Land uplift creates important meadow habitat and a potential original niche for grassland species

Abstract

Semi-natural grasslands have been severely affected by agricultural land-use change. However, the isostatic land adjustment following deglaciation in the Northern Hemisphere means that new land is continually being created in coastal areas. We modelled isostatic adjustment during the last 4000 years in a region of the Baltic coast to estimate the emergence of potential grassland habitat. We also compared the α and β diversity of existing managed and abandoned coastal meadows, and assessed their contribution to biodiversity at landscape scales. We estimated that half the 7866 km² of emerging land had the potential to become coastal meadow habitat, which is an order of magnitude larger than the total area of all valuable semi-natural grassland in the study region today. The small area of managed coastal habitat remaining was found to have a disproportionate influence on the richness of threatened species at landscape scales, but our results also show that continued management is essential for the maintenance of grassland biodiversity. Our combination of approaches identifies uplifted coastal meadows as an additional original niche for grassland plant species, while highlighting that low-intensity disturbance through grassland management is essential for the maintenance of diversity at multiple scales.

1. Introduction

Habitat destruction is the most serious threat to biodiversity today, with diversity losses having major consequences for ecosystem functioning and human society [1–3]. In Europe, twentieth-century agricultural intensification has led to a massive reduction in the cover of species-rich semi-natural pastures and meadows in rural areas [4,5], with consequences for plant biodiversity and ecosystem service provision [6,7]. Today, a large fraction of red-listed species are strongly associated with agricultural landscapes and the grasslands that remain [8,9]. One type of grassland of particular interest is the coastal meadow, listed as priority habitat in the EU habitat directive (92/43/EEC). Like other types of grassland in the region, these habitats face ongoing pressure from abandonment and agricultural land-use change [10].

The rapid loss of semi-natural grassland is at odds with their creation, which is a result of a long continuity of low-intensity management by grazing and haymaking [9,11]. Through analyses of historical maps, we have a good understanding of grassland management and change during the past few centuries [5,12]. However, there is an ongoing discussion regarding the origins of the species associated with such grasslands today, as well as the (pre-)historically naturally open areas which form their original niche. Fire, flooding and grazing by large herbivores have all been suggested as important historical sources of small-scale disturbance that provided habitat for light-demanding and short-statured plant species [13,14]. Human settlement and the development of agricultural systems is then thought to have provided larger and more stable areas of open habitat into which such plant species shifted their realized niche [15].

Modern-day habitat destruction means that the maintenance, restoration and re-creation of grasslands is an important priority for conservation management in Europe [16,17]. However, the process of shore displacement through glacial isostatic adjustment, whereby coastal land areas are slowly ‘rebounding', following the retreat of the ice-sheet since the Last Glacial Maximum 22 000 years ago, means that new land is actually being created in these ecologically important regions across the Northern Hemisphere [18,19]. A handful of previous studies have used coastal meadows as model systems, focusing on succession and the build-up of the vegetation and seed bank layers (e.g. [20–22]). Despite acknowledging differences between uplift areas with and without grassland management [23,24], there has been very little focus on the role of management in optimizing the potential for these new habitats to contribute to biodiversity at local and landscape scales.

While the grassland management of newly emerged land through livestock grazing and mowing might benefit biodiversity today, such land areas may have provided naturally open habitat for species in the past. In this study, we aim to investigate isostatic adjustment as a natural process creating grassland habitat and an original niche for grassland specialist species, and to identify the biodiversity and conservation value of remaining effectively managed coastal meadow habitats. We first use a simple shore displacement model to estimate the magnitude and rate of emergence of potential coastal meadow habitat during the last 4000 years in a region on the Baltic coast of Sweden. We then compare the established vegetation of managed and abandoned coastal meadows to examine the impact of low-intensity disturbance on plant biodiversity and community build-up. Finally, we evaluate the contribution that managed coastal meadows can have for plant species richness at larger spatial scales. We predict that: (1) a substantial amount of potentially valuable meadow habitat has been created through isostatic adjustment; (2) grazing on coastal meadows increases plant diversity at local scales; and (3) that the presence of managed coastal meadow habitat has a positive contribution to plant species richness at larger scales, providing potential habitat for plant species not available elsewhere in the landscape.

2. Material and methods

(a). Study region

The historical provinces of Södermanland (8388 km²) and Uppland (12 676 km²) are located on the east coast of Sweden, with the city of Stockholm located where the two provinces meet (59°19'28″ N 18°4'21″ E; figure 1). Following the end of the last Ice Age, evidence of agricultural and fishing communities exists in this region from approximately 3000 years ago [25]. After ice retreated from the area, the isostatic adjustment was rapid and the land has risen more than 160 m in the study area. At present, the isostatic adjustment in the study region is between 2 and 4 mm per year. Depending on shore topography, the effects can be quite dramatic along a relatively flat relief and in shallow bays, where several metres of new land can become exposed within a few years. The Baltic Sea is not influenced by tidal water. Precipitation in the region is approximately 450–600 mm per year, and mean temperature in January is −2 to −5°C and around 16° C in July (data from Swedish Metrological and Hydrological Institute, SMHI; https://www.smhi.se/klimatdata). Specific site descriptions are given in the separate sub-sections, below.

Figure 1.

The two regions (a) Uppland and (b) Södertörn where the isostatic adjustment was modelled over the last 4000 years. Local plant diversity was measured in transects in 10 managed and 10 abandoned coastal meadows (stars), while landscape-scale plant species richness was derived from the flora of Södermanland province (dashed line). Background topography from national digital elevation model © Lantmäteriet.

(b). Isostatic adjustment

We modelled the isostatic adjustment for the past 4000 years, corresponding to the settlement of humans in the region, as inferred by the appearance of crop species in the pollen record [26,27]. Uplift was calculated separately for the coast of Uppland and the island of Södertörn in Södermanland (figure 1), due to differences in adjustment rate with latitude. Uplift from 2000 BC until AD 2000 was modelled using a database of lake sediment data, where the time of lake isolation was calculated using diatom analysis and ¹⁴C dating of the sediment [28–30]. Diatoms in lake sediments show the lake's cut-off point from the sea, and combined with lake elevation can produce a simple shore displacement model, which uses interpolation to estimate the year in which each pixel in the study area was exposed from the sea during the last 4000 years. This year was then assigned to the century of emergence. Then, national LIDAR elevation data (resolution 4 m²; resampled to 100 m² for data management purposes using the nearest neighbour method) was used to calculate land surface slope angle. The shore displacement model was then overlain with the slope data to separately identify uplifted areas with 0–1° and 0–2° slopes. Such flat areas could potentially act as meadow habitat, as opposed to areas with steeper slopes, which are indicative of exposed bedrock. Two ranges of slopes are used to give an indication of potential grassland with both a more restrictive and a more generous definition of flat land. Raster maps for surface water (lakes and sea) were then created from national vector maps (Uppland: terrain map, 1 : 50 000; Södertörn: property map, 1 : 10 000) using the pixel centre method. These were then used to remove areas interpolated as being uplifted, but that were actually still under water. Analyses were carried out in ArcGIS v. 10 (ESRI, Redlands, USA).

(c). Plant biodiversity

(i). Local scale

In the summer of 2012 we surveyed plant species occurrences in 20 coastal meadows (sites; figure 1). These grasslands have never been synthetically fertilized, drained or tilled for crop production. Meadows are typically dominated by clonal species and stress tolerators, where grasses dominate close to the sea and herbs higher up were the soils are drier, and physical disturbance (ice scouring) and salinity is lower [10]. Ten of our sites are currently managed by grazing and 10 have been abandoned. All sites were previously managed as meadow according to land-use maps available from around 1900 (Swedish: Häradskartan), and remained open habitat in later land-use maps from the 1950s (Swedish: Ekonomiska kartan), indicating that abandonment occurred during the past 50 or so years. Managed sites were subject to low-intensity grazing by cattle (sward height around 15 cm). At each site, twelve 1 m² plots were arranged along a transect running perpendicular to the shoreline in which the presence of all vascular plants were recorded. Plots were placed on each side of the high water mark (defined as the boundary of the presence of Ranunculus acris, combined with the presence of seaweed and reed debris) and then 2, 4, 8 and 16 m away from the high water mark in each direction. An additional plot was placed at the water's edge (or the closest point with vegetation cover) and one was also placed at the upper end of the coastal meadow furthest from the water. Each plot was placed randomly across the 20 m width of the transect (see electronic supplementary material, figure S1). The position along the transect was used as a proxy for relative plot age, where 1 was the plot closest to the water and 12 furthest from the shore. One particularly narrow managed meadow had only 10 plots in its transect, resulting in a total of 238 plots in the whole investigation.

All biodiversity analyses were carried out using R 3.0–3.2 [31]. Before assessing the effects of management and plot age on α diversity (defined here as species richness, i.e. the number of species within a plot, site or landscape), we needed to check for spatial autocorrelation at both the plot and the site level. Moran's I (R function: Moran.I in the ape package [32]) was calculated, finding significant correlation of α diversity in plots situated closer together according to the GPS position taken at each plot (Moran's I = 0.7, p = <0.001), but not sites (centroid of all plots within a site; Moran's I = −0.15, p = 0.8). For plot-level analyses, we therefore used quasi-generalized linear mixed models (GLMMs) with Poisson distributions, corrected for overdispersion using penalised quasi-likelihood with a log link (function glmmPQL in the MASS package [33]). A model explaining α diversity was created using both management, relative plot age and their interaction as fixed effects, with site as a random effect to control for inherent similarity within sites. Non-significant variables (p = >0.05) were dropped in a stepwise manner in order of descending p-values until only significant effects remained, along with any non-significant component variables of significant interactions. At the site level, we used Wilcoxon–Mann–Whitney tests (function: wilcoxon.test), identifying whether α diversity was higher in the managed than the abandoned meadows at both the plot and the site level.

Differences in β diversity between managed and abandoned plots and sites were assessed using multivariate homogeneity of groups dispersions, which plots community distance matrices in multivariate space, calculating the centroid point of each group of communities (in our case managed or abandoned; function betadisper in the vegan package [34,35]). We used Raup–Crick dissimilarity of species presence–absence at the plot level and Chao dissimilarity of the number of plots in which each species was observed at the site level using vegan's vegdist function. Raup–Crick and Chao dissimilarity are designed to account for potential missing species using presence–absence and abundance data, respectively. Before calculating β diversity, the Chao and Raup–Crick distance matrices were assessed for spatial autocorrelation using Mantel tests (function mantel in vegan) comparing geographical (plot locations, site centroids) and community distance, finding no effect (plots: Mantel's r = 0.037, significance = 0.39; sites: Mantel's r = 0.041, significance = 0.26). The effect of management on β diversity was then tested by conducting analysis of variance on the distances between managed and abandoned communities and their respective group centroid points (function: anova). Species lists were compared against the 2015 Swedish Red List and the list of 35 typical and characteristic species of Baltic coastal meadows [36,37].

(ii). Landscape scale

We used the publicly available data from the plant biodiversity atlas over Södermanland [38] combined with the Swedish government's database of ecologically valuable grasslands (TUVA; http://www.sjv.se/tuva) to assess the influence of the area of coastal meadows on plant species richness in grid squares 5 × 5 km in size. First, the free map of Sweden (1 : 1 000 000; http://www.lantmateriet.se/en/Maps-and-geographic-information/Maps/Sverigekartor) was used to identify all land areas (including islands) and coastal boundaries in Södermanland. Next, the GIS file over the TUVA grasslands was downloaded, and cross-referenced with the online database to identify which grasslands contained the Baltic meadow vegetation type, and the area of this grassland type within each meadow. These three layers (coast, land and grassland) were then split into the 5 × 5 km grid of the plant biodiversity atlas data. To avoid confounding effects of small areas of land containing relatively few species, and to ensure that only truly coastal areas were compared, grid squares containing at least 6.25 km² of land area (one-quarter grid square) and 5 km of coastline (equivalent to one side of the grid square) were retained for analysis, resulting in 62 grid squares, containing 377 ha of the total 570 ha of Baltic coastal meadow in the province of Södermanland. GIS data preparation was carried out using PostGIS [39].

Predictors of total species richness and the number of red-listed species were then analysed using the same stepwise quasi-GLMM method described above. Because of collinearity between our predictor variables land area and coast length (Spearman's ρ = −0.64 p = <0.01; function: cor.test), only land area and coastal meadow area (both zero-mean scaled) and their interaction were included in the full models. Grid square identity was included as a random variable due to the detection of spatial autocorrelation (centre of grid square) for both species richness (Moran's I = 0.04, p = <0.001) and the number of red-listed species (Moran's I = 0.06, p = <0.001).

3. Results

(a). Isostatic adjustment

Our model of land uplift calculates that 7866 km² of land has emerged in Uppland and Södertörn during the last 4000 years (figure 2). Using flat topography as an indicator of potential grassland habitat, it shows that around 2000 to 4000 km² of coastal meadows may have been created during this time period, at a rate of around 50–150 km² per century. During the twentieth century, when the fragmentation of grasslands has been most severe, 48–80 km² of potential coastal meadow habitat has emerged from the sea.

Figure 2.

Modelled land uplift and the emergence of potential meadow for the Uppland study region during the past 4000 years. Map colours show approximate time of (a) land emergence and (b,c) potential meadow, i.e. emerging land with a slope of (b) 0–2° and (c) 0–1°. Black indicates areas that were either above sea level 4000 years ago or that remain still below sea level today. (d) Modelled rate of uplift per century for both Uppland and Södertörn.

(b). Plant biodiversity

A total of 137 plant species were recorded in the local scale field survey. Managed sites had a significantly higher α diversity at the site level, with a mean ± s.d. of 37 ± 10.62 in managed sites and 14.4 ± 6.4 in abandoned sites (Wilcoxon–Mann–Whitney W = 98, p = <0.001; figure 3a). Management also had a large effect on plot-level α diversity, with a mean ± s.d. of 10.77 ± 4.79, compared with 4.64 ± 2.61 in abandoned meadows (GLMM coefficient (managed) = +0.88, p = 0.001; figure 3a). Relative plot age was also included in the final model, having a positive effect on α diversity (figure 3b; GLMM coefficient = +0.063, p = <0.001). In addition to α diversity, β diversity was also higher between managed plots than abandoned plots (F_1,236 = 13.79, p = <0.001), but for sites, β diversity was higher between abandoned meadows (F_1,18 = 5.28, p = 0.03). All 10 managed sites contained Baltic coastal meadow species, with 12 of these specialists observed in our plots. Only five such species were found in our abandoned sites, although it was in an abandoned plot that the only red-listed species in the study was observed. At the landscape scale, the total number of species in a grid square was only predicted by land area (GLMM coefficient 0.033, p = 0.04). The number of red-listed species was most strongly related to the cover of coastal meadows, with their importance increasing with decreasing total land area within a grid square. The full model was retained, consisting of meadow area (GLMM coefficient 0.24, p = <0.01), land area (GLMM coefficient 0.003, p = 0.96) and their interaction (GLMM coefficient −0.32, p = 0.02).

Figure 3.

Effect of management on plant species richness in 10 managed (shaded) and 10 abandoned (white) coastal meadows in southern Sweden. (a) Plot- (1 m²) and site-level α diversity, showing median, quartiles, range and outliers. Asterisks indicate significant differences between abandoned and managed grasslands at both the plot (GLMM) and site (Wilcoxon–Mann–Whitney) level. (b) Increases in plot-level α diversity with relative age, where plot age ranges from 1 (closest to shore) to 12 (furthest from shore; see electronic supplementary material, figure S1). Trend line calculated using R's lowess function.

4. Discussion

(a). Natural grassland creation

The deglaciation that occurred in the Baltic approximately 10 000 years ago has resulted in the emergence of potential grassland habitat through isostatic adjustment. During the last 4000 years, we estimate that 7900 km² of new land has emerged in our study region, with 48–80 km² estimated to have emerged in the twentieth century. This is compared with the total of 0.3 km² Baltic coastal meadow and 265 km² total managed valuable grassland habitat across the whole study region. Although our basic model gives only an approximation of the rate of land uplift, it is very illustrative to highlight that in our study area, many square kilometres of land emerges from the sea every century. This process has occurred in many regions across the Northern Hemisphere where there is an isostatic adjustment, and particularly where shorelines are relatively flat.

This large area of natural grassland habitat may have provided an important niche for the low-competitive grassland species found in today's semi-natural grasslands. The naturally open areas that have previously been suggested as pre-agricultural habitats for grassland plants, such those arising following storm damage, flooding and fire [14], will have been relatively isolated and localized in space, with more permanent grassland only occurring following the introduction of agriculture to a region [15]. Our results indicate that isostatic adjustment has resulted in the gradual and continuous creation of large and well-connected areas of naturally open land. However, our results also show that regular disturbance is required for the maintenance of species-rich plant communities. Disturbance from physical stresses present at the seashore [40], along with the presence of wild ungulates which are at least understood to be able to maintain (if not create) open habitat [14,41], may have provided enough disturbance to maintain openness in the early Holocene. Therefore, we consider that coastal grasslands may have been overlooked as another potential original niche of grassland plant species. With the arrival of farming, these coastal grasslands will have been regularly managed, with their salty grass considered a valuable commodity [42]. Although it is known that open areas of vegetation were present in the northern European landscape prior to human settlement [26,27], it would be interesting to compare directly coastal and inland areas to see if an imprint of higher grassland cover due to land uplift is detectable in the pollen record.

(b). Grassland management and biodiversity

Although it is not known to what extent different factors contributed to openness in the pre-agricultural landscape, the results from our field survey show that grassland management is required in contemporary coastal meadows to maintain species-rich target plant communities at local and landscape scales. We found that livestock grazing in coastal meadows resulted in more plant α and β diversity and a more effective build-up of biodiversity along the transect compared with meadows that were abandoned around 50 years ago at the plot level. Surprisingly, the low-biodiversity communities of the abandoned meadows were more variable among sites. It is well known that grassland management is beneficial to plant diversity compared with unmanaged and abandoned sites, with examples from a range of environments in Europe and elsewhere [43–45]. In addition to preventing the dominance of highly competitive species, grassland management can promote seed dispersal through the actions of grazing ungulates and meadow managers [46,47], while the presence of livestock also provides gaps for establishment [48]. The combination of these factors has resulted in different types of semi-natural grasslands being among the most species-rich habitat types in the world at small scales [49]. However, in coastal meadow habitats, the relationship between grassland management and biodiversity has not always been so clear [24,50]. It is possible that in these habitats, the effects of grassland management are especially dependent on the character of management employed, as has been shown in a dune system [51].

The number of species per plot increased significantly with distance from the shore in both managed and abandoned coastal meadows. This has previously been shown in other studies in similar environments [22,24]. However, we also showed that meadows subject to grassland management experience a more effective build-up of biodiversity in time than those left to develop following abandonment. While the increase in land area resulting from isostatic adjustment provides new habitat area even in landscapes lacking grassland management [52], our results indicate that regular disturbance is required to develop and maintain niche space for coastal meadow communities. In the distant past, this may have been provided through the actions of wild herbivores. Today however, regular and effective management is required if these areas of potential grassland habitat emerging from the sea are to support species-rich plant communities.

In addition to promoting local-scale plant biodiversity, grassland management of coastal meadow habitats is also important for species richness at the landscape scale. As expected from the species–area relationship [53], the amount of land in coastal areas is an important determinant for the total number of species present. However, for the number of red-listed species, the area of coastal meadow was more important than total land area. The effect was small, but these grasslands only make up a small fraction of the total land area within each grid square. This relationship reflects the threatened nature of these habitats, which provide a home for species not able to compete elsewhere. Of course, a number of other factors are important determinants of plant biodiversity, such as landscape composition and heterogeneity, the presence of other core habitats and land-use history [5,54]. This was reflected in the fact that the importance of coastal meadows diminished with increasing land area within a grid square, as a larger area can support a greater variety of habitats. Nevertheless, our results indicate that the presence of small, diverse habitats such as coastal meadows can have disproportionate positive effects on biodiversity at relatively large spatial scales.

(c). Concluding remarks

Our modelling exercise revealed the extent to which potential grassland habitat has emerged through isostatic adjustment during the mid- to late Holocene. Compared with the present-day area of coastal meadow habitat, many more times potential grassland area emerges each century, being more comparable with the total extent of valuable grassland existing in the whole region. We think that this continuous creation of open land has been an overlooked original niche for grassland plant species. However, we also showed that low-intensity disturbance—today provided by grazing livestock—is necessary to maintain and maximize diversity in these habitats, which are seen as a priority for conservation management at the international level. Coastal meadows are important and protected habitats for biodiversity of different organism groups [7] (EU habitat directive 92/43/EEC), and we have shown that managed coastal meadows are important for plant biodiversity at multiple spatial scales. Unfortunately, coastal areas in general are increasingly threatened by changes in land use and climate [10]. Therefore, efforts to manage existing coastal meadows and restore suitable areas should be prioritized, as well as long-term plans for the management of future grassland that is currently under water.

Fieldwork permissions

All fieldwork was carried out under the Swedish Right of Public Access to the countryside.

Data accessibility

Local- and landscape-scale plant diversity data, georeferenced raster files showing total uplift (shore displacement model) and the emergence of potential grassland habitat, and summary tables for these rasters, are available at http://dx.doi.org/10.17045/sthlmuni.5514127 [55]. External data used in the analyses are cited in line.

Authors' contributions

S.A.O.C. conceived the study, supervised data collection and uplift modelling, and contributed to writing the manuscript. A.G.A. analysed the biodiversity data and led the writing of the manuscript.

Competing interests

We declare we have no competing interests.

Funding

The study was financed by the strategic research program EkoKlim at Stockholm University. A.G.A. is also supported by the Swedish research council Formas (2015-1065). The Swedish LifeWatch Analysis portal, from which the biodiversity-atlas data were extracted is funded by the Swedish Research Council and the Swedish Environmental Protection Agency through the Swedish LifeWatch Project—grant no. 829-2009-6278.