The use of functional traits as a tool in evaluating restorations of peatlands

Summary Introduction Fens are minerotrophic peatlands with the water table located very close to the ground level. The water flows through the peatland by internal seepage and occasional regions of surface overflow (Rydin and Jeglum, 2006). The composition of plant communities found in fens depends on a complex interplay of pH (Sjörs and Gunnarsson, 2002, Wheeler and Proctor, 2000, Økland et al., 2001), nutrient availability (Pauli et al., 2002, van der Hoek et al., 2004), light availability (Kotowski et al., 2001, Kotowski and van Diggelen, 2004) and water table depth (Kotowski et al., 2001, Mälson et al., 2008). This interplay makes fen ecosystems vulnerable to changes in any of the above mentioned factors, and often a modification of one variable alters another variable. An example of this is drainage of fens, which lowers the water-table, which in turn aerates the previously anoxic peat, resulting in mineralization of the peat, with release of nutrients (Turner and Haygarth, 2001). This causes increased primary productivity (Joyce, 2001), which in turn lowers the light availability (Kotowski and van Diggelen, 2004), which has a detrimental effect on characteristic fen species (Kotowski and van Diggelen, 2004, Sundberg, 2012, Vermeer and Berendse, 1983). The lowered water-table increases the influence of rain-water. This results in acidification (van Diggelen et al., 2006), usually followed by an increase in peat mosses (Sphagnum spp.) which causes a positive feedback of further acidification, due to ion-exchange between Sphagnum spp. and its local surrounding (Rydin and Jeglum, 2006). Degradation and restoration During the 20th-century large areas of peatlands in Europe were drained for agricultural production, forestry purposes or peat extraction (Vasander et al., 2003). In 1997 it was reported that 62 % of Europe’s mires (peat accumulating peatlands) have been lost, and only 5 % is protected (Joosten, 1997). Attempts to counteract this development include conservation and restoration, although even the protected areas include peatlands that are in strong need of restoration (Sundberg, 2006). Ecological restoration as defined by the Society for Ecological Restoration is the process of assisting the recovery of an ecosystem that has been degraded damaged or destroyed (Society for Ecological Restoration International Science & Policy Working Group, 2004). Ecological restoration of fens aims at restoring both the abiotic factors (hydrology, light-availability and nutrient-concentration) and the biotic factors (species assemblage). Expressed in terms of community assembly rules, the restoration aims to alter the environmental filters (abiotic and biotic) that limits the restoration site from developing species communities similar to those found in analogous undisturbed ecosystems (further on called ‘reference ecosystems’ or ‘reference sites’). In moderately degraded fens, i.e. cases of moderate drainage for forestry purposes, a common restoration method is the combination of rewetting by blocking drainage ditches and tree-cutting to restore the light-conditions (Haapalehto et al., 2011, Hedberg et al., 2012, Laine et al., 2011, Lanta et al., 2006, Mälson et al., 2008, Mälson et al., 2010). In heavily degraded fens, were the peat has mineralized, resulting in excess nutrient-concentration, an increasingly applied restoration method is topsoil removal. This restoration method has the capacity to increase the ground-water level, remove the unwanted seed-bank, increase light-conditions and remove excess nutrients (Klimkowska et al., 2010, Patzelt et al., 2001, Rasran et al., 2007, Tallowin and Smith, 2001). In cases where the original species community has vanished, and dispersal limitation makes natural colonization into the restoration-site improbable, species have to be introduced in order to restore the species-community. Among methods used to introduce communities of species, direct seeding (Fraser and Madson, 2008, Tallowin and Smith, 2001) and haytransfer from reference-meadows (Hölzel and Otte, 2003, Klimkowska et al., 2010, Patzelt et al., 2001, Rasran et al., 2007) are commonly used. Due to the mentioned multitude of factors influencing the species community in a fen (e.g. hydrology, light, pH and nutrient concentration), ecological restoration is difficult, and restoration sites rarely reach the conditions of the reference sites (Moreno-Mateos et al., 2012). The relative success of ecological restoration has mostly been measured in terms of numbers of target species present, indices that combine species identity and relative abundance of each species (Shannon, 1948, Simpson, 1949) or ordination methods based on species identity and abundance. These methods can provide information regarding how the species composition at the restoration site compares to that of the reference site, but provide no ecological explanation for any similarity or difference that is detected. Secondly, an evaluation based on species identities loses its usefulness outside the geographical distribution-range of the studied species. The advent of functional diversity A solution to the limits of the species-identity focused analysis is to turn the attention to the functional characteristics the species possess. In functional ecology functional diversity measures have been put forward as a measure that takes into account the functions of species (Diaz and Cabido, 2001, Díaz et al., 2007, Garnier et al., 2004, Garnier et al., 2007, Laliberté and Legendre, 2010, Mason et al., 2005, Villéger et al., 2008). Although functional diversity appears to have the potential to highlight if and how a restoration site differs in functional composition from the reference site, as well as describing the environmental filters that influence the outcome of the restoration project, the methods have not been applied to restoration ecology. Considering that species composition in fens is influenced by several ecological filters, i.e. anoxia (a strong abiotic filter) (Kotowski et al., 2010) and competition for light (a biotic filter) (Kotowski and van Diggelen, 2004), the application of a functional analysis for traits corresponding to these filters may be useful for ecologists when analysing if the species community in the restoration site is functionally similar to the species community in the reference site, and how specific traits are filtered by the restoration measures. With this tool ecologists would have the potential to, based on functional analyses of similar restoration projects, determine which target species are likely to establish, and adapt restoration measures in order to increase the chance of successful establishment of species carrying certain traits. My research focused on the potential value of a functional diversity analysis in the analysis of fen restoration. The value lies in the potential in obtaining information on how restoration actions modify the environmental filters that only species with certain traitvalues can pass, and how the species community in the restoration site differs in functional composition from the species community in the reference site. A difference in the functional composition between the species communities should indicate a difference in the environmental filters, and thereby pinpoint factors that constrain the restoration site from reaching the state of the reference-site. A variety of functional diversity indices exists today. I have limited my research to 4 indices that, based on their structure, should have the potential to provide information on environmental filters in fen restoration. These four indices are the three indices presented by Mason et al. (2005) (functional richness, functional evenness and functional divergence), as well as functional dispersion (Laliberté and Legendre, 2010). Apart from these, the Community Weighted Means of the functional traits are analysed, as well as the functional group abundance. The three indices put forward by Mason et al. (2005) and placed into a multivariate context by Villéger et al. (2008) (functional richness, functional evenness and functional divergence) measure the size of the filled niche space, the evenness of biomass distribution within a niche space and the abundance distribution within a niche space respectively. These three indices all describe different facets of functional diversity. However, Laliberté and Legendre pointed out that the connection to the convex-hull makes functional richness sensitive to outliers, and that functional richness does not take into consideration the relative abundance of species. Functional evenness and functional divergence on the other hand include the relative abundance of species, but lack information on the distribution of species in the trait space. Their solution to this was the index functional dispersion, which is a measure of the dispersion of trait values in a trait space, defined as the average distance of all species to the centroid of all species in a trait space (Laliberté and Legendre, 2010). Description of the research conducted for this thesis This thesis consists of four individual publications, in which various aspects of the restoration of fens have been explored. The first publication is a meta-analysis of the diversity and relative success of species introduction methods in restoration of fens and semi natural grasslands, carried out by conducting an extensive search for studies that used introduction of species as a part of ecological restoration. The introduction-methods used were described and quantified, and the success of each method was evaluated based on how many of the introduced species that established in the restoration site. When available, the authors written comments about their view of the restoration were included as a variable in the evaluation. The analysis showed that in many cases species introduction through the transfer of hay from donor-meadows is a successful method for restoring species communities. This is not to say that the method is superior to any other species introduction method, since external factors may considerably change which method is most suitable. An example of this is that direct seeding or hay spread may not be recommended if the species are rare, have low germination rate or low seedling survival. Further the application of hay-transfer in the restoration of Calowanie fen showed that hay transfer without ground-disturbance (such as top-soil removal) has very little impact on the species community, and if the degradation is moderate, as in the Swedish sites, then complete removal of the vegetation may not be recommended. Following this a Swedish fen restoration project covering three fens that were drained for forestry purposes in the 1950s and restored in 2002 by cutting planted trees and blocking drainage ditches was analysed both through a functional group-focused analysis (publication 2) and through a functional trait diversity analysis (publication 3). The vegetation changes in the three Swedish restoration sites were monitored between 2002-2010 and compared to a reference-site that was monitored between 1978-1979 prior to it was drained. The application of a functional group-focused analysis in the Swedish fen restoration projects (publication 2) clearly showed that both rewetting and tree cutting increased the cover of Sphagnum, wetland bryophytes and sedges. Tree cutting increased the cover of grasses, wetland vascular plants and juvenile trees. Both treatments resulted in an increase in species richness, and the combination of the two treatments resulted in the highest speciesrichness. However, the rich-fen specialists that were the target of the restoration did not recover, most likely due to dispersal limitation. By applying a functional trait analysis (publication 3) we could detect that the restoration sites had a higher functional richness, a higher functional dispersion and a higher canopy height than the reference-sites, which indicates a too relaxed filter. Tree cutting resulted in an increase in functional richness and functional dispersion, which indicates a lowered habitat filter after the removal of the shading canopy. As a contrast to the Swedish restoration project, I analysed the outcome of an ecological restoration of Calowanie fen located 30 km south-east of Warsaw that previously had been drained for agricultural production (publication 4). Fen restoration in Calowanie was conducted in 2008 by removing the degraded top-soil, and introducing target species from donor meadows via hay-transfer. The top-soil removal site was divided up into belts, and hay was dispersed on every second belt, providing the possibility to analyse the effect of top-soil removal with and without hay-dispersal. As a control, hay from the donor meadows were also dispersed on control plots where no top-soil removal was conducted. I monitored permanent plots located in all treatment-types (top-soil removal with and without haytransfer and control-plots with and without hay-transfer), as well as in donor meadows. Within the top-soil removal sites, plots were distributed in such a way that the gradient of relative water table depth was covered so that each plot’s distance to the ground-water level could be included as an explanatory variable in the community analysis. The outcome of the restoration was analysed both through a classical species-identity focused analysis and through a functional trait analysis. A multivariate RDA analysis of the restoration in Calowanie fen (publication 4) showed clearly that the top-soil removal site and the reference-site were separated from the control site along the first ordination axis, which followed a wetness gradient. Hay-dispersal had very little influence on the species community compared to the effect of top-soil removal. We could also through the classical analysis detect that characteristic species responded significantly in abundance to the ground-water level. By applying a functional trait analysis we could detect that the groundwater level imposes a strong habitat filter revealed by a significant decrease in functional richness and functional dispersion with increasing ground-water level. At high ground-water level this habitat filter selects for capacity for clonal lateral spread, high Ellenberg moisture values, capacity for hydrochorous dispersal and low specific leaf area and lack of capacity for autochorous dispersal. Further, we could detect that the restoration site differs in trait composition from the reference site by having a significantly lower proportion of species that disperse through autochory, which was probably caused already at the harvesting of the hay from the donor meadows, as indicated by the functional trait analysis of species present as seeds in the harvested hay. Among the species groups that are disfavoured by this are sedges, which were a target species group of the restoration. Conclusions My research has taken place in two countries in sites located up to 960 km from each other. The research sites differ in geological history, land-use history, degree of degradation, climatic conditions and length and start of vegetation season. As a consequence, there are also differences in species pools as well. Whereas the Swedish sites had a large bryophyte cover, including an abundance of Sphagnum species in two out of the three Swedish sites, the Calowanie fen was dominated by grasses, sedges and herbs spanning from species typical for heavily degraded fens that were found in the unrestored control sites, to species typical for wet meadows and fens found in restored sites. Limnic species could be found in the deepest parts of the restored sites. Large differences in species-composition between geographic areas can cause considerable challenges in drawing site-independent conclusions regarding which species are likely to benefit from a specific restoration action. By applying a functional approach we switch the focus from which species will benefit from the restoration measures to which trait values and trait composition will species favoured by the restoration measures have. This switch of focus enables general site-independent conclusions to be drawn despite differences in species pools. Having data on how species possessing different trait-values are filtered by specific restoration actions, allows us to better predict how specific species are likely to respond to a specific restoration action, as long as trait-data for these species are available. With the functional-group based analysis we could in the Swedish fenrestorations conclude that certain groups of species benefitted from one or both of the restoration actions - tree cutting and rewetting. We could also detect that the species richness was higher in the restored sites than in the reference sites. We could however not find any ecological explanation to why this was the case. With the functional-trait analysis we could detect that the habitat filter in the restoration site was relaxed when the shading by canopy was removed. This weaker habitat filter allowed species of previously disfavoured trait-composition to flourish. Rewetting can influence productivity both by decreasing productivity by imposing anoxia, or increasing productivity by the release of phosphorous following rewetting of degraded peat. The higher canopy in the restoration site compared to the reference site indicates that productivity has increased due to release of phosphorous following rewetting of areas where previous drainage have caused the breakdown of peat, making nitrogen and phosphorous readily available. Similarly the functional-trait analysis of the restoration of Calowanie fen allowed us to get an ecological explanation to the results provided by the species based analysis. Together, the Swedish and Polish studies provide examples of how the functional trait analysis can be used in analysing the effect of restoration measures. Rather than being a replacement for the traditional analysis, it is a complement that provides an ecological explanation to why a specific species is likely to be favoured or disfavoured by a specific restoration action. With the combination of the traditional species-identity focused analysis and the functional-trait analysis we have the possibility to conduct restoration analyses that provides results that are both detailed in terms of effect on species in the study, but also relevant outside the distribution range of the present species, thus bringing added value to the science of ecological restoration.