Interaction of Nitrogen, Phosphorus, and Potassium Fertilisation and Precipitation on the Nitrogen Use Efficiency of Rainfed Grass

Abstract

Nitrogen (N) fertilisers should be utilised as efficiently as possible. In addition to N fertiliser doses, nitrogen use efficiency (NUE) is influenced by other factors. The effects of four different N, phosphorus (P), and potassium (K) supply levels (0–3) and rainfall periods (dry, normal, wet) were investigated on NUE indices in six selected years of a field experiment. Generally, rainfall and N had the strongest effects. N × rainfall supply interaction increased biomass production from 0.915 t ha⁻¹ (N0, dry) to 10.4 t ha⁻¹ (N3, wet). The N balance varied between −47.3 kg ha⁻¹ (N0, wet) and 218 kg ha⁻¹ (N3, dry). The N uptake per unit N of fertiliser (FNRE) was mainly determined by the P × rainfall interaction, varying between 26.13% (P0, dry) and 83.73% (P3, wet). Both the biomass increment per unit N of fertiliser (agronomic efficiency—AE) and the biomass production per unit N uptake (internal efficiency—IE) were mainly influenced by the N × rainfall interaction, with AE decreasing from 43.7 kg kg⁻¹ (N1, wet) to 10.6 kg kg⁻¹ (N3, dry) and IE from 114 kg kg⁻¹ (N0, normal) to 45.9 kg kg⁻¹ (N3, wet). Both P and, to a lesser extent, K had a significant positive effect on these indices. The N dose of 200 kg ha⁻¹ year⁻¹, the P₂O₅ supply of 153 mg kg⁻¹, and the K₂O supply of 279 mg kg⁻¹ proved to be optimal in terms of NUE indices.

1. Introduction

Among the nutrients that can be taken up from the soil, N is the most significant limiting factor for plant biomass production in terrestrial ecosystems and agriculturally cultivated areas. Of the N sources, N fertilisers are the most important, but significant amounts can also come from other sources. For example, in Europe, N deposition from the atmosphere can be 1–20 kg ha⁻¹ year⁻¹ depending on prevailing wind patterns, rainfall, and proximity to industrialised or densely populated areas, and similar amounts can be released to the soil from irrigation water depending on its source. Legumes and N-fixing bacteria can also fix significant amounts of N if conditions are favourable [1].

N fertilisers are the largest N input, since the N demand of crops is most often met by fertilisers in agricultural production. With the progressively increasing use of N fertilisers, maize grain yield increased from 4 t ha⁻¹ in 1964 to over 8 t ha⁻¹ in 1994 in the USA [2], so the increasing global food demand justifies the use of N fertilisers. However, the primary source of N losses is also the use of N fertilisers, most commonly occurring in the form of soil erosion and nitrate leaching [1]. Excessive use has environmental [3] and health [4] impacts, and is also economically disadvantageous, so the use of N fertilisers requires due care. Applying lower-than-optimal doses, on the other hand, may lead to soil depletion, as well as result in a lower-than-achievable crop yield [5,6]. Due to the mobile nature of N, the time of application is also crucial [7].

A commonly used term for the utilisation of N or N fertilisers is nitrogen use efficiency (NUE). NUE has been conventionally used as the inverse of the plant tissue N concentration, i.e., the amount of tissue that can be produced from a given unit of N [8]. However, due to the different growth and storage strategies and the length of the growing season of different plants [9], as well as agronomic considerations, more useful measures were developed and NUE became more of a catch-all concept, so specific indices are often used instead of NUE. Congreves et al. [10] describe 23 different NUE indices. The NUE indices most often fall into one of the following three groups: (i) indices characterised by the yield/fertiliser ratio (agronomic efficiency or partial factor productivity), (ii) indices characterised by the N uptake/fertiliser ratio (uptake or recovery efficiency), or (iii) indices characterised by the yield/N uptake ratio (utilisation efficiency) [11].

Within a given farm or experiment, NUE generally decreases with the amount of N applied [12]. However, at the global level, the opposite trend has been observed. Countries or regions characterised by high-nitrogen-input systems with high yield have operated with greater efficiency [13]. This could be due to a number of reasons, such as modern crop varieties; better mechanisation, extension service, and soil conditions; diverse macro- and micronutrient supplies; irrigation; or better rainfall supply [14,15]. Thus, in addition to the applied N rates, other factors also play a crucial role in N utilisation, and a deeper understanding of the impact of these factors is essential from a nutritional, economic, and environmental point of view, so that the available N resource, which, in the case of agricultural land, is mostly N fertiliser, is properly utilised and as little of it as possible is wasted [16].

Precipitation is a double-edged sword for NUE indices. Optimal plant biomass production requires the right amount and distribution of precipitation, but as the amount of precipitation increases, the N leaching loss rate may increase [17]. The effect of N fertiliser rates on NUE is a well-researched topic, but the effect or interaction of water availability and other nutrients on NUE is less so, especially in the case of grassland. To the best of our knowledge, the effects of four different N, P, and K treatment levels combined with three rainfall scenarios have not yet been studied. The importance of grassland is great because of its role in animal feeding. By fertilising grassland areas with N, the improvements achieved in the N supply of the livestock system ultimately also appear in the human diet [18].

Based on previous research results, it was assumed that NUE indicators would decrease in parallel with increasing N fertiliser rates and decreasing precipitation. It was further hypothesised that the improved P and K supplies in the soil would increase the NUE indices, and interactions between rainfall and macroelements would be characteristic. It was aimed to determine the optimal N fertiliser treatment, as well as the soil P and K supply levels and rainfall amount that would contribute to optimal NUE values under relatively high grass biomass production.

2. Materials and Methods

The grass was established in a long-term field experiment started in 1973 at the Nagyhörcsök Research Station of the Institute for Soil Sciences, Centre for Agricultural Research, HUN-REN (N 46°51′56.84″; E 18°31′10.17″; alt. 140 m a.s.l.). After various field crops had been studied, the grass was sown in autumn 2000. The grass experiment is located inside the Research Station, surrounded by other experiments set up with field crops. The Research Station site is surrounded by arable fields used for the cultivation of commercial crops, typically corn and wheat. There is no grazing on the sward. The herbivorous wild animals found in the area typically feed on the crops and do not affect the biomass production of the grass.

The soil of the site was calcareous loamy chernozem (Calcaric Phaeosem) with the following parameters in the 0–20 cm layer: pH(H₂O): 8.24, pH(KCl): 7.39, CEC: 28.3 meq 100 g⁻¹, organic matter: 3.04%, CaCO₃: 4.27%. The total and the mineral N content in the soil were 2031 mg kg⁻¹ and 15.3 mg kg⁻¹, respectively [19].

The experimental plots were arranged in a full factorial design involving the possible combinations of three factors: N, P, and K treatments at four different (0 = unfertilised control, 1 = moderate, 2 = high, and 3 = very high supply) levels, resulting in a total of 64 different treatments with 2 replicates in a total of 128 plots. The size of the plots was 6 m by 6 m. The form of the applied fertilisers was Ca-ammonium nitrate (N), superphosphate (P), and potassium chloride (K). The 0, 100, 200, and 300 kg ha⁻¹ yr⁻¹ N treatments corresponding to N0, N1, N2, and N3 levels were applied in two equal doses in autumn and in spring. P and K fertilisers were applied in autumn of 1999 at P0 = 0, P1 = 500, P2 = 1000, and P3 = 1500 kg ha⁻¹ of P₂O₅ and K0 = 0, K1 = 500, K2 = 1000, and K3 = 1500 kg ha⁻¹ of K₂O to provide fertilisation for the following 10–15 years, since the incorporation of these solid fertilisers into the soil is not possible without damaging the sward. Every five years, the P and K concentrations of the soil in the plots were measured to ensure that the nutrient supply levels remained significantly different from one another. The applied fertiliser treatments and the “plant available” ammonium lactate-soluble (AL) P and K contents of the 0–20 cm soil layer are shown in

Table 1. Fertiliser treatments and the ammonium lactate-soluble P₂O₅ and K₂O contents in the 0–20 cm layer of the soil.

	N, P and K Treatment Levels	0	1	2	3
Treatments	N kg ha−1 year−1	0	100	200	300
	P2O5 kg ha−1 in 1999	0	500	1000	1500
	K2O kg ha−1 in 1999	0	500	1000	1500
Nutrient	AL-P2O5 mg kg−1 in 2000	66	153	333	542
concentration in soil	AL-K2O mg kg−1 in 2000	135	193	279	390

. Accordingly, these fertilisation treatment levels will hereafter be referred to as N0, N1, N2, N3, P0, P1, P2, P3, K0, K1, K2, and K3.

The study also included the effect of rainfall amount on NUE indices. The elemental analysis of the grass biomass was carried out between 2001 and 2010, so years that were typically dry, normal, or wet were selected. Two mowings were performed every year except 2003, when only one mowing was performed due to drought. The first mowing took place in the last week of May and the second in the first week of September. Primarily, the amount of rainfall in the months of March and April affected the yield of the first cut of this sward [20], and the total annual biomass production was mainly determined by the amount of precipitation from March to July. Therefore, dry, normal, and wet years were identified based on the amount of precipitation from 1 March until 31 July during the years 2001–2010. For the dry, normal, and wet categories, two typical years were selected that could be clearly separated from each other in terms of precipitation. The following years were involved in the analysis with rainfall from 1 March through 31 July: dry period: 2009—153.2 mm and 2003—163.5 mm, normal period: 2002—206.4 mm and 2008—229.5 mm, wet period: 2001—253.1 mm and 2005—259.2 mm (Figure 1).

The 10-year (2001–2010) average precipitation sum at the experimental site between 1 March and 31 July was 238.1 mm, while the annual sum was 550.8 mm. Precipitation was measured at the research station with a Hellmann rain gauge with an accuracy of 0.1 mm. The precipitation amount was recorded daily at 7 a.m.

The 10-year average temperatures in March, April, May, June, and July were 7.5, 13.3, 18.5, 21.8, and 23.8 °C, respectively. The annual average temperature was 12.2 °C. Certain months of the dry years also had high temperatures. In April 2009, 0 mm of precipitation was accompanied by the warmest average temperature of 16 °C in the years examined, while in 2003, the temperatures of 21.1 and 25.4 °C measured in May and June were the highest compared to the other years.

The following soil analyses were carried out: soil pH(H₂O) in a 1:2.5 soil:water suspension, pH(KCl) [21], total N content [22], mineral N content [23], organic matter content [24], cation exchange capacity (CEC) [25], CaCO₃ content [26], and ammonium acetate-lactate soluble (AL-P₂O₅ and AL-K₂O) plant-available P and K fractions [27].

The aboveground vegetation was harvested at a height of 4 cm above the ground. After mowing, the fresh biomass was weighed, and then it was dried for ten days at 30 °C in a drying room to measure dry biomass. Hereafter, biomass will refer to the harvested dry aboveground biomass.

The Kjeldahl method [28] was used to determine the plant N element content from the dried and finely ground plant samples.

N uptake was calculated by multiplying the dry biomass production by the N concentration of the biomass:

$$\mathrm{N}\mathrm{uptake}\left[{\mathrm{kg}\mathrm{ha}}^{-1}\right]=\mathrm{Biomass}\left[{\mathrm{kg}\mathrm{ha}}^{-1}\right]\times {\mathrm{Biomass}}_{\mathrm{Nconc}}\left[{\mathrm{g}\mathrm{kg}}^{-1}\right]\times 1000$$

(1)

N balance was calculated by subtracting the amount of N applied with fertiliser from N uptake by biomass:

$$\mathrm{N}\mathrm{balance}\left[{\mathrm{kg}\mathrm{ha}}^{-1}\right]=\mathrm{Fertili}\mathrm{s}\mathrm{er}\mathrm{N}\left[{\mathrm{kg}\mathrm{ha}}^{-1}\right]-\mathrm{N}\mathrm{uptake}\left[{\mathrm{kg}\mathrm{ha}}^{-1}\right]$$

(2)

For the characterisation of NUE, the fertiliser-N recovery efficiency (FNRE), agronomic efficiency (AE), and internal efficiency (IE) indices were calculated according to Dobermann [29] and Congreves et al. [10].

FNRE, the percentage of fertiliser N taken up by the plant, was calculated by subtracting the N uptake of N control dry biomass from the N uptake of N treated dry biomass, and then dividing this difference by the applied amount of N via fertiliser:

$$\mathrm{FNRE}[\%]={\displaystyle \frac{{\mathrm{NU}}_{\mathrm{Ntreated}}[{\mathrm{kg}\mathrm{ha}}^{-1}]-{\mathrm{NU}}_{\mathrm{Ncontrol}}[{\mathrm{kg}\mathrm{ha}}^{-1}]}{\mathrm{Fertili}\mathrm{s}\mathrm{er}\mathrm{N}[{\mathrm{kg}\mathrm{ha}}^{-1}]}}\times 100$$

(3)

AE is the biomass increment caused by fertiliser N compared to the unfertilised control, and was calculated by subtracting the N control dry biomass production from the N treated dry biomass production, and then dividing this difference by the applied amount of N via fertiliser:

$$\mathrm{AE}\left[\mathrm{kg}\mathrm{biomass}\mathrm{increase}\mathrm{per}\mathrm{kg}\mathrm{N}\mathrm{applied}\right]={\displaystyle \frac{{\mathrm{Biomass}}_{\mathrm{Ntreated}}[\mathrm{kg}]-{\mathrm{Biomass}}_{\mathrm{Ncontrol}}[\mathrm{kg}]}{\mathrm{Fertili}\mathrm{s}\mathrm{er}\mathrm{N}[\mathrm{kg}]}}$$

(4)

IE, the conversion of the N taken up by the plant into biomass, was calculated by dividing the dry biomass production by the N uptake of the biomass:

$$\mathrm{IE}\left[\mathrm{kg}\mathrm{biomass}\mathrm{per}\mathrm{kg}\mathrm{N}\mathrm{uptake}\right]={\displaystyle \frac{\mathrm{Biomass}[\mathrm{kg}]}{\mathrm{N}\mathrm{uptake}[\mathrm{kg}]}}$$

(5)

A uniform grass seed mix was used when planting the sward with the following rates: Phleum pratense: 19%, Festuca pratensis: 18%, Phalaris arundinacea: 15%, Lolium perenne 13%, Festuca arundinacea: 12%, Dactylis glomerata: 9%, Festuca rubra: 8%, Agropyron cristatum: 6%. The different treatments resulted in different proportions of grassland species and new species were introduced. The relative abundance of the grass species in the examined period can be found in a previous paper [30].

The effect of nutrient supply levels and rainfall periods on dry biomass production, N uptake, FNRE, AE, and IE was evaluated with a mixed-effects linear model ANOVA, where the fixed effects were the N, P, and K nutrient supply levels and rainfall periods and the random effect was the two years selected within each rainfall period. Adding the random effect significantly improved the model for each dependent variable. The main effects and interactions were examined in all possible combinations. Significant differences between treatment groups were tested with Tukey’s HSD post hoc test at p < 0.05. In order to obtain a more precise understanding of the effect of the investigated factors, we calculated the partial eta-squared effect sizes for the effects or interactions that proved to be significant. Partial eta-squared is the proportion of variance linked to an explanatory variable that is not explained by other variables in the model. The following rule applies to the interpretation of effect sizes: <0.01 = very small, 0.01–0.06 = small, 0.06–0.14 = medium, >0.14 = large [31]. The statistical evaluation of the data was conducted with R software (version 4.4.0) [32], and the figures were made with the ggplot2 package (version 3.5.1) [33].

3. Results

N, P, and rainfall had a significant main effect on each of the parameters studied, while K had no significant effect on FNRE or IE (

Table 2. Significance of F-values for each source of variation.

Source of Variation	Biomass	N Balance	FNRE	AE	IE
N	***	***	***	***	***
P	***	***	***	***	**
K	***	**	n.s.	***	n.s.
Rainfall	*	*	*	*	*
N:P	**	***	n.s.	n.s.	n.s.
N:K	*	**	*	n.s.	n.s.
P:K	n.s.	n.s.	n.s.	***	n.s.
N:Rainfall	***	***	*	***	***
P:Rainfall	*	**	***	n.s.	n.s.
K:Rainfall	n.s.	n.s.	n.s.	n.s.	n.s.

Significant effects are marked with *** p < 0.001, ** p < 0.01, and * p < 0.05. n.s. indicates non-significant effect. Three- and four-factorial interactions had no significant effect.

). The interactions of the factors were also significant in several cases, which are analysed in detail below. In the following, only the effects or interactions that produced a significant difference are described.

3.1. Biomass

Biomass production was most affected by rainfall (Figure 2). The average value of 2.88 t ha⁻¹ in dry years increased to 5.30 t ha⁻¹ in normal years and to 8.15 t ha⁻¹ in wet years, averaged over the fertiliser treatments. Based on the partial eta-squared values, the effect of N was not much lower than that of rainfall. Biomass production increased from 2.12 t ha⁻¹ at the N0 treatment level to 5.48 t ha⁻¹ at the N1 level, 6.94 t ha⁻¹ at the N2 level, and 7.23 t ha⁻¹ at the N3 level. For P, the P1 treatment increased the P0 control to 5.87 t ha⁻¹ from 4.42 t ha⁻¹, but treatments above the P1 level did not result in a notable increase in biomass production. The main effect of K had the lowest impact on biomass production; with increasing K treatment, biomass increased from 5.00 t ha⁻¹ at K0 to 5.72 t ha⁻¹ at K3. Of the interactions, N × rainfall was the strongest.

Figure 3 shows the significant interactions. The interaction of N × P (Figure 3A) on biomass was manifested in the fact that biomass at the N1–N3 levels increased to a greater extent in the higher P treatments; in particular, the P0 level resulted in a significantly lower biomass than the other P levels. Furthermore, increasing N treatments differed significantly from N0 to N2, while there was no detectable difference between N2 and N3. Figure 3B shows that the effect of K treatments on biomass also tended to be significantly different at the higher N2–N3 levels. In addition, at the K3 treatment level, the effects of all four N levels were significantly different from each other, while at lower K levels, there was no detectable difference between the N2 and N3 levels. The N × rainfall interaction (Figure 3C) resulted in the largest differences in biomass. The lowest biomass production of 0.915 t ha⁻¹ occurred in the dry period at the N0 level and the highest of 10.4 t ha⁻¹ occurred in the wet period at the N3 level. However, wet and dry rainfall periods differed significantly only for treatments N1–N3 and not for N0. The interaction of P with the rainfall (Figure 3D) was manifested in the fact that biomass production at the P2 level in the dry period (2.89 t ha⁻¹) was not significantly different from the P0 treatment (2.25 t ha⁻¹), while it was significantly different in the other rainfall periods. In addition to increasing rainfall, the P1 treatment caused the greatest increase in biomass compared to P0, while higher P levels led to no considerable increase in biomass.

3.2. N Balance

The N balance was most strongly affected by N (Figure 4). The N balance increased from −23.2 kg ha⁻¹ at the N0 treatment level to 23.8 kg ha⁻¹ at the N1 level, 70.4 kg ha⁻¹ at the N2 level, and 149 kg ha⁻¹ at the N3 level. The effect of rainfall was not far behind that of N. The value of 99.3 kg ha⁻¹ in dry years decreased to 71.2 in normal years and to −5.39 kg ha⁻¹ in wet years, averaged over the fertiliser treatments. For P, there was a decrease from P0 (69.9 kg ha⁻¹) to P1 (51.8 kg ha⁻¹) and from P2 (53.8 kg ha⁻¹) to P3 (44.5 kg ha⁻¹). The main effect of K had the lowest effect on the N balance. With increasing K treatment, the N balance decreased from 62.1 kg ha⁻¹ at K0 to 49.2 kg ha⁻¹ at K3.

The N balance increased with N treatment and was significantly increased by each N level (Figure 5A). The interaction of N × P was manifested in the fact that the N balance at the N2 and N3 levels was significantly lower as a result of the P1–P3 treatments compared to P0. Figure 5B shows that the effect of the K treatments on the N balance was also manifested at the N2 and N3 levels, reducing the value of the N balance as in the case of P. The N × rainfall interaction (Figure 5C) resulted in the largest differences in N balance, with the lowest N balance of −47.3 kg ha⁻¹ in the wet period in the N0 treatment and the highest of 218 kg ha⁻¹ in the dry period in the N3 treatment. However, the wet and dry rainfall periods differed significantly only for the N3 treatment. The interaction of P with the rainfall (Figure 5D) was reflected in the fact that the P3 treatment in the dry period and the P1–P3 treatments in the wet period resulted in a significantly lower N balance compared to the P0 treatment, whereas in the normal period there was no significant difference between the P treatments.

3.3. FNRE—Fertiliser-N Recovery Efficiency

FNRE was most strongly affected by rainfall (Figure 6). Its value was 28.1% in the dry period, 44.5% in the normal period, and 73.6% in the wet period, averaged over the treatments. The second strongest effect was produced by P. The 41.2% value in the P0 treatment increased to 50.6% in the P1 treatment, decreased to 48.0% in the P2 treatment, and reached its highest level of 55.2% in the P3 treatment. The value of 50.9% in the N1 treatment increased to 52.7% in the N2 treatment and decreased to 42.6% in the N3 treatment. The main effect of K was not significant.

The N × K interaction in Figure 7A shows that the effects of K were significantly different at higher N2 and N3 levels, with higher K2 and K3 levels typically resulting in higher FNRE, but the effect was not fully consistent, i.e., at the N2 level, the K2 treatment produced significantly higher FNRE (57.0%) than K0 (47.8%), while at the N3 level, the FNRE was significantly higher in the K3 treatment (48.5%) than in the K2 (39.3%). At the K3 treatment level, increasing N treatments did not cause a significant decrease in FNRE, whereas at lower K levels, FNRE was significantly decreased by the N3 treatment. In the case of the N × rainfall interaction (Figure 7B), the rainfall periods were clearly separated, but only the dry and wet periods were significantly different. In the wet period, the N3 treatment resulted in a significantly lower FNRE (63.1%) compared to N1 (80.1%) and N2 (77.7%), while in the normal and dry rainfall categories there was only a significant decrease between the N2 and N3 treatments. The strongest interaction was observed in the case of P × rainfall (Figure 7C), with the lowest FNRE of 26.1% occurring in the dry period for the P0 treatment and the highest of 83.7% in the wet period for the P3 treatment. The P × rainfall interaction also showed a distinct separation between the rainfall periods, but this was only significant between the dry and wet periods at the P1 level and above. The P treatment levels increased the FNRE, but this was only significant for the normal and wet periods. For normal rainfall, there was a significant increase between P0 (39.6%) and P3 (48.9%), while for the wet period, there was a significant increase between P0 (57.9%) and P1 (78.0%).

3.4. AE—Agronomic Efficiency

Similarly to FNRE, AE was most strongly affected by rainfall, increasing from 14.1 kg kg⁻¹ in the dry period to 29.7 kg kg⁻¹ in the normal period and 31.9 kg kg⁻¹ in the wet period (Figure 8). N had the second strongest effect, causing a decrease from 34.4 kg kg⁻¹ in N1 to 24.1 kg kg⁻¹ in N2 and 17.1 kg kg⁻¹ in N3 treatment. The AE of 20.8 kg kg⁻¹ at the P0 level increased to 26.3 kg kg⁻¹ at P1, 25.1 kg kg⁻¹ at P2, and 28.7 kg kg⁻¹ at the P3 level. The main effect of K had the lowest impact on AE, with an increase from 21.9 kg kg⁻¹ at K0 to 25.0 at the K1 level, whereas at K2 and K3, its values were 27.6 and 26.4 kg kg⁻¹, respectively. Of all the interactions, N × rainfall had the biggest effect.

The interaction of P × K on AE is shown in Figure 9A. The P treatments increased AE, which was also enhanced by the K1–K3 treatments, but at the K0 level, P had no significant effect. The effect of K could be characterised as fluctuating, but higher K levels mostly resulted in significantly higher AE values. The N × rainfall interaction is shown in Figure 9B. AE was the highest at the N1 level in the wet (43.7 kg kg⁻¹) and normal (42.7 kg kg⁻¹) rainfall periods, being significantly higher than for the dry (16.8 kg kg⁻¹) rainfall period, but at higher N levels, the difference was not significant. In the case of the wet and normal rainfall periods, a significant decrease occurred at both the N2 and N3 levels, but only at the N3 level in the case of the dry period (10.6 kg kg⁻¹).

3.5. IE—Internal Efficiency

IE was most affected by rainfall (Figure 10). The value was similar in the dry and wet rainfall periods, at 68.6 and 62.8 kg kg⁻¹, respectively, while in the normal period, it was higher, at 82.3 kg kg⁻¹, averaged over the fertiliser treatments. The effect of N was not much lower. The IE of 101.1 kg kg⁻¹ in the N0 control treatment was consistently reduced by increasing N doses, being as low as 49.6 kg kg⁻¹ in the N3 treatment. The main effect of P on IE was relatively low, and neither the main effect nor the interaction of K with other factors was significant.

The interaction of N × rainfall on IE is shown in Figure 11A. IE was highest for the normal rainfall period at the N0 level (114 kg kg⁻¹), and there was a significant difference between the normal and wet periods at the N1–N3 treatment levels. Regardless of the rainfall period, the value of IE was significantly lower at the N2 level than at the N0 or N1 levels, with a further significant decrease in the N3 treatment. The lowest value was obtained in the wet period in the N3 treatment (45.9 kg kg⁻¹). The P treatment (Figure 11B) significantly improved IE, but only between the P0 and P1 levels, while above the P1 level, there was only a slight decrease.

4. Discussion

Among the macronutrients, N generally has the strongest effect on biomass in the case of both crops and grassland [34,35]. However, rainfall has a decisive influence on the biomass production of unfertilised grassland [36]. Also, according to Sinclair and Rufty [37], N and the availability of water are the main limiting factors for crop yield increases, as confirmed by the results of the present experiment. Based on the partial eta-squared effect sizes, it was found that favourable rainfall amount and distribution had an even greater effect on biomass production than N, although the strength of the two effects was similar (Figure 2). The N treatment increased the biomass production up to a dose of 200 kg ha⁻¹, but it was significantly higher in the wet period than in the dry period (Figure 3). A contrary result was obtained by Wang et al. [38], who found that N application had a much stronger effect on barley biomass production than water supply.

The N balance is a simple but very useful indicator for monitoring the fate of N, from which it is possible to deduce the optimal doses of fertiliser to be applied, thus reducing N losses and increasing sustainability [39]. A moderate positive N balance is usually required for adequate biomass production or crop yields. Xu et al. [40] found that the N balance was optimal between 44.4 and 64.9 kg ha⁻¹ in winter wheat–summer maize rotation systems, whereas for lower N balances the yield may be reduced to an undesirable extent. In the present experiment, there was mostly no significant biomass increase above the N2 treatment (Figure 3), and the N balance for the N2 level was 70.4 kg ha⁻¹, averaged over the rainfall categories. However, there was a marked but not significant difference between the different rainfall periods. The value of 126 kg ha⁻¹ in the dry period decreased to 89.8 kg ha⁻¹ in the normal period and to −4.33 kg ha⁻¹ in the wet period (Figure 5), due to higher biomass production, which resulted in higher plant N uptake from the soil.

The FNRE index expresses the efficiency of plant N uptake from fertiliser N, which is important in the case of grass due to the amount of N available for grazing animals per unit area, which is built into animal protein and thus contributes to human nutrition. According to Dobermann [29], the FNRE is 30–50% for cereals and 50–80% in well-managed cropping systems, in the case of low levels of N use or low soil N supplies. This is in line with the values obtained in this experiment, which fluctuated, in particular due to the effect of precipitation, between 28.1% (dry) and 73.6% (wet). Cassman et al. [2] reports that in the rice–wheat systems of India, the FNRE was 18% in one year and 49% the next because of weather conditions. The FNRE values of winter wheat and summer maize were also affected by the precipitation amount [41]. The FNRE index is sensitive to plant N uptake, which is the product of biomass production and the N concentration in the biomass. Rainfall also had the strongest effect on biomass production, and a previous study showed that plant N concentration did not decrease in the case of higher rainfall amounts [30], which might have been expected due to the dilution effect [42]. Therefore, in this experiment, the FNRE was quite highly dependent on the biomass production. Interestingly, the partial eta-squared values indicated that P had a greater effect on FNRE than N (Figure 6). P treatment was also found to have a positive effect on the FNRE for guinea grass (Panicum maximum), increasing it from 34% to 54–56% [43]. Egan et al. [44] also confirmed the increasing effect of P on FNRE in a grassland experiment. Similarly, in the present experiment, the FNRE was 41.2% at the P0 level and 55.2% at the P3 level, but the range was much more extreme when rainfall was taken into consideration, with values of 26.1% for the P0 treatment in the dry period and 83.7% for the P3 treatment in the wet period (Figure 7). Duan et al. [45] also obtained significant P effects compared to the K treatment for wheat and corn crops, and P is mentioned as a key factor in relation to NUE improvement. However, in accordance with Liebig’s law of minimum, the effect of P, K, or other elements on a given soil depends primarily on the supply of that element in the soil, as well as on the needs of the plant. In the present experiment, due to the characteristics of the calcareous loamy soil, P had a stronger effect than K. In a sandy soil, however, stronger K effects may be expected [46].

AE is affected not only by crop yield or biomass production but also by N treatment. In the present experiment, its average value was 25.2 kg kg⁻¹, which is favourable according to Dobermann [29], who classified values above 25 kg kg⁻¹ as typical for well-managed systems. Partial eta-squared analysis showed that AE was most affected by rainfall and N (Figure 8). In the N1 and N2 treatments, in the normal and wet rainfall periods, the AE values exceeded 25 kg kg⁻¹. However, in the N3 treatment and in the dry period, the values remained below this limit (Figure 9). In a meta-analysis study, Liang et al. [47] found that the average value of AE was 20.8 kg kg⁻¹, while the most marked differences were caused by the N application rate, crop, type of N fertiliser, and amount of precipitation. The AE values were higher for combined N, P, and K treatments compared to N alone, but the difference was not significant, so the effect of P and K was only moderate. In the present experiment, P and K treatments had significant effects and interactions on AE, but this effect may vary depending on the soil supply and plant requirements.

IE was clearly most affected by rainfall and N, but also by P treatment (Figure 10). The strong negative effect of N was also confirmed in other grass experiments, in addition to the positive effect of P treatment [43]. The IE values in this experiment were relatively high, averaging 71.2 kg kg⁻¹, which was higher than the typical 55–65 kg kg⁻¹ [29], but at the N2 and N3 treatment levels, the values were mostly below 55 kg kg⁻¹ in the dry and wet rainfall periods (Figure 11). Interestingly, the normal rainfall period was clearly separated from the wet and dry rainfall periods, which were similar to each other. This suggests that in the wet period, N uptake increases with biomass production, and in the dry period, N uptake and biomass production decrease in parallel, while in the normal period, the plant can make the best use of the N uptake for biomass production.

The P and K treatments had a positive effect or interaction on almost all the NUE indices examined. This may be because, in addition to the above-ground biomass increase, the root biomass could certainly have also increased. The increase in root length and mass could have improved N uptake, and thus also the NUE indices. The effect of P and K treatment on increasing root biomass and N uptake, and thereby improving NUE indices, has also been confirmed by other research [48]. Liu et al. [49] found that in rice plants, the ratio of plant N uptake in relation to the total length and weight of the root increased due to adequate P supply; furthermore, P and K increased the activity of N-metabolising enzymes.

NUE indices can typically be controlled by the interactions of biological processes as well [50]. The changes in IE and other NUE indices may also be influenced by the fact that, as mentioned above, although a uniform grass seed mix was used when the grassland was established, different treatments resulted in different proportions of species cover and new species were spontaneously colonised [51]. Therefore, in nutrient-poor treatments, species tolerant of nutrient-poor conditions were dominant, and in medium-to-well-supplied plots, species that could use nutrients well were dominant, and their nutrient utilisation and growing strategies were different [8]. Even different grass cultivars can have different NUEs [52,53]. It is well known that in the dry period, nutrient uptake by the root is limited, and in the wet period, it is enhanced, but this can also lead to N leaching. It is reasonable to assume that the species that colonise each treatment can make the most of nutrient conditions under average or normal rainfall, while both too much and too little rainfall may reduce their efficiency.

5. Conclusions

Dry biomass production increased up to the N2 (200 kg ha⁻¹ year⁻¹) level in all three rainfall periods. N treatments only caused significant FNRE reductions above N2. AE showed a significant decrease between N1 and N2 and between N2 and N3 for both wet and normal rainfall periods, while in the dry rainfall period, it showed a significant decrease only at N3 compared to N1 and N2. This suggests that, of the N levels used in this experiment, N2 dose may be optimal for all three rainfall periods, but of course, better rainfall supply is more favourable. To a lesser extent, P and K also improved biomass production and NUE indices. P improved all the indicators studied, but typically or to a significant extent only up to the P1 level, i.e., above a supply level of 153 mg kg⁻¹ P₂O₅ in the soil, there was no improvement in biomass or NUE. In contrast, increasing K supplies enhanced biomass production and, at the N3 level, FNRE, while the N balance decreased up to the highest K3 level, corresponding to 390 mg kg⁻¹ K₂O, but to a lesser extent compared to P. Considering the recommended N2 treatment, the K2 level, corresponding to 279 mg kg⁻¹ K₂O, was optimal from the point of view of the FNRE index.

This long-term field experiment made it possible to investigate the effects of N, P, and K treatments on NUE indices in years with different rainfall. The results may help to increase the NUE of grasslands, to avoid unnecessary fertiliser application, and to increase the body weight of grazing animals more efficiently. The limitation of the applicability of the results is that in the case of different grass species composition, different soil parameters, and different climatic conditions, the effects of the tested treatments may differ and a different treatment combination may be optimal. Measuring N leaching and volatilisation under more controlled conditions would provide additional valuable data, which could be goals of further research.