Introduction

Efficient use of nitrogen (N) requires careful matching of N supply to crop demand. Often application of mineral fertilisers is combined with application of organic amendments such as farmyard manure (FYM), slurries, and crop residues (also called organic manures, organic fertilisers or organic inputs). N in organic amendments generally has a lower availability to crops than N in mineral fertilisers, mainly depending on the C:N ratio of the amendment (Flavel and Murphy 2006). Therefore, N in organic amendments must be carefully matched with mineral fertiliser N application to avoid leaching while making sure sufficient N is available for crop growth. This requires the characterization of the organic amendments by their Nitrogen Fertiliser Replacement Value (NFRV), also called the Mineral Fertiliser Equivalent (Jensen 2013).

NFRV can be based on the amount of mineral fertiliser N which is substituted by an amount of organic amendment-N (kg/kg) while attaining the same crop yield (Herron and Erhart 1965; Schröder 2005a; Schilling 1987). Crop yields can be expressed as fresh matter weights (FM), dry matter weights (DM) or N contents (N yields, Jensen 2013). These are all valid procedures, with the difference that values of NFRV based on N yields are often slightly lower than those based on DM or FM weights (Jensen 2013). N in organic amendments is always accompanied by other nutrients, such as phosphorus (P), potassium (K) or sulphur (S) which also affect crop yields. It is therefore important to exclude these effects when estimating values of NFRV based on yields (either DM, FM or N yields, Schröder 2005a).

Lory et al. (1995) have suggested to calculate NFRV using economic optimal N rates (mineral fertiliser N application rates at which marginal crop yields offset marginal fertiliser costs) with and without organic amendments. Yields at economic N rates with and without organic amendments might however differ, which hampers comparison at equal yield levels. In addition, economic N rates are dependent on prices of fertilisers and harvested products, which makes estimations less robust through time and space.

Another manner to determine NFRV of organic amendments is by using isotope dilution techniques. Using 15N labelled materials, the fate of N from either organic amendments or mineral fertilisers is measured among plant and soil fractions and compared (Diekmann et al. 1993; Janzen et al. 1990). Additional methods to calculate NFRV (such as the analysis of near infrared reflectance spectra of organic amendments) have been proposed but need further development (Delin et al. 2012).

Values for NFRV differ when estimated in the first year of application of organic amendments (short-term NFRV) or after repeated applications and several years (long-term NFRV, Gutser et al. 2005; Schröder 2005b), with higher values found for long-term NFRV. For FYM, NFRV ranges between 0.10 and 0.70 (Birkmose 2009; Jensen 2013; Pikula et al. 2016; Webb et al. 2013). For slurry, NFRV ranges between 0.20 and 0.90 (Birkmose 2009; Delin et al. 2012; Jensen 2013; Kundler et al. 1989; Langmeier et al. 2002; Webb et al. 2013). For straw, NFRV has been estimated to be around zero (Dhillon and Dev (1984) as cited by Katyal 1993). For green manures, NFRV has been estimated ca. 0.4 (Janzen et al. 1990), but this will probably depend on the species of green manure cultivated.

Factors known to affect NFRV (at a given dose of the amendment) include the form of N in the amendment, crop type cultivated, soil type, method of application, time of application and the manuring history which may govern N retention and losses (Birkmose 2009; Jensen 2013; Katyal 1993; Kundler et al. 1989; Webb et al. 2013). Here we evaluate the effect of an additional factor on NFRV which is currently not taken into account: the total N supply. We address the following research question: Does NFRV of organic amendments depend on total N supply?

Materials and methods

To answer the research question, values for long-term NFRV were calculated based on yield response curves (either DM or FM weights of marketable products) at low and high total N supply. This required data from field trials where different N rates were applied with and without organic amendments. In such a set-up, mineral fertiliser N saved with an addition of organic amendment at low mineral fertiliser rates (N saved low) can be compared with mineral fertiliser N saved by adding the same amount of organic amendment at high mineral fertiliser rates (N saved high). In these cases, sufficient P and K has to be applied to ensure these macro-nutrients are not a yield limiting factor and affect the calculation of NFRV of organic amendments. Organic amendments however do not only supply N, P and K but also other nutrients, such as sulphur (S). Over the past decades, S was often not yield limiting due to the ample supply from the atmosphere (Eriksen 2009). More recently this has changed but not always adjusted for in field experiments.

In a recent meta-analysis of long term experiments in Europe (Hijbeek et al. 2017), a database was compiled with data from 20 long term experiments. All these 20 experiments fulfilled the following six criteria: (1) at least four increasing levels of mineral fertiliser N rates without organic amendments; (2) at least four increasing mineral fertiliser N rates with organic amendments; (3) P and K applied in ample amounts on all fields; (4) at least five years of yield data; (5) if crops are cultivated in rotation, yield data available for at least two rotation cycles; (6) yield data reported for individual crop types (no aggregated data of whole-rotation yield output were used). Assessing NFRV called for an additional requirement particular to this study: (7) N contents of organic amendments must be known.

Using these seven criteria, we assembled a total of 38 data sets (with multiple crops and organic amendment types) from eight experiments in five countries across Europe (Table 1). The experiments ranged in duration between 15 and 44 years. These data, therefore, can be used to consider long-term NFRV, including the greater part of the residual effect of manure application (Hernández et al. 2013; Jensen 2013; Schröder 2005b). The setup of these experiments implies that mineral fertiliser N rates are always confounded with total N supply. We therefore use the term ‘total N supply’.

Table 1 Details of eight experiments included in the meta-analysis
Full size table

For each data set, two yield response curves to mineral fertiliser-N were fitted, one without (Y0) and one with organic amendments (YA), following George (1984), see Fig. 1.

$$Y = a + b*0.99^{N} + c*N + \varepsilon$$
(1)

In Eq. 1, Y is crop yield (t/ha) expressed in either fresh or dry matter weight of the marketable product, N is the mineral fertiliser N rate (kg N/ha), a, b and c are parameters to be fitted and ε is the error term.

Fig. 1
figure 1

Illustration of calculating mineral fertiliser N savings from organic amendments at low total N supply (Nsaved low) and at high total N supply (Nsaved high). Y0 is the yield response curve without organic amendments. YA is the yield response curve with organic amendments. At the left yields are shown from maize cultivated in Iasi (Romania) between 1984 and 1992, with and without farmyard manure. At the right yields are shown from winter barley cultivated in Keszthely (Hungary) between 1999 and 2010, with and without farmyard manure

Full size image

For each pair of curves, the amount of mineral fertiliser saved by using an organic amendment was calculated at low and high total N supply (N saved low and N saved high). N saved low was defined as the amount of mineral fertiliser N required without organic amendments to match the yield obtained with the amendment alone (N = 0, Eq. 2):

$$Y^{0} \left( {N_{saved\, low} } \right) = Y^{A} \left( {N = 0} \right)$$
(2)

To calculate N saved high, the highest yield level which was reached by both response curves within the experimental set-up was determined for each pair of response curves and called \(Y_{ \hbox{max} }^{0,A}\). This could be the maximum of one of the two response curves (as in Fig. 1a) or the highest yield reached within the experimental set-up (as in Fig. 1b). Accordingly, \(N_{ \hbox{max} }^{0}\) was defined as the mineral fertiliser N rate needed to reach \(Y_{ \hbox{max} }^{0,A}\) without organic amendments. Following, N saved high is the reduction in mineral N fertiliser when reaching \(Y_{ \hbox{max} }^{0,A}\) with organic amendments:

$$Y^{A} \left( {N_{max}^{0} - N_{saved\,\,high} } \right) = Y_{max}^{0,A}$$
(3)

Thus, N saved low represents the reduction of mineral fertiliser when using organic amendments at low total N supply. N saved high represents the reduction of mineral fertiliser when using organic amendments at high total N supply. In some cases, additional mineral fertiliser N was added to the organic amendments to allow for decomposition of straw or successful cultivation of green manures. If so, this was included in the calculation of N saved low and N saved high.

To find NFRV low and NFRV high, both N saved low and N saved high were divided by the total N content of the organic amendments (Eqs. 4, 5).

$$NFRV_{low} = \frac{{N_{saved\,low} }}{Total\,N\,organic\,amendment}$$
(4)
$$NFRV_{high} = \frac{{N_{saved\,high} }}{Total\,N\,organic\,amendment}$$
(5)

Following, for each pair of response curves, the difference between NFRV low and NFRV high was calculated. Statistical models based on the function lme (linear mixed-effects model) of package nlme (Pinheiro et al. 2015) were used to find the least squares (LS) means for NFRV low and NFRV high per type of organic amendment. These models can account for random effects (Konstantopoulos 2011). In this analysis, the eight experiments were included as random effects. All data were processed in R (version 3.2.5; R Core Team 2015).

Results

N saved low and N saved high for FYM were calculated for seventeen pairs of response curves (Fig. 2). Mean NFRV low of FYM was 0.53 (±0.26–95% Confidence Interval (CI)) and mean NFRV high of FYM was 1.12 (±0.71–95% CI) (n = 17, Table 2). NFRV high was 2.13 times larger than NFRV low (p = 0.04, Table 2).

Fig. 2
figure 2

Yield response curves to mineral fertiliser, with and without FYM. Blue arrows pointing to the left indicate a saving of mineral fertiliser N when applying FYM. Blue arrows pointing to the right indicate additional mineral fertiliser N is needed when applying FYM. a winter wheat in Bologna 1990–2000; b sugar beet (DM) in Iasi 1986–2006; c maize (DM) in Iasi 1984–1992; d winter barley in Keszthely 1999–2010; e potatoes in Lukavec 2002–2007; f winter barley in Lukavec 2002–2007; g sugar beet (DM) in Muencheberg 1985–1993; h potatoes (DM) in Muencheberg 1983–1999; i winter rye (DM) in Muencheberg 1996–2004; j winter wheat (DM) in Muencheberg 1984–2002: k maize (DM) in Muencheberg 1995–2003; l spring barley (DM) in Muencheberg 1978–1982; m winter barley (DM) in Speyer 1994–1999; n winter wheat (DM) in Speyer 1994–1999; o sugar beet in Sproda 1999–2010; p spring barley (86% DM) in Sproda 1999–2010; q winter wheat (86% DM) in Sproda 1999-2010. (Color figure online)

Full size image
Table 2 Long-term NFRV low and NFRV high for different types of organic amendment
Full size table

For slurry, only one dataset was available (Fig. 3). For this dataset, NFRV low was 0.35 and NFRV high was 0.58. The ratio between NFRV high and NFRV low was 1.65 (n = 1, Table 2).

Fig. 3
figure 3

Yield response curves of winter wheat to mineral fertiliser, with and without slurry applied in Bologna (1999–2010). Blue arrows pointing to the left indicate saving of mineral fertiliser N when applying slurry at low and high total N supply. (Color figure online)

Full size image

For straw, one-third of the NFRV values was negative at low total N supply, while almost half (47%) of the NFRV values for straw was negative at high total N supply (n = 15, Fig. 4). Mean NFRV low for straw was 0.12 (±0.36–95% CI) and mean NFRV high was 0.35 (±1.67–95% CI). NFRV high for straw was 3.07 times larger than mean NFRV low, but not significantly so due to the large variation in the effects of straw on mineral fertiliser requirements (p = 0.79).

Fig. 4
figure 4

Yield response curves to mineral fertiliser, with and without straw application. Blue arrows pointing to the left indicate a saving of mineral fertiliser N when applying straw. Blue arrows pointing to the right indicate additional mineral fertiliser N is needed when applying straw. a maize in Bologna 1991–2001; b winter wheat in Bologna 1990–2000; c potatoes (FM) in Methau 1999–2010; d spring barley (86% DM) in Methau 1999–2010; e sugar beet in Methau 1999-2010; f winter wheat (86% DM) in Methau 1999–2010; g maize (DM) in Muencheberg 1995–2003; h potatoes (DM) in Muencheberg 1983–1999; i spring barley (DM) in Muencheberg 1978–1982; j sugar beet (DM) in Muencheberg 1985–1993; k winter rye (DM) in Muencheberg 1996–2004; l winter wheat (DM) in Muencheberg 1984–2002; m spring barley (86% DM) in Sproda 1999–2010; n sugar beet in Sproda 1999–2010; o winter what (86% DM) in Sproda 1999–2010. (Color figure online)

Full size image

In five data sets, a combination of straw and green residues (green manures and/or beet leaves) was used as an organic amendment (Fig. 5). In all five cases, additional mineral fertiliser N was added to support decomposition of straw and/or cultivation of green manures, which was accounted for. At low total N supply, combining straw and green residues had a negative NFRV in two cases. At high total N supply, negative NFRV did not occur for a combination of straw and green residues in our dataset. Mean NFRV low was 0.14 (±0.39–95% CI) and mean NFRV high was 0.91 (±1.21–95% CI). For the combination of straw and green residues, NFRV high was 6.6 times larger than NFRV low, but not significantly different, due to the small sample size (Table 2).

Fig. 5
figure 5

Yield response curves to mineral fertiliser, with and without a combination of straw and green residues. X-axes indicate mineral fertiliser N, including additional mineral fertiliser N added for the cultivation of green manures or decomposition of straw (at each site on average 16.7 kg N/ha/year). Blue arrows pointing to the left indicate a saving of mineral fertiliser N for a combination of straw and green residues. Blue arrows pointing to the right indicate additional mineral fertiliser N is needed for a combination of straw and green residues. a maize (DM) in Iasi 1984–1992; b sugar beet (DM) in Iasi 1986–2006; c sugar beet in Speyer 1994–1999; d winter barley (DM) in Speyer 1994–1999; e winter wheat (DM) in Speyer 1994–1999. (Color figure online)

Full size image

Discussion

For all types of organic amendments, mean NFRV was higher at high total N supply than at low total N supply. Only for FYM the difference was significant (p = 0.04). For the other types of organic amendment either the sample sizes were too small or the variation too large for detecting significant differences.

Currently, values for NFRV are often estimated by applying organic amendments without extra mineral fertilisers. In Europe, farmers often apply a mixture of organic amendments and mineral fertilisers, thus having a higher total N supply (Potter et al. 2010). We compared values for NFRV at high total N supply (either the highest yield within the experimental set-up or the maximum of a response curve) with values of NFRV for organic amendments at low total N supply (at the start of the response curves). In practice, farmers will (1) operate within environmental restrictions on N applications and (2) try to achieve an economic optimum. Therefore, most often farmers will target yields somewhat below the yield maximum of the N response curves, which will generally be closer to the values we calculated for NFRV high than for NFRV low.

Crop residues are often not used by farmers to replace mineral fertilisers as they have a high C:N ratio and might require N for decomposition (Knapp et al. 1983). In these cases, crop residues have a negative NFRV. Straw had a negative NFRV in one-third (NFRV low) to half (NFRV high) of the cases in our study. A combination of green manures and straw had a negative NFRV in two-fifth (NFRVlow) to none (NFRV high) of the cases.

Our findings are based on 38 datasets from eight long term experiments. Besides straw and FYM, however, there were only a limited number of observations per organic amendment type (Table 2). Further work is needed to assess to what extent NFRV differs between low and high total N supply for the other amendment types, for different soil and climate conditions and in combination with different crop types. Such an exercise would require a much larger dataset, preferably with N contents of organic amendments available.

A number of mechanisms might change the shape of the yield response curve and therefore cause a difference between the values of NFRV low and NFRV high, on which we will now elaborate. A first mechanism could be that organic amendments provide other benefits than N to the crops (other nutrients or soil improvements), which has not been adjusted for in the mineral fertiliser only plots. In this case, the calculated values of NFRV cannot be solely attributed to N but include additional yield effects (Janssen 2002; Wadman et al. 1987). Elimination of the limitation(s) that cause such additional yield effect (e.g. by supplying a mineral S fertilizer) would result in a steeper mineral fertilizer response (i.e. a better conversion of applied mineral fertiliser N into DM or FM yield). Consequently, NFRV high would decrease and approach NFRV low . A second alternative explanation is that addition of mineral fertiliser might increase available N from the amendment (or vice versa). Third, addition of mineral fertiliser could increase uptake efficiency of N from organic amendment (or vice versa). Finally, a combination of above mechanisms might take place at the same time.

In our study, savings of mineral fertiliser when using organic amendments were compared at low and high total N supply. At low N supply, using only organic amendments was compared with using only mineral fertiliser. At high total N supply, using a combination of organic amendments and mineral fertiliser was compared with using only mineral fertiliser. The larger value of NFRV at high total N supply could suggest that the NFRV of organic amendments is higher when organic inputs are combined with mineral fertiliser. This suggestion could be further investigated by calculating NFRV of organic amendments when applied in different amounts. If validated, this would open a new perspective on the advantage of distributing available organic amendments among many farmers in a given region: the region would require less fertiliser to produce the same yield output, than when the available organic amendments were concentrated in few farms.

In this study, we have used yield response curves expressed in either dry or fresh matter to calculate values for NFRV low and NFRV high. Other methods exist to calculate NFRV, such as based on N content (Jensen 2013) or using isotope dilution techniques (Diekmann et al. 1993; Janzen et al. 1990). An exploration of different methods into NFRVlow and NFRVhigh might give further insights into possible mechanisms.

If our findings can be further generalized, the observed contrast in NFRV between the respective N ranges may have practical implications for fertiliser recommendations, depending on the mechanisms underlying the contrast. Similarly, the contrast might justify adjustment of statutory values for fertiliser equivalency coefficients as used to regulate N use in the Action Programmes—under the Nitrates Directive—of various EU member states.

Conclusions

Currently, values for NFRV are often based on experiments where crop yields in plots with only organic amendments (no mineral fertilisers) are compared with crop yields in plots with only mineral fertiliser, at relatively low total N supply. In many European countries, however, farmers operate at high total N supply. NFRV coefficients play a key role in fertiliser recommendation systems and tools (e.g. MANNER-NPK, Nicholson et al. 2013) and various national Action Programmes in response to the EU Nitrates Directive (EEC 1991). For example, current Action Programmes in the UK, Denmark and the Netherlands use NFRV values of 0.10, 0.45 and 0.30–0.60 kg/kg, respectively for total N in FYM, and values of 0.45, 0.75 and 0.60–0.80 kg/kg for pig slurry (Dalgaard et al. 2014; RVO 2014; UK 2015).Footnote 1 These values are often a political compromise, but based on short or long-term estimations of NFRV. These statutory values define, in combination with N application standards, the maximum amounts of N fertiliser that farmers in those countries may apply, depending on crop and soil type.

This meta-analysis of eight long term experiments on different types of organic amendments shows that NFRV of FYM was roughly up to two times larger at high than at low total N supply rate. Currently, NFRV is usually assessed at the ‘lower end’ of the N response curve. Considering that farmers in Europe normally operate at relatively high mineral fertilisation rates, the use of NFRV determined at the ‘lower’ end of the N response curve may underestimate NFRV of organic amendments. When using these lower values to estimate the N fertilizer replacement value of organic amendments, complementary mineral N fertiliser requirements are estimated to be larger than actually needed for achieving certain yields. This might lead to overuse, lower N use efficiency and more losses to the environment.