Plant phosphorus nutrition indicators evaluated in agricultural grasslands managed at different intensities

Europ. J. Agronomy 44 (2013) 67–77

Contents lists available at SciVerse ScienceDirect

European Journal of Agronomy journal homepage: www.elsevier.com/locate/eja

Plant phosphorus nutrition indicators evaluated in agricultural grasslands managed at different intensities F. Liebisch a,∗ , E.K. Bünemann a , O. Huguenin-Elie b , B. Jeangros c , E. Frossard a , A. Oberson a a b c

Institute of Agricultural Sciences (IAS), Swiss Federal Institute of Technology Zurich (ETH), Group of Plant Nutrition, Eschikon 33, 8315 Lindau, Switzerland Research Station Agroscope Reckenholz-Tänikon (ART), Reckenholzstrasse 191, 8046 Zürich, Switzerland Research Station Agroscope Changins-Wädenswil (ACW), Route de Duillier 50, Case Postale 1012, CH-1260 Nyon 1, Switzerland

a r t i c l e

i n f o

Article history: Received 11 November 2011 Received in revised form 31 July 2012 Accepted 13 August 2012 Keywords: Grassland Management Plant mineral analysis Phosphorus nutrition diagnosis Nutrient ratio Phosphorus nutrition index

a b s t r a c t A precise assessment of the phosphorus (P) nutrition status of plants is necessary for an efficient P management in agricultural grasslands. Plant mineral analysis is a tool to identify the nutrition status of grasslands and several P indicators derived thereof are available. However, the interpretation of plant P indicators in grassland samples is complex due to variation in botanical composition, changing nutrient concentrations during growth and interactions between nutrients. The aim of this study was to compare indicators on the P nutrition status of plants and eventually to improve their application in agricultural grasslands. We studied three agricultural grassland types in Switzerland that were managed at different intensity either for high biodiversity or high forage production. Each grassland type was for 5–25 years subjected to treatments of low, recommended or high rates of P, nitrogen (N) and potassium (K) fertilizer. The recommended P rates ranged from 15 to 48 kg P ha−1 year−1 . We measured plant P, N and K concentrations in the aboveground biomass of grasses, legumes and forbs. The evaluated P indicators were: P concentration, N:P and K:P ratios and the P nutrition index (PNI). The PNI is calculated as P concentration divided by the linear relationship describing optimal sward P concentration as a function of N concentration (PNI = 100P/(0.06 × 5N + 1.5)). We observed significant yield reduction compared to the expected yield only when one or more nutrients were omitted from fertilization. Fertilizer P input higher than recommended did not significantly increase yields above the yield expected for the respective management and altitude. Under P limiting conditions, forbs and legumes had significantly higher P concentrations than grasses. Additionally, the proportion of legumes affected the P indicators integrating N. Therefore, we used the P indicators in the grass fraction, which was always the main botanical fraction. In grasses all P indicators differentiated between P fertilized and non-P-fertilized treatments. Concentrations of P from 2.1 to 3.0 mg g−1 indicated sufficient P supply, while yield was reduced at lower and not increased at higher concentrations, suggesting luxurious P consumption. Our results for N:P and K:P suggested optimal ranges of 5.5–9.0 and 6.0–10.5, respectively. The PNI showed a clear differentiation between deficient, sufficient or surplus fertilizer inputs. For the precise and correct interpretation of the plant P nutrition status in agricultural grasslands under different management, we propose to use the PNI in the grass fraction. Finally, the interpretation of the indicators was valid for the agricultural grasslands managed at different intensities in spite that the grass fraction was composed by different species. © 2012 Elsevier B.V. All rights reserved.

1. Introduction

Abbreviations: PNI, phosphorus nutrition index; V, Les Verrières; W, Watt; B, Baldegg; DM, dry matter; PCO2 , available soil phosphorus extracted with CO2 saturated water; P0, no P fertilization; P0.5, less P fertilization than recommended; P1, recommended P fertilization; P2, higher P fertilization than recommended. ∗ Corresponding author. Present address: Institute of Agricultural Sciences (IAS), Swiss Federal Institute of Technology Zurich (ETH), Group of Crop Sciences, Universitätsstrasse 2, 8092 Zurich, Switzerland. Tel.: +41 44 6323323; fax: +41 44 6321143. E-mail addresses: [email protected] (F. Liebisch), [email protected] (E.K. Bünemann), [email protected] (O. Huguenin-Elie), [email protected] (B. Jeangros), [email protected] (E. Frossard), [email protected] (A. Oberson). 1161-0301/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.eja.2012.08.004

Agricultural grasslands cover a large area in Europe (Peeters, 2009). Losses of P from intensively managed grasslands were identified as a main source of P to surface waters (Haygarth et al., 2006; Pilgrim et al., 2010). In addition, the decline of biodiversity in grasslands has been related to the accumulation of P in soils (Wassen et al., 2005; Ceulemans et al., 2011). Nowadays, grasslands managed at lower intensities, i.e. with lower nutrient inputs and lower yield expectation, are used for biodiversity conservation and thus are often subsidized by governments (Isselstein et al., 2005). On intensively managed grasslands, high forage production at the same time as minimizing P losses remains a challenge (Haygarth

68

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

et al., 2006) which requires a correct assessment of the plant P status. Plant mineral analysis is used as a standalone or supportive tool for nutrient management in agriculture, especially in perennial crops such as grasslands. Thresholds and interpretation schemes of mineral nutrient concentrations in grassland samples were mostly established by means of nutrient addition experiments, where a nutrient is added until the maximum yield is obtained (Black, 1993; Duru and Thélier-Huché, 1997). Such experiments might be of restricted use to identify thresholds for grasslands managed with lower yield expectations. In grasslands, measured P concentrations are interpreted using either critical P concentrations below which dry matter (DM) production is limited by P or concentration ranges that indicate sufficient P supply. To include the interaction of P with other elements, nutrient ratios (N:P and K:P, Venterink et al., 2003) and other combined nutrient indicators such as the P nutrition index (PNI; Duru and Thélier-Huché, 1997) have been developed. The PNI is calculated by dividing the measured P concentration by the linear relationship describing optimal P concentration as a function of sward N concentration. This approach accounts for the fact that both, P and N concentration in plant tissue decrease with increasing DM production, but to different degree (Duru and Thélier-Huché, 1997). Plant analysis in grasslands is mostly done on bulk samples (Flisch et al., 2009) or sometimes on specific species such as perennial ryegrass (Lolium perenne) or white clover (Trifolium repens) (Høgh-Jensen and Schjoerring, 2010). Critical values and optimal ranges of concentrations and ratios are still under discussion because they might differ for species, grassland type, management and local climate conditions (Whitehead, 2000). The aim of our study was to evaluate plant P indicators derived from plant mineral analysis of aboveground biomass in agricultural permanent grasslands which were managed either for high agricultural production or for high biodiversity. Our hypothesis was that plant P indicators would take different values if high agricultural production was to be achieved than if biodiversity was to be conserved. The investigated plant P indicators were simple P concentrations, P in relation to N and K, and the PNI. We analysed the three botanical fractions grasses, legumes and forbs separately to assess the effect of botanical composition on the different plant P indicators. To identify deficient, sufficient and surplus P nutrition of the swards, we evaluated the plant P indicators in relation to long term fertilizer application and yield response relative to the yield expected for the respective management and altitude. Finally, we aimed at proposing critical values to improve the interpretation of plant P analysis in permanent grasslands managed at different intensities.

2. Materials and methods 2.1. Experimental locations Three locations with different cutting frequency, fertilizer input and environmental conditions were selected for this study: (1) north of Les Verrières (V) in the Jura mountains (46◦ 56 18 N, 6◦ 27 38 E, 1140 m above sea level), (2) Watt (W) near Zürich (47◦ 26 44 N, 8◦ 29 32 E, 500 m above sea level) and (3) in the eastern area of the Lake Baldegg (B) watershed (near 47◦ 11 55 N, 8◦ 31 01 E, 560–780 m above sea level). The locations V and W are long-term fertilizer trials on agricultural grasslands managed for biodiversity conservation, each with different rates of N, P and K applications since 1992. Location B is an agricultural region dominated by livestock production and intensive grassland management for forage production with high rates of animal manure application.

The ten year mean annual temperature, precipitation and number of vegetation days of the locations were 5.8, 9.8 and 10.0 ◦ C, 1400, 1077 and 1200 mm and 184, 257 and 256 days year−1 for V, W and B, respectively. Vegetation days were defined as the number of days with daily mean temperature above 5 ◦ C. Climatic data were obtained from a nearby meteorological station run by MeteoSwiss, the Federal Office of Meteorology and Climatology (IDAWEB, 2009). Location V is a subalpine mesic grassland. The soil is a slightly acid Cambisol as defined by FAO (1998) which has a low base saturation (<50%). The soil texture in the top 10 cm is a silty loam (30% clay and 15% sand). According to Klötzli et al. (2010), the vegetation can be described as a Festuco-agrostietum association (species rich grassland with 30–50 angiosperm species). The generally low agronomic value of this sward is due to a short vegetation period resulting in low DM production (two harvests per year) and the late first harvest (in order to let plants flower and produce seeds) which produces low quality forage. However, its ecological value is high due to high plant species numbers and the presence of montane species. Location W is a well-drained mesic grassland habitat established on a medium deep (50–70 cm) Cambisol (FAO, 1998) with medium base saturation (50–80%). The soil texture in the top 10 cm is a sandy loam (22% clay and 44% sand). The vegetation is an Arrhenaterion-elatioris association with 30–40 angiosperm species. It is of relatively high agronomic and ecological value because of its relatively high DM yields and forage quality (of the 2nd and 3rd re-growths) combined with high plant species numbers (Huguenin-Elie et al., 2006a). The grasslands at B are mesic with high nutrient inputs and four to six harvests per year, mainly used for silage. The soils are Eutric Cambisols as defined by FAO (1998) with a high base saturation (60–95%). The texture ranges from loamy sand to silty loam (16–26% clay and 23–53% sand). We characterized the vegetation as a high yielding Trifolio-lolion association (Klötzli et al., 2010) with 10–15 angiosperm species. These grasslands are of high agronomic value because of their high biomass production with high forage quality produced by four to six harvests per season taken during vegetative growth. 2.2. Management and fertilizer treatments The two fertilizer trials in V and W were managed by research institutions (Agroscope Changins-Wädenswil and Agroscope Reckenholz-Tänikon, respectively). Farmers managed the on-farm grasslands in B. The fertilizer trials in V and W were established in 1992. The on-farm swards were fertilized and harvested similarly since over 20 years and in the case of N1P1K2org/min since five years. The swards in V, W and B were cut two, three and five times, respectively, in 2008 (Table 1). Each treatment had four replicate plots of 2 m × 5 m which were in V and W randomized in four blocks. On-farm in B, four plots of the same size were arranged as spatial replicates within the investigated grassland, with at least 5 m distance between plots. In one case in B (N1P1K2org/min), a fifth plot was installed because of local heterogeneity of botanical composition observed before the first harvest. The nutrient input of the treatments was based on the fertilizer recommendation by Flisch et al. (2009), with the recommended P rate designed to replace the P removal by the expected yield for a given management intensity and a given altitude. The following nomenclature was used: N, P and K stand for the three nutrients followed by 0, 0.5, 1 and 2 for no, less, equal and more fertilizer applied than recommended. The suffixes org and org/min stand for the use of slurry and combined mineral fertilizer and slurry application, respectively. In treatments without suffix only mineral fertilizers were applied. The recommended fertilization rates at level 1 range from 15 to 48 kg P ha−1 (Table 1) depending on the P removal by the

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

69

Table 1 Management, fertilizer treatments and recommended fertilization (Flisch et al., 2009) of the experimental grasslands at the investigated locations Les Verrières (V), Watt (W) and Baldegg (B) in 2008. The treatments are labelled using the following nomenclature: V, W or B for the location and N, P and K for the three nutrients followed by 0, 0.5, 1 and 2 for no, less, equal or more fertilizer than recommended. Management (altitude) and location

Treatment

Fertilizer

Harvests (cuts year−1 )

Yield expectation (t ha−1 year−1 )

Fertilization (kg ha−1 year−1 ) N

Medium intensity (1000–1500 m above sea level) Recommendation V-N1P0K0 Les Verrières V-N1P0.5K0.5 V-N1P1K1 V-N1P2K2 Low intensity (<700 m above sea level) Recommendation W-N0P0K0 Watt W-N1P0K1 W-N1P1K0 W-N1P1K1 W-N0P1K1 W-N1P1K2org Intensive (>∼700 m above sea level) Recommendation Baldegg B-N1P1K2org Intensive (<600 m above sea level) Recommendation B-N1P1K2org/min Baldegg B-N2P2K1org I B-N2P2K1org II a b c

P

K

2 2 2 2 2

5.0

Mineral Mineral Mineral Mineral

25 25 25 25 25

15 0 9 17 26

79 0 29 58 116

3 3 3 3 3 3 3

6.5

Mineral Mineral Mineral Mineral Mineral Dairy slurrya

40 0 45 45 45 0 55

17 0 0 17 17 17 17

79 0 83 0 83 83 102

5 5

11.5

Dairy slurrya , b

130–150 192

39 31

228 548

5–6 5 5 5

13.5

Dairya , b Pig slurrya , c Pig slurrya , c

150–180 48 144+27 24 210 116 210 116

280 411 256 256

Estimated according to Flisch et al. (2009) assuming 1:1 dilution of slurry by water. Assumed availability of 3.0, 0.8 and 6.6 kg m−3 representing 70, 100 and 100% of the total content for N, P and K respectively (Flisch et al., 2009). Assumed availability of 4.2, 1.7 and 4.3 kg m−3 representing 70, 100 and 100% of the total content for N, P and K respectively (Flisch et al., 2009).

expected yield. The expected yields (Table 1) of Flisch et al. (2009) are based on data records of the Swiss federal research stations obtained on farm and in trials. The expected yields in V and W do not represent the maximum potential yields but yields that can be expected under reduced management intensity in order to sustain high species diversity. In V, four treatments with increasing mineral P and K applications and similar N supply were chosen from a randomized split block experiment. Fertilizers were applied as ammonium nitrate after the first harvest and as super-phosphate and potassium chloride in spring or in autumn after the last harvest. In W, five mineral fertilizer treatments from a randomized block experiment (Huguenin-Elie et al., 2006b) and an organically fertilized treatment from another experiment at the same location were chosen. Fertilizers were applied as ammonium nitrate (split in three applications: in spring, after the first and after the second harvest), superphosphate and potassium chloride (both applied in spring). In the organically fertilized treatment, dairy slurry was applied in spring and after the first harvest. In B, two grasslands were fertilized as recommended by Flisch et al. (2009) and two received more P than recommended (Table 1). Dairy or pig slurry produced on the respective farm was usually broadcast applied at the onset of the vegetation in spring and after each harvest. The amount of available nutrients applied with the slurries (Table 1) was estimated according to slurry type, number of applications and animal husbandry (Flisch et al., 2009). In the year of study (2008), there was no application of slurry at the onset of the vegetation on the investigated grasslands in B. In N1P1K2org/min no slurry was applied after the second harvest, but 27 kg N ha−1 were supplied as ammonium nitrate. 2.3. Basic soil characteristics We sampled the 0–5 cm topsoil layer after the first regular harvest to analyse P availability. We took twelve soil cores with an inner diameter of 2.5 cm distributed over each experimental plot and mixed them into one composite sample per plot. Soil P availability was determined on air-dried soils using a Swiss reference

method (FAL, 1996) which extracts P with CO2 saturated water (PCO2 , Table 2). In the extract P was measured by malachite green colorimetry (Ohno and Zibilske, 1991). The treatments were classified by P availability classes as suggested in Flisch et al. (2009). Total P was extracted from dried samples after incineration (550 ◦ C for 8 h) with 0.5 M H2 SO4 at room temperature. P was measured by colorimetry as described above. Total N and carbon (C) were analysed on dried and pulverized soil samples using a NCS elemental analyser (Flash EA 1112 Series NCS analyser, Thermo Electron Corporation, Waltham, USA). 2.4. Botanical composition The botanical composition was recorded in May 2007 in W and in April and June 2008 in B and V, respectively. A species inventory was made for each plot and species abundance was estimated using classes of yield proportion and visual estimation of soil cover according to Dietl (1995). The three grassland habitats are typical anthropogenically formed grasslands with characteristic vegetation (Table 3) under the respective site properties, climate conditions and management. A detailed species inventory is provided as supplementary information. 2.5. Plant sampling and analysis We sampled plant biomass at a defined vegetative growth stage when plants have a high nutrient demand. The growth stage was determined by stem elongation to emergence of the inflorescence described as stage two to three in Jeangros et al. (2001) of Orchard grass (Dactylis glomerata) which was present at all locations. Sampling generally took place 1–4 days prior to regular harvests, except for the first sampling in V and W which was a few weeks before the first regular harvest to get plant material at the appropriate growth stage. The DM yield was measured at the regular harvest time. For the sampling a randomly selected area of 50 cm × 50 cm with at least 50 cm distance to the plot border was cut 4 cm above the ground using electric scissors. Sampling on the same spot during

70

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

Table 2 Selected soil properties in the treatments at the investigated locations. The clay, silt and sand contents of the soils were determined in the 0–10 cm soil layer. The pH, total carbon (Ctot), nitrogen (Ntot), phosphorus (Ptot) and available phosphorus extracted with CO2 saturated water (PCO2 ) were determined in the 0–5 cm soil layer. Location

Treatment

pHa

Ctotb (g kg−1 )

Ntotb (g kg−1 )

PCO2 c (mg kg−1 )

Ptotd (mg kg−1 )

Les Verrières (V)

N1P0K0 N1P0.5K0.5 N1P1K1 N1P2K2 N0P0K0 N1P0K1 N1P1K0 N1P1K1 N0P1K1 N1P1K2org N1P1K2org N1P1K2org/min N2P2K1org I N2P2K1org II

6.0 5.9 5.8 5.5 6.1 5.5 6.0 5.8 5.7 6.8 6.8 6.5 6.9 7.5

39.3 39.1 36.3 35.7 28.5 29.2 28.8 30.6 29.6 38.8 40.1 32.4 40.5 59.8

4.0 4.1 3.9 3.8 3.0 3.0 3.1 3.1 3.1 4.2 4.1 3.4 4.2 4.9

0.20 0.25 0.39 0.55 0.37 0.34 1.26 0.84 0.86 0.96 4.60 1.77 16.70 10.70

565 688 732 841 580 555 679 658 730 713 1574 1609 2103 2505

Watt (W)

Baldegg (B)

a b c d

In H2 O, ratio soil:H2 O = 1:2.5. Measured by NCS elemental analyser (Thermo Electron Corporation, Waltham, USA). According to Swiss reference methods in (FAL, 1996). Extraction with 0.5 M H2 SO4 after incineration.

Table 3 Selected parameters for botanical composition of the investigated treatments. Proportion (in %)a

Dominant speciesb

Location

Treatment

Forbs

Total

Grasses

Legumes

Forbs

Les Verrières

N1P0K0

72

8

20

31

6

4

21

N1P0.5K0.5

69

14

17

32

7

5

20

N1P1K1

61

21

18

31

7

5

20

N1P2K2

60

19

21

31

7

5

19

N0P0K0

51

11

37

32

9

5

17

N1P0K1

66

6

28

29

8

5

17

N1P1K0

65

2

33

30

9

5

17

N1P1K1

56

14

30

32

10

6

16

N0P1K1

53

27

20

31

9

6

16

N1P1K2org

72

7

20

28

9

5

15

N1P1K2org

51

11

38

15

5

1

8

N1P1K2org/min

59

3

38

15

5

2

8

N2P2K1org I

78

14

7

10

4

1

6

N2P2K1org II

87

5

8

10

4

1

6

Grasses

Watt

Baldegg

Legumes

Species number

Festuca rubra, Anthoxanthum odoratum, Agrostis capillaris, Pimpinella major, Ranunculus ssp., Rumex acetosa, Taraxacum officinale Festuca rubra, Anthoxanthum odoratum, Agrostis capillaris, Dactylis glomerata, Lathyrus pratensis, Ranunculus ssp., Rumex acetosa Festuca rubra, Anthoxanthum odoratum, Agrostis capillaris, Dactylis glomerata, Lathyrus pratensis, Ranunculus ssp., Rumex acetosa Festuca rubra, Anthoxanthum odoratum, Dactylis glomerata, Lathyrus pratensis, Ranunculus ssp., Rumex acetosa Holcus lanatus, Arrhenatherum elatius, Anthoxanthum odoratum, Dactylis glomerata, Festuca rubra, Galium album Holcus lanatus, Arrhenatherum elatius, Anthoxanthum odoratum, Dactylis glomerata, Festuca rubra, Galium album Holcus lanatus, Arrhenatherum elatius, Anthoxanthum odoratum, Dactylis glomerata, Trisetum flavescens, Galium album Holcus lanatus, Arrhenatherum elatius, Anthoxanthum odoratum, Dactylis glomerata, Trisetum flavescens, Galium album Holcus lanatus, Arrhenatherum elatius, Anthoxanthum odoratum, Trisetum flavescens, Trifolium pratense, Galium album Holcus lanatus, Arrhenatherum elatius, Anthoxanthum odoratum, Trisetum flavescens, Trifolium pratense, Galium album Lolium perenne, Taraxacum officinale, Trifolium repens, Alopecurus pratensis Lolium perenne, Festuca rubra, Taraxacum officinale, Lolium multiflorum, Poa pratensis Lolium perenne, Taraxacum officinale, Trifolium repens Lolium perenne, Trifolium repens, Poa pratensis

a

Calculated as the annual average dry matter of all harvests. b Dominant species were derived from visual estimation prior to the first harvest. The proportion of each dominant species was higher than 5% of ground cover and the sum of the dominant species shown covered between 60 and 98%.

later harvests was avoided. In W the biomass sampling was at the same level as and well correlated to whole plot harvests (r2 = 93) in 2007 and 2008, thus we considered the sampling method applicable for V and B as well. Immediately after cutting, the plant material was separated into the three botanical fractions grasses, legumes and forbs. We weighed the plant material after drying at 60 ◦ C for

three days. The DM of the three fractions and consecutive harvests was summed for total DM yield. Samples were milled (particle size ∼1 mm) using a cutting mill (Retsch GmbH, Germany). Subsequently, we pulverized a subsample using a ball mill (Retsch GmbH, Germany). For P and K analysis, we incinerated milled samples at 550 ◦ C for 8 h and solubilized the ashes in 15 M nitric acid at room

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

temperature. The concentrations were measured with an ICP-MS (Agilent, USA). The N concentration was measured by dry combustion with a NCS elemental analyser (Flash EA 1112 Series NCS analyser, Thermo Electron Corporation, Waltham, USA). We report the nutrient concentrations on a DM basis for the three botanical fractions grasses, legumes and forbs. 2.6. Calculations We converted measured total DM yields into relative yields using the expected yields of Flisch et al. (2009) (Table 1). By taking them as 100% reference, we assume that they represent the optimum yield for a given location and management. For the interpretation of the P nutrition indicators, we considered that indicator values related to relative yields ranging from 90 to 110% indicated sufficient P supply. Bulk nutrient concentrations were calculated as weighted means as shown for NCbulk in Eq. (1): NCbulk =

(NCgrasses × DMgrasses + NClegumes × DMlegumes + NCforbs × DMforbs ) (DMgrasses + DMlegumes + DMforbs )

(1)

where NC and DM are the nutrient concentrations (in mg g−1 ) and the dry matter (in g), respectively. The P concentrations were multiplied with yield for each harvest and fraction separately and summed as total P removal. The P removal was subtracted from the P fertilizer input to obtain the annual P balance. The P content in the slurries was estimated according to Flisch et al. (2009) assuming 1:1 dilution of slurry by water. Nutrient ratios N:P and K:P were calculated using the measured concentrations for fractions and the calculated bulk concentrations, respectively. The PNI was calculated according to the formula proposed by Duru and Thélier-Huché (1997): PNI =

100 × Pm 0.065 × Nm + 1.5

71

applied the Holm adjustment for p-values for multiple comparisons. Prior to analysis, we transformed percentage data of relative yields to their arc sin. Correlation coefficients () and their level of significance were calculated by the Pearson product-moment correlation. Trendlines were fitted to graphs in SigmaPlot for Windows version 11.0 (Systat Software Inc., 2008) using non-linear regression. We used a polynomial-linear, exponential rise to maximum and Gaussian peak procedure to fit the linear, exponential and peak trendlines, respectively. The coefficient of determination (r2 ) was derived for the used function. Significance levels are: ns. = not significant, *p < 0.05, **p < 0.01 and ***p < 0.001. 3. Results 3.1. Dry matter yield Mean total DM yields ranged from 4.3 to 14.8 t ha−1 year−1 (Fig. 1). The average yields of P1 treatments (V 5.2 t ha−1 , W 6.7 t ha−1 , and B 13.6 t ha−1 ) were close to the expected yields (Table 1). In general, differences were more pronounced between management intensities and thus location than between treatments within each location. Mean relative yields ranged from 59% in W-N0P0K0 to 109% in B-N1P1K2org/min (LSD = 23%). We observed significant effects of fertilizer treatments within the locations V and W. Treatments fertilized as or higher than recommended were close to the expected yields. A significant yield reduction was found in treatments, in which one or more nutrients had been omitted, with a reduction in relative yield in V-N1P0K0, W-N1P1K0 and W-N1P0K1 by 16, 22 and 19%, respectively. The most severe reduction by about 40% occurred in W-N0P0K0 that

(2)

where Pm and Nm are the measured P and N concentrations in mg g−1 and the PNI is unitless. Although the critical curve has been established for grass alone and grass dominated swards we used Eq. (2) to calculate the bulk sward PNI (PNIbulk ) in order to test the implications of the botanical composition. For PNIbulk we used the weighted mean Nmbulk and Pmbulk concentration as calculated by Eq. (1). The PNI in non-leguminous plants (PNInonleg ) was calculated with Eq. (2), using the weighted mean Nmnonleg and Pmnonleg concentrations of grasses and forbs calculated as shown for Nmnonleg in Eq. (3). Nmnonleg =

(NCgrasses × DMgrasses + NCforbs × DMforbs ) (DMgrasses + DMforbs )

(3)

For derivation of lower and upper thresholds for P concentrations and PNI we used the range between 90 and 100% relative yield and values were rounded to one decimal place and to the decade, respectively. For N:P and K:P we used the range reflected by treatments with at least 90% relative yield and values were rounded to half or full integer. 2.7. Statistical analysis Statistical analyses were performed with R version 2.8.1 (R-Development-Core-Team, 2008). We analysed differences in nutrient concentrations between botanical fractions and harvests for individual treatments using a two factorial ANOVA followed by comparison of mean values using the Student–Newman–Keuls test (SNK). A three way ANOVA including fertilizer treatment as a factor was only applicable for V and the mineral fertilized treatments in W, respectively. If the F-test was significant at p < 0.001, a test for least significant difference (LSD) with an alpha of 0.05 was conducted using the library agricolae v. 1.09 (Mendiburu de, 2009). We always

Fig. 1. Mean annual input-removal P balance at the sward level and dry matter yield for the botanical fractions grasses, legumes and forbs in 2008 as affected by management (location) and fertilizer treatment. Error bars indicate standard deviation. Grey horizontal lines show the expected yield for each management. Different letters indicate significant differences of the P balance and total annual yield within a location (LSD test at alpha = 0.05, with Holm adjusted p values).

72

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

Fig. 2. Left: Effect of legume proportion on the PNI in bulk biomass (PNIbulk ), shown as the difference between PNIbulk and PNI in non-leguminous plants (PNInonleg ). Right: Effect of forb proportion on the PNI in non-leguminous samples, shown as the difference between PNInonleg and PNI in grasses (PNIgrasses ). The Pearson correlation coefficient () in legumes and in forbs is shown for all samples.

received no fertilizer at all for 16 years. In B, we observed no effects of fertilization on relative yield. 3.2. Soil P availability and P balances In general, the PCO2 extracted from soils increased in the order V < W < B (Table 2). Within each location, PCO2 increased with P fertilizer application level. Flisch et al. (2009) interpret PCO2 using the five soil P status classes very low, low, adequate, high and enriched in P. Applying this interpretation, all treatments of location V had a low soil P status except V1P2K2 which was adequately P supplied. In W, the P fertilized treatments were classified as adequate whereas P availability was low in P0 treatments. In B, the soil P status of B-N1P1K2org/min was classified as high and the other three treatments as P enriched. The estimated annual P balance in 2008 (Fig. 1) ranged from a deficit of −14 in BN1P1K2org/min to a surplus of +85 kg ha−1 year−1 in B-N2P2K1org I. Except B-N1P1K2org/min, the P balance of each P1 treatment was not significantly different from zero. In contrast, P treatments without or with P fertilization higher than recommended were significantly lower or higher than zero, respectively. 3.3. Botanical composition The botanical fractions grasses, legumes and forbs were composed of different species and species numbers, depending on the location (Table 3), with similar total species number at W and V. Grasses represented the main fraction in all grasslands and fertilizer treatments (Fig. 1). The highest grass proportions were observed in B-N2P2K1org I + II (78 and 87%) and V-N1P0K0 and W-N1P1K2org (both 72%). The highest proportions of legumes were found in the fertilizer treatments V-N1P1K1, V-N1P2K2 and W-N0P1K1 with balanced or high P and K application.

legumes showed significantly higher P concentrations than grasses (non-significant trend for forbs in V-N1P0K0). Under P1 fertilization, P concentrations tended to be lower in legumes than in grasses and forbs. As expected, the N concentrations in legumes were always higher than in grasses and forbs. The N concentrations of grasses and forbs were usually similar. The K concentrations were generally highest in grasses (except in V-N1P0K0, B-N1P1K2org and B-N2P2K1org I) and lowest in legumes. The coefficient of variation of P and K concentration within each treatment was lower in grasses than in legumes and forbs (Table 4). The coefficients of variation of P concentration for the total dataset were 30%, 35% and 21% for grasses, forbs and legumes, respectively. The differences in P, N and K concentration in the botanical fraction were reflected in the N:P and K:P ratios (data not shown). The N:P ratios were significantly higher in legumes than in the other two fractions, with ranges of mean N:P of treatments from 10.0 to 16.5 in legumes compared to 5.5 to 11.5 and 5.0 to 10.0 in grasses and forbs, respectively. For K:P, the differences were less pronounced, but K:P tended to be lower in legumes (2.0–8.0) than in grasses (4.0–12.0), while it was intermediate in forbs (3.5–10.0). The significant change in PNI caused by legumes and forbs is shown in Fig. 2. The presence of legumes decreased the PNIbulk by a value equivalent to about half of the legume proportion. For example, a legume fraction of 40% in the sward DM lowered the PNI by 20. Additionally, we observed a significant effect of the forb fraction to increase PNInonleg . The difference between PNInonleg and PNIgrasses reflected a high variation of PNInonleg caused by the presence of forbs. The changes of PNInonleg ranged from −8 to 35. Since grasses are the most important yield component in the investigated grasslands we will in the following detailed comparison of nutrient indicators only present results on the grass fraction. 3.5. Nutrient indicators in grasses

3.4. Nutrient concentrations in botanical fractions Fertilization with P, N or K increased the concentrations of the particular nutrient in the aboveground plant biomass of all botanical fractions (Table 4). In general, N concentrations were less variable over time and between fractions than P and K concentrations which differed significantly between fractions (p < 0.0001). Different fertilizer effects on P concentrations in the botanical fractions were indicated by a significant interaction (p < 0.0001) between treatment and botanical fraction. When P supply was omitted (W-N0P0K0, W-N1P0K1 and V-N1P0K0), forbs and

Phosphorus concentrations in grasses were significantly affected by locations (p < 0.0001) and treatments (p < 0.001, Table 4). Generally, B had higher P concentrations than V and W. Concentrations of the treatments V-N1P0K0, W-N0P0K0 and WN1P0K1 were 1.7 mg g−1 on average. In contrast, P input at level 2 in V-N1P2K2 and B-N2P2K1org I + II did not result in significantly higher concentrations than in the treatments P1 fertilized as recommended. Mean N concentrations in grasses were generally higher in B than in V and W (Table 4), but we observed no treatment differences within a location (except for W-N0P0K0 and V-N1P0K0 which

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

73

Table 4 Mean N, P and K concentrations (g kg−1 ) and coefficients of variation (CV) in the aboveground biomass of the botanical fractions grasses (G), legumes (L) and forbs (F). The total number of replicates (n) consists of four spatial replicates in the field (except N1P1K1org/min: five replicates) and two, three and five harvests in 2008 in Les Verrières, Watt and Baldegg, respectively. Absence of legumes or forbs reduced the number of replicates in some cases. Different letters indicate significant differences between the botanical fractions within a treatment according to a SNK test at p < 0.01. Location

Les Verrières

Treatment

N1P0K0

N1P0.5K0.5

N1P1K1

N1P2K2

Watt

N0P0K0

N1P0K1

N1P1K0

N1P1K1

N0P1K1

N1P1K2org

Baldegg

N1P1K2org

N1P1K2org/min

N2P2K1org I

N2P2K1org II

Botanical fraction

G L F G L F G L F G L F G L F G L F G L F G L F G L F G L F G L F G L F G L F G L F

Nitrogen

Phosphorus

Potassium

n

Mean (g kg−1 )

p < 0.01

CV%

n

Mean (g kg−1 )

p < 0.01

CV%

n

Mean (g kg−1 )

p < 0.01

CV%

8 8 8 8 8 8 8 8 8 8 8 8 12 12 12 12 10 12 12 5 12 12 12 12 12 12 12 12 12 12 20 19 20 25 18 25 20 16 13 20 13 15

19.4 36.2 19.2 19.9 35.1 19.3 20.7 37.0 19.6 19.6 35.9 19.3 17.7 29.5 17.9 20.1 37.3 19.5 20.7 33.3 20.8 19.6 31.3 20.4 20.0 28.7 19.9 21.2 32.5 21.0 25.1 42.1 25.2 28.1 37.9 30.5 29.0 39.2 32.4 28.7 34.5 30.1

b a b b a b b a b b a b b a b b a b b a b b a b b a b b a b b a b c a b b a b b a b

8 8 12 13 1 14 13 1 13 12 1 15 17 18 2 16 8 4 7 14 13 13 7 3 5 21 2 3 18 9 29 12 26 17 17 19 31 20 23 18 19 14

8 8 8 8 8 8 8 8 8 8 8 8 12 12 12 12 11 12 12 12 12 12 12 12 12 12 12 12 12 12 20 19 20 25 19 25 20 16 13 20 13 15

1.7 2.4 1.8 2.6 2.7 2.8 3.3 3.1 3.2 3.5 3.2 3.4 1.7 2.3 2.0 1.7 1.9 2.1 3.1 3.5 3.4 3.2 2.9 3.2 3.4 2.8 3.3 2.4 2.5 2.5 4.1 3.9 5.0 3.9 3.7 4.4 4.5 3.4 4.8 4.3 3.1 4.9

b a b a a a a a a a a a c a b b a a a a a a b a a b a a a a b b a b b a a b a a b a

20 53 52 9 51 52 12 52 51 14 52 52 20 51 45 23 44 46 19 47 44 26 46 45 23 46 44 26 47 45 14 18 24 15 21 11 17 12 5 12 13 16

8 8 8 8 8 8 8 8 8 8 8 8 12 12 12 12 11 12 12 12 12 12 12 12 12 12 12 12 12 12 20 19 20 25 19 25 20 16 13 20 13 15

12.0 7.9 9.9 15.7 9.4 14.7 19.0 11.9 16.1 22.8 20.7 24.7 12.9 8.8 10.5 21.3 12.7 18.2 13.0 7.5 11.9 21.2 11.8 16.0 20.9 12.2 16.6 25.0 15.6 20.4 30.9 24.8 35.0 29.9 25.4 31.5 29.8 28.4 37.3 29.2 20.6 28.8

a b ab a b a a c b ab b a a b b a b a a b a a c b a c b a c b a b a ab b a ab b a a b a

14 52 56 10 51 52 16 53 53 7 54 54 17 50 44 16 52 47 22 57 50 24 64 50 22 59 49 20 51 45 12 12 15 9 7 10 28 17 21 17 41 26

were significantly lower). Similar to P and N, K concentrations were higher in B than in V and W (Table 4) and K concentrations were significantly higher in K fertilized than in unfertilized treatments. Concentrations of K in the K-over fertilized treatments (V-N1P2K2, B-N1P1K2org and B-N1P1K2org/min) were not significantly higher than in the treatments where K was applied as recommended. The mean N:P ratios in grasses in all P0 treatments were higher than 9 (Fig. 3). They were lower than 7.5 in all treatments where P was applied except in W-N1P1K2org. The minimum and maximum mean K:P ratios were found in treatments W-N1P1K0 and W-N1P0K1, respectively, where K or P were omitted. We found no significant difference in K:P ratios between treatments in which both P and K were fertilized as recommended and treatments with surplus fertilization of P or K or both. For the PNI we first verified the linear relationship of P to N proposed by Duru and Thélier-Huché (1997). To this end, we related N and P concentrations (in mg g−1 ) of the grass samples from all treatments with sufficient K input for each of the three grassland types separately. Treatments with K concentrations above 20 mg g−1 in grasses were considered to be sufficiently K supplied (Bailey et al., 1997a,b). The slopes and intercepts of the resulting

trend lines were not significantly different. The combination of all data points resulted in the function: P = 0.07 × N + 2.0 (r2 = 0.61, p < 0.0001, n = 130, P and N concentrations in mg g−1 ). Neither the slope nor the intercept were significantly different from the function proposed by Duru and Thélier-Huché (1997) which we used to calculate the PNI. The mean PNI in grasses ranged from 62 to 137 (Fig. 4). The PNI of P0 treatments was less than 80 and significantly lower than for treatments with P application. Fertilization at the recommended rate resulted in a mean PNI between 80 and 120, with the exception of B-N1P1K2org which had a PNI of 132. The P2 treatments resulted in mean PNI values above 120 which were however not significantly different from the PNI obtained for P1 treatments. 3.6. Phosphorus indicators in relation to relative yield The relationship of P concentrations and PNI of grasses to relative yield showed a “Mitscherlich” type of function (Mitscherlich, 1928), with strong increase at low indicator values and no change of relative yield at high values (Fig. 4). We found no significant yield

74

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

Fig. 3. Mean N:P and K:P ratios in aboveground biomass of grasses sampled in vegetative growth stage as affected by P fertilization (white, grey and black for P0, P1 and P2, respectively) and management (location = symbol) and their relationship to relative annual yield. Horizontal lines represent 100% (solid) and 90% (dotted) of relative yield, respectively. Dashed vertical grey lines represent the critical ratios for K:P and N:P ratios in our study. Error bars indicate standard deviation. Treatment differences are shown as least significant differences according to a LSD test with Holm adjusted p values (alpha = 0.05). For the relationship to relative yield the Pearson correlation ˆ coefficient () is shown. Trendlines were fitted as Gaussian peak (3 parameters, y = a × e(−0.5×((x−x0)/b)2)) , excluding the labelled treatments.

increase above a P concentration of 2.1 mg P g−1 and for a PNI equal to or higher than 80. The relationship of nutrient ratios to relative yield was reflected by a Gaussian curve. In the case of the N:P ratio, a linear relationship could also be applied but had a lower coefficient of determination than the Gaussian relationship. A N:P ratio above 9.0 was related to yield reduction (Fig. 3). For K:P ratios we found yield limiting treatments over the whole range. However, the treatments with relative yields above 90% always had mean K:P ratios between 6.0 (V-N1P2K2) and 10.5 (W-N1P1K2org). The correlation coefficients between relative yield and N:P ratio or PNI were higher ( > 0.80) than between relative yield and P concentration ( = 0.64). 4. Discussion 4.1. Yields and soil phosphorus availability reflect long term fertilization The DM yields in treatments fertilized as recommended or higher were close (±10%) to the yield expected for the respective management and altitude (Flisch et al., 2009). Relative yields of less than 90% of the expected yield illustrated insufficient P and K supply in treatments where P, K or both nutrients were omitted. Yield responses reflected the long term fertilization regime, the available soil PCO2 (see supplementary material) and the annual P balance. The higher yield of W-N1P1K1 than W-N1P0K1 demonstrated the limitation of sward yield by P alone. In B, no treatment had a yield lower than expected because there was no P0 or P0.5 treatment. Because DM yields of the P1 treatments without colimitation met the reference yields given by Flisch et al. (2009) at the three locations, we consider P1 treatments as sufficiently P supplied. Consequently, P0 and P2 treatments would represent P deficiency and a P surplus status, respectively. 4.2. Botanical composition reflects long term fertilization regime

Fig. 4. Mean P concentration (mg g−1 ) and PNI in aboveground biomass of grasses sampled in vegetative growth stage as affected by P fertilization (white, grey and black for P0, P1 and P2, respectively) and management (location = symbol) and their relationship to relative annual yield. The PNI was calculated as proposed by Duru and Thélier-Huché (1997). Horizontal lines represent 100% (solid) and 90% (dotted) of relative yield, respectively. Dashed vertical grey lines represent the critical P concentrations of 2.1 and 3.0 mg g−1 and PNIs of 80 and 120, respectively. Error bars indicate standard deviation. Treatment differences are shown as least significant differences according to a LSD test with Holm adjusted p values (alpha = 0.05). For the relationship to relative yield the Pearson correlation coefficient () is shown. Trendlines were fitted as exponential rise to maximum (3 parameters, y = y0 + a × (1 − e(−b×x) )), excluding the labelled treatments.

Grassland management at locations V and W combined production targets with biodiversity management. Species numbers especially of forbs and legumes were higher than at B where the purpose was production of high quality forage with P input rates which should for ecological reasons not exceed the P removal by harvested forage. The different management intensity applied to W and B, i.e. locations with similar altitude, affected the grass species composition, with dominance of high yielding ryegrass at B. At each location and with each treatment, grasses constituted the main botanical fraction. The grass fractions were highest in treatments with high N fertilization, either absolutely or in relation to P and K. Many authors reported that increased N supply fosters grasses

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

(Høgh-Jensen and Schjoerring, 2010; Ledgard et al., 1996; Jeangros, 1993). Inversely, P and K fertilization without N application decreased the grass and increased the legume fraction confirming earlier findings of Thöni (1964, 1982). The absence of P and/or K fertilization, however, reduced the legume fraction. For forbs, our findings were not consistent between treatments and locations, which might be due the large variability of the reactions of individual forb species to different fertilizer regimes (Whitehead, 2000). 4.3. Botanical fractions affect interpretation of sward phosphorus nutrition status Mineral nutrient concentrations in grassland plants change during growth (Thöni, 1964; Whitehead, 2000; Daccord et al., 2001). By sampling at a defined vegetative growth stage of grasses we limited differences in nutrient concentration induced by growth stage. However, legumes and forbs may have been at different growth stages at the different harvests. This may be one reason why the within treatment variation of P concentration in legumes and forbs was higher than in the grass fraction. In contrast, the coefficient of variation of P concentration was lowest for the legumes when taking the complete dataset, suggesting that P concentrations of legumes were less affected by P fertilization and season than P concentrations of grasses and forbs. Nevertheless, the differences of P, N and K concentrations between harvests were analogous in the three botanical fractions (data not shown). Therefore, they could largely be ascribed to timing of fertilizer application and seasonal differences of nutrient availability. We observed substantial differences of P, N and K concentrations between botanical fractions, as previously reported by Whitehead (2000). While previous works reported higher P concentrations in legumes than in grasses under limiting conditions for P (Gallet et al., 2003) and N (Jouany et al., 2005), we confirmed this observation for treatments without severe N deficiency, i.e. which had yields of at least 90% of the expected yield. For PNI in bulk sward samples we observed a negative relationship to the total legume proportion causing an underestimation of sward P status. Jouany et al. (2005) suggested a correction of the bulk sward PNI for sward legume proportion. In our study correction by the addition of half of the total legume proportion (PNIcorrected = PNIbulk + 0.5 × %legumes) was appropriate. Similarly, N:P ratios in bulk samples underestimated the P nutrition status in presence of legumes due to the higher N concentrations (data not shown). Additionally, we observed a highly variable and generally positive effect of forbs on the PNI when the non-leguminous fraction was used. Although relatively small, this effect was significant and could lead to an overestimation of P nutrition status when high proportions of forbs are present. Such an overestimation could be critical when the P nutrition status is close to deficient. Since grasses are the most important yield component, had lower variation of P concentration than legumes and forbs within a given treatment and reliably reflected the P availability in grasslands, we recommend the PNI or element ratio to be determined on the grass fraction sampled at a defined vegetative stage. 4.4. Interpretation of the phosphorus indicators in the grass fraction 4.4.1. Critical phosphorus concentrations The P concentration of 2.1 mg g−1 in grasses above which we found no significant yield increase fell into the range of 2.0–6.0 mg P g−1 proposed by Whitehead (2000) as sufficient for perennial ryegrass. However, it was below the critical concentration of 2.6 mg g−1 suggested for ryegrass by Bailey et al. (1997b). In our study of a gradient of management intensity and different dominant grass species, the same critical P concentration seems

75

to apply as critical P value for grasslands managed for high biodiversity such as V and W and for grasslands intensively managed for high biomass production such as B. This is in line with Hill et al. (2005) who report more pronounced differences among temperate grassland species for external (i.e. available soil P status) than internal (i.e. P concentration in plant tissue) P requirement. However, the range of ±10% of expected yield that we considered as satisfactory might not be appropriate to all grassland management purposes. If 10% yield decreases cannot be accepted, a critical value of 2.4 or 3.0 mg P g−1 corresponding to 95 and 100% of relative yield, respectively, could be used. Further, the interpretation of nutrient concentrations in relation to forage quality differs from that for yield response. For example, roughage should have a minimum P concentration of 3.0 mg g−1 to cover the requirements of dairy animals (Arrigo, 1999). We restricted the interpretation of P to its function as an essential element potentially limiting grassland biomass production. However, P concentrations of 3.0 mg g−1 were reached by recommended P fertilization at each location and the maximum of the response curve in our study was observed at a P concentration of 3.5 mg g−1 . Phosphorus concentrations higher than 3.0 mg g−1 were not related to significantly higher yields. Such luxurious consumption thus indicates inefficient use of fertilizer P. 4.4.2. Nutrient ratios The mean N:P and K:P ratios in treatments yielding more than 90% of the expected yield always covered a range of values because the decrease in nutrient concentration with increasing DM is nutrient specific (Duru and Thélier-Huché, 1997). The optimal range without significant yield limitation caused by deficiency of one or the other nutrient is delimited by a lower and a higher critical ratio. For both N:P and K:P, the higher critical ratio can be used as a threshold to identify P deficiency. Although the fitted Gaussian function reflects the relationship of the two nutrient ratios to relative yield it was hampered by an imbalanced number or missing treatments. Especially the N:P ratio where the lower intersection of the fitted function meets the 90% relative yield line could not be confirmed by experimental treatments. Thus, we used the experimental values of the highest and the lowest mean ratios of the treatments with corresponding relative yields over 90% to define the optimal range. The optimal N:P range of 5.5–9.0 included the norm N:P ratio of 9.0 from Bailey et al. (1997a) reported for ryegrass, but was lower than the critical ratios of 14.5 reported by Venterink et al. (2003) and of 14 and 16 reported by Koerselman and Meuleman (1996) for mixed species samples of grasslands managed at low intensity. The norm N:P ratio suggested for ryegrass by Bailey et al. (1997a) separated all P fertilized treatments from the unfertilized treatments, which qualified it as a critical N:P ratio to identify P deficiency in our study. It was also close to the intersection of the fitted curve with the line of 90% relative yield which is at 9.5. For N, we could not rule out a critical ratio because no exclusively N limited treatment was included in the experiment. Potentially N limited treatments identified by N concentrations lower than 20 mg g−1 (Bailey et al., 1997b) were co limited by P and or had relative yields above 90% indicating that the critical N:P ratio for N deficiency might be equal or lower than 5.5. For K:P, the peak shape of the fitted curve reflected the yield decreases caused exclusively by P or K deficiency. Between 6.0 and 10.5, treatments were either not limited, or co-limited by P and K. The norm ratio of 8.8 identified by Bailey et al. (1997a) was higher than the mean K:P (7.3) and the maximum K:P (8.1) of the fitted curve in our experiment, but was still within the optimal range suggested by the treatments with relative yields above 90%. Both N:P and K:P were not suitable to rule out the treatments over-fertilized by both nutrients, indicating that these ratios cannot be used to identify luxurious P nutrition. Likewise, they could

76

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77

not be applied to identify a P deficiency when both nutrients were deficient, as shown by treatments W-N0P0K0 and V-N1P0K0 which fell into the optimal range. While a large variation of critical N:P (and K:P) ratios has been reported for semi-natural grasslands (Koerselman and Meuleman, 1996; Fujita et al., 2010), we found the same critical N:P and K:P ratios to apply to all three studied grassland types, in spite that they were dominated by different grass species and were managed at different intensities. This highlights the importance of sampling at a defined vegetative growth stage, as usually done for agricultural grasslands (Dampney, 1992; Whitehead, 2000). 4.4.3. Phosphorus nutrition index In our study we found only non-significant deviations from the equation of Duru and Thélier-Huché (1997) in the P to N relationships of the grass fractions within the locations and the combined data for all locations. Showing that the equation from Duru and Thélier-Huché (1997) is applicable to the grass samples from different management intensities, i.e. agricultural grassland composed by different grass species, provided that samples are taken at same growth stages of the grasses. A PNI of 80 corresponded to relative yields of 90% which we considered satisfactory. However, a latent N deficiency was suggested by the relatively low N concentrations in the mineral fertilized treatments in W. This was not in conflict with the management purpose of the grassland at this location. A PNI higher than 120 was related to yields equal to or higher than expected. Further, PNI correctly expressed the expected P nutrition status for the respective P balance. Only in B-N1P1K2org a luxurious P nutrition was indicated by the PNI which was not reflected by the annual P balance. Possibly, we underestimated the P input with slurry in the B-N1P1K2org treatment because some pigs were also raised on that farm. Pig slurry contains more P than dairy slurry (Flisch et al., 2009) and both slurry types were mixed in the storage tank. This assumption is supported by the high PCO2 found in the soil under this treatment. The soil P status of the similarly managed treatment in B-N1P1K2org/min was lower, which might result from the higher yield and N application. Overall, our results confirm the applicability of the PNI as a P nutrition indicator in grasslands managed for different purposes when grass samples in the vegetative stage are used. 5. Conclusions The proportion of the grasses, legumes and forbs on DM in agricultural grasslands managed at different intensities for different purposes was highly variable and affected the interpretation of plant nutrition indicators. The analysis of the grass fraction sampled at vegetative growth stage improved the accuracy of the evaluation of P nutrition status of the grasslands and a common interpretation could be derived in spite that the grass fraction of the different locations/intensities was composed by different grass species. All P indicators, allowed to distinguish between P deficient and P fertilized treatments when other nutrients were not limited. Phosphorus concentrations from 2.1 to 3.0 mg g−1 could be interpreted as sufficient for plant growth and optimal for efficient P use in agricultural grasslands. However, the two P indicators integrating N (N:P and PNI) correlated better to relative yield than P concentration. For N:P and K:P ratios we deduced optimal ranges of 5.5–9.0 and 6.0–10.5, respectively. Both ratios failed to show distinct P, N or K limitation in case of co-limitation and could not identify luxurious P nutrition. The consideration of nutrient specific indicators in addition to P indicators is needed to reveal co-limitation correctly. The PNI resulted in a clear treatment separation, allowing differentiation between deficient, sufficient and surplus P fertilization. The P indicator thresholds were the same for grasslands used for high

agricultural production as for grasslands managed for biodiversity conservation. Acknowledgements We thank the State Secretariat of Research and Education (SFB) for funding this project and the COST action 869 for their financial support. We also thank the farmers for access to their fields and their cooperation and Isabelle Falconi-Bürgi and Franz Stadelmann from Canton Lucerne, Department for Agriculture and Forest, for their advice. We acknowledge Cornelia Bühlmann, Monika Scheller and the technicians from the involved institutes for their practical support and Lukas Rosinus and Juan Herrera for their statistical advice. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.eja.2012.08.004. References Arrigo, Y., 1999. Fütterungsempfehlungen und Nährwerttabellen für Wiederkäuer, 4th ed. Eidgenössische Forschungsanstalt für Nutztiere (Posieux), Zollikofen. Bailey, J.S., Beattie, J.A.M., Kilpatrick, D.J., 1997a. The diagnosis and recommendation integrated system (DRIS) for diagnosing the nutrient status of grassland swards: I. Model establishment. Plant and Soil 197, 127–135. Bailey, J.S., Cushnahan, A., Beattie, J.A.M., 1997b. The diagnosis and recommendation integrated system (DRIS) for diagnosing the nutrient status of grassland swards: II. Model calibration and validation. Plant and Soil 197, 137–147. Black, C.A., 1993. Soil Fertility Evaluation and Control, 2nd ed. CRC Press Inc., Boca Raton. Ceulemans, T., Merckx, R., Hens, M., Honnay, O., 2011. A trait-based analysis of the role of phosphorus vs. nitrogen enrichment in plant species loss across North-west European grasslands. Journal of Applied Ecology 48, 1155–1163. Daccord, R., Arrigo, Y., Kessler, J., Jeangros, B., Scehovic, J., Schubiger, F.X., Lehmenn, J., 2001. Nährwert von Wiesenpflanzen: Gehalt an Ca, P, Mg und K. AGRARForschung 8, 264–269. Dampney, P.M.R., 1992. The effect of timing and rate of potash application on the yield and herbage composition of grass grown for silage. Grass & Forage Science 47, 280–289. Dietl, W., 1995. Wandel der Wiesenvegetation im Schweizer Mittelland. Zeitschrift für Ökologie und Naturschutz 4, 239–249. Duru, M., Thélier Huché, L., 1997. N and P–K status of herbages: use for diagnosis of grasslands. In: Inra (Ed.), Diagnostic Procedures for Crop N Management. Les colloques de l’Inra, pp. 125–138. FAL, FAC, RAC, 1996. Bodenuntersuchungen für die Düngungsberatung. Schweizerische Referenzmethoden der eidgenössischen landwirtschaftlichen Forschungsanstalten Agroscope, vol. 1. FAO, 1998. World Reference Base for Soil Resources. FAO, ISRIC and ISSS, Rome. Flisch, R., Sinaj, S., Charles, R., Richner, W., 2009. Grundlagen für die Düngung im Acker- und Futterbau. AGRARForschung 16, 1–100. Fujita, Y., Robroek, B.J.M., de Ruiter, P.C., 2010. Increased N affects P uptake of eight grassland species: the role of root surface phosphatase activity. Oikos 119, 1665–1673. Gallet, A., Flisch, R., Ryser, J.P., Nösberger, J., Frossard, E., Sinaj, S., 2003. Uptake of residual phosphate and freshly applied diammonium phosphate by Lolium perenne and Trifolium repens. Journal of Plant Nutrition and Soil Science 166, 557–567. Haygarth, P.M., Bilotta, G.S., Bol, R., Brazier, R.E., Butler, P.J., Freer, J., Gimbert, L.J., Granger, S.J., Krueger, T., Macleod, C.J.A., Naden, P., Old, G., Quinton, J.N., Smith, B., Worsfold, P., 2006. Processes affecting transfer of sediment and colloids, with associated phosphorus, from intensively farmed grasslands: an overview of key issues. Hydrological Processes 20, 4407–4413. Hill, J.O., Simpson, R.J., Wood, J.T., Moore, A.D., Chapman, D.F., 2005. The phosphorus and nitrogen requirements of temperate pasture species and their influence on grassland botanical composition. Australian Journal of Agricultural Research 56, 1027–1039. Høgh-Jensen, H., Schjoerring, J.K., 2010. Interactions between nitrogen, phosphorus and potassium determine growth and N2 -fixation in white clover and rye grass leys. Nutrient Cycling in Agroecosystems 87, 327–338. Huguenin-Elie, O., Gago, R., Stutz, C., Lüscher, A., Kessler, W.,2006a. Long-term effects of fertilisation on herbage composition, yield and quality of an Arrhenatherion˜ para type meadow. In: Grassland Science in Europe, vol. 11. Sociedad Espanola el Estudio de los Pastos (SEEP), Madrid, pp. 550–552. Huguenin-Elie, O., Stutz, C., Lüscher, A., Kessler, W., Gago, R., 2006b. Einfluss von Stickstoff und Kalium auf eine Fromentalwiese. AGRARForschung 13, 62–67.

F. Liebisch et al. / Europ. J. Agronomy 44 (2013) 67–77 IDAWEB, 2009. Meteo Swiss, the Swiss Federal Office of Meteorology and Climatology. Isselstein, J., Jeangros, B., Pavlu, V., 2005. Agronomic aspects of biodiversity targeted management of temperate grasslands in Europe. Agronomy Research 3, 139–151. Jeangros, B., 1993. Prairies permanentes en montagne. Effets de la fréquence des coupes et de la fertilisation azotée sur la composition botanique. Revue Suisse d’Agriculture 25 (6), 345–360. Jeangros, B., Scehovic, J., Schubiger, F.X., Lehmann, J., Arrigo, Y., Daccord, R., 2001. Nährwert von Wiesenpflanzen: Trockensubstanz-, Rohprotein- und Zuckergehalte. AGRARForschung 8, 1–8. Jouany, C., Cruz, P., Theau, J.P., Petibon, P., Foucras, J., Duru, M., 2005. Diagnostic du statut de nutrition phosphatée et potassique des prairies naturelles en présence de légumineuses. Fourages 184, 547– 556. Klötzli, F., Dietl, W., Marti, K., Schubiger-Bossard, C., Gian-Reto, W., 2010. Vegetation Europas. Das Offenland im vegetationskundlich-ökologischen Überblick, 1st ed. hep Verlag ag, Bern. Koerselman, W., Meuleman, A.F.M., 1996. The vegetation N:P ratio: a new tool to detect the nature of nutrient limitation. The Journal of Applied Ecology 33, 1441–1450. Ledgard, S.F., Sprosen, M.S., Steele, K.W., 1996. Nitrogen fixation by nine white clover cultivars in grazed pasture, as affected by nitrogen fertilization. Plant and Soil 178, 193–203. Mendiburu de, F., 2009. Agricolae – Statistical Procedures for Agricultural Research Using R, http://tarwi.lamolina.edu.pe/∼fmendiburu. Mitscherlich, E.A., 1928. Die zweite Annäherung des Wirkungsgesetzes der Wachstumsfaktoren. Zeitschrift für Pflanzenernährung, Düngung, Bodenkunde 12, 273–282.

77

Ohno, R., Zibilske, L.M., 1991. Determination of low concentrations of phosphorus in soil extracts using malachite green. Soil Science Society of America Journal 55, 892–895. Peeters, A., 2009. Importance, evolution, environmental impact and future challenges of grasslands and grassland-based systems in Europe. Grassland Science 55, 113–125. Pilgrim, E.S., Macleod, C.J.A., Blackwell, M.S.A., Bol, R., Hogan, D.V., Chadwick, D.R., Cardenas, L., Misselbrook, T.H., Haygarth, P.M., Brazier, R.E., Hobbs, P., Hodgson, C., Jarvis, S., Dungait, J., Murray, P.J., Firbank, L.G., 2010. Interactions among agricultural production and other ecosystem services delivered from European temperate grassland systems. In: Donald, L.S. (Ed.), Advances in Agronomy. Academic Press, Newark, USA, pp. 117–154. R-Development-Core-Team, 2008. R. A language and environment for statistical computing. In: R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. URL http://www.R-project.org. Systat Software Inc., 2008. SigmaPlot for Windows Version 11.0, build 11.0.0.75. Thöni, E., 1964. Über den Einfluss von Düngung und Schnitthäufigkeit auf den Pflanzenbestand und den Mineralstoffgehalt des Ertrages einer feuchten Fromentalwiese. Dissertation (Prom. Nr. 3505). ETH Zürich. Thöni, E., 1982. Die Wirkung von Schnitthäufigkeit und Stickstoffdüngung auf verschiedene Mischungstypen im Kunstfutterbau. Schweizerische Landwirtschaftliche Forschung 21, 111–126. Venterink, H.O., Wassen, M.J., Verkroost, A.W.M., de Ruiter, P.C., 2003. Species richness-productivity patterns differ between N-, P-, and K-limited wetlands. Ecology 84, 2191–2199. Wassen, M.J., Venterink, H.O., Lapshina, E.D., Tanneberger, F., 2005. Endangered plants persist under phosphorus limitation. Nature 437, 547–550. Whitehead, D.C., 2000. Nutrient Elements in Grassland: Soil–Plant–Animal Relationships. Cabi, New York.