Self‐facilitation and negative species interactions could drive microscale vegetation mosaic in a floating fen

Aim: The formation of a local vegetation mosaic may be attributed to local variation in abiotic environmental conditions. Recent research, however, indicates that self-facilitating organisms and negative species interactions may be a driving factor. In this study, we explore whether heterogeneous geohydrological conditions or vegetation feedbacks and interactions could be responsible for a vegetation mosaic of rich and poor fen species. Location: Lake Aturtaun, Roundstone Bog, Ireland. Methods: In a floating fen, transects were set out to analyze the relation between vegetation type and rock–peat distance and porewater electrical conductivity. Furthermore, three distinct vegetation types were studied: rich fen, poor fen and patches of poor fen within rich fen vegetation. Biogeochemical measurements were conducted in a vertical profile to distinguish abiotic conditions of distinct vegetation types. Results: Geohydrological conditions may drive the distribution of poor and rich fen species at a larger scale in the floating fen, due to the supply of minerotrophic groundwater. Interestingly, both rich and poor fen vegetation occurred in a mosaic, when electrical conductivity values at 50 cm depth were between 300 µS/cm and 450 µS/cm. Although environmental conditions were homogeneous at 50 cm, they differed markedly between rich and poor fen vegetation at 10 cm depth. Specifically, our measurements indicate that poor fen vegetation lowered porewater alkalinity, bicarbonate concentrations and pH. No effects of rich fen vegetation at 10 cm depth on biogeochemistry was measured. However, rich fen litter had a higher mineralization rate than poor fen litter, which increases the influence of minerotrophic water in rich fen habitat. Conclusions: These results strengthen our hypothesis that species can drive formation of vegetation mosaics under environmentally homogeneous conditions in a floating fen. Positive intraspecific self-facilitating mechanisms and negative species interactions could be responsible for a stable coexistence of species, even leading to local ecosystem engineering by the species, explaining the local vegetation mosaic at the microscale level in a floating fen.


| INTRODUC TI ON
Peatlands can be complex with large variations in structure and hydrology Lamers et al., 2015;Rydin & Jeglum, 2006). Globally, the occurrence and characteristics of peatlands are primarily driven by climate and geomorphological processes (Joosten & Clarke, 2002). On a regional scale, peatlands often consist of large complexes that include rich fen, poor fen and bog vegetation (Joosten & Clarke, 2002;Rydin & Jeglum, 2006), which can be related to heterogeneity in environmental conditions, such as hydrology, acid buffering capacity and nutrient supply . Differences in environmental conditions can result in local vegetation mosaics within one ecosystem, resulting in distinct dominating species depending on prevailing abiotic conditions (Kuhry, Nicholson, Gignac, Vitt, & Bayley, 1993). However, next to abiotic influences, biotic interactions could play an important role in the formation of a local patchy landscape Rietkerk & van de Koppel, 2008), as vegetation mosaics have also been observed in environmentally homogeneous conditions. Species can on the one hand overcome adverse environmental conditions by means of self-facilitation and on the other hand can even change these environmental conditions by ecosystem engineering.
Overcoming adverse environmental conditions by means of self-facilitation can play an important role when a species invades in an ecosystem or during succession (Callaway, 1995;Holmgren, Scheffer, & Huston, 1997). Many examples of this concept have been found in a diverse range of ecosystems with plants exposed to stress, such as heat or drought (Callaway, 1995). While high plant densities can lead to competition for nutrients, space or light (Stachowicz, 2001), the establishment of new conspecific seedlings under stressful conditions will mostly take place when plants are growing in high densities and facilitate their survival. For example, the canopy of "nurse plants" facilitates establishment of seedlings in dry conditions by alleviating environmental stress (Holmgren et al., 1997). In a range of ecosystem types, such as salt marshes and seagrass ecosystems, negative species interactions and intraspecific self-facilitation processes have been identified that resulted in the coexistence of two dominating species (van der Heide et al., 2012; Van Wesenbeeck, Koppel, Herman, Bakker, & Bouma, 2007). Even though environmental conditions were similar, both negative species interactions and intraspecific self-facilitation induced bistability of dominating species.
Another concept that can play an important role in vegetation patterning is ecosystem engineering, where organisms, either deliberately or inadvertently, modify their physical habitat (Jones, Lawton, & Shachak, 1994. Paleoecological studies have shown that peatland ecosystems can exhibit bistability, such as in hummock-hollow formation over time Moore, 1977;Walker & Walker, 1961). These small-scale patterns are remarkably stable and resilient to changes in environmental conditions (Belyea & Clymo, 2001), which is often the result of habitat-modifying properties of peat mosses (Nungesser, 2003) of which a number of mechanisms have been identified (van Breemen, 1995). Many modeling studies describe bistability in peatlands Nungesser, 2003;Rietkerk, Dekker, Wassen, Verkroost, & Bierkens, 2004). However, there are few empirical studies on this subject because of the large time scale of peat formation and patterning Gunnarsson, Malmer, & Rydin, 2002).
In this study, we empirically explored whether heterogeneous geohydrological conditions or vegetation interactions could be responsible for a vegetation mosaic of rich and poor fen species in a floating fen. For this, we studied distinct rich and poor fen vegetation in a floating mire that seemed to be bistable for at least 40 years (Figure 1; Appendix S1; Van der Maarel & Roozen, 1975;van Groenendael, Hochstenbach, Mansveld, & Roozen, 1975): poor fen vegetation dominated by Sphagnum spp. and rich fen vegetation dominated by Schoenus nigricans. We hypothesized that bistability of poor and rich fen species is caused by vegetation interactions that overcome environmental stress, increase their own density and exclude other species. Next to self-facilitation, we also hypothesized that ecosystem engineering of both poor and rich fen species could affect the local environment and contributed to the vegetation mosaic.

| Study site
Lake Aturtaun in Roundstone Bog, Connemara, Ireland

| Geohydrology
Geohydrology of the floating fen was studied by determining the hydraulic head and electrical conductivity (EC) of the porewater throughout the mire, and by measuring rock depth below the peat surface. The calcareous rock below the peat in Roundstone Bog is known to enrich groundwater and surface water by dissolution of calcium and bicarbonate (Grootjans et al., 2016;Jenkin, Fallick, & Leake, 1992). In 2014, the hydraulic head was measured with piezometers that were placed evenly distributed throughout the floating fen (n = 10). In 2018, EC was measured at two depths (10 and 50 cm) across the fen in four transects, which consisted of 11 plots (Figure 1b), using a 2-m long EC probe calibrated with a handheld EC meter (Tetracon® 325, WTW electrode, pH/Cond 3,320 multimeter, Wissenschaftlich-Technische Werkstätten, Weilheim, Germany). The distance between peat surface and underlying rock layer was determined using a 4-m long PVC-tube in the same transects in 2018.

| Biogeochemistry
In each vegetation plot, pore water was sampled in a vertical profile to assess the potential influence of vegetation and water below the float- Total Inorganic Carbon (TIC) concentration in peat soil pore water was measured using infrared gas analyses (IRGA, ABB Advance Optima, Frankfurt, Germany). Carbon dioxide levels in pore water were calculated from TIC concentrations, temperature, pH, and carbonic acid equilibrium constants (K a ) (Dickson & Millero, 1987), according to the equations below (Stumm & Morgan, 1996): The pH and alkalinity (cf. Roelofs, 1983) were determined on the sampling day using a handheld meter (Multi 340i meter, Wissenschaftlich-Technische Werkstätten GmbH) connected to a pH probe (Orion 9156BNWP; Thermo Fisher Scientific).
In 2014, peat soil samples of a known volume were taken in every vegetation plot at approximately 10 and 50 cm depth with a half-cylinder chamber peat corer (50 cm long, Ø 5 cm). Subsamples were dried at 70°C for 48 hr to determine soil dry weight and bulk density. Bioavailable phosphorus (P) (Olsen-P extract;Olsen, 1954)   and methane (CH 4 ) production rate with an infrared carbon analyzer (IRGA; ABB Analytical). Base saturation was estimated using NaCl extraction (see above). Concentrations of cations displaced by Na were measured by ICP (and pH measurement for H + ) and used as a proxy for base saturation (BS) (Kleijn, Bekker, Bobbink, Graaf, & Roelofs, 2008).

| Statistical analyses
Normal distribution of the residuals and homogeneity of variance of the data were tested with the Shapiro-Wilk test, Q-Q plots and Levene's test, and when necessary data were transformed    Figure 2a,b). Proximity of the rock layer to the peat surface was significantly correlated to EC (R 2 = 0.42, p < 0.001; Figure 2c). Nevertheless, there was considerable overlap between EC values and species occurrence. When EC values ranged between ~300 µS/cm and 450 µS/cm, both rich and poor fen vegetation occurred (Figure 2c, gray area). At 10 cm depth, conductivity was higher compared to values found at 50 cm depth, but there was less variability in the spatial EC pattern and the relationship with rock depth was less strong (Appendix S4). We quantified upward water seepage in the entire floating fen, with an average hydraulic head of 1.6 ± 0.3 cm.

| Biogeochemistry
Nutrient concentrations in the floating fen were low and we did not find any differences in the depth profile or between vegetation types. The average total dissolved phosphorus concentration was 2.6 ± 0.4 µmol/L, and NH  Figure 3). In the poor fen vegetation, pH at 10 cm depth did not statistically differ compared to +5 cm and to 50 cm depth, but the pH was significantly lower at +5 cm compared to 50 cm depth. Alkalinity was lowest in the hummock and the upper soil layer and was significantly higher at 50 cm depth. HCO Anaerobic CO 2 production rates of the rich and poor fen litter differed significantly (Figure 4). No CH 4 was detected in the samples.

| Habitat modification by poor fen vegetation
Poor fen species, specifically peat mosses (Sphagnum spp.), generate positive feedbacks enabling them to create acidic, nutrient-poor, F I G U R E 3 (a) Porewater pH; (b) alkalinity (mEq/L); (c) bicarbonate (µmol/L); and (d) base saturation (%) at +5 (hummock), −10 and −50 cm in poor fen (n = 4 ± SE), patches of poor fen within rich fen vegetation (n = 3 ± SE) and rich fen (n = 3 ± SE) vegetation plots. Note that the yaxis of the pH (a) starts at 4. Significant differences are indicated by either capitalized (between vegetation types at −10 cm), capitalized and underlined (between vegetation types at −10 cm), non-capitalized (depth within poor fen vegetation) or bold non-capitalized (depth within poor fen in rich fen vegetation) and underlined letters (depth within rich fen vegetation) F I G U R E 4 CO 2 production rates of poor fen (n = 4 ± SE) and rich fen (n = 4 ± SE) peat (mmol CO 2 /L fresh soil/day) after five days of anaerobic incubation with peat collected in June 2018. No CH 4 was detected. Significance level is indicated by **<0.01 van BERGEn Et al. cold and anoxic conditions. These conditions positively affect peat accumulation (leading to ombrotrophication) and stimulate peat moss dominance (van Breemen, 1995). We clearly observed acidification of the upper soil layer in peat moss plots in the floating fen, which has been observed in many studies and habitats (Clymo, 1964;Cusell et al., 2015;Hájek & Adamec, 2009;Harpenslager, van Dijk, et al., 2015;Soudzilovskaia et al., 2010;van den Elzen et al., 2017). Although we observed upward seepage of water from below the floating mire in the entire floating fen (average hydraulic head of 1.6 ± 0.3 cm), together with a high base saturation in all vegetation types and soil layers (>90%), poor fen vegetation still had a significantly lower pH, alkalinity and HCO − 3 concentration in the hummock and at 10 cm depth (Figure 3). Only in patches of poor fen within rich fen vegetation, no significant differences were found in the vertical pH profile, but they showed the same trend as the poor fen vegetation (Figure 3), indicating patches of poor fen within rich fen vegetation had more difficulties with acidifying their environment. Soudzilovskaia et al. (2010) suggested that active release of protons by Sphagnum species was not an important mechanism of peat acidification during the shift from fen to bog.
However, peat mosses in poor fens form a hummock, because they produce high amounts of secondary metabolites (e.g., phenolic compounds), which result in slower decomposition rates compared to rich fen species (Clymo, 1964;Clymo & Hayward, 1982;Mettrop, Cusell, Kooijman, & Lamers, 2014;Verhoeven & Liefveld, 1997). We did not identify a significant difference in seepage of HCO − 3 -rich groundwater throughout the fen, but likely the hummock formed a groundwater mound that limited the supply of acid-neutralizing HCO − 3 -rich groundwater. The slower decomposition in poor fen was indeed confirmed by the low CO 2 production in poor fen litter, mainly consisting of peat moss biomass (1.40 ± 0.07 mmol CO 2 /L fresh soil/day). Low poor fen decomposition rates together with relatively high production rates result in a high net peat production leading to ombrotrophication. With increasing distance from buffered conditions combined with the ability of peat mosses to retain rainwater, the influence of nutrient-poor and less buffered rainwater increases, and acids produced are less easily neutralized (Bootsma et al., 2002;Granath et al., 2010;Soudzilovskaia et al., 2010;van Breemen, 1995). Nevertheless, we still found the highest bio-available P concentration in poor fen vegetation in the upper soil layer, which may be explained by either the fact that sphagnum lacks roots able to mobilize this P fraction or by self-facilitating feedbacks of poor fen species that cause a buoyancy-driven upward flow of underlying water with an extra supply of nutrients Rappoldt et al., 2003).

| Habitat modification by rich fen vegetation
Rich fen litter, mainly consisting of Schoenus nigricans litter, showed a 1.8 times higher potential CO 2 production rate than poor fen litter. Additionally, we observed no biogeochemical trends in pH, alkalinity, HCO − 3 concentrations and BS along the vertical depth profile. We expect this to be caused by self-facilitating feedbacks of rich fen vegetation. High anaerobic decomposition rates generate alkalinity  and mineralization will result in a relatively high nutrient availability. This high nutrient availability will stimulate the growth rate and height of the rich fen vegetation, making it a strong competitor (e.g., for light), specifically for the generally short-growing poor fen vegetation. Furthermore, fast decomposition rates result in a low organic matter accumulation rate (Bragazza, Buttler, Siegenthaler, & Mitchell, 2009;Lamers et al., 2000;Scheffer, Van Logtestijn, & Verhoeven, 2001;Verhoeven & Toth, 1995) and in this way, a habitat with close proximity to baserich groundwater is maintained with environmental conditions favoring rich fen vegetation growth (Tyler, 1979).

| Competition for light: mutual exclusion
The spatially segregated occurrence of rich and poor fen vegetation in dense tussocks and hummocks indicates that next to self-facilitating feedbacks, competitive strategies are preventing invasion of the contrasting vegetation. For example, peat mosses form acidic rainwater lenses, which diminishes the influence of base-rich groundwater. The rich fen vegetation does not prefer these abiotic conditions (Tyler, 1979). It has been shown that acidic conditions decrease germination of rich fen vegetation, specifically Schoenus nigricans, though the exact mechanism remains unclear (Boatman, 1962;Clymo & Hayward, 1982). Inversely, abiotic conditions that favor rich fen vegetation growth, such as a high pH, high alkalinity and high HCO − 3 concentrations (Tyler, 1979), negatively affect poor fen vegetation performance (Harpenslager, van den Elzen, et al., 2015;Vicherová, Hájek, & Hájek, 2015;Vicherová, Hájek, Šmilauer, & Hájek, 2017).

| Overcoming mutual exclusion
Within an EC range of ~300-450 mEq/L, patches of poor fen within rich fen vegetation (e.g., peat mosses) were able to survive negative effects from upward mineral-rich water seepage on density, as judged from the intermediate pH, alkalinity and bicarbonate concentration compared to the other vegetation plots (either poor or rich fen, Figure 3). However, no clear succession from rich fen vegetation into a Sphagnum-dominated bog took place in this floating fen (Appendix S1; this study; Van der Maarel & Roozen, 1975;van Groenendael et al., 1975). peatland ecosystem (Grootjans et al., 2016). The observed mosaic of rich and poor fen vegetation could originate from a short period of alleviating environmental stressors: a window of opportunity (sensu Balke, Herman, & Bouma, 2014), as generally poor fen vegetation is unable to establish in areas influenced by upwelling and inundation of HCO − 3rich groundwater (Lamers, Smolders, & Roelofs, 2002;Vicherová et al., 2015Vicherová et al., , 2017. Therefore, we expect that a period of drought resulted in lower groundwater levels, which temporarily alleviated stress enabling establishment of poor fen vegetation inside rich fen vegetation. Simultaneously, reduced groundwater pressure likely lowered alkalinity and bicarbonate concentrations in the porewater. During drought, peat mosses are less affected by HCO − 3 toxicity, which can enable the poor fen vegetation to become locally dominant (Granath et al., 2010).
After having reached a critical density and size, feedbacks created by the peat moss vegetation itself (acidification, rainwater retention and peat accumulation), further reduce the negative impact of HCO − 3 on the poor fen vegetation .
During drought, oxygen can penetrate the soil and potentially decrease the acid-neutralizing capacity and pH as a result of acid production by aerobic microbial redox processes (Stumm & Morgan, 1996). Peat mosses are able to retain rain water during drought via the ability to store water in their hyaline cells in order to regulate capillary pressure that transports water from below and to reduce hydraulic conductivity of the peat layer, preventing lateral and vertical water losses (Clymo & Hayward, 1982;Päivänen, 1973;Rydin & Jeglum, 2006;Schipperges & Rydin, 1998). Rich fen species are much more prone to drought, because they lack these mechanisms (Bakker, van Bodegom, Nelissen, Aerts, & Ernst, 2007;Mettrop et al., 2015), which gives them a competitive disadvantage in periods with reduced groundwater pressure.

| Conceptual model
Here, based on empirical data, we present a conceptual model for a window of opportunity due to a short period of drought for the rise of a stable vegetation mosaic of rich and poor fen vegetation.
When peat mosses are well-established, they are able to overcome HCO − 3 stress by means of density-dependent feedbacks, including acidification and the formation of hummocks that retain poorly buffered rainwater (Granath et al., 2010;Hájková et al., 2012). Similarly, this situation would occur when HCO − 3 -rich groundwater pressure increases again after a period of drought that temporarily alleviated stress for peat mosses so their density could increase. At the same time, when HCO − 3 -rich groundwater pressure increases, rich fen vegetation gains a competitive advantage and the ability to outcompete poor fen vegetation (Granath et al., 2010;Hájková et al., 2012).
Succession from rich to poor fen species likely stagnates and instead of reaching a climax stage of succession, a vegetation mosaic could emerge in the floating fen ( Figure 5). We expect that as conditions remain within certain boundaries (e.g., EC values of ~300-450 mEq/L in the deeper soil layer), coexistence of poor and rich fen vegetation is possible and maintained due to self-facilitating feedbacks and mutual F I G U R E 5 Conceptual model of a floating fen showing the development of a rich fen vegetation-dominated state (1: left situation) towards a stable mosaic of rich and poor fen vegetation (2 and 3: right situation). During drought as a window of opportunity (WoO), the influence of HCO − 3 -rich groundwater is reduced (from dashed to dotted line) and poor fen vegetation colonizes the floating fen (2). When poor fen vegetation density increases, self-facilitating feedbacks cause ombrotrophication and poor fen vegetation is able to overcome HCO − 3 stress when groundwater pressure increases (red arrow). Simultaneously, when groundwater pressure is high again, rich fen vegetation (3) is able to outcompete poor fen vegetation again with self-facilitating feedbacks such as fast decomposition, resulting in close proximity to HCO exclusion. Currently, causal relationships are lacking and further experiments should focus on elucidating this mechanism.

| Conclusion
Our findings strengthen our hypothesis that multiple showed that differences in environmental conditions were related to vegetation mosaics at a microscale (Hájek, Hekera, & Hájková, 2002).
Here, we provide a possible explanation for the mosaic of poor and rich fen species in a floating fen based on empirical data. Interestingly, self-facilitation of species not only helps them to cope with environmental stress, but also engineers their direct environment (upper soil layer) and likely results in a stable coexistence. We expect this principle to play an important role in the resilience of fen ecosystems and therefore further research should elucidate the mechanism underlying vegetation mosaics in similar environmental conditions.

ACK N OWLED G EM ENTS
The authors thank Vera van Berlo for supporting practical work, Emiel Brouwer for help with determining moss species, and Sebastian Krosse and Roy Peters for helping with chemical analyses. We report no conflict of interest.