Haymaking complemented by moderate disturbances can sustain and restore species‑rich alpine to subalpine grasslands

The high plant diversity in alpine to subalpine grasslands is threatened by the abandonment of land use. In addition, changing environmental conditions might lead to vegetation shifts even when traditional land use is maintained, as observed in grasslands in Switzerland during the last decades. Maintaining and restoring the diversity of such grasslands might therefore require modified management methods. We conducted a six-year experiment to assess the responses of plant species richness, mean ecological indicator values, and vegetation composition to five management treatments, including scraping as additional management measure: haymaking (in autumn), haymaking complemented by scraping (i.e. manual raking) in autumn, haymaking complemented by scraping in spring, only scraping in spring, and abandonment of land use. We hypothesized that haymaking complemented by scraping in either season would remove additional biomass and increase species richness by creating open patches that can reduce inter-specific competition and promote species establishment. We found positive effects of haymaking complemented by scraping on plant species richness and habitat quality, indicated by the increased mean indicator value for light. Abandonment showed the opposite effects and increased mean indicator values for nutrients. Interest-ingly, haymaking combined with scraping in autumn promoted the development of the vegetation towards the composition similar to the resident vegetation type. Our findings show that extensive land use is essential to maintain species-rich alpine to subalpine grasslands. Further, they imply that modified land use can compensate for the negative developments such as reduced habitat quality and species richness caused by environmental changes and help restore the vegetation.


Introduction
Central European grasslands, in particular mountain grasslands that are dominated by herbaceous species and are situated at high elevations (Schirpke et al. 2017), harbor a high biodiversity of various taxa (Dengler et al. 2014;Boch et al. 2020).The species diversity and composition of these grasslands developed during a long period of traditional low-intensity land use (Stöcklin et al. 2007;Ellenberg and Leuschner 2010;Hejcman et al. 2013).However, this diversity has become threatened, and many grassland-associated plant species have decreased in abundance while only few, mostly dominant species increased in abundance, leading to homogenization of communities over the last century (Jandt et al. 2022).The main reasons for biodiversity decline and altered species composition in Central European grasslands are land-use intensification including fertilization, higher stocking densities and increased mowing frequencies (Tasser and Tappeiner 2002;Fischer et al. 2008;Allan et al. 2014;Gossner et al. 2016;Visconti et al. 2018;Weber et al. 2023) and the abandonment of land-use, often in areas that are remote, isolated or difficult to access with machinery, such as steep slopes (MacDonald et al. 2000;Orlandi et al. 2016;Valkó et al. 2018;Elliott et al. 2023).Traditional land use in mountain regions also aligns with cultural values and plays a vital role in the support of local livelihoods (Montenegro-Díaz et al. 2022).It comprises either pasturing or haymaking.The traditional method of haymaking involves manually mowing by scythe and subsequent raking for baling the hay.Today, these traditional methods have been largely replaced by methods involving machinery and are only applied on very steep mountain slopes where it is not possible to use any machinery.In addition, the use of these grasslands was increasingly abandoned because their management is very labor intensive, often dangerous (e.g.wild hay meadows on very steep slopes) and was insufficiently subsidized by federal environmental agencies (Stöcklin et al. 2007).Abandonment promotes forest encroachment leading to changes in environmental conditions (e.g.litter accumulation, shading, microclimate).This in turn can lead to shifts in species composition towards just a few dominant species and the loss of typical grassland species, reflected by a transition from one vegetation type to another (MacDonald et al. 2000;van den Berg et al. 2011;Orlandi et al. 2016;Valkó et al. 2018;Boch et al. 2019a;Zehnder et al. 2020).In the past decades, Switzerland made investments for the environment in the agricultural sector including management regulations, contracts with farmers and direct payments to compensate lower yield in extensively managed grasslands.This should help preventing further decreases of area and habitat quality of valuable sites (BAFU 2017).
However, global-change-induced shifts in environmental conditions might additionally cause species losses and altered species composition, leading to transitions to different vegetation types, even when traditional land-use regimes are maintained (e.g.Dise et al. 2011;Lyons et al. 2023).In temperate regions, changes towards more fertile conditions because of large-scale fertilization and atmospheric nitrogen deposition have often been reported (Roth et al. 2019), promoting fast-growing and dominant species and leading to vegetation shifts.This development is challenging and can require additional measures, such as adapted land-use regimes, when the goal is to maintain and restore the species composition and the diversity of particular vegetation types (Lyons et al. 2023).Besides drastic measures, one opportunity can be the introduction of an additional management event or modified management methods, such as an additional rather early cut or sward disturbance, either with machinery or by a late low intensity grazing event.This can help remove biomass and nutrients, reduce the growth of dominant and fast-growing plant species early in the season, prevent litter accumulation, create open patches, and activate the soil seedbank (Bobbink and Willems 1993;Turner et al. 2006;Foster et al. 2007;Klaus et al. 2018;Zhang et al. 2020;Lyons et al. 2023).However, an additional cutting event is rather cost intensive, as it requires another visit to the site over several years.In addition, it might lead to species losses by reducing the time to develop flowers and seeds of some species.In contrast, methods of disturbing the sward and topsoil, such as modified scarifying of the lawn, open up the soil and remove litter and dense layers of moss as the material is subsequently removed.This could be a low-cost alternative because it can complement mowing and potentially be applied on the same day after the mowing.This sward disturbance can be repeated after few years as required.It can be applied either manually by scraping the ground with a rigid rake or using machinery such as a harrow in better accessible sites.However, effects of a sward disturbance approach on plant species richness and composition have not yet been studied experimentally in mountain grasslands.In addition, it has not been tested experimentally whether mowing complemented by scraping combines the positive effects of the two approaches, leading to greater plant species richness.Finally, it is unknown whether the timing of scraping is important, i.e. whether it has different effects when applied in spring or in autumn.It can be hypothesized, that scraping in spring in addition to a mowing event in autumn might remove more biomass and nutrients than a combined management event in autumn.
In grassland-related studies, species richness is a commonly used measure of biodiversity and habitat quality.However, total species richness is not necessarily a suitable indicator to reflect environmental changes, as values mainly peak under environmental conditions with intermediate productivity and disturbance (Grime 1973;Huston 2014;Bergauer et al. 2022).In addition, temporal patterns in species richness may be biased, as such measures are prone to observer differences (Archaux 2009;Burg et al. 2015), and it has been shown that species lists of vegetation plots that were compiled by different observers usually differ from each other to a certain degree (Kapfer et al. 2017;Boch et al. 2022).Further, it has been shown that prior experience and training in conducting vegetation surveys can particularly affect species numbers (because of a greater ability to detect and identify species) and cover estimates over time (reviewed by Morrison 2016).Mean ecological indicator values are an alternative tool that is widely used and considered appropriate to characterize site conditions and temporal changes of habitat quality in vegetation plots (Diekmann 2003;Bergamini et al. 2019).Ecological indicator values describe the realized niche of a plant taxon on an ordinal scale (Landolt et al. 2010).Averaging values over all taxa at a given site yields information on the site's environmental conditions (Tölgyesi et al. 2014).Mean indicator values have been proposed to be relatively robust to various samplingrelated shortcomings, including relocation error of repeated plots (Boch et al. 2019b) and observer differences in species lists and training effects (Futschik et al. 2020;Boch et al. 2022), and they can thus provide a reliable estimate of ecological changes, such as temporal shifts in vegetation composition.In grasslands, nutrient-poor and high-light conditions are often associated with low competition and large numbers of specialized species that can cope with unfavorable conditions, indicated by high mean stress indicator values.Further, high mean ruderality indicator values can indicate disturbance-related establishment of pioneer species (Landolt et al. 2010;Küchler et al. 2018).For biodiversity conservation in various grassland types, maintaining and restoring such conditions have been proposed to be essential (e.g.Härdtle et al. 2006;Wamelink et al. 2005).
In the alpine to subalpine grasslands of the Schynige Platte Alpine Botanical Garden in the Bernese uplands (Switzerland), species numbers have decreased and vegetation changes have occurred during the last decades (Hegg and Schaffner 2012).Even though traditional haymaking has been largely maintained since the establishment of the garden in 1927, in order to convey the cultural values of this form of land use to the garden visitors, i.e. one cut per year in autumn (scythe haymaking) with subsequent raking for baling the hay, the character of the sites has visibly changed increasingly towards nutrient-rich grasslands whose composition resemble what has been described as Polygono-Trisetion or Adenostyletalia communities (see Delarze et al. 2015).Possible reasons might be the climatewarming driven prolonged vegetation period, but also nitrogen deposition that are reported to be higher than threshold levels in susceptible ecosystems, especially at high elevation, leading to changes in species composition and turnover in Swiss grasslands in the past decades (Roth et al. 2019).Here, we investigated whether manual scraping with a rigid rake might serve as sward-disturbance to remove litter and the dense moss cover and method to complement haymaking as a measure to enhance plant diversity and support the alpine grassland vegetation communities that were present when the Alpine Botanical garden was established.We conducted a six-year experiment, testing the response of plant species richness and mean ecological indicator values to management treatments ranging from combined haymaking and scraping, to haymaking or scraping alone, to complete abandonment.We further tested whether the treatments led to changes in species composition, i.e. shifts to different vegetation types.
We addressed the following questions: 1) How do plant species richness and mean indicator values respond to abandonment and modified haymaking methods that include scraping?2) Do abandonment and modified haymaking methods that include scraping as sward-disturbance method lead to shifts in plant species composition towards different vegetation types?
We expected that scraping as additional disturbance treatment would remove more biomass and nutrients than haymaking alone, leading to restoration of the Caricion ferrugineae communities.In addition, we predicted that scraping would reduce inter-specific competition and create open patches, promoting seed germination and species establishment.

Study site
The Schynige Platte Alpine Botanical Garden is situated in the Bernese uplands (Switzerland; 46°39′12″N, 7°54′40″E) at 1980-2010 m a.s.l.The mean annual temperature is approximately 1 °C, and the mean annual precipitation is about 1800 mm (Spiegelberger et al. 2006;Hegg and Schaffner 2012).The growing season starts after snowmelt, usually in the beginning of June.In the beginning of the nineteenth century the site was closely above the natural tree line (Lüdi 1936).Besides the cultivated plant communities in the garden beds, the composition of the resident alpine to subalpine grassland communities of the Schynige Platte is similar to what has been described as the Caricion ferrugineae alliance, intermixed with species which had been attributed to the Seslerion variae, Nardion strictae and Polygono-Trisetion alliances (Lüdi 1948).The grasslands are unfertilized and have been cattle grazed before the garden was established.After the establishment of the garden in 1927, the grasslands were mown once per year in autumn.Mowing was done with a scythe, with subsequent raking for haymaking.From 1980 to 1996 a string trimmer was used for mowing.This reduced species richness and increased grass cover, possibly because the string trimmer crushed the cut biomass into pieces so small that they were not fully removed during haymaking and in turn formed a thick litter layer on the ground.Since 1996, the grasslands have again been mown by scythe, with subsequent raking in September (Hegg and Schaffner 2012) using a traditional wooden hay rake for haymaking in order to convey the cultural values of this form of land use to the garden visitors.

Experimental design and vegetation sampling
In 2009, we selected six grassland sites in the Alpine Botanical Garden that were large enough for our purposes.The sites had different expositions (one north, three east, two south and one west exposed slope) and varied from 25° to 50° slope inclination.In each site we established five 2 m × 2 m plots, which we permanently marked below ground with metal pegs.In each of the sites we then randomly assigned five treatments to the five plots: (1) haymaking in September, (2) haymaking complemented by scraping in September, (3) haymaking in September complemented by scraping in June, (4) only scraping in June, and (5) abandonment of land use, i.e. no haymaking or scraping (Fig. 1).For scraping, we used a heavy-duty bow rake for soil, which removes litter and dense bryophyte layers and also penetrates the uppermost 0.5 cm of the soil.We maintained these treatments from September 2009 until September 2015.Each year in July (2010)(2011)(2012)(2013)(2014)(2015), we recorded all vascular plant species in all 30 plots and estimated the percentage of the area covered by each species.

Characterization of vegetation types
We assigned all relevés from all consecutive years to vegetation types, using the modified scoring system of Eggenberg and Bornand (2023) with the Vegedaz program (Küchler 2023).This scoring system creates a value indicating the similarity of a vegetation plot to a particular vegetation type based on the sum of different values for different diagnostic species that are present in a plot.Delarze et al. (2015) listed taxa (vascular plants, algae, bryophytes, lichens) that are representative of one or more vegetation types occurring in Switzerland and at the same time distinguished four types of diagnostic species: (1) character species, i.e. species that mainly occur in that particular vegetation type (code of vegetation type printed in italic font), (2) dominant character species, i.e. character species that often dominate the vegetation (code of vegetation type printed in italic bold font), (3) accompanying species, i.e. species that frequently occur in the particular vegetation type (code of vegetation type printed in normal font), and (4) dominant accompanying species, i.e. species that frequently occur in the particular vegetation type and dominate the vegetation (code of vegetation type printed in bold font).When applied to a species list associated with a vegetation plot, the scoring system counts four points for each character species, eight points for each dominant character species (if dominant in the plot), one point for each accompanying species, and two points for each dominant accompanying species (if dominant in the plot).The system uses an abundance threshold of 10% cover to decide if a species is dominant in the plot.For example, a species marked as dominant and having a plot cover <10% is given four points when listed as a character species but one point when listed as an accompanying species, and a species marked as dominant and having a plot cover >10% is given eight points when listed as a character species but two points when listed as an accompanying species.To differentiate the abundances of species within the plots, we slightly modified this method by linearly weighting the species abundance.For instance, a character species with 50% cover was given six points (instead of eight), while one with 75% cover was given seven points.In this way, we calculated a score for all relevés for all consecutive years for the four major vegetation types which had been described as Caricion ferrugineae, Seslerion variae, Polygono-Trisetion, Mesobromion erecti and Nardion strictae (Delarze et al. 2015).

Mean ecological indicator values
We calculated mean ecological indicator values for nutrients, light, temperature, moisture, competition, stress and ruderality for each plot and for all consecutive years based on Landolt et al. (2010), using the Vegedaz program (Küchler 2023).In contrast to the ordinal scale from one to five for nutrient and light values, the information on competition, stress and ruderality in Landolt et al. (2010), which is based on Grime's C-S-R triangle (Grime 1974), was transformed to a numerical value ranging from 0 to 3. For instance, a species with 'ccc' in Landolt et al. (2010), indicating the strongest competitive ability, was assigned a value of 3 for competition and a value of 0 for both stress and ruderality, whereas a taxon with 'css' was assigned a competition value of 1, a stress value of 2, and a ruderality value of 0. We used the arithmetic mean of the indicator values, rather than the cover-weighted means, as results of the two approaches were qualitatively similar.

Statistical analysis
We performed all statistical tests using R version 4.2.2 (R Core Team 2022).We used linear mixed models with orthogonal contrasts (lme4 package; Bates et al. 2015) to test whether there were changes in species richness, mean indicator values, TypoCH scores of vegetation types and the cover of individual species (only species with both, an overall frequency of ≥25% and a mean cover of ≥1% across plots and years) across time and among the treatments, and to assess whether these changes differed significantly from the mean of all other treatments.We therefore computed the differences between the relevés of each year and the overall mean of the same year, and tested these differences for trends, separately for the treatments.For illustration purposes, we added regression lines to the line plots (Fig. 2), and the slopes of these regressions to the result Tables (Tables 1, 2; Tab.S1).The reader may note that these slopes might differ from the trend values of the linear mixed models (Tables 1, 2; Tab.S1).

Results
Across all years and plots, we recorded 176 vascular plant taxa in total (excluding uncertain species identifications).The number of species per plot ranged from 32 to 71, with an average of 47.0 (±8.8 SD).However, we recorded higher species numbers over time (average +1.3 species per plot when comparing the species numbers of first and last year).

How does plant species richness respond to abandonment and modified haymaking methods?
Overall, species richness increased stronger in plots combining scraping in spring and haymaking in autumn than in the other treatments (average +2.2 species per plot over time; Fig. 2).A similar trend of increasing species richness in plots combining haymaking and scraping in autumn (average +1.9 species) was statistically not significant (P > 0.05).In addition, there was an increase in species numbers in abandonment plots that was lower than the mean (average +0.7 species), but this difference was only marginally significant (Table 1, Fig. 2).

How do mean indicator values respond to abandonment and modified haymaking methods?
Mean indicator values were hardly affected by the treatments.The mean indicator for light remained constant in plots combining haymaking and scraping in autumn but was significantly greater than the mean because all other treatments showed a decreasing trend.In addition, the mean  indicator value for nutrients increased significantly in abandoned plots.All other relationships between treatments and mean indicator values did not significantly differ from the mean; this included decreasing but non-significant trends of the mean indicator value for nutrients in plots of both treatments combining haymaking and scraping (Table 2, Fig. 3), as well as decreasing but non-significant trends of stress and increasing competition in abandoned plots (Table 2, Fig. 4).In addition, disturbance by scraping did not lead to the establishment of additional ruderal species, as indicated by the non-significant relationships between the treatments including scraping and the mean indicator value for ruderality (Table 2, Fig. 4).

Do abandonment and modified haymaking methods lead to shifts in plant species composition towards different vegetation types?
Despite visually obvious changes towards taller species and higher grass cover in the plots in which biomass was not removed, i.e. the abandonment and scraping only treatments, our model revealed few significant effects of the treatments on the vegetation composition, i.e. regarding shifts of the cover of single species (Tab.S1) and shifts among vegetation types (Table 1, Fig. 5).However, while haymaking combined with scraping in autumn led to the development of the vegetation towards composition similar to what has been described as the Caricion ferrugineae alliance, abandonment had the opposite effect and reduced species which has been mentioned to be associated with Caricion ferrugineae.In addition, scraping applied alone hampered species which have been attributed to what has been described as the Polygono-Trisetion and Nardion strictae alliances (Table 1, Fig. 5).Regarding the cover shifts of single species (Tab.S1), we found that abandonment promoted long leaved graminoids and competitive broad-leaved forbs (e.g.Carex ferruginea, Festuca rubra, Gentiana lutea), while the cover of smaller statured species declined (e.g.Globularia nudicaulis, Homogyne alpina).Scraping alone had no positive effects on the cover of any species but also negatively affected the cover of many small-statured and early flowering species with low disturbance tolerance (e.g.Anthyllis vulneraria, Hippocrepis comosa, Ligusticum mutellina, Soldanella alpina, Trollius europaeus).Interestingly, the two treatments combining mowing and scraping consistently reduced the cover of species with low mowing tolerance and the ones that were promoted by abandonment.However, while scraping and mowing in autumn did not promote the cover of any species, scraping in spring and mowing in autumn positively affected the cover small statured forbs (e.g.Helianthemum nummularium, Homogyne alpina, Scabiosa lucida).

General diversity patterns
With an average of 47 species per 4 m 2 plot, the alpine to subalpine grasslands in our research area can be considered species rich, compared with the average of 43 species in 10 m 2 alpine to subalpine grassland plots of comparable grasslands in Switzerland (see Elyno-Seslerietea class in GrassPlot Diversity Explorer: https:// edgg.org/ datab ases/ Grass landD ivers ityEx plorer; N = 350; Biurrun et al. 2021).Also Dengler et al. (2020) reported an average of about 43 species in 10 m 2 plots from alpine grasslands across the Palaearctic biogeographic realm.

Possible observer effects and limitations of detecting temporal changes of habitat quality
We found an overall increase in species numbers over time in all treatments, which can likely be attributed to an observer training effect (e.g.Morrison 2016).It might well be that we got accustomed with the local flora during working many years in a row at the site, which might have improved our ability to better detect local species and to discover potentially new species in our plots.As mean ecological indicator values have been shown to be relatively robust to various sampling-related shortcomings, including pseudoturnover, i.e. observer differences in species lists, and training effects (Futschik et al. 2020;Boch et al. 2022), we consider these values the most reliable estimate of ecological changes including temporal shifts in vegetation composition, particularly in species-rich vegetation types like in our experiment on Schynige Platte.Nevertheless, we believe that our approach of comparing changes in species numbers between treatments over time gives a reliable representation of changes because the observer training effect can be considered being equivalent among treatments.After six years, we found only few significant effects on species richness and habitat quality, as indicated by the changes in mean ecological indicator values.We are aware that having more replicates might have resulted in more pronounced treatment effects.In addition, only six years is a rather short experiment duration to find clear treatment effects under alpine to subalpine climatic conditions with a very short vegetation period.Although the vegetation in the abandoned and only scraped plots, i.e. plots with no biomass removal, changed very obviously visually, our vegetation analysis revealed few effects of the treatments on the vegetation composition.Thus, we found only few transitions among vegetation types.

Response of plant species richness, mean ecological indicator values, and species composition to abandonment and modified haymaking methods
In Western Europe, most grasslands below the alpine zone are secondary grasslands, originating from timber harvesting, forest clearing to establish arable fields and permanent grasslands, or grazing of woodlands (Boch et al. 2020).(Bohn et al. 2004;Boch et al. 2020).
The abandonment of land use has therefore been identified as one of the major threats for grassland plant biodiversity, particularly in secondary grasslands of temperate regions (Visconti et al. 2018;Dengler et al. 2020).With abandonment, no biomass is removed anymore, and successional dynamics lead to the promotion of a small number of competitive and often dominant plant species that are often characterized by high nutrient indicator values.This was confirmed by our findings of abandonment increasing the cover of long leaved graminoids and competitive broad-leaved forbs and decreasing the cover of smaller statured species.
A possible reason might be the increasing competition for light that can hinder the germination and establishment of light-demanding grassland species and cause the competitive exclusion of small-statured, specialized and often threatened species in the sub-canopy.This can even lead to decreasing plant species richness in the longer term (Maurer et al. 2006;Deák et al. 2016;Valkó et al. 2018;Boch et al. 2019a;Zehnder et al. 2020;Elliott et al. 2023).In calcareous grasslands in the UK, van den Berg et al. ( 2011) found increased mean nutrient indicator values after land-use extensification.Similarly, Boch et al. (2019a) found abandonment and wood encroachment to be positively related to mean nutrient indicator values and negatively related to the proportion of specialist species.These patterns were also reflected in our findings: in comparison to the treatments combining haymaking and scraping, abandonment tended to negatively affect species richness and increase mean nutrient and competition indicator values.This points to a loss of habitat quality and a shift in the plant community composition (Landolt et al. 2010;Küchler et al. 2018).In addition, the vegetation change towards tall-grass communities after landuse abandonment could increase the probability of natural hazards on steep slopes, such as landslides and (until shrub encroachment) avalanches (Tasser et al. 2003;Stöcklin et al. 2007).Thus, our results suggest that land use by humans is essential to maintain the high diversity in these alpine to subalpine grasslands, to retain their cultural heritage, and to prevent natural hazards (MacDonald et al. 2000;Maurer et al. 2006;Stöcklin et al. 2007).
In the alpine to subalpine grasslands on Schynige Platte, the land use applied since the establishment of the Alpine Botanical Garden, consisting of one cut per year in autumn (by scythe), seems to be inadequate for maintaining the species richness and composition, as species numbers have decreased and the vegetation community has changed towards more nutrient-rich grasslands during the last decades (Hegg and Schaffner 2012).This might be due to altered environmental conditions, such as climate change and nitrogen deposition.For instance, climate warming might lead to species becoming dominant because the conditions are more suitable for them or to the establishment of new dominant species as upward range shifts occur (Steinbauer et al. 2018).Besides climate warming, nitrogen deposition has been suggested to have had strong effects on species composition and turnover in Swiss grasslands in the past decades (Roth et al. 2019).During the six years of our experiment, we found little evidence of detrimental effects of traditional land use on habitat quality, as indicated by the non-significant effects on the mean ecological indicator values, and the even positive developments in comparison to the abandonment treatment.
Regarding species richness, we found a stronger increase in the plots combining haymaking and scraping than in the other treatments.Scraping can be seen as a moderate top-soil disturbance, removing more biomass and nutrients than haymaking alone, reducing interspecific competition to some degree, and creating open patches which can in turn promote seed germination and species establishment.However, the increase in species numbers in this treatment cannot be attributed to the establishment of ruderal species that are favored by disturbance, as we found no significant differences among the treatments in the mean indicator values for ruderality.In line with our findings, Müller et al. (2014) and Klaus et al. (2018) found higher species richness after experimental top-soil disturbance than in undisturbed controls in German grasslands.In addition, as indicated by the development of the mean ecological indicator values in our study, haymaking combined with scraping tended to have positive effects on habitat quality, towards nutrient-poor and highlight conditions that are suitable for specialized species.Although to a lesser degree, this positive development was also visible in the species composition, as the cover of species with low mowing tolerance and of species that were promoted by abandonment consistently decreased with combined mowing and scraping.In addition, scraping in spring and mowing in autumn positively affected the cover small statured forbs.This might be because the growth of early and fast-growing species was disrupted by spring scraping which might have created high-light conditions, promoting less competitive species (Zhang et al. 2020).In contrast, haymaking and scraping in autumn promoted the development towards a vegetation composition which has been attributed to the Caricion ferrugineae alliance.However, as the vegetation period is rather short in alpine to subalpine grasslands and vegetation responses are therefore slow, our results might have been more pronounced if the treatments had been maintained for a period longer than six years.Furthermore, whether such a disturbance measure needs to be applied over several years or only once when e.g.litter and moss cover has been built and whether a similar development could also be achieved by e.g.extensive grazing combined with haymaking, which would be even more practical in some cases, or the use of machinery, needs to be investigated in future studies.

Conclusions and practical implementation
Our findings of negative abandonment effects highlight the need to maintain management for biodiversity conservation in alpine to subalpine grasslands.Maintaining or even increasing subsidies for additional measures paid by the federal agencies are therefore needed as incentives for farmers to maintain extensive forms of land use and preserve this cultural heritage.Our results further imply that modifying haymaking methods, in our case with top-soil disturbances, can compensate for the negative developments due to altered environmental conditions and could therefore be a practicable alternative to restore the diversity and habitat conditions of alpine to subalpine grasslands.From a practical point of view, using machinery is the most widespread method for haymaking.Especially on a larger scale, machines would be needed for moderate topsoil disturbance.For this purpose, various tillage implements are available for all tractor types, ranging from large harrows for easily accessible areas to smaller solutions for single-axle tractors that can be used even on steep slopes.

Fig. 1
Fig. 1 Photos of the six sites in the Schynige Platte Alpine Botanical Garden (Bernese uplands; Switzerland), each with five plots that were randomly assigned management treatments: (1) haymaking in Sep-

Fig. 2
Fig. 2 Development of species richness (±se) over six years, showing the regression lines separated by management treatment

Fig. 4 Fig. 5
Fig. 4 Development of mean CSR indicator values over six years, separated by management treatment

Table 1
Summary of the linear mixed models with orthogonal contrasts, showing the effects of management treatments on species richness and scores of similarity of the observed vegetation composition to five vegetation types across time.Significant differences (P < 0.05) and marginally significant differences (P < 0.1) are indicated with bold values of slope of difference (from regression lines), trend of difference (temporal trend) and p-values

Table 2
Summary of the linear mixed models with orthogonal contrasts, showing the effects of management treatments on mean ecological indicator values across time.Significant differences (P < 0.05) are indicated with bold values of slope of difference (from regression lines), trend of difference (temporal trend) and p-values