Data analysis
We assigned the recorded vascular plant species to four morphological groups: short-lived forbs, short-lived graminoids, perennial forbs and perennial graminoids. The full list of the recorded species and their assignment to morphological groups are provided in Appendix 1.
We used two proxies for conservation values: Shannon diversity and naturalness score. We calculated the Shannon diversity of the vascular plant species in each plot. For expressing the naturalness of the vegetation, we classified the recorded plant species into Social Behaviour Types (SBT) according to the classification system of Borhidi (1995). The classification system assigns a naturalness value to each SBT category, ranging from -3 (AC – adventive competitors) to +10 (Su – unique specialist species). We calculated cover-weighted naturalness scores for all plots.
We used two proxies of forage quality: a Hungarian classification system (Balázs 1949) that takes multiple palatability criteria into account and specific leaf area (SLA), which is a proxy for hydrated and more attractive leaves. We classified the species according to their forage quality based on the classification system of Balázs (1949). Forage quality scores range from -3 (toxic plants) to +8 (highly valuable forage plants). The average of the reported SLA values was derived from a regional database (E.-Vojtkó et al. 2020) and for the species not represented in the regional database, the LEDA database (Kleyer et al. 2008) was used. We calculated cover-weighted forage quality and SLA scores for all plots.
We characterised the availability of floral resources with two proxies: flowering period and the cover of insect-pollinated species. Flowering period was calculated as the number of months when a species is flowering using the work of Király (2009). We calculated the community-weighted means (CWM) of flowering period for all plots.
We used repeated measures general linear models (RM-GLMs) for testing the effect of disturbance (two levels: disturbed plots, intact grasslands; fixed factor), year (two levels: 2020, 2021; repeated measures factor) and season (two levels: spring, summer; repeated measures factor). All possible interactions between the factors were included in the models. Dependent variables were: total vegetation cover, cryptogam cover, perennial forb cover, perennial graminoid cover, short-lived forb cover, short-lived graminoid cover, Shannon diversity, naturalness score, forage quality score, specific leaf area, CWM of flowering period and the cover of insect-pollinated plants. All univariate statistics were calculated using the GLM repeated measures command in IBM SPSS Statistics v. 20.0 (Armonk, NY: IBM Corp).
To assess the species composition of the vegetation in the disturbed plots and intact grasslands across the two study years and two seasons, non-metric multidimensional scaling (NMDS) based on percentage cover of the species was calculated using CANOCO 5.0 program (Ter Braak & Šmilauer 2012). For the multivariate analysis, quadrats from the same plot, season and year were averaged.
To investigate if plant abundances differed among i) disturbance types (disturbed plots vs. intact grasslands), ii) years (2020 vs. 2021) and iii) seasons (spring vs. summer), we applied permutational analysis of variance (PERMANOVA). To do so, first we calculated a continuous distance matrix using the raw abundance records of each of the 68 plant species, considering Bray-Curtis distance coefficients. In the following step, we fitted six PERMANOVA tests treating types, years, and seasons as well as their interactions as grouping factors in comparison with abundance distances, employing the PERMANOVA function available in the ”PERMANOVA” package (Vicente-Gonzalez & Vicente-Villardon 2021) of the R statistical programming environment (R Core Team 2021).