenced by the slope steepness (Fig. 3). Withincreasing slope steepness, higher ridges reach their greatest soilconservation capacity slightly later than lower ridges do, and thissoil conservation capacity is completely lost much later. Liu et al.(2013) observed a significant negative effect of the ridge height oninterrill–rill erosion, while when contour failure occurred ridgeheight aggravated the sediment yield significantly but reducedrunoff generation (Liu et al., 2014). Under seepage conditions, theridge height significantly and positively affected the runoff atp < 0.01, possibly because more water was stored in the higherridge (i.e., deeper furrow) system at the end of the seepageduration, and thus more runoff could be generated in the followingrainfall experiment. A positive effect of ridge height was alsoobserved in this experiment, similar, albeit not significantly similarto the results of Liu et al. (2014), indicating that contour failure-induced erosion can be positively affected by ridge heightregardless of whether it takes place under drainage or seepageconditions. Ridge height had a diverse effect on the runoff and thesediment yield. With increasing ridge height values, the runoff wasimproved continuously at an increasing rate, with the minimumvalue occurring at a ridge height of 6.7 cm, whereas the sedimentyield exhibited a convex curve, with the maximum valuesoccurring at a ridge height of 12 cm (Fig. 4b) (Table 5). Thisdiverse effect should be carefully considered when choosing theoptimum ridge height.

58 Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59

4. Conclusions

The effects of the row grade, field slope, and ridge height on therunoff and the sediment yield in contour ridge systems underseepage conditions were studied under simulated rainfall. Anorthogonal rotatable central composite design was used to testdifferent combinations of these factors, and regression modelswere used to quantify the effects of these factors. The soil erosionprocess could be divided into four sub-processes: interrill,headward, contour failure, and rill erosion periods. The lattertwo erosion periods produced vast sediment yield. Second-orderpolynomial regression models predicted the sediment yield betterthan they did the runoff, with determined coefficients of 0.74 and0.56, respectively. Interactions among the row grade, field slope,and ridge height did not exert significant effects on the runoff orsediment yield, even at p < 0.1. Compared with the runoff, the rowgrade and field slope had a greater effect on the sediment yield,which is represented as a convex curve with a sediment yield peakand derived related factor value. The ridge height was the onlyfactor that had a significant positive effect on the runoff, with anincreasing effect rate. Our results indicated that avoiding the codevalues at which the sediment yield value peaked would reduce soilloss in practice.

In this study, during the water supply period, the seepagevolume per time first increased and then stabilized when the soilwas saturated with water. To reduce the influence of the seepagechange on soil erosion in the rainfall simulation, a nearly steadyseepage rate was created by supplying water to the furrow for60 min. However, the water saturation regime in the field is not aphenomenon that will occur during all rainfall events. How thesefactors influence soil erosion under an unsteady seepage raterequires further investigation. For the seepage itself, preciselymeasuring the flow rate during rainfall events is still a challenge.Seepage-induced changes in the pore-water pressure and soilerodibility on row sideslopes should also be considered moreclosely, which could help to reveal the erosion mechanism. Inaddition, the sieved soil used here aimed to eliminate the influenceof plant roots or stones, which could lead to preferential flow.Confounded by preferential flow, the seepage amount and its effecton soil erosion would be more difficult to assess in the field.

Acknowledgement

Financial support for this research was provided by the NationalNatural Science Foundation of China (Nos. 41101263 and41301292).

References

Al-Madhhachi, A.T., Hanson, G.J., Fox, G.A., Tyagi, A.K., Bulut, R., 2013. Measuring soilerodibility using laboratory mini JETs. Trans. ASABE 56 (3), 901–910.

Al-Madhhachi, A.T., Fox, G.A., Hanson, G.J., Tyagi, A.K., Bulut, R., 2014. Mechanisticdetachment rate model to predict soil erodibility due to fluvial and seepageforces. J. Hydraul. Eng. 140 (5), 040140101–0401401010.

Barton, A.P., Fullen, M.A., Mitchell, D.J., Hocking, T.J., Liu, L., Wu Bo, Z., Zheng, Y., Xia,Z.Y., 2004. Effects of soil conservation measures on erosion rates and cropproductivity on subtropical Ultisols in Yunnan Province, China. Agric. Ecosyst.Environ. 104 (2), 343–357.

Chu-Agor, M.L., Fox, G.A., Cancienne, R.M., Wilson, G.V., 2008. Seepage causedtension failures and erosion undercutting of hillslopes. J. Hydrol. 359 (3–4),247–259.

Domínguez, J.R., González, T., Palo, P., Sánchez-Martín, J., 2010. Anodic oxidation ofketoprofen on boron-doped diamond (BDD) electrodes. Role of operativeparameters. Chem. Eng. J. 162 (3), 1012–1018.

Fang, N.F., Shi, Z.H., Li, L., Guo, Z.L., Liu, Q.J., Ai, L., 2012. The effects of rainfall regimesand land use changes on runoff and soil loss in a small mountainous watershed.Catena 99, 1–8.

Flanagan, D., Livingston, S., 1995. Water Erosion Prediction Project (WEPP) UserSummary-NSERL Report No. 11. USDA-ARS National Soil Erosion ResearchLaboratory, West Lafayette, IN.

Fohrer, N., Berkenhagen, J., Hecker, J.M., Rudolph, A., 1999. Changing soil and surfaceconditions during rainfall: single rainstorm/subsequent rainstorms. Catena 37(3–4), 355–375.

Fox, G.A., Chu-Agor, M.L., Wilson, G.V., 2007a. Erosion of noncohesive sediment byground water seepage: lysimeter experiments and stability modeling. Soil Sci.Soc. Am. J. 71 (6), 1822–1830.

Fox, G.A., Wilson, G.V., Simon, A., Langendoen, E.J., Akay, O., Fuchs, J.W., 2007b.Measuring streambank erosion due to ground water seepage: correlation tobank pore water pressure, precipitation and stream stage. Earth Surf. Proc. Land.32 (10), 1558–1573.

Fox, G.A., Wilson, G.V., 2010. The role of subsurface flow in hillslope and streambank erosion: a review. Soil Sci. Soc. Am. J. 74 (3), 717–733.

Gabbard, D., Huang, C., Norton, L., Steinhardt, G., 1998. Landscape position, surfacehydraulic gradients and erosion processes. Earth Surf. Proc. Land. 23 (1),83–93.

Hadjmohammadi, M., Sharifi, V., 2012. Simultaneous optimization of the resolutionand analysis time of flavonoids in reverse phase liquid chromatography usingDerringer’s desirability function. J. Chromatogr. B 880, 34–41.

Hanson, G.J., Hunt, S.L., 2007. Lessons learned using laboratory JET method tomeasure soil erodibility of compacted soils. Appl. Eng. Agric. 23 (3), 305–312.

Hanson, G.J., Simon, A., 2001. Erodibility of cohesive streambeds in the loess area ofthe midwestern USA. Hydrol. Process. 15 (1), 23–38.

Hatfield, J.L., Allmaras, R.R., Rehm, G.W., Lowery, B., 1998. Ridge tillage for corn andsoybean production: environmental quality impacts. Soil Till. Res. 48 (3), 145–154.

Hessel, R., Messing, I., Liding, C., Ritsema, C., Stolte, J., 2003. Soil erosion simulationsof land use scenarios for a small Loess Plateau catchment. Catena 54 (1), 289–302.

Hou, X.W., Wu, J.J., Xu, J.M., Tang, C.X., 2009. Interactive effects of lead andbensulfuron-methyl on decomposition of 14C-glucose in paddy soils.Pedosphere 19 (5), 577–587.

Huang, C.H., Laften, J.M., 1996. Seepage and soil erosion for a clay loam soil. Soil Sci.Soc. Am. J. 60 (2), 408–416.

Jiang, N., Zhang, A.Z., Yang, R.Q., Zhang, Y.C., 2011. An experimental approach tooptimize several processing conditions when extruding soybeans. Anim. FeedSci. Technol. 170 (3–4), 277–283.

Karmaker, T., Dutta, S., 2013. Modeling seepage erosion and bank retreat in acomposite river bank. J. Hydrol. 476 (0), 178–187.

Ke, L., Takahashi, A., 2012. Strength reduction of cohesionless soil due to internalerosion induced by one-dimensional upward seepage flow. Soils Found. 52 (4),698–711.

Kinnell, P.I.A., 2000. The effect of slope length on sediment concentrationsassociated with side-slope erosion. Soil Sci. Soc. Am. J. 64 (3), 1004–1008.

Liu, Q.J., Shi, Z.H., Yu, X.X., Zhang, H.Y., 2014. Influence of microtopography, ridgegeometry and rainfall intensity on soil erosion induced by contouring failure.Soil Till. Res. 136, 1–8.

Liu, Q.J., Zhang, H.Y., An, J., Wu, Y.Z., 2013. Soil erosion processes on row sideslopeswithin contour ridging systems. Catena . http://dx.doi.org/10.1016/j.catena.2013.11.013.

Meyer, L.D., Harmon, W.C., 1987. How row-sideslope length and steepness affectsideslope erosion. Trans. ASAE 32 (2), 639–644.

Midgley, T.L., Fox, G.A., Wilson, G.V., Heeren, D.M., Langendoen, E.J., Simon, A., 2013.Seepage-induced streambank erosion and instability: in situ constant-headexperiments. J. Hydrol. Eng. 18 (10), 1200–1210.

Nouwakpo, S.K., Huang, C.H., 2012. The role of subsurface hydrology in soil erosionand channel network development on a laboratory hillslope. Soil Sci. Soc. Am. J.76 (4), 1197–1211.

Nouwakpo, S.K., Huang, C.H., Bowling, L., Owens, P., 2010. Impact of verticalhydraulic gradient on rill erodibility and critical shear stress. Soil Sci. Soc. Am. J74 (6), 1914–1921.

Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., Yoder, D.C., 1997. Predictingsoil erosion by water. A guide to conservation planning with the RevisedUniversal Soil Loss Equation (RUSLE). Agriculture Handbook 703. USDA,Washington, D.C.

Shi, Z.H., Cai, C.F., Ding, S.W., Wang, T.W., Chow, T.L., 2004. Soil conservationplanning at the small watershed level using RUSLE with GIS: a case study in theThree Gorge area of China. Catena 55 (1), 33–48.

Shi, Z.H., Yue, B.J., Wang, L., Fang, N.F., Wang, D., Wu, F.Z., 2013. Effects of mulch coverrate on interrill erosion processes and the size selectivity of eroded sediment onsteep slopes. Soil Sci. Soc. Am. J. 77 (1), 257–267.

St-Pierre, N.R., Weiss, W.P., 2009. Technical note: designing and analyzingquantitative factorial experiments. J. Dairy Sci. 92 (9), 4581–4588.

Stevens, C.J., Quinton, J.N., Bailey, A.P., Deasy, C., Silgram, M., Jackson, D.R., 2009. Theeffects of minimal tillage, contour cultivation and in-field vegetative barriers onsoil erosion and phosphorus loss. Soil Till. Res. 106 (1), 145–151.

Tang, Q.Y., Feng, M.G., 2006. Data Analysis System – Experimental Design, StatisticalAnalysis and Model Optimization. China Scientific Press, Beijing, China.

USDA-ARS, 2008. User’s reference guide, Revised Universal Soil Loss EquationVersion 2, http://www.ars.usda.gov/sp2UserFiles/Place/64080510/RUSLE/RUSLE2_User_Ref_Guide.pdf, March 1, 2013.

Valentin, C., Poesen, J., Li, Y., 2005. Gully erosion: impacts, factors and control.Catena 63 (2–3), 132–153.

Vandamme, J., Zou, Q., 2013. Investigation of slope instability induced by seepageand erosion by a particle method. Comput. Geotech. 48, 9–20.

Wilson, B.N., 1993. Evaluation of a fundamental based detachment model. Trans.ASAE 36 (4), 1115–1122.

Q.J. Liu et al. / Soil & Tillage Research 147 (2015) 50–59 59

Wilson, G.V., Periketi, R.K., Fox, G.A., Dabney, S.M., Shields, F.D., Cullum, R.F., 2007.Soil properties controlling seepage erosion contributions to streambank failure.Earth Surf. Process. Land. 32 (3), 447–459.

Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses. Aguide to conservation planning. Agriculture Handbook 537. USDA,Washington, D.C.

Xie, Y., Lin, X.J., Liu, Y.N., Zheng, Y.Z., Liu, B.Y., Zhang, G.H., 2008. Calibration ofsimulated rainfall and its spatial distribution for trough rainfall simulator. Bull.Soil Water Conserv. 28 (4), 1–6 (In Chinese with English abstract).

Zhang, G.H., Liu, B.Y., Li, P.K., 2007. Principles and properties of artificial troughrainfall simulator. Bull. Soil Water Conserv. 27 (6), 56–60 (In Chinese withEnglish abstract).

Zheng, F.L., Huang, C.H., Norton, L.D., 2000. Vertical hydraulic gradient and run-onwater and sediment effects on erosion processes and sediment regimes. Soil Sci.Soc. Am. J. 64 (1), 4–11.

Zhou, X.L., Wang, H., Chen, Q.G., Ren, J.Z., 2007. Coupling effects of depth of film-bottomed tillage and amount of irrigation and nitrogen fertilizer on springwheat yield. Soil Till. Res. 94 (1), 251–261.