Variability of interrill erosion at low slopes

EARTH SURFACE PROCESSES AND LANDFORMS Earth Surf. Process. Landforms 36, 97–106 (2011) Copyright © 2010 John Wiley & Sons, Ltd. Published online 5 July 2010 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/esp.2024 Variability of interrill erosion at low slopes A. Armstrong,1* J. N. Quinton,1 B. C. P. Heng2 and J. H. Chandler1 Lancaster Environment Centre, Lancaster University, Lancaster, UK 2 Department of Civil and Building Engineering, Loughborough University, Loughborough, UK 1 Received 15 October 2009; Revised 11 February 2010; Accepted 23 February 2010 *Correspondence to: Alona Armstrong, Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, UK. E-mail: [email protected] ABSTRACT: Numerous models and risk assessments have been developed in order to estimate soil erosion from agricultural land, with some including estimates of nutrient and contaminant transfer. Many of these models have a slope term as a control over particle transfer, with increased transfer associated with increased slopes. This is based on data collected over a wide range of slopes and using relatively small soil flumes and physical principals, i.e. the role of gravity in splash transport and flow. This study uses laboratory rainfall simulation on a large soil flume to investigate interrill soil erosion of a silt loam under a rainfall intensity of 47 mm h−1 on 3%, 6% and 9% slopes, which are representative of agricultural land in much of northwest Europe. The results show: (1) wide variation in runoff and sediment concentration data from replicate experiments, which indicates the complexities in interrill soil erosion processes; and (2) that at low slopes processes related to surface area connectivity, soil saturation, flow patterns and water depth may dominant over those related to gravity. Consequently, this questions the use of risk assessments and soil erosion models with a dominant slope term when assessing soil erosion from agricultural land at low slopes. Copyright © 2010 John Wiley & Sons, Ltd. KEYWORDS: microtopography; water depth; surface connectivity; rainfall simulation; particle size Introduction Soil is a key resource and understanding its erosion has become increasingly important given growing food demands, the increasing awareness of the impacts of diffuse pollution downstream and the introduction of legislation such as the European Union Water Framework Directive. One of the dominant sources of soil erosion is agricultural land, which is thought to contribute up to 50% of diffuse pollution to water courses (Defra, 2002). In order to understand, and therefore reduce erosion, much attention has been given to the dominant controls on erosion from agricultural land. Variables such as rainfall (Huang, 1995), soil type (Bradford and Foster, 1996; Ben-Hur and Wakindiki, 2004), ground cover (Snelder and Bryan, 1995) and slope (Fox and Bryan, 1999) have all been demonstrated to affect erosion rates. Of these factors slope and rainfall are among the most universal and feature in many models and risk assessment procedures (De Roo et al., 1996; Morgan et al., 1998; Renschler et al., 1999). Traditionally, soil erosion is thought to increase with slope. This has been demonstrated in the field (Chaplot and Le Bissonnais, 2000) and in the laboratory (Assouline and BenHur, 2006). The process behind increased erosion at higher slopes is attributed to the increased velocities which increase stream power and the preferential movement of splashed particles in the downslope direction (Ghadiri and Payne, 1988). However, there are complexities which can cause exceptions to this accepted relationship including variation in infiltration and surface flow characteristics, such as (1) the development of flow threads, (2) source area connectivity, (3) soil saturation and therefore cohesive strength, and (4) water depth and therefore obstruction of sediment movement and efficiency of splash detachment and transport. These complexities will alter the slope impact on the detachment and transport of soil by flow and rain splash. Infiltration affects erosion-slope relationships by controlling the discharge. Infiltration should be greater on shallower slopes as water depth is greater which increases the pressure head. In addition, higher infiltration rates on lower slopes have been attributed to greater hydraulic conductivities around stable mounds, more of which are submerged given the greater water depth (Fox et al., 1997). However, these accepted physical relationships between slope and infiltration are complicated by surface sealing, which reduces the infiltration rate. Surface seals have been observed to develop slower and less extensively on steeper slopes and therefore infiltration rates are higher (Poesen, 1984), although after sufficient time surface seals can develop to the same extent on steeper slopes (Luk et al., 1993). Variation in surface flow characteristics will also impact soil erosion. Interrill flow is often assumed to be evenly distributed sheet flow. However, in practice overland flow comprises of a distribution of flow velocities with more, narrower and faster flow threads found on steeper slopes and less, wider, and slower flowing flow threads on gentler slopes (Fox et al., 1997). Furthermore, Bryan (1979) noted that the hydraulics of 98 A. ARMSTRONG ET AL. overland flow varied with slope angle. These differences in flow regime impact both the amount and size distribution of sediment transported. Related to the flow pattern, is the connectivity of the soil surface, the degree of saturation and obstruction of sediment movement. Given the broader flows associated with lower slopes a greater proportion of the soil surface is likely to be connected compared with steeper slopes. While the greater linkage of source areas with the channel has the potential to allow the transportation of a greater amount of sediment, the flow energy will be lower, leading to lower sediment concentrations and the transport of finer sediment. In addition to the extent of the flow, slope will impact water depth producing the greatest depths on shallower slopes. While no study has aimed to examine the impact of flow depth on obstruction of particle movement, and consequently erosion rates, there is anecdotal evidence in Parsons et al. (1998) which suggests that at low flow and rainfall energies, sediment transport distances are greater on shallower slopes. Finally, another complexity is that with greater connectivity and deeper flows on shallower slopes a higher proportion of the soil will be saturated. Given that saturation has been shown to decrease cohesive strength and therefore potentially increase erosion (Gao et al., 2003) sediment supply may be increased at lower slopes. Consequently, although historically slope and soil erosion are considered to be positively related, processes relating to the amount and nature of the runoff, in terms of its extent, depth, and velocity distribution, may cause higher soil erosion rates on lower slopes. Given variation in rainfall, soil type and surface cover it is difficult to test such slope-erosion relations in the field. Therefore, several studies examining slope controls on erosion have been undertaken using laboratory rainfall simulation. However, these studies look at a wide range of slope angles (Table I), most of which are far in excess of those typical in agricultural fields in northwest Europe. While the conclusions of these studies all state that soil erosion increases with slope some illustrate exceptions at the lowest slopes (Helming et al., 1998; Römkens et al., 2002; Assouline and Ben-Hur, 2006). Furthermore, most of the existing studies used soil flumes with an area of <2 m2 (Table I). Although, such flumes have advantages: they are easier to work with and can be used under drip-type simulators, it is possible that processes which dominate in field environments may not occur in these smaller soil boxes, especially those which relate to the distribution and nature of shallow flows which, as outlined earlier, can impact the relationship between slope and soil erosion. Due to the small flume sizes and unrepresentative range of slopes used in existing studies the fundamental underpinning of soil erosion on slope in various models and risk assessments Table I. may be unsound for assessing erosion from arable agricultural land. This study has been designed to address this research gap by testing the slope–erosion relationship on slope angles representative of agricultural land using a larger soil flume (3·9 × 1·4 m) to avoid scale issues associated with small soil flumes. In addition to monitoring runoff volumes and sediment loss the extent of runoff and the particle size distribution of the sediment were monitored to allow insight into processes responsible for relations between slope and soil erosion. Methods In order to test the relationship between slope and soil erosion rates a laboratory rainfall simulator and a 3·9 m long by 1·4 m wide soil flume were used. The slope was set at 3%, 6% and 9% and the experiments run in triplicate. Along each side of the soil flume a 15 cm exclusion zone was created to prevent boundary effects. The flume was layered with 20 cm of gravel, a sheet of fine mesh, 20 cm of sand and 20 cm of soil. The soil used was a silt loam with 4·6% clay, 49·9% silt and 45·5% sand (primary particle size distribution) and was screened to 10 mm. The soil was field moist, added to the flume in known volumes, compacted to a bulk density of 1·3 g cm−3 and prepared as a seed bed. At the end of each experiment the top 4 cm of the soil was removed and fresh soil packed to the same density. Between each slope the subsoil was turned over, raked, and reprofiled to assure similar subsurface conditions between slope angles. Soil moisture was measured using six Delta-T ML2-x theta probes and recorded using a DL6 data logger. The theta probes were inserted 7 cm below the surface 0·5, 1·5, 2·5 and 3·5 m from the top of the flume and 12 cm below the surface 0·5 and 3·5 m from the top of the flume. Dry bulk density measurements were made between experimental runs using a 5 cm long cylinder with a 6 cm diameter. The rainfall was generated using a pumped rainfall system. Pressure regulators maintained an even pressure of 0·45 bar to four Fulljet ½ HH 40WSQ nozzles via solenoid valves as described in Strauss et al. (2000). The solenoid valves were controlled by a PC to turn on for two seconds and off for six seconds, generating rainfall with an average intensity of 47 mm h−1 with a standard deviation of 2·5. The spatial variability was reduced by altering nozzle positions and gave a Christensen uniformity coefficient of 70% (Christensen, 1942). De-ionized water was used for all runs to ensure no variability in input water quality. Runoff was collected using a runoff trough with a metal plate lip inserted into the soil. Discharge was monitored at the end of the flume every two seconds using weighing scales (Ohaus Defender 3000 Hybrid) with 0·02 kg resolution. The impact of Summary of rainfall simulation studies with the rainfall rate, slope angle and soil flume size Reference Fox and Bryan (1999) Bradford and Foster (1996) Huang (1995) Meyer and Harmon (1989) Ben-Hur and Wakindiki (2004) Helming et al. (1998) Huang (1998) Wan and El-Swaify (1998) Assouline and Ben-Hur (2006) Römkens et al. (2002) Copyright © 2010 John Wiley & Sons, Ltd. Rainfall rate (mm h−1) 49·1 72 50, 70, 14, 27, 40 15, 30, 30, 60, 45, 60, 24, 60 15, 30, 100 56, 115 45, 60 90 90, 135 45, 60 Slope (%) Flume size (m) 2·5, 11·5, 20·5, 30, 40 9, 20 4, 5, 9, 20 5, 10, 20, 30 9, 15, 20, 25 2, 8, 17 5, 10, 15 4, 9, 18, 27, 36 5, 9, 15, 20, 25 2, 8, 17 1·0 × 1·02 × 1·2 × 0·15, 0·30, 0·45 0·5 × 3·7 × 5× 0·6 × 0·5 × 3·7 × 0·4 1·02 1·2 & 0·60 × 0·30 0·3 0·60 0·6 0·3 0·3 0·60 Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011) VARIABILITY OF INTERRILL EROSION AT LOW SLOPES and 9% experiments, respectively. The soil moistures were not significantly different except those between the 3% and the 6% runs (p < 0·05). The differences in soil moisture between runs at the same slope were very similar: within 0·02 m3 m−3. No relationship was found between the mean soil moisture for each run and peak sediment concentration, time to peak sediment concentration, discharge at peak sediment concentration, time to steady state discharge, steady state discharge, total soil loss and mean sediment concentration. Runoff The runoff response varied in terms of the time between start of rainfall and start of runoff, steady state discharge and the time it took them to reach steady state (Figure 1). The differences in all three of these variables were not attributable to slope (p > 0·05) (Table II). Surface wetness Observation of the overland flow during the experiments indicated that there was lower surface connectivity and more flow 3% 20 .06 15 .04 10 .02 5 0 12 .06 10 .04 8 .02 Discharge, l/s 6% Sediment concentration, g/l sediment concentration on the density of the soil water mixture was considered and found to be minimal. The first sample was taken on the onset of runoff and then four at 30 second intervals, four at one minute intervals, four at two minute intervals, four at four minute intervals, three at six minute intervals and two at 10 minute intervals, giving a total of 22 samples. The rain pulsing was found to significantly impact both the sediment concentration and particle size distribution of samples (Armstrong and Quinton, 2009), therefore, the samples were taken in multiples of eight seconds (the duration of the rain pulse cycle). The number of multiples was dependent on the runoff: samples were taken until approximately 800 ml of water was sampled. Immediately after sample collection an aliquot was taken and used to determine the effective particle size distribution using a Malvern MasterSizer 2000MU. Each sample was measured three times and 21 samples were run three times (resulting in nine measurements) to allow the stability of the particle size distribution to be assessed. The contribution of the >0·5 mm fraction (the measurement limit of the MasterSizer is 1 mm, but to minimize sub-sampling bias and reduce aggregate breakdown a threshold size of 0·5 mm was selected) was calculated by passing the remained of the sample through a 0·5 mm sieve. The sample was carefully poured across the sieve to minimize aggregate breakdown. The sediment on the sieve was dried and weighed and the volume of the remainder of the sample was measured, emptied into a pre-weighed beaker, placed in the oven at 105 °C until dry, cooled, and weighed. The sediment concentration was calculated by summing the >0·5 mm and the <0·5 mm fractions. The primary particle size distribution of the parent soil was also analysed using the MasterSizer, with the organics removed prior to analysis (Gale and Hoare, 1991). Oblique stereo-photographs of the flume surface were taken at the end of each experiment, when the flume had drained, using two 10 megapixel Nikon D90 cameras mounted on camera arms attached to the ceiling. Twelve target points were located around the sides of the flume and surveyed using a total station. A digital elevation model (DEM) was produced using Leica Photogrammetry Suite 9.0 (Chandler et al., 2005; Heng et al., in press), an orthophotograph was draped over it and the soil surface categorized as ponding (smooth water surface with no aggregates visible), water-dominated (some aggregates visible but dominated by water), soil-dominated (many aggregates surrounded by water) or soil only (no visible water) in plan form and the areas calculated in ArcGIS 9.3. The DEMs were also used to create contour maps in Leica Photogrammetry Suite 9.0. The Kruskall–Wallis H test (Conover, 1999) was used to assess for significant differences between soil moisture, steady state discharge, time to steady state discharge, surface wetness, sediment concentration and sediment size with relation to slope. Analysis was undertaken using the “kwallis2” command in Stata10 (StataCorp, 2007) which allows the groups between which there are significant differences to be identified. The total amount of soil loss for each experiment was calculated by multiplying the sediment concentration by the discharge by the number of seconds between samples. 99 6 0 9% 40 .06 30 .04 20 .02 10 Results 0 Soil properties 2000 4000 6000 Time since runoff, seconds 3 The mean soil bulk densities were 1·33 g cm for each slope with values between 1·29 and 1·40 g cm3. The mean soil moistures from the six theta probes over the period of sample collection were 0·43, 0·37 and 0·40 m3 m−3 for the 3%, 6% Copyright © 2010 John Wiley & Sons, Ltd. 0 0 Figure 1. Sediment concentration and discharge for each of the experimental runs by slope. Filled in symbols denote discharge and open symbols denote sediment concentration. Note variable y-axis scales. Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011) 100 A. ARMSTRONG ET AL. Table II. Summary parameters for each experimental run Experiment 3% 3% 3% 6% 6% 6% 9% 9% 9% slope, slope, slope, slope, slope, slope, slope, slope, slope, run run run run run run run run run 1 2 3 1 2 3 1 2 3 Moisture (m3 m−3) Time to steady state discharge (seconds) Steady state discharge (l s−1) Time to runoff (seconds) Ponding (%) Water-dominated (%) Soil-dominated (%) Soil only (%) 0·44 0·43 0·43 0·35 0·37 0·36 0·40 0·40 0·41 1882 3532 2096 2126 2156 2292 3872 3180 2470 0·043 0·056 0·044 0·048 0·036 0·054 0·050 0·052 0·058 869 668 214 297 244 118 349 328 206 14·3 16·2 12·3 1·0 2·4 0·7 0·6 0·7 1·9 41·0 45·5 35·9 15·5 11·1 12·6 10·5 14·1 11·2 3·7 3·7 2·5 9·6 3·3 4·6 7·0 4·0 6·2 41·1 34·5 49·3 73·9 83·2 82·0 81·9 81·2 80·6 Figure 2. Typical surface area connectivity and flow patterns on the (a) 3%, (b) 6%, and (c) 9% slopes. More threaded flow was evident on the 6% and 9% slopes compared with the 3% slope. This figure is available in colour online at wileyonlinelibrary.com threads on the steeper slopes, with the difference being far greater between the 3% and 6% slopes compared with the 3% and 9% slopes (Figure 2). Quantification of the proportions of the flume covered by ponding water, water-dominated, soil-dominated and soil only (Table II) indicated that there was a significantly higher (p < 0·05) proportion of ponded water and water-dominated surfaces on the 3% slopes compared with the 6% and 9% slopes. Furthermore, the proportion of soil-dominated area was significantly lower (p < 0·05) on the 3% slopes compared with the 6% and 9% slopes and there was no significant difference in the proportion of soil only areas between the slopes. Sediment concentration The sediment concentrations were highly variable between and within the slope treatments (Figure 1). However, up to a discharge of 0·02 l s−1, the highest sediment concentrations were mostly on the 3% slope and the lowest on the 9% slopes (Figure 1). After 0·02 l s−1 there was no discernable pattern between concentration and slope (Figure 1). In contrast, when graphed as hysteresis plots with rescaled axis there was a clear pattern in the sediment concentration-discharge form with slope (Figure 3). Generally, the 3% runs were characterized by a peak in sediment concentration at the beginning of the run followed by a gradual decline, the 6% runs by a sharp increase in sediment concentration and an oscillating decline, and the 9% runs by a slower increase in sediment concentration and a gradual smooth decline (Figure 3). The differences in sediment response were evaluated by comparing (1) time, Copyright © 2010 John Wiley & Sons, Ltd. (2) instantaneous discharge, and (3) cumulative discharge at the point of maximum sediment concentration. This showed that sediment concentrations peak more rapidly at the lower slopes (mean times of 519, 623 and 1717 seconds for the 3%, 6% and 9% slopes, respectively) with statistically significant differences between the 3% and 9% (p < 0·05) and 6% and 9% (p < 0·05) slopes. The instantaneous discharge associated with the peak sediment concentrations was higher on the higher slopes (mean discharges of 0·008, 0·026 and 0·031 l s−1 for the 3%, 6% and 9% slopes, respectively) with significant differences between the 3% and 6% slopes (p < 0·05) and 3% and 9% slopes (p < 0·05). Finally, the cumulative discharge associated with the peak sediment concentrations were also higher on the higher slopes (means of 4·0, 12·8 and 32·5 l for the 3%, 6% and 9% slopes, respectively) with significant differences between the 3% and 9% slopes (p < 0·05). In order to assess the effect of slope at steady state, or approaching steady state erosion the sediment concentration of the last samples from each run was compared, but no significant differences were found (Table III). Sediment size The sediment size distribution of the eroded sediment is significantly finer than that of the parent soil, even though the parent soil distribution was determined after disaggregation (dispersion with sodium hexametaphosphate) and organic matter removal, whereas the eroded sediment was not treated (Figure 4). The <0·5 mm and >0·5 mm fractions were analysed separately to reduce bias given the different determination Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011) VARIABILITY OF INTERRILL EROSION AT LOW SLOPES 3%, 1 3%, 2 15 15 10 10 5 5 101 3%, 3 20 15 Sediment concentration, g/l 10 5 6%, 1 6%, 2 6%, 3 12 10 9 8 7 6 10 10 8 6 5 9%, 1 9%, 2 9 14 8 12 7 10 6 8 9%, 3 35 30 25 20 15 6 5 0 .02 .04 .06 0 .02 .04 .06 0 .02 .04 .06 Discharge, l/s Figure 3. Plots of sediment concentration as a function of discharge by slope (3%, 6% and 9%) and run (1, 2, and 3). Note variable x and y axis scales. Table III. Summary of measured variables from the last sample from each run and if there was a significant difference with relation to slope (p < 0·05) Variable Time since runoff (seconds) Sediment concentration (g l−1) Discharge (l s−1) Total discharge (l) Proportion sediment >0·5 mm Mean diameter (volume) (microns) Mean diameter (surface) (microns) d (0·1) (microns) d (0·5) (microns) d (0·9) (microns) Percentage clay Percentage silt Percentage sand Difference NS NS NS NS 3& 3& 3& 3& 3& 3& 3& 3& 3& 6 6, 6 6 6 6, 6 6, 6, 3&9 Total soil loss 3&9 3&9 3&9 Note: The slope with the highest value is highlighted in italic typeface· NS = not significant. methods employed. Examination of the <0·5 mm particle size data, both graphically and numerically, indicates that, the particle size coarsens through time on each run but there are no trends with slope and the variability between repeats is high (Figure 5 and Table IV). During the 9% slopes the median particle size initially fined and then coarsened: a trend not apparent on the other slopes (Figure 5). In terms of temporal dynamics of the <0·5 mm fraction examination of the first sample from each run indicates that the sediment transported on the 6% experiments was coarser than those transported on the 3% and 9% experiments (Figure 4a). In contrast, the sediment transported in the last sample of each experiment, when the system was approaching steady state, was notably finer for the 3% runs and there was no discernable difference between the 6% and 9% data (Figure 4b). These observations concur with statistical analysis of the last samples from each run: sediment transported on the 3% slopes was consistently significantly finer than that transported on the 6% and 9% slopes regardless of size parameter used (Table IV). Copyright © 2010 John Wiley & Sons, Ltd. The concentrations of the >0·5 mm fraction determined by sieves are also variable between repeat experiments. There is less >0·5 mm fraction transported on the 3% slope compared with the 6% slope (p < 0·05), but more is transported on the 6% slope compared with the 9% slope but the differences are not significant (p > 0·05). At the steeper slopes a positive relationship between discharge and concentration of the >0·5 mm fraction is evident at discharges above 0·35 l s−1 (Figure 5). Examination of the total amount of soil loss during each experiment indicates there are limited differences between the slopes (Figure 6), with mean loss of 1·6, 1·5 and 2·3 kg for the 3%, 6% and 9% slopes, respectively. If the last run at the 9% slope is discounted (soil loss was far greater than on any of the other runs) then the mean total soil loss is 1·4 kg, thus total soil loss decreases with slope angle, although the differences are not statistically significant. Discussion The results of these experiments indicate that at slope angles representative of agricultural land soil erosion does not increase with slope as traditionally assumed and that the runoff and erosion response is very variable. While numerous studies demonstrate the positive relationship between slope and erosion, closer inspection indicate that some corroborate our findings at low slopes. Helming et al. (1998) investigated soil loss from slopes with different surface roughness with each experiment involving subsequent rainstorms with intensities decreasing from 60 to 15 mm h−1 on the same soil surface. They found, for a smooth surface, that the soil loss from the 2% slope was greater than from the 8% slope (0·12 compared with 0·07 kg m−2) during the first rainfall event (60 mm h−1). Römkens et al. (2002) used the same experimental set-up and soil as Helming et al. (1998) and started with 15 mm h−1 rainfall and built up to 60 mm h−1. In this case there was no measureable difference in the sediment yield Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011) 102 A. ARMSTRONG ET AL. 100 (a) 80 60 40 3% 6% 9% Parent soil 20 0 1 10 100 Cumulative mass percentage Cumulative mass percentage 100 (b) 80 60 40 3% 6% 9% Parent soil 20 0 1000 1 10 Particle size, microns 100 1000 Particle size, microns Figure 4. Particle size distribution for the parent soil and (a) first and (b) last sample in each experimental run. (b) (a) 3% 3% .6 9 8 .4 7 .2 6 5 10 15 20 6% D0.5, microns 12 10 8 6 0 5 10 15 20 9% Concentration of sediment >0.5 mm, g/l 0 0 0 .04 .06 .04 .06 .04 .06 6% 1.5 1 .5 0 0 .02 9% 10 2 9 1.5 8 .02 1 7 .5 6 0 0 5 10 15 20 Run 2 .02 Discharge, l/s Sample Run 1 0 Run 3 Run 1 Run 2 Run 3 Figure 5. Change in the (a) d(0·5) sediment size and (b) contribution of the >0·5 mm size fraction through the runs for each slope. from the 2% and 8% slopes during the 15 mm h−1 rainfall and for the second event (30 mm h−1) the sediment yield was higher from the 2% slope compared with the 8% slope (0·13 and 0·08 kg m−2, respectively). Finally, Assouline and Ben-Hur (2006) and Bryan (1979) found that there was no perceivable slope control: Assouline and Ben-Hur (2006) found no difference in sediment concentrations from 5% and 9% slopes and Copyright © 2010 John Wiley & Sons, Ltd. Bryan (1979) found no slope control during a series of experiments on dry and saturated soil using eight soils and 10 slope angles (3° to 30°). In addition Parsons et al. (1998) examined transport distances of 3 mm particles along a fixed bed flume (particle diameters between 1 and 2 mm) at slopes of 3·5°, 5·5°, and 10° for a range of rainfall (0·00–0·85 J m−2 s−1) and flow energies (0·05–0·50 J m−2 s−1), which were varied indeEarth Surf. Process. Landforms, Vol. 36, 97–106 (2011) VARIABILITY OF INTERRILL EROSION AT LOW SLOPES 103 Table IV. Summary of particle size data for every fifth sample: volume-weighted mean diameter, d(0·1), d(0·5), d(0·9) and fraction <0·5 mm (mean and percentiles are from the <0·5 mm fraction given bias introduced by combining the sieve and MasterSizer data) Mean (microns) Sample Run 1 d(0·1) (microns) d(0·5) (microns) Run 2 Run 3 Run 1 Run 2 Run 3 Run 1 Slope, 3% 1 9·2 5 9·6 10 9·6 15 16·1 20 12·9 9·4 15·8 16·3 19·5 16·2 10·7 8·9 11·4 16·7 27·1 2·4 2·6 2·3 2·4 2·5 2·4 2·4 2·5 2·6 2·6 2·1 2·5 2·3 2·4 2·5 6·3 6·9 6·7 7·4 7·3 Slope, 6% 1 10·8 5 9·8 10 16·6 15 16·7 20 19·9 10·9 11·2 14·2 15·3 21·7 9·6 10·3 11·9 17·9 18·9 2·4 2·3 2·5 2·6 2·8 2·5 2·2 2·5 2·5 2·8 2·4 2·2 2·4 2·5 2·5 Slope, 9% 1 9·9 5 9·8 10 14·2 15 14·4 20 16·8 14·8 9·7 10·7 14·2 18·0 9·6 10·3 11·9 17·9 18·9 2·1 2·0 2·2 2·4 2·6 2·1 2·1 2·3 2·4 2·6 2·4 2·2 2·4 2·5 2·5 5 Run 2 d(0·9) (microns) Percentage <0·5 mm Run 3 Run 1 Run 2 Run 3 Run 1 Run 2 Run 3 6·2 7·4 7·5 8·5 8·3 5·9 6·4 6·7 7·2 8·2 17·0 18·6 20·1 25·3 23·0 16·7 30·7 30·1 35·4 33·4 19·4 17·1 21·4 29·9 45·3 99·8 99·7 99·3 97·7 99·4 100·0 99·3 96·8 97·4 95·4 99·6 99·7 99·4 98·8 98·3 6·9 6·8 8·1 8·7 9·9 6·9 6·6 7·5 8·2 10·6 6·3 6·3 7·0 8·0 8·4 22·2 20·7 31·4 34·6 43·3 21·6 23·3 29·1 31·7 50·3 18·1 20·2 23·8 36·9 41·1 99·4 99·4 97·7 96·0 87·8 98·6 98·3 97·3 91·5 87·3 98·8 97·6 97·7 89·7 95·4 5·9 5·6 6·4 7·6 8·9 6·0 5·6 6·5 7·8 8·9 6·3 6·3 7·0 8·0 8·4 20·1 19·4 25·8 29·1 36·7 22·6 18·7 20·8 29·1 39·3 18·1 20·2 23·8 36·9 41·1 99·6 99·2 96·0 95·8 94·8 99·7 98·7 97·8 97·0 96·1 99·5 99·6 99·4 98·1 95·0 220 Total soil loss Total flow 3 3 3 200 2 1 3 2 180 1 3 2 Total flow, l Total soil loss, kg 4 160 3 2 1 3 2 1 2 2 1 1 1 3 6 140 9 Slope, % Figure 6. Total soil loss and discharge from each experimental run. pendently from slope. Their data showed that particles travelled further on the 3·5° slopes than the 5·5° slopes at the lower flow energies (<0·30 J m−2 s−1). Furthermore, at a flow energy of 0·15 J m−2 s−1 and rainfall energy of up to 0·24 J m−2 s−1 transport distances were greater on the 3·5° slope than on the 10° slope. These experiments used various soil types and therefore suggest that our findings may be applicable to other soil types. The runoff records from our experiments were highly variable in terms of the time for runoff to commence, steady state discharge and the speed at which it was attained and statistical analysis suggested slope was not a dominant control. The inconsistencies in steady state runoff must be due to the variability in: rainfall intensity; infiltration rates, variable soil properties including surface seal development; varying amounts of Copyright © 2010 John Wiley & Sons, Ltd. water flowing in and out of the 15 cm exclusion zone; or a combination these factors. The mean rainfall rate was 47 mm h−1 with a standard deviation of 2·5 and is therefore unlikely to be the dominant cause of the variability in runoff. Variations in soil properties occurred despite careful preparation of the flume and a stable bulk density. While there was some variability in soil moisture the differences were no greater than 0·02 m3 m−3 between repeats. The maximum difference in moisture content between slopes was 0·09 m3 m−3, but the only significant difference was between the 3% and 6% runs. Furthermore, examination of the average moisture content of each run did not show any pattern with peak sediment concentration, time to peak sediment concentration, discharge at peak sediment concentration, time to steady state discharge, steady state discharge, total soil loss or mean Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011) 104 A. ARMSTRONG ET AL. sediment concentration. Therefore, while variable soil moisture has been shown to impact on soil erosion (Le Bissonnais et al., 1995) we believe the small variability had limited impact in these experiments. We have no direct measurements of surface seal development, however, observations of the soil surface structure suggest that its development across the flume was not uniform and that it may, therefore, have played an important role in controlling the variability of the discharge response. DEMs of the flume surface after the experiments generated using the digital photogrammetry indicated that some overland flow measured at the end of the flume originated from the 15 cm exclusion zone (Heng et al., in press). The variability in discharge was greater than that commonly found in flume studies (Fox and Bryan, 1999), but such studies generally use smaller flumes and a wider range of slopes (Table I). Furthermore, at lower slopes variability discharge records from some studies is high: Assouline and Ben-Hur (2006), who used a 0·3 by 0·5 m soil flume, report variation between replicates with overlapping standard errors at 5% and 9% slopes. While the variability makes interpretation of the data more complex, it is also more synonymous of results obtained from field environments, see for example Wendt et al. (1986), and it is processes occurring in these environments and their relative importance which we aim to understand. Despite the variability in runoff records, characterization of the surface ponding on the flume surface gave similar results between repeats and different results between slopes. The differences were far greater between the 3% and 6% slopes, compared with the 6% and 9% slopes. This concurs with the results of Fox et al. (1997) who also documented much more water ponding on lower slopes: 400 cm3 of water storage on 1·5° slopes, 175 cm3 on 6·5° slopes and 150 cm3 on 11·5° slopes. However, the distribution of patches varied between runs due to slight differences in microtopography and, together with development of a surface seal, and varying contributions from the exclusion zones explains why time to steady state discharge varies. In addition to the quantification of surface conditions, observation of the flume surface suggested that the flow patterns were different: more, thinner, faster flow threads occurred on the 6% and 9% experiments compared with the 3% experiments. These differences in surface conditions can be used to explain the sediment results as they control: (1) source area connectivity; (2) saturation; (3) flow patterns; and (4) depth of flow, and therefore sediment detachment and transport. Although not statistically significant there was a negative trend between total soil loss and slope (Figure 6). This did not correspond with total discharge, which is often the dominant control on sediment, nutrient or contaminant transfer. While there are limited data points and the differences are not statistically significant these data suggest there may be a dominant slope control, dominated by source area connectivity, flow patterns and depth of flow rather than discharge. These findings are corroborated by Bryan (1979) who also found that discharge was not the principal control over sediment concentrations and attributed the variability in sediment concentration to variation in the soil surface, water depth, drop diameter and interactions between them. However, this is the response from the onset of rainfall to approaching or at steady state and this should be borne in mind if comparing with models as they are generally based on steady state conditions. A slope impact was clearly evident in the forms of the relationships between sediment concentration and discharge, despite variations in the magnitude of these variables (Figure 3). The difference are less notable between the 6% and 9% slopes. This reflects the same pattern as found in the Copyright © 2010 John Wiley & Sons, Ltd. surface wetness analysis and observations of the flow network, and thus can also be attributed to surface connectivity, saturation, flow patterns and water depth. Other potential explanations appear less likely: although soil moisture differed significantly between the 3% and 6% slopes the difference in hysteresis loop form was most notable between the 3% and 9% slopes suggesting that soil moisture was not the cause. Varying flow patterns with slope may be a key variable affecting the sediment concentration discharge relationship as it impacts both the surface connectivity and the flow velocities. Interill flow is often conceptualized as even shallow flow but in reality flow threads exist (Dunkerley, 2004) and consequently there is a distribution of velocities, as opposed to an equal velocity across the soil surface. Dunkerley (2004) investigated this in a field setting and found that actual velocities were significantly different from the mean velocity: flow thread speeds were commonly 2·5 times greater than the mean and up to 6–7 times greater. Furthermore, Bryan (1979) found that the turbulence of flow altered with slope. Consequently, the flow thread pattern will impact the transport capability of the flow, affecting both the concentration and the size distribution. A difference in flow thread pattern was evident during the experiments, with more narrow faster flow threads associated with the steeper slopes compared to broader shallower flow on the 3% slopes (Figure 2). Consequently, the difference in the form of the sediment concentration–discharge relations could be partially attributed to the efficiency of the different velocity distributions associated with each slope and the differences in surface connectivity which results. It was not possible to reliably measure the velocity distributions as intrusive measurement techniques [e.g. salt-gauging (Planchon et al., 2005)] could not be used given their potential impacts on the flow and sediment transfer and tracing techniques, such as dyes, need correction factors and can be very inaccurate (Li et al., 1996). However, observation of the flume during the experiments indicated that larger particles were transported in the faster flow threads, but it was not possible to assess differences in concentrations between the slower flowing areas and the faster flow threads without disturbing the flow. These observations are supported by the particle size data which indicates that while the size distribution of the last samples were similar for the 6% and 9% experiments, the distribution was notably finer for the 3% experiments, which had broader slower flows. However it may also indicate that more coarse material was splashed into the flow at the steeper slopes given splash action has been shown to supply coarser sediment at a range of slopes (Wan and El-Swaify, 1998). Consequently, if the dynamics of the erosion are considered it is possible to explain the high initial sediment concentrations followed by a decline on the 3% slope by considering the surface flow characteristics. The significantly greater proportion of ponded and water-dominated areas resulted in increased connectivity, reduced soils cohesive strength and increased aggregate breakdown by slaking and thus greater sediment transport. Furthermore, the greater water depth on the 3% slope will have also influenced sediment transfer by controlling the efficiency of splash detachment and transfer (Kinnell, 1991) and also the ease with which particles can move through the water column. The efficiency of splash detachment and the ease with which particles can move through the water column were difficult to observe during the experiments, but the change in splash detachment efficiency and transport is well documented (Moss and Green, 1983; Ferrera and Singer, 1985; Kinnell, 1991). The impact of the depth of the water column and subsequently the ease with which particles can move through the water is less well docuEarth Surf. Process. Landforms, Vol. 36, 97–106 (2011) VARIABILITY OF INTERRILL EROSION AT LOW SLOPES mented. However, Parsons et al. (1998) examined the travel distances of particles using a fixed-bed flume and their data show that at the lower rainfall energies and flow energies particles travelled further on the lower slopes. Consequently, the only feasible explanation is that the increased water depths at the lower slopes enabled the particles to travel downslope less hindered than on the steeper slopes with lower water depths. Analysis of the particle size data give further insight into processes occurring during the experiments. Firstly, the particle size distribution of all eroded sediment was significantly finer than that of the parent soil (even though the parent soil was analysed for its primary particle size) and this enrichment of fines concurs with the results of many soil erosion studies (Quinton et al., 2001) and will have important implications for the transfer of contaminants and nutrients. The particle size distribution of the eroded sediment may have coarsened through time which we attribute to increase in discharge, exhaustion of the supply of fines or the development of a shielding layer (Heilig et al., 2001). The sediment particle size data can be used to infer the balance between wash and splash processes and changes in sediment regime with regard to slope. The particle size distributions of the <0·5 mm fraction of the first samples from different slopes were comparable, although the concentrations were significantly higher on the 3% slopes (Figure 4). The results of Wan and El-Swaify (1998), who examined splash and wash erosion of a silty clay at 4–36% slopes and rainfall of 65–135 mm h−1, illustrated that the particle size distribution of splash and wash sediment is significantly different, thus suggesting that the balance of wash and splash erosion was the same for all slopes in our experiments. However, the coarse sediment fraction (>0·5 mm) was significantly lower on the 3% slope compared with the 6% and 9% slopes, indicating different dominant processes may have been operating. This could be due to increased surface ponding at the 3% slopes resulting in sedimentation of the larger particles or the higher saturation causing the breakdown of aggregates producing a finer sediment supply on the 3% slopes. Furthermore, splash transport may have been more dominant at the steeper slopes: larger particles were transported by splash into the narrow flow threads, which had faster velocities than flow threads on the 3% slope, and were transported to the end of the flume. Alternatively, the higher stream powers on the steeper slopes may have increased transportation of larger particles. Another possible mechanism causing this trend is the operating of the third form of particle transport, splash creep, (Asadi et al. 2007), which is expected to be more dominant at higher slopes given the influence of gravity. While more coarse sediment was transported on the 6% slope compared with the 3% slope, a greater amount of coarse sediment was transported on the 6% compared with the 9% slope. This change in direction of trend could be explained by a threshold grain size above which slope is a dominant control or may be related to water depth and the ease with which particles can travel unhindered as observed by Parsons et al. (1998). Alternatively, as the median particle size of the <0·5 mm fraction initially declined on the 9% slopes and then increased (Figure 5) there could be greater aggregate breakdown on the higher slopes given the increased energy as a result of gravity and shallower water depths. In order to allow inferences to be made regarding the role of slope on sediment size the last samples from each run were statistically analysed when the system was at or approaching steady state erosion. This highlighted a clear slope impact on the particle size distribution of eroded sediment: the only variables which were significantly different between slopes Copyright © 2010 John Wiley & Sons, Ltd. 105 were those related to sediment size (Table III). Sediment was consistently finer on the 3% slopes. The differences were only significant between the 3% and steeper slopes which is the same pattern as found in the analysis of surface water coverage suggesting source area connectivity, saturation, flow patterns and flow depth may all have promoted the finer particle size through selective removal of fines, breakdown of aggregates, insufficient flow energy to transport coarser particles, and sedimentation of larger particles. Wan and El-Swaify (1998) examined the mean diameter of particles transported and found that size did not increase with slope at rainfall intensities <45 mm h−1. The rainfall rate in this study was 47 mm h−1 so, while a slope impact was evident between the 3% and steeper slopes for it to be apparent between the 6% and 9% slopes higher rainfall intensities may be required. Furthermore, as postulated for the results from the first sample of each run, the coarser sediment found on steeper slopes may reflect the balance of splash and wash processes (Wan and El-Swaify, 1998). One 3% slope run has a discernibly coarser particle size than the other two, although it is still finer than the 6% and 9% runs (Figure 4b). This is explained by a depression, identified by the contour maps derived from the DEMs, towards the end of the flume for the two runs with the finer coarser particle size distribution within which the coarser particles settled out of suspension. This highlights the importance of microtopography on sediment transfer. Conclusion These experiments indicated that interrill soil erosion of a silt loam at low slopes is highly variable. Differences in steady state discharge, which showed no relationship with slope, are attributed to variation in topography and infiltration, as a result of variable surface sealing, soil properties and contribution of water from the exclusion zone. The rate at which steady state discharge was attained also varied independently of slope and is attributed to variation in surface sealing and connectivity controlled by the microtopography. Sediment concentration was also highly variable but there were characteristic hysteresis loops for each slope, with the largest difference in form between the 3% and steeper slopes. Furthermore, finer sediment was generally associated with the 3% slopes and there were limited differences in particle size between the 6% and 9% slopes. These trends concur with the analysis of surface water coverages and therefore the differences in sediment dynamics are attributed to the affects of varying surface connectivity, saturation, flow thread patterns and flow depths, all of which are strongly influenced by microtopography. It is not possible to determine the relative dominance of these factors as they are interrelated and could not be controlled independently in our experiments. As previously mentioned there is evidence of this change in dominant processes at lower slopes in existing research (Helming et al., 1998; Römkens et al., 2002; Assouline and Ben-Hur, 2006) but it is not highlighted. These findings point to the complexity of sediment transport on slopes representative of arable agricultural land for a silt loam. They indicate that the interaction between surface topography and slope may be critical in controlling the development of flow pathways and the transport of soil particles. This has implications for soil erosion models and risk assessments, many of which include slope as a key driver. Furthermore, it will also be of significance for models and quantification of pollutant transport, which is strongly controlled by particle size distribution. Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011) 106 A. ARMSTRONG ET AL. Acknowledgements—This research was carried out as part of NERC funded project (NE/E007015/1). The authors would like to thank the anonymous reviewers for their helpful comments, Brenda Cookson for her assistance in the laboratory and Graham Sander, Cecil Scott, and Andrew Wheatley for useful discussions. References Armstrong A, Quinton JN. 2009. Pumped rainfall simulators: the impact of rain pulses on sediment concentration and size. Earth Surface Processes and Landforms 34: 1310–1314. Asadi H, Ghadiri H, Rose CW, Rouhipour H. 2007. Interrill soil erosion processes and their interaction on low slopes. Earth Surface Processes and Landforms 32: 711–724. Assouline S, Ben-Hur M. 2006. Effects of rainfall intensity and slope gradient on the dynamics of interrill erosion during soil surface sealing. CATENA 66: 211–220. Ben-Hur M, Wakindiki IIC. 2004. Soil mineralogy and slope effects on infiltration, interrill erosion, and slope factor. Water Resources Research 40: W03303. Bradford JM, Foster GR. 1996. Interrill soil erosion and slope steepness factors. Soil Science Society of America Journal 60: 909–915. Bryan RB. 1979. The influence of slope angle on soil entrainment by sheetwash and rainsplash. Earth Surface Processes 4: 43–58. Chandler JH, Fryer JG, Jack A. 2005. Metric capabilities of low-cost digital cameras for close range surface measurement. The Photogrammetric Record 20: 12–26. Chaplot V, Le Bissonnais Y. 2000. Field measurements of interrill erosion under different slopes and plot sizes. Earth Surface Processes and Landforms 25: 145–153. Christensen JE. 1942. Irrigation by sprinkling. In University of California Agricultural Experiment Station Bulletin, 670. University of California: Berkeley, CA; 124. Conover WJ. 1999. Practical Nonparametric Statistics. Wiley: New York. De Roo APJ, Wesseling CG, Ritsema CJ. 1996. LISEM: a single-event physically based hydrological and soil erosion model for drainage basins. I: Theory, input and output. Hydrological Processes 10: 1107–1117. Department for Environment, Food and Rural Affairs (Defra). 2002. Agriculture and Water: A Diffuse Pollution Review. Defra: London; 115. Dunkerley D. 2004. Flow threads in surface run-off: implications for the assessment of flow properties and friction coefficients in soil erosion and hydraulics investigations. Earth Surface Processes and Landforms 29: 1011–1026. Ferrera AG, Singer MJ. 1985. Energy dissipation for water drop impact into shallow pools. Soil Science Society of America Journal 49: 1537–1542. Fox DM, Bryan RB. 1999. The relationship of soil loss by interrill erosion to slope gradient. CATENA 38: 211–222. Fox DM, Bryan RB, Price AG. 1997. The influence of slope angle on final infiltration rate for interrill conditions. Geoderma 80: 181–194. Gale SJ, Hoare PG. 1991. Quaternary Sediments. Belhaven Press: London. Gao B, Walter MT, Steenhuis TS, Parlange JY, Nakano K, Rose CW, Hogarth WL. 2003. Investigating ponding depth and soil detachability for a mechanistic erosion model using a simple experiment. Journal of Hydrology 277: 116–124. Ghadiri H, Payne D. 1988. The formation and characteristics of splash following raindrop impact on soil. Journal of Soil Science 39: 563–575. Heilig A, DeBruyn D, Walter MT, Rose CW, Parlange JY, Steenhuis TS, Sander GC, Hairsine PB, Hogarth WL, Walker LP. 2001. Testing a mechanistic soil erosion model with a simple experiment. Journal of Hydrology 244: 9–16. Copyright © 2010 John Wiley & Sons, Ltd. Helming K, Romkens MJM, Prasad SN. 1998. Surface roughness related processes of runoff and soil loss: a flume study. Soil Science Society of America Journal 62: 243–250. Heng BCP, Chandler JH, Armstrong A. In press. Applying close range digital photogrammetry in soil erosion studies. Photogrammetric Record. Huang C.-h. 1995. Empirical analysis of slope and runoff for sediment delivery from interrill areas. Soil Science Society of America Journal 59: 982–990. Huang C-h. Sediment regimes under different slope and surface hydrologic conditions. Soil Sci Soc Am J 1998; 62: 423–430. Kinnell PIA. 1991. Effect of flow depth on sediment transport induced by raindrops impacting shallow flows. Transactions of the American Society of Agricultural Engineers 34: 161–168. Le Bissonnais Y, Renaux B, Delouche H. 1995. Interactions between soil properties and moisture content in crust formation, runoff and interrill erosion from tilled loess soils. CATENA 25: 33–46. Li G, Abrahams ADJ, Atkinson JF. 1996. Correction factors in the determination of mean velocity of overland flow. Earth Surface Processes and Landforms 21: 509–515. Luk SH, Cai Q, Wang GP. 1993. Effects of surface crusting and slope gradient on soil and water losses in the hilly loess region, North China. CATENA 24: 29–45. Meyer LD, Harmon WC. How row-sideslope length and steepness affect sideslope erosion. Transactions of the American Society of Agricultural Engineers 1989; 32: 639–644. Morgan RPC, Quinton JN, Smith RE, Govers G, Poesen JWA, Auerswald K, Chisci G, Torri D, Styczen ME. 1998. The European Soil Erosion Model (EUROSEM): a dynamic approach for predicting sediment transport from fields and small catchments. Earth Surface Processes and Landforms 23: 527–544. Moss AJ, Green P. 1983. Movement of solids in air and water by raindrop impact. Effects of drop-size and water-depth variations. Australian Journal of Soil Research 21: 257–269. Parsons AJ, Stromberg SGL, Greener M. 1998. Sediment-transport competence of rain-impacted interrill overland flow. Earth Surface Processes and Landforms 23: 365–375. Planchon O, Silvera N, Gimenez R, Favis-Mortlock D, Wainwright J, Le Bissonnais Y, Govers G. 2005. An automated salt-tracing gauge for flow-velocity measurement. Earth Surface Processes and Landforms 30: 833–844. Poesen J. 1984. The influence of slope angle on infiltration rate and Hortonian overland flow volume. Zeitschrift für Geomorpholgie 49: 117–131. Quinton JN, Catt JA, Hess TM. 2001. The selective removal of phosphorus from soil: is event size important? Journal of Environmental Quality 30: 538–545. Renschler CS, Mannaerts C, Diekkrüger B. 1999. Evaluating spatial and temporal variability in soil erosion risk – rainfall erosivity and soil loss ratios in Andalusia, Spain. CATENA 34: 209–225. Römkens MJM, Helming K, Prasad SN. 2002. Soil erosion under different rainfall intensities, surface roughness, and soil water regimes. CATENA 46: 103–123. Snelder DJ, Bryan RB. 1995. The use of rainfall simulation tests to assess the influence of vegetation density on soil loss on degraded rangelands in the Baringo District, Kenya. CATENA 25: 105–116. StataCorp. 2007. Stata Statistical Software: Release 10. Stata Press: College Station, TX. Strauss P, Pitty J, Pfeffer M, Mentler A. 2000. Rainfall simulation for outdoor experiments. In Current Research Methods to Assess the Environmental Fate of Pesticides, Jamet P, Cornejo J (eds). INRA Editions: Paris; 329–333. Wan Y, El-Swaify SA. 1998. Characterizing interrill sediment size by partitioning splash and wash processes. Soil Science Society of America Journal 62: 430–437. Wendt RC, Alberts EE, Hjelmfelt AT. 1986. Variability of runoff and soil loss from fallow experimental plots. Soil Science Society of America Journal 50: 730–736. Earth Surf. Process. Landforms, Vol. 36, 97–106 (2011)