Towards a trait-based ecology of wetland vegetation
Summary
- Functional traits mechanistically capture plant responses to environmental gradients as well as plant effects on ecosystem functioning. Yet most trait-based theory stems from terrestrial systems and extension to other habitats can provide new insights.
- Wetlands differ from terrestrial systems in conditions (e.g. soil water saturation, anoxia, pH extremes), plant adaptations (e.g. aerenchyma, clonality, ubiquity of bryophytes) and important processes (e.g. denitrification, peat accumulation, methane emission). Wetland plant adaptations and trait (co-)variation can be situated along major plant trait trade-off axes (e.g. the resource economics spectrum), but soil saturation represents a complex stress gradient beyond a simple extension of commonly studied water availability gradients.
- Traits that affect ecosystem functioning overlap with patterns in terrestrial systems. But wetland-specific traits that mediate plant effects on soil redox conditions, microbial communities and on water flow, as well as trait spectra of mosses, vary among wetland types.
- Synthesis. With increasing availability of quantitative plant traits a trait-based ecology of wetlands is emerging, with the potential to advance process-based understanding and prediction. We provide an interactive cause-and-effect framework that may guide research efforts to disentangle the multiple interacting processes involved in scaling from environmental conditions to ecosystem functioning via plant communities.
Introduction
Trait-based framework
Trait-based ecology is a way of addressing and integrating questions ranging from evolutionary theory to ecosystem science by recasting them in terms of phenotypic traits. The field has grown exponentially in the last decade, owing to this capacity for synthesis and the promise of generalization and prediction (Shipley et al. 2016). Traits represent a common currency to study the mechanistic basis of species occurrence as a function of the environment (via response traits, Lavorel & Garnier 2002) as well as species effects on ecosystem functioning and services (via effect traits, Díaz et al. 2007; Lavorel 2013). The majority of this theory has been developed in terrestrial habitats, especially grasslands (De Bello et al. 2010). In wetlands, environment-trait and trait-process linkages are still comparatively less well-studied. There is a need to establish how far trait-process linkages can be generalized between ecosystems.
Wetlands, drivers, types
Wetlands are intermediate habitats that have commonalities with and differ from both terrestrial and aquatic habitats. They are distinguished by the presence of water, either at the surface or within the root zone, which leads to poorly aerated soils and predominance of biota that can tolerate wet and reducing conditions. Major primary producers in wetlands are bryophytes and vascular plants with mostly terrestrial growth forms rooted in soil or peat, as opposed to algae in fully aquatic habitats (Van der Valk 2006). Wetlands can thus be defined as habitats ‘that are inundated or saturated by water at a frequency and for a duration sufficient to support a prevalence of vegetation typically adapted for life in saturated soil conditions’ (Mitsch & Gosselink 2015). Wetlands are characterized by properties (e.g. anoxia, highly diverse redox conditions, highly organic soils, sometimes extreme pH) and processes (e.g. denitrification, carbon sequestration in peat, methanogenesis) that are uncommon in terrestrial habitats, which makes them an ideal natural laboratory to examine and extend established trait-based theory. From the wide variety of wetland types, we here focus on three broad classes of inland freshwater wetlands with sparse or absent tree cover: bog, fen and marsh. Bogs and fens are both peatlands, whereas marshes are peat forming to a lesser and variable degree (Rydin & Jeglum 2013). The dominant vegetation differs between wetland types, depending on hydrological and chemical conditions. In peatlands, habitat-trait interactions are particularly strong: not only must present species have traits to endure the particular conditions (soil saturation, low pH values and low nutrient availability), but these conditions are also strongly influenced by the traits of dominant species.
Wetland ecosystem services
Wetlands are notable for the provision of three regulating service complexes that are important at regional to global scales (Maltby & Acreman 2011; Turner, Georgiou & Fisher 2011): (i) Water flow regulation including water storage and flood attenuation; (ii) climate regulation in the form of energy exchange, carbon sequestration and greenhouse gas (GHG) emissions; and (iii) water quality regulation by biogeochemical transformations, retention or removal of excess nutrients from runoff. We use the term ‘service complex’ to emphasize the fact that what is commonly termed a service (e.g. water flow regulation) generally consists of multiple aspects and potential benefits that may trade-off with each other or even be mutually exclusive. When attempting to unravel the ecology underlying service delivery, it is essential to be clear on which aspect is of interest. Wetland type, landscape configuration and hydrological connectivity (Bullock & Acreman 2003; Maltby 2009; Maltby & Acreman 2011) set the stage for which ecosystem services can be delivered and how they are distributed across landscapes.
Ecosystem services (ES) are increasingly focused on in policy and management (Martinez-Harms et al. 2015). This situation demands a clear understanding of the biophysical controls of ES potential to inform management decisions (Lavorel 2013; Bennett et al. 2015). What constitutes an ES depends on both the ecological potential as well as human demand (Bennett et al. 2015). We here focus on ecological ES potential, i.e. the capacity of ecosystems to generate and maintain potentially beneficial processes irrespective of actual human demand or use. Plant functional traits have emerged as a promising link from species to ecosystem processes that underpin ES generation (Díaz et al. 2007; Lavorel 2013; Lamarque et al. 2014).
Our starting point is the assumption that the drivers determining wetland type cause a distinct functional signature in constituent plant species, and that this local plant community affects most of the ecosystem processes involved in service generation. The aim is to establish a qualitative framework to guide and facilitate work on scaling from environmental conditions to ecosystem functioning and ES potential for wetlands. A future goal would be to move towards quantification, but at this stage this is only possible for a few links. We review environmental drivers that determine the wetland type (bog, fen, marsh). We ask which traits are typical of wetland plants and how commonly measured traits behave in wetland environments. We then identify the properties and processes that underlie the three service complexes of climate regulation, water flow and water quality regulation, and review the knowledge base for trait effects on these, highlighting knowledge gaps and research needs. The main output is an online interactive summary of the literature in which causes and effects are traced from large-scale environmental drivers that define local conditions, which in turn select for trait values with final effects on ecosystem processes and services.
From drivers to plant traits
Drivers of wetland type and key environmental gradients
The fundamental drivers of wetland type are the source of water, which influences water chemical composition, and the associated hydroperiod (depth, frequency and duration of flooding). The three wetland types – bogs, fens and marshes – can thus conveniently be delineated based on the predominant source of incoming water (Fig. 1a).
Water source and hydroperiod mediate the more proximate environmental drivers nutrient availability, pH, the average water level and temporal variation thereof. The relative importance of proximate drivers varies between wetland types. In bogs, nutrient availability is low, whereas in fens and marshes receiving surface runoff it can vary from very low to high. pH is low in bogs (also due to biotic feedback effects) and can be high in fens that depend on mineral-rich surface or groundwater. Water-table depth fluctuates the most in intermittently or seasonally flooded marshes. Often found along the edges of rivers or lakes, they can become completely inundated and mechanical disturbance from water flow is an additional factor. The common feature of all wetlands is the occurrence of saturated soils that develop frequent or prolonged anoxic conditions. This puts wetlands among the few ecosystems that provide a reduced environment for microbial interactions, enabling anaerobic processes such as denitrification, iron reduction, sulphate reduction and methanogenesis as well as slowing down decomposition rates (Mitsch & Gosselink 2015).
Wetlands are heterogeneous also at the local scale, reflecting regional differences. They often feature a varied microtopography that arises from the interplay of abiotic and biotic processes (such as peat formation). Microenvironments are characterized by differences in water-table depth, oxygen levels and pH, which lead to variation in different processes also at the local scale (Fig. 1b).
Wetland plant adaptations and trait syndromes
Patterns of trait variation within habitats and along broader environmental gradients at least partly reflect adaptation (Reich et al. 2003). Figure 2 summarizes traits that mitigate for conditions that can be adverse in wetlands: drought, flooding, nutrient scarcity and extreme pH. Some traits vary in a correlated fashion, suggesting underlying evolutionary or physiological trade-offs that constrain the trait space available to plants (e.g. the leaf economics spectrum, Wright et al. 2004) and reduce all possible variation to a limited number of axes of specialization (Westoby et al. 2002). The plant economics spectrum (PES) aims to explain trait (co-)variation and distribution along environmental gradients based on evidence of covariation between root, stem and leaf structural and biochemical traits that respond to nutrients, carbon acquisition (light) and water in a correlated fashion (Wright et al. 2004; Freschet et al. 2010; Reich 2014). The proposition is that traits align along a continuous axis of fast to slow resource turnover rates: if one trait is ‘fast’ (e.g. high photosynthetic capacity) then other traits must be fast also (e.g. nutrient uptake rates) (Reich 2014). The pivotal drivers in the PES (nutrients, light and water availability) are naturally shared by wetland plants, but wetlands can be situated at extreme ends of these gradients (e.g. excessive water availability in flooded conditions, very low nutrient availability in bogs and oligotrophic fens) and are structured by additional factors not included in PES relationships (pH, mechanical disturbance).
Trait responses to soil saturation
High water-tables in wetlands result in soil oxygen depletion and reducing conditions, changing nutrient availability and leading to an accumulation of decay-inhibiting phenolics (Freeman, Ostle & Kang 2001). The major plant adaptation to anoxia is the formation of aerenchyma, i.e. intercellular spaces that allow for transport of oxygen through the plant down to the roots, allowing for root respiration and oxygenation and thus detoxification of the rhizosphere (Colmer & Voesenek 2009; Pezeshki & DeLaune 2012) (Fig. 1b). Internal pressure gradients due to humidity or temperature differences may enhance this effect (Sorrell & Hawes 2010), as do adventitious roots, hypertrophied lenticels, and pneumatophores in trees (Catford & Jansson 2014). To avoid excessive oxygen loss to the soil and enhance downward oxygen diffusion, many wetland plants form a barrier to radial oxygen loss (ROL) in their roots (Colmer & Voesenek 2009). Non-wetland plants do not exhibit ROL barriers and have much less aerenchyma than wetland plants.
High water availability could cause comparatively high transpiration rates and low water use efficiency (WUE) in wetland plants (Behrendt et al. 2000) but this possibility has to our knowledge not been extensively studied. Paradoxically, soil saturation can cause drought stress and favour corresponding adaptations and responses. This is because anoxic conditions in saturated soils reduce root conductance, thereby obstructing water uptake; if evaporative demand on the shoot meanwhile stays constant, the plant experiences drought stress (Poorter et al. 2009). Some wetland plants exhibit (plastic) water-conserving mechanisms, such as temporary stomatal closure (Pezeshki & DeLaune 2012) or lower stomatal pore area (Savage & Cavender-Bares 2012). Wetland plants thus show a range of strategies to optimize or control WUE (Fig. 2).
PES traits have been linked to coarse climatic gradients of water availability, i.e. the balance between potential evapotranspiration and precipitation (Wright et al. 2004), but the effects of (long-term) waterlogging and resulting oxygen stress on PES traits are not well understood. High rates of water uptake and transpiration would correspond to the fast end of the spectrum (Reich 2014), so that one would expect marsh species to exhibit an acquisitive trait syndrome. With regard to water relations, the PES would predict concurrent high hydraulic conductance, lower wood density, high leaf area per shoot (sapwood area), high photosynthetic capacity as well as lower leaf thickness and leaf dry matter content (LDMC) (Reich et al. 2003; Reich 2014). As detailed below, wood density can decrease with wetness, which indicates a trend towards the acquisitive end of the spectrum. Other traits, however, tend towards the opposite, more conservative habit [notably specific leaf area (SLA), leaf nitrogen content (LNC), and LDMC].
Several studies report that high wetness increases root porosity (Ordoñez et al. 2010; Douma et al. 2012; Martina & von Ende 2012; Savage & Cavender-Bares 2012; Baastrup-Spohr et al. 2015), for example through the formation of aerenchyma. Obligate wetland species have recently been shown to have higher mean root aerenchymatous volume than facultative wetland species (Tanentzap & Lee 2016). Aerenchyma may be non-adaptive or even maladaptive under dry conditions owing to a trade-off between root porosity and root mechanical strength (Striker et al. 2007). Stem specific density (SSD) in woody plants decreases with wetness (Douma et al. 2012; Martina & von Ende 2012), which is predicted to correlate with higher stem hydraulic conductivity (Reich 2014). Below-ground allocation tends to decrease with permanent waterlogging (Webb, Wallis & Stewardson 2012), in contrast to its increase with mechanical disturbance from water flow (Gurnell 2014).
Leaf nitrogen content decreases with increasing wetness, a pattern interpreted as due to decreasing N availability with prolonged soil saturation (Ordoñez et al. 2010; Douma et al. 2012). Relationships of SLA with increasing wetness have been reported as positive (Ordoñez et al. 2010), non-significant (Douma et al. 2012; Savage & Cavender-Bares 2012), or negative (Poorter et al. 2009; Violle et al. 2011; Wright & Sutton-Grier 2012). Baastrup-Spohr et al. (2015) found a hump-shaped response of SLA along a broader wetness gradient. Reduced SLA at high wetness may be the result of an energetic trade-off with other structural or physiological traits, or an adaptation to high water levels in itself. In flooding-sensitive plants, high wetness reduces SLA because of drought stress resulting from reduced root water uptake in anaerobic soils (Poorter et al. 2009). In tolerant plants, aerenchyma could alleviate this effect to some degree and allow for leaves with larger SLA.
Zelnik & Čarni (2008) instead investigated the fraction of Grime's plant strategies along a wetness gradient (elevation above water level). These Competition-Stress-Ruderal strategies are in turn correlated with traits (Grime et al. 1997). The strongest responses were found for increased prevalence of CS strategies (competitive and stress-tolerant) and decreased prevalence of CR strategies (competitive and ruderal) with higher wetness. Dominating leaf traits for these strategies (Pierce et al. 2013) would suggest that higher wetness favours higher LDMC while reducing SLA.
Conditions of frequently prolonged soil saturation might thus favour a conservative habit. With regard to the PES, flooding and soil saturation are not an extension of the usually examined water availability gradient based on precipitation, but represent a stress gradient that prohibits dominance of the most acquisitive trait syndromes. Specific adaptations to high wetness may incur additional costs and constraints.
Trait responses to water-table fluctuations and flooding
Irregular flooding can cause temporary submergence of plants and mechanical disturbance. Responses to submergence include physiological down-regulation, leaf dimorphism, aerenchyma formation or rapid shoot elongation (Colmer & Voesenek 2009; Webb, Wallis & Stewardson 2012; Catford & Jansson 2014). Two strategies have been described: tolerance and escape (termed Low Oxygen Quiescence and Escape Syndrome, respectively; reviewed in Colmer & Voesenek 2009). Tolerance is characterized by economical energy usage, increased anaerobic metabolism and up-regulated biochemical protection from harmful cellular changes. Shoots cease to elongate during submergence. This strategy is advantageous during short term submergence. Escape is characterized by re-orientation of growth direction and increase in shoot growth, preservation or development of aerenchyma and structures to facilitate underwater gas exchange such as ‘aquatic’ roots. Development of new shoot organs that resemble aquatic leaves with higher SLA, thinner cell walls and cuticles, and chloroplasts close to the epidermis have also been observed (Mommer & Visser 2005). This strategy is advantageous during long-term submergence (Colmer & Voesenek 2009). Amphibious plants show a greater increase in SLA upon submergence than other species, pointing to the adaptive value of plastic responses (Poorter et al. 2009).
A high degree of phenotypic plasticity as adaptation to highly variable environments (water-table fluctuations) could hypothetically be the rule in wetland plants and cause comparatively large intraspecific trait variability in such species, both physiologically and morphologically (Albert et al. 2011). Significant plasticity of root aerenchyma volume can be adaptive under increased flooding by reducing plant biomass loss (Tanentzap & Lee 2016). The degree of plasticity of different traits hence could be indicative of their importance as response traits, although plasticity of a trait does not always necessarily imply that it is adaptive (Van Kleunen & Fischer 2005).
Adaptations to mechanical disturbance from flooding include relatively high below-ground biomass allocation, extensive rhizomes, high stem flexibility and narrow leaves (Colmer & Voesenek 2009; Sosnová, van Diggelen & Klimešová 2010; Catford & Jansson 2014; Gurnell 2014). In a global study, the three components of leaf mechanical resistance (work to shear, force to punch and force-to-tear) all were correlated with the inverse of SLA (1/SLA = LMA, Leaf Mass per Area), bridging leaf mechanical properties with resource uptake traits (Onoda et al. 2011). Sclerophylly (stiff, tough, leathery leaves; Read et al. 2006), traditionally interpreted as an adaptation to low water availability, may thus also arise in response to mechanical stress. But such long-lived leaves are also a syndrome of nutrient conservation in bogs (Rydin & Jeglum 2013).
Clonal reproduction is widespread in wetlands (Sosnová, van Diggelen & Klimešová 2010) and may serve several functions, such as protection and anchorage in highly disturbed habitats but also resource acquisition and storage in nutrient-poor environments (Cornelissen et al. 2014) as well as increased regeneration capacity after disturbance (Gurnell 2014).
Where wetland plants fall out of the spectrum
The PES posits for any species the convergence of traits across all plant organs onto a particular position along the slow-fast spectrum (Reich 2014). Yet there appear to be deviations from these patterns within and between organs, possibly mediated by differing plasticity between traits and in response to different resources. For example, leaf trait covariation does not always follow PES predictions: in an experiment with wetland plants, leaf nitrogen showed no relationship with SLA at given resource levels nor was it strongly coupled to photosynthesis, leading to questions about the generality of PES predictions (Wright & Sutton-Grier 2012).
Across species, reported decreases in wood or SSD with wetness (Ordoñez et al. 2010; Douma et al. 2012; Savage & Cavender-Bares 2012) are consistent with the PES, but the predicted resulting increase in relative growth rate (RGR) due to higher stem hydraulic conductivity does not seem to occur (Douma et al. 2012). Similarly, while across a range of both terrestrial and wetland species, both root depth and relative growth rate decline with increasing wetness (Douma et al. 2012), root elongation rate among willow species decreases with increasing wetness but with no concomitant change in whole plant RGR (Savage & Cavender-Bares 2012). These results could point to either compensatory changes in other traits or to effects of confounding environmental factors.
Within species, differential plasticity of traits (genotypic or phenotypic) may further modulate predicted patterns of trait covariation. In an experiment with reed canary grass (Phalaris arundinacea), below-ground biomass decreased with wetness under low soil nitrogen, but increased with wetness under high soil nitrogen; additionally, both below- and above-ground biomass were greater under high versus low light availability (Martina & von Ende 2012). This three-way interaction between wetness, nitrogen and light indicates significant modulation of PES-predicted trait covariation by abiotic conditions. Such alterations do not necessarily undermine the validity of an underlying economics spectrum, but they do point to a range of variation within PES relations that should be accounted for (Albert et al. 2011).
Intraspecific variability adds a third dimension to trait covariation
The degree of phenotypic plasticity can be viewed as a trait under genetic control and selection and might be quantifiable (Matesanz, Gianoli & Valladares 2010). Plasticity seems to vary depending on the driver, the species and the traits. First, with regard to drivers, plasticity of most traits was greatest in response to light, followed by fertility and wetness in reed canary grass (Martina & von Ende 2012). These three drivers are further expected to interact: N availability can decrease with wetness under long-term saturation and oxygen depletion, which reduce N mineralization rates and increase N losses through denitrification (Ordoñez et al. 2010), and light competition should increase with fertility and higher above-ground biomass. Second, species differ in their plasticity, due to either genotypic variability or phenotypic plasticity. Even if responses of traits to wetness and fertility tend to be in the same direction, they can differ in magnitude among species (Wright & Sutton-Grier 2012). The relative position of species along trait axes, however, might be conserved and community traits remain predictable. Third, traits themselves may differ in their plasticity. In terrestrial plants, intraspecific variability is known to be higher for leaf nitrogen and phosphorus concentration than for SLA, and smallest for LDMC and maximum height (Albert et al. 2011; Kazakou et al. 2014). While plastic responses of leaf chemistry to fertility could be expected, light availability should theoretically have the potential to alter also SLA and/or height. High spatial and temporal variability in wetland conditions could hypothetically cause relatively high intraspecific variability in wetland plants. Unravelling the patterns behind trait plasticity is a major research task required to complement the general predictions laid out in the PES.
A plant economics spectrum for bryophytes?
Wetlands are also important habitats for bryophytes, and especially peat mosses (Sphagnum spp.) dominate in bogs and oligotrophic fens. The chemical composition of mosses creates recalcitrant litter. Mosses have high water holding capacity contributing to waterlogging and anoxia, and tend to have high cation exchange capacity (CEC), which generates acidity through proton release from cation exchange sites (Cornelissen et al. 2007; Turetsky et al. 2012).
Encouragingly, trade-off axes similar to the PES of terrestrial plants are now being uncovered also for bryophytes, which will facilitate the inclusion of this important group into trait-based models (Bengtsson, Granath & Rydin 2016). In Sphagna for example, an analogous economics spectrum is discernible, from relatively fast turnover species, characterized by high apex (capitulum) biomass, less dense canopies, higher maximum photosynthetic rates and litter mass loss, to slow turnover species with opposite traits (Rice, Aclander & Hanson 2008; Laing et al. 2014). Differences among Sphagnum subgenera indicate phylogenetic conservation of the traits, but there are additional differences in species typical for different wetland habitats (Bengtsson, Granath & Rydin 2016).
Because bryophytes differ from vascular plants in structure and function, the question arises which traits, or indeed which organs, are comparable. For example, maximum photosynthetic rate (Amax) may be a good correlate of RGR in both vascular plants and bryophytes, but while SLA is a good proxy trait for Amax in the former it is not meaningful in (Cornelissen et al. 2007). Bryophyte photosynthesis additionally strongly depends on shoot hydration, influenced by whole canopy architecture in mosses growing above the water-table. Some authors therefore use the bryophyte canopy as analogue to vascular plant leaves. Canopy mass per area (CMA) indeed covaries negatively with nitrogen concentration and photosynthetic rate, as does LMA in vascular plants (Waite & Sack 2011).
Plant trait effects on ecosystem service potential
The following sections lay out a framework that links wetland plant effect traits to ecosystem properties and processes involved in the generation of the three wetland services of water flow regulation, water quality regulation, and climate regulation. The potential for wetlands to generate these services depends on the presence and balance of multiple ecosystem processes, which in turn are influenced by multiple traits. An online interactive figure (see Fig. S1, Supporting Information) summarizes these complex links, whereas Fig. 3 provides a static version. The links depicted in Fig. 3 all have some support in the literature, but ambiguous links are highlighted and the level of uncertainty is presented (see Table S1 and Fig. S1) as an illustration of research needs. One example is how local pH relates to tissue pH, and how that in turn affects acidification and decomposition. Other uncertainties include the role of clonality, and the relationships between nutrient availability and structural descriptors of vegetation. Some of these traits clearly are required to cope with particular environmental conditions (response traits), others are included mainly for their role as effect traits with potential relevance to ecosystem functioning. Effect traits may still vary along environmental gradients due to covariation or trade-offs with response traits. In our model, we focus on qualitative relationships, where most of the links are fairly well supported by the literature. However, there is still a lack of quantitative data on the strength and relative effect size of many of the links in Fig. 3, which are required to accurately model and predict ES levels. Yet we hope that our model will inspire researchers to fill those gaps, and to this end Fig. S1 and Table S1 will be openly available for additions with quantitative information.
Trait effects on water flow regulation
The water balance of any ecosystem is determined by input (precipitation, surface and ground water recharge) and output terms (evapotranspiration, surface and ground water discharge) (Rydin & Jeglum 2013). Plants directly affect the outflow terms via the reduction in surface flow velocity, the partitioning of precipitation via canopy interception, evaporation and the promotion of infiltration, and via water uptake, storage and transpiration (Brauman et al. 2007; Nepf 2012; Gurnell 2014; Waddington et al. 2015). Indirect effects on these processes are via vegetation-mediated effects on soil properties such as soil stability, water holding capacity, temperature, porosity and hydraulic conductivity (Eviner & Chapin 2003; Ehrenfeld, Ravit & Elgersma 2005; Rydin & Jeglum 2013).
Surface flow regulation is most influenced by size-related and structural traits. Flow resistance in riparian systems depends on canopy height, density (depending on above-ground biomass, individuals per area) and structural heterogeneity (growth form diversity) (Gurnell 2014). Due to competition for light, increased canopy height might correlate with higher density and leaf area index (LAI) (Falster et al. 2011), characteristics that would synergistically increase flow resistance. Woody plants show generally higher surface flow resistance than herbaceous vegetation (Dosskey et al. 2010), and clonal plants in marshes can decrease flow velocity due to particularly high shoot density (Bouma et al. 2013) as well as strong anchorage and resistance to uprooting (Gurnell 2014). Anchorage and soil stabilization also depend on root architecture, depth and root tensile strength (Bardgett, Mommer & De Vries 2014).
Vegetation roughness coefficients used in hydrodynamic modelling comprise density, elasticity, shape and bendiness of the vegetation (Nepf 2012; Bouma et al. 2013; Camporeale et al. 2013). Establishing mechanistic relationships between more widely available trait data and such coefficients would greatly benefit hydrological modelling studies (Camporeale et al. 2013). For example, leaf length of herbaceous wetland plants follows biomechanical expectations by exhibiting also greater leaf thickness and LDMC (Vernescu & Ryser 2009). Leaf mass per area, i.e. the inverse of SLA (a routinely measured trait), correlates linearly with mechanical resistance in many species across ecosystems (Onoda et al. 2011). Leaf shape also affects flow resistance: dissected or ramified leaves reduce flow velocity more than streamlined single-bladed leaves (Nepf 2012). Plants that alter surface microtopography by forming tussocks or hummocks also reduce near-surface flow velocity (Bouma et al. 2013), as does a large amount of litter cover, particularly coarse woody debris (Corenblit et al. 2014).
In addition to topographic factors, wetland water storage capacity is increased by the prevalence of mosses and highly organic soils (peat) with high pore volumes. Live moss canopies effectively function as sponges due to their intra- and extracellular water holding capacities. Morphological characteristics (leaf size and shape, the volume of specialized water holding hyaline cells, branch architecture) partly predict intra- and extracellular water holding capacity of individual peat moss (Sphagnum) shoots (Rice, Aclander & Hanson 2008; Rydin & Jeglum 2013). For whole canopy water retention, we expect positive effects of extracellular capillary and water holding capacity, shoot density and biomass per area, and negative effects of moss canopy height (Elumeeva et al. 2011; Michel et al. 2012). In raised parts of peatlands (hummocks) the Sphagnum community is dominated by species with smaller apex (i.e. capitulum) biomass and higher shoot density, which should exhibit higher canopy water retention capacity (Laing et al. 2014).
For soil water retention, the large capillary forces of many small pores are more effective than few large pores (Eviner & Chapin 2003). Large root biomass (root length density and depth) and clonal organs (spacers and rhizomes) increase soil porosity and soil organic matter content (Bardgett, Mommer & De Vries 2014; Cornelissen et al. 2014), enhancing soil pore volume and water holding capacity. Moss litter enhances permeability of soils due to slow decomposition rates and enhanced organic matter content (Turetsky et al. 2010), leading to a positive feedback between moss dominance and soil moisture.
Plant water uptake and evapotranspiration return water to the atmosphere, decrease soil saturation levels and regenerate space for infiltration. Dense canopies of species with large leaf area generally intercept precipitation and increase evaporation (Eviner & Chapin 2003). Water loss to the atmosphere via plant uptake and transpiration can be as high as a 50% of inflowing waters in high biomass stands (Howard-Williams 1985). Comparatively higher transpiration rates and lower WUE could hypothetically be the rule for wetland plants, which only occasionally have to cope with water scarcity. Indeed, some wetland plants have lower WUE than terrestrial species (Behrendt et al. 2000), but it appears unknown whether this is a general pattern.
Trait effects on water quality regulation
Plants affect wetland nutrient retention directly via nutrient uptake and storage as well as indirectly by mediation of physicochemical controls (sedimentation, adsorption) and microbial processes (decomposition, denitrification) that both affect and are affected by soil redox state (Saunders & Kalff 2001; Fisher & Acreman 2004; Dosskey et al. 2010).
Plant primary production binds nutrients and carbon in biomass. Sequestration in accreting peat or the release through decomposition depend on both plant or litter characteristics and on soil properties (temperature, soil moisture, acidity and oxygen levels). Productivity, growth and decomposability can be linked to generalized plant resource uptake strategies subjected to the fundamental trade-off between rapid resource acquisition and conservation of resources (Grime et al. 1997; Reich 2014). Plants at the acquisitive end of the PES facilitate rapid nutrient and carbon uptake during the growing season, but live shorter and show faster litter decomposition and remineralization (Wardle et al. 2004; Cornwell et al. 2008; De Deyn, Cornelissen & Bardgett 2008; Freschet, Aerts & Cornelissen 2012). Some forms of clonality may aid in longer term nutrient storage, e.g. via resorption of nutrients from senescing shoots to rhizomes or tubers (Cornelissen et al. 2014). Plants at the conservative end of the spectrum, such as evergreens common in bogs and nutrient-poor fens, grow more slowly but show lower litter decomposability and hence contribute to long-term sequestration of both nutrients and carbon (De Deyn, Cornelissen & Bardgett 2008; Freschet, Aerts & Cornelissen 2012).
Mycorrhizal association aids in resource acquisition with implications for nutrient and carbon cycling (Cornelissen et al. 2001). Different mycorrhizal types correspond to positions along the economics spectrum from slow to fast turnover in the order ericoid mycorrhiza (ERM), ectomycorrhiza (ECM) and arbuscular mycorrhiza (AM) (Cornelissen et al. 2001; Bardgett, Mommer & De Vries 2014). Mycorrhizal association may be less common in plants with alternative resource acquisition strategies (e.g. clonal growth), and in wet or acidic environments that are unfavourable to mycorrhizal fungi (Hempel et al. 2013). In hummocks and lawns of nutrient-poor bogs and fens, however, ericoid mycorrhizal associations with dwarf shrubs are likely to be important for nutrient retention and recycling.
Where oxygen is depleted, denitrification is the primary contributing process to nitrogen removal, returning gaseous dinitrogen (N2) to the atmosphere (Verhoeven et al. 2006). If incomplete, as e.g. under high levels of nitrate loading, low pH or incomplete anoxia, denitrification also results in the production of the potent GHG nitrous oxide (N2O) (Verhoeven et al. 2006; Liu & Greaver 2009). Plants mediate denitrification by the quantity and quality of soil carbon inputs that provide the energy source for denitrifying bacteria (McGill, Sutton-Grier & Wright 2010). C input quantity can be estimated by above-ground and below-ground plant biomass, and C input quality is inversely related to C : N ratios of shoots and roots (McGill, Sutton-Grier & Wright 2010; Sutton-Grier et al. 2011; Sutton-Grier, Wright & Richardson 2013) and hence the resource economics spectrum. Root exudates also contribute high quality C to microbial metabolism and thereby fuel denitrification (Zhai et al. 2013; Bardgett, Mommer & De Vries 2014). Aerenchymatic tissue and high root porosity cause soil oxygenation via ROL, which can inhibit denitrification (Sutton-Grier, Wright & Richardson 2013). The relevance of denitrification to water quality regulation depends on N loading rates, wetland type and landscape position. Denitrification is important in marshes and small-scale wetlands embedded in agricultural landscapes (Blackwell & Pilgrim 2011), whereas bogs and most fens are severely nitrogen-limited and exhibit net nitrogen retention (e.g. Wiedermann et al. 2007).
Trait effects on climate regulation
Plants mediate climate regulation directly by the uptake and long-term storage of atmospheric CO2-C in standing and dead biomass and indirectly via effects on decomposition and microbial processes that regulate GHGs (i.e. methanogenesis, methanotrophy and denitrification) (Fig. 3). Traits linked directly to growth and decomposition are captured by the plant economics spectrum (e.g. SLA, leaf nutrient concentration); decomposition additionally depends heavily on tissue carbon chemistry. Important indirect negative effects on decomposition in peatlands are via species traits that lead to maintenance of soil saturation and soil water acidification, enabling the accumulation of vast deposits of peat (Limpens et al. 2008; Turetsky et al. 2012; Rydin & Jeglum 2013).
The enzymatic latch hypothesis (Freeman, Ostle & Kang 2001) proposes the activity of a single enzyme, phenol oxidase, to be the primary driver of low decomposition rates in peatlands. Specifically, the typical peatland conditions of low oxygen availability (Freeman, Ostle & Kang 2001), low pH (Williams, Shingara & Yavbitt 2000), and/or low temperature (Pinsonneault, Moore & Roulet v) prevent phenol oxidase from degrading phenolic compounds, the accumulation of which then inhibits other major degradative enzymes (hydrolases). The main control on phenol oxidase activity may vary between peatlands, and could further involve direct effects of vegetation, for example via the provisioning of substrate to phenol oxidase (Zhang et al. 2010).
Peat accumulation is highest in moss-dominated wetlands. Mosses, and especially peat mosses (Sphagnum spp.), show the lowest decomposition rates of all functional groups, often due to their constitutive chemistry, particularly structural carbohydrates and lignin-like compounds, uronic acids and phenolic compounds (Cornelissen et al. 2007; Turetsky et al. 2008; Lang et al. 2009). Tissue pH has been proposed as an integrative trait capturing secondary chemistry in both bryophytes and vascular plants (Cornelissen et al. 2011). While it may correlate with lower decomposability (Freschet, Aerts & Cornelissen 2012), its role for reducing decomposability in addition to the major factor of anoxia is unclear. As in vascular plants, high photosynthetic capacity in Sphagnum species also correlates with higher litter decomposability (Laing et al. 2014). Sphagnum peat accumulation is thus determined by habitat-trait interactions, with faster growth in wet open habitats vs. higher water holding capacity and lower decomposability due to production of secondary metabolites in hummock species (Turetsky et al. 2008; Laing et al. 2014).
Vegetation affects methane emission from wetlands via the same pathways as denitrification (provision of carbon substrate to microbes and effects on soil redox conditions) and additionally by acting as conduit through which gas can bypass layers with active methanotrophs and escape to the atmosphere (Laanbroek 2010). High primary productivity, labile litter and high rates of root exudation enhance carbon supply to microbial metabolism and methanogenesis (Whiting & Chanton 1993; Nilsson et al. 2001; Koelbener et al. 2010). Aerenchymatous plant tissues have two conflicting effects on methane emission: soil oxygenation can inhibit methanogenesis or promote oxidation of methane, but aerenchyma is also a main export pathway for methane to the atmosphere. However, the degree of rhizospheric CH4 oxidation that prevents methane from escaping has been found to be the primary control of CH4 emission (Ström, Mastepanov & Christensen 2005), and may be correlated with root biomass (Bouchard et al. 2007). Experimental tests of effects of carbon input traits and soil oxygenation traits on methane emission showed that increased soil oxygenation by productive plants with high root porosity could offset the potential stimulation of methanogenesis from additional carbon inputs (Sutton-Grier & Megonigal 2011).
Trait trade-offs can scale up to ecosystem service trade-offs
Fundamental trade-offs at the trait level may scale up to trade-offs or synergies among ecosystem properties and processes and ultimately ES (Lavorel & Grigulis 2012). Wetlands dominated by species at the fast end of the resource economics spectrum are expected to have high nutrient retention capacity due to high uptake rates and production of labile litter, causing high decomposition rates, which may deplete oxygen and fuel higher denitrification rates (Fig. S1). At the same time, faster decomposition and mineralization imply a lower potential for long-term carbon sequestration. Here, the same trait values (fast growth, high SLA and LNC) have contrasting effects on two different ES. Nutrient retention and C sequestration capacity in wetlands are thus expected to trade-off at the ecosystem level as a consequence of a species-level trade-off. In this case, the PES appears to be useful to explain trait covariation and its effect on ecosystem level properties.
The relative magnitude of the processes that comprise an ES also matters (Butterfield et al. 2016). Carbon sequestration depends on the balance between primary production and decomposition. Net ecosystem exchange of carbon, often the main determinant of carbon accumulation in peat, seems primarily controlled by total plant productivity (Lund et al. 2010). Yet in many wetlands, plant trade-offs that lead to low litter decomposability in slow-growing species (Turetsky et al. 2008; Freschet, Aerts & Cornelissen 2012) lead to lower productivity but nevertheless also greater net carbon sequestration.
Feedbacks and thresholds
An important aspect not addressed in the online figure and Fig. 3 are potential feedbacks from traits and processes to drivers. Not least feedbacks via ecohydrological processes will propagate through the ecosystem (Waddington et al. 2015). For example, waterlogging and resulting anoxia enhance both nutrient removal and carbon sequestration by enabling denitrification and slowing down decomposition. However, the key wetland plant adaptation to these conditions, i.e. the formation of aerenchyma, causes local soil oxygenation, which counteracts anoxia and might alter these relations. Soil oxygenation decreases denitrification rates by redirecting the terminal oxidation of electron carriers from denitrification to aerobic decomposition and thereby negatively affects wetland nutrient removal capacity in fast turnover systems. In slow turnover systems, increased soil oxygenation may enhance heterotrophic decomposition rates and thus decrease carbon sequestration capacity.
There are also feedbacks among processes that may positively affect ES potential. For example, Sphagna are able to invade nutrient-rich fens by positive feedbacks between lower decomposition rates, peat accumulation and increasing distance to groundwater, which causes acidification and can trigger a rapid ecosystem shift from fen to bog that increases carbon sequestration capacity massively (Granath, Strengbom & Rydin 2010; Soudzilovskaia et al. 2010).
Concluding remarks: towards a trait-based ecology for wetlands
Since ES have become a focus of management efforts (Keeler et al. 2012; Bennett et al. 2015; Martinez-Harms et al. 2015), a complete picture of the underlying ecology is required to provide meaningful recommendations. Trait-based approaches can capture the links from environmental drivers to plant communities and onwards to the ecosystem processes that underlie ES delivery (McGill et al. 2006; Díaz et al. 2007; Suding et al. 2008; De Bello et al. 2010; Lavorel & Grigulis 2012).
We have indicated that the role of some species traits, as well as some of their links to ecosystem functions, are uncertain or ambiguous. Our data and model will be openly available for the scientific community to improve by modifying the information about the uncertainty and direction of responses and effects (Moor et al. 2017). In essence this means that our graph (Figs 3 and S1) illustrates our current understanding and is offered as a tool for including new data in a dynamic fashion. It can be seen as a step towards a model in which also quantitative relationships will be included. Quantifying the strengths of the links in the conceptualization provided here, will ultimately allow prediction of effects of vegetation changes due to management or global change on ecosystem functioning and service delivery (Kominoski et al. 2013; Moor, Hylander & Norberg 2015).
Trait-based approaches to plant effects on wetland functioning still lag behind compared to the knowledge base for other habitats, but as we show here, there have been advances during recent years (e.g. McGill, Sutton-Grier & Wright 2010; Sutton-Grier et al. 2011; Sutton-Grier, Wright & Richardson 2013). Bryophyte functional ecology is well under way (Cornelissen et al. 2007; Waite & Sack 2011; Laing et al. 2014; Bengtsson, Granath & Rydin 2016), but parallels to vascular plant traits should be further developed. How far postulated trait axes such as the plant economics spectrum (Reich 2014) can be generalized to wetlands is a research challenge (Wright & Sutton-Grier 2012). Wetland-specific drivers (especially waterlogging) and adaptations might alter predicted relationships but offer opportunities for extension and refinement of theory.
Traits have the potential to become the fundamental unit of ecology, a common currency that can unify mechanistic knowledge from diverse ecosystems. At the same time a more detailed focus on ecosystem-specific traits is needed to understand the functional variation within a particular ecosystem type, such as wetlands. This might alter predicted relationships and offer opportunities for extension and refinement of theory.
We will leave the reader with a few outstanding questions, which might serve as a guide to future research:
- How does trait covariation predicted by the plant economics spectrum differ between wetland and terrestrial environments?
- Is the degree of plasticity in response to nutrient availability, hydrology, and temperature higher in wetland than in terrestrial plants?
- How can we better express traits that are comparable between mosses and vascular plants?
- What is the relative effect size of plant impacts vs. abiotic effects on ecosystem processes and ES potential in wetlands?
Authors’ contributions
H.M. and J.N. initiated the paper; H.M. conducted the initial review; H.M., H.R. and J.N. led the writing of the manuscript; J.N. conceived and coded the interactive figure; K.H., M.B.N. and R.L. contributed substantially with their expertise; all authors gave final approval for publication.
Acknowledgements
This research was performed under the Strategic Research Program EkoKlim, Stockholm University. The reflections of Andrew Tanentzap, Peter Vesk and two anonymous referees are greatly appreciated.
Data accessibility
The supporting information Table S1 (containing data underlying Figs 3 and S1) and scripts to generate and modify the interactive Figure S1 are available on Github, at https://github.com/zahachtah/WetlandTraitDriver (Moor et al. 2017).
