Grazed bite item diversity of beef cows grazing semi-natural grassland in relation to herbage nutritive value and the actually available diversity of bite items

Abstract

Heterogeneous extensive grassland offers herbivores a variation of potential bite items (BI) from which they can select. But there is a lack of knowledge on the relationship between the actually available (BI_a) and the grazed (BI_g) BI diversity albeit this information is essential to understand grass sward composition. We assessed the BI_a in the sward and the BI_g to evaluate the BI diversity under two grazing intensities in two seasons in the replicated long-term grazing experiment Forbioben. The BI_a diversity was measured pre-grazing using a modified sward stick (200 measurement points per 1-ha paddock) whereby at each point the botanical composition, phenological stage, color and height were recorded and consequently structured into BI_a. Herbage samples were taken for each BI and analyzed via near-infrared-reflectance spectroscopy for the concentration of crude protein (CP), acid-detergent fiber in the organic matter (ADF), in vitro digestibility of the organic matter (ivdOM) and metabolizable energy (ME) and labeled as grazed and non-grazed. For the BI_g diversity each cow was observed in the morning and afternoon four times, 2 min each on 1 d each season using an app for counting the bites and following video analysis to determine the composition of the BI_g. A reduction in growth rates during the hot and dry summer might have caused the decline of BI_a diversity of 25% under moderate and 22% under lenient grazing towards autumn (p < 0.0001). This also influenced the BI_g as in spring 50% and 53% of the BI_a diversity were grazed under moderate and lenient grazing, respectively. In autumn only 36% and 28% were grazed, respectively corresponding well to the generally lower herbage nutritive value of BI in autumn as the offered grassland resource changed over time and provided greater ivdOM and ME in spring compared to autumn (p < 0.0001). This shows a greater selectivity from the limited choice in autumn. All cows were able to maintain a good quality diet irrespective of grazing intensity which is underlined by the fact that the digestibility of ingested herbage was the same among the grazing intensities but greater in spring than in autumn.

Heterogeneous and species-rich grassland can offer grazing cows a wide variety of bite items. We wanted to analyze this diversity in terms of composition and quality and how the diversity and quality affect the selectivity of beef cows on pasture.

INTRODUCTION

Central European grassland can be very heterogeneous containing a diverse array of species-richness when the management aims at preservation of grassland biodiversity (de Vries and Daleboudt, 1994; Verhulst et al., 2004; Marriott et al., 2009; Pakeman et al., 2019) by controlling the dynamics of plant species composition through extensive grazing with beef cattle (Molina-Maroto and Pérez-Marín, 2021).

Extensive grazing management usually provides feed in excess leading to the development of short, frequently defoliated areas and avoided areas with taller vegetation over time (Dumont et al., 2012). Biomass removal through grazing decreases light limitation and this has been identified as the most likely driver of patch-type differences in plant diversity (Borer et al., 2014). Short patches have therefore a greater diversity than tall patches (Obermeyer et al., 2022; Komainda et al., 2023). As a result, diversity measures at the pasture scale were greatest with the highest proportion of regularly defoliated short patches, as induced by a high but still extensive grazing intensity, rather than when the contributions of short and tall patches were equal (Tonn et al., 2019), such as under lower grazing intensity. In terms of the herbage nutritive value (Allen et al., 2011) regularly defoliated leafy short patches provide higher herbage nutritive values than tall patches (Clausecker et al., 2024) but herbage nutritive value varies also in response to the grazing pressure and seasonality as a result of the phenological development (Pavlů et al., 2006, 2021; Kleinebecker et al., 2011).

Overall, diverse patchy grassland offers a range of potential bite items (BI) that represent a variation in botanical diversity, plant organs, phenological development and chemical composition (Rutter, 2006; Fraser et al., 2022). The utilization of this BI diversity by grazing herbivores determines how species richness in grassland develops in the long-term (Dumont and Tallowin, 2011). If minor sward constituents forming potential BI are grazed preferentially these may become lost from the grass sward. On the other hand, if cows select preferentially from dominating sward constituents, minor sward components can compete and increase their overall contribution. Finding the grazing intensity that keeps the grassland open and manifests itself in the greatest diversity is still a matter of debate (Chabuz et al., 2019). Understanding cows’ views on diversity through BI selection in relation to BI provision could help refine management and make predictions of expected outcomes.

Some studies assessed the BI choice of cattle in grassland (Rutter et al., 2004; Dumont et al., 2007; Orr et al., 2012) or selection of distinct plant species (Cuchillo-Hilario et al., 2018). Others have investigated the quality of grazed forage species (Orr et al., 2004; Soder et al., 2009) and a few studies have assessed both quality and BI choice (de Vries and Daleboudt, 1994; Orr et al., 2014; Koczura et al., 2019). Overall, animals select BI containing preferred plant species (de Vries and Daleboudt, 1994; Dumont et al., 2007; Bolzan et al., 2020) of greater feed value than the pasture on average provides (Horadagoda et al., 2009; Edouard et al., 2010; Galyean and Gunter, 2016). However, it is unclear whether and how the selectivity of BI and their diversity changes with alterations of actual BI diversity such as under varied grazing intensity or the season which may also interact with one another.

Short grass swards with their greater plant species diversity and greater average herbage nutritive value theoretically allow the use of a greater BI diversity because herbage nutritive value is not limiting BI choice especially towards the end of the grazing season when more generative tall patches are present under less intensive grazing. The proportion of grazed BI relative to the BI availability should therefore be larger under higher compared to lower grazing intensity because selectivity is lower in the latter. However, due to the negative sward height-bite mass relationship (Ungar, 1996; Rook, 2000), the use of the entire diversity may be limited from constraints to maintain instantaneous herbage intake rate (Prache and Delagarde, 2011). On the other hand, cattle on short swards might be forced to increase BI diversity to maintain instantaneous herbage intake. The selection of BI in response to BI diversity has in general rarely been considered (Bolzan et al., 2020). Our research aims were therefore to assess the BI diversity and choice of BI in relation to grazing intensity and season which contrast in herbage nutritive value. In the present study, the term BI diversity refers to the number of all different BI found in the paddocks pre-grazing (available BI = BI_a) or those BI that were ingested during grazing (grazed BI = BI_g). Therefore, each BI consists of a combination of the botanical composition (grass, herbs, legumes and all their possible combinations), the phenology (vegetative, generative), the color (green, mixed, brown), and the extended sward height (short ≤ 10 cm, tall > 10 cm) and are referred to as unique BI following the approach of Farruggia et al. (2006). The BI diversity comprises the actual bite item (Bi_a) diversity, i.e., all BI that are available for the cows to graze in the respective pasture, and the grazed bite item (BI_g) diversity, i.e., all BI that were actually grazed by the cows. The main hypotheses were that i) shorter grass swards as found under higher grazing intensity allows a greater BI_a diversity, that ii) the BI_g diversity corresponds to the BI_a diversity and iii) that environmental conditions change both the BI_a and BI_g diversity in the same way.

MATERIAL AND METHODS

The study is in accordance with the German legal and ethical requirements of appropriate animal procedures. The consultation of the Institutional Animal Welfare Body is documented under no. E5/20 by the Animal Welfare Officer of the University of Goettingen.

Study Site, Setup and Weather Conditions

The study was conducted in the 2022 grazing season during two periods (between May and November) as part of the long-term cattle grazing experiment Forbioben which is located at the experimental farm of the University of Göttingen in Relliehausen, Solling Uplands, Lower Saxony, Germany (51°46’55.9 “N, 9°42’11.9”E), 265 to 340 m above sea level. The vegetation is a moderately species-rich semi-natural grassland, and it is representative of the species association known as Cynosurion cristati (Runge, 1973). The soil type is a Vertic Cambisol. The experiment was established in 2002 as a one-factorial experiment and has been managed in its current state since 2005. It compares three different grazing intensities (moderate (M), lenient (L), and very lenient (VL)) as the experimental treatment. The three grazing intensities are replicated three times and arranged in a randomized block design of three paddocks (each 1 ha) per grazing intensity. For the current study, the six paddock belonging to the grazing intensities M and L were used. The grazing intensities are controlled through different target sward heights, which are 6 cm in M and 12 cm in L. Cattle are put on the pasture when the target sward height is exceeded and removed when it falls below. No fertilizers, pesticides or other management options to improve the sward were used since at least 10 yr before the start of the experiment.

Temperature and precipitation data was retrieved from the station of the German Weather Service in Bevern (Holzminden) (DWD, 2022) approximately 21 km from the study side. Monthly mean temperature values during the study period were nearly always warmer (except during September) than the long-term average (Table 1). The monthly precipitation sum was always below the long-term average. The year 2022 represents one of the hottest and driest years recorded and is part of a sequence of dry years that occurred between 2018 and 2022 (Knutzen et al., 2023; Zhang et al., 2023; Gharun et al., 2024; Wang et al., 2024).

Table 1.

The monthly average temperature (°C) and precipitation sum (mm) from May to November 2022 for the station of the German Weather Service in Bevern (Holzminden). The long-term monthly mean temperatures and precipitation sums from 1991 to 2020 are shown in parentheses (DWD, 2022)

Month	Temperature °C	Precipitation mm
May	14.8 (13.5)	51.6 (60.0)
June	18.4 (16.5)	47.3 (65.0)
July	19.1 (18.5)	34.8 (78.0)
August	20.5 (18.2)	55.4 (70.0)
September	14.0 (14.2)	51.4 (58.0)
October	12.4 (9.8)	53.6 (67.0)
November	7.2 (5.6)	31.3 (61.0)

Animals

Grazing livestock were non-lactating Fleckvieh suckler cows (6.3 ± 3.20 yr in age, 641 ± 38.3 kg live weight in spring, 4 ± 3.1 calvings, means ± SD). Before each grazing period, the cows were weighed and randomly distributed to the paddocks according to their live weight in order to achieve similar groups. In spring, the grazing period lasted from May 11 to June 20 in grazing intensity M and from May 20 to June 15 in grazing intensity L. Three cows were turned out on each paddock. In autumn, the grazing period lasted from October 12 to November 4 in both grazing intensities treatments. Three and two cows were turned out on each paddock in grazing intensity M and L, respectively. The resulting stocking densities in livestock units (LU = 500 kg live weight ha⁻¹) were 3.9 ± 0.06 LU ha⁻¹ for both grazing intensities in spring. In autumn stocking densities were 4.5 ± 0.03 LU ha⁻¹ and 3.0 ± 0.05 LU ha⁻¹ in the M and L grazing intensities, respectively. Before and between the two grazing periods the cows were kept on similar grassland adjacent to the experimental paddocks where they were supplied with hay and straw during periods of low grassland growth.

Pre-Grazing Herbage on Offer

A rising plate meter was used every 2 wk to measure the compressed sward height (CSH, cm) on 50 points per paddock using a rising plate meter of 30 cm diameter and 200 g plate weight (Castle, 1976) to assess the target sward height. The herbage on offer was assessed by using a double sampling approach throughout the grazing season on a monthly basis. Therefore, CSH was measured on eight places per paddock and date. Afterwards the standing aboveground herbage biomass was cut to ground level. Samples were immediately cooled and transferred to the laboratory for drying to constant weight (at least 48 hours, 60 °C). The herbage dry matter samples determined in this way for each CSH measurement were used to obtain a linear regression model between the standing aboveground herbage biomass and CSH to predict the herbage on offer per paddock and date for all the remaining CSH measurements (Correll et al., 2003; Şahin Demirbağ et al., 2009). In this way, values for the pre-grazing herbage on offer were obtained.

Botanical Diversity on Pasture-Scale

The cover of all vascular plant species was visually estimated in May 2022 using the method of Braun-Blanquet (1964) in ten 1 m² observation quadrats per paddock. The mean number of species richness per m² as well as the Shannon diversity were calculated (Magurran, 2003). The dataset contains 60 botanical evaluations (6 paddocks x 10 subplots).

Assessment of Actual and Grazed Bite Item Diversity

Before the start of the study, we determined the composition of a BI as based on the characteristics of the grassland on the basis of what the cow can perceive. Based on previous studies of the bite area, a size of 100 cm² was determined for one BI (Laca et al., 1992; Ungar and Ravid, 1999). Figure 1 shows the different BI diversities and the factors that influence them, as well as what the BI consist of and how they were recorded. This is explained in more detail in the following sections.

Figure 1.

Schematic overview of the bite item diversities and their influencing factors as well as their measurement.

Actual Bite Item Diversity

The BI_a diversity reflects all unique BI and their percentage proportion that are available within one pasture environment pre-grazing and in the present case consequently in one experimental grazing paddock. For the purpose of the present study, it was assessed along two randomly placed transects per paddock (200 points per paddock) using a sward stick with a 10 × 10 cm steel frame adjusted to it. For each measurement point along the transect the characteristics of the BI was noted. In this way the number of unique BI per paddock was immediately determined in the field. Thereafter, herbage samples of each BI_a found per paddock were collected for herbage nutritive value analyses. For this, samples were taken by collecting 10 to 15 single hand-pickings of the upper third to half of the extended sward height in order to mimic the grazing animal. The 10 to 15 single pickings belonging to one BI were bulked into one sample. The BI specific samples were taken at random across the paddock and date.

Grazed Bite Item Diversity

To assess the BI_g diversity (chosen BI in the process of grazing), each cow was observed for four times, 2 min each in the morning and in the afternoon of each period (16 min per cow and period). The cows were observed by filming the mouth (Canon Eos 1100D) while using the mobile phone app obslog (Hamidi and Hamidi, 2022) in parallel to count all bites, differentiated between short (≤ 10 cm) and tall (>10 cm). The app obslog was specifically designed for the purpose of the present study. A screenshot of the apps surface is shown in Supplementary Material S1. For the purpose of the study, all cows had to be observed and filmed while grazing. Therefore, the order of cows per paddock was chosen at random and paddocks were filmed in a random manner depending on where the cows were grazing. At the start of each filming interval the time as well as the number of the interval was announced on the video. Each video was labeled according to the cow ID, the date, time and number of interval and stored to an external data storage. The app used delivered a date and time stamp for each click as well as the total number of clicks in one interval. At the end of each interval a stop button was clicked, signaling the end of an interval. Furthermore, after each interval the data was saved, and a .csv file was generated via the app which was also stored on an external data storage unit until further analysis. Via subsequent video analysis, the botanical composition, the color and phenological stage for each bite was assessed visually in order to obtain the number and proportion of unique BI in accordance with the BI_a diversity. Therefore, the VLC plus player (3.0.18 Vetinari) was used and videos were played in slow motion (0.50 x speed) for analyses. Due to the time stamp of the .csv file from the app, the time announced at the beginning of each video by the recording person as well as the audio of the cows taking a bite, each bite in the video could be matched to those in the .csv file from the animal observation. In this way the grazed BI could be classified in accordance to the ones assessed as BI_a in pastures. Besides the total number of unique BI, the relative BI_g diversity was calculated as proportion of unique BI_g per number of unique BI_a. The same person used the digital camera throughout the study period, while another person conducted the animal observation using the app to count the bites. Based on the video analyses the obtained BI_a (including those intended for herbage nutritive value analyses) could be subsequently labeled as grazed or non-grazed. Whether a BI was grazed or non-grazed is named ingested in the following for data analyses (see below).

Calculation of Bite Item Diversity and Selection

For each unique BI_a, we determined its proportion across all BI_a (a). The same was done for the BI_g (g). Then the Jacobs selectivity index (JSI) was calculated after Jacobs (1974):

$${\displaystyle \begin{array}{l}{\displaystyle JSI=(g-a)/(g+a-2\times g\times a).}\end{array}}$$

The JSI can reach a value between −1 and 1. Here, −1 means that the BI is grazed less than it is available in the sward, 0 means that the BI is grazed to the same extent as it is available and 1 means that it is preferentially selected. For the calculation of the JSI, only BI that were found as both BI_a and BI_g could be considered, which means that 50.9% of the BI_a were included in spring and 55.8% in autumn.

Herbage Nutritive Value on the Pasture-Scale and of Single Bite Items

The samples taken by hand plucking were immediately cooled after sampling and weighted before and after drying them at 60 °C for 48 hours in a forced air-oven. Afterwards the samples were milled in a two-step procedure, first to pass a 4-mm and then a 1-mm screen (Retsch SM 300 & Retsch ZM 300, Retsch GmbH, Haan, Germany). After milling, samples were analyzed using near-infrared-reflectance spectroscopy (NIRS) by scanning each sample twice on a Phoenix 5000 (BlueSun Scientific Inc, MD, USA). In order to obtain a weighted mean value of the paddock-based herbage nutritive value pre-grazing, the frequency of each unique BI_a was calculated per paddock, multiplied with the concentration of each herbage nutritive value parameter and the values were accumulated per paddock.

For the purpose of the present study, concentrations of crude protein (CP), the ash-free acid-detergent fiber in the organic matter (ADF), crude fat (CL) and crude fiber (CF) as well as crude ash (CA) and enzymatically insoluble organic matter (EISOM) were analyzed. Concentrations were then processed and predicted using a large calibration data set stored on a central server (VDLUFA Qualitätssicherung, NIRS GmbH, Kassel, Germany). The data sets for NIRS of the quality parameters CP, ADF, CL and CF contained 3169, 1088, 949 and 2676 calibration samples, respectively. Standard errors of calibration for CP, ADF, CL and CF were 0.76%, 1.32%, 0.3% and 1.22% and the corresponding standard errors of cross validation were 0.77%, 1.35%, 0.31% and 1.24%. In total 231 samples were analyzed. The metabolizable energy (ME MJ kg⁻¹ DM) was calculated after Losand et al. (2007):

$${\displaystyle \begin{array}{l}{\displaystyle \begin{array}{llll}{\displaystyle ME\left(MJk{g}^{-1}DM\right){\displaystyle }}& =5.51+0.00827\times ESOM\left(gk{g}^{-1}DM\right)& -0.0051\times CA\left(gk{g}^{-1}DM\right)& {\displaystyle +0.0251\times CL\left(gk{g}^{-1}DM\right),}\end{array}}\end{array}}$$

where ESOM is the enzymatic solubility of the organic matter and is calculated as

$${\displaystyle \begin{array}{lll}{\displaystyle ESOM\left(gk{g}^{-1}DM\right)=}& 1,000-EISOM\left(gk{g}^{-1}DM\right)& -CA\left(gk{g}^{-1}DM\right)\end{array}}$$

The in vitro digestibility of the organic matter (ivdOM) was calculated after Schmidt et al. (2004):

$${\displaystyle \begin{array}{lllll}{\displaystyle ivdOM\left(\mathrm{\%}\right)=}& 100\times (940-CA\left(gk{g}^{-1}DM\right)& -0.62\times EISOM\left(gk{g}^{-1}DM\right)& -0.000221\times EISOM{\left(gk{g}^{-1}DM\right)}^2)/& (1,000-CA\left(gk{g}^{-1}DM\right)).\end{array}}$$

Faecal Sample Analysis

In addition, faecal samples were taken from each cow during each season after being in the paddock for at least 3 d. Samples were collected immediately after defecation. The samples were immediately cooled after collection and subsequently stored at -18 °C. The samples were then defrosted and weighed before and after drying at 60 °C for 48 hours in a forced air-oven. Subsequently, the samples were milled to 1 mm and a subsample to 0.2 mm. The 1 mm samples were used to determine the CA concentration by weighing 1 g per sample and drying it at 105 °C for 24 hours whereafter it was placed in a muffle furnace (Nabertherm, Lilienthal) at 550 °C for 24 hours. Samples were weighed after each step and the ash concentration was used to calculate the organic matter of the faeces. The 0.2 mm samples were used for total carbon and nitrogen (C/N) analysis. The analysis of a subsample of 13 to 15 mg was done using elemental analysis (vario EL cube, Elementar, Langenseibold) by double sampling. The N concentration in the organic matter of faeces was then used to calculate the digestibility of the ingested herbage after Schmidt et al. (1999):

$${\displaystyle \begin{array}{lll}{\displaystyle dOM\left(\mathrm{\%}\right)=}& 95.90-460/N_{faeces}\left(gk{g}^{-1}DM\right)& -0.1582GD+0.00062G{D}^2,\end{array}}$$

where GD are the grazing days of the year. This method is commonly applied in order to calculate the ingested quality of herbage on grazed pastures (e.g., Smit et al., 2005; Schneider et al., 2011).

Data Analysis

The following target variables were obtained for the purpose of the present study: herbage on offer, Shannon diversity, plant species richness, the pasture-scale herbage nutritive value (CP, ADF, ME, ivdOM), dOM of ingested herbage, the BI_a and BI_g diversity, BI quality and the JSI.

All data were processed with RStudio (R Version 1.4.1717, 2021, The R Foundation for Statistical Computing, Vienna, Austria).

We analyzed each parameter using linear mixed effects models (package “nlme”) (Pinheiro et al., 2023) assuming fixed effects and interactions and all models were checked to meet the requirements of a normal distribution of residuals and variance homogeneity. All models included paddock nested in block as a random effect to account for the repeated measures design.

The following model structure was used for analyses of BI_a, BI_g, logit-transformed relative BI_g and herbage on offer pre-grazing as well as the weighed mean herbage nutritive value per paddock (CP, ADF, ME, ivdOM):

$${\displaystyle \begin{array}{l}{\displaystyle Y_{ij}=\mu +{\alpha }_i+{\beta }_j+{\left(\alpha \beta \right)}_{ij}+{\epsilon }_{ij},}\end{array}}$$

where Y_ij is the observation of the j-th grazing intensity in the i-th season, µ is the overall mean, α_i is the fixed effect of the i-th season, β_j is the fixed effect of the j-th grazing intensity, (αβ)_ij is the interaction effect of the i-th season and j-th grazing intensity, and ε_ij is the normally distributed error associated with the Y_ij observation. Separate variances per period were allowed for relative BI_g, per grazing intensity for ADF, per grazing intensity and of season for ME and ivdOM in order to meet the assumption of variance homogeneity.

For the Shannon diversity index and plant species richness the following model was used:

$${\displaystyle \begin{array}{l}{\displaystyle Y_i=\mu +{\alpha }_i+{\epsilon }_i,}\end{array}}$$

where Y_i is the observation of the i-th grazing intensity, µ is the overall mean, α_i is the fixed effect of the i-th grazing intensity, and ε_i is the normally distributed error associated with the Y_i observation.

For all BI quality parameters (CP, ADF, ME, ivdOM) the following model was used:

$${\displaystyle \begin{array}{l}{\displaystyle Y_{ijk}=\mu +{\alpha }_i+{\beta }_j+{\gamma }_k+{\left(\alpha \beta \gamma \right)}_{ijk}+{\epsilon }_{ijk},}\end{array}}$$

where Y_ijk is the observation of the k-th ingestion under the j-th grazing intensity in the i-th season, µ is the overall mean, α_i is the fixed effect of the i-th season, β_j is the fixed effect of the j-th grazing intensity, γ_k is the fixed effect of the k-th ingestion, (αβγ)_ijk is the interaction effect of the i-th season, j-th grazing intensity and k-th ingestion, and ε_ijk is the normally distributed error associated with the Y_ijk observation. The factor ingestion consequently served as factor to distinguish grazed from non-grazed BI. For ADF a separate variance was allowed per period and for ME and ivdOM for the factor ingestion in order to avoid variance heterogeneity.

For the Jacobs selectivity index the following model was used:

$${\displaystyle \begin{array}{l}{\displaystyle Y_{ij}=\mu +{\alpha }_i+{\beta }_j+{\epsilon }_{ij},}\end{array}}$$

where Y_ij is the observation of the j-th BI in the i-th grazing, µ is the overall mean, α_i is the fixed effect of the i-th grazing intensity, β_j is the fixed effect of the j-th BI, and ε_ij is the normally distributed error associated with the Y_ij observation.

For all parameters the final model was chosen, as based on the AIC for small samples sizes, and then tested for significance with marginal Wald tests. Comparisons of means were done posthoc for significant influencing variables using pairwise comparisons according to Tukey’s HSD test in the “emmeans” package (Lenth et al., 2024).

RESULTS

Availability and Quality of Herbage on Offer on Pasture-Scale

Differences in the pre-grazing standing herbage on offer depended on the season in interaction with the grazing intensity (Table 2). Shannon diversity and species richness did not differ significantly on pasture level (Table 2). The concentrations of CP and ADF differed depending on the season and the concentration of ME was not affected by any of the tested effects (Table 2). The ivdOM differed depending on the grazing intensity and the dOM of ingested herbage, as obtained from faecal samples, differed depending on the season (Table 2).

Table 2.

Output of marginal wald tests for the pre-grazing herbage on offer, Shannon diversity (m⁻²), plant species richness (m⁻²) and the concentrations of crude protein (CP, g kg⁻¹ DM), acid-detergent fiber in the organic matter (ADF, g kg⁻¹ DM), metabolic energy (ME, MJ kg⁻¹ DM), the in vitro organic matter digestibility (ivdOM, %) and the organic matter diegestibility of ingested herbage (dOM, %). Given are degrees of freedom, F- and p-values

	denDF	F-value	p-value
Herbage on offer (t DM ha−1)
Grazing intensity	2	3.40	0.2066
Season	4	166.66	0.0002
Grazing intensity*season	4	66.00	0.0012
Shannon diversity (m−2)
Grazing intensity	2	2.54	0.2518
Plant species richness (m−2)
Grazing intensity	2	2.70	0.2424
CP (g kg−1 DM)
Grazing intensity	2	4.99	0.1551
season	4	7.78	0.0494
Grazing intensity*season	4	0.04	0.8587
ADF (g kg−1 DM)
Grazing intensity	2	2.53	0.2524
Season	4	11.29	0.0283
Grazing intensity*season	4	2.37	0.1983
ME (MJ kg−1 DM)
Grazing intensity	2	17.42	0.0857
Season	4	0.16	0.5027
Grazing intensity*season	4	0.04	0.7257
ivdOM (%)
Grazing intensity	2	19.88	0.0468
Season	4	0.04	0.8596
Grazing intensity*season	4	0.02	0.9066
dOM of ingested herbage (%)
Grazing intensity	25	4.22	0.0506
Season	1	206.19	0.0443
Grazing intensity*season	1	2.15	0.3814

Herbage on offer in spring was higher in grazing intensity M than in grazing intensity L, but did not differ among grazing intensities in autumn. In grazing intensity L, herbage on offer differed between seasons, whereas this was not the case in grazing intensity M (Table 3).

Table 3.

Estimated means (± se) for the pre-grazing herbage on offer (t ha⁻¹ DM), the concentrations of crude protein (CP, g kg⁻¹ DM), acid-detergent fiber in the organic matter (ADF, g kg⁻¹ DM), the in vitro organic matter digestibility (ivdOM, %) and the organic matter diegestibility of ingested herbage (dOM, %). Different letters indicate significant differences among means of the grazing intensities within season and letters in parentheses indicate significant differences among seasons within grazing intensities for the herbage on offer and different letters indicate significant differences among means within the other parameters (p < 0.05)

Parameter	Season	Grazing intensity	Value
Herbage on offer (t DM ha−1)	Spring	M	1.9 ± 0.16 a (a)
		L	3.5 ± 0.16 b (a)
	Autumn	M	1.7 ± 0.16 a (a)
		L	2.1 ± 0.16 a (b)
CP (g kg−1 DM)	Spring	-	123 ± 8.2 a
	Autumn	-	175 ± 11.7 b
ADF (g kg−1 DM)	Spring	-	208 ± 12.6 a
	Autumn	-	296 ± 12.6 b
ivdOM (%)	-	M	68.1 ± 1.63 a
	-	L	53.1 ± 17.06 a
dOM of ingested herbage (%)	Spring	-	76.3 ± 0.092 b
	Autumn	-	69.8 ± 0.91 a

The CP concentration was significantly greater in autumn compared to spring which also applied to the ADF concentration (Table 3). The ivdOM did not differ among the seasons (Table 3). The dOM of ingested herbage, on the other hand, was significantly higher in spring compared to autumn and there was no significant difference among grazing intensity treatments (Table 3).

Bite Item Diversity

Differences in the BI_a and BI_g diversity depended on the season in interaction with the grazing intensity (Table 4). Differences in the relative BI_g diversity depended on the season (Table 4).

Table 4.

Output of marginal wald tests for the actual (BI_a), grazed (BI_g) and relative BI_g diversity. Given are degrees of freedom, F- and p-values

	denDF	F-value	p-value
BIa diversity
Grazing intensity	2	18.24	0.0507
Season	223	554.88	<0.0001
Grazing intensity*season	223	54.57	<0.0001
BIg diversity
Grazing intensity	2	1.45	0.3516
Season	91	115.92	<0.0001
Grazing intensity*season	91	4.61	0.0344
Relative BIg diversity
Grazing intensity	2	1.01	0.4200
Season	4	9.98	0.0342
Grazing intensity*season	4	0.42	0.5536

The BI_a and BI_g diversity were greater in spring compared to autumn for both grazing intensities (Table 5). However, only in spring a significant difference in the BI_a was found between the two grazing intensities with higher values under M compared to L (Table 5). The relative BI_g diversity did not differ within the seasons among grazing intensities, but was greater in spring compared to autumn in grazing intensity L (Table 5).

Table 5.

Estimated means (± se) for the actual (BI_a), grazed (BI_g) and relative BI_g diversity of unique bite items as presented in an interaction of grazing intensity*season. Different letters indicate significant differences among means of the grazing intensities within season and letters in parentheses indicate significant differences among means of seasons within grazing intensities (p < 0.05)

Diversity	Season	Grazing intensity	Number
BIa	Spring	M	24.6 ± 2.13 b (b)
		L	19.1 ± 2.13 a (b)
	Autumn	M	18.6 ± 2.14 a (a)
		L	14.9 ± 2.14 a (a)
BIg	Spring	M	12.4 ± 0.81 a (b)
		L	10.3 ± 0.82 a (b)
	Autumn	M	6.3 ± 0.84 a (a)
		L	5.5 ± 0.87 a (a)
Relative BIg	Spring	M	0.50 ± 0.03 a (a)
		L	0.53 ± 0.04 a (b)
	Autumn	M	0.36 ± 0.05 a (a)
		L	0.28 ± 0.13 a (a)

Bite Item Quality

The differences observed in the concentrations of CP, ADF and ME depended mainly on the season in interaction with the ingestion (Table 6). The same was found for the concentration of ivdOM (Table 6).

Table 6.

Output of marginal wald tests on the concentrations of crude protein (CP, g kg⁻¹ DM), acid-detergent fiber in the organic matter (ADF, g kg⁻¹ DM), metabolic energy (ME, MJ kg⁻¹ DM) and the in vitro organic matter digestibility (ivdOM, % DM⁻¹) of the individual bite items. Given are degrees of freedom, F- and p-values

	denDF	F-value	p-value
CP (g kg−1 DM)
Season	219	4.26	0.0402
Grazing intensity (GI)	2	4.80	0.1597
Ingestion	219	23.25	<0.0001
Season*GI	219	0.69	0.4071
Season*ingestion	219	7.82	0.0056
GI*ingestion	219	0.23	0.6285
Season*GI*ingestion	219	0.07	0.786
ADF (g kg−1 DM)
Season	219	2.74	0.0992
Grazing intensity (GI)	2	2.61	0.2474
Ingestion	219	12.00	0.0006
Season*GI	219	0.93	0.3363
Season*ingestion	219	4.57	0.0337
GI*ingestion	219	0.04	0.8393
Season*GI*ingestion	219	0.37	0.5414
ME (MJ kg−1 DM)
Season	219	17.06	<0.0001
Grazing intensity (GI)	2	16.60	0.0699
Ingestion	219	35.59	<0.0001
Season*GI	219	2.45	0.2111
Season*ingestion	219	7.70	0.0079
GI*ingestion	219	0.02	0.9863
Season*GI*ingestion	219	0.49	0.5116
ivdOM (% DM−1)
Season	219	37.12	<0.0001
Grazing intensity (GI)	2	16.32	0.0561
Ingestion	219	32.47	<0.0001
Season*GI	219	3.03	0.0832
Season*ingestion	219	6.76	0.0100
GI*ingestion	219	0.03	0.8525
Season*GI*ingestion	219	0.52	0.4711

The BI_g had a significantly greater CP concentration in autumn compared to spring. The CP concentration of BI_g was in both seasons greater than the non-grazed BI (Table 7). On the other hand, the ADF concentration was in both seasons lower in the BI_g than in the non-grazed BI (Table 7). In autumn, the non-grazed BI had higher ADF compared to spring (Table 7). The concentration of ME was greater in the BI_g during both seasons compared to the non-grazed BI and greater in spring compared to autumn for both the BI_g and non-grazed BI (Table 7). The same pattern as described for the concentration of ME is visible for the ivdOM (Table 7).

Table 7.

Estimated means (± se) of crude protein (CP, g kg⁻¹ DM), acid-detergent fiber in the organic matter (ADF, g kg⁻¹ DM), metabolic energy (ME, MJ kg⁻¹ DM) and the in vitro organic matter digestibility (ivdOM, % DM⁻¹) for the interaction between season and ingestion averaged over all paddocks. Different letters indicate significant differences among means of ingestion levels within seasons and letters in parentheses indicate significant differences among means of ingestion levels between seasons (p < 0.05)

Quality parameter	Season	Ingestion	Value
CP (g kg−1 DM)	Spring	yes	169 ± 6.13 b (a)
		no	152 ± 6.25 a (a)
	Autumn	yes	204 ± 8.06 b (b)
		no	136 ± 6.11 a (b)
ADF (g kg−1 DM)	Spring	yes	240 ± 5.90 a (a)
		no	267 ± 6.05 b (a)
	Autumn	yes	267 ± 16.92 a (a)
		no	378 ± 11.60 b (b)
ME (MJ kg−1 DM)	Spring	yes	11.75 ± 0.107 b (a)
		no	11.41 ± 0.171 a (a)
	Autumn	yes	11.28 ± 0.165 b (b)
		no	9.94 ± 0.167 a (b)
ivdOM (% DM−1)	Spring	yes	77.4 ± 1.15 b (a)
		no	74.2 ± 1.78 a (a)
	Autumn	yes	71.3 ± 1.25 b (b)
		no	58.3 ± 1.73 a (b)

Selection of Bite Items

The JSI in spring and autumn was affected by the BI (p < 0.0001) but not by grazing intensity (spring: p = 0.9399; autumn: p = 0.8579).

In spring twelve BI showed a JSI > 0 with nine of them representing short BI of which the short G_L_gen_g (0.98 ± 0.274; LSmea ± se) and the short G_veg_g (0.91 ± 0.251) BI had the greatest JSI (Figure 2). This was followed by tall G_L_gen_g (0.84 ± 0.251), short L_gen_g (0.80 ± 0.274), tall G_gen_g (0.79 ± 0.251) and short G_L_veg_g (0.51 ± 0.233) (Figure 2). While the tall BI were all generative, the phenological distribution within short BI was relatively balanced with five vegetative and four generative ones (Figure 2). Of the selected BI with values > 0, seven included grass in them and nine contained legumes (Figure 2). The lowest JSI had the tall G_veg_mix (−1.00 ± 0.251) followed by the tall H_veg_g (−0.98 ± 0.274), tall H_gen_g (−0.94 ± 0.251), short H_gen_g (−0.93 ± 0.251) and tall G_H_veg_g (−0.89 ± 0.251) (Figure 2).

Figure 2.

Estimated means (± se) of the Jacobs selectivity index of grazed bite items in spring and autumn as separated into short (≤ 10 cm) and tall (>10 cm) bite items. For bite item description see Figure 1. Different letters indicate significant differences among means of bite items within seasons (p < 0.05).

In autumn no tall BI showed a JSI > 0 (Figure 2). For the short BI G_veg_g had the greatest JSI (0.86 ± 0.098), followed by G_H_veg_g (0.73 ± 0.098) (Figure 2). All other BI had a JSI < 0 (Figure 2).

DISCUSSION

The choice of BI is a central component of the interaction between herbivores and the sward (Edouard et al., 2010). Understanding the selection process and identifying the underlying factors contributes to an understanding of the sustainable conservation of species-rich grassland. Thereby the bite is the smallest unit that is accessible in the interaction process between the grazing animal and the sward (Bailey et al., 1996). We therefore assessed the BI diversity in terms of the number of unique BI_a actually found in the pastures pre-grazing and in terms of the number of unique BI_g grazed by the cows.

Herbage on Offer and Herbage Nutritive Value on the Paddock-Scale

The grassland was grazed for a total of 266.1 ± 4.06 livestock unit grazing days (LUGD) ha⁻¹ year⁻¹ in grazing intensity M and 194.5 ± 1.23 LUGD ha⁻¹ year⁻¹ in grazing intensity L during 2022. These values are 35.5% and 18.4% lower than the long-term values as reported in a previous study on the same experimental site by Grinnell et al. (2023). Other studies have already shown that species-rich pastures can also produce higher yields than less species-rich pastures, without more undesirable species or bare ground (Baker et al., 2023). Furthermore, we found that only in spring the herbage on offer was greater in the lenient compared to the moderate grazing intensity. In addition, the herbage on offer differed only in the lenient grazing intensity between the seasons as a consequence of lower values found in autumn compared to spring. These values are likely to be due to weather conditions in 2022 being warmer and drier than the long-term average, leading to a corresponding reduction in growth rates (Signarbieux and Feller, 2012). The different seasonal growing conditions further influenced not only the herbage on offer but also its nutritive value, resulting in greater CP and ADF concentrations in autumn compared to spring whereas there were no differences for the ivdOM. However, the faecal samples show that the cows were able to achieve a greater dOM of ingested herbage in spring compared to autumn indicating further that there are seasonal differences in the offered grass sward and that the cows had different feed resources available to them. All cows were able to maintain a good quality diet irrespective of grazing intensity which is underlined by the fact that the digestibility of ingested herbage was the same among the grazing intensities but greater in spring than in autumn. However, all these above-mentioned parameters describe the pasture-scale sward composition although previous studies show that in extensive grassland the herbage is not grazed evenly and a spatial and temporal variability of the vegetation is provided (Dumont et al., 2012; Ludvíková et al., 2015; Tonn et al., 2019).

Bite Item Diversity and Grazing Intensity

The grazing period showed a clear effect on the BI_a and BI_g diversity as we found more BI_a in both stocking intensities in spring compared to autumn. In spring 80% of the BI_a measurements were made up by 14 unique BI. On the contrary, in autumn 80% were represented by only 8 unique BI. At the same time, 80% of BI_g comprised 11 and 2 BI in spring and autumn, respectively. This can be explained by the fact that more generative BI were found in spring compared to autumn (38.9% vs 11.3%) and that there are pheno-morphological changes in the vegetation during the grazing season (Vymazalová et al., 2014). Further, we found a decline of BI_a diversity of 25% under moderate and 22% under lenient grazing towards autumn. This influenced the BI_g as in spring 50% and 53% of the BI_a diversity were grazed under moderate and lenient grazing, respectively. In autumn only 36% and 28% were grazed, respectively. From this it can be concluded that not only were there less BI available in autumn than in spring, but that there was also greater selection from the limited choice in autumn. This is also consistent with the findings of Nota et al. (2024), who found that the relative BI consumption decreases as BI_a abundance declines. Tonn et al. (2019) found on the same experiment a greater plant species diversity in short patches compared to tall patches. This would explain the greater BI_a diversity we found in spring under moderate grazing compared to lenient grazing. Although we do not show the biting rate per step here, we saw fewer bites per step in autumn compared to spring and under lenient grazing in the corresponding data (not shown), which suggests greater selectivity for preferred BI under less intensive grazing. This can further be explained by the greater short:tall-patch ratio in the moderate grazing intensity compared to the lenient one (Tonn et al., 2019) and that cows prefer to ingest short and leafy vegetation (Kohler et al., 2006) so that the greater extent of short patches under moderate grazing made it easier for cows to select their preferred BI. It is likely that short patch BI varied more under moderate grazing than under lenient grazing because of a greater short patch area extent per pasture (Obermeyer et al., 2022). The grazing intensity consequently showed overall no clear effect on the BI_a and also the BI_g diversity, which is in line with the results from other studies (Dumont et al., 2009; Orr et al., 2012). One reason to explain the similar BI_g diversity may be that despite differences in stocking rates between moderate and lenient grazing, each grazing individual might still have a similar BI diversity available to it to make choices. Although the absolute number of BI_a was greater under moderate grazing in spring, the proportion of BI_g relative to the availability was not different to the lenient grazing intensity. This might be due to the negative sward height-bite mass relationship (Ungar, 1996; Rook, 2000) which might limit the use of the entire diversity from constraints to maintain instantaneous herbage intake rate (Prache and Delagarde, 2011). An answer to this requires further studies testing the effect of a large gradient in BI_a diversity under controlled but varied sward heights.

The Herbage Nutritive Value of Bite Items and Their Selectivity

The BI_g had greater CP and ME concentrations as well as a greater ivdOM and lower ADF concentration compared to the non-grazed BI, which implies that the cows selected for the BI with a higher nutritional quality. On average the ivdOM of the BI_g were 1.8% and 13.8% greater than the ivdOM of the BI_a in spring and autumn, respectively. This indicates that it is relatively easy for cows to find BI with an adequate ivdOM in spring, whereas fewer BI reached this level in autumn. However, this shows the potential of extensive grassland systems to supply individual livestock with sufficient herbage nutritive value. Koczura et al. (2019) found similar results for the ivdOM in their study with three dairy cow breeds in mountain pastures. From this it can be deduced that the cows can perceive differences in the quality of the BI and recognize differences which leads to selection. This is further exemplified by the comparison of slopes of the linear regressions of BI_g diversity on ME concentration or BI_a diversity that indicated a shallower increase for the latter one (slope ME = 4.11 vs. slope BI_a = 0.56, p < 0.05) (Supplementary Material S2). This indicates the cows chose a greater BI diversity in response to the BI_a diversity but less than in response to herbage nutritive value of the BI. In connection with the lower herbage nutritive value in autumn, selectivity increased and BI_g diversity declined.

Moreover, the BI selection of cows can change during the grazing season due to changes in nutritional requirements (Rook et al., 2004; Rutter, 2010) but also due to the spatial and temporal variability in the grassland. However, when analyzing the differences between the BI_g and non-grazed BI the CP concentration increased by 21% from spring to autumn in the BI_g while it decreased by 11% in the non-grazed BI, further showing the ability of cows to select for the BI with a higher feeding value in this comparably species-rich grassland. The ivdOM and ME declined slightly as a consequence of greater ADF concentrations over time. Due to the changes in the BI quality in the sward during the grazing season the cows had to adapt their selection process (Cuchillo-Hilario et al., 2018) over the grazing season. Nevertheless, grazing species-rich grassland can also help to improve the production performance of the animals (Beaucarne et al., 2025).

In spring the cows in our study showed an overall preference for BI containing legumes (75% of the BI with a JSI > 0), with their greater CP concentration in accordance with dairy cows studied by Roca-Fernández et al. (2016) and Hesselmann et al. (2025). However, the cows also grazed in tall and generative patches possibly as a reaction to declining regrowth rates of short patches during grazing (Dumont et al., 1995; Ebeling et al., 2020). Tall patches are in general more difficult to ingest (Pauler et al., 2020) due to a lower digestibility (Duru et al., 2008; Gardarin et al., 2014). In our study generative BI were only ingested in spring. The high number of BI_g in the generative stage in spring was caused by flowering legume BI which are still more digestible than vegetative grass (Beever and Mould, 2000). There is also evidence that cows have individual preferences which can change over the course of the grazing season and that they actively select their preferred BI (Hesselmann et al., 2025). In the present study, the cows actively selected BI that included grass (in spring 7 BI and in autumn 2 BI with a JSI > 0). This pattern can be explained by co-occurrence of grasses and legumes in short patch areas (Orr et al., 2012) with dominance of green laminae of grasses resulting from repeated defoliation that attract cows (Koczura et al., 2019).

CONCLUSION

Species-rich grassland allows suckler cows to choose from a variety of distinct bite items, with the herbage nutritive value of the BI as well as the season playing a crucial role in the choice. The animals appear to be highly adaptable to the changing conditions in the grassland and their own needs also influence their feed intake decisions, so that they are able to maintain the quality of the BI they consume. Future research should focus further on the botanical composition of the bite items in terms of plant species and whether the presence or absence of distinct species within bite items have an influence on the animal choices.

Author Contributions

Caroline Siede (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing—original draft, Writing—review & editing), Alina Juch (Data curation, Formal analysis, Investigation, Writing—original draft), Wiebke Pohlmann (Data curation, Formal analysis, Investigation, Writing—original draft), Dina Hamidi (Methodology, Writing—review & editing), Johannes Isselstein (Conceptualization, Funding acquisition, Methodology, Resources, Supervision, Writing—review & editing), and Martin Komainda (Conceptualization, Formal analysis, Methodology, Supervision, Writing—review & editing)

Conflict of Interest statement

The authors declare no conflicts of interest.

Financial Support Statement

This study was supported by funds of the EU H2020 SUPER-G project under the project ID 774124 and it was conducted in the frame of a DFG funded project (FKZ 467394361).