Introduction

Organic manures cause residual nitrogen (N) effects after the year of their application (Whitmore and Schröder 1996; Sörensen and Amato 2002; Sörensen 2004; Sörensen and Thomsen 2005; Daudén et al. 2005; Schröder et al. 2005a). However, the N fertilizer replacement values (NFRV) of manures communicated to farmers, are commonly based on N becoming available during the first growing season only (Schröder and Stevens 2004; Gutser et al. 2005; Reijs et al. 2007). If N provided through earlier applications is not accounted for in one way or another, manured crops may thus receive too much N. A greater awareness of the residual N effect is required now that cropping systems in the developed world must be redirected towards situations closer to N limitation for the benefit of the environment (Addiscott 2005). However, farmers point at the reverse side of this coin: when N recommendations are based on response trials executed on fields that were amply manured in the recent past, yield penalties resulting from withheld N may be underestimated (Motavalli et al. 1992; Schröder 2005). A better insight in these residual effects is hence also needed from this perspective.

Soil based indicators giving growers reliable estimates of the mineralizable N potential, are lacking so far (Powlson 1997; Curtin and McCallum 2004). Instead, models predicting available N as a function of previous organic N inputs may be helpful. An example of such a model was presented by Schröder et al. (2005a). Their model was calibrated with data from a medium term experiment in which cattle slurry was annually applied to silage maize. The study indicated that the relative decomposition rate (RDR) of the organic N in slurry amounted to 0.25–0.33 year−1. The parameter setting of their model did not need any adjustments to achieve a satisfactory match between observed and simulated N yields of maize in an independent data set (Schröder et al. 2006). However, lack of regular soil perturbation, may render lower RDR’s for manures applied to permanent grassland (Silgram and Shepherd 1997). Moreover, RDR’s are affected by the nature of the organic compounds in manure as well (Chae and Tabatabai 1986; Chadwick et al. 2000). Both aspects justify additional research.

The NFRV of manures is not just determined by the RDR of the organic N fraction but also by the fate of the mineral N fraction which mainly consists of ammonium N. This form of N can be easily lost due to volatilization as ammonia (NH3). To prevent this loss, slurry injection has become compulsory on grasslands in The Netherlands (Huijsmans 1999). Some farmers question this obligation, arguing that the benefits of injection in terms of a reduced NH3 loss are countered by the detrimental effects of injection on soil and sod quality and thus on N utilization. Their views are not supported by recent research (Schils and Kok 2003), but there is an ongoing demand for additional evidence (Sonneveld et al. 2007).

This paper reports results of a field experiment on cut grassland directed at the assessment of the NFRV of various types of manures in the year of application and in subsequent years. The paper also includes testing of a simple numerical model allowing exploration of the long term consequences of manure applications on available N and yield.

Materials and methods

Field experiment

In the spring of 2002 an experiment on the N fertilizer replacement values (NFRV) of manures was started on a free draining slightly loamy, moderately fine sandy soil (FAO classification ‘cambic podzol’) with an organic matter content of approximately 4% and a pH of 5.5 (0–20 cm) at the Experimental Dairy Farm ‘De Marke’ in Hengelo, The Netherlands (52°03″N, 6°18″E). Additional soil fertility indicators are presented in Table 1. The experiment was established in permanent (since 1994) grassland with a sward dominated (>80%) by Lolium perenne L. Dicots including white clover were chemically controlled for an unbiased assessment of the NFRV (Schröder 2005). The treatments installed from 2002 comprised a control receiving no N at all (CONT), and treatments receiving 300 kg total N ha−1 year−1 as surface applied calcium ammonium nitrate (CANs), as sod injected anaerobically digested cattle slurry (S1ADi) or sod injected untreated cattle slurry (S1UNi) from the De Marke farm, as sod injected (S2UNi) or surface applied (S2UNs) untreated cattle slurry from a second farm, and as surface applied solid cattle manure (FYMs) from another farm. The ratio of mineral N to total N in these treatments ranged from 100% in CANs, 59% in S1ADi, 49–52% in the other slurries, to 12% in the FYMs (Table 2). Estimated carbon (C) to organic N ratio’s were reasonably similar (range 14.8–15.8) in all manure types, but the C to total N ratio of FYMs was much larger due to the low ammonium N concentration. Application rates of mineral P2O5 (triple super phosphate) and K2O (muriate of K) were adjusted to the different amounts of these elements present in manure to make sure that these elements would not be limiting yields (Table 2). Other nutrients were considered amply available (http://www.bemestingsadvies.nl).

Table 1 Soil fertility indicators at the onset of the experiment in 2002
Full size table
Table 2 Application rate (kg ha−1 year−1) per treatment and some additional characteristics of the used manures
Full size table

Farm yard manure was applied by hand once a year in the second half of February. The mineral fertilizer N and slurry applications were split over three dates (around April 1, second half of May after first grass cut, second half of June after second grass cut) in a 120 to 100 to 80 ratio, based on N contents. Slurries were applied with a precision tanker designed for field trials, equipped with either four splash plates (effective total application width exactly 2.75 m, plate distance 70 cm) or injection tines (effective total application width 2.75 m, tine distance 25 cm, tine depth 5–7 cm).

An annually increasing part of each plot was left unfertilized until no N at all was applied to the grassland in 2006. This provided an increasing number of diverging treatments i.e. plots that had been N fertilized or manured for four (2002–2005), three (2002–2004), two (2002–2003), one (2002) or no year at all. Thus, the share of plots from which N was withheld increased and eventually comprised plots that had not received any N for 1–5 years. This allowed us to discern the residual N effects and first year NFRV’s.

Treatments were embedded in a randomised block design with four replicates. Manure type (including CANs) was the main factor arranged in plots within blocks, the number of years left unfertilized was the superimposed splitting factor within plots. Size of these subplots measured 12 m × 2.75 m.

Grass was cut four to five times a year in the middle of May, second half of June, first week of August, around September 1 and the second half of October. All treatments were harvested on the same day using a Haldrup forage harvester especially designed for trials. The area cut was the inner 10 × 1.5 m of each subplot. After assessing the fresh weight per subplot, the harvest was spread over a wooden panel. From this heap a random sample of approximately 500 g was taken with an auger (content 280 cm3) to determine the dry matter (DM) content. This sample was dried for 16 h at 70°C and homogenized after grinding it in a 1 mm mill. Of this material a 3 g sample was taken to assess the remaining moisture content by drying for 4 h at 103°C. From the remainder a 0.9 g sample was taken to assess the total N contents according to Dumas (Macro N, Foss Heraeus).

The apparent N recovery (ANR) of manure or fertilizer was calculated as the difference of the N yield of a manured (or fertilized) crop and a control crop, and expressed as a percentage of the total N input from manure. The cumulative ANR of a manure applied in year n was calculated as the summed differences (over the years n + m, m being 0 and 1, 0, 1 and 2, or 0, 1, 2 and 3) of the N yields of crops that had been manured (or fertilized) for the last time in year n and crops that had become controls for the first time in year n, expressed as a percentage of the total N input from manure (or fertilizer) in year n. The NFRV of manures was defined as the ANR of manure, expressed as a percentage of the ANR of mineral fertilizer N in the CANs treatment (Schröder 2005). A mineral fertilizer N treatment, needed for the calculation of NFRV’s, was not included in 2006. The impact of residual N effects on the NFRV of manures could hence not be determined in 2006. Consequently, NFRV’s taking account of the first year residual effect were based on the manure batches of 2002, 2003 and 2004, only. Similarly, NFRV’s also taking account of the second year residual effect were based on just the batches of 2002 and 2003, those also taking account of the third year residual effect had to be based on the batch of 2002 only.

Around March 1, prior to fertilizer and manure applications, and immediately after the last cut, soil samples were taken (eight cores of 153 cm3 each per subplot) for determination of soil mineral N (SMN, including NO3 and NH4) in the upper 60 cm soil layer. After homogenizing this sample, 50 ml soil material was added to 100 ml 0.01 M CaCl2, shaken during 1 h at 20°C and subsequently filtered. SMN in this filtrate was determined using a continuous flow analyser (TRAACS 800, Bran and Luebbe).

At the onset of the experiment (spring 2002) the soil was sampled (0–30 cm) to determine total N (Kjeldahl) and organic matter (by loss on ignition) contents by taking 12 cores of 94 cm3 each per block. This observation was repeated after four cropping seasons (spring 2006, eight cores per subplot) in fields that had continuously received one of the manures (or fertilizer-N) or had been control. To estimate the size of the soil N pool, we multiplied the total soil N content by the corresponding bulk density. Bulk density was assumed to depend on the C-content of the soil (Whitmore et al. 1992), adopting a C-concentration in the soil of 59 kg per 100 kg organic matter (Lagreid et al. 1999).

Weather data were collected at the site. Average temperatures during the growing season (March–October) ranged from 12.2°C in 2002 to 14.2°C in 2006. The summed precipitation (including irrigation) in this period ranged from 494 mm in 2002 to 591 mm in 2006 (Table 3).

Table 3 Mean daily air temperature, cumulative precipitation and irrigation from May 1 to October 31
Full size table

Model

We used the data from the experiment to test a simple N model. The basic version of this model has been developed by Wolf et al. (1989) and was tested by Wolf and Van Keulen (1989); Osmond et al. (1992), and Schröder and Van Keulen (1997). Schröder et al. (2005a) extended the model with an additional N pool representing the easily available N compounds such as those found in slurries. This version of the model was successfully tested with an independent data set (Schröder et al. 2006). In the present version of the model we made a more explicit distinction between the organic-N and ammonium-N fraction of manure to allow the model to be used for different types of manures. We also decided to allow for different RDR’s of organic manure-N in the year of application (RDRnorg1) and the years after application (RDRnorg2) to account for initial immobilization (Sörensen 2004) or gradual changes in recalcitrance. In addition to the stable organic soil-N pool (SON) and labile organic soil-N pool (LON) of Wolf et al. (1989), we thus discern a pool of manure N (Norg) of which a fraction RDRnorg1 is mineralized in the year of application (Norg1) after which a fraction RDRnorg2 annually mineralizes from the remaining pool (Norg2) in the years after application (Fig. 1).

Fig. 1
figure 1

Diagram of the model showing the flows of N from fertilizer (Nfert), rain (Nrain), manure (Nm + Norg) and labile (LON) and stable soil organic N pools (SON), to losses (Nloss), to the crop (Ncrop), to LON and SON (RDR’s referring to the relative decomposition rates of the organic N pools involved)

Full size image

In the model, ammonium-N in manure (Nm), mineral fertilizer-N (Nfert), atmospheric deposition (Nrain), and mineralized N from the organic N in manure (Norg) are allocated to losses (Nloss), to the harvestable crop (Ncrop) or to LON, according to transfer coefficients (Table 4). Similarly, N mineralized from LON is allocated to Nloss, Ncrop and SON (Fig. 1). The allocation of fertilizer N to harvestable crop N and LON was based on Schröder et al. (2007), the complement becoming allocated to Nloss. The transfer efficiency of Nm to crops and LON is lower, because 10% (when injected) to 70% (when surface applied) of the ammonium N may volatilize as NH3 upon application (Huijsmans 1999). The utilisation of N liberated from LON and Norg by crops is also lower than that of N in Nfert, as we assume that 5% of the mineralization occurs outside the growing season of grass, becoming allocated to Nloss. Thus we intended to acknowledge that the quality of the grass swards during our commonly wet winters may become insufficient to intercept all the N that may mineralize during brief mild periods. The yearly input via deposition was set at 36 kg N ha−1 (source: National Institute of Public Health and the Environment RIVM, Bilthoven, The Netherlands). Of the N in Nrain 25% is supposedly deposited outside the growing season of grass and allocated to Nloss (Schröder and Van Keulen 1997). The N that is not inherently lost from Nm and Nrain, and from mineralized LON and Norg, as described above, is allocated to Ncrop, LON and Nloss in the same way as Nfert.

Table 4 Transfer coefficients (kg kg−1) of N in atmospheric deposition (Nrain), in mineral fertilizers (Nfert), ammonium-N in manure (Nm), in organic N in manure mineralized in the year of application (Norg1) or the subsequent years (Norg2), in labile soil organic N (LON) and in stable soil organic N (SON), to losses to air and soil below the rooted soil profile (Nloss = sum of volatilization, denitrification and leaching), to harvestable crops (Ncrop), to LON and to SON, used in the simulation of direct and residual N effects of cattle manures and fertilizers applied to cut grassland
Full size table

From Wolf et al. (1989) we adopted that the SON to LON ratio is 3 in an equilibrium situation and that the coefficient for the transfer of N from LON to SON is 0.15. This assumption implies that RDRson is 20 times smaller than that RDRlon (Wolf et al. 1989). Note that this fixed ratio of RDR’s still allows that the sizes of LON and SON do not have a fixed ratio as long as systems are moving to a new equilibrium.

Trends in the organic N pool (see results section), made us doubt to use the observed pool size of 2002 as a starting point of our simulations. Instead we used the average of the observations in 2002 and 2006, according to which LON + SON amounted to 4,570 kg N ha−1 (0–30 cm), after correction for residual N originating from the estimated 196 kg total-N ha−1 applied as undigested cattle slurry in the year before the experiment started.

Testing of the model was restricted to the weighted RDRlon + son (as the two are linked, as indicated before) and the RDR of Norg (RDRnorg1, RDRnorg2). RDRlon+son was calculated according to Wolf et al. (1989), based on the N yields of the long term (2002–2006) control in 2004, 2005 and 2006 whilst correcting for the contribution from Nrain, the surmised SON/LON ratio, and the estimated size of the soil organic N pool. Accordingly, a RDRlon+son of 0.02 year−1 was calculated which is in agreement with Coleman and Jenkinson (1996). In addition to this value we also ran the model with a RDRlon + son of 0.03 year−1, as suggested by the analysis of Schröder et al. (2005a). The model was tested with combinations of RDRnorg1 values (0.00, 0.10, 0.20, 0.33 and 0.50 year−1, the latter reflecting the present composition underlying the recommendation for untreated cattle slurry in The Netherlands [Beijer and Westhoek 1996]) and RDRnorg2 values (0.10, 0.20 and 0.33 year−1), thus trying to find a combination that could satisfactorily simulate the observed N yields. The goodness of fit between simulated and observed data was evaluated with the indices proposed by Addiscott and Whitmore (1987). A good fit between simulated and observed data does not necessarily imply that the model correctly predicts the residual effect of manure whenever a data set includes manure treatments; the good fit could originate from a correct prediction of the first year NFRV rather than of residual N effects. Therefore, we first analysed plots not receiving any N in a given year (just reacting to residual N). We excluded the control plot of 2002 from this analysis because application rates and manure composition in the years preceding our experiment were estimated instead of recorded. Subsequently, we repeated the analysis after extending the data set with treatments that were manured in a given year. Subsequently the model was used to explore the consequences of repeated manure applications on their long term NFRV’s and on the impact of suboptimal fertilizer N applications on yield depressions.

According to the literature (e.g. Hart et al. 1993) organic inputs to the soil may gradually become more and more recalcitrant, eventually resulting in RDR’s close to the RDRlon + son. If our observations over, at most, four residual years would not confirm these diminishing RDR’s, it could thus be incorrect to apply the RDRnorg2, as calculated via the above procedure, to years beyond the 4 years for which the model was actually calibrated. We have hence decided to explore the long-term NFRV’s and yield depressions according to a scenario in which we assumed the RDR to remain constant over time (‘constant RDR’) and a scenario in which we assumed the RDR to diminish (‘diminishing RDR’). In the latter scenario RDRnorg1 > RDRnorg2 with RDRnorg2 gradually dropping in steps of 0.01 year−1 from its estimated value during the first four residual years to our estimated value of RDRlon + son.

Results

Field experiment

Dry matter yields of cut grassland were strongly related to their N yields (Fig. 2) which, in turn, responded positively (P < 0.05) to fertilizer N and manure in all years. The results indicated that manure also had a positive effect on N yields after the year of application, whereas such an effect was not observed where mineral fertilizer N (CANs) had been applied in the previous year (Table 5). This residual N effect of manure was strongest in the first year after application and faded away afterwards.

Fig. 2
figure 2

Observed relationship between the N yield and the dry matter (DM) yield of cut grassland, the solid line reflecting the fitted relationship, the straight dotted lines reflecting set N concentrations in grass dry matter

Full size image
Table 5 Annual nitrogen yield of cut grassland (kg N ha−1), as affected by manure type and manuring history
Full size table

In the spring of 2003 the content of soil mineral N (SMN, 0–60 cm) of FYMs exceeded SMN of CANs by 10 kg N ha−1 (P < 0.05). Differences at all other moments amounted to less than 2 kg N ha−1, on average (range −7 to +6), and were not significant (Fig. 3). SMN of the other manure types (data not shown) fell in between the values of CANs and FYMs. Consequently, residual N effects of manures must have resulted from mineralization within the growing season rather than from carry-over of residual SMN from previous seasons.

Fig. 3
figure 3

Observed soil mineral N (SMN, 0–60 cm layer), as affected by farm yard manure relative to calcium ammonium nitrate applied at annual rates of 300 kg total N ha−1 and the unfertilized control (error bars referring to standard errors of means, n = 4)

Full size image

As the amounts of applied total N slightly deviated from the intended annual rate of 300 kg N ha−1 (apart from CANs treatments, Table 2) and the N yields of controls differed per main plot (Table 5), comparisons of effects of the separate manure types must be made relative to their exact rates, i.e. by comparing the ANR’s. The experimental design allowed us to determine the cumulative ANR over two consecutive years (n, n + 1) for four manure batches (2002, 2003, 2004 and 2005), the cumulative ANR over 3 years (n, n + 1, n + 2) for three batches (2002, 2003 and 2004), and the cumulative ANR over 2 years (n, n + 1) for two batches (2002, 2003). Residual N effects of manures in the fifth year after application were small (Table 5) and ANR’s including this year are not shown separately. Our analysis of N recoveries indicated that the ANR decreased in the order CANs > (S1ADi, S1UNi, S2UNi) > (S2UNs, FYMs), generally showing a positive relationship between the ANR and the mineral N to total N ratio of the manure. Observed ANR of the CANs treatment (Table 6) confirmed the efficient transfer of Nfert to Ncrop assumed by us in Table 4. ANR’s of manure-N increased with time contrary to the ANR of mineral fertilizer-N. This increment was not any larger for FYM and after 4 years ANR had thus not yet reached the level attained by the injected slurries. ANR of injected slurry (S2UNi) exceeded ANR of surface applied slurry (S2UNs).

Table 6 Cumulative apparent nitrogen recovery (ANR, %) from manure in cut grassland after 1 or 2 years (average of the four batches applied in 2002, 2003, 2004 and 2005), after 1, 2 or 3 years (average of the three batches applied in 2002, 2003 and 2004), and after 1, 2, 3 or 4 years (average of the two batches applied in 2002 and 2003) as affected by the manure type
Full size table

The fate of the N not recovered in harvested crops (i.e. approximately (100 − ANR) / 100 × 300 kg N ha−1 year−1) could not be fully elucidated. A fraction of the applied ammonium must inevitably have volatilized from the manures, especially when surface applied (S2UNs, FYMs). Accumulation as SMN was insignificant (Fig. 3), leaving leaching, denitrification and storage in soil organic N as potential output terms. Comparison of total soil N pools (0–30 cm layer) in the springs of 2002 and 2006 pointed at depletion rather than accumulation of soil organic N (Fig. 4). It must be noted that the calculated annual decline rates of 150 kg N ha−1 (CONT, CANs) to 300 kg N ha−1 (FYMs) of this soil pool are unrealistically high. The errors associated with their measurement strongly question the reliability of their absolute values. The relative decline in FYMs was smaller than in CONT and CANs. This suggests that some of the N input of FYMs had been stored in organic matter as yet. The size of the N pools of other treatments (data not shown) fell in between the values of CANs and FYMs.

Fig. 4
figure 4

Size of the N pool in the upper 30 cm soil layer at the onset of the experiment and after 5 years, as affected by four annual applications of fertilizer N (calcium ammonium nitrate, CANs) or farm yard manure (FYMs), relative to the control (error bars referring to standard errors of means, n = 4)

Full size image

The average NFRV of digested slurry amounted to 58% in the first year. Including the residual N effect in the second year (first residual year), increased the NFRV by 3% ( = 62 − 59%) to 61%. Including the residual N effect in the third and fourth year, increased the NFRV by another 2% ( = 64 − 62%) and 4% ( = 71 − 67%), respectively. Corresponding NFRV’s of undigested slurry in the first, the first two, the first three and the first 4 years, averaged 51–53%, 58–61%, 61–63% and 64–67%, respectively (Table 7). The NFRV of surface applied undigested S2UNs was much lower. There were no signs that the low first year NFRV of S2UNs was compensated by increased residual contributions. FYM’s showed the lowest NFRV, their residual contributions to the NFRV in terms of percentages not being any larger than those of slurries. In general the data could not confirm that the gain in NFRV diminished with each next year that was included in the calculation.

Table 7 N fertilizer replacement value (NFRV, kg N per 100 kg N applied) of manures applied to cut grassland after 1 year (average of the four batches applied in 2002, 2003, 2004 and 2005), after 1 or 2 years (average of the three batches applied in 2002, 2003 and 2004), after 1, 2 or 3 years (average of the two batches applied in 2002 and 2003), and after 1, 2, 3 or 4 years (average of the batch applied in 2002)
Full size table

Model

Simulations overestimated the observed N yields of cut grassland responding to residual N by, on average, 23 kg N ha−1 year−1 when a RDRlon + son of 0.03 year−1 was adopted (data not shown), whereas this difference was 6 kg N ha−1 year−1 when the RDR was set at 0.02 year−1. This made us decide to continue our simulations with the latter value.

This analysis indicated that any value of RDRnorg1 between 0.10 and 0.33 combined with a RDRnorg2 of 0.10 year−1, yielded the best compromise of relatively low mean differences between observed and simulated N yields and relatively high correlations between both. Values of 0.00 or 0.50 year−1 for RDRnorg1 and values of 0.20 or 0.33 year−1 for RDRnorg2 made the model perform less. No clear distinction could be found between the two contrasting manure types FYMs and S1ADi (Table 8, other manures not shown separately) which made us decide to pool the data (Table 9).

Table 8 Statistical indices for the goodness of fit between observed and simulated N yields of cut grassland resulting from just the residual N effects (RES) of surface applied farmyard manure (FYMs) and injected anaerobically digested slurry (S1ADi), as related to the relative decomposition rates (RDR) of organic N during the year of application (RDRnorg1, year−1) and during subsequent years (RDRnorg2, year−1)
Full size table
Table 9 Statistical indices for the goodness of fit between observed and simulated N yields of cut grassland resulting from just the residual N effects (RES) and from ‘first year’ and residual N effects together (TOTAL) of manures (the various types being pooled), as related to the relative decomposition rates (RDR) of organic N during the year of application (RDRnorg1, year−1) and during subsequent years (RDRnorg2, year−1)
Full size table

N yields of fertilized grassland were systematically underestimated by the model in the first year (2002), as illustrated with the run in which RDRnorg1 was set at 0.33 and RDRnorg2 at 0.10 year−1 (Fig. 5). This deviation could not be attributed to atypical weather conditions in 2002 (Table 2). When all treatments were pooled, absolute differences between observed and simulated N yields resulting from residual N effects only, were below 20 kg N ha−1 year−1 in approximately 90% of the cases when the value of RDRnorg2 was limited to 0.10 year−1 (Table 9).

Fig. 5
figure 5

Observed N yields of cut grassland versus simulated N yields, adopting a relative decomposition rates (RDR, year−1) of 0.02 for (combined labile and stable) soil organic N in the upper 30 cm layer (LON + SON), 0.33 for the organic N in manure decomposing during the year of application (Norg1) and 0.10 for the remaining organic N (Norg2) decomposing during subsequent years (arrow, indicating fertilized treatments in 2002)

Full size image

Our calculations on the implications of annually repeated manure applications indicated that it would take 7 yearly applications to raise the NFRV of undigested slurry from a simulated initial 55–60% to 75% in the ‘diminishing RDR’ scenario, adopting a RDRnorg1 of 0.33 year−1, a RDRnorg2 of 0.10 year−1 during the first four residual years and a RDRnorg2 gradually dropping to 0.02 year−1 in the subsequent 8 years. In the ‘constant RDR’ scenario, adopting a RDRnorg1 = RDRnorg2 of 0.10 year−1, such an increase would need 10 yearly applications. It would take around 20 (‘constant RDR’) to 40 (‘diminishing RDR’) yearly applications to raise the NFRV of farm yard manure from an simulated initial 20–30% to a similar value. Note that the NFRV of FYM during the first few years of applications was better simulated with a RDRnorg1 of 0.33 year−1 than with 0.10 year−1 (Fig. 6).

Fig. 6
figure 6

Observed and simulated N fertilizer replacement value (kg N 100 kg N−1) of injected cattle slurry (mean composition of S1UN and S2UN) and surface applied farm yard manure (FYM), simulations following a constant decomposition rate of organic manure-N (0.10 year−1) or a diminishing decomposition rate (0.33 year−1 in year of application, 0.10 year−1 in the subsequent 4 years, decreasing by 0.01 per year to a final value of 0.02 year−1 afterwards)

Full size image

Both the observed and the simulated NFRV of anaerobically digested slurry were higher in the first year of application, as a consequence of the conversion of organic N into ammonium N (Table 2). The corresponding lower annual input of organic N affected the residual N effect of digested slurry negatively according to our simulations, which was more or less supported by the observations (Table 7). The observed difference in NFRV of digested and undigested slurry hence declined, the larger the number of years taken into account (Fig. 7).

Fig. 7
figure 7

Observed and simulated N fertilizer replacement value (kg N 100 kg N−1) of injected untreated cattle slurry (S1UNi) and injected anaerobically digested cattle slurry (S1ADi), simulations following a constant decomposition rate of organic manure-N (0.10 year−1)

Full size image

Simulated total N losses could not be confronted with observations because losses were not measured in the experiment. The simulated change of the soil N pool (0–30 cm) amounted to −48, −11, and +63 kg N ha−1 year−1 for CONT, CANs and FYMs treatments, respectively (according to ‘diminishing RDR’ scenario with an adopted RDRnorg1, RDRnorg2 and RDRlon+son = 0.33, 0.10 and 0.02 year−1, respectively). These simulated changes did not agree with the observations, as indicated in the previous section, although the observed differences between treatments were reasonably predicted.

Repeated applications of manures impact on the N supplying power of soils and thus on the N yield of crops from which N is withheld. This is shown by the simulated yield depression of unfertilized grassland that has either been manured for 30 years at a rate of 250 kg undigested slurry-N ha−1 year−1 complemented with mineral fertilizer N up to a level to attain a harvestable yield of 300 kg N ha−1 year−1, or has been fertilized up to this yield level with mineral fertilizer N only (Fig. 8).

Fig. 8
figure 8

Simulated N yield reduction of unfertilized grasslands that have either been manured for 30 years at a rate of 250 kg undigested slurry-N ha−1 complemented with mineral fertilizer N up to a level to attain a harvestable yield of 300 kg N ha−1 year−1, or have been fertilized up to this yield level with mineral fertilizer N only, simulations following a constant decomposition rate of organic manure-N (0.10 year−1) or a diminishing decomposition rate (0.33 year−1 in year of application, 0.10 year−1 in the subsequent 4 years, decreasing by 0.01 per year to a final value of 0.02 year−1 afterwards)

Full size image

Discussion

Nitrogen pollution per unit farm produce is determined my many interacting factors, among which the utilization of applied manure N is one (Schröder et al. 2005b). Incorrect estimates of the short-term and residual N effects of manure could thus have a negative effect on resource utilization. Residual N effects of applied manure do not become fully expressed in experiments lasting just 1 year, either because manured treatments initially lack the N contributions from residual N of prior applications when the site has not been manured at equally high rates, or because control treatments benefit as much as manured treatments from residual N originating from the period before the experiment started. The observed NFRV’s of undigested cattle slurries in the year of application, were in good agreement with short term experiments elsewhere (Schils and Kok 2003; Reijs et al. 2007). The observed NFRV of surface applied slurry was lower than that of injected slurry, indicating that a negative effect of the probable NH3 loss on N yields was apparently not compensated by a positive effect of the avoided mechanical damage resulting from the injection tines. This observation is in agreement with the findings of Schils and Kok (2003).

The establishment of ‘manuring histories’ in our experiment also clearly demonstrated residual N effects of manure. However, due to the limited duration of the experiment we were not able to fully appreciate the long-term consequences. We have tried to circumvent this shortcoming by combining the results with a simple model. This model assumes a first order decay of organic manure and allows for different RDR’s in the year of application and in the following years.

The model did not succeed in predicting the observed dynamics of the soil N pool. Probably, this was due to the impossibility of accurately assessing small changes of the soil N pool against a large background value, as earlier encountered by Schröder et al. (2005a).

The yield response of cut grassland to fertilizer N and manures, however, was reasonably well simulated, bearing in mind that weather effects were not incorporated into the model. Our analysis indicated that the decomposition rate of manures may be larger in the year of application than in subsequent years. Our results suggest that the RDR of organic manure N in the year of application (RDRnorg1) is closer to values around 0.20 than 0.50 year−1. The latter value is as yet the assumption of extension services in The Netherlands, based on the results of hydrolysis of Norg in sulphuric acid (Sluijsmans and Kolenbrander 1976). The residual N effect of manures was better predicted by adopting a RDR in later years (RDRnorg2) of 0.10 than of 0.20 year−1 or more. These RDR values are lower than the 0.25–0.33 year−1 derived from two recent experiments in arable crops on similar soil types (Schröder et al. 2006). Although it must be noted that the present model description differed slightly from that of Schröder et al. (2006), this could imply that the absence of tillage reduces decomposition rates. An analysis of the existing literature on manure decomposition, as presented in Schröder et al. (2005a), indicates that the observed RDR’s of manure range from as little as 0.10 to as much as 0.50 year−1. This broad range does not question the difference in principle between short term and long term NFRV’s of manures; the value of the RDR does, however, affect the number of years needed to attain equilibrium between annual inputs of organic N and the annual decomposition of this material. Differences in RDR are probably attributable to differences in climate and soil type (Hassink 1994; Sörensen and Amato 2002) and the composition of manures as a result of the use of bedding material and diet composition (Chadwick et al. 2000; Kebreab et al. 2001; Reijs et al. 2007).

Sörensen and Amato (2002) presented data from which residual N effects of pig slurry of 5 and 2% of the total N input could be calculated for the first and second year after the year of application, respectively, according to the difference method as used in the present paper. An isotope substitution study of Sörensen and Thomsen (2005) pointed at a residual N effect of pig slurry of 3% of the total N input in the first year following the year of application. Sörensen (2004) reported residual effects of cattle slurry on the N uptake of barley followed by a cover crop. The residual effect on N yields amounted to 5–7% and 5% of the total N input in the first and second year, respectively, when effects on barley and a following cover crop were pooled to approximate the length of our grassland growing season. These estimates may be conservative as their test crops received the recommended mineral N rate, making them less responsive to residual N. The previous values from the literature are in good agreement with our residual effects of undigested cattle slurries of 1–10% (first residual year), 2–5% (second residual year), and 1–4% (third residual year) of the total N input (Table 7), although our slurries contained more Norg per kg total N. When related to the input of organic N, the values observed by Sörensen (2004) corresponded with an apparent decay of 0.11–0.16 year−1 of the applied amount of organic N in the slurry. Again, this is in reasonable agreement with our 0.10 year−1.

Our experiments and the subsequent modelling both indicated that manures with a relatively large first year NFRV (e.g. anaerobically digested slurry) exhibited a relatively small residual N effect, compared to the untreated injected slurries and, to a lesser extent, farm yard manure. This trend is in agreement with simulations of Berntsen et al. (2007), although their absolute values of residual N effects (expressed as the NFRV of manures applied to barley after 30 annual applications) were smaller than ours.

Whatever the exact number of years needed to attain equilibrium between annual organic N inputs and their full apparent decay, the implication of this is that the initially perceived differences in NFRV between anaerobically digested slurry, untreated slurry and farm yard manure, will largely disappear with time. It is hence questionable to base preferences for certain manure types on just their short-term NFRV which is, self-evidently, closely linked to the ammonium content and the C to total N ratio (Reijs et al. 2007), provided that NH3 loss is avoided. If residual N effects are ignored, the long-term environmental impact may be underestimated too. Consequently, it seems short-sighted to call seemingly inert vegetable and garden waste composts ‘environmentally benign’, just because their initial NFRV is low (Nevens and Reheul 2003). Any manure with a narrow mineral N to total N ratio can in the long run be associated with an untimely release of N and this could increase leaching losses. This may in particular apply to arable rotations with long fallow periods (Sörensen 2004).

The present plea to take better account of the long term NFRV when deciding on application rates, has yet another implication. When N recommendations for a crop are partly based on trials that were carried out on sites to which manures have been amply applied before the experiments started, the true N requirements of that crop will be underestimated. One may reason that these combined biases of underestimated NFRV’s and N requirements, will compensate each other in our advices to farmers. However, site-specific N management requires a better disentanglement of both shortcomings. If not, the long term consequences of reduced N rates for yields will not be fully acknowledged (Motavalli et al. 1992). These consequences of residual N effects deserve attention too, in addition to a correct estimate of the long term NFRV of manure. Farming systems are by nature dynamic due to changing economic and environmental drivers. Crop yields and environmental quality do not respond instantaneously to changed practices, however. As such farming systems are always underway to new equilibria without ever reaching them really. This contributes to uncertainties. The present model may help to improve our quantitative understanding of the impact of former and future nutrient management strategies, thus reducing these uncertainties.