Effects of irrigation intensity on preferential solute transport in a stony soil

ABSTRACT

If irrigation intensity exceeds soil infiltration capacity, water may flow preferentially down cracks and large pores. In this situation, solute transport will involve only a fraction () of the soil’s water and leaching rate may be affected. To assess whether irrigation intensity affects preferential solute flow, an experiment was performed at Lincoln using 12 steel-encased lysimeters with a Lismore Stony Silt Loam soil under two irrigation intensities, 5 and 20 mm h^–1. Burns’ equation was used to describe the measurements of non-reactive tracer concentration as a function of drainage. Under dry antecedent moisture conditions, bromide transport was not significantly different under the different irrigation rates, even though strong preferential leaching occurred, with of 0.23. For chloride, was 0.85 and 0.58, for 5 and 20 mm h^–1 respectively, sufficient evidence to confirm the effect of irrigation intensity (P < 0.05). By assuming to be 1.00 for the median rainfall at Lincoln, an exponential function was fitted to the data, suggesting a lower limit of 0.35 for under moist conditions. Implications for nutrient leaching are discussed.

KEYWORDS:

• Burns’ equation
• lysimeter
• nutrient loss risk
• preferential flow
• solute leaching

Introduction

In soils, water is a solvent for plant nutrients and other agricultural chemicals. After rainfall or irrigation, nutrients applied on the surface are dissolved and transported by water infiltrating into the soil where they are redistributed and subsequently taken up by plants or microorganisms. When the storage capacity of the soil is limited or the water flow rate is high, nutrients will be transported faster down the profile, limiting the opportunity for plant uptake. Once leached below the root zone, the nutrients cannot be taken up and have the potential to contaminate the groundwater and/or surface water bodies. The balance between storage and transport of water and nutrients in the soil is therefore of high importance when determining the risk of environmental impact of different land uses and management practices. This balance can be substantially affected by the occurrence of preferential flow.

Preferential flow is a term used to describe the effects of non-uniform flow of water through soils (Clothier et al. 2008; Nimmo 2012; Beven & Germann 2013). When preferential flow occurs, part of the soil is not involved in the transport of water or solutes, momentarily reducing the storage capacity and increasing the flow rate, which means that leaching of surface applied solutes is much more likely. The occurrence of preferential flow is commonly linked to high intensities of rainfall or irrigation, surpassing the soil’s infiltration capacity, and is often an issue when the soil moisture is near saturation. In such conditions, water flows at great velocities through large pores, cracks, old roots and earthworm channels, bypassing most of the soil. There is also evidence that preferential flow occurs in many soils under unsaturated conditions (McLeod et al. 2008; Carrick 2009; Nimmo 2012). Different soil properties, such as aggregate level, and phenomena such as entrapped air and water repellence, can limit the access of infiltrating water to parts of the soil, and if rainfall or irrigation intensities are high enough, preferential flow is likely to occur. In such situations, incoming water and solutes contained therein will move through preferred pathways and around aggregates to a greater depth than would occur if the entire soil were involved in absorbing and transporting water and solutes.

In the Canterbury Plains of New Zealand, there is evidence of preferential flow for some soils (Fraser 1992; Silva et al. 2000; Toor et al. 2005; Pang et al. 2008; Carrick 2009; Carrick et al. 2014; McLeod et al. 2014). Research has been limited, but it indicates that stony soils deserve particular attention. With stones, soils have less space for retaining and transporting water and solutes, which can result in higher leaching rates. The risk for leaching losses can be greatly exacerbated in these soils if preferential flow also occurs. For instance, lysimeter studies at Lincoln University have shown that under similar irrigation and nutrient management the fraction of applied nitrogen that was leached from a Lismore stony soil was 2.4 times larger than that from a stone-free Templeton soil (Di & Cameron 2005, 2007). Preferential flow has been shown to be more important on stony Lismore soil than on deep stone-free Templeton soil (Pang et al. 2008) and it has also been shown to occur in other stony soils of Canterbury (Toor et al. 2005; Carrick et al. 2014; McLeod et al. 2014). Stony soils are common in Canterbury and North Otago (Kear et al. 1967; Carrick et al. 2013), regions that have been experiencing substantial land-use change driven by the increased use of irrigation. It has been estimated that Lismore soils cover approximately 40% of the irrigated land in the Canterbury Plains and a further 25%–35% of the area has similar shallow or stony soils (Carrick et al. 2013; Van der Weerden et al. 2014). Across the Canterbury region, leaching losses of nitrate were estimated at 20 million kg N per annum in 2010, a value that has been predicted to have doubled since 1990 (Dymond et al. 2013). Concern over such environmental impact of agricultural activities is increasing and has become a major focus for research and policy development (e.g. Lilburne et al. 2010; ECan 2011).

The majority of the irrigated land in New Zealand is in the South Island, with Canterbury, at 444,800 ha, representing approximately 60% of all irrigated area (Statistics New Zealand 2012). The area of irrigated land is increasing with an estimated potential for doubling the current extent in Canterbury (Saunders & Saunders 2012). The increased use of irrigation over the past 20 years is linked to the intensification of farming systems, particularly dairy farming (Carrick et al. 2013; Van der Weerden et al. 2014). Spray systems, especially centre-pivot, are predominant, representing 88% of the area irrigated in Canterbury, while flood systems represent approximately 10% and this area has been decreasing (Statistics New Zealand 2012). Centre-pivot systems are the most popular choice for implementing new irrigation schemes and to replace old border-dyke systems, a move that has been driven by the need for increasing water-use efficiency. Although centre-pivots allow a good level of control over the amount of water applied (Evans et al. 2013; Hedley 2015), their very nature implies that the instantaneous intensity of application varies considerably along its length. It is important to distinguish between the average irrigation rate, which represents the amount of water applied in a given time divided by the area irrigated, and the instantaneous irrigation intensity, hereafter simply called irrigation intensity, which is the rate of irrigation applied by each nozzle (INZ 2015). For a centre-pivot, the linear speed of each nozzle increases as the distance from the centre point increases; consequently, the irrigation intensity has to be much larger in the outer edge of the area irrigated. As an example, a system deploying an average of 5 mm day^–1 reaches an intensity of 20 mm h^–1 at a radius of about 200 m, and will need to reach intensities of about 100 mm h^–1 for the sprinklers 1000 m away from the centre (INZ 2007; Powers 2012). These irrigation intensities are much greater than 0.5 mm h^–1, the median hourly rainfall intensity for the years 2000–2014 in Lincoln (data from Cliflo, NIWA 2015). Such high irrigation intensities in soils with potential for preferential flow are likely to greatly elevate the risk for leaching losses.

Evidence for preferential flow can be obtained by several different methods, although quantifying it can be difficult (Kung et al. 2000; Clothier et al. 2008; Carrick 2009). Leaching experiments in soil cores or lysimeters using tracers (i.e. non-reactive solutes) are often used to obtain temporal water and solute flow data. Modelling analyses of such data allow algorithms for preferential flow to be developed and tested. While the mobile–immobile water paradigm is the most common approach (Simunek et al. 2003; Kohne et al. 2006; Pang et al. 2008), it can be difficult to employ as it requires detailed data sets for the determination of its many parameters. Alternatively, simpler empirical approaches can be useful in situations where detailed soil data are not readily available (Addiscott 1993; Scotter et al. 1993; Monaghan & Smith 2004). The Burns’ equation is one such model. Initially developed as a simple description of solute transport, the Burns’ equation has been reinterpreted and shown to be able to characterise the transport of solute under preferential flow (Scotter et al. 1993). This elegant and effective approach has been shown to describe accurately the results of preferential solute transport from both laboratory and field experiments (e.g. Heng & White 1996; Magesan et al. 1999).

Simple approaches such as the Burns’ equation can be used to assess the potential for environmental impacts from intensive land use under irrigation. Given the expansion of irrigation in the Canterbury region, especially on to vulnerable soils (i.e. shallow and stony soils with potential for preferential flow), the effects of irrigation intensity on the risk of leaching losses requires a better understanding. For these reason, we performed a controlled experiment where tracers were applied under contrasting irrigation intensities to a stony soil in large undisturbed lysimeters. Our objectives were to verify the presence of preferential solute flow in a stony silt loam soil, assess whether or not irrigation intensity affects solute preferential flow in this soil, and by analysing the data, describe the potential for nutrient leaching losses from stony soils under irrigation.

Materials and methods

Lysimeter experiment

The experiment was conducted at the lysimeter facility of Plant and Food Research at Lincoln in December 2014. The facility has a movable rain shelter that was used to prevent rainfall on the lysimeters during the experiment. The lysimeters were collected early that year from a dairy farm near Methven, South Island, following the procedure described by Cameron et al. (1992); 12 lysimeters, each of 500 mm diameter and 700 mm length were collected (effective size of the soil monolith was assumed to be 480 mm in diameter and 675 mm in length). The soil was a Lismore Stony Silt Loam (Pallic Firm Brown; Hewitt 2010), with an average of 225 mm of total pore space and 95 mm of plant available water over the lysimeters’ depth (details of average soil properties are given in Table 1). No suction was applied to the bottom of the lysimeters. The original ryegrass/white clover sward was kept in the lysimeters and trimmed on three occasions prior to the experiment. Over this period, the lysimeters were exposed to rain most of the time and received no irrigation, except when the equipment was tested. At the time of the experiment, the plants exhibited signs of strong water stress and were not actively growing.

Table 1. Selected soil properties of the Lismore Stony Silt Loam used in the lysimeter experiment. Measurements from samples collected during lysimeter’s extraction.

Horizon	Ap1	Ap2	Bw	Bt
Depth (m)	0.00–0.10	0.10–0.30	0.30–0.50	0.50–0.70
Functional horizon^1	tSLw	tSLw	VLw	VAf
Stone content (%)	17	26	54	64
Sand^2 (%)	27	25	33	67
Clay^2 (%)	28	27	25	13
Bulk density^2 (kg m^–3)	940	1190	1100	1070
Porosity^2 (% vol)	57	55	58	58
Field capacity^3 (% vol)	43	40	40	29
Hydraulic conductivity^3 (m h^–1)	8.8	3.0	1.2	3.2
Carbon content^2 (%)	3.6	2.3	1.1	0.9

¹Based on Webb (2003) and Webb & Lilburne (2011).

²Values given are for the fine earth fraction.

³Values estimated using pedo-transfer functions (Cichota et al. 2013).

The experiment included two irrigation treatments: average application intensities of 5 mm h^–1 and 20 mm h^–1. The irrigation system used micro sprinklers (Quick TeeJet QJ200 nozzle bodies with FL-5VC nozzles, TeeJet Technologies) in each lysimeter. The sprinklers were calibrated and controlled individually to deliver an even volume of water throughout the experiment. To attain 5 mm h^–1 rates, the sprinklers were on for 2 s and off for approximately 170 seconds, while for 20 mm h^–1, the sprinklers were on for 2 s and off for 43 seconds. Each of the lysimeters received an application of bromide (as KBr, at a rate equivalent to 50 kg ha^–1) at the start of the experiment, at dry antecedent soil moisture conditions, and chloride (as KCl, at a rate of 400 kg ha^–1) under moist antecedent conditions, when a volume approximately equal to one liquid-filled pore-volume had drained (about 11.7 and 47.3 h after the start of the experiment, for 20 and 5 mm h^–1 treatments, respectively). Furthermore, the bromide was applied in two ways: half of the lysimeters received bromide 1 day prior to the start of the irrigation, allowing some equilibrium with the soil matrix; the remaining lysimeters received bromide just before the irrigation started. The tracers were diluted in 500 mL of water for bromide and 1000 mL for chloride; the solutions were carefully applied to achieve an even distribution over the lysimeter surface using spray bottles over a period of time to mimic the corresponding irrigation intensity. The irrigation was kept constant during the experiment, which lasted 33 h for the 20 mm h^–1 treatment and 130 h for the 5 mm h^–1 treatment (corresponding to about three pore-volumes of leachate collected). This ensured that almost all the applied tracer was recovered. The drainage rate was measured using a tipping spoon mechanism (PCB 9602, Pronamic) and, in each lysimeter, soil moisture probes (SM300, Delta-T Devices Ltd) were also installed at depths of 50, 150, 330 and 500 mm. The leachate was collected in 10 L containers, which were weighed and emptied at regular intervals (every 30 min for the lysimeters receiving 20 mm h^–1 of irrigation and every 2 h for the lysimeters receiving 5 mm h^–1). At each sampling time, a 20 mL sample was collected and stored refrigerated at 4 °C for chemical analyses. The samples were analysed using ion exchange chromatography (QuikChem 8500, Lachat).

Data analyses and modelling

The drainage data were analysed for consistency and to ensure that the irrigation had been uniform across all lysimeters. Except for the initial samples and those immediately after the application of chloride, every second leachate sample collected was analysed for tracer concentration. The concentrations of tracers in the unanalysed samples were estimated by interpolation. The quantity of each tracer leached at each collection time was then calculated by multiplying the drainage volume by the tracer concentration.

To describe the leaching series for each tracer, formulations of the Burns’ equation following the approach presented by Scotter et al. (1993) were used. Solutions for different initial and boundary conditions were tested given the importance of selecting the appropriate boundary condition, as showed by Magesan et al. (1999). Anticipating our results, only the two most appropriate solutions are presented here.

In the function used for bromide, the amount of tracer applied, (kg/ha), is assumed to be at the soil surface and the irrigation water is free of the tracer. The leaching intensity, (kg ha^–1), is a function of cumulative drainage, (mm), and is given by:(1)

where (mm) is the lysimeter’s depth, (m³ m^–3) is the soil water content at field capacity, and represents the fraction of water that effectively takes part in the solute transport. The product is often referred to as the fractional transport volume (Scotter et al. 1993). The original version of the Burns’ equation assumed that all the soil’s water was involved in solute transport ( = 1.0).

To describe the chloride application, the boundary condition is defined by the concentration in the irrigation water being equal to 0 except for a given period over which the solute was applied (15 min for the lysimeters under 20 mm h^–1 and 1 h for the lysimeter at 5 mm h^–1). The concentration in the irrigation water is given by the amount of tracer, (kg ha^–1), divided by the volume of water with which it was applied, (mm). The solution is slightly more complex:(2)

The two solutions of the Burns’ equation were fitted to the data from the lysimeter experiment using the non-linear least squares (nls) method of the R statistical package (R Core Team 2015). The only parameter that had to be fitted was . The goodness-of-fit was also evaluated by the root mean squared error (RMSE) and the coefficient of determination (R²); the irrigation intensity treatment effect was tested using Student’s t test with 5% probability.

Results

Lysimeter experiments

Drainage was recorded relatively soon after the irrigation began, at an average of 1.4 h for the lysimeters under 20 mm h^–1 and 8 h for the lysimeters irrigated at 5 mm h^–1 (i.e. after 28 and 40 mm of irrigation). In two lysimeters, one from each treatment, drainage started considerably earlier (55 min and 6.2 h), suggesting possible bypass (leakage) through cracks in the soil or along the lysimeter wall. The soil had been quite dry at the start of the experiment, with the topsoil near wilting point. This ‘leakage’ was short lived and had minimal effect on the amount drained and no noticeable effects on leachate concentration. In all the lysimeters, soon after the drainage began, the rate increased sharply and reached an approximately steady rate after about half of a pore-volume had been drained. While the soil water content also increased after irrigation began, a plateau value was not reached until approximately one pore-volume had been drained (Figure 1). This is indicative that water flow was strongly dominated by preferential flow paths, with a slower wetting of the soil matrix, possibly reflecting the presence of soil hydrophobicity in the dry initial conditions. The potential for hydrophobicity has been shown to be widespread in New Zealand soils (Wallis et al. 1991; Carrick et al. 2011; Deurer et al. 2011).

Figure 1. Comparison of drainage rate and soil water content measured at four depths, average from six replicates of each irrigation treatment (irrigation rate of either: A, 5 mm h^–1 or B, 20 mm h^–1). Note that water content values are qualitative only as the probes were not specifically calibrated for this soil.

Bromide was detected in all samples and the concentration was highest in the first leachate collected or soon afterwards. It then decreased steadily, reaching nearly 0 at the end of the experiment for both irrigation intensities (Figure 2). There was no difference in the pattern of bromide leaching between lysimeters that received bromide 1 day prior to the experiment, compared with those that received it just before the irrigation commenced. Moreover, the concentrations of bromide in the leachate from the lysimeters under irrigation of 5 mm h^–1 were similar to those under 20 mm h^–1.

Figure 2. Concentrations of bromide in the leachate of lysimeters irrigated with 5 mm h^–1 (A) or 20 mm h^–1 (B). Bromide was applied in the soil surface 1 day prior to the experiments (●) or just before the experiment started (■).

Chloride was also detected in all samples, as it was present in the irrigation water (where the concentration averaged 20 mg L^–1). The chloride concentration of the leachate increased soon after the pulse was applied (after about 40 mm and 20 mm of irrigation had been applied for the 5 mm h^–1 and 20 mm h^–1 treatments, respectively), reaching a peak when less than half of a soil pore-volume has been drained (Figure 3). The peak happened earlier for the lysimeters receiving 20 mm h^–1 than for those receiving 5 mm h^–1, and although the peak concentrations appeared quite similar, the difference was statistically significant.

Figure 3. Concentrations of chloride in the leachate of lysimeters irrigated with 5 mm h^–1 (A) or 20 mm h^–1 (B). Chloride was applied when drainage amounted to about 180 mm.

Modelling analyses

The Burns’ equation was able to describe the experimental data well. When Equation (1) was fitted to the bromide data, it described about 96% of the variability with an RMSE of approximately 4% of the mean (Figure 4). The values for were found to be 0.23 and 0.24 (Table 2) for the 5 mm h^–1 and 20 mm h^–1 treatments, respectively, and the two values were not significantly different. These values indicate that only a small fraction of the soil’s water was responsible for the transport of the solute. Note that the soil was still quite dry at this stage of the experiment (Figure 1).

Figure 4. Average cumulative amount of bromide leached from lysimeters irrigated with 5 mm h^–1 (A) or 20 mm h^–1 (B). ■ indicate measured values, with whiskers representing the standard deviation of six replicates; the line is the Burns’ model using Equation (1). Also shown the measures for goodness-of-fit.

Table 2. Estimated values, and standard deviation, for the fraction of the soil water that takes part on solute transport (parameter f_t on Burns’ equation), for two irrigation intensities and two non-reactive tracers. Bromide was applied on dry soil while chloride on moist soil. See text for more details.

Irrigation intensity	Bromide	Chloride
5 mm h^–1	0.23 ± 0.003	0.85 ± 0.021
20 mm h^–1	0.24 ± 0.003	0.58 ± 0.008

Equation (2) fitted to the chloride data accounted for 96%–99% of the variability, with RMSE again at approximately 4% of the mean (Figure 5). The effect of irrigation intensity on the fitted model was significant. The fraction of the water that takes part in the solute transport () was significantly smaller for the lysimeters irrigated at 20 mm h^–1, 0.58, than those at 5 mm h^–1, 0.85 (Table 2). The effective fractional water volume approached the default value, that is, the water content at field capacity (Scotter et al. 1993), as the irrigation intensity decreased.

Figure 5. Average cumulative amount of chloride leached from lysimeters irrigated with 5 mm h^–1 (A) or 20 mm h^–1 (B). ■ indicate measured values, with whiskers representing the standard deviation of six replicates; the line is the Burns’ model using Equation (2). Also shown the measures for goodness-of-fit. Chloride pulse was applied when drainage amounted to about 180 mm.

Based on the analyses of the chloride data, we postulate that the value of will approach unity at a suitably low irrigation intensity, i.e. all soil water takes part in solute transport at very low water flows. As a proxy for the water application intensity when is close to 1, we determined the median hourly rainfall intensity at Lincoln, the experimental site (which was 0.5 mm h^–1 for records between 1999 and 2014 from a nearby weather station, data not shown). An exponential function fitted well the relationship described by the three pairs of values and irrigation (or rainfall) intensity data (also fitted using the nls method in R; Figure 6). The relationship suggests that preferential flow of solutes decreases at increasing rates as the irrigation intensity increases. It was estimated that, for the stony soil used in this experiment, under moist conditions, there is a lower limit of 0.35 for the value of . The variation in becomes very small after it reaches about 0.40, when irrigation intensities are approximately 50 mm h^–1.

Figure 6. Relationship between the fraction of water involved in solute transport () and irrigation intensity (). ■ indicate the results from fitting the Burns’ equation to chloride data from lysimeters irrigated with 5 mm h^–1 and 20 mm h^–1. □ represents the assumed value for (1.0) at median rainfall intensity for Lincoln (0.5 mm h^–1).

Discussion

Clear evidence for the transport of solutes via preferential flow was found in the experimental data. First, the fast onset of drainage and the stabilisation of the water flow before the soil water content reached equilibrium (Figure 1) are expected when the soil’s water is moving preferentially through a limited fraction of the pore space (Nimmo 2012). This is corroborated by the detection of bromide applied to the soil surface in the first sample of drained water (Figure 2). Moreover, the concentration of chloride, applied when the flow regime was near steady-state, peaked much earlier than would be expected if all the soil pore space was involved in the transport of solutes (Figure 3)—it would be at approximately one pore-volume. Results from other studies with similar New Zealand soils also found strong evidence for preferential flow (Silva et al. 2000; Pang et al. 2008; Carrick et al. 2014; McLeod et al. 2014).

The Burns’ equation provided a good description of the experimental data and was used to characterise the extent of preferential flow following the interpretation of Scotter et al. (1993). This approach is simpler than most standard models, requiring only basic soil properties and has only one fitting parameter () which has the merit of a practical physical interpretation: it is the fractional amount of water that takes part in solute transport. By examining Equations (1) and (2), it is apparent that smaller values of correspond to smaller drainage volume needed to induce solute leaching from the soil surface and consequently higher risk of losses. Based on the value of estimated from the 12 lysimeters (Table 2), the null hypothesis that irrigation intensity does not affect preferential solute flow was tested. The results for the chloride data were conclusive: the leaching rate is greater under higher irrigation intensity and thus the null hypothesis was rejected. The bromide data did not show significant differences, this was most likely because the soil was exceptionally dry at the start of the experiment. The results for bromide represent a case study in a contrasting initial condition to that of chloride, where the soil was near field capacity (Figure 1). The smaller values for found for bromide suggest an effect of antecedent soil moisture, which was also observed by Carrick (2009) in a deep loamy soil. Previous studies in New Zealand have typically used a standardised antecedent moisture condition near field capacity (Pang et al. 2008; McLeod et al. 2014). Results from this study suggest that further investigation is required to unravel any potential relationship between preferential flow behaviour and variation in antecedent soil moisture.

Our results add further evidence to that provided by previous studies of the importance of preferential flow process (e.g. Fraser 1992; McLeod et al. 1998; Pang et al. 2008; Carrick 2009) and the use of Burns’ equation to characterise it is promising. The results indicate that the risk for nutrient losses from irrigated soils that have a tendency for preferential flow would increase as the irrigation intensity increases. Irrigation is expanding on Canterbury Plains, especially in vulnerable stony soils (Carrick et al. 2013), therefore, greater attention should be paid to the management of irrigation systems to avoid unwanted environmental impact. By design, centre-pivots, the most commonly used irrigation system in New Zealand, have variable intensities along their length. The irrigation intensity should increase linearly along the length of a centre-pivot in order to deliver a constant irrigation amount (Powers 2012). High application intensity at the end of the irrigation line suggests higher levels of preferential flow and consequently a potentially larger risk of leaching losses. Moreover, the area irrigated by a sprinkler increases exponentially with the distance from the centre-point, so the actual risk for nutrient leaching of the whole irrigation system compounds quickly for larger centre-pivots. In this situation, the margin for errors on management becomes narrow and the number of mitigation options is very limited.

The results from the chloride trace data of this study, which best mimic a well-watered paddock, suggest an exponential relationship between irrigation intensity and the fraction of the soil’s water that takes part in solute transport (Figure 6). This relationship is based on only three points, so it warrants further research with a wider range of irrigation intensities. While the shape of this relationship should be studied further, including different soil types, the general trend shown in Figure 6 is sound: the value of should get closer to 1 as irrigation intensity approaches 0 (no preferential flow), the data show it will decrease at high intensities, but cannot go below 0. Based on the analyses of the chloride data of our study, it can be inferred that the value of will plateau out, with only small variations when the irrigation intensity is around 50 mm h^–1 or greater. The fraction of the soil water content effectively transporting water would reach a minimum of 0.35 for the stony soil used in this experiment. Note that the value of when used as in Equation (1) (which represents situations such as fertiliser applications) can be also interpreted as the relative speed at which leaching happens. This means that for the Lismore stony soil the volume of water applied required to induce drainage under high application intensities would be 35% smaller than that under very low intensity. Thus, leaching of a given amount of solute at the end of a large centre-pivot would only take about 35% of the time taken under typical average rainfall.

Conclusions

Results from a lysimeter experiment using a Lismore Stony Silt Loam soil under two irrigation intensities verified the occurrence of preferential flow. Using two non-reactive tracers (bromide and chloride), it was shown that antecedent soil moisture condition and irrigation intensity affected the extent of preferential solute flow. The data were successfully described using appropriate solutions of Burns’ equation, which was used to quantify the fraction of the soil pore space () involved in the transport of solutes. Strong preferential flow with , averaging 0.23, was verified under dry antecedent moisture conditions using bromide tracer, but with no effect from irrigation intensity. With the soil near field capacity, the value of estimated using chloride data decreased from 0.85 when irrigation intensity was 5 mm h^–1 to 0.58 under 20 mm h^–1. This difference was sufficient to confirm a significant (P < 0.05) effect of irrigation intensity on preferential flow. Therefore, the risk of leaching losses through preferential flow in irrigated soils should increase as irrigation intensity increases. Based on the analysis of chloride leaching, an exponential function was fitted to the data of versus irrigation intensity; this suggests that under moist conditions, the value of for the stony soil studied has a lower limit of 0.35, reached at high irrigation intensities. This relationship warrants further research under a wider range of irrigation intensities and in different soils.

Acknowledgements

This work was conducted through the Land Use Change and Intensification programme of Plant & Food Research. We also wish to thank Trevor Knight, Samuel Dennis, Frank Tabley and Graeme Rogers for helping on the collection and set-up of lysimeters or the conduction of the experiment. Thanks also to Chris Dunlop for the chemical analyses.

Disclosure statement

No potential conflict of interest was reported by the authors.