Does National Wetland Inventory class consistently identify vegetation and edaphic differences in Oregon tidal wetlands?

Abstract

Accurately mapping, modeling, and managing the diversity of wetlands present in estuaries often relies on habitat classification systems that consistently identify differences in biotic structure or other ecosystem characteristics between classes. We used field data from four Oregon estuaries to test for differences in vegetation structure and edaphic characteristics among three tidal emergent marsh classes derived from National Wetlands Inventory (NWI) data: low marsh, high marsh, and palustrine tidal marsh. Independently of NWI class, we also evaluated the number and types of plant assemblages present and how edaphic variables, non-native plant cover, and plant species richness varied among them. Pore water salinity varied most strongly across marsh classes, with sediment carbon and nitrogen content, grain size and marsh surface elevation showing smaller differences. Cover of common vascular plant species differed between marsh classes and overall vegetation composition was somewhat distinct among marsh types. High marsh had the largest species pools. However, plot-level plant diversity was similar among marsh classes. Non-native species cover was highest in palustrine and high marshes. The marshes in the study contained a large number of plant assemblages with most occurring across more than one marsh class. The more common assemblages occurred along a continuum of tidal elevation, soil salinity, and edaphic characteristics, with varying plant richness and non-native cover. Our data suggest that NWI classes are useful for differentiating several general features of Oregon tidal marsh structure, but that more detailed information on plant assemblages found within those wetland classes would allow more precise characterization of additional wetland features such as edaphic conditions and plant diversity.

Introduction

Estuarine wetlands support a number of important coastal processes including primary productivity and carbon sequestration (Chmura et al. 2003; Irving et al. 2011), shoreline stabilization (Shepard et al. 2011), and nutrient uptake (Jordan et al. 2011; Findlay and Fischer 2013). Coastal wetlands are also key habitats for fish, providing areas for refuge, spawning, rearing, and foraging (Cattrijsse and Hampel 2006; Davis et al. 2012). However, sustained provision of these ecosystem functions and services is threatened by habitat loss and degradation and may be further compromised by climate-related changes such as sea-level rise (Brittain and Craft 2012; Zedler and Kercher 2005).

Due to the importance and vulnerability of tidal wetlands to climate and anthropogenic stressors, there is a significant need to map and assess their status worldwide (Finlayson et al. 1999). At regional, national, or international scales, use of a consistent classification framework for mapping or evaluating wetland resources allows for informed evaluation of spatial and temporal trends in wetland area, composition, or status. For instance, by using a consistent framework, wetlands from disparate regions can be studied collectively to assess trends in habitat extent or condition across large spatial scales (e.g., Dahl 2011; EPA 2016). Moreover, effective classification frameworks may allow scientists or natural resource managers to infer ecosystem structure or function in specific geographic areas where more comprehensive ecological assessments are lacking.

One long-standing classification scheme for wetlands in the United States is the National Wetlands Inventory (NWI) of the US Fish and Wildlife Service (Cowardin et al. 1979). A hierarchical system, the NWI classification framework utilizes aerial photography (Matthews et al. 2016) to divide freshwater, brackish, and marine wetlands into specific classes for habitat mapping. In tidally-influenced estuaries, wetland classes include intertidal unconsolidated sand and mud (mudflats), regularly and irregularly flooded emergent marsh, scrub-shrub wetlands, tidal swamps, and aquatic beds. Since its development, the NWI framework has expanded to include additional descriptors that may allow users to infer wetland functionality and ecosystem services. For instance, Tiner (2003) used local and regional expert knowledge to develop correlations between NWI classes and wetland function. NWI has been used for a wide array of purposes, including development of environmental indicators to evaluate the condition of specific sites or watersheds (Tiner 2004), assessment of change in coastal wetland extent (Carle 2011), and provision of spatial vegetation inputs for sea-level rise modeling (Geselbracht et al. 2015).

More recently, the Coastal and Marine Ecological Classification Standard (CMECS) led by the National Oceanic and Atmospheric Administration (NOAA) was designed to classify all coastal habitat in the United States for conservation prioritization. CMECS is also a hierarchical classification system that integrates data on geologic, physiographic, oceanographic, and biological attributes of coastal areas (Keefer et al. 2008).

While classification systems such as NWI and CMECS are important tools for conducting ecological analyses at large spatial scales along the coast, their utility for assessing ecosystem structure, function, or services is contingent on the degree to which they accurately and consistently identify wetland areas (Matthews et al. 2016) and describe differences in biotic assemblages or environmental conditions among wetland classes. Their ability to translate ecological information across spatial scales may be compromised if wetland classes from one region bear little structural or functional resemblance to the same classes from other regions. They may also have insufficient information content if biological or abiotic characteristics of individual classes overlap considerably with other classes.

In this study we tested the degree to which three tidal emergent marsh classes derived from the NWI classification system corresponded with differences in edaphic characteristics, tidal elevation, and vegetation composition along the Oregon coast. In a previous paper (Janousek and Folger 2014) we focused on analyzing linkages between environmental variables (including abiotic and landscape factors) and plant species presence, assemblage composition, and diversity in these wetlands. In the present study, we use the same dataset to test the consistency of the NWI-based classification, examining the degree to which wetland surface elevation, sediment characteristics, plant cover, assemblage type, non-native plant cover, and plant richness differed among major tidal marsh classes located in the Pacific Northwest. Additionally, we evaluated the number and type of plant assemblages present in Oregon tidal marshes irrespective of NWI wetland class to examine their distribution across different marsh classes and to test for differences in sediment properties and vegetation characteristics among the major assemblage types we observed.

Materials and methods

Study design.

We sampled vegetation, elevation, and sediment properties in the Yaquina, Netarts, Alsea, and Coquille estuaries during the summer of 2010 as described in Janousek and Folger (2014). These estuaries represent a range of estuarine types present in the Pacific Northwest, from a marine-dominated lagoon with little freshwater input (Netarts Bay) to a drowned river mouth (Yaquina) to the more river-dominated systems of Alsea and Coquille (Lee and Brown 2009). In each of the estuaries, we used a 2009 GIS layer of NWI wetland classes to randomly obtain a sample of replicate plots in three vegetated wetland classes typical of tidal wetlands in Oregon estuaries: (i) regularly‐flooded estuarine marsh (NWI code E2EMN; hereafter denoted “low estuarine”), (ii) irregularly‐flooded estuarine marsh (E2EMP; hereafter referred to as “high estuarine”), and (iii) seasonally flooded-fresh tidal palustrine and temporarily flooded-fresh tidal palustrine marsh (PEMR and PEMS; hereafter called “tidal palustrine”). Low and high estuarine marsh was common in all four estuaries. Tidal palustrine marsh area was relatively limited in the estuaries, so we pooled the two NWI classes into a single tidal palustrine marsh class for analyses (Table 1).

Table 1.

Landscape attributes of the Oregon estuaries sampled in this study. The normalized freshwater flow index quantifies annual river flow per estuary area with larger values indicating more river‐dominated estuaries (Lee and Brown 2009). Numbers in parentheses following tidal marsh habitat areas are the number of plots sampled in this study for each wetland class.

			Tidal marsh habitat area (ha)
Estuary	Estuary description	Normalized freshwater flow (m3 m−2 yr−1)	Low estuarine (E2EMN)	High estuarine (E2EMP)	Tidal palustrine (PEMR+PEMS)
Netarts	Bar-built estuary, northern OR	11	62.9 (12)	38.4 (8)	2.2 (2)
Yaquina	Drowned river mouth, central OR	63	108.0 (30)	127.0 (22)	22.9 (24)
Alsea	Drowned river mouth, central OR	211	138.3 (13)	84.1 (4)	24.4 (7)
Coquille	Drowned river mouth, southern OR	695	54.5 (19)	61.7 (10)	4.7 (0)
Total plots sampled			74	44	33

In the field we located GIS-selected plots with a resource‐grade Trimble GPS unit (horizontal accuracies <1 m). At most of these locations, we sampled one or two additional plots at random distances along transects perpendicular to the elevation gradient. In the field we determined all plot locations by GPS and later determined their NWI class by overlaying plots on a 2011 NWI GIS layer. The analyses used in this study do not include any plots that occurred in woody tidal wetland (e.g., tidal scrub-shrub) or any emergent marsh plots that were misclassified by NWI as non-emergent wetland (e.g., as mudflat). NWI polygons containing our sampled plots varied in number and size within each estuary. We usually sampled at least several polygons per marsh class per estuary. Plot numbers within each individual NWI polygon ranged from 1 to 12 (mean = 3.4). We sampled polygons from near the estuary mouth to fresher tidally-influenced areas upriver.

Field sampling.

We sampled vegetation and sediments from June to September 2010. At each random location we determined percent cover of common vascular plants in a 1 m² plot, averaging usually two independent observations per plot. Species with < 1% cover were ignored by each individual observer. We considered only the upper most layer of vegetation in the plot, so total cover in each plot summed to 100%. We also determined vascular plant species presence or absence in a 0.25 m² plot nested within the larger plot to evaluate species richness; in this instance, species in all canopy layers were examined to obtain a total species count.

Within each 1.0 m² plot, we collected two surface soil samples (5 cm depth; 2.5 cm diameter; subsamples pooled and homogenized) during the summer to determine sediment characteristics including grain size, organic matter content, and pore water salinity. We re-sampled the same plots between November 2010 and March 2011 to obtain winter soil salinities. Within the 0.25 m² plot, we assessed light transmission through the vascular plant canopy in four replicate measurements (2 spots x 2 replicates) using a spherical LiCOR PAR sensor by measuring incident light immediately above the sediment surface and at about waist height (above the emergent marsh plant canopy). At the edge of the 1 m² plot, we obtained the geodetic elevation of the soil surface with a survey-grade Trimble 4700 GPS receiver. Instrument vertical precision (standard deviation) was 0.012 m, as determined with repeated measurements at South Beach tidal benchmark S743.

Sample processing.

We determined summer and winter soil pore water salinities with a hand-held Reichert refractometer (Cambridge Instruments, Buffalo, NY) by pressing sediment into Whatman paper filters fitted inside plastic syringes. We determined the fraction of sand (≥63 μm) versus silt + clay (<63 μm) with a Horiba LA-950 A2 particle size analyzer after initial digestion of samples with 30% hydrogen peroxide to remove organic matter (Lee and Brown 2009, Janousek and Folger 2012). From the same cores we also determined total organic carbon (TOC) and total organic nitrogen (TON) with a Carlo Erba EA 1108 elemental analyzer following removal of large roots from the samples (Janousek and Folger 2014). We post-processed raw GPS data with OPUS on-line software (https://www.ngs.noaa.gov/OPUS/) to obtain UTM 10N, NAD83(CORS96) horizontal and NAVD88 (with Geoid09) vertical positions. In the Alsea, Coquille and Yaquina estuaries we converted NAVD88 geodetic elevations to local mean-higher high water (MHHW) by applying estuary-specific correction factors after measuring NAVD88 elevations at NOAA tidal benchmarks. In Netarts Bay, we determined the relationship between MHHW and NAVD88 with a combination of empirical water-level measurements and GPS measurements made with a Trimble R8 GPS receiver (Janousek and Folger 2014).

Statistical analyses.

We conducted analyses in R (v.2.13 to 3.1; R Development Core Team). All statistical comparisons among emergent marsh classes focused on differences among the three marsh classes defined for this study: low marsh (NWI class E2EMN, n = 74), high marsh (class E2EMP, n = 44), palustrine marsh (classes PEMR and PEMS, n = 29 and n = 4 respectively). Specific analyses had slightly varying sample sizes due to some instances of missing data. We used one-factor Kruskal-Wallis ANOVA and a posteriori pair-wise comparisons of marsh classes to test for differences in marsh surface elevation and edaphic variables (e.g., salinity) and percent cover of the nine most frequently-occurring vascular plant species in the 1 m² plots (multiple testing was adjusted with the Benjamini and Hochberg method at α = 0.05). We evaluated how much variance was explained by wetland class for each response variable by computing a general effect size statistic, η² = χ²/(N-1)*100% (Serlin et al. 1982), for each Kruskal-Wallis test where η² is a metric of percent variance explained that ranges from 0% to 100%.

We examined species richness patterns based on presence/absence data collected in 0.25 m² plots. We tested for differences in plot-level plant richness among marsh classes with parametric ANOVA and pair-wise comparisons with Tukey’s adjustment using package “lsmeans” after confirming homogeneity of variances with Levene’s test (package “lawstat”). We estimated the percent variance in richness accounted for by wetland class by computing the effect size from terms in the ANOVA model: ω² = (SS_factor – df_factor*MS_error)/(SS_total + MS_error)*100%, where SS = sum of squares, df = degrees of freedom, and MS = mean square (ω² is analogous in scale and concept to η² above; Grissom & Kim 2012). We also evaluated differences in species pools among the three wetland classes with species accumulation curves using 500 permutations to generate curves with 95% confidence intervals (package “vegan”).

We also tested for differences in total percent cover of non-native species among wetland classes with a Kruskal-Wallis test and a posteriori pair-wise comparisons. We classified species as native or non-native using local floristic sources (Jefferson 1975; Cooke 1997; Kozloff 2005; Jaster et al. 2016). We computed total non-native cover as the summed cover values of all likely non-native species occurring at least once in each 1 m² plot, including Agrostis stolonifera, Ammophila arenaria, Cirsium arvense, C. vulgare, Cotula coronopifolia, Cynosurus echinatus, Erechtites minima, Holcus lanatus, Lotus corniculatus, Phalaris arundinacea, Plantago coronopus and Schedonorus arundinaceus. Most of these 12 species have been treated consistently in the literature as non-native in Oregon, but Agrostis stolonifera and Phalaris arundinacea may have some native populations (Kozloff 2005; Jaster et al. 2016). Additionally, we treated all Agrostis occurrences as the dominant A. stolonifera, though it is possible that additional species are present in the flora. We did not include species in non-native cover totals when their status was equivocal (Juncus gerardii; Kozloff 2005; Jaster et al. 2016, USDA 2017), or where they were only identified to genus level (e.g., Spergularia, Rubus).

We used percent cover data to calculate Bray-Curtis dissimilarities among plots (58 taxa from the 1 m² plots identified to at least the genus level) and classify vegetation into assemblage types (hierarchical cluster analysis) and evaluate differences in overall composition among the three marsh classes (non-metric multidimensional scaling, NMDS) with package ‘vegan’. With the cluster analysis we delineated the number and types of wetland plant assemblages in the tidal marshes independently of their NWI classification. We used average linkage with the function ‘hclust’ in the base package of R, grouping assemblages at the 60% dissimilarity level. For the NMDS analysis, we used the function ‘metaMDS’ on the Bray-Curtis dissimilarity matrix from the square root-transformed and Wisconsin standardized percent cover of all species (n = 146; Oksanen 2015). We selected the best fitting two-dimensional ordination after 20,000 restarts. We tested for differences in overall species composition between wetland classes (globally and pair-wise comparisons) with function “adonis” in “vegan” using 9999 permutations.

The cluster analysis yielded 25 plant assemblages, eight of which were represented by at least five vegetation plots in the dataset. For these eight assemblages, we tested for differences in tidal elevation, light transmission through plant canopies, and edaphic variables with one factor ANOVA and least squared a posteriori means tests (summer and winter pore water salinity and light transmission data were natural log-transformed since this improved homoscedasticity according to Levene’s test). We also tested for differences in plant species richness and total percent cover of non-native species among the assemblages with Kruskal-Wallis ANOVAs since transformation did not improve homoscedasticity (Levene’s test) followed by pair-wise comparisons among assemblages with the Benjamini and Hochberg method. We determined effect sizes on response variables due to assemblage identity (ω² or η²) as described above.

Results

Elevation, edaphic characteristics, and plant cover.

Wetland surface elevation and soil characteristics differed among the three major emergent marsh types (Table 2). Low estuarine marsh plots occurred at lower elevations relative to local MHHW on average than did high estuarine or tidal palustrine marsh classes. However, elevations ranged broadly within each marsh class with considerable overlap. Summer and winter pore water salinities were highest in low estuarine marsh, intermediate in high estuarine marsh, and lowest in tidal palustrine marsh, with significant differences among all wetland classes. Overall, winter pore water salinities were 56‐70% lower than summer values. Surface sediments in low estuarine marsh contained less TOC and TON than high estuarine or tidal palustrine marshes which were similar. Sand, silt, and clay content also differed significantly between low and high estuarine marsh, but the relative magnitude of differences was small, and neither class differed from tidal palustrine marsh. Among the abiotic factors examined, wetland class explained a relatively large percentage of variability in winter-time pore water salinity (η² = 51.1%) and summer-time pore water salinity (η² = 36.2%). However, less variability in total sediment nitrogen (η² = 11.6%), total sediment carbon (η² = 10.3%), or marsh surface elevation, sand, silt, and clay content (each <10%) was explained by wetland class.

Table 2.

Sediment characteristics among three tidal marsh classes in Oregon estuaries (mean ± SE).

	Tidal marsh habitat class			Kruskal-Wallis ANOVA (df =2)
Edaphic characteristic	Low estuarine	High estuarine	Tidal palustrine	χ2	P	η2 (%)
Marsh elevation (m, MHHW)	−0.106 ± 0.0434a	0.134 ± 0.0451b	0.091 ± 0.0515b	13.0	0.001	8.8
Summer pore water salinity	23.2 ± 1.46a	14.3 ± 1.53b	5.9 ± 0.83c	53.2	<0.0001	36.2
Winter pore water salinity	10.3 ± 0.68a	5.5 ± 0.68b	1.74 ± 0.162c	75.6	<0.0001	51.1
Total sediment carbon (%)	10.5 ± 0.80a	13.4 ± 1.11b	15.4 ± 1.13b	15.4	0.0005	10.3
Total sediment nitrogen (%)	0.72 ± 0.048a	0.90 ± 0.060b	1.06 ± 0.075b	17.4	0.0002	11.6
Sand content (%)	5.6 ± 0.84a	4.4 ± 1.26b	6.5 ± 2.39ab	7.8	0.02	5.2
Silt content (%)	79.9 ± 0.55a	77.9 ± 0.80b	78.0 ± 1.82ab	8.4	0.01	5.6
Clay content (%)	14.4 ± 0.58a	17.7 ± 1.03b	15.5 ± 0.97ab	9.6	0.008	6.4

Total average vascular plant cover exceeded 85% of plot area in all three marsh classes. The percent cover of nine common plant species varied across the three marsh classes (Figure 1; Supplemental table 1). Species with higher cover in low estuarine marsh included Sarcocornia perennis, Distichlis spicata, Jaumea carnosa, and Triglochin spp. Deschampsia cespitosa reached its greatest cover in high estuarine marsh. Carex lyngbyei occurred in all emergent marsh types, but had its greatest cover in tidal palustrine marsh. Although a single species often dominated cover within any individual 1 m² plot, no single species ever averaged more than 21% cover overall within any specific marsh class. Light transmission through plant canopies to the sediment surface was significantly different among all marsh classes (χ² = 27.2, df = 2, P < 0.0001, η² = 18.2%; Supplemental Figure 1), with the lowest light penetration in tidal palustrine marsh suggesting that it had the most dense vegetation coverage.

Figure 1.

Percent cover of nine common marsh plant species in low estuarine, high estuarine, and tidal palustrine marshes in four Oregon estuaries (means ± SE). Statistics (χ², P, η²) show results of one-factor Kruskal Wallis tests. Wetland classes sharing the same letter did not have significantly different cover.

Species richness.

We identified 55 vascular plant taxa to the species or genus level in the 0.25 m² plots, with 60 taxa having at least 1% cover in the larger 1.0 m² plots (Supplemental table 1). Macroalgae, bryophytes, and lichens were also present in the sampled plots, though the latter two groups of organisms typically occurred as epiphytes on other vegetation or on nurse logs at elevations likely above tidal influence. Vascular plant species richness at the (0.25 m²) plot level differed overall by wetland class (F_2,147 = 3.9, P = 0.02, ω² = 3.7%, Figure 2A), but the only significant pair-wise difference was greater species richness in high estuarine marsh compared with tidal palustrine marsh. High estuarine marsh had larger species pools than low estuarine marsh as suggested by non-overlapping confidence intervals for species accumulations curves (Figure 2B), but differences between tidal palustrine marsh and the other classes were less clear due to lower sample size.

Figure 2.

Variability in plot-level (A) and cumulative (B) plant species richness among tidal marsh classes. In (A), boxplot whiskers show the full range of values. Species accumulation curves in B are bracketed by 95% confidence intervals. Wetland classes sharing the same letter were not significantly different.

Common non-native species occurring in the wetlands included Agrostis stolonifera (occurred in 34% of all 1 m² plots), Schedonorus arundinaceus (7%), Phalaris arundinacea (5%), and Cotula coronopifolia (2%). Total non-native cover in 1 m² plots differed significantly among marsh classes, with greatest cover in tidal palustrine marsh and lowest cover in low estuarine marsh (χ ² = 33.3, df = 2, P < 0.0001, N = 147, η² = 22.8%; Figure 3).

Figure 3.

Differences in total non-native species cover among tidal marsh classes (mean ± SE). Wetland classes sharing the same letter were not significantly different. Boxplot whiskers show the full range of values.

Species assemblage identification and ordination.

Cluster analysis based on percent cover data indicated that 25 different assemblages were present in the tidal marsh dataset at a cutoff of 60% dissimilarity (Figure 4). Eight assemblages were represented by at least five plots in the data set (Supplemental table 2). The most commonly occurring assemblage was dominated by Deschampsia cespitosa with Juncus balticus ssp. ater, Distichlis spicata or another species present as a co‐dominant (assemblage F). Another common assemblage was a mixture of Sarcocornia perennis, Jaumea carnosa and Distichlis spicata (H). Agrostis stolonifera, together with one of several other co-dominant species, also commonly occurred in the marshes we sampled (assemblage C). Less commonly occurring assemblages (not shown in Supplemental table 2) included those dominated by Jaumea carnosa, Schedonorus arundinaceus, Spergularia spp., Sarcocornia perennis, Triglochin maritimum with Sarcocornia perennis, or Phalaris arundinacea.

Figure 4.

Vascular plant assemblages in Oregon tidal marshes according to hierarchical clustering with average linkage. Capital letters A-H denote the eight most common assemblages in the data set, described further in Supplemental Table 2.

There was only a partial correspondence between assemblage identity and emergent tidal marsh class. Many of the common assemblages were found in at least two of the three wetland classes examined in this study (Supplemental table 2). In particular, assemblages C and D showed a poor correspondence with tidal marsh class since they occurred fairly evenly across all three marsh classes. Assemblage G was exclusively found in low estuarine marsh. Wetland classes differed to some degree in overall species composition (NMDS ordination; Figure 5A). The ordination showed partial separation of vegetation composition between low estuarine, high estuarine, and tidal palustrine marsh, but also suggested substantial overlap in composition among wetland classes. The compositional differences between classes were statistically significant, but relatively weak in magnitude (ADONIS test on three classes: F_2,143 = 9.1, R² = 0.11, P < 0.0001). In pair-wise tests, all three classes were statistically distinct from each other (all P < 0.0001), with the greatest difference between low estuarine and tidal palustrine marsh (R² = 0.13). The ordination also indicated differences among the 8 common marsh plant assemblages identified in the cluster analysis, although there were also instances of overlap among assemblages (Figure 5B).

Figure 5.

Non-metric multidimensional scaling plots of plant community composition in Oregon tidal marshes. Panels show the same plots classified by (A) wetland class and (B) by assemblage type identified in the cluster analysis (see Figure 4). Grey stars in panel B represent plots that did not belong to any of the eight-most common assemblages.

Common plant assemblages differed overall in tidal elevation (F_7,102 = 6.1, P < 0.0001, ω² = 24.6%), summer pore water salinity (F_7,101 = 6.4, P < 0.0001, ω² = 25.7%), winter pore water salinity (F_7,102 = 11.2, P < 0.0001, ω² = 39.3%), sediment TOC (F_7,102 = 3.9, P = 0.0008, ω² = 15.7%), sediment TON (F_7,102 = 3.1, P = 0.005, ω² = 11.9%), and light transmission through plant canopies (F_7,102 = 6.5, P < 0.0001, ω² = 25.8%) (Figure 6A–F). Among these six abiotic factors, pore water salinity tended to differ most strongly among assemblages. For example, summer pore water salinity was much higher in assemblages F, G, and H than in A, B, and E. Pair-wise comparisons also showed that assemblages tended to occur along continuums of edaphic conditions. Assemblages also had overall differences in vascular plant species richness (χ² = 25.1, df = 7, P = 0.0007, η² = 23.2%; Figure 7A) although many individual pairs of assemblages were not statistically distinct. Cover of non-native species varied strongly by assemblage type (χ² = 80.3, df = 7, P < 0.0001, η² = 73.7%) with assemblage C having very high non-native cover and assemblages E-H having very low cover (Figure 7B).

Figure 6.

Differences in wetland surface elevation, edaphic characteristics, and light attenuation through plant canopies among eight common tidal marsh assemblages in Oregon. Thick horizontal bars show assemblage medians, boxes show 25 and 75% quantiles of the data distribution, and whiskers show the full range of values observed within each assemblage. Assemblages sharing the same letters are not significantly different (Tukey’s HSD test at α = 0.05). TOC = total organic carbon; TON = total organic nitrogen.

Figure 7.

Differences in (A) vascular plant species richness (in 0.25 m² plots), and (B) total non-native species cover among eight common marsh assemblages identified in the study. Thick horizontal bars show assemblage medians, boxes show 25 and 75% quantiles of the data distribution, and whiskers show the full range of values observed within each assemblage.

Discussion

Standardized wetland classification systems such as the National Wetlands Inventory have a number of important potential applications in coastal wetland ecology and management including constructing wetland inventories, characterizing the diversity of estuarine wetland types, and assessing ecosystem change across large spatial and temporal scales. However, the utility of these classification systems for such purposes is contingent on the degree to which they accurately and consistently distinguish wetland from non-wetland habitat (Guidugli-Cook et al. 2017) and characterize ecologically-meaningful differences among classes. Other approaches to classifying coastal wetlands could include the use of indicator plant species or assemblages, or classifications based on gradients of elevation, salinity, or soil type. We discuss the relative utility of NWI classes for characterizing the structure of tidal marshes in Oregon, how the descriptive capacity of NWI classes compares with a classification based on vegetation assemblages, and implications for using NWI wetland classes to map or model wetland change along the Pacific Northwest coast.

NWI descriptive power in tidal emergent marshes.

Our examination of abiotic and biotic characteristics in three common emergent marsh types in Oregon estuaries suggested that NWI-based wetland classes had varying success in describing differences among classes. We found large differences among wetland classes for parameters such as soil pore water salinity and cover of many individual plant species, but much smaller differences in tidal elevation, sediment grain size, plot-level plant species richness, and overall species composition. Low marsh had higher pore water salinities than high marsh or tidal palustrine marsh, consistent with the fact that this wetland class is usually found in closer proximity to the ocean and the influence of tides. Notably, both average summer and winter pore water salinities in tidal palustrine marsh (5.9 and 1.7 ppt respectively) exceeded the 0.5 ppt threshold for the palustrine-estuarine division used by NWI, suggesting that at least some of the tidal palustrine marshes in Oregon should be re‐classified as estuarine habitat (e.g., Brophy 2013).

NWI-based classes poorly characterized spatial differences in tidal elevation, with class identity explaining only 9% of the variability in marsh surface elevation among all plots (Table 2). Similarly, only 5–12% of total variability in sediment nitrogen, carbon, sand, silt, and clay content was associated with wetland class. These differences among classes were statistically significant due to large sample sizes, but not particularly large ecologically. Thus, inferring differences in sediment organic carbon stocks or sediment type at the estuary or regional scale by using NWI class may have low accuracy.

Wetland classes more effectively characterized differences in vegetation composition. The percent cover of several individual plant species varied by class, with Sarcocornia perennis, Jaumea carnosa and Distichlis spicata having a clear affinity for more saline wetlands, and Carex lyngbyei having greater abundance in fresher marshes (Janousek and Folger 2014). The relative abundance of these species could serve as useful rapid indicators of marshes of varying sediment pore water salinity, whether along vertical gradients within individual sites, or among sites varying in distance from the mouth of an estuary. A few common species along the Oregon coast like Juncus balticus ssp. ater were common in all marsh classes, and may be less effective indicators of different estuarine wetland types.

In terms of overall vegetation composition and diversity, wetland classes correlated with differences in regional species pools. Cumulative species pools were greater in high marsh than in low marsh (Figure 2B), indicating that NWI could be helpful for assessing spatial patterns of gamma diversity across estuarine landscapes which could be useful for wetland management or restoration. At the plot level however, all three marsh classes had similar species richness (Figure 2), despite the known presence of positive relationships between tidal elevation and plant richness in Oregon tidal wetlands (Janousek and Folger 2014) and in other temperate regions (Kunza and Pennings 2008), and salinity relationships with species richness (Więski et al. 2010). There were also only relatively small differences in overall plant composition among wetland classes (Figure 5A). Salinity and tidal elevation are two of the major drivers of plant composition (Watson and Byrne 2009; Weilhoefer et al. 2013; Janousek and Folger 2014), so our finding that NWI class adequately described spatial differences in salinity, but not tidal elevation, may help explain its mixed success at characterizing differences in vegetation composition in this study. In Oregon tidal marshes, the spatial scale at which NWI mapping has been implemented fails to capture important gradients in marsh surface elevation as well as smaller-scale wetland topographic heterogeneity (e.g., small tidal channels, depressions in the marsh surface, or nurse logs) that is known to affect plant composition and diversity (Fogel et al. 2004; Morzaria‐Luna et al. 2004).

Our findings of mixed success in NWI’s ability to characterize differences in tidal marsh structure in Oregon has implications for using standardized wetland classes to infer spatio-temporal differences in wetland function. On the one hand, our finding of higher non-native species cover in high marsh and tidal palustrine marsh relative to low marsh (Figure 3) suggests that fresher tidal wetlands may be more susceptible to invasion by non-native species in the Pacific Northwest (e.g., Weilhoefer et al. 2013), and that NWI class might be a useful proxy for determining tidal marsh vulnerability to invasion by non-native species at larger spatial scales. On the other hand, it may be difficult to infer spatial differences in ecosystem function across these landscapes due to changes in species richness or composition (e.g., Hooper et al. 2005; Cardinale et al. 2012) since these aspects of vegetation structure were fairly similar among classes.

An assemblage-based approach.

Characterization of wetland structure based on vegetation assemblage type could provide an alternative means of classifying and mapping tidal wetland diversity, especially for the important attributes of wetland elevation and vegetation composition. Using plant cover data and an intermediate cutoff in compositional dissimilarity to define assemblages, we found a large number of assemblage types present in Oregon’s tidal wetlands, similar to earlier studies (Jefferson 1975; Eilers 1975; Weilhoefer et al. 2013). Assemblage identity appeared to be relatively effective at characterizing differences in marsh surface elevation, plot-level species richness, and non-native species cover.

Although characterization of wetlands by the number and type of plant assemblages present may provide a useful supplement to NWI for describing wetland structure, it also presents some distinct challenges. First, emergent marsh vegetation assemblages are likely to exist across a continuum, rather than as discrete types that can always be easily classified in the field. For instance, although several of our common assemblage types (E, G, H) closely matched assemblages identified in previous attempts to classify the Oregon flora (Jefferson 1975, Eilers 1975, Frenkel et al. 1981, Taylor 1983, Frenkel and Morlan 1991), these sources identified additional assemblages or combinations of common species not well captured in our sampling. Second, effective use of assemblage data to classify wetland types would require consistent classification of community types across a region, an increasing challenge at larger spatial scales, where changes in regional species composition may lead to novel associations of species. Finally, obtaining assemblage-level data may require increased investment in field reconnaissance by trained personnel who can identify common wetland species in situ, or who can generate maps of vegetation distribution from multi-spectral data (Tuxen et al. 2011; Moffett and Gorelick 2013).

Implications for use of NWI in wetland landscape studies.

The National Wetlands Inventory is a valuable dataset documenting wetland type and extent across the United States, with a potential range of applications. For instance, landscape ecologists have used NWI data to assess coastal ecosystem function (Tiner 2003) or estimate coastal sediment carbon and nitrogen loss due to wetland drowning (Zhong and Xu 2011). NWI data are often used as an input for the “Sea-level Affecting Marshes Model” (SLAMM), a tool to project the future distribution of wetland types under different sea-level rise scenarios (e.g., Geselbracht et al. 2015; Stralberg et al. 2011). The strengths and weaknesses of the NWI classification system will impact the accuracy of other applications in which it is used. Studies primarily in freshwater wetlands have highlighted some of the weaknesses of NWI, including incorrect assignment of wetlands to specific classes, underestimates of the spatial extent of larger wetlands, or failure to include very small wetlands in landscape inventories (Matthews et al. 2016, Guidugli-Cook et al. 2017). These challenges may be due to the age of available NWI maps for some areas, lack of ground-truthing to verify mapping efforts, or the scale at which NWI mapping is conducted. In the Oregon estuaries we studied, NWI weaknesses appeared to be due more to class incongruence with in situ differences in elevation or vegetation, rather than misclassification of vegetated versus non-vegetated wetland areas.

In the coastal zone, informative sea-level rise modeling requires accurate information on tidal elevation, salinity, and other coastal processes such as sediment deposition and productivity to project changes in wetland extent or type (Schile et al. 2014; Thorne et al. 2015). A number of coastal models are available for projecting sea-level rise impacts to coastal wetlands (Schile et al. 2014; Swanson et al. 2014), but SLAMM is a popular model since it can be run with minimal user inputs if site-specific parameter data are lacking. For land-use cover data, SLAMM can incorporate NWI polygons which are then translated to simplified coastal habitat categories including “tidal fresh marsh”, “regularly flooded marsh”, “irregularly flooded marsh”, and “tidal swamp” (Clough et al. 2010). Based on our finding that NWI marsh classes may provide inadequate information on tidal marsh elevation, species richness, or edaphic characteristics in Oregon, we caution against making strong inferences on changes in coastal wetland composition when sea-level rise projections are made with SLAMM and NWI data. This seems particularly relevant to putative differences between “regularly flooded marsh” and “irregularly flooded marsh”.

Finally, we suggest two ways in which NWI data could be supplemented with other data sources to better describe structure in Oregon tidal wetlands. First, in cases where elevation data are available (e.g., LiDAR raster or field measurements), existing NWI polygons (e.g., “E2EMN” and “E2EMP”) could be subdivided into elevation classes. Since NWI relatively effectively provides information on soil salinity regime, the addition of elevation data would help users characterize differences among tidal marshes along both major ecological axes (Janousek and Folger 2014). In the case of SLAMM, if the user has a relatively precise digital elevation model (DEM) as input into the model, it may be more profitable to first run SLAMM with sea-level projections of interest, and then use the final DEM (rather than solely relying on the model’s output of land cover classes) to refine maps of future wetland type based on locally-customized relationships between elevation and wetland type. Second, vegetation data could be incorporated into spatial maps of estuarine marshes in the Pacific Northwest, either as assemblage type or by dominant species. These supplemental data may provide more information on wetland elevation, species richness, non-native species abundance, and edaphic conditions than NWI class attribution alone.

Conclusions.

High species and assemblage diversity is characteristic of the Oregon tidal wetland flora (Weilhoefer et al. 2013), challenging broad-scale classification schemes such as NWI to adequately characterize the diversity of wetland types present in the region or their complex spatial configuration across gradients of elevation and salinity. In Oregon tidal marshes, NWI-based marsh classes provide useful information about pore water salinity regime, overall plant species pools, cover of select common species, and average non-native species cover, but they are less informative of spatial differences in marsh elevation, sediment grain size and organic content, or plot-level species richness. Thus landscape-scale applications of NWI for coastal modeling, resource assessment, or change analysis in the Pacific Northwest should be conducted with these limitations in mind. Data collected on vegetation composition, whether gathered remotely or by rapid assessment methods, may be a useful supplementary source of information on spatial variation in tidal wetland condition or function at the landscape scale.