A total of 33 plant species were recorded with perennial ryegrass Lolium perenne present in 93% of fields (Table S2; Fig. 3). Dung density, taken as a proxy for livestock grazing intensity, was unaffected by habitat, management or their interaction (Table 1a). Sward height was significantly (10%) taller in conventionally managed than AES fields (Table 1b; Fig. 2a). Plant species richness was significantly (32%) higher, and rare plant species richness was also significantly (91%) higher in semi-improved than improved grasslands (Table 1c, d; Fig. 2b, c).
A total of 17% of plant community composition was explained by the first two dbRDA ordination axes, suggesting weak structuring indicative of highly even communities (Table 2; Fig. 3). The first dbRDA axis had the lowest cover of perennial ryegrass and the highest cover of, for example, Sphagnum spp. mosses, crested dogstail Cynosurus cristatus and Yorkshire fog Holcus lanatus (= less intensively managed grassland), whilst the second axis had the lowest cover of Sphagnum spp. mosses and the highest cover of ryegrass, Yorkshire fog and broad-leaved dock Rumex obtusifolius (= more intensively managed). Permutational ANOVA tests to assess the significance of the constraining variables within the dbRDA axes suggested that grassland plant community composition varied between management (Fdf = 1,86 = 2.352, p = 0.014), habitat (Fdf = 1,86 = 11.718, p = 0.001) and their interaction (Fdf = 1,86 = 4.011, p = 0.001) (Table S3).
A total of 2,018 individual invertebrates belonging to 70 families (of which 45 were classed as ‘rare’, i.e. < 10% abundance) were identified (Table S1). Araneae (spiders) represented 4% of the total invertebrate families and were most abundant (869 individuals) with the Linyphiidae (money spiders) being the most common family. Coleoptera (beetles) had 12 families (16%) and notably were also common (405 individuals). Diptera (flies) were the most taxon rich order with 24 families (34%) but had lower abundance (355 individuals).
All invertebrate metrics exhibited significant spatial autocorrelation being negatively associated with PCNM2 (Table 3); thus, all plots of estimated marginal means neutralised this effect by fitting PCNM2 at its average value when assessing management and habitat effects (Fig. 2). Accounting for spatial autocorrelation, total and rare invertebrate abundance were both negatively associated with conventional management (Table 3a, b) being 4% and 218% higher in AES fields, respectively, although large % differences were driven by a small actual number increase (Fig. 2d, e), whilst total abundance was positively associated with plant dbRDA axis 2 (Table 3)a, b. More invertebrates were found in plant communities with lower dominance of ryegrass and greater coverage of native grasses, such as Yorkshire fog (Holcus lanatus). Total and rare invertebrate family richness were both negatively associated with conventional management (Table 3c, d) being 17% and 14% higher in AES fields, respectively (Fig. 2f, g). We accept hypothesis 1 that plant and invertebrate diversity is greater in semi-improved than improved grasslands but reject hypothesis 2 that AES impacts would have greatest effect in improved grasslands where extensification had the greatest potential to effect invertebrate diversity and abundance. The effect of agri-environment scheme management was consistent between grassland types.
Agri-environment scheme management and habitat effects
Improved and semi-improved grasslands are dominated by perennial ryegrass (Lolium perenne), which is the most commonly sown agricultural grass with greater cover (near monoculture) in improved grasslands. Such grasslands are typically drained, rejuvenated regularly by ploughing and reseeding, and receive annual applications of fertiliser, usually in the form of slurry. Semi-improved grassland swards, whilst still dominated by ryegrass, also contain native grasses, sedges, mosses and weeds indicative of wetter conditions associated with less drainage. Semi-improved grasslands are infrequently or never reseeded but might receive fertilisation albeit less frequently than improved grassland (given the wetter conditions). Semi-improved grasslands are usually grazed with lower livestock densities which may also encourage species-rich swards by creating a more heterogeneous structure (Marrs et al. 2004; Fraser et al. 2009). In contrast, improved grasslands are either intensively grazed or regularly harvested for fodder in the form of silage creating more homogenous swards. Nevertheless, in this study, dung densities that were measured as a potential proxy for grazing density did not differ significantly between semi-improved and improved grasslands; although positively associated with semi-improved grassland, this was not significant (p = 0.077). It may be that conventional and improved grassland fields were used for forage more often than livestock extensive grazing. If harvested for silage, grass is typically allowed to grow long prior to harvest resulting in increased sward heights as observed here. The resulting competition might reduce plant species richness in favour of fast growing ryegrass (Jefferson 2005). However, the timing of harvesting went unmeasured in this study, and as such, we cannot definitively state that it is a direct result of management actions.
Multivariate analysis suggested that a considerable portion of variation in plant community composition was left unexplained, suggesting that unmeasured environmental variables as well as a number of local scale processes might play a significant role in determining community patterns. Extending the number of quadrats used to sample may alleviate this in the future. Nevertheless, agricultural grasslands are highly even, homogenous communities compared to highly rich and diverse set-aside areas or areas of Special Scientific Interest, so low structuring (limited covariance between species) is to be expected. Nevertheless, statistical modelling suggested that habitat rather than management was the main driver in differences between the plant communities indicative of their intensification over the long-term (grasslands in this study were at least 10 years old). Notably, AES management as part of the Northern Ireland Countryside Management Scheme (NICMS) was in place for a minimum of 5 years, whilst more generally, AESs do not run for more than 10 years (Lennox and Armsworth 2011).
Total and rare invertebrate abundances and family-level richness were unrelated to grassland type (improved or semi-improved) but were all significantly higher in AES than conventionally managed fields associated with swards indicative of wetter conditions with lower dominance of ryegrass and greater coverage of native grasses suggestive of lower intensity management. Although the results have shown clear effects of management on invertebrate biodiversity, we cannot conclude this as a causative relationship without before-and-after data. Other processes, such as seasonal and micro-climate variation in invertebrate communities, may also play a large role in dictating abundance and diversity. Additionally, other measures such as artificial intelligence or genetic metabarcoding techniques could serve as species identification tools which could reduce variation in future studies.
Carabidae ground beetles and Staphylinidae rove beetles were the most abundant families within the Coleoptera being large polyphagous predators with diversified generalist feeding habits commonly associated with open farmland where they provide biological control of pests such as slugs (Tillman et al. 2012). Linyphiidae money spiders contributed to the majority of individuals from the order Araneae, which are widespread due to their ballooning dispersal ability, allowing them to colonise field centres more rapidly than other spiders (Gallé et al. 2018). The Linyphiidae are small-bodied sheet weavers and provide biological control of pests such as aphids. Sphaeroceridae dung flies, Lonchopteridae pointed-wing flies and Phoridae hump-backed flies were most abundant within the Diptera. The majority are saprophagous and are involved in the decomposition of organic matter (Castelli et al. 2020). In the wet, cool upland grasslands studied here, this may help accelerate nutrient cycling, which could otherwise be slower via bacteria or fungal channels alone. Thus, the major invertebrate groups of upland grasslands, which were more taxa-rich under AES management, deliver key (agro)ecosystem services integral to the functioning of healthy farmland, reducing the need for chemical herbicidal and insecticidal inputs. Indeed, reduction of such inputs should be an integral component of AES measures (although it is not here), which in other studies have been associated with increased biodiversity (Fuentes-Montemayor et al. 2011; Anderson et al. 2013; Fritch et al. 2017; Humbert et al. 2021). We observed no interaction between habitat and management on invertebrate abundance or family-level richness suggesting AES management had a consistent effect on semi-improved and improved grasslands.