The application of filter-based assembly models in the restoration of sand grasslands (2018-2023)
The aim of the present project was to integrate and scale up the local knowledge gained on sand grassland restoration in Hungary within the framework of filter-based assembly models in order to help the elaboration of a future national restoration strategy. We re-sampled previous restoration experiments (expansion in time), and local results were completed with landscape scale factors that might influence the outcome of restoration measures (expansion in space). We hypothesized that (1) regeneration potential was higher in landscapes with less intensive land use; (2) the dispersal filter represented a stronger constraint in landscapes with more intensive land use; (3) niche availability constraints were less important than dispersal limitation in oligotrophic systems; (4) the biotic filter became more relevant as the vegetation closes and in landscapes with high invasive pressure; (5) restoration success increased if targets are within the range of potential natural vegetation (PNV) types.
We have re-sampled five restoration experiments following the original sampling protocols to obtain long-term data. We have also sampled the main habitat types (primary grassland, secondary grassland, plantation, arable land, and native forest) within the 500 m radius landscape buffer around each experimental site. In addition, we performed a detailed census for invasive alien species along 8 ×100 m transects around each site. We also made some field visits to estimate restoration success related to shrub control in the Kiskunság National Park related to the MPNV-restoration target selection topic. We won a grant (ELKH SA-66/2021) to do some additional samplings in other seed introduction experiments carried out by other experts within the Kiskunság National Park, where we estimated the success of restoration interventions compared to control and the composition of the surrounding landscape.
We investigated the spatial regeneration trajectories of open and closed steppe and Poplar-juniper sand dune forests and thickets based on local, neighboring, and old-field regeneration capacity estimates of the Hungarian Habitat Mapping Database (MÉTA) for the whole country. The most vulnerable type of vegetation was closed grasslands, where spontaneous recovery was the most limited (Csákvári et al. 2019). We estimated the impact of environmental factors (naturalness, abiotic factors, and land cover) in determining the regeneration success. Our results confirm that better restoration results can be achieved in the vicinity of larger (semi-)natural areas, but the specific site conditions must also be taken into account during prioritization, e.g., seasonality of the local precipitation and the sand content of the soil (Csákvári et al. 2021). We used this knowledge together with analyzing spatial regeneration trajectories (the comparison between regeneration locally, at neighboring sites, and on abandoned old-fields) to prioritize restoration efforts for these vegetation types (Csákvári et al. 2022). Country-wide GIS-based spatial regeneration maps of sandy habitats were finalized and published. The final results were built in a decision-supporting system that can localize and determine the extent of priority areas for sand grassland restoration based on Multiple Potential Natural Vegetation models and abiotic and biotic conditions. The decision support system was tested for the Danube-Tisza Interfluve.
We have found that the dispersal is a strong filter for the spontaneous recovery of sand grasslands on degraded land. Without introducing target species, restoration efforts (e.g., mowing or carbon amendment only) might fail to achieve long-term success, depending on the availability of propagules of target and invasive species in the landscape (Halassy et al. 2019, 2021, Reis et al. 2021). Tree plantations of non-native species were found to be the main source of invasion, and secondary grasslands developed after cropland abandonment could be both sources of target and invasive species (Reis et al. 2022). Distance to primary grasslands or the estimated amount of target propagules did not have any significant impact on restoration progress, but the higher abundance of invasive species in the 500 m landscape buffer and the proximity to tree plantations had a significant negative effect (Reis et al. 2022). Within a shorter distance (100 m) from the restoration sites, invasive pressure affected the abundance of perennial invasive species interacting with time, but no distance-related effect on invasion was found (Sáradi et al. under review). The fact that seeding was found to be the best method for restoring sand grassland (Llumiquinga et al. 2021, Reis et al. 2022, 2023) also shows that dispersal has a key role to play in the assembly processes in the studied system. We delineated and tested possible seed transfer zones for restoration purposes (Cevallos et al. 2020, 2021).
We found that abiotic environmental conditions also play an important role in the regeneration of sand grasslands. With climate change, increasing temperatures and more frequent summer droughts can be expected that can severely affect restoration efforts. Drought in the year of sowing can result in weak seedling establishment (Kövendi-Jakó et al. 2021). Droughts also lead to shifts in vegetation composition during long-term vegetation development, indirectly promoting the dominance of invasive species over target species, because invasive plants seem to better recover after drought events (Krpán 2023). Soil properties remained similar in the long term. The manipulation of soil available nitrogen by carbon amendment was of limited importance for the restoration success, resulting in lower cover of vascular plants and cryptogams compared to control, but no differences were found for target or invasive alien species (Halassy et al. 2021) or species of different nitrogen requirements (Seyidova 2020).
Long-term vegetation development trajectories show that restoration sites approach reference conditions, but do not reach full success in 10–25 years. This is due to the fact that although the cover of target species increases with time, the cover of invasive species can also increase depending on the studied restorative treatments and the landscape context. This is primarily due to an increase in the cover of perennial invasive species (mostly Asclepias syriaca), while the cover of annual invasive species (dominantly Ambrosia artemisiifolia, Conyza canadensis) generally decrease with time (Sáradi et al. under review). Restorative methods that keep the vegetation open, like carbon amendment and, even more so, mowing, create windows for colonization, but the new sites can be occupied by invasive species if they are present in the landscape (Halassy et al. 2019, 2021, Reis et al 2021). Introducing target species reduces the availability of recruitment niches and besides increasing target cover, increases the resistance of restoration to invasion (Llumiquinga et al. 2021, Reis et al. 2022, 2023). The introduction of native species can be implemented through the transfer of commercially available species, species collected in the wild or hay with varying success in terms of vegetation cover and diversity (Kövendi-Jakó et al. 2019) or planting in the case of woody species (Halassy et al. 2020). Seeds from ex situ seed banks can also be used for the establishment of sand grassland species, as they maintain their viability in the short term as an adaptation to the local arid environment (Kövendi-Jakó et al. 2021).
We collected restoration interventions at a national level between 2002 and 2016 and evaluated the achievements regarding Aichi Target 15 (Török et al. 2019). We used this dataset with some additional new projects to investigate the relationship between the necessary post-treatment intensity of grasslands restored by shrub removal and the range of potential local native vegetation types according to the Multiple Potential Natural Vegetation (MPNV) model's predictions. We confirmed the utility of MPNV in defining sites that have low potential for grasslands, while high potential for forests and thus are likely to require continuous management and thus continuous human and financial efforts after grassland restoration. We propose MPNV to be used to set local restoration targets and estimate associated maintenance needs, and to identify self-sustainable restoration targets in general. The related manuscript (Vörös et al.) is under review in Restoration Ecology (D1 journal).
Implications for practice
Based on our results, we propose large-scale grassland restoration on abandoned agricultural fields as a cost-effective and sustainable restoration target, where present in the range of potential natural vegetation, instead of industrial tree plantations and afforestation with non-native species.
Multiple potential natural vegetation models can help to focus the spatial prioritization of grassland restoration over forest restoration at the national scale by detecting areas with high grassland and low forest potentiality where restoration of open habitats is more feasible and sustainable. It can also be used to set local restoration targets and estimate maintenance needs.
Regeneration potential estimates (or the environmental factors that correlate well with regeneration capacity) help optimize the spatial allocation of restorative interventions. The type of suggested restoration interventions can be also prioritized based on the spatial regeneration trajectories.
Priority should be given to locations with good regeneration capacity where passive restoration can be used. The presence of large remnants of (semi-) native vegetation indicates better regeneration and lower restoration (intervention) intensity, while the presence and proximity of non-native tree plantations are threats to recovery and should be either avoided or managed with high-intensity interventions.
In fragmented landscapes, the best method for restoring sand grasslands is the introduction of target species that boosts the establishment of target species and controls non-native invasion. Sowing commercially available species, harvested species, including seeds stored in an ex situ seed bank, hay transfer, or planting are equally well suited for this purpose. The creation of small species introduction units, also referred to as "establishment windows" from where the species can spread
to the whole site can be applied as a cheaper solution for large-scale restoration. It is recommended to use multi-year, scheduled seeding to reduce the negative impacts of particularly dry years.
Where existing vegetation limits the restoration of sand grassland, low-intensity mowing or grazing should be applied, especially to control woody encroachment – involving non-native species, e.g., Robinia pseudoacacia – and to open up space for colonization. The new spaces can be occupied also by invasive species; therefore, we suggest applying mowing (grazing) in combination with other treatments, e.g., the introduction of target species or targeted control of invasive species.
Carbon amendment can be used right after the abandonment of arable cultivation to temporarily reduce soil available nitrogen and keep the vegetation open in order to supplement seeding, but only on the small scale and especially when N fertilizers were previously applied causing high nutrient levels in the soil. Once the vegetation established, applications should be ceased.
Restorative treatments can have the highest influence on the success of sand grassland restoration, exceeding the impact of landscape factors and time. Targeting the dispersal, abiotic, and biotic filters in parallel would improve the effectiveness of restoration.