The plants of temperate salt marshes are exposed to multiple stresses such as frequent flooding associated with salinity, temporary anoxia in the sediment, and possibly sulfide and manganese toxicity together with mechanical stress and sediment coverage (Leuschner and Ellenberg, 2017). Despite these constraints, some saltmarsh plants such as cordgrass (Spartina spp.) and sea purslane (Atriplex portulacoides) are known to be highly productive (Smith et al., 1979; Schubauer and Hopkinson, 1984; Bouchard and Lefeuvre, 2000). Many saltmarsh species develop extensive root systems and it has been found that plant biomass and productivity may be larger below- than above-ground in these environments (Valiela et al., 1976; Smith et al., 1979; Schubauer and Hopkinson, 1984; Groenendijk and Vinklievaart, 1987; Tripathee and Schaefer, 2015). For example, 50–90% of the productivity of Spartina alterniflora was found to be contributed by root and rhizome growth and turnover in a salt marsh in the eastern United States (Valiela et al., 1976; Darby and Turner, 2008). This suggests that a large part of the soil organic carbon contained in saltmarsh sediments is derived from roots, and below-ground productivity is an important factor in the carbon and nutrient cycles of these semi-aquatic ecosystems. Due to their short lifespan and rapid turnover, fine and very fine roots (diameters <2 mm) act as an important sink for carbohydrates supplied by photosynthesis (Jackson et al., 1997; Gill and Jackson, 2000). However, most studies on the saltmarsh below-ground compartment focus on the total (i.e., fine and coarse root, and rhizome) biomass, while only few studies have addressed the structure and dynamics of roots <2 mm in diameter, i.e., the fraction of the root system which likely is responsible for water and nutrient uptake.

Coastal salt marshes are extreme habitats, which require specific adaptations of the plants that colonize them. Species growing in the frequently inundated lower zone of the marsh have to cope with anoxia and reducing conditions in the soil. This environment may trigger the formation of aerenchyma in roots and rhizomes, which facilitate oxygen supply, and foster the development of strategies to exclude and excrete salt (Rozema et al., 1981, 1985). Root aerenchyma can increase the plant's capacity to detoxify potentially harmful ions such as S²⁻, Fe²⁺ or Mn²⁺ in the rhizosphere (Lee, 1999, 2003; Maricle and Lee, 2002). In the upper marsh, stress from inundation is less frequent, but plant growth may additionally be limited by nitrogen shortage (Valiela and Teal, 1974; Kiehl et al., 1997; Levine et al., 1998; van Wijnen and Bakker, 1999). The large below-ground biomass often found in salt marshes is thus not surprising, as it may be needed to secure nutrient and water acquisition, and to anchor the plants in a relatively unstable sediment. Root:shoot mass ratios exceeding unity are also found in other stressful environments such as nutrient-poor or dry grasslands and cold tundra ecosystems (Jackson et al., 1996; Leuschner et al., 2013).

Small-scale heterogeneity is a characteristic feature of many temperate saltmarsh ecosystems. Even minor elevation differences in the salt marsh may cause great spatial differences in inundation frequency, water level height, salinity and the degree of soil anoxia, and thus in the conditions for root growth in the sediment. This is also reflected in the zonation of saltmarsh communities (Bakker, 2014; Leuschner and Ellenberg, 2017), with salinity and tidal inundation as the main factors driving species distribution across the elevation gradient (Cooper, 1982; Armstrong et al., 1985; Rozema et al., 1985; Ungar, 1998).

Root system studies across elevation and water level gradients and in different sediment types should reflect the small-scale vegetation mosaic in salt marshes and may display the associated plant strategies to cope with varying environmental constraints. An example illustrating species differences is the study of Bouma et al. (2002) in a Dutch salt marsh, who found root longevity to be shorter in the highly competitive upper marsh grass Elytrigia atherica than in the more stress-tolerant grass Spartina anglica from the lowermost pioneer zone. Results of Steudel et al. (2011) and Ford et al. (2016) suggest that plant species richness, which can vary between one species per plot at the most stressful sites and more than 10 species at higher elevation, could also influence the root mass of saltmarsh communities, modifying the influence of abiotic factors. Also for other ecosystems it is known that plant diversity increases root biomass (Mommer et al., 2010; Mueller et al., 2013; Eisenhauer et al., 2017). Resource capture in more diverse communities may for example be enhanced by adjusting the depth distribution of roots between species (i.e., “complementarity effect”; Loreau and Hector, 2001; Cardinale et al., 2007; Mommer et al., 2010). Furthermore, the “selection effect” may increase biomass production in more diverse communities, when a very productive species dominates the biomass of the species mixture (Loreau and Hector, 2001).

In this study, we examined the variation in fine root mass, fine root depth distribution and fine root morphology across elevational gradients in two common types of North Sea tidal salt marshes, a barrier island marsh with mostly sandy sediment and a foreland salt marsh with silty-clayey sediment. Due to these differences in geomorphology, sites assumedly differ in nutrient availability and further soil properties, such as bulk density, enabling the investigation of these parameters on fine root traits. The variation in elevation and inundation frequency was addressed by transects reaching from the low-elevation pioneer zone with dominant Spartina stands with daily inundation to the high-elevation upper salt marsh with dominant Elytrigia swards that experience flooding only 4–8 times per month. We focused on fine roots <2 mm in diameter due to their relevance for resource uptake, while larger roots and rhizomes with primarily storage, conduction and anchorage function were not considered. We searched for those abiotic and biotic factors (inundation frequency, salinity, soil texture, soil nutrient and element content, species diversity), which exert the largest influence on fine root mass and morphology in the studied salt marshes. We tested the hypotheses that (i) fine root mass is on average greater in the sandy than the silty sediments due to higher nutrient availability in the latter, (ii) the decrease of rooting depth with increasing soil depth is less pronounced in the pioneer zone due to an assumed greater adaptive potential of the inhabiting species to the anoxic sediment conditions, (iii) the likely more stress-tolerant species of the frequently inundated pioneer zone have more robust roots with lower specific root area (SRA) and specific root length (SRL), but higher root tissue density (RTD) than the more competitive species of the upper marsh, and (iv) plant species richness has a positive effect on the root mass of the community.

Our fine root inventory in three common saltmarsh communities at two geomorphologically different saltmarsh sites clearly showed differences in fine root mass between the three saltmarsh communities across the elevational gradient, but also between the two sites. At both study sites, the greatest fine root mass was recorded in the lower salt marsh, which correlated with highest plant diversity in these communities, where Atriplex portulacoides co-existed with about 10 other herb and grass species. We assume that in this lower saltmarsh community, a considerable diversity of below-ground space occupation strategies exists, as annuals (Salicornia spp.), small herbs (Spergularia media), tall herbs (e.g., Aster tripolium, Triglochin maritima), grasses (Puccinellia maritima, Spartina anglica, Festuca rubra) and dwarf shrubs (Atriplex portulacoides) share the rooted soil volume and likely partition space as explained by the “complementarity effect” (Loreau and Hector, 2001). In contrast, the upper saltmarsh community is nearly exclusively occupied by dense stands of Elytrigia atherica, and the pioneer zone is largely dominated by Spartina anglica with only low cover of additional grasses or herbs, which leaves less room for root space partitioning. This positive correlation between diversity and fine root mass in agreement with our hypothesis (iv) matches results from other root studies in salt marshes and other ecosystems and indicates an existing effect of biodiversity (Hooper et al., 2005; Cardinale et al., 2006). For example, Ford et al. (2016) found plant species richness to be a significant predictor of root biomass in a salt marsh in the UK. Similarly, in the mesocosm experiment of Sullivan et al. (2007) with various saltmarsh species, a significant increase in root biomass and also in the root:shoot ratio was recorded with increasing species diversity. These different reports suggest that plant diversity is a driver of fine root mass in temperate saltmarsh communities, in agreement with the findings in other non-saline grassland habitats (Mommer et al., 2010; Mueller et al., 2013; Eisenhauer et al., 2017; Weisser et al., 2017).

The lower saltmarsh zone plays a crucial role for sediment trapping and stabilization in the foreland bordering dykes and islands along the north-west European Wadden Sea coast. Our root data suggest that the relatively species-rich communities with abundant A. portulacoides are effective sediment stabilizers, since plant diversity was found to increase root density. This matches the observation that plant species richness enhances soil stability and erosion protection in salt marshes and other grasslands (Coops et al., 1996; Chen et al., 2012; Gould et al., 2016). Supporting previous studies (Chen et al., 2012), we suggest that A. portulacoides plays a role as key species in stabilizing sediment by its extensive root system in ungrazed saltmarsh sites.

The multiple regression analysis and the direct comparison of the three community types (zones) and two sediment types (sites) suggest that the principal abiotic factors salinity and inundation frequency only have a relatively small influence on fine root mass. The Na content of the sediment as a proxy of inundation duration showed a positive (though relatively weak) influence on root mass in the total sample (and in Westerhever), and root mass and root density reached a minimum in the upper marsh of both sites, where the exposure to salinity and inundation was lowest. This pattern is best explained by a dominant effect of plant species identity and species richness on community fine root mass, while effects of salinity (and anoxia) are diminished by effective adaptation of the species. It appears that productive species like S. anglica and A. portulacoides with specific adaptations to salinity and anoxia replace less adapted species in the lower marsh and pioneer zone, where they are capable of establishing large root systems. The photosynthesis of the C4 grass Spartina is rarely limited by salinity (Longstreth and Strain, 1977) and the species is capable of oxidizing sulfide in the roots and rhizosphere by means of effective O₂ transport to the sediment through its aerenchyma (Lee, 1999, 2003; Maricle and Lee, 2002). Therefore, S. anglica is a highly productive species (exceeding 6,000 g d.m. m⁻² yr⁻¹; Long and Woolhouse, 1979) despite the mostly anoxic sediment. A. portulacoides is also highly tolerant of salinity (Redondo-Gómez et al., 2007) which is due to the accumulation of quaternary ammonium compounds for osmoregulation (Stewart and Lee, 1974; Rozema et al., 1985). The effective adaptation of the pioneer zone species (and also of the lower marsh zone taxa) to the adverse conditions in this environment is also reflected in a higher proportion of root mass (31–35%) in the subsoil (20–40 cm depth) of the pioneer zone compared to the upper salt marsh (17–19%). This is in accordance with our hypothesis (ii) in which we postulated a deeper root system in the more frequently inundated pioneer zone with well-adapted species. Interestingly, there was a great difference in fine root density between sites at 10 cm soil depth (Figure 4). It may be speculated that the very high fine root density in this soil depth found at Spiekeroog is related to the soil texture and associated lower plant-available P concentrations in the soil which may lead to a species-specific increase in fine root mass to compensate for limited P in the lower salt marsh. An earlier study by Strieckmann (1989, unpubl. data; Supplementary Material 4) also found no root density decrease down to 24 cm depth in other north German S. anglica marshes. In contrast, the fine root density decrease from the topsoil to 20–40 cm was significant in the upper salt marsh of both sites, indicating that E. atherica is more sensitive to waterlogging (Armstrong et al., 1985; Schröder et al., 2002; Veeneklaas et al., 2013) and thus develops a shallower distribution of fine roots.

Our data also suggest that higher nutrient availability tends to reduce fine root mass, as it was postulated in our first hypothesis. Fine root mass was higher by roughly 25%, and root area index even by about 50% in the sandy Spiekeroog plots compared to the silt- and clay-rich Westerhever plots. This may indicate that the plants in the sandy sediment need to produce larger root systems, with especially large surface area, to compensate for the generally lower concentrations of plant-available nutrients found in the sandy sediments of the Spiekeroog salt marsh. Sediment stability might also be responsible for differences in fine root mass: a higher root mass in the sandy Spiekeroog sediments could be required for plant anchorage, as fine-grained clay soils resist erosion better than sandy soils (van Eerdt, 1985; Allen, 1989). Since bulk density was higher at the sandy Spiekeroog site providing less air and water space, and less space for root proliferation, especially in the deeper soil horizon, it may be assumed that the need of fine roots for nutrient capture and plant stability exceeds the disadvantages for root growth in soil with higher bulk density, furthermore indicating the good adaptation of roots to the abiotic conditions.

Comparison of our fine root mass data to other studies on below-ground biomass in salt marshes is limited due to several reasons. First, it has to be kept in mind that we measured total fine root mass and not fine root biomass, even though our figures should be close to biomass according to our live:dead ratio assessment under the microscope in a number of subsamples. While a few authors explicitly sampled only fine root biomass, various studies do not state, whether live and dead roots were separated, and the term “biomass” is sometimes used for total root mass (biomass + necromass), which makes comparison to other studies difficult. Second, other authors used different criteria for defining root mass or biomass, sometimes including coarse roots and rhizomes as well, or retrieving root mass with sieves of greater mesh size than we did. This will increase or decrease root biomass figures, thus leading to deviating results. Groenendijk and Vinklievaart (1987) investigated total below-ground biomass in 0–60 cm in a Dutch salt marsh and obtained much higher profile totals than we did, which may only be partly explained by the lower profile depth investigated in our study (0–40 cm). Nevertheless, the same root mass distribution patterns along the saltmarsh elevational gradient became visible in the Dutch study: Greatest below-ground biomass totals were recorded in the A. portulacoides-dominated lower marsh (mean: 13,338 g m⁻²), while the lowest biomass occurred in the upper marsh dominated by E. atherica (mean: 7,763 g m⁻²). Comparison of our data from the S. anglica stands at our study sites (site means of 1,500–2,000 g m⁻²) with data from Spartina alterniflora-dominated salt marshes on the east coast of the United States indicates a similar root biomass average, but larger variation among sites (600–11,000 g m⁻²; Smith et al., 1979; Windham et al., 2003; Tripathee and Schaefer, 2015). Unpublished root biomass data from a grazed and an ungrazed salt marsh in Schleswig-Holstein (Germany) of Strieckmann (1989, unpublished) range between 1,000 and 5,000 g m⁻² for the profile to 24 cm (Table 6, Supplementary Material 4). In contrast to our study, the greatest below-ground biomass in the ungrazed site was recorded in the plots dominated by S. anglica and A. tripolium, while the minimum (~1,500 g m⁻²) occurred in the A. portulacoides stands. Similar to our plots in Westerhever, the upper saltmarsh community with E. atherica dominance was the only stand with a root biomass:aboveground biomass ratio <1 (Table 6).

The fact that we considered only fine roots (<2 mm in diameter) and not coarse roots and rhizomes, may explain differences in root biomass totals among different studies in Spartina marshes. For example, Darby and Turner (2008) found a mean root biomass of 753 g m⁻², but a mean rhizome biomass of 1,952 g m⁻² of S. alterniflora in a Louisiana salt marsh, matching findings of Schubauer and Hopkinson (1984). Data of total belowground biomass (including larger root diameters) are important for carbon cycle studies, but less informative when the below-ground absorptive surface of plants and communities is assessed. Several studies indicate that the standing fine root biomass (or mass) in salt marshes varies considerably with season (Groenendijk and Vinklievaart, 1987; Steinke et al., 1996; Darby and Turner, 2008), which may also explain differences between studies.

In our study, we did not distinguish between dead and live fine roots, as we found the proportion of dead fine roots to be low (<10%) in all inspected samples. Similar live:dead ratios were reported by Groenendijk and Vinklievaart (1987) in a Dutch salt marsh (<15% non-living roots), while the proportion of below-ground necromass was greater in certain North American salt marshes, exceeding root biomass (Valiela et al., 1976; Schubauer and Hopkinson, 1984; Darby and Turner, 2008). Apart from differences in the root diameter considered and likely variance in root mortality rates, this may be a consequence of slow root decomposition rates as was found in a Dutch salt marsh (Buth, 1987). Direct observation of fine root dynamics with rhizoscopes may be needed to unravel the causes of different root live:dead ratios in anoxic sediments.

A comparison of the fine root mass of salt marshes with that of other grassland or herbaceous communities indicates that the multi-stress conditions in this saline environment demand for high carbohydrate investment in below-ground organs by the plants. According to root mass data compiled by Leuschner and Ellenberg (2017), non-saline mesic to moist temperate grasslands have root masses in the topsoil (mostly 0–15 cm) of 500–2,000 g m⁻² matching the root mass figures of our upper salt marsh, while being smaller than the root masses found in the lower marsh and the pioneer zone. Jackson et al. (1997) give a global fine root mass mean of 1,510 g m⁻² for temperate grasslands, which is also lower than the root masses found in our lower salt marsh and in other Spartina stands in the northern hemisphere. Similarly, our root surface area totals (RAI values >300 m² m⁻²) were several times larger than the mean RAI recorded for temperate grasslands by Jackson et al. (1997) (79.1 m² m⁻²).

One might expect that the substantial variation in inundation frequency, sediment anoxia and salinity found in the different saltmarsh zones of Spiekeroog and Westerhever should lead to pronounced differences in fine root morphology in the different communities (hypothesis iii). Bouma et al. (2002) hypothesized that the saltmarsh plants of the pioneer zone have slower-growing, more stress-tolerant roots, while the plants of the upper salt marsh should be faster growing, which might result in a higher RTD of the former. We found elevated tissue densities in both the upper marsh and the pioneer zone of the Westerhever site, but no difference across the Spiekeroog gradient and thus no consistent RTD pattern in our study. Similarly, community differences in SRL and SRA were not consistent across the two elevational gradients; mean fine root diameter was remarkably constant in our samples. We conclude from these inconsistent patterns that fine root morphology is largely under the control of plant species and their specific adaptations to the adverse conditions (i.e., the formation of aerenchyma), while a more general pattern of root morphology did not emerge in the studied saltmarsh communities. This contradicts hypothesis (iii) but matches the findings of Bouma et al. (2001, 2002) of an only weak responsiveness of the root architecture of three halophytic grass species to nitrogen supply, inundation and oxygen content of the sediment and no relationship between root longevity and tissue density. Root N concentration peaked at both sites in the lower salt marsh, possibly reflecting species-specific N uptake patterns.

Our study in two salt marshes with contrasting geomorphology shows that fine root mass is relatively high and fine root surface area large in comparison to grassland ecosystems in terrestrial habitats. The plants of the frequently inundated lower marsh and pioneer zone seem to be well adapted to this stressful environment, allowing some specialist species such as Spartina to establish a root system with deep penetration of the sediment for good anchorage and nutrient supply. These adaptations of characteristic species appear to control the fine root mass of the saltmarsh communities, overriding effects of environmental stress. As in other habitats with extreme environmental conditions, a higher number of plant species tends to increase root biomass and seems to promote soil exploration through root space partitioning by functionally different species. Studies on fine root dynamics and root function in terms of water and nutrient uptake are needed for a mechanistic understanding of the below-ground compartment of saltmarsh communities and its sensitivity to environmental change.

CL, DH, and RR conceived and designed the research project. TD and RR performed field and laboratory work. DH, TD, and RR analyzed the data. CL and RR wrote the manuscript. All authors approved the final version of the manuscript.