Predicting climate change effects on wetland ecosystem services using species distribution modeling and plant functional traits

Our goal was to estimate potential biomass density from the abundance data of the VMI inventory. The presence points were split into three datasets—one for each of the three abundance classes (single individuals, frequent, and dominant), and Maxent was run separately on occurrences of each abundance class. For the modeled occurrence probabilities of the three abundance classes, a weighted average was calculated as approximation of a prediction of relative abundance (weights: single = 0.01, frequent = 0.1, dominant = 1). While simple occurrence data could already provide reasonable estimates of abundance (Van Couwenberghe et al. 2013), this method allowed us to model an occurrence value that is more closely related to estimated field abundance than the usual output of Maxent. Yet we treat the results as a basis for discussion rather than a solid prediction.

The mass ratio hypothesis poses that the community mean trait value weighted for relative biomass (CWMT) is the strongest driver of ecosystem processes (Grime 1998). The relationship between the estimate of abundance from the above SDM approach and the potential relative biomass of each species in the community can be inferred by means of general allometric scaling relationships between species height, individual biomass, and density. Specifically, height is proportional to biomass M ^0.264 (Niklas and Enquist 2001), and maximum abundance N scales with M ^−0.757 (Belgrano et al. 2002). Putting these relationships together (Electronic Supplementary Material), we estimated maximum potential biomass (kg dry weight m⁻²) from height (m) as B _max = 1.769e05*(H/b)^(−0.757/a) (where a = 0.264, b = 2.58; for trees: a = 0.345, b = 3.71).

Based on this potential relative biomass, we arrived at first guiding estimates of potential changes in relevant CWMT values in bogs, fens, and riparian wetland communities across the NDB. Changes in CWMT values were calculated separately for the field and shrub layers within each wetland type, for each 500-m cell across the NDB, and averaged within the five identified regions of community change.

Note that relative biomass in each cell represents proportions of species according to modeled environmental suitability; translation to local communities (habitat filter and competitive effects) is beyond the scope of this study. The relative contribution of each species to CWMT changes was assessed via the standardized mean squared error resulting from excluding that species from calculations (Electronic Supplementary Material, Table S5). Since trees form a very different functional group than field or shrub layer species and due to their low species number, we treat them separately. As we had neither trait data nor size estimates of bryophytes, effects of distributional changes of bryophytes (in our case, only mosses) are discussed based on estimated relative abundance and compositional shifts only.

SLA increases in all wetland types, most strongly in the riparian field layer, and in the field layer of bogs (Fig. 2a, absolute CWMT values in Electronic Supplementary Material, Table S6), suggesting a trend to higher net primary production (NPP, due to increased relative growth rates) and faster nutrient turnover (higher nutrient uptake and decomposition rates). In the fen field layer, differences are minor, whereas among shrub layers, SLA is predicted to increase most in fens, with only small changes in bogs and riparian wetlands (Fig. 2b).

In the field layer, CH change is minor in bogs and riparian wetlands, but pronounced in fens, particularly so in regions 1, 2 (Fig. 2c), suggesting larger standing biomass, higher potential canopy interception and transpiration rates, as well as increased flow resistance. CH increases in the shrub layer of bogs and fens, whereas riparian shrub CH shows regionally varied responses (Fig. 2d).

The RD decreases in field layers of bogs and riparian wetlands, whereas in fens, RD is predicted to increase in the field layer (Fig. 2e), consistent with the pattern of change in rhizomatous species (Fig. 3d, e), suggesting larger belowground biomass in fens. Also in the shrub layer, RD increases in fens (Fig. 2f). In shrub layers of bogs and riparian wetlands, RD overall tends to decrease, but shows a spatially more variable pattern (Fig. 2f), again due to above-mentioned regional differences in relative changes of R. tomentosum, S. cinerea ,and S. caprea.

Shifts to higher proportions of herbaceous plants in bogs and riparian wetlands (Fig. 3a, c), and to more summer green species in the fen field layer (Fig. 3h), mirror increases in SLA. The decrease of rhizomatous species in bogs (Fig. 3d) might decrease nutrient storage in belowground biomass but could slow peat decomposition via reduced soil oxygenation, while increases in tussock-forming species would benefit water flow resistance; in fens, the opposite pattern emerges (Fig. 3e). Increases in persistent green species in bogs and the fen shrub layer (Fig. 3g, h) suggest more recalcitrant litter with lower decomposition rates, slightly counteracting SLA increases. The shift from nonmycorrhizal to more facultative mycorrhizal species in all field layers (Fig. 4a–d) indicates higher potential for nutrient uptake and increased plant productivity, but this might be counteracted by decreases in obligatory mycorrhizal species among shrubs in fens and riparian wetlands (Fig. 4d, f).

Among hydrophytes, more lifeforms with rigid structure become dominant at the expense of mainly elastic lifeforms (Fig. 5a), except in region 1. In terms of space occupancy, a clear shift toward larger species is observed (Fig. 5b), positively affecting particle retention. Both trends indicate higher potential water flow resistance.

Among bog species, the wet hollow species Sphagnum majus and S. cuspidatum and the hummock species S. rubellum are predicted to decrease, mostly in the western NDB (Electronic Supplementary Material, Fig. S3). This might decrease the productivity of bogs since hollow species are suggested to be more productive than hummock species (Gunnarson 2005). On the other hand, only small changes are predicted for other dominant species in the NDB (e.g., the hummock species S. fuscum and S. capillifolium). In fens, Calliergonella cuspidata and Sphagnum fallax are predicted to increase noticeably throughout NDB; both species are also favoured by increased nutrient inputs (not modeled), thus potential increases in nitrogen deposition might further emphasize this change (Limpens et al. 2011). Scorpidium revolvens (rich fens) and Sphagnum pulchrum (poor fens) decline, both fen species typical of wetter parts. Also Sphagnum imbricatum declines.

Maximum potential biomass is predicted to increase for most tree species (Pinus sylvestris, Betula pubescens, Fraxinus excelsior, and Alnus glutinosa). Only Alnus incana and Picea abies are predicted to decrease. Due to long generation times and uncertainties in estimating realized biomass for trees, we here restrict the discussion to one particular species that might serve as an example: A. glutinosa increases strongly across the NDB, overcompensating the northward shift of Alnus incana. It features key effect traits (nitrogen fixation symbionts, deep soil oxygenation, high inundation tolerance, and flow resistance) with implications for nutrient retention and flood attenuation.

Water flow regulation is an important aspect of wetland functioning and benefits humans not only by attenuating the devastating effects of floods, but also by filtering particles and aiding sedimentation (contributing to nutrient retention) as well as maintaining soil water saturation (reducing decomposition rates and aiding in C sequestration) (Zedler 2003). The largest effect of vegetation on water flow regulation in riparian wetlands is via physical resistance, slowing down of surface waters, and temporary storage. Among hydrophytes, the shift toward larger and more rigid species indicates an increased potential for flow resistance and the attenuation of floodwater peaks within streams (Nepf 2012). In riparian wetlands, a relative increase of tussock-forming graminoids, along with the ubiquitous increase of A. glutinosa points to an increased potential of vegetation to slow down overflowing river waters and decrease flood damage (Claessens et al. 2010; Bouma et al. 2013). Increased CH in fens and bogs could theoretically cause longer water-retention times, higher canopy interception, infiltration and transpiration rates, and thus a generally stronger role in the hydrological cycle. Given these relationships, the flood attenuation service would benefit from the predicted changes. In fens and bogs, however, the species-specific waterholding capacity of bryophytes, for which data are insufficient, is important for flow regulation (Rydin and Jeglum 2013). A shift from Sphagnum dominated to Calliergonella dominated fens might decrease the waterholding capacity and hence increase downstream flood risk.

Major plant effects on the retention and removal of nutrients are via mediation of plant uptake and storage, denitrification (direct loss of N₂ to the atmosphere), and facilitation of sedimentation (Saunders and Kalff 2001). Mycorrhizal symbiosis generally increases nutrient uptake rates and can increase plant productivity and standing biomass (Hempel et al. 2013). The overall shift toward higher proportions of mycorrhizal species in all field layers might thus lead to higher standing biomass and potentially nutrient storage, even if most species are facultatively mycorrhizal. Mycorrhizal infection in facultative species is known to decrease with higher soil water saturation (Muthukumar et al. 2004), such that general decreases in precipitation could be positive for symbioses. Among shrubs, however, obligatory mycorrhizal species decrease, so the overall effect of proportional shifts in mycorrhizal types is inconclusive. The observed trend for increases in SLA indicates an increase of fast-growing species, higher community NPP and larger above-ground biomass and litter amount, i.e., overall increased nutrient uptake. On the other hand, the usually higher leaf nutrient content of high-SLA species implies more easily decomposable litter, faster decomposition rates, and reduced storage times (Cornwell et al. 2008).

Denitrification increases under input of more and higher quality litter, i.e., more and better usable carbon sources for denitrifying microbes (Sutton-Grier et al. 2013). Denitrification itself depends on anoxic conditions, but its rates are the highest at the interface between anoxic and oxygenated layers, where denitrification is coupled to nitrification most efficiently (Mitsch and Gosselink 2007). Plants that are able to oxygenate the rhizosphere around their roots should thus be overall beneficial to denitrification. Assuming that RD is proportional to the volume of soil a plant can oxygenate, higher observed RDs and proportion of rhizomatous species in fens (but not bogs or riparian wetlands) would increase denitrification rates. Better indicators for plant soil oxygenation potential would be root porosity or aerenchymatous tissue in plant species. Most sedges, rushes, and particular functional keystone species like Phragmites are characterized by large internal gas-spaces, which allow for diffusion of gases through the plant and oxygenation of the rhizosphere. The observed decrease of relative biomass of some sedges might reduce this effect, but other large species known to contribute substantially to soil oxygenation increase (Phragmites australis, Typha latifolia) (Jackson and Armstrong 1999). Denitrification rates as well as nutrient-uptake rates (but not long-term storage) might thus overall increase, and therewith the removal of nitrogen from downstream waters.

Vascular plants influence C storage via NPP (potentially increasing with larger SLA and CH), the net ecosystem carbon exchange rate (likely to increase with higher NPP and decomposition rates) and peat properties (Limpens et al. 2008). Increased decomposition due to higher-quality litter (as well as higher temperatures) is likely to have a stronger negative effect on C storage than on nutrient retention. Increased proportion of rhizomatous growth forms and RD in fens could be positive for C storage if coincident with greater belowground biomass (Cronk and Fennessy 2001), or negative via higher soil/peat oxygenation that could fuel decomposition. The formation and composition of peat, the major long-term C reservoir in wetlands, is dominated by bryophytes, especially Sphagna (Schaepman-Strub et al. 2008) due to their effects on the physicochemical environment (maintenance of high water table, low oxygen availability to decomposers, low temperature, and acidification) and low litter decomposability (De Deyn et al. 2008). Predicted decreases in precipitation, potentially leading to lower water tables (and higher rates of heterotrophic respiration), along with concurrent effects on mosses are here therefore of greater concern than shifts in traits of vascular plants. Out of 28 moss species, only three were predicted to increase substantially in abundance. The increase of the two nutrient-tolerant fen species C. cuspidata and Sphagnum fallax, and of the relatively less densely growing S. squarrosum, could mean a shift to less recalcitrant peat. Long-term C sequestration might thus decrease, even if short-term CO₂ uptake and C storage increase.

Wetlands are also a significant source of the potent greenhouse gas methane (Mitra et al. 2005). Methanogenesis is dependent on anaerobic conditions, labile litter providing accessible C substrate for methanogens, and the export pathway (Mitsch and Gosselink 2007). Decreasing precipitation and warming should generally decrease water table levels, resulting in a deeper aerobic layer, which would reduce methanogenesis but increase CO₂ release from heterotrophic respiration. More easily decomposable litter, as produced by plants with higher SLA and faster growth rates, could fuel methanogenesis in the same way as denitrification but will have little effect if it gets decomposed before reaching the anaerobic layer. The most important effect of vascular plants on methane emissions is by providing an export pathway through aerenchymatous plant tissue, whereby methane bypasses the aerobic soil layer where it could be oxidized. Sedge cover has been found to be a strong predictor of methane efflux via plants, but other aerenchymatous plants (e.g., Juncus spp.) and plants with open gas flow-through systems (Phragmites, Typha) have the same effect (Vasander and Kettunen 2006). As the majority of our study species are able to develop aerenchyma, and as relative increases in some plants that can transport methane are balanced by decreases in others, this effect may not change much. Lower water tables would result in overall lower methane production.

Estimation of the overall effect of vegetation changes on C storage and greenhouse gas emissions is difficult, depends on the timescale considered, and might be more strongly influenced by abiotic factors (nitrogen deposition, climate; Limpens et al. 2008) and land use change (especially drainage). In the case of the scenario of this study, both CO₂ uptake and release are likely to increase in the short term, whereas methane emission most likely will decrease.