A 6-year comparison of nitrate leaching from grass/clover and N-fertilized grass pastures grazed by sheep

Transcription

1 Journal of Agricultural Science, Cambridge (1998), 131, Cambridge University Press Printed in the United Kingdom 39 A 6-year comparison of nitrate leaching from grass/clover and N-fertilized grass pastures grazed by sheep S. P. CUTTLE*, R. V. SCURLOCK AND B. M. S. DAVIES Institute of Grassland and Environmental Research, Plas Gogerddan, Aberystwyth, Ceredigion, SY23 3EB, UK (Revised MS received 31 December 1997) SUMMARY Nitrate leaching was measured over a 3-year period from rotationally grazed perennial ryegrass (Lolium perenne L.) pasture receiving 200 kg fertilizer-n ha and from similarly grazed ryegrass white clover (Trifolium repens L.) pasture that received no N fertilizer. The results are discussed together with those from the same plots in the preceding 3 years when they were stocked continuously. Under both managements, the numbers of grazing sheep were adjusted on the basis of the quantity of herbage available on the plots. During the whole 6 years, mean nitrate concentrations in soil water collected by porous cup samplers remained below the European Union limit of 11 3 mgn l except for the fertilized grass plots in year 5 of the study. Quantities of nitrate leached ranged from 6 to 34 kg ha per year from the grass clover plots and 2 46 kg ha from the fertilized plots. Leaching losses from both types of pasture were positively correlated with the numbers of lamb grazing days in the later part of the grazing season. This relationship and the high spatial variability associated with the measurements indicated that N derived from excreta was the main source of leached nitrate. It was concluded that, where pastures of equal productivity are compared, similar quantities of N are likely to be leached from grass clover swards as from grass swards receiving N fertilizer. INTRODUCTION Leaching from agricultural land is a major contributor of nitrate to groundwater and surface water catchments. Enhanced concentrations of nitrate increase the risk of eutrophication of water bodies and adversely affect the quality of drinking water supplies where concentrations approach or exceed the European Union limit of 11 3 mg nitrate-n l (EEC 1980). Much of the agricultural loss of nitrate originates from fertilized grassland. Grass itself is efficient at utilizing applied N and, under suitable conditions, annual fertilizer rates of kg N ha may be applied to cut swards without excessive leaching. However, introduction of the grazing animal reduces the efficiency of N-use and increases the risk of loss (Ryden et al. 1984). Ruminants excrete over 75% of the N they ingest (Jarvis et al. 1989) and grazing animals return this N directly to the pasture in localized patches of dung and urine. The high concentrations of N in urine patches are readily * To whom all correspondence should be addressed. steve.cuttle bbsrc.ac.uk hydrolysed in the soil to mineral forms which are susceptible to loss by leaching and other pathways (Haynes & Williams 1993). Nitrogen in dung is mineralized less rapidly and is thought to be less important as a source of leached nitrate (Lantinga et al. 1987; Deenen & Middelkoop 1992). Nitrate losses from grassland generally increase with increasing fertilizer inputs and stocking rates (Barraclough et al. 1992; Garrett et al. 1992) so that high losses are most frequently associated with the more intensively managed pastures. Clover-based swards, in which N is supplied by biological fixation of atmospheric N, have been presented as a more environmentally acceptable and sustainable alternative to fertilizer-based swards. However, it is unclear whether their potential benefits include a reduced risk of nitrate leaching. As N-fixation is generally insufficient to support the stocking rates that can be maintained with high fertilizer inputs, stock numbers and leaching losses may be expected to be less than those from intensively managed grassland. This was demonstrated in studies reported by Ryden et al. (1984) in which much less nitrate was leached from an unfertilized grass clover pasture than from a N-

2 40 S. P. CUTTLE, R. V. SCURLOCK AND B. M. S. DAVIES fertilized grass monoculture receiving 420 kg N ha. Other comparisons with pastures receiving more moderate rates of N fertilizer have also shown benefits of reduced nitrate losses from the clover-based pastures (Scholefield et al. 1992; Owens et al. 1994). However, investigations elsewhere have demonstrated that significant leaching of nitrate can occur from clover-rich pastures (Baber & Wilson 1972; Ball 1982) and from pure clover swards (Macduff et al. 1990). Parsons et al. (1991) questioned the desirability of encouraging a high proportion of clover in swards in view of the increased risk of N loss. It is unclear to what extent the lower leaching losses that have been reported are simply a result of reduced stocking rates or whether clover-based pastures offer additional benefits that are greater than those that could be achieved by reducing N inputs to fertilized pastures. This paper presents results of an investigation to compare nitrate leaching from an unfertilized grass clover pasture with that from a similarly managed grass pasture receiving a moderate rate of N fertilizer. The fertilizer rate was initially selected to match the productivity of the fertilized sward to that of the grass clover pasture; although, in practice, the productivity of both swards varied considerably in the different years of the investigation. Nitrate leaching was measured for 6 years between 1987 and 1993, initially for 3 years when the swards were stocked continuously and then for a further 3 years under a system of rotational stocking. Results from the first 3 years of the study were reported by Cuttle et al. (1992). The present paper describes the results for the following 3 years and analyses the findings for the 6-year period as a whole. MATERIALS AND METHODS Field site and pasture management The experimental site was located at an altitude of 35 m at the Institute of Grassland and Environmental Research, near Aberystwyth (Grid reference SN ) on an area of stony, well-drained fine loam soils (Dystric Cambisols) of the Denbigh 1 association (Rudeforth et al. 1984). The field was ploughed and reseeded in September 1985 to provide three 0 4 ha plots containing a mixed perennial ryegrass (Lolium perenne L., cv. Chieftain) white clover (Trifolium repens L., cv. Menna) sward and three similar plots with a pure ryegrass sward. Between 1987 and 1989, the plots were grazed continuously by ewes and lambs from March to June, then with lambs to November or December. Stock numbers were adjusted to maintain a target sward surface height of mm. Fertilizer was applied to the grass-only plots to supply a total of 150 kg N ha in 1987 and 200 kg N ha in later years. No N fertilizer was applied to the grass clover plots. Prior to the start of measurements in 1987, 20 ceramic cup samplers (Grossmann & Udluft 1991) were installed in a regular array in each plot. The majority of samplers were inserted with the cup at 1-m depth but in areas of shallow soil, installation depths varied between 0 5 and 1 m. Drainage volumes were measured as the mean of the volumes from four zero-tension lysimeters (0 5 m surface area, m depth) (Belford 1979) containing undisturbed monoliths of the grass clover sward and underlying soil collected from areas adjoining the experimental plots. Meteorological data were obtained from an automatic weather station in the study field and from a manually recorded station 0 8 km from the site. Further details of the site and the initial phase of the experiment are provided by Cuttle et al. (1992). The proportion of clover in the grass clover swards declined progressively during the first three years of the investigation and by the third year, stocking rates had fallen to half those on the fertilized grass plots. This was attributed, in part, to the system of continuous stocking. In sheep-grazed pastures, the persistency of medium leaf-sized varieties of white clover, such as Menna, is considered to be favoured by rotational rather than continuous grazing (Evans et al. 1992; Swift et al. 1992). In 1990, management of the grass clover and N-fertilized grass plots was therefore changed to a 6-week rotational grazing system. In addition, further clover seed (cv. Menna) was broadcast over the grass clover plots in June 1990 at a rate of 2 5 kg ha. For the introduction of rotational grazing, additional fences were erected to divide all plots into halves and provide six subplots within each pasture type. This allowed subplots to be grazed in turn for 1 week followed by a 5-week regrowth period without grazing. Plots were grazed between April and November or December, with five complete grazing circuits followed by a final, more rapid circuit to remove any excess herbage at the end of the growing season. During the first four circuits, ammonium nitrate fertilizer was applied to each grass subplot 2 weeks before it was due to be grazed; each plot receiving 60, 50, 50 and 40 kg N ha at successive applications to provide a nominal total of 200 kg N ha during the year (Table 1). No fertilizer was applied during the later circuits, after the beginning of September. In 1992 the rotation was modified to allow better control of grazing during the period of maximum herbage growth. Between May and July, both halves of the original plots were grazed as a single unit, effectively halving the number of plots in the rotation and allowing four 3-week circuits of one week grazing followed by a 2-week regrowth period. During these more rapid cycles, N fertilizer continued to be applied to the grass plots 2 weeks before grazing but at half the rate of the equivalent applications in previous years so as not to increase the total quantity of N applied. The number of sheep was adjusted at the start of each grazing period on the basis of the

3 Nitrate leaching from pastures 41 Table 1. Mean clover contents in grass clover plots and fertilizer inputs to grass plots in Years 1 6 with numbers of grazing days before and after weaning. Each ewe lamb unit in the pre-weaning period is equivalent to one ewe plus twin lambs Continuous stocking Rotational stocking Year 1 Year 2 Year 3 Year 4 Year 5 Year 6 Grass clover pasture Mean clover content (%)* 35 (4 5) 11 (2 0) 4 (1 3) 19 (5 6) 24 (0 4) 8 (1 5) Grazing days ha Pre-weaning (ewe lamb units) Post-weaning (lambs) Fertilized grass pasture N fertilizer (kg N ha) Grazing days ha Pre-weaning (ewe lamb units) Post-weaning (lambs) * Mean clover content (April September) as percentage of herbage dry matter. quantity of herbage available in the subplot that was to be grazed. Sward surface height was measured immediately before the start of grazing and converted to an estimate of herbage mass using a previously determined relationship between mass and sward height. The number of sheep required to graze this quantity of herbage within a 7-day period was calculated from the expected daily intake of ewes and lambs based on their liveweight. The calculations included an allowance for herbage growth during the week and a residual herbage mass at the end of grazing equivalent to a sward surface height of mm. During the course of the experiment, small patches of clover had developed in the fertilized grass plots and in May 1992 the affected areas were sprayed with a selective herbicide to eliminate most of the clover from these plots. Other aspects of management and experimental procedures were kept as similar as possible to those in the first 3 years of the investigation. As in previous years, phosphorus and potassium fertilizers and lime were applied to all plots as necessary on the basis of annual soil analyses to maintain the nutrient content and ph of the soil at recommended values (MAFF 1988). No N fertilizer was applied to the grass clover plots. Soil water samples were collected weekly whenever the soil was sufficiently moist to provide samples from the majority of samplers. Water samples were stored at 4 C prior to analysis, which was normally completed within 4 days of sample collection. Concentrations of nitrate nitrite-n were determined directly on the samples using a segmented flow analyser and a colorimetric method adapted from Henriksen & Selmer-Olsen (1970). Occasional analyses of randomly selected samples indicated that nitrite contents were negligible and the results are therefore reported as nitrate-n concentrations. Concentrations of ammonium-n were similarly determined using a method adapted from Crooke & Simpson (1971). Analysis of the data In the earlier paper (Cuttle et al. 1992), results for the first 3 years of the investigation were described on the basis of hydrological years, between April and the following March, referred to as Years 1 to 3. This convention is continued in the present paper for Years 4 to 6. As in the earlier years, where drainage continued beyond the end of the hydrological year, annual nitrate losses are calculated up to the end of drainage rather than to the end of March. Quantities of nitrate leached were calculated separately for each sampler and sampling date as the product of the nitrate concentration and corresponding weekly drainage volume. The total quantity leached at each position during the winter was then calculated as the sum of these weekly values. On the few occasions during winter when individual samplers failed to collect a water sample, concentrations were estimated as the mean of the concentrations from the particular sampler in the weeks immediately preceding and following the period for which data were unavailable. Where samplers failed to provide water samples over an extended period, they were excluded from the calculations for that period. Overall leaching losses were calculated as the mean of values of all 60 samplers within a sward type. Means were estimated using Sichel s estimator, x s, assuming a log-normal distribution for each set of 60

4 42 S. P. CUTTLE, R. V. SCURLOCK AND B. M. S. DAVIES values (Sichel 1952; Parkin & Robinson 1992). Four additional samplers had been installed in each plot in Year 3 to provide data from camping areas where sheep appeared to congregate and which would be expected to receive particularly high returns of excreta. The introduction of rotational grazing in Year 4 appeared to modify the animals behavioural patterns such that these positions were no longer representative of camping areas and data from these samplers are therefore omitted from the calculations in this paper. RESULTS Pasture and livestock performance Results for Years 4 6 are described in detail with comparative data for Years 1 3 included in the relevant tables. The clover content of the grass clover plots in Years 4, 5 and 6 was consistently greater than under the continuous stocking management in Year 3 (Table 1). The content was greatest in Year 5 when clover was uniformly distributed throughout the plots, unlike the previous year when there were considerable differences between plots. The greatest stock-carrying capacity of the grass clover swards, expressed as the number of sheep grazing days (i.e. number of sheep number of days for which they were grazing), coincided with the maximum clover content in Year 5. Although there was less clover in all plots in the final year, the number of grazing days in Year 6 was greater than in Year 4. High sheep numbers in the pre-weaning period in Year 6 suggested a carry-over of fertility from the preceding year. Although the overall mean clover content between Years 4 and 6 was similar to the mean for Years 1 3, the overall stocking rate was c. 30% greater than under con- tinuous stocking. Numbers of grazing days for the N- fertilized grass plots were greater than those for the grass clover plots in Years 4, 5 and 6 (Table 1). However, the pattern of variation was similar in both types of pasture, with the highest stocking in Year 5 and lowest in Year 4. Sheep numbers on the fertilized plots were consistently greater under rotational grazing than under continuous stocking, in both the pre- and post-weaning periods. In Year 4, pastures were grazed from 26 March to 23 November but because of excess grass it was necessary to return sheep to graze the grass plots for a further period between 4 and 14 December. The grazing periods in Years 5 and 6 were the same for both types of pasture; from 25 March to 29 November and from 1 April to 25 November, respectively. Rainfall and drainage volumes The annual rainfall recorded between April and March was similar in Years 4 and 5 but slightly greater in the final year (Table 2). Differences between years are more evident if summer and winter rainfall are considered separately. The April September period in Year 4 was appreciably drier than in the following years. There were 179 days during the summer with no drainage from the lysimeters, compared with 88 and 71 days in Years 5 and 6, respectively. Unlike other years, drainage had ceased in March, before the start of Year 4 and it then restarted later in the autumn. Potential soil moisture deficits continued to develop later into the summer so that the maximum deficit occurred later and was ultimately greater than in other years. The smallest moisture deficit occurred in Year 5, even though there was slightly less rainfall that summer than in Year 6. Table 2. Summer and winter rainfall and maximum potential soil moisture deficit (from 1 March) in Years 1 6 with mean quantities of drainage from the lysimeters. The drainage period refers to periods of continuous drainage, except in Year 5 when drainage started on 29 July and then temporarily ceased between 28 August and 25 September. In Year 6, drainage continued after the end of measurements on 1 April Year April September Rainfall (mm) October March Maximum potential soil moisture deficit (mm) Total drainage (mm) ( S.E.) (D.F. 3) Drainage period Continuous stocking 1 ( ) ( 9) 28 Sep 26 Apr 2 ( ) ( 4) 20 Jul 17 May 3 ( ) ( 11) 11 Oct 28 Mar Rotational stocking 4 ( ) ( 9) 23 Sep 30 May 5 ( ) ( 12) 29 Jul 4 Jun 6 ( ) ( 16) 14 Aug 1 Apr

5 Nitrate leaching from pastures (a) Mean nitrate-n concentration (mg N/l) (b) (c) Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Month Fig. 1. Mean nitrate-n concentrations (Sichel s estimator) in soil water samples from the grass clover plots ( ) and fertilized grass plots ( ) in(a) Year 4 ( ), (b) Year 5 ( ) and (c) Year 6 ( ). Bars are only shown for those weeks when there was drainage from the lysimeters. Drainage from the lysimeters also restarted earlier than in other years. Winter rainfall, between October and March, was least in Year 5 and resulted in the smallest drainage volume from the lysimeters. Winter rainfall in Year 4 was similar to that in Year 6 and drainage volumes were similar in both years. In all three years, drainage continued beyond the end of March. In Years 4 and 5, it did not end until late May or early June of the following hydrological year. Although measurements in the final year were not continued after the end of March, rainfall in April, May and June was similar to or greater than in previous years, indicating that drainage would have again continued into May or later. Lysimeter drainage volumes and summer and winter rainfall values in Years 4 6 were within the ranges recorded in Years 1 3 of the experiment. Nitrate concentrations and quantities leached As in the initial phase of the investigation, spatial patterns of leaching were highly variable with widely differing nitrate concentrations in water collected from different sampling positions. Within a set of samples collected on a single date in winter, nitrate concentrations could range from 0 01 to 100 mg N l.

6 44 S. P. CUTTLE, R. V. SCURLOCK AND B. M. S. DAVIES This variability was characteristic of samples from both the grass clover and fertilized grass plots. Temporal patterns of variation for individual sampling positions were more predictable. Nitrate concentrations were low in summer and increased to a peak in mid-winter before declining to their initial values in spring. The spatial variability arose from differences in the magnitude of this winter peak. Data sets were positively skewed and generally conformed to lognormal distributions. Mean nitrate concentrations (x ), assuming a log-normal distribution for both s pasture types, are shown in Fig. 1. Mean concentrations from the fertilized grass plots were consistently greater than those from the grass clover plots. In both types of plot, the highest concentrations occurred in Year 5. Concentrations of ammonium-n in soil water samples remained 0 05 mg N l in all years and did not contribute significantly to the mean N loss from either pasture type. The widely differing nitrate concentrations resulted in similarly wide ranges in the estimates of nitrate leached at different sampling positions. Losses in Years 4 6 ranged from 5 to mg nitrate-n m for individual samplers in the grass clover pasture and from 4 to mg m in the case of the fertilized grass pasture. Data sets were positively skewed in each year of the study (Table 3). Application of D Agostino s method to test for normality of the ln-transformed data (Parkin & Robinson 1992) demonstrated that all data sets conformed to lognormal distributions (P 0 05) with the exception of values for the grass clover pasture in Year 3. Deviations from the theoretical distribution were not sufficiently consistent to justify the use of a more complex model and mean values were therefore calculated using Sichel s estimator (x ), assuming a s log-normal distribution in all cases. Mean values of the quantities of nitrate leached from the grass clover and fertilized grass pastures in Years 4 6 are shown in Table 3. In all three years, the mean loss from the grass clover plots was less than from the grass pasture. The range of values for the annual leaching loss from the grass clover plots was similar to that measured under continuous stocking in Years 1 3 (Table 3). Similarly, losses from the fertilized grass pasture under rotational stocking were equal to or greater than those measured under continuous stocking in Years 1 3. The mean values presented in Table 3 underestimate the true loss as they do not include an allowance for the greater leaching of N from camping areas. The error is likely to be greatest in Years 1 3 under continuous stocking as there would have been a greater tendency for sheep to camp than in the more densely stocked, rotationally grazed pastures (Haynes & Williams 1993). Measurements of losses from camping areas in Year 3 indicated a possible underestimate of up to 5 kg N ha per year in the case Table 3. Estimates of the mean quantities of nitrate-n leached from the grass clover and fertilized grass plots in Years 1 6, expressed as Sichel s estimate of the mean (x s ) Quantity of nitrate leached (kg N ha) Coefficient of skewness 95% Confidence limits Mean (x ) Lower Upper s Grass clover pasture Continuous stocking Year Year Year Rotational stocking Year Year Year Fertilized grass pasture Continuous stocking Year Year Year Rotational stocking Year Year Year of the continuously grazed plots (Cuttle et al. 1992). Losses in the final year would also be underestimated because measurements finished at the end of March rather than continuing until drainage ceased. This error is likely to be small. Except in Year 1 when there was direct leaching of the fertilizer applied in spring, the additional N loss in drainage after March did not exceed 7% of the total annual loss and averaged only 4%. DISCUSSION Changes in the productivity of the fertilized grass and grass clover swards Throughout the investigation, stocking rates were adjusted on the basis of the quantity of herbage available to the sheep. The number of grazing days therefore represents a suitable measure for comparing the overall productivity of the pastures in different years and is independent of the change from continuous to rotational stocking. The stock-carrying capacity of the fertilized grass plots was greater in Years 4 6 than in the first 3 years of the experiment (Table 1). The low output in Year 1 can to some extent be attributed to the smaller fertilizer input but not in the following 2 years when the fertilizer rate

7 Nitrate leaching from pastures 45 was equal to that in Years 4 6. The increase in stocking rates coincided with the change from continuous to rotational grazing but was greater than would be expected to occur solely from the change in management. Frame (1992) observed that in most comparisons of grazing systems, where other factors were equal, there was very little difference in animal output from continuously and from rotationally stocked pastures. In the present investigation, visual observations suggested that the increased productivity may have been associated with increasing net mineralization of soil N in the years after reseeding of the pasture. In the initial years, the fertilized grass was frequently pale green or yellow and appeared to be N deficient compared to a more uniform, darker green in later years. The field had been cropped with cereals for several years before the swards were sown for the current experiment and this may have depleted the soil organic matter sufficiently to restrict the supply of mineralized N in Years 1 and 2. In later years there were indications that moisture stress limited productivity in dry summers. There was also considerable variation in the stockcarrying capacity of the grass clover plots during the 6-year study although this variation appeared to be independent of that exhibited by the fertilized plots. Productivity was greatest in Year 1 and showed no evidence of the initial restriction to growth that was evident in the fertilized grass plots. Stocking rates tended to be highest in those years when there was a high proportion of clover in the sward but this relationship was not statistically significant (P 0 05). A more dominant factor appeared to be the availability of soil moisture. The number of grazing days in the post weaning period was negatively correlated with the maximum potential soil moisture deficit during the year (P 0 01). Although the mean stocking rate during the period of rotational grazing was greater than under continuous grazing, rates were too variable to be able to attribute the increase to the change in management. Leaching from the N-fertilized grass pasture Mean nitrate concentrations in soil water from the fertilized grass plots exceeded the European Union limit of 11 3 mg N l during much of the winter in Year 5 but remained below this value in all other years of the investigation. Previous estimates (Cuttle et al. 1992) indicated that where calculations included an allowance for the greater loss of N from camping areas, mean concentrations also approached this limit for a number of weeks during the winter of Year 3. The wide range of nitrate losses measured at different sampling positions is similar to the marked spatial variability found in measurements of soil mineral-n in grazed pastures (White et al. 1987; Afzal & Adams 1992; Haynes & Williams 1993). This heterogeneity has been attributed to the uneven deposition of excreta, with large losses occurring from those areas of soil where urine was deposited during the grazing season. If urine patches represent the major source of leached nitrate, then the N loss from the pasture as a whole will be largely determined by the proportion of the pasture area that is affected by urine and thus by the number of grazing stock. In the present study, annual leaching losses were positively correlated (P 0 05) with the number of lamb grazing days between weaning in late June and the end of grazing but were not correlated with sheep numbers in the pre-weaning period. The highest losses coincided with the period of rotational grazing in Years 4 6 but there was no indication that the change from continuous to rotational grazing had a direct influence on the quantity of nitrate leached, other than through the impact of the greater stock numbers in these later years. These results are in contrast to those reported by Brock et al. (1990) for sheep-grazed pastures in New Zealand where leaching losses under set-stocking were 2 3 times those from rotational grazing. Annual leaching losses were not significantly correlated with either the total winter drainage volume or the maximum potential soil moisture deficit during the summer (P 0 05). Assuming piston flow and solute displacement, winter rainfall in all years was sufficient to have leached nitrate to below sampling depth and the total quantity of nitrate leached would therefore be expected to be independent of drainage volume. The absence of a relationship between N loss and the degree of moisture stress is in contrast to results reported by other workers. Scholefield et al. (1993) measured nitrate leaching from fertilized, cattle-grazed swards on a clay soil in SW England. Over a 7-year period, there was a positive relationship between the quantity of nitrate leached and the maximum potential soil moisture deficit during the grazing season. Other studies have also reported increased leaching of nitrate in winters following summer droughts (Garwood & Tyson 1977; Webster & Dowdell 1984). These effects have been ascribed to the greater accumulation of nitrate in the soil in dry years as a result of reduced uptake by the sward, reduced denitrification and an increased flush of mineralization on rewetting the soil. However, these increases may be offset by increased volatilization of ammonia from urine patches in hot, dry conditions (Haynes & Williams 1993) and by a reduction in the total number of urine patches if herbage growth and the numbers of grazing animals are restricted by drought. Dry summers had a greater impact on stocking rates at the Aberystwyth site than in the study reported by Scholefield et al. (1993), in spite of much greater soil moisture deficits in the latter experiment (Tyson et al. 1992). Reductions in stocking

8 46 S. P. CUTTLE, R. V. SCURLOCK AND B. M. S. DAVIES rates and in the area affected by urine may have been sufficient to counteract those factors tending to increase losses in dry years and explain the absence of a relationship between N leaching and soil moisture at the Aberystwyth site. Leaching from the grass clover pasture Mean nitrate concentrations in water samples from the grass clover plots never exceeded the European Union limit of 11 3 mg N l although peak concentrations approached this value for brief periods in Years 1 and 5. Concentrations were consistently greater than those from the fertilized plots during the first two winters but this pattern was reversed from Year 3 onwards. As with the fertilized grass pasture, the quantity of N leached from the grass clover plots was positively correlated with the number of grazing days between July and December (P 0 05), indicating that the proportion of pasture affected by excreta was a dominant factor determining losses. However, N losses were most closely correlated with the proportion of clover in the sward (P 0 01). This positive correlation may have resulted from the statistical interaction between clover content, stocking rate and N loss but may also be indicative of direct leaching of N from clover. The proportion of clover frequently declines during winter (Harris et al. 1983; Woledge et al. 1990) and the death and decomposition of clover roots and nodules may add to the nitrate that is available for leaching. The potential for clover monocultures to release significant quantities of N was demonstrated in lysimeter studies by Low (1973) but losses are likely to be less in mixed swards where grass is present to utilize any N that is released during the growing season (Jones et al. 1974). During the present investigation, there were reductions in the clover content of the sward that were indicative of clover death but the spatial patterns of N loss suggest that this did not contribute significantly to the quantities of N leached. In those years when it was abundant, clover was relatively uniformly distributed in the sward and any loss from this source would be expected to affect nitrate concentrations at the majority of sampling positions. In practice, the mean loss in all years was largely determined by high losses from a few sampling positions, assumed to be indicative of urine patches, rather than from a change in the overall, background loss. Nitrate losses from the grass clover plots were not significantly correlated with maximum soil moisture deficits or with winter drainage volumes. Comparison of leaching from the fertilized grass and grass clover pasture Leaching measurements in both types of pasture were characterized by high spatial variability and skewed N leached (kg N/ha) Lamb grazing days/ha (July December) Fig. 2. Relationship between the quantity of nitrate leached and total number of lamb grazing days between July and December for the grass clover ( ) and fertilized grass plots ( ) in Years 1 6. The dashed line denotes the fitted regression line of the combined data sets from the two pastures (see text). distributions indicating that excess N from excreta was the main source of nitrate in both cases. Nitrate losses from both types of pasture were positively correlated with the numbers of grazing days and could be described by the following equations for the regression of N leached on number of grazing days between July and December: Fertilized grass pasture: N leached (kg ha) Grazing days 11 4 S.E. ( ) ( 10 85) (r 0 733, D.F. 4, P 0 05) Grass clover pasture: N leached (kg ha) Grazing days 5 9 S.E. ( ) ( 7 60) (r 0 676, D.F. 4, P 0 05) Neither the gradients nor intercepts of the two lines differed significantly, indicating that both relationships could be described by a single expression (Fig. 2): N leached (kg ha) Grazing days 9 5 S.E. ( ) ( 5 94) (r 0 718, D.F. 10, P 0 001) The similarities between the separate responses indicate that where grass clover pastures are compared with fertilized pastures of equal productivity, similar quantities of N will be leached from both types of pasture, irrespective of whether N was supplied as fertilizer or by biological fixation. In spite of these relationships between stocking rates and losses, the number of grazing days is a coarse measure of the quantity of urine-derived N that is likely to be available for leaching when the soil returns to field capacity in the autumn. In particular,

9 Nitrate leaching from pastures 47 Table 4. Correlations between the quantity of nitrate leached and the number of lamb grazing days during different portions of the post-weaning grazing period: calculated for the combined data from the grass clover and fertilized grass pastures (D.F. 10) Period Correlation coefficient (r) P Lamb days in period as % of July December total Jul Dec Aug Dec Sep Dec Oct Dec Nov Dec Jul Nov Jul Oct Jul Sep Jul Aug it assumes a similar contribution from all urine patches, irrespective of the date on which they were deposited. Where urine is deposited early in the season, gaseous losses and plant uptake may be sufficient to remove most of the N from the soil before leaching commences. Urinations later in the grazing season, when opportunities for uptake are reduced, will leave more N available for leaching. This was demonstrated in a previous experiment in which artificial urine was applied to small plots to simulate deposition of sheep urine on different dates between July and November (Cuttle & Bourne 1993). The quantity of N remaining in the soil at the start of leaching was inversely related to the number of days since urine application. Only 3% of the N from urine applied in July remained in the soil in November. Similarly, Thomas et al. (1988) found that enhanced contents of mineral-n persisted in soil for days after applications of sheep urine to an upland grass sward and Sherwood & Fanning (1989) reported that where cattle urine was applied to pasture, only applications made after August resulted in increased N concentrations in the soil in December. Thus, in the present experiment, it would be expected that correlations between N loss and sheep numbers would be improved by restricting consideration to grazing days in the later part of the season when the impact of urinations would be greatest. In practice, this was not so; progressively excluding grazing days for July, August and September reduced correlation coefficients compared with those for the July December period, whereas excluding grazing days from December, November and October had little effect (Table 4). These relationships were investigated further using more accurate estimates of the area affected by urine, calculated from a simple model based on the numbers of sheep, their weights, published values of the area of a typical urine patch and number of urinations per day. The content of urine-n in the soil was set to decline at a linear rate as indicated by the study of Cuttle & Bourne (1993). In this case, the model overestimated the impact of grazing later in the season and estimates of soil nitrate were poorly correlated with measured losses. Correlation coefficients were improved by reducing the rate at which urine-n was assumed to decline in the soil but the improved correlations were no better than those obtained simply using numbers of lamb grazing days between July and December. These relationships suggest that lamb numbers in July and August were more important in determining the magnitude of nitrate loss than those at the end of the grazing season. In part, this is a reflection of the greater numbers of sheep on the plots earlier in the summer and their greater contribution to the total grazing days (Table 4). Hack-ten Broeke et al. (1996) used more sophisticated models of excretion, leaching and N dynamics to examine the contribution of excreted N to nitrate leaching from cattle-grazed swards in the Netherlands. The simulations only partially explained the measured variability that was assumed to be associated with urine patches: high concentrations of nitrate were attributed to urinations in September and later but also to overlapping of urine patches. Although stocking rates and urine returns appear to be the major factors determining nitrate leaching, there are differences between clover- and fertilizerbased pastures which may influence the relative quantities of N leached from these swards. The similar quantities of nitrate leached from both types of pasture in the present investigation, and also in a similar experiment at an upland site (Cuttle et al. 1996), indicate that the net effects of any such differences are small. However, the measurements are subject to the uncertainties arising from the high spatial variability present in grazed pastures and may not be sufficiently precise to detect differences that do exist. Additional information from detailed examination of the data for individual water samplers provided no evidence of differences in the processes controlling nitrate leaching in the two types of pasture. Over the 6 years, the ranges and distributions of the point measurements of nitrate loss were broadly similar for the grass clover and fertilized grass plots. There was no indication of consistent differences in the upper part of the range that might be expected to arise if inhibition of N-fixation reduced further inputs of N to urine patches in grass clover pastures (Ledgard et al. 1982; Marriott et al. 1987; Boursier et al. 1989). In these circumstances, losses would be less than from urine patches in fertilized pasture where subsequent fertilizer applications would add to the excess N already available for leaching. Similarly,

10 48 S. P. CUTTLE, R. V. SCURLOCK AND B. M. S. DAVIES these comparisons provided no evidence that the N content of urine differed in the two types of pasture (Jarvis 1992). Other processes, such as the accumulation of fertilizer-n during drought or the release of N from clover roots, may influence losses from the pasture area as a whole, rather than acting specifically on areas affected by urine. Differences in background leaching, as represented by the majority of water samplers, were small and indicate that these processes did not have a dominant influence on the overall quantities of N that were leached. The only evidence of a direct loss of fertilizer-n occurred in spring at the end of Year 1 when heavy rain fell shortly after fertilizer had been applied. Although this represented a loss of only 6 kg N ha, it was responsible for most of the total of 7 kg ha leached from the fertilized grass plots during the year (Cuttle et al. 1992). Such losses may be of greater significance in pastures on poorly drained soils where there is a greater risk of surface run-off (e.g. Cuttle & James 1995). Clearly, this form of direct loss would not occur from grass clover pastures that did not receive N fertilizer. In a similar investigation, Scholefield & Tyson (1992) compared nitrate leaching from cattle-grazed grass clover plots and from equivalent N-fertilized grass plots. Less nitrate was leached from the grass clover pasture but animal output per ha was also lower and, as in the present study, it was concluded that at equal levels of animal production, leaching losses were likely to be similar for both types of pasture. Hutchings & Kristensen (1995) developed a model of herbage growth and N dynamics in grazed pasture and compared nitrate leaching from grass clover and fertilized grass pastures. Predicted leaching losses from grass clover pasture were less than from fertilized grass but differences were small at herbage yields corresponding to those of clover-based pastures in the UK and for the quantities of N leaching measured in the current investigation. The authors concluded that the small differences in the quantities of N leached and the large spatial differences may explain why field investigations have failed to detect significant differences between the two types of pasture. These conclusions are in accord with those of the present investigation which demonstrated that nitrate losses from grass clover pastures are likely to be similar to those from fertilizer-based pastures of equal productivity but also demonstrated the high spatial variability and imprecision associated with these measurements. In spite of the similarity between the quantities of nitrate leached from the two types of pasture, grass clover swards can contribute towards reducing nitrate leaching from grassland. The savings in fertilizer costs that accompany the change to clover-based swards increase the economic viability of this form of less-intensive grazing system (Doyle & Bevan 1996) and favour their more widespread adoption. Reductions in nitrate leaching will occur as a result of the reduction in grazing intensity rather than any intrinsic difference in the quantities of N leached from equivalent clover- and fertilizer-based pastures. This work was funded by the Ministry of Agriculture, Fisheries and Food under contract CSA We are grateful to the farm staff at Aberystwyth for their care and management of the sheep during the 6-year period, to members of the scientific staff who contributed advice and assistance and to ICI Fertilizers plc who donated the fertilizers used in the investigation. AFZAL, M.& ADAMS, W. A. (1992). Heterogeneity of soil mineral nitrogen in pasture grazed by cattle. Soil Science Society of America Journal 56, BABER, H. L.& WILSON, A. T. (1972). Nitrate pollution of ground water in the Waikato Region. Journal of the New Zealand Institute of Chemistry 36, BALL, P. R. (1982). Nitrogen balances in intensively managed pasture systems. In Nitrogen Balances in New Zealand Ecosystems, Proceedings of Workshop, Palmerston North, May 1980, pp New Zealand: DSIR. BARRACLOUGH, D., JARVIS, S. C., DAVIES, G. P.& WILLIAMS, J. (1992). The relation between fertilizer nitrogen applications and nitrate leaching from grazed grassland. Soil Use and Management 8, BELFORD, R. K. (1979). Collection and evaluation of large soil monoliths for soil and crop studies. Journal of Soil Science 30, BOURSIER, P. J., RAGUSE, C. A.& TAGGARD, K. L. (1989). Growth and nitrogen-fixing responses of subterranean clover to application and subsequent removal of ammonium nitrate. Crop Science 29, REFERENCES BROCK, J. L., BALL, P. R.& CARRAN, R. A. (1990). Impacts of management on leaching of nitrate from pastures. Proceedings of the New Zealand Grassland Association 52, CROOKE, W. M.& SIMPSON, W. E. (1971). Determination of ammonium in Kjeldahl digests of crops by an automated procedure. Journal of the Science of Food and Agriculture 22, CUTTLE, S. P.& BOURNE, P. C. (1993). Uptake and leaching of nitrogen from artificial urine applied to grassland on different dates during the growing season. Plant and Soil 150, CUTTLE, S. P.& JAMES, A. R. (1995). Leaching of lime and fertilisers from a reseeded upland pasture on a stagnogley soil in mid-wales. Agricultural Water Management 28, CUTTLE, S. P., HALLARD, M., DANIEL, G. & SCURLOCK, R. V. (1992). Nitrate leaching from sheep-grazed grass clover and fertilized grass pastures. Journal of Agricultural Science, Cambridge 119, CUTTLE, S. P., HALLARD, M., GILL, E. K. & SCURLOCK,

11 Nitrate leaching from pastures 49 R. V. (1996). Nitrate leaching from sheep-grazed upland pastures in Wales. Journal of Agricultural Science, Cambridge 127, DEENEN, P. J. A. G.& MIDDELKOOP, N. (1992). Effects of cattle dung and urine on nitrogen uptake and yield of perennial ryegrass. Netherlands Journal of Agricultural Science 40, DOYLE, C. J. & BEVAN, K. (1996). Economic effects of legume-based grassland systems. In Legumes in Sustainable Farming Systems. BGS Occasional Symposium No. 30 (Ed. D. Younie), pp Reading: British Grassland Society. EUROPEAN ECONOMIC COMMUNITY (1980). Council Directive on the Quality of Water for Human Consumption. Official J. No EEC L229,11. EVANS, D. R., WILLIAMS, T. A. & EVANS, S. A. (1992). Evaluation of white clover varieties under grazing and their role in farm systems. Grass and Forage Science 47, FRAME, J. (1992). Improved Grassland Management. Ipswich: Farming Press. GARRETT, M. K., WATSON, C. J., JORDAN, C., STEEN, R. W. J.& SMITH, R. V. (1992). The Nitrogen Economy of Grazed Grassland. Proceedings No Peterborough: The Fertiliser Society. GARWOOD, E. A. & TYSON, K. C. (1977). High loss of nitrogen in drainage from soil under grass following a prolonged period of low rainfall. Journal of Agricultural Science, Cambridge 89, GROSSMANN, J.& UDLUFT, P. (1991). The extraction of soil water by the suction-cup method: a review. Journal of Soil Science 42, HACK-TEN BROEKE, M. J. D., DE GROOT, W. J. M. & DIJKSTRA, J. P. (1996). Impact of excreted nitrogen by grazing cattle on nitrate leaching. Soil Use and Management 12, HARRIS, W., RHODES, I.& MEE, S. S. (1983). Observations on environmental and genotypic influences on the overwintering of white clover. Journal of Applied Ecology 20, HAYNES, R. J.& WILLIAMS, P. H. (1993). Nutrient cycling and soil fertility in the grazed pasture ecosystem. Advances in Agronomy 49, HENRIKSEN, A.& SELMER-OLSEN, A. R. (1970). Automatic methods for determining nitrate and nitrite in water and soil extracts. Analyst 95, HUTCHINGS, N. J. & KRISTENSEN, I. S. (1995). Modelling mineral nitrogen accumulation in grazed pasture: will more nitrogen leach from fertilized grass than unfertilized grass clover? Grass and Forage Science 50, JARVIS, S. C. (1992). Nitrogen cycling and losses in clover based pastures. Herba 5, JARVIS, S. C., HATCH, D. J.& ROBERTS, D. H. (1989). The effects of grassland management on nitrogen losses from grazed swards through ammonia volatilization; the relationship to excretal N returns from cattle. Journal of Agricultural Science, Cambridge 112, JONES, M. B., STREET, J. E. & WILLIAMS, W. A. (1974). Leaching and uptake of nitrogen applied to annual grass and clover-grass mixtures in lysimeters. Agronomy Journal 66, LANTINGA, E. A., KEUNING, J. A., GROENWOLD, J. & DEENEN, P. J. A. G. (1987). Distribution of excreted nitrogen by grazing cattle and its effects on sward quality, herbage production and utilization. In Animal Manure on Grassland and Fodder Crops: Fertilizer or Waste? (Eds H. G. van der Meer, R. J. Unwin, T. A. van Dijk & G. C. Ennik), pp Dordrecht, The Netherlands: Martinus Nijhoff Publishers. LEDGARD, S. F., STEELE, K. W. & SAUNDERS, W. H. M. (1982). Effects of cow urine and its major constituents on pasture properties. New Zealand Journal of Agricultural Research 25, LOW, A. J. (1973). Nitrate and ammonium nitrogen concentration in water draining through soil monoliths in lysimeters cropped with grass or clover or uncropped. Journal of the Science of Food and Agriculture 24, MACDUFF, J. H., JARVIS, S.C.&ROBERTS, D. H. (1990). Nitrates: leaching from grazed grassland systems. In Nitrates Agriculture Eau (Ed. R. Calvet), pp Paris: INRA. MARRIOTT, C. A., SMITH,M.A.&BAIRD, M. A. (1987). The effect of sheep urine on clover performance in a grazed upland sward. Journal of Agricultural Science, Cambridge 109, MINISTRY OF AGRICULTURE, FISHERIES AND FOOD (1988). Fertilizer Recommendations for Agricultural and Horticultural Crops. Reference Book 209. London: HMSO. OWENS, L. B., EDWARDS, W. M. & VAN KEUREN, R. W. (1994). Groundwater nitrate levels under fertilized grass and grass-legume pastures. Journal of Environmental Quality 23, PARKIN, T. B. & ROBINSON, J. A. (1992). Analysis of lognormal data. Advances in Soil Science 20, PARSONS, A. J., ORR, R. J., PENNING, P. D.& LOCKYER, D. R. (FOR RYDEN, J. C.) (1991). Uptake, cycling and fate of nitrogen in grass clover swards continuously grazed by sheep. Journal of Agricultural Science, Cambridge 116, RUDEFORTH, C. C., HARTNUP, R., LEA, J. W., THOMPSON, T. R. E. & WRIGHT, P. S. (1984). Soils and their Use in Wales. Bulletin No. 11. Harpenden: Soil Survey of England and Wales. RYDEN, J. C., BALL, P. R.& GARWOOD, E. A. (1984). Nitrate leaching from grassland. Nature (London) 311, SCHOLEFIELD, D.& TYSON, K. C. (1992). Comparing the levels of nitrate leaching from grass clover and N-fertilized grass swards grazed with beef cattle. In Proceedings of the 14th General Meeting of the European Grassland Federation, Lahti, Finland 1992, pp SCHOLEFIELD, D., TYSON, K. C., GARWOOD, E. A., ARMSTRONG, A. C., HAWKINS, J.& STONE, A. C. (1993). Nitrate leaching from grazed grassland lysimeters: effects of fertilizer input, field drainage, age of sward and patterns of weather. Journal of Soil Science 44, SHERWOOD, M.& FANNING, A. (1989). Leaching of nitrate from simulated urine patches. In Management Systems to Reduce Impact of Nitrates (Ed. J. C. Germon), pp Barking, Essex: Elsevier Science Publishers. SICHEL, H. S. (1952). New methods in the statistical analysis of mine sampling data. Transactions of the London Institute of Mining and Metallurgy 61, SWIFT, G., MORRISON, M. W., CLELAND, A. T., SMITH- TAYLOR, C. A. B.& DICKSON, J. M. (1992). Comparison of white clover varieties under cutting and grazing. Grass and Forage Science 47, THOMAS, R. J., LOGAN, K. A. B., IRONSIDE, A. D.& BOLTON,