Human-induced environmental transformations are leading to biodiversity loss and ecosystem degradation (IPBES 2019).

In rural areas, two contrasting processes are affecting local biodiversity and ecosystem functioning, namely agricultural intensification and land abandonment, especially in remote, less productive areas (Cramer et al. 2008; Plieninger et al. 2014).

While the impacts of intensification have been extensively studied (see e.g., Tscharntke et al. 2005; Rakosy et al. 2022), comparatively little is known about land abandonment. The exclusion of human intervention can trigger 'passive rewilding' (Navarro and Pereira 2015), which facilitates the restoration of natural ecosystems and opens new opportunities for conservation. Rewilded areas can provide habitat for species (Boitani and Linnell 2015); forest regrowth can enhance carbon sequestration (Pugh et al. 2019), promote soil recovery (Pointereau et al. 2008) and help regulate the hydrological cycle (Stoate et al. 2009). However, while potentially beneficial, these processes often come at the expense of semi-natural open habitats, particularly pastures and grasslands, which are often of great conservation concern (Carboni et al. 2015).

This is especially true for semi-natural grasslands, which are secondary formations, created and maintained through centuries of traditional and low-intensity human practices (Valkó et al. 2018) and have become part of the traditional agricultural landscape. Among the semi-natural grassland types, dry grasslands are unique species-rich assemblages (e.g., Wilson et al. 2012) that harbour a high diversity of plant species and provide habitat for higher trophic level species, especially arthropods (Zulka et al. 2014; Labadessa et al. 2015). Not only do they have high conservation value, but they are also recognised for providing regulating, provisioning, and cultural ecosystem services (Tscharntke et al. 2005; Bengtsson et al. 2019).

Like all other semi-natural habitats, semi-natural dry grasslands are particularly vulnerable to abandonment (Schrautzer et al. 2009). After abandonment or under use, semi-natural grasslands undergo natural vegetation dynamics. Although successional patterns can be stochastic (Chesson 2012) and vary from site to site, they generally result in species turnover and/or changes in the three-dimensional architecture of the plant community (Pickett et al. 2003). In temperate Europe, the pattern of species turnover is associated with the loss of specialised grassland species (Gustavsson et al. 2007), the rapid establishment and spread of competitive graminoids (Carboni et al. 2015), and the increase of herbaceous and shrub species typical of forest edges, and tree seedlings (Schrautzer et al. 2009).

Several studies have investigated changes in plant species composition and structure during the succession process (e.g., Bonanomi and Allegrezza 2004; Habel et al. 2013), as well as the relationships between plant community composition and structure and higher trophic level species (e.g., Elliott et al. 2023), both from taxonomic (e.g., species richness) and functional perspectives (e.g., life history traits or ecological strategies; e.g., Kelemen et al. 2017). Overall, research has shown that abandoned grasslands have lower diversity than extensively managed semi-natural grasslands (Pykälä et al. 2005; Klimek et al. 2007). Conversely, species richness of higher trophic levels has been reported to increase, at least in the early stages of succession (Öckinger et al. 2006), as higher and structurally heterogeneous plant communities facilitate the occurrence of more diverse arthropod communities (Kruess and Tscharntke 2002), which respond strongly to habitat changes (Öckinger et al. 2006; Colom et al. 2021).

Although the relationship between grassland succession and plant and animal species richness and composition has been widely studied (e.g., Habel et al. 2013; Elliott et al 2023), little is known about how grassland succession affects interactions between species belonging to different trophic levels. Among the multiple biotic interactions, animal-mediated pollination has received considerable attention in semi-natural dry grasslands, as they provide resources (i.e., nectar and pollen), breeding, nesting and overwintering habitats for pollinators and effectively promote their conservation (Fantinato et al. 2021). Furthermore, plant-pollinator interactions have been shown to contribute to the assemblage and maintenance of grassland communities (Benadi and Pauw 2018; Fantinato et al. 2018, 2019a). Therefore, plant species turnover and changes in vegetation structure following grassland abandonment are likely to have noticeable effects on species dynamics, pollination interactions and pollination service (Fantinato et al. 2019b). Plant species turnover may lead to changes in floral resources availability on grasslands, e.g., by reducing the total number of entomophilous species by favouring competitive wind-pollinated grasses. Changes in plant species composition could thus have a major impact on pollination, as changes in the type or quantity of floral rewards pose a major threat to pollinators (Goulnik et al. 2021). Understanding the effects of grassland abandonment on floral resources and pollination interactions can therefore help inform management plans for semi-natural grasslands that ensure the maintenance of a diverse plant and pollinator community and the services they provide.

In light of the above, this research aimed to answer the following questions: (i) How does the quantity of different types of floral resources (i.e., number of flowers, nectar volume, and number of pollen grains) change during succession of semi-natural dry grasslands? (ii) How does the species richness of plants and pollinators and their interactions change during the succession of semi-natural dry grasslands?

The sampled plots had different cover of plant species of forest edges, varying from 0.33% to 90.21% (mean ± SD; 29.26% ± 24.21%), indicating that the dry grassland communities recorded in the sampled plots were at different stages of succession. Plant species of forest edges that firstly occurred in the plots were Brachypodium rupestre, Asparagus acutifolius, Teucrium chamaedrys, Geranium sanguineum and Cervaria rivini. As soon as succession progressed (namely the cover of plant species of forest edges increased), seedlings of shrubs and trees also occurred, such as Cornus sanguinea, Rosa canina, Spartium junceum, Fraxinus ornus, and Quercus pubescens.

The quantity of floral resources varied greatly between the sampled plots. The number of flowers varied from 0.00 to 18,512.60 flowers per plot (mean ± SD; 465.38 ± 1570.45), the volume of nectar varied from 0.00 µl to 2,885.28 µl (mean ± SD; 75.44 µl ± 283.52 µl), while the number of pollen grains varied from 0.00 to 912,733,383.23 (mean ± SD; 2,148,660.32 ± 8,371,645.95).

The relationship between the number of flowers, the volume of nectar and the number of pollen grains with the cover of plant species of forest edges were all significantly hump-shaped, suggesting that regardless of the type of floral resource, all peaked at intermediate values of cover of plant species of forest edges (Table 1; Fig. 2).

Overall, 42 animal-pollinated plant species and 76 pollinator species were recorded in sampled plots. The richness of animal-pollinated plants in bloom per plot varied from 0.00 to 7.00 (mean ± SD; 1.51 ± 1.63); most frequent animal-pollinated plants in bloom were Thymus pulegioides (59% of sampled plots), Helianthemum nummularium subsp. obscurum (56%), Globularia bisnagarica (52%) and Stachys recta (48%). The richness of pollinator species varied from 0.00 to 8.00 (mean ± SD; 1.00 ± 1.50); the most frequent pollinator species were Apis mellifera (63% of sampled plots), Bombus hortorum (56%), Epicometis hirta (48%), Episyrphus balteatus (41%) and Eristalis tenax (41%). The number of pollination contacts varied from 0.00 to 16.00 (mean ± SD; 1.73 ± 3.04). The most visited plant species per plot were Globularia bisnagarica (mean ± SD; 7.70 ± 8.36), Potentilla pusilla (mean ± SD; 4.66 ± 2.88), Pilosella officinarum (mean ± SD; 2.85 ± 3.07) and Geranium sanguineum (mean ± SD; 2.15 ± 2.65).

The relationship between the richness of animal-pollinated plants in bloom with the cover of plant species of forest edges was significantly hump-shaped (Table 1; Fig. 2), suggesting that the peak of richness of animal-pollinated plants in bloom was at low-intermediate values of cover of plant species of forest edges. Finally, the richness of pollinator species and the number of pollination contacts were significantly related to the same covariates, i.e. the probability of absence of pollinator species and of pollination contacts was negatively associated with the number of flowers (ln-transformed; Table 2; Fig. 3), with the probability of absence decreasing to zero once at least ten flowers were present, while the expected richness of pollinator species and the expected number of pollination contacts were positively related to the richness of animal-pollinated plants in bloom (ln-transformed; Table 2; Fig. 3). In other words, the probability of pollinator presence and pollination contacts depended on the number of flowers in the plot, while the expected richness of pollinators and the expected number of pollination contacts depended on the richness of animal-pollinated plants.

The abandonment of traditional management practices has been shown to lead to significant changes in the environmental characteristics and structural attributes of dry grassland communities (Valkó et al. 2018).

In the present study, we have shown that succession of dry grasslands after abandonment affects structural properties of plant communities and has critical implication for a crucial function in terrestrial ecosystems, namely animal-mediated pollination, even in correspondence of the first dynamic stages.

Patterns of animal-mediated pollination interactions at community level are related to the type and quantity of floral resources (Fantinato et al. 2021). Our results have shown that the quantity of floral resources provided at the community level changed during dry grassland succession. In particular, the quantity of floral resources peaked at low-intermediate values of percentage cover of plant species of forest edges, regardless of the type of floral resource. Floral traits such as the number of flowers, the volume of nectar and the number of pollen grains produced have been shown to be genetically regulated (e.g., Tsuchimatsu et al. 2020), but they also respond to local environmental conditions. Roth et al. (2023), for example, have shown that plants produce fewer flowers per plant under drought conditions. Furthermore, the flowers tend to become smaller (Kuppler and Kotowska 2021), bloom for a shorter time (Turner 1993) and produce less nectar (Phillips et al. 2018). In addition to soil moisture, soil nutrients have also been shown to influence floral traits and are positively correlated with flower size, nectar concentration and volume, and pollen grain number (Vaudo et al. 2022). In the early stages of grassland abandonment, the accumulation of litter leads to increased soil moisture and nutrient availability (e.g., Hassan et al. 2021), which can promote flower production as well as nectar volume and pollen grain number (e.g., Plos et al. 2023; Roth et al. 2023). However, as succession progresses, the spread of competing graminoids such as Brachypodium rupestre (Bonanomi and Allegrezza 2004) may ultimately lead to the exclusion of animal-pollinated plant species from the community, reducing the quantity of floral resources.

Dry grasslands are biodiversity hotspots harbouring a large diversity of animal-pollinated species (Fantinato et al. 2021). In our study, we have shown that succession following the abandonment of traditional management practices affects the richness of animal-pollinated species in bloom. Consistently with previous studies that found a hump-shaped relationship between vascular plant species richness and grassland succession (Kesting et al. 2015), we showed that the richness of animal-pollinated plants peaked at low-intermediate levels of cover of plant species of forest edges. In the early stages of grassland succession, the presence of plant species of forest edges in the community may initially increase overall plant richness. However, once dense stands of clonally growing graminoids dominate the community, the richness of plant species rapidly decreases. In addition, the accumulation of organic matter and increasing soil moisture promoted by grassland abandonment can alter soil fertility and nutrient cycling over time, further affecting the suitability of the habitat for grassland species (Deng et al. 2016). As the succession process progresses, increasing woody plant density and abundance reduces the overall grassland species diversity (Schrautzer et al. 2009).

Interestingly, the richness of pollinator species and the number of pollination contacts were not directly related to the percentage cover of plant species of forest edges, but rather indirectly, as they were significantly related to the number of flowers and the richness of animal-pollinated plants in bloom. Our results showed that the probability of absence of pollinators and of pollination contacts decreased to zero once at least ten flowers were present. In other words, the probability of pollinator presence and pollination contact was significantly related to the presence of flowers. Although the presence of floral resources such as nectar and pollen may ultimately lead pollinators to develop floral fidelity and therefore continue to visit flowers of the same species because they have learned that floral resources are present (Brosi 2016), it is attractive features of flowers (e.g., flower size, shape, colour, floral scent, etc.) that initially enable pollinators to locate floral resources (van der Kooi et al. 2023). This could explain why neither the volume of nectar nor the number of pollen grains influenced the probability of absence (or, conversely, presence) of pollinators and of pollination contacts.

The richness of plant species then significantly influenced the richness of pollinator species and the number of pollination contacts. These results can be explained by the fact that taxonomic richness usually positively correlates with functional richness (e.g., Dovrat et al. 2021), i.e., the richer the community of animal-pollinated plants, the higher the probability of plant species that differ in their floral traits and in the type and quantity of floral resources. This in turn would attract pollinators with different feeding preferences and flower handling abilities, ultimately increasing the richness of pollinator species and the chance of pollination contacts.

The abandonment of traditional management practices and the subsequent succession of dry grasslands towards forest edges has been shown to lead to biodiversity conservation issues and ecosystem changes, such as changes in soil properties and grassland productivity. In the present study, we have shown that the succession of dry grasslands also affects animal-mediated pollination, even in correspondence of early stages. Although animal-pollinated shrubs (e.g., Cornus sanguinea, Rosa canina, Spartium junceum) and trees (e.g., Fraxinus ornus) can provide floral resources, they only flower for a limited period of the year (usually in early spring) and rarely for more than a month. Beside showing restricted blooming periods, in some cases, shrubs and trees show high degrees of pollinator specialisation, namely, their floral morphology can be effectively handled only by a narrow group of pollinator species. This is the case, for example, of S. junceum, which shows mass blooming in late spring. However, the high degree of pollination specialisation of S. junceum, resulting from the complexity of its floral morphology and the thickness of its flowers, which allow only a few pollinators to forage, means that it occupies a peripheral position in the network of pollination interactions (i.e., it cannot sustain a broad community of pollinators on its own). Ultimately, this means that the contribution of dry grasslands to pollinator conservation cannot be replaced by shrub and forest communities.

Our results could provide useful insights for planning management practices that optimise the conservation of plants and pollinators in dry grasslands as well as pollination interactions.

The hump-shaped relationships that both the richness of animal-pollinated plants and the quantity of floral resources evidenced with the cover of plant species of forest edges suggest that the first dynamic stages ensure both an increase in plant species richness and in the quantity of floral resources supplied. Such a situation cannot be achieved through complete abandonment or even irregular management, that over time predictably lead to passive rewilding and grassland loss. Rather, improving grassland heterogeneity, leaving spatially scattered small areas where the frequency of mowing is temporarily slowed down to create conditions of early succession, can increase the number of niches for plant and animal species and improve the pollination function in dry grasslands.

This approach allows to create and maintain conditions of early succession, that contribute to increase the number of niches for plant and animal species and improve the pollination function in dry grasslands.