Grassland Restoration Drives Strong Multitrophic Biodiversity Recovery, but Climate Extremes Jeopardize Drought‐Sensitive Species

ABSTRACT

Semi‐natural grasslands, Europe's most biodiverse ecosystems, are rapidly declining due to agricultural intensification, abandonment, and afforestation, leading to biodiversity loss and reduced ecosystem function. Despite their ecological value, grasslands are often overlooked, while afforestation, for instance, is prioritized for climate mitigation. This study assessed the effects of grassland abandonment, afforestation, and ecological restoration on multitrophic species richness and soil conditions. We used Estonian semi‐natural calcareous grasslands (alvars) as a model system. Results showed that historically overgrown and afforested grasslands recover fast and rapidly become species‐rich after woody plant removal and low‐intensity grazing reinstatement. These grasslands also serve as stable carbon sinks, storing soil carbon at levels comparable to afforested grassland areas, with restoration having no negative impact on carbon storage. Multitrophic species richness responded to restoration in three main ways: fast responders (plants, pollinators, birds) recovered relatively quickly, slow responders (lichens, bryophytes, soil fungi) showed little to no short‐term change, and drought‐sensitive species (ground‐dwelling arthropods) declined due to prolonged drought, which also potentially overshadowed the impact of restoration. Grassland restoration is vital for biodiversity, the sustainable supply of ecosystem services, and climate resilience, with long‐term monitoring needed to track recovery.

Restoration of abandoned and afforested semi‐natural grasslands resulted in rapid grassland recovery and an increase of multitrophic species richness on average. Although, when looking at specific organism groups, species responded to restoration in three main ways: fast response, where species recovered relatively quickly; slow response, where organism groups showed little to no short‐term change in species richness; and drought‐sensitive response, where change in richness and community composition patterns changed mostly due to prolonged drought, which overshadowed the impact of restoration. Additionally, these grasslands also serve as stable carbon sinks, storing soil carbon at levels comparable to afforested grassland areas, with restoration having no negative impact.

1. Introduction

Semi‐natural grasslands, shaped by centuries of low‐intensity grazing and mowing, are among Europe's most biodiverse ecosystems. They host many species characteristic of open habitats (Wilson et al. 2012) and support functional biodiversity, essential for agricultural production (Moonen and Bàrberi 2008). Biodiversity underpins vital ecosystem functions and services, and species‐rich grasslands play an important role in sustaining ecosystem multifunctionality by simultaneously providing multiple ecosystem services (Prangel et al. 2023).

Throughout the Quaternary glacial–interglacial cycles, the creation and continuity of patchy open landscapes in Europe were mostly attributed to native megaherbivores and natural disturbance regimes (Pearce et al. 2025). However, in the present day, human influence and domestic animals have taken over the role of these large herbivores in semi‐natural grassland management (Pärtel et al. 2005). For centuries, European landscapes have been heavily influenced by human activity, and open herb‐dominated habitats have persisted and coevolved under moderate grazing or mowing (Pearce et al. 2025; Svenning 2002). However, the species‐rich communities that developed due to traditional agricultural practices are now threatened by agricultural intensification, abandonment, and secondary succession or afforestation of the historical low‐intensity grasslands (Török et al. 2018). A drastic decline in grassland area has been reported in all of Europe. For instance, different Estonian semi‐natural grasslands have lost more than 95% of their historical spread, and many of the remaining patches are in need of restoration (Helm and Toussaint 2020). The rapid decrease in the grassland area has led to an extinction debt (Helm et al. 2006) where species are still persisting in unsuitable conditions but, with no improvement in habitat quality, are inevitably prone to extinction in the future (Saar et al. 2012). Extinction debt can be viewed as a warning call and poses a time window of opportunity to restore the threatened habitats and ‘avoid the payment’ of debt (Kuussaari et al. 2009). In general, a significant decrease in species richness has been detected in Europe and globally, especially in the taxa adapted to open habitats (Jandt et al. 2022; Lengyel et al. 2023; Staude et al. 2023). For instance, a large‐scale decline in the richness and abundance of arthropods (Seibold et al. 2019) and a significant reduction in the abundance of grassland birds (Douglas et al. 2023) have been documented due to grassland habitat loss.

Despite the high nature value of grasslands, significant loss of habitat area, and documented declines in species richness and abundance, grassland protection still needs more representation in conservation policies. Moreover, the persistence of remaining grasslands is additionally threatened by increasing focus on afforestation and tree planting as a nature‐based solution for mitigating climate change (Temperton et al. 2019). However, research demonstrates that extensively managed species‐rich grasslands also serve as stable and effective carbon sinks (Lindborg et al. 2023; Prangel et al. 2023; Soussana et al. 2004; Yang et al. 2019), with most of the carbon stored in soil, one of the largest and more stable terrestrial reservoirs of carbon (Bai and Cotrufo 2022). Furthermore, an increasing number of studies have discussed the connection between biodiversity and climate change mitigation, and there have been indications that restoring and enhancing the number of plant species in degraded grasslands can lead to increased carbon sequestration (Hungate et al. 2017; Steinbeiss et al. 2008; Yang et al. 2019).

Given the concerning state of biodiversity and grassland condition, along with an increased demand for efficient ecosystem service provision and climate change mitigation, large‐scale restoration efforts and long‐term impact monitoring are essential to slow down the trend of extinction. There are still relatively few large‐scale restoration projects with long‐term monitoring and maintenance schemes, or the monitoring of restoration success usually concentrates only on one or two organism groups, most commonly vascular plants (Jandt et al. 2022; Rinella et al. 2016; Smith et al. 2017). In order to better understand how ecosystems respond to restoration, we need more complex monitoring encompassing a wider range of biota covering multiple trophic levels, for example, “multidiversity” and inclusion of the relationship with soil condition and carbon storage.

The grassland restoration project “LIFE to alvars” (LIFE13 NAT/EE/000082) was launched in Estonia in 2014 to mitigate the grassland biodiversity losses and increase the livelihood of rural communities. The project focused on the restoration of alvar grasslands (calcareous grasslands on thin soil on limestone bedrock), which have specific flora and fauna and have been shown to exhibit extinction debt (Helm et al. 2006) because of the drastic habitat loss over the last century. By the end of the project in 2019, 2500 ha of historical calcareous alvar grasslands were restored and taken into management (extensive grazing) with long‐term management contracts. In collaboration with the restoration project, we established a comprehensive, multitrophic monitoring scheme covering vascular plants, bryophytes, lichens, spiders, myriapods, bumblebees, butterflies, birds, and arbuscular mycorrhizal fungi, with the inclusion of monitoring soil condition and environmental factors. We systematically surveyed soil conditions and multitrophic species richness before and after restoration, providing a robust and extensive assessment of biodiversity recovery. This study addresses three main questions:

1. How do the species composition and diversity of different organism groups evolve during the restoration recovery phase (within 0.5–5 years after restoration)?

2. Does the recovery of grassland species richness depend on the initial degradation intensity (still relatively open grassland, overgrown grassland, afforested grassland)?

3. How do different organism groups respond to restoration impact? Do they respond in a relatively similar manner, or are there any taxon‐specific patterns?

2. Material and Methods

2.1. Study Area

Alvars are semi‐natural calcareous grasslands on Ordovician or Silurian limestone outcrop areas in northern Europe, mainly in Sweden and Estonia (6280* Nordic alvar and Precambrian calcareous flatrocks). Alvars are characterized by a thin soil layer (on average < 20 cm) with relatively high pH ~7 (Zobel and Kont 1992). These shallow soils often dry out completely during the summer, while the poor drainage capacity of limestone bedrock leads to flooding in autumn and spring. The soil type is rendzic leptosol, where only the humus layer can usually be found (Pärtel, Mändla, and Zobel 1999). The specific environmental conditions have led to a rich, specialized flora and fauna with a high proportion of protected species (Reitalu et al. 2014; Rosén and Maarel 2000). There are three main alvar grassland types based on environmental conditions and the most common plant species: temporarily wet Molinetum/Sesleria type, dry Avenetum type with thicker soil, and very dry and thin‐soiled Festucetum type (Pärtel, Kalamees, et al. 1999).

Similarly to other semi‐natural grasslands in Europe, land use changes over the last century have led to a drastic decrease in alvar grassland area. In Estonia, the decrease in alvar grasslands is mainly related to two factors: (1) abandonment of traditional grazing practices leading to encroachment of shrubs (mostly Juniperus communis and Corylus avellana ) and trees, and (2) afforestation with Scots pine ( Pinus sylvestris ) in the 1960s and 1970s. Significant alvar grassland area on the northern coast was also lost to urban development (Kalamees et al. 2012).

2.2. Restoration Process

The restoration efforts of the “LIFE to alvars” project (LIFE13 NAT/EE/000082) focused on the historical calcareous grassland area in western Estonia, where the grassland recovery potential was the highest. The selection of grasslands to be restored was based on historical records and scientific research (Helm et al. 2006; Pärtel, Kalamees, et al. 1999; Pärtel, Mändla, and Zobel 1999; Pärtel et al. 1998; Zobel et al. 1996). Only grasslands with a known history of extensive management and relatively good current landscape connectivity were chosen for the project. For example, grasslands in northern Estonia are very fragmented, and the remaining small patches are under constant pressure from urban expansion, making them not feasible for restoration efforts. Restoration actions were carried out between 2015 and 2019 and included shrub cutting, tree felling, and stump milling. After removal of the excessive woody vegetation, the expected remaining shrub cover was 30%–50% (in some subsites also below 30%), and only single scattered trees were left. All woody biomass was removed from the sites or burned on the spot. An important part of the restoration project was reinstating the traditional moderate grazing regimen by cattle, sheep, or horses to guarantee the maintenance of restored areas. After the reinstatement of the grazing regime, grasslands were left to recover naturally, and no additional restoration methods were used (like seeding, hay transfer, etc.), except for the removal of shrub water sprouts in some more overgrown subsites a couple of years post‐restoration.

2.3. Experimental Design

In connection with the “LIFE to alvars” (LIFE13 NAT/EE/000082) restoration project, a monitoring scheme was established with the intent to track the long‐term ecological consequences of overgrowing, afforestation, and grassland restoration. This article is based on the first two ecological surveys conducted as part of the monitoring—one before and one after the restoration. The monitoring scheme included 29 restoration sites, where restoration was implemented, and 7 unrestored comparison sites with no restoration actions performed. Altogether 36 study sites on three Estonian islands (Saaremaa, Muhu, Hiiumaa) and in the western part of mainland Estonia (Figure 1). Prior to the restoration, we identified three different grassland condition classes (subsites) according to their degradation level in each study site: Open, Overgrown, and Afforested grassland subsite. Open subsite represents abandoned grassland in the first stage of degradation but still in good condition, characterized by a relatively short herb layer and moderate (up to 30%–60%) shrub cover (mostly J. communis ). Overgrown subsite represents a former open grassland area that has been overgrown by dense shrubs with more than 60% shrub cover (mostly J. communis ) due to abandonment. Afforested subsite represents the previously open grassland area where P. sylvestris was planted during the afforestation campaign in the 1960s‐70s, now with tree cover of at least 50%. Additionally, we chose a control subsite near the restoration site, outside the restoration project boundaries. Control subsite represents abandoned alvar grasslands close to the restoration area, which were still in relatively good condition (similar to the open subsite) but were not going to be restored, and where the course of continuing abandonment can be monitored in the future. We chose to have a control subsite, in addition to the unrestored comparison study sites, since there were only seven of those, and they might not reflect representative results for every study site's conditions. In conclusion, we had two types of controls—the unrestored comparison sites with three subsites (open, overgrown, and afforested) and control subsites near restoration sites.

FIGURE 1.

Study sites: positioning of grassland restoration and unrestored comparison sites in western Estonia (above), and example of positioning of the subsites before and after restoration in Paope restoration site with aerial photo comparison (Estonian Land and Spatial Development Board) showing the restoration impact (below). Yellow dashed polygons indicate the boundaries of the LIFE to Alvars restoration project. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

All study sites were chosen in 2014–2015 prior to the restoration actions, and surveyed twice, before (between 2014 and 2016) and after the restoration actions (2019). The original plan was to survey precisely the same fixed subsites before and after restoration. However, because the pre‐restoration survey was done before the final restoration agreements, there were subsites that eventually were not restored. In those restoration sites, we chose comparable after‐restoration subsites in the restored parts of the same restoration area. Therefore, our sampling scheme is a mixture of repeated measurements from exactly the same locations and of sites with pre‐ and post‐restoration surveys done from nearby locations with similar communities. In the restoration study sites, 66 subsites were surveyed both before and after in precisely the same position (20 Open, 14 Overgrown, 11 Afforested, and 21 Control), 50 subsites were surveyed only before (9 Open, 15 Overgrown, 18 Afforested, and 4 Control) and 45 were surveyed only after restoration (10 Open, 15 Overgrown, 14 Afforested, and 3 Control). So altogether in the restoration sites, our survey included 29 open, 29 overgrown, 29 afforested, and 25 control subsites pre‐restoration; 30 open, 29 overgrown, 29 afforested, and 24 control subsites post‐restoration. In unrestored comparison sites, all subsites (7 Open, 7 Overgrown, and 7 Afforested) were surveyed in the same locations.

2.4. Surveying Different Organism Groups

In each subsite, we surveyed different taxa in close proximity to a fixed, permanently marked point, which was chosen to be reflective of the average subsite condition. The exact monitoring methods depend on the particularities of each organism group (Table 1). More detailed information on the survey methods can be found in the Supporting Information.

TABLE 1.

Organism groups surveyed in Estonian alvar grasslands before and after restoration, with a short summary of the monitoring methodology.

Taxa	Method	Surveyed indicators	Survey time
Vascular plants	10 m radius plot around the permanently marked point of the subsite Plant biomass sampling gathered from a 20 × 20 cm quadrat within the 10 m radius plot	Number of species, species identity and abundance within the 10 m radius plot Weight of dry plant biomass (g/cm2)	Summer, during the height of the vegetation period Pre‐restoration: 2014–2015 Post‐restoration: 2019
Bryophytes	10 m radius plot around the permanently marked point of the subsite (same as for vascular plants)	Number of species within the 10 m radius plot	Pre‐restoration: 2014–2016 Post‐restoration: 2019
Ground lichens	10 m radius plot around the permanently marked point of the subsite (same as for vascular plants)	Number of species within the 10 m radius plot	Pre‐restoration: summer 2015, 2016 Post‐restoration: summer 2019
Epiphytic lichens	Three junipers were chosen within the 10 m radius plot	Number of species on the selected junipers within the 10 m radius plot	Pre‐restoration: summer 2015, 2016 Post‐restoration: summer 2019
Bumblebees	Transect walk method (Pollard 1977) –individuals counted and identified on a 100 m long and two meters wide transect for 20 min Transect close to the 10 m radius plot	Abundance and species richness (identified on‐site and in the lab) of bumblebees interacting with flowers	Pre‐restoration: 12–22/07/2015 and 16–23/07/2016 Post‐restoration: 11–21/07/2019
Butterflies	Transect walk method (Pollard 1977) –individuals counted on a 250 m long transect for 5 min Transect close to the 10 m radius plot	Abundance and species richness of butterflies interacting with flowers	Pre‐restoration: 15‐26/07/2015, 5‐9/08/2015, 3‐14/06/2016 Post‐restoration: 4‐9/06/2019, 4‐12/07/2019, 2‐8/08/2019
Birds	Point‐count method (Bibby et al. 1992)–individuals counted in a 100‐m radius area for 5 min in the mornings between 5 and 10 am. The exact location of the surveys differed slightly from the subsites of other organism groups to allow sufficient space between the locations	Abundance and species richness of breeding birds	Pre‐restoration: May and June 2015‐2016, early morning Post‐restoration: May and June 2019, early morning
Spiders and myriapods (millipedes, centipedes)	Pitfall trap method (Hohbein and Conway 2018)–10 traps of 200 mL plastic cups, 70 mm diameter with 50 mL NaCl solution and covered with raised plywood plates, traps in use for 10 days. Trap transect was placed close to the 10 m radius plot.	Abundance and species richness of spiders and myriapods	Pre‐restoration: 18/06‐09/07/2015 And in five study areas 07‐16/06/2016 Post‐restoration: 29/06–17/07/2019
Arbuscular mycorrhizal (AM) fungi	9 teaspoons of soil was collected from the middle of the 10 m radius plot. DNA extracted from 5 g of dried soil using a DNeasy PowerMax Soil DNA Isolation Kit (Qiagen). AM fungal sequences were amplified from soil DNA extracts using primers specific for AM fungi from the small‐subunit (SSU) ribosomal RNA gene: WANDA (Dumbrell et al. 2011) and AML2 (Lee et al. 2008). Sequencing was performed on the Illumina MiSeq platform, using a 2 × 300 bp paired‐end sequencing approach. All sequencing was carried out at Asper Biogene LLC (Tartu, Estonia)	AM fungal sequences were matched against virtual taxa (phylogenetically defined taxonomic units) in the MaarjAM database (Öpik et al. 2010)	Soil samples collected Pre‐restoration: summer 2014–2016 Post‐restoration: summer 2019

2.5. Soil Properties

Soil samples for soil properties were taken from five randomly chosen spots from the topsoil (up to 10 cm) within a 10 m radius plot in each subsite. The five samples were mixed (~500 g), air‐dried, and stored until the analysis. Soil organic matter content (SOM, %), pH (in KCl), Ca, K, Mg, and P content (mg/kg) were estimated in each soil sample. Plant‐available Ca, K, Mg, and P contents were measured following the Mehlich 3 method (Mehlich 1984). Soil organic matter content was estimated using weight loss on ignition: 5 g of air‐dry soil was dried at 105°C for 2 h, incinerated in a muffle furnace at 360°C for 2 h, and the incineration ashes were weighed to find out the soil sample weight loss on ignition (LOI). Soil organic matter (SOM%) was calculated as SOM% = LOI/dry weight of soil × 100.

2.6. Statistical Analysis

We calculated the total multitrophic species richness of all surveyed species groups–“multidiversity”–including vascular plants, bryophytes, bumblebees, butterflies, spiders, myriapods, ground lichens, and AM fungi operational taxonomic units. Birds and epiphytic lichens were excluded from the total richness because these groups were not surveyed in all subsites. We used two methods to calculate the multitrophic species richness: standardized and raw richness. In the case of standardized multitrophic species richness, richness values of all species groups were divided by the maximum richness of the respective species group (Allan et al. 2014). Raw richness is simply the sum of all detected taxa.

To study whether soil chemical properties (SOM, pH, Ca, K, Mg, and P) have changed in relation to restoration, we used Wilcoxon's rank sum test with Bonferroni correction of p‐values to compare soil properties before and after restoration.

To characterize the change in different organism groups from before to after restoration, we used species richness and ordination analysis within the different organism groups (vascular plants, bryophytes, lichens, butterflies, bumblebees, birds, spiders, myriapods, AM fungi). In the case of vascular plants, bryophytes, and birds, we also calculated species richness separately for grassland specialist species to further test whether the restoration leads to the recovery of grassland‐specific diversity.

Richness change before and after restoration was studied using R programming language (R Core Team 2024) and Generalized Linear Mixed Effects (GLME) models with Poisson error distribution with packages “lme4” (Bates et al. 2015) and “emmeans” (Lenth 2022). For standardized multidiversity, a Linear Mixed Effects (LME) model was used. “Subsite” nested in “Study area” was used as a random factor; “Subsite” (Open, Overgrown, Afforested, Control), “Period” (Before restoration, After restoration), and their interaction term were used as fixed explanatory variables. To specifically test which subsites showed significant richness change after restoration compared to before restoration, we used Estimated Marginal Means (EMM) and “Before restoration” versus “After restoration” contrast. We used separate analyses (GLME model followed by EMM) for restoration sites and unrestored comparison sites with similar model structures. In addition to testing the before‐after contrast, we used EMM to test the difference among subsites before and after restoration. With the latter test, the assumption was that the subsites are significantly different from each other before restoration and that successful recovery of communities would lead to subsites being similar to each other in terms of ecological diversity.

To analyze the statistical difference in community composition, we performed a perMANOVA using the packages vegan (Oksanen et al. 2022) and pairwiseAdonis (Martinez Arbizu 2020). To characterize the community change from before to after restoration, we used Nonmetric Multidimensional Scaling (NMDS) using a package “vegan” (Oksanen et al. 2022) command metaMDS() for NMDS analyses. This is a robust technique for indirect gradient analysis suitable for community ecological studies (Minchin 1987). We used Bray‐Curtis distance as a distance metric for NMDS. For the vascular plants, bryophytes, and lichens, we based the NMDS on presence/absence data. In the case of the birds, butterflies, bumblebees, spiders, and myriapods, we used counts of individuals as abundance estimates in NMDS. In the case of AM fungi, we used the percentage of each taxon from the total number of sequences as an abundance estimate. On the NMDS plots, we used group ellipses with a standard error and confidence value of 0.99 to illustrate changes from before restoration to after restoration in community composition.

3. Results

3.1. Effect of Grassland Restoration on Multitrophic Species Richness

On average, the standardized multitrophic species richness increased significantly post‐restoration compared to pre‐restoration survey results in previously afforested and overgrown grasslands, while we recorded a decrease in species richness in the open and control subsites (Figure 2). When we used the raw sum species richness for the analysis, the decrease in species richness in the open and control subsites was also present, though marginal, and when we excluded the drought‐sensitive species groups (spiders and myriapods) from the raw sum multitrophic richness comparison, the restoration showed no negative impact on the species richness in open and control subsites and in unrestored comparison sites (Table 2, Figure S1). As we presumed, the species richness post‐restoration became more similar between the previously different subsites (among subsites contrast), compared to pre‐restoration, where in the majority of the observed taxa groups, the species richness differed significantly between the subsites (Table S2). Although the changes in richness patterns are still transitional and not happening at the same speed in all of the species groups, that was also reflected in the species composition analysis results, where many species groups still had significantly different species composition between the subsites post‐restoration (Table S4).

FIGURE 2.

Changes in multitrophic species richness (standardized richness) in relation to alvar grassland restoration. Left panel, restoration sites, and right panel, unrestored comparison sites before and after restoration. Significance values are indicated as follows: ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05.

TABLE 2.

Multitrophic species richness values (x̄ average richness) of the different grassland condition classes (open, overgrown, afforested, and control) in the restoration sites in pre‐ and post‐restoration comparisons (Before/After).

Taxa	Open grassland		Overgrown		Afforested		Control
	x̄ richness	z‐ratio	x̄ richness	z‐ratio	x̄ richness	z‐ratio	x̄ richness	z‐ratio
	Before/After	p‐value	Before/After	p‐value	Before/After	p‐value	Before/After	p‐value
Summed multitrophic species richness	128/121	−1.9 ns (p = 0.06)	111/120	2.4*	82/115	9.5***	139/131	−1.8 ns (p = 0.07)
Without drought‐sensitive species	112/115	0.9 ns	98/113	4.6 ***	69/109	11.8***	122/123	0.3 ns
Vascular plants	49/54	2.9**	49/63	5.7***	38/68	12.1***	62/61	0.1 ns
Strict alvar specialists	15/15	0.4 ns	11/15	4.2***	7/13	7.3***	16/16	−0.3 ns
Broad grassland specialists	36/40	2.1*	33/43	5.5***	21/42	12.4***	44/43	−0.1 ns
Plant biomass g/m 2	381/158	***	218/167	ns	100/155	**	307/219	**
Bryophytes	18/19	0.3 ns	16/17	0.9ns	13/14	1.5ns	15/17	1.4 ns
Grassland specialists	12/12	0.5 ns	7/9	2.6**	3/5	3.5***	9/10	1.1 ns
Epiphytic lichens	24/20	−2.4*	18/19	0.03 ns	13/13	−0.3 ns	29/21	−2.4*
Ground lichens	8/8	0.7 ns	2/3	0.5 ns	0/0	−0.3 ns	6/6	−0.4 ns
Bumblebees	4/4	−0.1 ns	2/4	4.8***	1/4	7.2***	4/5	0.9ns
Butterflies	6/4	−2.5*	1/4	5.6***	1/4	6.3***	7/5	−1.9 ns (p = 0.05)
Spiders	13/5	−8.4***	9/5	−5.7***	10/6	−4.4***	13/7	−6.4***
Myriapods	2/1	−1.5 ns	2/1	−3.7***	2/1	−3.2**	2/1	−2.4*
Birds	6/7	1.6 ns	7/9	1.3 ns	7/6	−0.7 ns	NA	NA
Open habitat birds	4/4	0.8 ns	3/6	2.0*	1/4	6.3***	NA	NA
AM fungi OTU	28/26	−1.5 ns	26/23	−1.6 ns	14/17	2.6**	27/27	−0.04 ns

Note: Positive z‐ratio values indicate an increase in species numbers post‐restoration, and negative values indicate decreased species richness post‐restoration. Significance is indicated as follows: p‐value = ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05. Species richness for unrestored comparison sites can be found in the Supporting Information (Table S1).

3.2. Soil Properties

We recorded no difference between the subsites before restoration (open grassland, overgrown, afforested) in their soil properties (Ca, K, Mg, P content mg/kg, and soil organic matter %) nor any significant changes in the soil properties in any of the subsites when comparing the pre‐ to post‐restoration grasslands (Figure S2).

3.3. Richness and Community Composition of Different Organism Groups

Vascular plants showed a significant increase in species richness in open, overgrown, and afforested subsites post‐restoration, with the highest increase (40%) in previously afforested grasslands (Table 2, Figure 3B). In addition, the PERMANOVA test and NMDS ordination plot illustrate a significant change in vascular plant community composition post‐restoration compared to pre‐restoration results (Table S3), with a trend toward greater similarity between the previously overgrown or afforested subsites and the open and control subsites, even though the species composition still remains different (Figure 3A, Table S4). The species richness of strict alvar grassland vascular plant specialists significantly increased in previously overgrown and afforested subsites, while generalist grassland plant species richness significantly increased in all but control subsites post‐restoration (Table 2, Figure S3). Post‐restoration, the community composition of the strict alvar grassland vascular plant specialists is similar between the open, overgrown, and control subsites, while it differed significantly between these subsites pre‐restoration (Table S4).

FIGURE 3.

Changes in the vascular plant community in relation to alvar grassland restoration. (A) Vascular plant community composition reflected by nonmetric multidimensional scaling (stress value 0.23), ellipses with standard error and confidence value of 0.99 are indicated. Both ordination figures are based on the same NMDS, positions of restoration sites are illustrated on the left figure, and positions of unrestored comparison sites on the right figure. (B) Vascular plant species richness before and after restoration in different subsites. Significance values are indicated as follows: ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05.

In the case of bryophytes, there was no significant change in species richness (Table 2, Figure 4B). However, the PERMANOVA test and NMDS ordination plot indicate that the species composition of bryophytes in previously overgrown and afforested subsites has changed and become slightly more similar to that of open and control subsites post‐restoration (Figure 4A, Tables S3 and S4), and we observed an increase in the richness of bryophyte species characteristic of open grassland habitats (Figure S4).

FIGURE 4.

Changes in bryophyte community in relation to alvar grassland restoration. (A) Bryophyte community composition reflected by nonmetric multidimensional scaling (stress value 0.25), ellipses with standard error and confidence value of 0.99 are indicated. Both ordination figures are based on the same NMDS, the positions of restoration sites are illustrated on the left figure, and the positions of unrestored comparison sites on the right figure. (B) Bryophyte species richness before and after restoration in different subsites. Significance values are indicated as follows: ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05.

While epiphytic lichens showed some change in species richness in open and control subsites post‐restoration (Table 2, Figure S5), ground lichens expressed no response to restoration (Table 2, Figure S6). The PERMANOVA test and NDMS ordination plots showed no clear changes in species composition when comparing the pre‐ and post‐restoration ground and epiphytic lichen communities (Figures S5 and S6, Table S3).

Species richness of bumblebees increased significantly in previously overgrown and afforested subsites in restoration areas, whilst decreasing in open and overgrown subsites in unrestored comparison sites post‐restoration (Table 2, Figure 5B). PERMANOVA test and NMDS ordination plot of community composition indicate that bumblebee communities between the subsites were vastly different from each other before restoration and that species composition converged and clustered closer together after the restoration (Figure 5A, Tables S3 and S4).

FIGURE 5.

Changes in the bumblebee community in relation to calcareous grassland restoration. (A) Bumblebee community composition reflected by nonmetric multidimensional scaling (stress value 0.22), ellipses with standard error and confidence value of 0.99 are indicated. Both ordination figures are based on the same NMDS, the positions of restoration sites are illustrated on the left figure, and the positions of unrestored comparison sites on the right figure. (B) Bumblebee species richness before and after restoration in different subsites. Significance values are indicated as follows: ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05.

Butterfly species richness increased significantly post‐restoration in previously overgrown and afforested subsites, while decreasing slightly in open subsites (Table 2, Figure S7). Similarly to bumblebees, the PERMANOVA test and NMDS figure of the species composition of butterflies in different subsites indicate a tendency of butterfly community convergence after the restoration (Table S4). However, the NMDS analysis also indicates an overall shift in species composition across all subsites when comparing pre‐ and post‐restoration periods (Figure S7).

There was no significant difference between pre‐ and post‐restoration species richness of all bird species in any of the subsites (Table 2, Figure 6B). However, the PERMANOVA analysis of community composition indicated a significant difference in species composition in overgrown and afforested subsites post‐restoration compared to pre‐restoration results (Figure 6A, Table S3). There was a significant increase in species richness of birds characteristic of open habitats (Figure 6C).

FIGURE 6.

Changes in bird community in relation to alvar grassland restoration. (A) Bird community composition reflected by nonmetric multidimensional scaling (stress value 0.19), ellipses with standard error and confidence value of 0.99 are indicated. Both ordination figures are based on the same NMDS, positions of restoration sites are illustrated on the left figure, and positions of unrestored comparison sites on the right figure. (B) Overall bird species richness and C: Open habitat specialist bird species richness before and after restoration in different subsites. Significance values are indicated as follows: ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05.

Species richness of both ground‐dwelling spiders and myriapods decreased significantly post‐restoration in all of the subsites, including in the unrestored comparison sites (Table 2, Figure 7B, Figure S8). On average, spider species richness declined by 50% and abundance by 60% between the two surveys. In addition, the PERMANOVA test and NMDS ordination plots indicated that the species composition of spiders systematically changed in all of the subsites post‐restoration. While before restoration, the species composition of spiders in overgrown and afforested subsites was more similar to each other and different from open and control subsites, then after restoration, species composition was similar in all subsites, yet also significantly different from pre‐restoration open and control subsite species composition (Figure 7A, Tables S3 and S4). There were no significant changes in myriapod species composition post‐restoration compared to pre‐restoration results (Figure S8).

FIGURE 7.

Changes in ground‐dwelling spider community in relation to alvar grassland restoration in Estonia. (A) Ground‐dwelling spider community composition reflected by nonmetric multidimensional scaling (stress value 0.21), ellipses with standard error and confidence value of 0.99 are indicated. Both ordination figures are based on the same NMDS, positions of restoration sites are illustrated on the left figure, and positions of unrestored comparison sites on the right figure. (B) Ground‐dwelling spider species richness before and after restoration in different subsites. Significance values are indicated as follows: ***p ≤ 0.001, **0.001 < p ≤ 0.01; *0.01 < p < 0.05, ns p ≥ 0.05.

The number of AM fungi operational taxonomic units (OTUs) significantly increased post‐restoration in afforested subsites, but was still significantly lower than in other subsites (Table 2, Figure S9). The PERMANOVA test and NMDS ordination plots indicated that the post‐restoration community composition of AM fungi changed in previously afforested subsites and became more similar to the overgrown subsites than it was before restoration (Figure S9, Table S3). However, the post‐restoration community composition in previously afforested and overgrown subsites remained different from the open and control subsites (Table S4).

4. Discussion

Restoring abandoned and afforested grasslands by removing excessive woody vegetation and reinstating a traditional grazing regime showed an overall positive impact on the multitrophic species richness, also known as “multidiversity” (Allan et al. 2014). Most organism groups responded positively to the restoration efforts, with increased species richness and/or a shift toward open grassland community composition, demonstrating the potential to reverse biodiversity loss (Kuussaari et al. 2009). We identified three primary response groups to restoration (and disturbance): fast responders, slow responders, and drought‐sensitive (disturbance‐sensitive) species. The fast responders–vascular plants, butterflies, bumblebees, and birds–showed a rapid and significant increase in species richness and/or a total shift in species composition, with grassland habitat specialists re‐establishing post‐restoration. The slow responders—lichens, bryophytes, and AM fungi—exhibited minimal or delayed recovery, indicating that some species groups require longer timeframes or targeted interventions to fully rebound. Drought‐sensitive groups, namely ground‐dwelling spiders and myriapods, experienced sharp declines. For these species groups, the potential benefits of restoration were strongly overshadowed by extreme and prolonged drought periods in the spring and summer of 2018.

Previously afforested and shrub‐encroached grasslands had, on average, a stronger reaction to restoration activities expressed by a higher gain in species richness than the already open grasslands, which even showed some losses in certain species groups. This can be explained by changes in light availability and more intense soil disturbance in overgrown and afforested sites than in already open regions, where mostly just moderate grazing was added as a restoration measure. Soil disturbance is recognized as one of the important drivers for plant diversity (Schnoor et al. 2015), and removing shading‐dominant woody species opened up free space for germination and colonization by grassland species. We assume that the species decline in open and control grasslands was the result of other external factors rather than restoration effects.

Restoration had no significant impact on soil properties, including soil organic matter content, which remained consistent across subsites (open, overgrown, and afforested grassland). While clearcutting can increase microbial activity in soil organic matter and accelerate soil carbon loss (Povilaitienė et al. 2022), open alvar grasslands have been shown to exhibit lower microbial activity than afforested alvar grasslands (Prangel et al. 2023). We speculate that this, combined with rapid vascular plant recolonization, could contribute to the stable carbon storage post‐restoration. Additionally, since we did not record any differences in soil conditions prior to the restoration between the different grassland conditions, we could assume that there will not be any significant changes happening to the soil in the future either.

Our findings confirm that timely restoration can reverse species loss and mitigate existing extinction debt, preventing permanent species declines in the future. However, delayed responses in some taxa and changes in environmental factors highlight the importance of long‐term monitoring and adaptive management to ensure full ecological recovery. With increasing risks from extreme weather events like drought, integrating climate‐adaptive strategies into restoration efforts is essential to enhance ecosystem resilience and sustain biodiversity gains.

4.1. Fast Responders

Vascular plants are one of the key fast responders, acting as good indicators for assessing restoration and recovery success (Tölgyesi et al. 2019; Török et al. 2011) even in the early stages of recovery (Zobel et al. 1996). We could see a rapid recolonization of vascular plant species already within a year after restoration. While in the early stages of recovery, the increase in species richness can often be attributed to an increase in generalist species (Waldén and Lindborg 2016), our findings also revealed a significant increase in grassland and, more specifically, alvar grassland plant specialist richness. Full recovery is often constrained by seed limitation due to degraded seed banks or habitat fragmentation (Bakker and Berendse 1999; Jacquemyn et al. 2011; Winsa et al. 2015) even after extended periods post‐restoration (Smith et al. 2017). However, alvar grasslands show a strong recovery potential and fairly persistent seedbanks (Kalamees et al. 2012), yielding promising results. The restoration sites in the “LIFE to alvars” project were also carefully chosen in areas where large historical alvars were distributed, but plant communities were shown to exhibit extinction debt, a delay in reaction to habitat changes (Helm et al. 2006). For example, prior to the restoration, open and overgrown grasslands exhibited no difference in plant species richness (both generalist and alvar specialist), even though there were significant differences in habitat conditions like the level of shrub encroachment or tree cover (Helm et al. 2006; Prangel et al. 2023). The results of the current study indicate that in habitats where species still endure due to historical legacy effects, the communities could be relatively easy to restore. Conversely, in regions where the species are lost or where local environmental conditions are drastically altered (like in arable fields), restoration becomes more challenging and costly because soil conditions might need to be altered or reintroduction of species by seeds or hay transfer is necessary (Knight and Overbeck 2021; Török et al. 2011).

Additionally, we observed a rapid shift toward grassland species composition across various trophic levels, not just in plants. For instance, while bird species richness remained similar, species composition changed significantly, with open habitat specialists recolonizing the areas and replacing the species favoring closed canopies. These are very welcome and promising results since grassland and cropland birds have been declining rapidly for the past decade and will continue the trend of decline without successful restoration efforts (Reif et al. 2024).

Restoring plant communities is a prerequisite to the restoration of pollinator communities (Sexton and Emery 2020). For example, in agricultural landscapes, sowing seed mixtures of flowering plants is a well‐known measure to bring back the pollinators (Buhk et al. 2018; Marja et al. 2019), especially where the landscape is highly homogenous or food resources for pollinators are scarce (Bihaly et al. 2024). Both bumblebee and butterfly abundance and richness have also been shown to be positively impacted by moderate grazing (Pöyry et al. 2004). In our results, the bumblebee and butterfly richness increased significantly in previously overgrown and afforested grasslands, but there was a slight decrease in butterflies in open grasslands and a decrease in bumblebees in unrestored comparison sites. We speculate that this could also be influenced by the drought, causing a decrease in flowering and nectar resources (Phillips et al. 2018), not necessarily the result of restoration and added grazing. Butterfly communities were also quite different between the survey years, irrespective of the grassland condition, and there can be a considerable turnover between consecutive observation years related to stochastic influences and weather (Dennis et al. 2010). Further monitoring would help to disentangle whether the richness decline in open areas and the community compositional change are due to restoration and grassland management or caused by other environmental factors.

4.2. Slow Responders

The slow responders—lichens, bryophytes, arbuscular mycorrhizal fungi—exhibited no significant changes in species richness and composition within the monitored timeframe (up to 5 years since restoration), or the response was fairly slow. This corroborates previous findings of lichens and bryophytes exhibiting a time lag in response to habitat change and slow community recovery (Hämäläinen and Fahrig 2024). In our study, bryophytes exhibited some recovery, mostly in the previously afforested areas, where the species connected to closed canopy habitats were becoming replaced by open habitat specialists. This response was comparable to birds, with the difference that while birds did a full community composition swap within the same timeframe, bryophytes associated with open habitats just started to inhabit the restored regions. Hence, we also categorised bryophytes as slightly slower responders to habitat change, needing more recovery time.

Interestingly, a legacy effect was observed in the diversity and community composition of arbuscular mycorrhizal fungi (AM fungi), reflecting the enduring influence of previous overgrowth and afforestation on soil biota. This delayed response in soil organisms to habitat changes can be explained by the passenger/driver hypothesis, suggesting that the AM fungi act as passengers, recolonizing the sites only after the plant communities, as their potential symbionts, have already established (Neuenkamp et al. 2018). The current diversity of soil fungi still bears the imprint of past overgrowth and afforestation, which had altered plant and fungal communities through increased colonization by ectomycorrhizal fungi associated with coniferous woody habitats (Wang et al. 2019). This legacy effect is especially evident in afforested sites, where changes in soil conditions and plant symbionts were more pronounced than in shrub‐encroached regions. Even though the AM fungal diversity increased post‐restoration compared to pre‐restoration results, the diversity is still significantly lower than in previously overgrown and open grasslands.

This contrast between fast and slow responders highlights the varying rates of recovery across different species groups and ecosystem functions as a whole, putting an emphasis on the importance of including several species groups in long‐term monitoring schemes and decision‐making plans.

4.3. Drought‐Sensitive Species

The recovery of grassland species can be hindered by concurrent disturbances and unfavorable biotic or abiotic factors in the restoration areas (Brudvig et al. 2017). Our study design enabled us to disentangle restoration effects from other external disturbances, revealing a severe decline in ground‐dwelling spider communities (spiders and myriapods) across all grassland condition classes, irrespective of the restoration efforts. This collapse of arthropod communities is most likely a result of prolonged drought during the spring and summer of 2018, 1 year before the arthropod sampling in 2019 (Prangel et al. 2024). Drought is known to have a strong impact on different taxa (Peterson et al. 2021; Phillips et al. 2018; Rosén 1995), especially when combined with other disturbances, such as restoration activities. Restoration can temporarily cause reduced plant cover and increased bare ground, making these areas more susceptible to drought and loss of topsoil due to wind erosion. This vulnerability is exacerbated when extreme weather events, such as prolonged droughts, coincide with early recovery phases (Buisson et al. 2015). Our results align with previous findings, demonstrating a severe decline in ground‐dwelling arthropod populations following prolonged periods of disturbance (Framption et al. 2000; Peterson et al. 2021). Additionally, we could also see a decrease in bumblebees and vascular plant biomass in grassland areas not affected by restoration and grazing, indicating the impact of drought. Given the increasing frequency of such anomalies in the current context of climate change, this challenge is only expected to intensify (Masson‐Delmotte et al. 2021). The compounded effects of drought on already stressed ecosystems might require more adaptive management strategies, like creating windbreaks and keeping the bare ground to an absolute minimum post‐restoration by nurse populations or light mulching, which also helps maintain soil moisture, in addition to erosion control (Fan et al. 2023). Furthermore, inoculating the restoration sites with native mycorrhizae could help to speed up the recovery of soil systems supporting above‐belowground interactions (Vahter et al. 2020).

4.4. Restoration Success, Impact Monitoring, and Limitations

The restoration progress in our study areas can be considered successful and is moving in the desired direction. However, the relatively short recovery period since the initial restoration actions (0.5–5 years since restoration) was not sufficient to cover the dynamics of slow‐responding species and environmental stochasticity. Also, future changes in soil properties remain possible due to the slow response rate of soil processes and carbon cycling (Poeplau et al. 2011). Long‐term monitoring is crucial for assessing restoration success and understanding the full trajectory of ecosystem recovery (Resch et al. 2021; Török et al. 2021). Helsen et al. (2013) documented significant recovery of specialist species richness within 15 years of restoration, suggesting that similar outcomes for grasslands are possible given time. Encouragingly, several species specialized in open habitats are already strongly establishing themselves in the restoration areas of the current study, reinforcing this potential. Regardless, for the recovery of the system to continue and to keep improving, it will also need a stable and long‐term management plan, since, without human interference, these systems will start to degrade once again (Prangel et al. 2023). Upkeep of the old management contracts and development of new policies on sound management and funding are essential. While restoration success in these calcareous grasslands suggests a high recovery potential, challenges like seed limitation, drought, and soil legacy effects in more degraded regions necessitate more complex restoration measures, adaptive management, and sustained long‐term monitoring (Olsson and Ödman 2014; Rasran et al. 2007; Török et al. 2011).

A multidiversity‐focused approach must be prioritized to track both rapid and delayed responses, ensuring that slow‐recovering taxa are not overlooked. Given the time lag in species recovery, monitoring should extend to at least 10 years post‐restoration to capture long‐term dynamics and potential secondary succession effects (Prangel et al. 2023; Tölgyesi et al. 2022). That is also something that our study lacked—the long‐term monitoring results due to the limited project timeline. Further post‐restoration monitoring would be needed to capture the full recovery of these grasslands. Currently, we are able to show the initial direction that the recovery is taking, how effective it has been, and whether we would need to make any adjustments to better the outcome (Coon et al. 2021). Early post‐restoration results for grassland recovery may lead to misinterpretation of the results, since species with a lag in recovery could still be reflecting the pre‐restoration or transitional disturbed ecosystem state rather than the expected climax community (DeCock et al. 2022). Additionally, our study could have profited from a higher number of study sites in unrestored comparison sites and reference sites (the managed high‐quality grassland), which could have added an even higher degree of confidence to the results.

As we progress through the Decade on Ecosystem Restoration (United Nations Environment Agency 2019) and work toward the goals of Nature Restoration Law in Europe to restore ecosystems to good ecological condition, grassland restoration stands out as a high‐impact conservation priority. Overgrown and afforested grasslands, particularly on low‐fertility soils, hold immense recovery potential, supporting high biodiversity and ecosystem multifunctionality—benefits that have the potential to outweigh those of wood production and carbon sequestration alone (Prangel et al. 2024). Semi‐natural species‐rich grasslands exhibit great and stable carbon storage capabilities while supporting multiple other essential ecosystem services like pollination, cultural heritage preservation, etc. Restoration policies must reflect these ecological trade‐offs and prioritize grasslands alongside forests in global climate and biodiversity strategies.

Author Contributions

Elisabeth Prangel: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing – original draft. Triin Reitalu: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing – review and editing. Liis Kasari‐Toussaint: investigation, writing – review and editing. Riho Marja: investigation, writing – review and editing. Inga Jüriado: investigation, writing – review and editing. Tiiu Kupper: investigation, writing – review and editing. Nele Ingerpuu: investigation, writing – review and editing. Ede Oja: investigation, writing – review and editing. Anu Tiitsaar: investigation, writing – review and editing. Reet Karise: investigation, writing – review and editing. Villu Soon: investigation. Krista Takkis: writing – review and editing. Liis Keerberg: investigation. Mart Meriste: investigation. Aveliina Helm: conceptualization, funding acquisition, project administration, resources, supervision, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: gcb70496‐sup‐0001‐supinfo.docx.

Prangel, E. , Reitalu T., Kasari‐Toussaint L., et al. 2025. “Grassland Restoration Drives Strong Multitrophic Biodiversity Recovery, but Climate Extremes Jeopardize Drought‐Sensitive Species.” Global Change Biology 31, no. 9: e70496. 10.1111/gcb.70496.

Data Availability Statement

The data that support the findings of this study are openly available in Dryad (datadryad.org).https://doi.org/10.5061/dryad.pg4f4qs1b.