Nitrogen losses from grass ley after slurry application – surface broadcasting vs . injection

While many small dairy farms have shut down milk production, the livestock density and number of livestock farms have increased in certain regions in western and central Finland during recent decades. At present, most dairy farms prefer almost continuous grass cultivation to crop rotation with cereals and grasses. Consequently, slurry is spread onto fields of silage grass instead of using earlier methods where slurry was applied to cereal fields before autumn ploughing or before spring tillage. Due to the soil wetness and risk of soil compacNitrogen losses from grass ley after slurry application – surface broadcasting vs. injection


Introduction
While many small dairy farms have shut down milk production, the livestock density and number of livestock farms have increased in certain regions in western and central Finland during recent dec-ades.At present, most dairy farms prefer almost continuous grass cultivation to crop rotation with cereals and grasses.Consequently, slurry is spread onto fields of silage grass instead of using earlier methods where slurry was applied to cereal fields before autumn ploughing or before spring tillage.Due to the soil wetness and risk of soil compac-tion in spring, however, only mineral fertilizer is often surface applied to grass in the beginning of the growing season whereas cattle slurry is applied after the first cut.If the growing season is extremely rainy, it may not be possible to spread slurry with heavy machinery on wet soils in summer.Then the slurry tanks are emptied in the autumn to provide storage capacity for the winter months.
In Finland, slurry is applied to grassland surface either by conventional broadcasting or by more recently adopted band spreading and trailing shoe techniques, whereas injection is used to apply slurry below the soil surface.Surface application is an easy and cheap process but it leaves the manure prone to NH 3 volatilization (Braschkat et al. 1997, Mattila andJoki-Tokola 2003) and surface runoff (Turtola and Kemppainen 1998).Injection of slurry might be better environmentally but it is more expensive and more difficult than broadcast application.Top-dress fertilization of grass fields with mineral fertilizers is also a typical complementary method.
The purpose of this study was to compare two different slurry application methods -surface broadcasting and injection -on grass fields.The former method is a cheap and commonly used practice on most dairy farms whereas the latter is considered as a difficult method to use, particularly on stony soils.In this study, we investigated whether slurry injection could be recommended in given environmental conditions in boreal climates.Losses of total nitrogen (TN), ammonium N (NH 4 + -N, hereafter NH 4 -N) and nitrate N (NO 3 --N, hereafter NO 3 -N) to surface runoff water from the surfaceapplied slurry were compared to losses from injected slurry or mineral fertilization on a grass field.Knowledge of ammonia (NH 3 ) losses to air due to the methods in cool autumn weather was also lacking.Nitrogen uptake by grass was measured for N balances.The amounts of soil mineral N (SMN; NH 4 -N plus NO 3 -N) at different depths were also measured and N balances were calculated to allow an estimation of the risk for NO 3 leaching.

The experimental field
The study was performed on an eight-plot experimental field (0.34 ha; Uusi-Kämppä and Heinonen-Tanski 2008) located in Jokioinen, south west Finland (60°49'N 23°30'E).The area had a long-term  mean annual precipitation of 607 mm and mean annual temperature of 4.3 °C, with the mean temperatures of the coldest (February) and the warmest (July) months being -6.5 and 16.1 o C, respectively (Drebs et al. 2002).The soil was classified as Typic Cryaquept (Soil Survey Staff 1996) containing 61% clay in the plough layer.The concentrations of Ca, K, Mg and P in the plough layer were at a satisfactory or good level.
The experimental plots with slopes of 0.9-1.7% were isolated from each other by plastic film to a depth of 0.6 m and by soil banks.Uncultivated 10-m wide buffer zones were established at the lower edge of the plots since buffers (mostly 3-m or 15-m wide) are typical on Finnish fields.Tenmetre wide buffer area in the upper edges of the plots of total length of 70 m, and 0.5-m (1998)(1999)(2000) or 1. 5-m (1996-1997) wide borders on both sides of the plots were also untreated, with neither soil nor plant sampling, nor slurry application, due to difficulties to drive and work with tractors and spreaders on those areas on the narrow plots.The grass ley on the experimental field consisted for the most part of timothy (Phleum pratense) and meadow fescue (Festuca pratensis) sown in June 1995.The grass ley was cut twice a year, with the first cut always in June and the second cut in late August (1996), September (1997, 1998) or early October (2000).

Treatments and applications
The experimental treatments were as follows: 1. Surface broadcasting (SB) of cattle slurry onto the grass ley (three replicates); 2. Injection (IN) of cattle slurry into the grass ley (depth of 0.05-0.1 m; three replicates); and 3. Mineral fertilization (MF) -top-dress fertilization onto the grass ley (two replicates).
Slurry was applied annually to grass ley after the first cut in 1996-1997 (Phase I) and biannually after the first and the second cuts in 1998-2000 (Phase II).In Phase I, the application rates of mineral N (160 kg ha -1 yr -1 including NH 4 -N of slurry and NH 4 -N and NO 3 -N of mineral fertilizer) and total P (36 kg ha -1 yr -1 ) represented the nutrient amounts allowed by Finnish 'good agricultural practice' and the average used on most Finnish farms.In Phase II, the corresponding amounts were 230 and 66 kg ha -1 yr -1 for mineral N and total P, respectively.In autumn, slurry amounts of 33-42 t ha -1 were applied, although the maximum allowed autumn slurry amount at that time was 30 t ha -1 (Finlex 1998).In fact, 120-160 kg ha -1 more TN in slurry was applied than allowed in the nitrate directive (170 kg TN ha -1 yr -1 ) to detect possible environmental risks due to over-dosing of manure.More details about P applications in slurry and fertilizers, slurry properties, and storage tanks have been presented by Uusi-Kämppä and Heinonen-Tanski (2008).
In Phase I (annual slurry application in June, 1996June, -1997)), cattle slurry (34-61 t ha -1 , Table 1) was applied to an area of 3 m × 50 m by a "Vogelsang" spreader on slurry plots after the first grass cut in June.Slurry was either applied to the soil surface with a band spreading unit equipped with a small splash plate under each hose, or injected with an injector that had 10 tines with 0.3 m spacing, each equipped with a disc coulter and a press wheel (Kapuinen 1998).During Phase II (biannual slurry application in June andOctober, 1998-2000) slurry was applied by a "Teho-Lotina" spreader that had an injector with 0.47 m tine spacing and disc coulters but no press wheels.Broadcast spreading was carried out by holding the injector up while each tine was equipped with a small splash plate.The slurry amounts were slightly higher in the IN than SB plots due to a lower driving speed during injection.In autumn 2000, the field was ploughed three days after the slurry application, when the ammonia volatilization measurements had been finished.
Mineral fertilizer was spread by a "Juko" fertilizer drill to all plots in spring and to MF plots after the first cut in June.In spring 1996, NK fertilizer (20% N and 15% K; Table 1) was surface applied to all plots.Since then ammonium nitrate fertilizer (26% N) was spread in spring, except in spring 1999 and spring 2000 when NPK fertilizer (20% N, 4% P and 7% K) was spread on the MF plots only.Due to half of the NH 4 -N in slurry spread in autumn was assumed to be available for plants in the following spring (Ministry of Agriculture and Forestry 1998), 39 and 31 kg ha -1 less fertilizer N was applied to slurry plots than to MF plots in spring 1999 and in spring 2000, respectively.On the MF plots, NPK fertilizer (20% N, 4% P and 7% K) was surface applied every summer, except in the first summer, when NPK fertilizer (18% N, 5% P, 10% K) was spread.

Measurement of ammonia volatilization
Volatilization of NH 3 was measured after the autumn applications of slurry (SB and IN) in 1999 and 2000 by the equilibrium concentration technique, also called the "JTI method" (Svensson 1994).The method uses passive diffusional NH 3 samplers that are placed on treated areas both in ambient air and under ventilated chambers.The ammonia volatilization rate in ambient air is calculated from the amounts of NH 3 absorbed by the samplers.Air temperature is used to calculate the diffusion coefficient of NH 3 .The concentration of NH 3 inside the chambers was used as a measure of NH 3 volatilization potential without the effect of varying wind conditions in ambient air.
Ammonia volatilization was measured in the daytime starting at 5-15 min after the application of slurry and lasting for 2.75-4 h divided into two consecutive periods.On the following two days, NH 3 measurement began about 24 h and about 48 h after the slurry application and lasted for 3.5-5 h on each of the days.The measurement was carried out in the three replicate plots of both SB and IN.Two chambers and two ambient air sampler holders were placed on each plot.Air temperature was measured with a thermohygrograph at about 0.2 m height and wind speed was measured with a cup anemometer at 2 m height.The volatilization of NH 3 between measurement periods was interpolated by calculating the average emission values before and after an interval and correcting it based on the temperature and wind speed that prevailed during the interval.The procedure is described in detail by Malgeryd (1996).

Water sampling and analyses
Surface and near-surface runoff (referred to hereafter as surface runoff) to a depth of 0.3 m was collected in a modified collector trench planned by Puustinen (1994) at the lower end of each plot and fed by pipes into 8 plastic tanks (2.0 m 3 ) buried in the soil.Water volume was measured by flow meters (Oy Tekno-Monta Ab, JOT-company, 1992) and representative subsamples were taken through samplers (Fig. 1) for laboratory analyses when the tanks were emptied.Water was sampled 16-27 times per year, with most samplings in spring and autumn.The time interval between water samplings in peak runoff periods varied from a day to two weeks, depending on rains and snowmelts.
The volume of runoff water was calculated from the whole plot area, whereas the N losses were calculated from the slurry applied source area.On the border areas, the mean TN losses through surface runoff were estimated to be negligible (ca 0.5 kg ha -1 yr -1 ) according to the TN concentrations of surface runoff water measured earlier on nearby plots under unfertilized grass (Turtola and Paajanen 1995).In spring 1997, the surface runoff results from one SB plot and one MF plot were omitted, due to freezing of the outlet pipes.Precipitation was measured at Jokioinen Observatory, Finnish Meterological Institute, situated 0.5 km from the field.
Water samples were stored in polyethylene bottles for periods from a few days to a few weeks in the dark (4°C) before determining the nutrient concentrations.The storage time probably did not have a large impact on the concentrations of TN and NO 3 -N, but the concentrations of NH 4 -N may have decreased during the prolonged storage (Turtola 1989).For the determinations of NH 4 -N and NO 3 -N, the samples were filtered through a membrane filter (0.2 µm) and analysed with a Skalar autoanalyser according to Finnish standard methods (SFS 3030, SFS 3032).The concentration of TN was determined from unfiltered water samples by oxidation of N compounds to NO 3 in alkaline solution (SFS 3031).

Soil sampling and analyses
Because the drainage water was not measured, both the amounts of NH 4 -N and NO 3 -N as well as their sum (SMN) in the 0-60 cm soil layers were used to indicate the risk of N leaching from the grass ley.Soil samples were taken separately from each plot in spring and autumn before the application of mineral fertilizer or slurry (Uusi-Kämppä and Heinonen-Tanski 2008).The samples taken in spring 1997 were omitted because the field had been fertilized a few days earlier.
Soil samples were frozen immediately after the sampling.For NH 4 -N and NO 3 -N analyses, soils were thawed overnight (4°C), and 40 ml of moist soil was subsequently extracted with 100 ml of 2 M KCl for 16 hours (Sippola and Yläranta 1985).After filtration, concentrations of NH 4 -N and NO 3 -N were measured with a Skalar autoanalyser.The concentrations of TN and carbon (C) were determined using the C-N-autoanalyser (LecoCN-2000, Leco Corporation, St.Joseph, MI, USA).

Other samplings and calculations of nitrogen balances
Slurry samples were taken during spreading and analysed for concentrations of TN (Kjeldahl) and NH 4 -N as described by Mattila and Joki-Tokola (2003).Above-ground biomass was sampled before harvesting the grass.Samples (0.64 m 2 ) were collected from each plot so that the grass was cut leaving a stubble of 1 cm.Plant samples were dried at 60°C overnight for TN analysis with a LECO analyser and at 105°C for dry matter (DM) determination.
Field N balance was estimated as the difference between N inputs and outputs (Equation 1).The N uptake of grass, ammonia volatilization and TN in surface runoff were considered as outputs in the calculations.Ammonia volatilization from summer-applied slurry was estimated to be 40% of the applied NH 4 -N for surface application and 0.4% for injection, based on the results of Mattila and Joki-Tokola (2003).Volatilization from autumnapplied slurry was taken from the results of the NH 3 measurements carried out in this study.Ammonia volatilization from mineral fertilizer, in turn, was estimated to be 1.6% of the applied N (Grönroos et al. 2009).Equation 1:

Statistical analyses
Amounts of grass yield and biomass N in grass as well as the amounts of NO 3 -N, NH 4 -N and SMN in the soil (0-0.6 m) were analysed statistically using a mixed model, where treatment, sampling date and their interactions were used as fixed effects while block, block x treatment and block x sampling date were used as random effects.The soil data were log-transformed before analysis because of skewed distributions.
For statistical analyses of the surface runoff results, log-transformation was used for the TN and NH 4 -N values.The data from the two study phases were analysed together using a mixed model whereby study, treatment, and their interactions were used as fixed effects, whereas block, block × treatment, and block × study were used as random effects.Each block included two or three adjacent plots with different treatments.Soil, plant and runoff analyses were performed using an SAS/MIXED procedure.
The effect of application technique on NH 3 concentration in chambers was studied for each measurement period with analysis of variance according to a randomized complete block arrangement (Steel and Torrie 1981) with three replications.The effect was considered significant with p values <0.05.The analysis was carried out with the GLM procedure of SAS statistical software version 6.12.

Dry Matter and Nitrogen Uptake of Grass
In Phase I, the mean DM grass yields (8.0-9.3 t ha -1 yr -1 ) and N uptakes (160-200 kg ha -1 yr -1 ) were higher in the MF plots than in the slurry treated plots (Table 2).There were no significant differences in the DM yields or N uptakes between treatments in the first cuts, probably since all the treatments received the same amount of fertilizer N in spring.
In contrast, in the second cuts, the DM yields and N uptakes were statistically (p < 0.05) lower in the slurry treated plots than in the MF plots, although the same amount of mineral N (ca 80 kg ha -1 ) was spread in all treatments.
In Phase II (biannual slurry application), the mean DM yields (5.5-7.0 t ha -1 yr -1 ) and N uptakes (90-125 kg ha -1 yr -1 ) were lower than in Phase I, although the mean applications of mineral N were 44, 53 and 20 kg ha -1 yr -1 higher on the SB, IN and MF plots, respectively, than in Phase I.This time, however, three-fourths of the applied mineral N originated from cattle slurry on the SB and IN plots, whereas in Phase I, half of the mineral N was from slurry and half from mineral fertilizer.As in Phase I, there were no statistical differences in uptake between the treatments in the first cuts, but in the second cuts, the N uptake was statistically higher (p < 0.05) in the MF and IN plots than in the SB plots in 1998-1999.

Ammonia volatilization
The NH 3 volatilization from SB was considerable, which is indicated both by NH 3 volatilization rates in the ambient air and by NH 3 concentrations in the chambers (Table 3).Over IN, chamber concentrations of NH 3 were low and the volatilization rates in ambient air were close to zero and often slightly negative, which may indicate deposition of NH 3 that drifted from SB.Despite this disturbance, it can be concluded that the volatilization of NH 3 from injected slurry was small compared with broadcast slurry.To obtain undisturbed results, slurry injection and the subsequent NH 3 measurement should have been carried out before broadcasting.However, different timing of the applications would have compromised the comparison of SB and IN by making a difference in the weather conditions at application and during a few days thereafter.
There are also earlier studies showing that injection of slurry into soil effectively prevents NH 3 volatilization (e.g.Frost 1994, Dosch andGutser 1996).Most of the previous work has been done with mass balance or wind tunnel techniques, but the JTI method used in this study has proven to give results comparable with other methods (Misselbrook et al. 2005b).Mattila and Joki-Tokola (2003) used the same JTI equipment as in this study, and measured negligible NH 3 volatilization from cattle slurry injected to ley in summer, and a 40% loss, on average, of NH 4 -N from broadcast slurry.The results reported here indicate that surface application in autumn may cause high losses despite the lower temperature.In cooler weather, the volatilization rate is lower, but total losses may still be considerable, as also observed by Sommer et al. (1991).The effect of temperature on NH 3 volatilization is interconnected with many other factors such as solar radiation, air humidity, soil moisture content and drying of manure after application.Temperature as such has not always proven an important factor in determining NH 3 volatilization from applied manure (e.g.Braschkat et al. 1997, Sommer and Olesen 2000, Misselbrook et al. 2005a).
On the SB plots, ammonia volatilization was the largest measured single N flow into the environment (15% of TN application and 24% of the mineral N).The NH 3 volatilization was highest on the application day and decreased rapidly during the following two days (Table 3).The decrease is assumed to result from a rainfall and a reduction in the concentration of NH 4 -N in the slurry although not measured after the application.Ammonia volatilization was higher in 2000 than in  53) 125 (70) † The grass was not harvested.
Different letters in the same row indicate a significant difference between treatments (p < 0.05).SB = surface broadcasting of slurry, IN = slurry injection, MF = mineral fertilization 1999, which may have resulted at least partly from higher ambient temperatures (Table 3) and a higher DM content of the slurry: 6.9 and 8.3% in 1999 and 2000, respectively (Uusi-Kämppä and Heinonen-Tanski 2008).

Amount of surface runoff
The mean annual precipitation during the experiment was 626 mm (586-673 mm) which is near the longterm  average of 607 mm.On our field, the surface runoff was 10-20% of the precipitation (Table 4).The mean annual surface runoff (110 mm) in the Phase II was comparable to the surface runoff (110 mm) on a nearby clay soil under timothy and red clover in September 1992-August 1993 (Uusitalo et al. 2007).In Phase I, surface runoff (64 mm) was only half of that was measured in Phase II.On a coarse-textured pasture soil, Saarijärvi et al. (2007) measured surface runoff of 66-107 mm which was around 40% of the total runoff and 15% of the average precipitation in Eastern Finland.The measured volumes of surface runoff on our field agreed quite well with these findings, indicating that there has been deep percolation (drainflow) as well.However, if the drainage system does not function well or there is no drainage system, the surface runoff can be multifold compared to volumes of drainflow from well-drained grass fields (Turtola andPaajanen 1995, Bilotta et al. 2008).

Nitrogen losses in surface runoff
Owing to the relatively small amounts of fertilizer and slurry N in Phase I and lack of heavy rainfall after the slurry applications in summer, losses of TN, NH 4 -N and NO 3 -N in surface runoff were negligible from all treatments over the 18-month monitoring period (Table 4).In fact, the volumes of surface Superscripts denote statistically significant differences.Volatilization from injected slurry is excluded, because it was close to zero and may have been affected by NH 3 drifting from broadcast slurry.Nitrogen loss values include measured emissions from all the four periods and estimated emissions during their intervals.Precipitation between the end of a measurement period and the start of the next period is in parenthesis.SB = surface broadcasting of slurry, IN = slurry injection, MF = mineral fertilization.
runoff and N losses were higher before slurry applications during the snow melting in spring 1996.In Phase II, over the 36-month monitoring period the cumulative losses of NH 4 -N and TN, 5.2 kg ha -1 and 20 kg ha -1 , respectively, were still relatively small in surface runoff from the plots with slurry broadcasting (Table 4).The TN losses were small, although the slurry TN rates exceeded the currently allowed maximum amount of 170 kg ha -1 yr -1 .
Injection further reduced the originally small surface runoff losses of NH 4 -N and TN by 83% (p < 0.001) and 34% (p < 0.01), respectively, compared with surface broadcasting, although a little more slurry TN (20-30 kg ha -1 yr -1 ) was spread on the IN plots.On a fine sandy soil, Turtola and Kemppainen (1998) measured great annual N losses in surface runoff from grass with autumn broadcast slurry, 16-36 kg ha -1 yr -1 and 7.7-22 kg ha -1 yr -1 for TN and NH 4 -N, respectively.In their study, however, the amount of TN applied in autumn was one-third higher and the volumes of surface runoff were three times greater than in Phase II of our study.In Norway, Uhlen (1978) reported that surface runoff losses of TN and NH 4 -N were 8 and 4 kg ha -1 , respectively, during the next 14 months after autumn application of 60 t ha -1 semi-liquid cow manure (228 kg TN ha -1 ) to grass.On boreal pastures, too, the annual losses of TN in surface runoff were small (below 5 kg ha -1 ) in the study of Saarijärvi (2008), although the pastures often receive more N than silage grasses.
However, after slurry application (140-155 kg TN ha -1 ) to wet soil on October 16, 1998, followed with heavy rainfall (60 mm) and surface runoff (10 mm) during the next two weeks, the mean losses of TN and NH 4 -N in surface runoff from the SB plots were 9.3 and 3.5 kg ha -1 over 2.5 months, respectively (Fig. 2), being 47% of TN and 67% of NH 4 -N losses over the whole 3-year study phase.During three days after slurry application, incidental TN losses were highest, at 6.8, 0.5 and 0.1 kg ha -1 from the SB, IN and MF plots, respectively.Soon after slurry application, the mean TN concentration in surface runoff water was 92 mg l -1 for SB, but less for IN (7.6 mg l -1 ) and MF (1.2 mg l -1 ; Fig. 3).Since concentrations of NO 3 -N and NH 4 -N from SB plots were < 0.1 mg l -1 and < 51.1 mg l -1 , respectively, a large part of TN was in organic form.In June 1998, surface runoff (4 mm) from the grass stubble was also high with high rainfall (99 mm) but N losses .96.9 5.2 0.9 0.7 1.1 2.2 0.8 † n = number of samplings SB = surface broadcasting of slurry, IN = slurry injection, MF = mineral fertilization were negligible because slurry had not yet been applied (Fig. 2).
While the N losses were small on our experimental field with a slope of 0.9-1.7%,losses may be higher on steep slopes with heavy rainfalls soon after slurry application.The surface application of manure is no longer allowed on fields with an average slope of over 10% (Finlex 2000).Heathwaite et al. (1998) have also shown that the 10-m untreated buffer zone below the source area applied with cattle slurry reduced the TN load by 75% in surface runoff.Thus in our study, the 10-m buffer zone probably decreased nitrogen losses from all treatments.At present, nitrogen losses from slurry applied fields are mitigated, since the application of N fertilizers (including slurry N) is not allowed on areas closer than 5 metres to a watercourse.And along the width of the next five metres, surface application of N fertilizers is prohibited if the field slope exceeds two per cent (Finlex 2000).Even wider unmanured areas would be needed on field edges with steep slopes along lakes and rivers to decrease direct N losses in surface runoff from source fields to water.
The cumulative load of NO 3 -N in surface runoff was small in all treatments (0.8-2.2 kg ha -1 ), being highest in the IN plots over the 3-year study phase.The small NO 3 -N losses in surface runoff from grass are consistent with results from studies of Uhlen (1978), Turtola and Kemppainen (1998), Ridley et al. (2001), Smith et al (2001), andSaarijärvi (2008).Ploughing of grass soil in October 2000 increased slightly losses of NO 3 -N and TN in surface runoff but decreased NH 4 -N losses (Table 4).

Soil mineral nitrogen and nitrogen leaching
Although in Phase II slurry was spread to the grass ley in autumn, the SMN amounts measured in the following spring were only slightly higher (0-30 kg ha -1 ) or even lower (4-6 kg ha -1 ) than the amounts measured in autumn before the slurry applications.This demonstrates that the slurry N added in the autumn (105-155 kg TN; 60-80 kg NH 4 -N) might have volatilized, become converted to organic form in the soil or leached.In the IN plots, however, the SMN amounts in spring were significantly higher (p = 0.03) than in the SB plots, probably due to lower NH 3 volatilization and slightly higher N input.Also the NO 3 -N amounts were 6-7 kg ha -1 higher in the IN plots compared to SB plots in May 1999 and in October 1999 (p < 0.001).Cameron et al. (1996) observed that NO 3 -N leaching was consistently higher after subsurface injection of dairy pond sludge compared to surface application.According these results slurry injection may thus increase N leaching from grass fields.The summer season 1999 was fairly warm and dry and therefore only one grass yield could be harvested (Table 2).Hooda et al. (1998) and Scholefield et al. (1993) have reported that NO 3 -N leaching is higher after a dry and warm summer than after a wet and cool summer season, since in dry conditions nitrification may be high whereas denitrification and plant uptake of N can be lower than during cool and wet years.In October 1999 and April 2000, the NO 3 -N amounts in soil were 3-7 kg ha -1 higher than measured at other times in this study (Fig. 4) and, thus, there was a slightly higher risk for NO 3 -N leaching from the grassland.Immediately after ploughing in October 2000, the amount of was greater in the plots where slurry had been previously injected (p < 0.01) than in those in which slurry had been broadcast (Fig. 4).In the SB plots, part of the slurry N was volatilized as NH 3 and therefore also the NH 4 -N amounts in soil were smaller than in IN plots (p < 0.01).
The experiment continued for 5 years and before that the field had not received manure for years, which increased the capacity of the soil to retain excess slurry N. In this respect, the situation is often different on animal farms, where the same grass fields have been manured for decades.Moreover clay soil has a higher capacity for retaining NH 4 -N than coarse textured soils.Since most Finnish cattle farms are situated on areas with coarse textured soils, the risk for higher N leaching losses to water is more likely than in our study.

Nitrogen balance and fate of nitrogen
During the five study years, the cumulative field TN surpluses were 687, 971 and 65 kg ha -1 in the SB, IN and MF plots, respectively (Fig. 5).In Phase I, TN balances were negative on the MF plots.In Phase II, the amount of non-recovered N was extremely high, up to 58% (ca 210 kg ha -1 yr -1 ) and 72% (ca 280 kg ha -1 yr -1 ) of the TN input on the SB and IN plots, respectively (Fig. 5).According to Macdonald and Jones (2003), 20-70% of the N inputs to agricultural systems may be unaccounted for.Although denitrification was not measured it is obvious that large part of organic N applied in slurry was not mineralized and thus it was not recognized as SMN.In Canada, Bittman et al. (2007) estimated that ca 30% of applied manure-N was stored in soil organic matter.A significant amount of NH 4 -N in slurry might also have been microbially immobilized soon after application due to decomposition of fatty acids in slurry (Kirchmann andLundvall 1993, Sørensen andAmato 2002).According to the results of Huss-Danell and Chaia (2007), over 30 kg N ha -1 can be incorporated into grass roots in the northern part of Sweden.Pierzynski and Gehl (2005) showed that some of the N saved from NH 3 emissions may have been lost as N 2 O from slurry injected fields.In a Finnish study, however, only ca 0.7% of cattle slurry N incorporated with a disc was lost as N 2 O fluxes (Syväsalo et al. 2006, Perälä et al. 2006).Ammonium can also be fixed into clay minerals or nitrate can be leached into subsurface drains or ground water.

Conclusions
Despite cool autumn weather, a considerable portion (20-33%) of the surface-applied slurry NH 4 -N was lost through ammonia volatilization within a few days after application, but the injection of slurry into the soil effectively prevented this.Nitrogen losses in surface runoff from grass field applied with slurry were small during the five study years, except when heavy rainfall occurred after slurry application in autumn.Although high slurry N amounts were added to grass, nitrogen leaching risk was surprisingly small from clay soil.If over-dosing of manure would continue longer, however, the situation could be different.When moderate slurry amounts (as in Phase I) are applied in summer and by a technique with low NH 3 emissions most of the N is kept within the nutrient cycle of the farm.These study results can be directly applied to clay soils, whereas on coarse textured soils, the leaching losses may be higher than in this study.

Fig. 1 .
Fig.1.Sampling of surface runoff: 1, water from the collector tank flows through the flow meter; 2, the water sample drips into a pail; 3, the rest of the water flows through an outlet.

Fig. 2 .
Fig. 2. Cumulative losses of total nitrogen in surface runoff and periodic precipitation from summer 1996 to autumn 2000.Slurry applications are marked by arrows.(Au, autumn; Sp, spring; Su, summer)

Fig. 4 .
Fig. 4. Amounts of soil mineral nitrogen (±S.D.) at 0-60 cm during 1996-2000.The number of weeks passed between previous slurry application and soil sampling is shown inside of the bars.

Fig. 5 .
Fig. 5. Input of mineral N and non-mineral N and output of N (plant uptake, NH 3 volatilization and TN in surface runoff) during study Phases I and II.

Table 1 .
Application dates, amended plot area, amount of slurry, and total nitrogen (TN) applications in slurry (s) and mineral fertilizer (mf) in plots where slurry was surface broadcast (SB) or injected (IN) into soil and in mineral fertilized (MF) plots.Values in parenthesis indicate the application rate of mineral N in slurry and mineral fertilizer.

Table 2 .
Over-ground grass dry matter yields and biomass N. Percentage of biomass N from the previous mineral N application is given in parenthesis.

Table 3 .
Concentration of NH 3 in chambers on surface broadcasting (SB) and injection (IN) plots, NH 3 volatilization in ambient air and weather conditions during the measurement periods.

Table 4 .
Precipitation and means of surface runoff and losses of total nitrogen, ammonium nitrogen and nitrate nitrogen to surface runoff water.