Wetlands are important habitats, often threatened by drainage, eutrophication and suppression of grazing. In many countries, considerable resources are spent combatting scrub encroachment. Here, we hypothesize that encroachment may benefit biodiversity – especially under eutrophic conditions where asymmetric competition among plants compromises conservation targets. We studied the effects of scrub cover, nutrient levels and soil moisture on richness of vascular plants, bryophytes, soil fungi and microbes in open and overgrown wetlands. We also tested the effect of encroachment, eutrophication and soil moisture on indicators of conservation value (red-listed species, indicator species and uniqueness). Plant and bryophyte species richness peaked at low soil fertility, whereas soil fertility promoted soil microbes. Soil fungi responded negatively to increasing soil moisture. Lidar-derived variables reflecting degree of scrub cover had predominantly positive effects on species richness measures. Conservation value indicators had a negative relationship to soil fertility and a positive to encroachment. For plant indicator species, the negative effect of high nutrient levels was offset by encroachment, supporting our hypothesis of competitive release under shade. The positive effect of soil moisture on indicator species was strong in open habitats only. Nutrient poor mires and meadows host many rare species and require conservation management by grazing and natural hydrology. On former agricultural lands, where restoration of infertile conditions is unfeasible, we recommend rewilding with opportunities for encroachment towards semi-open willow scrub and swamp forest, with the prospect of high species richness in bryophytes, fungi and soil microbes and competitive release in the herb layer.

Study sites

As part of the present study, we conducted field inventories at 44 wetland sites. The data collection was designed to supplement data collected in the Biowide project, a nation-wide survey of biodiversity in Denmark (Brunbjerg et al. 2019). Biowide included a total of 130 study sites (40 m × 40 m), of which we included all 58 sites evaluated as moist or wet based on plant species composition and soil moisture measurements. The Biowide sites varied in woody species cover from open vegetation over heterogeneous scrub to closed-canopy forest. The Biowide sites also varied in nutrient status from infertile to fertile soils, but were selected to foremost include natural and semi-natural habitats. The additional 44 sites were chosen to increase data coverage of former agricultural land, semi-natural meadows and agriculturally improved meadows, as well as different levels of scrub encroachment. We did a stratified selection of sites according to succession/light availability (open, tall herb, shrubs, closed canopy), nutrient status (high/fertilized, mid, low/natural) and soil moisture (moist, wet). Sites were located across Denmark, preferably with a minimum distance of 500 m between sites (except two set of sites, where distances were 252 m and 491 m, Fig. 1a). The geographic dispersal of sites ensured sites from both carbon rich soils, clay and sand. Each site (40 m × 40 m) consisted of four 20 m x 20 m quadrants, each with a 5 m circular plot in the centre (Fig. 1b). 

To illustrate the coverage of the soil moisture, nutrient and encroachment gradients covered by the combined data, we compared site mean Ellenberg F, N and L values (Ellenberg et al. 1992) for all 5 m circle plots with a reference data set from national monitoring, using identical 5 m circular plots (59,227 sites from agricultural, semi-natural, natural open and forest vegetation, http://www.naturdata.dk) (Nygaard et al. 2017). Mean Ellenberg values were calculated for plots with more than five species present. In scatterplots of plot-mean Ellenberg values, 95 percentile convex hull polygons were drawn for the reference data set as well as the Biowide and wetland data set to visually evaluate the representativity of our data (Appendix A).

Biowide data collection

Collection of vascular plants and bryophyte presence/absence data: Vascular plants and bryophytes were inventoried by a trained botanist and exhaustive lists for the four 5 m circle plots were made for each site. In addition, all additional species in the quadrat, but outside the 5 m circles, were recorded. The inventory was done in summer 2014 with a few early spring vascular plant species added in May 2015 (Brunbjerg et al. 2019). We aggregate the four 5 m circle plots and additional species list to obtain a site species list for analyses.

Collection of soil eDNA data: As alternative measures of biodiversity, we used the richness of operational taxonomic units, i.e. OTUs (Blaxter et al. 2005) of fungi and (eukaryotic) soil microbes obtained from metabarcoding of soil-extracted eDNA (Frøslev et al. 2017, Frøslev et al. 2019). We collected soil samples from all sites for the eDNA inventory. Samples were taken in October-November 2014. At each site, we sampled 81 equally distant soil samples from the top c. 15 cm and pooled them after removal of coarse litter. We homogenized the soil by mixing with a drilling machine mounted with a mixing paddle. A subsample of soil for DNA extraction and metabarcoding was taken from the homogenized sample.

Soil moisture: soil moisture was measured using a FieldScout TDR 300 Soil Moisture Meter. Sixteen equally distanced measurements were taken in each 40 × 40 m site in May 2016 (spring/wet period). To cover the temporal variation in moisture the measurements were repeated in August 2016 (summer/dry period) (Brunbjerg et al. 2020).

Wetland sites data collection

All additional data collection specifically for the present study was done according to Biowide protocols (Brunbjerg et al. 2019), with the exception of the following: 1) presence/absence of early-spring plants species was not recorded on a separate visit, 2) soil samples were collected during the plant inventory in July-August 2018, i.e. not in November. Soil moisture was measured as in the Biowide project in July-August 2018.

The present dataset spans 102 sites covering most types of wetlands including agriculturally improved meadows, natural meadows, fens, bogs, reed swamps, meadows dominated by large herbaceous perennials, open wetlands with scattered willows and birches, willow thickets, birch forests and swamp forests (Appendix B).

DNA extraction and metabarcoding

For Biowide and wetland soil samples DNA was extracted and subjected to eDNA metabarcoding through DNA extraction, PCR amplification of genetic marker regions (DNA barcoding regions) and massive parallel sequencing on the Illumina MiSeq platform as described in Brunbjerg et al. (2019). For this study, we used high-throughput sequencing data from marker genes amplified with primers targeting eukaryotes (mainly soil microbes) (18S) and fungi (ITS2). OTU tables were constructed following the overall pipeline in Frøslev et al. (2017). For both fungi and eukaryotes, this consisted of an initial processing with DADA2 (ver. 1.8) (Callahan et al. 2016) to identify exact amplicon sequence variants (ESVs) including removal of chimeras. The preparation of the Biowide-fungal (ITS2) and Biowide-soil microbe (18S) eDNA datasets have been published in Fløjgaard et al. (2019) and Frøslev et al. (2019) respectively, although the fungal dataset was re-sequenced for this study (a detailed description of the sequencing data can be seen in Appendix C).

Lidar-based measures

We calculated 23 lidar variables to represent encroachment in the 102 sites. We used the same procedure as in Valdez et al. (2021). The calculations were based on the National lidar-based point cloud (recorded leafs-off, springs and autumns 2014-2015; light wavelength: 1550 nm; point density = 4-5 points/m2, vertical uncertainty: 5–10 cm) that is freely available from www.dataforsyningen.dk. The lidar point cloud was converted to a canopy height model by subtracting the terrain model (DTM) from the surface model (DSM). The final set of variables had a resolution of 1.5 m (except one, see below). For all lidar processing and calculation, we used the OPALS tools (Pfeifer et al. 2014) version 2.3.1 in a Python 2.7 environment, and we used the supplier classification of points into terrain and vegetation that came with the dataset originally. The means and standard deviations of the following lidar variable were calculated for a 30 m radius circle centered in each study site to reflect actual levels and their variability within sites and their immediate surroundings. For further details on calculation of lidar variables see Valdez et al. (2021). The set of lidar variables encompassed: potential solar radiation (mean and std), adjusted solar radiation (i.e., solar radiation adjusted for vegetation cover; mean and std), amplitude (uncalibrated, but corrected for aircraft type and seasonality, see Valdez et al. 2021), vegetation height (mean and std), vegetation cover (mean, std), mean vegetation density at 0-100 cm, 1 -3 m, 3 -10 m and 10 -50 m, canopy openness (mean, std), terrain openness (mean, std), terrain slope (mean, std),  echo ratio (i.e., canopy complexity; mean, std), heat load (std) and mean fine-scale (0.5m) terrain roughness (Appendix D).

Response variables

OTU richness

As alternative measures of biodiversity, we used the richness of operational taxonomic units, i.e. OTUs (Blaxter et al. 2005) of fungi and soil microbes from metabarcoding of soil eDNA (Frøslev et al. 2017, Frøslev et al. 2019). Classical data collection of fungi is time consuming and OTU richness has been found to resemble classical observed species richness at least for groups of macrofungi that are feasible to include in field inventories (Frøslev et al. 2019). We used OTU richness of fungi and soil microbes as response variables to reflect diversity of species groups not monitored otherwise in this project.

Red-listed species: Site richness of red-listed species (belonging to the categories Critically endangered, endangered, vulnerable, near-threatened and data deficient) was calculated for vascular plants and bryophytes based on the current national red list (Moeslund et al. 2019a, Redlist.au.dk).

Indicator species: Indicator species include vascular plant species considered moderately to very sensitive to habitat alteration (Fredshavn et al. 2010, see Appendix E). The list of indicator species (Fredshavn et al. 2010) was developed to indicate favorable conservation status according to the Habitats Directive (European Commission 2007). Common to these indicator species is a preference for infertile habitats (low Ellenberg N and high Grime’s S values, Grime 1979).

Biotic uniqueness: Uniquity is a scale-dependent metric of biodiversity reflecting how unique the biodiversity of a given site is compared to the gamma diversity across the containing region or collection of sampled sites (Ejrnæs et al. 2018). Uniquity can be calculated based on both observational data as well as non-annotated DNA-data (e.g., OTUs) and hence can reflect both species uniqueness and genetic uniqueness. Contrary to other biodiversity metrics, uniquity accounts for sampling bias and spatial scale. Due to the built-in weighting method, uniqueness of non-annotated DNA-data can be calculated corresponding to e.g. red-listed species (Ejrnæs et al. 2018). Here, we calculated fungi and soil microbe uniquity according to Ejrnæs et al. (2018) in order to reflect the unique site contribution of fungi and soil microbe DNA to the gamma diversity of the collection of sites. Uniquity calculations were based on fungal and soil microbe OTU matrices, site habitat classes and weights from the full Biowide data set (n=130, Brunbjerg et al. 2019) combined with the wetland data set (n=44). The parameter X was set to 1000.

Explanatory variables

Soil moisture: The trimmed mean of 16 measures per site was used to reflect site soil moisture. For Biowide sites, we used the trimmed mean in August. We detected a systematic discrepancy between moisture in Biowide sites (measured in 2016) and wetland sites (measured in 2018), which could be accounted for by the summer of 2018 being extremely dry. We therefore interpolated the soil moisture trimmed mean values for wetland sites using the predicted values from a k nearest neighbor regression using soil moisture trimmed mean in Biowide sites (n=130, Brunbjerg et al. 2019) as response variable, Ellenberg F values as explanatory variable and k = 11.

Soil fertility: Good and reliable field-based measures of nutrient availability are difficult to obtain, as nutrient availability is extremely variable across time and space (Andersen et al. 2013 and references herein). In contrast, the nutrient ratio (mean site Ellenberg N/mean site Ellenberg R, Ellenberg et al. 1992) has been found to reflect eutrophication in wetlands and be highly correlated with the number of typical species in fens (Andersen et al. 2013). For each site we cal­culated mean Ellenberg N and Ellenberg R values (plant-based bioindication of nutrient status and soil pH, respectively) (Ellenberg et al. 1992). The Ellenberg nutrient ratio was used to reflect eutrophication and the idea of the ratio is to account for the fact that natural nutrient availability in wetlands increases with pH. To avoid circularity in analyses, plant species included in the plant-based conservation indicators (red-listed plants, typical plants) were removed before calculating the nutrient ratios for each model in question.

Encroachment: We made a rough classification of sites into two groups (open vegetation and scrub/forest vegetation) based on field photos. The encroachment variable was coarse i.e. ‘open’ represented mainly open and herb dominated vegetation but sites with scattered small shrubs were also categorized as ‘open’ plots as long as the shrubs did not dominate the plot. The two level factor variable was used as explanatory variable.

References:

Andersen, D. K., B. Nygaard, J. R. Fredshavn, and R. Ejrnæs. 2013. Cost-effective assessment of conservation status of fens. Applied Vegetation Science 16:491-501.

Blaxter, M., J. Mann, T. Chapman, F. Thomas, C. Whitton, R. Floyd, and E. Abebe. 2005. Defining operational taxonomic units using DNA barcode data. Philosophical Transactions of the Royal Society B: Biological Sciences 360:1935-1943.

Brunbjerg, A. K., H. H. Bruun, L. Brøndum, A. T. Classen, L. Dalby, K. Fog, T. G. Frøslev, I. Goldberg, A. J. Hansen, M. D. D. Hansen, T. T. Høye, A. A. Illum, T. Læssøe, G. S. Newman, L. Skipper, U. Søchting, and R. Ejrnæs. 2019. A systematic survey of regional multi-taxon biodiversity: evaluating strategies and coverage. BMC Ecology 19:43.

Brunbjerg, A. K., H. H. Bruun, L. Dalby, A. T. Classen, C. Fløjgaard, T. G. Frøslev, O. L. Pryds Hansen, T. T. Høye, J. E. Moeslund, J.-C. Svenning, and R. Ejrnæs. 2020. Multi-taxon inventory reveals highly consistent biodiversity responses to ecospace variation. Oikos 129:1381-1392.

Callahan, B. J., P. J. McMurdie, M. J. Rosen, A. W. Han, A. J. A. Johnson, and S. P. Holmes. 2016. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Meth 13:581-583.

Ejrnæs, R., T. G. Frøslev, T. T. Høye, R. Kjøller, A. Oddershede, A. K. Brunbjerg, A. J. Hansen, and H. H. Bruun. 2018. Uniquity: A general metric for biotic uniqueness of sites. Biological conservation 225:98-105.

Ellenberg, H., H. E. Weber, R. Düll, V. Wirth, W. Werner, and D. Paulissen. 1992. Zeigerwerte von pflanzen in Mitteleuropa. Scripta Geobotanica 18:1-248.

European Commission. 2007. Interpretation manual of European Union habitats EUR27.

Fløjgaard, C., T. G. Frøslev, A. K. Brunbjerg, H. H. Bruun, J. Moeslund, A. J. Hansen, and R. Ejrnæs. 2019. Predicting provenance of forensic soil samples: Linking soil to ecological habitats by metabarcoding and supervised classification. Plos One 14:e0202844.

Fredshavn, J., R. Ejrnæs, and B. Nygaard. 2010. Teknisk anvisning for kortlægning af terrestriske naturtyper. TA-N3, Version 1.04. Fagdatacenter for Biodiversitet og Terrestriske Naturdata, Danmarks Miljøundersøgelser.

Frøslev, T. G., R. Kjøller, H. H. Bruun, R. Ejrnæs, A. K. Brunbjerg, C. Pietroni, and A. J. Hansen. 2017. Algorithm for post-clustering curation of DNA amplicon data yields reliable biodiversity estimates. Nature communications 8:1188.

Frøslev, T. G., R. Kjøller, H. H. Bruun, R. Ejrnæs, A. J. Hansen, T. Læssøe, and J. Heilmann-Clausen. 2019. Man against machine: Do fungal fruitbodies and eDNA give similar biodiversity assessments across broad environmental gradients? Biological conservation 233:201-212.

Grime, J. P. 1979. Plant strategies and vegetation processes. John Wiley & Sons, New York.

Moeslund, J. E., B. Nygaard, R. Ejrnæs, N. Bell, L. D. Bruun, R. Bygebjerg, H. Carl, J. Damgaard, E. Dylmer, M. Elmeros, K. Flensted, K. Fog, I. Goldberg, H. Gønget, F. Helsing, M. Holmen, P. Jørum, J. Lissner, T. Læssøe, H. B. Madsen, J. Misser, P. R. Møller, O. F. Nielsen, K. Olsen, J. Sterup, U. Søchting, P. Wiberg-Larsen, and P. Wind. 2019a. Den danske Rødliste. Aarhus Universitet, DCE – Nationalt Center  for Miljø og Energi. www.redlist.au.dk.

Nygaard, B., C. Damgaard, K. E. Nielsen, J. Bladt, and R. Ejrnæs. 2017. Terrestriske Naturtyper 2004 – 2015. NOVANA. Aarhus Universitet, DCE – Nationalt Center  for Miljø og Energi. www.novana.au.dk.

Pfeifer, N., G. Mandlburger, J. Otepka, and W. Karel. 2014. OPALS – A framework for Airborne Laser Scanning data analysis. Computers, Environment and Urban Systems 45:125-136.

Valdez, J. W., A. K. Brunbjerg, C. Fløjgaard, L. Dalby, K. K. Clausen, M. Pärtel, N. Pfeifer, M. Hollaus, M. H. Wimmer, R. Ejrnæs, and J. E. Moeslund. 2021. Relationships between macro-fungal dark diversity and habitat parameters using LiDAR. Fungal Ecology 51:101054.

Files can be accessed using Microsoft excel.