Wetland water quality patterns and anthropogenic pressure associations across the continental USA

Abstract

Anthropogenic impacts on lake and stream water quality are well established but have been much less studied in wetlands. Here we use data from the 2016 National Wetland Condition Assessment to characterize water quality and its relationship to anthropogenic pressure for inland wetlands across the conterminous USA. Water samples obtained from 525 inland wetlands spanned pH from <4 to >9 and 3 to 5 orders of magnitude in ionic strength (chloride, sulfate, conductivity), nutrients (total N and P), turbidity, planktonic chlorophyll, and dissolved organic carbon (DOC). Anthropogenic pressure levels were evaluated at two spatial scales – an adjacent scale scored from field checklists, and a catchment scale indicated by percent agricultural plus urban landcover. Pressure at the two spatial scales were uncorrelated and varied considerably across regions and wetland hydrogeomorphic types. Both adjacent- and catchment-scale pressure were associated with elevated ionic-strength metrics; chloride elevation was most evident in road-salt using states, and sulfate was strongly elevated in a few sites with coal mining nearby. Nutrients were elevated in association with catchment-scale pressure but concomitant changes were not seen in planktonic chlorophyll. Acidic pH and high DOC occurred primarily in upper Great Lakes and eastern seaboard sites having low anthropogenic pressure, suggesting natural organic acid sources. Ionic strength and nutrients increased with increasing catchment-scale pressure even in Flats and closed Depression and Lacustrine sites, which indicates connectivity to rather than isolation from upland anthropogenic landuse even for wetlands lacking inflowing streams.

Introduction:

The large-scale loss of wetlands brought about by decades of agricultural and urban development has been well documented (Dahl 1990, Davidson 2014; Hu et al. 2017), as has degradation of many of the remaining wetlands (Moomaw et al. 2018). However, attention to wetland degradation is primarily focused on their biological communities – plants, birds, amphibians, microbes (e.g., Findlay and Houlahan 1997; Galatowitsch 2018; Gaiser et al. 2015; Genet and Olsen 2008; Guntenspergen et al. 2002; Lougheed et al. 2008; Wright et al. 2006) – and not on changes to water quality. In part, this is because surface water is not consistently present and sampleable in wetlands, even though its evidence via water-logged soils and water-adapted plants is a key element of wetland determination (Mitch and Gosselink 2015). Further, water quality is often treated as a stressor to wetlands rather than a response itself (e.g., Ehrenfeld and Schneider 1993; Liston et al. 2008; Sierszen et al. 2006), or else the focus is on capacity of wetlands to ameliorate the quality of the water they intercept to the benefit of downstream systems (e.g., Cheng et al. 2020; Johnston 1991; Jordan et al. 2011; Zedler 2003). This only limited examination of wetland water quality responses is in distinct contrast to lakes, streams, and marine coastal waters where a plethora of studies have analyzed water quality as an ecological response and used water quality to rate waterbody conditions, infer ecosystem benefits, and reach protection and remediation decisions (reviews in Carpenter et al. 1998; Downing et al. 2021; Lintern et al. 2018; Schindler 2006, Smith 2003, Vadeboncoeur et al. 2003).

The basic premise of the analyses presented here is simple: use data collected in the National Wetland Condition Assessment (NWCA) to assess wetland water quality patterns in relation to anthropogenic pressures in the same way these are typically assessed in lakes and streams and coastal waters. Conceptually, this is not new as some previous studies have examined wetland water quality in this way. However, these studies have operated only over local to regional spatial scales and primarily in central and northeastern North America -- e.g., Minneapolis metropolitan area (Detenbeck et al. 1993), eastern Ontario (Houlahan and Findlay 2004), Laurentian Great Lakes (Crosbie and Chow-Fraser 1999, Morrice et al. 2008) – and fall short of covering many wetland types and geographic settings across the USA. To our knowledge, the NWCA -- spanning almost 25 degrees of latitude and 60 degrees of longitude across the conterminous United State – is geographically the most comprehensive dataset on wetland conditions available, and thus allows the investigation of wetland water quality versus anthropogenic pressure over larger spatial scales than ever before.

Here, we use water quality data collected as part of the 2016 NWCA (the second 5-year NWCA assessment cycle) to examine the range of water quality conditions found across USA and their association to wetland types, biogeographic regions, and anthropogenic pressures. We make an explicit comparison of anthropogenic impacts at the adjacent (wetland shoreline and buffer) scale versus the catchment scale, to help inform whether wetlands are or are not strongly connected to terrestrial landscapes. Water quality data were also collected in the 2011 NWCA (the first assessment cycle) and used to demonstrate that increasing levels of watershed agricultural landuse were associated with increasing levels of conductivity, nutrients, and plankton chlorophyll (Trebitz et al. 2019). The present study using the 2016 NWCA dataset extends this understanding to a broader set of wetlands, more resolved biogeographic regions and hydrogeomorphic categories, and several additional interesting water quality analytes.

Methods

Sampling design and water quality measurements

Water quality sampling was conducted as part of the 2016 cycle of the NWCA, which included wetlands across the conterminous USA (all states except Alaska and Hawaii). The NWCA spans both naturally occurring wetlands and wetlands created or restored as mitigation sites, but excludes wetlands constructed for wastewater treatment purposes. The bulk of sites were randomly selected using a spatially balanced design (Olsen et al. 2019), but to boost the number of putative low-disturbance sites NWCA also sampled hand-picked sites nominated for inclusion by partner agencies. We refer to the randomly selected sites as probability sites because each has a known selection probability that can be used as a weighting factor to generate wetland population estimates, whereas handpicked sites lack such probabilities and cannot be included in population estimates. The sampling frame for the 2016 random site selection was based on a combination of Wetland Status and Trends polygons (Dahl and Bergeson 2009) and National Wetland Inventory polygons (http://www.fws.gov/wetlands/Data/Mapper.html).

NWCA sampling is assessment-area based; sampling focuses on a 0.5 ha polygon encompassing a randomly selected point coordinate called the assessment area (AA). Typically, this AA is a circle with 40m radius (about 0.5-hectare area), but the field crew can lay it out in an oblong or irregular shape if necessary to stay within wetlands boundaries. The NWCA also collects some data (notably on anthropogenic stressors, see below) in a 100m wide band encircling the AA called the buffer; depending on wetland size and shape the buffer can fall entirely within the wetland, entirely within upland, or a mix of the two. If no surface water is found in the AA -- even if water is present elsewhere in the wetland -- the site is sampled for other NWCA parameters (vegetation, soil profile) but marked as not yielding a water chemistry sample; water inundation status is not considered in visit timing. Sites are not sampled if the AA lacks wetland-type vegetation, if the wetland is in active use for crop production or aquaculture, if landowners deny access permission, or if excessive effort is required to reach it (>1 day travel by foot or boat). Additional details on site selection and sampling design are available in USEPA (2016a). NWCA protocols include a second same-year sampling visit to 10% of sites to characterize temporal variability, but only first-visit data are analyzed here.

In brief, water quality sampling used a pole-mounted dipper to fill a 1L bottle that was filtered on-site (0.7-micron glass-fiber filter) for later chlorophyll-a analysis and a 1L cubitainer as basis for all other water chemistry analyses. To the extent possible within the AA, the water sample was taken away from inlets and outlets. The cubitainers and chlorophyll-a filters were chilled, placed in darkness, and express-shipped to analytical laboratories for later analysis. The bulk of the 2016 NWCA samples were analyzed by the ALS Middletown Pennsylvania laboratory, while samples collected in Nebraska, Utah, and Wisconsin were analyzed by the University of Nebraska-Lincoln Water Sciences laboratory, Utah Public Health laboratory, and Wisconsin State Laboratory of Hygiene, respectively. Laboratory analysis methods are summarized in Table 1 and detailed in USEPA (2015). Water quality analytes examined here are pH, specific conductance (COND), total nitrogen (TN), total phosphorus (TP), planktonic chlorophyll a (CHLA), turbidity (TURB), dissolved organic carbon (DOC), chloride (Cl^-), and sulfate (SO₄⁼). We do not present data on nitrate/nitrite and ammonia which lacked analyzable signal due to high below-detection rates. Measurements of TURB, DOC, Cl^-, and SO₄⁼ were newly added to the NWCA in 2016 while the other analytes were already measured in 2011.

Table 1.

Summary of water quality analytes in the 2016 NWCA. N is number of inland sites (out of 525) where data were obtained, detection limit target is per US EPA 2015 (actual values vary due to method differences among labs and dilution factors for some analytes).

Analyte (abbreviation)	Units	Method in brief	N	N below detect	Target detect limit	Median (range)
Conductivity (COND)	μS/cm	Lab meter, standardized to 25°C	525	none	1.0	158 (10 – 14,300)
Chloride (Cl-)	mg/L	Filtration, ion chromatography	524	3	0.1	4.2 (0.02 – 538)
Sulfate (SO4=)	mg/L	Filtration, ion chromatography	524	24	0.25	2.8 (0.03 – 7310)
Total nitrogen (TN)	μg/L	Persulfate digestion, colorimetric analysis	525	none	10	1200 (134 – 67,600)
Total phosphorus (TP)	μg/L	Persulfate digestion, colorimetric analysis	525	1	2.0	104 (2.5 – 18,500)
TN:TP ratio	--	Computed as molar ratio	525	n/a	n/a	64 (1.7 – 7240)
Chlorophyll-a (CHLA)	μg/L	Collect on filter, buffer with MgCO3, acetone extract, fluorometry	523	52	0.5	5.3 (0.2–2136)
Turbidity (TURB)	NTU	Lab meter	525	2	1.0	7.0 (0.1 – 2170)
Dissolved organic carbon (DOC)	mg/L	Filtration, persulfate oxidation, infrared detection	511	1	0.1	11.3 (0.1 – 222)
pH	--	Lab meter	525	n/a	n/a	7.5 (3.1 – 9.6)

Wetland inclusion and classification

Although NWCA samples both tidally influenced and freshwater wetlands, only data from freshwater wetlands are presented here. This is because water quality naturally differs between tidally influenced and freshwater wetlands, and because tidal inputs of broadly mixed oceanic water make immediate watershed effects difficult to resolve. The latter would also complicate inferences for wetlands receiving seiche-driven inputs from the Laurentian Great Lakes (Morrice et al. 2011) but, unlike the 2011 NWCA, no Great Lakes fringing wetlands were sampled in 2016. We also excluded from analysis one naturally brackish wetland on Utah’s Great Salt Lake and nine sites not classified as tidal but with sea-water influence suggested by ocean proximity and high conductivity (>1000 μS/cm). Five of these nine were situated on small coastal islands, while the other four were on low-lying mainland within a few kilometers of an ocean.

Our analyses of wetland water quality intersect two pertinent classification schemes, namely biogeographic region which organizes general climate and landform patterns across the USA (aggregates of level III ecoregions described in Omernik and Griffith 2014) and hydrogeomorphic type which classifies wetlands according to water source (HGM type hereafter; Brinson 1993, NRCS 2008). The biogeographic regions used for NWCA reporting (per US EPA 2016b) are 1) eastern mountains and upper midwest (EMU) spanning the New England states, the Appalachian mountains, the upper Great Lakes, and highlands of Oklahoma, Arkansas, and Missouri; 2) inland coastal plains (ICP) spanning the Gulf and Atlantic coasts and the lower Mississippi river region; 3) interior plains (PLN) spanning much of the central USA; 4) western mountains (WMT) spanning the wetter higher-elevation portions of the western states; and 5) xeric (XER) which comprises the largely arid remainder of the western states (Figure 1). Sites were classified into HGM types by the field crews, with classifications subsequently verified or corrected using aerial imagery. Our analyses used the HGM categories of ‘Riverine’, ‘Slope’, and ‘Flats’ as is, but recombined the categories of ‘Depression’ and ‘Lacustrine fringe’ based on whether there was a stream input (‘Dep/Lac-open’, hereafter) or not (‘Dep/Lac-closed’, hereafter). We used aerial imagery to look for presence vs. absence of inflowing streams for all lacustrine fringe wetlands and any depression wetlands where the field crew had checked the “inflow present” box. Most depression wetlands lacked stream inflow (~94%) and went into the Dep/Lac-closed category, while lacustrine fringe wetlands were split more evenly between having vs. lacking an inflowing stream (~60 vs. 40%).

Figure 1:

Map of sites sampled by the 2016 National Wetland Condition Assessment. Symbols indicate wetland hydrogeomorphic type except sites not yielding a water sample are marked with “X”, states are outlined in black, and red lines show division of states into west vs east and road-salt using (north) vs not (south). Background colors depict biogeographic regions: WMT=western mountains, XER=xeric, PLN=plains, EMU=eastern mountains and upper midwest, and ICP=interior coastal plains.

We also partitioned wetlands into two other geographical classifications for some analyses. One classification is eastern vs. western USA, with the dividing line placed at the western border of states transited by the Mississippi river (Figure 1). The second classification has to do with winter road salt usage, with states classified as substantially road-salt using or not based on maps in Hintz et al. (2022). We recognize there is considerable within-state variation in salt application but our classification is binary as indicated in Figure 1.

Anthropogenic pressure characterization

Anthropogenic activities potentially impacting wetland water quality are compared at two primary spatial scales – activities in close proximity to the sampled location referred to as the adjacent scale, and activities over the substantially larger surrounding catchment scale. We refer to these as ‘pressures’ to convey that their actual influence on water quality remains to be established. Field crew comments suggested that a third type of anthropogenic activity – mining – is also relevant for a small number of NWCA wetlands. We evaluated potential mining effects separately from general adjacent and catchment scale pressure (see analysis section).

We characterized adjacent-scale pressure from field checklists of human disturbances visible to the field crew in either the AA (intensity categories of low, medium, or high) or the surrounding 100m buffer (intensity inferred from position: inner, middle, or outer buffer-plot ring; details per USEPA 2016a). Our approach to scoring adjacent-scale pressure follows the physical alteration scoring procedure of Lomnicky et al. (2019) but uses only a subset of checklist items relevant to water chemistry (Online Supplement Table 1). Potential chemical alteration scores (CALT_NUT, CALT_SED, and CALT_SAL) were constructed by summing the relevant number of checkmarks in the AA and each of the 12 buffer plots weighted by their location (inside the AA weight =25, inner buffer ring=4, middle ring=2, outer ring=1), which assumes highest impact for disturbances within the AA and declining impacts with distance into the buffer. To correct for differences between scores of the three chemical alteration types arising simply from a different number of potential checklist items, we standardized them by dividing by the maximum value observed (187 for CALT_NUT, 268 for CALT_SED, and 78 for CALT_SAL) and multiplying by 100. We then computed CALT_CSUM as the sum of the three individual CALT scores. Finally, to facilitate graphical and statistical analyses, we divided CALT_NUT, CALT_SED, CALT_SAL, and CALT_CSUM into categories where zero was designated “minimal pressure” and non-zero scores were assigned to low, medium, or high by trisection of the data range.

We characterized catchment-scale pressure by intersecting the AA centerpoint coordinates with stream-based geospatial drainage polygons and imputing to the wetlands the percent agriculture plus urban landuse for those drainages (Online Supplement Table 1). The landuse classification is from the 30 m² pixel scale 2016 National Landcover Database (NLCD hereafter; Jin et al. 2019), and we obtained the drainage delineation from the StreamCat database, a publicly available U.S. EPA product that comes with NLCD attributes already compiled (Hill et al. 2015). We elected to use the immediate catchment as the drainage area for analyses – the area potentially contributing to any given stream segment via overland flow – rather than analyzing the full watershed which adds in the catchment of any segments further upstream. We recognize that stream-segment catchments may not equate exactly to the geographic area feeding water to Flats, Slope, or Dep/Lac-closed wetlands but no wetland-specific drainage area map is available for the conterminous USA. We classified sites as having minimal catchment-scale pressure if <10% agriculture + urban landuse based on plots showing no appreciable increase in COND or TP below 10% (Online Supplement Figure 1); sites above that threshold were assigned pressure levels of low (10 to <40%), medium (40–70%), or high (>70%) by trisection.

Analyses:

We proceeded through the analyses by: 1) first characterizing anthropogenic pressure levels in relation to wetland region and HGM type, then 2) characterizing the range of and relationship among the water quality analytes and how these differ by region and HGM type; then 3) describing how water quality distributions shift in association with general anthropogenic pressure levels; and finally 4) for a subset of the water quality analytes namely pH, Cl-, and TP, examining specifics of their relationship to adjacent and catchment-scale pressure. To infer anthropogenic pressure distributions across the conterminous USA, we used the site inclusion probabilities as areal weighting factors (per Stevens and Olson 2004) to extrapolate to the wetland population. However, all water quality distributions and relationships to vs. anthropogenic pressure were analyzed using only the actual (unweighted) site data. To enable the two spatial scales of anthropogenic pressure to be separated, we examined catchment pressure in the subset of sites where adjacent pressure was minimal or low, and conversely examined adjacent pressure where catchment pressure was minimal or low. Since a few of the wetlands had field crew comments indicating potential nearby mining and acid mine drainage is known to elevate sulfate in streams (Herlihy et al. 1990), we specifically examined patterns in SO₄⁼ in sites having potential mining impacts. Cases of elevated SO₄⁼ in conjunction with mining comments all fell within the Appalachia coal-mining region (per map in Strager et al. 2015), so we made a three-way comparison among sites having mining comments in that region, sites lacking mining comments but also in that region, and sites outside that region but in the same USA states as the first two groups.

Before statistical analyses, analyte concentrations below the laboratory detection limit were assigned a value of half that limit (per Hornung and Reed 1990). Except for pH, all water quality analytes were nonnormally distributed with a long right-hand tail, so we analyzed them using base 10 logarithmic transformation. There were no zero-values for any analytes so log-transformation did not eliminate any data points. To add to the ability to examine nutrient patterns, we computed TN:TP molar ratios (dividing concentrations by their respective molecular weights) and classified the results as indicative of nitrogen limitation if <20, phosphorus limitation if >50, and co-limitation in between (per Guildford and Hecky 2000).

Results

Distribution of wetlands yielding water quality data

The 2016 NWCA survey produced visit-1 assessment area (AA) water quality data from 525 inland wetlands, meaning wetlands that were not estuarine or tidally influenced. Every conterminous USA state had at least one inland wetland that yielded water quality data, with the best represented states being Kentucky (N=33), Minnesota (N=32), Colorado (N=26), Florida (N=25), Washington (N=24), Wisconsin (N=23), Montana (N=21), and Idaho (N=20). The east vs. west USA split divides the number of wetlands visited roughly in half (52% east, 48% west; Figure 1) but a higher percentage of sites yielded water samples in the east than the west (66% vs. 54%). Out of the 525 sites yielding water quality data, 12% were handpicked and the other 88% were probability sites; handpicked sites were most common in the EMU region (27%) and Dep/Lac-closed HGM type (20%) and least common in the ICP region (none) and Flats HGM type (4%; Table 2). We report on the handpicked vs. probability distinction because only probability sites can be used to infer population estimates for wetlands across the conterminous USA.

Table 2.

HGM types analyzed (rows), putative water sources for them, and number of sites yielding water quality data (probability + handpicked) in each biogeographic region (columns). Probability vs. handpicked are distinguished because only the former contribute to areal population estimates in Figures 2 and 3. Region acronyms are: EMU=eastern mountains and upper midwest, ICP=interior coastal plains, PLN=plains, WMT=western mountains, XER=xeric.

HGM type	water source	EMU	ICP	PLN	WMT	XER	Row N
Riverine	surface water	53+15	50+0	50+7	66+2	26+2	271
Dep/Lac-closed	groundwater & precip.	32+18	24+0	34+5	11+4	6+0	134
Dep/Lac-open	surface & groundwater	8+2	2+0	7+1	1+0	11+0	32
Flats	precipitation	8+2	31+0	4+0	3+0	1+0	49
Slope	groundwater	4+2	2+0	0+1	24+3	3+0	39
Column N		144	109	109	114	49	525

By region, wetlands were most likely to yield a water sample in the EMU (73% success rate) and least likely in the XER (37%), with WMT, ICP, and PLN intermediate (66, 62, and 55% respectively). By HGM type, sample yield rates were highest in Riverine wetlands (69%) and lowest in Flats (46%) with rates intermediate for Slope (53%) and Depression + Lacustrine (53%; we did not undertake closed vs. open classification for sites that did not yield a water sample). Per the HGM classification premise, predominant water sources are surface water for Riverine wetlands, groundwater for Slope wetlands, direct precipitation for Flats, surface + groundwater for Dep/Lac-open wetlands, and groundwater + precipitation for Dep/Lac-closed wetlands (Table 2). A sample size of at least N=5 was attained for all five biogeographic regions only in Riverine and Dep/Lac-closed HGM sites. Slope wetlands yielding water quality data were well represented only in the WMT region, Flats were well represented only in the ICP region, and Dep/Lac-open wetlands were generally sparse (Table 2).

Anthropogenic pressure

The three adjacent-scale pressure metrics CALT_NUT, CALT_SED, and CALT_SAL were intercorrelated (expected since they share some checklist items) but coefficients of 0.55 to 0.65 indicate some differences among them (Table 3). Catchment landcover ranged from 0 to 95% agricultural and 0 to 93% urban, but the two landcover types were uncorrelated and agriculture contributed most strongly to the combined percentage (Table 3) since cases of high urban landcover were few (>66% urban only at sites in Chicago, Indianapolis, and Kansas City). Out of the 525 inland wetlands with water quality data, 171 (32%) had minimal catchment-scale pressure (agriculture +urban <10%), 63 (12%) had minimal adjacent-scale pressure (CALT_CSUM=0), 93 (18%) had minimal pressure at both scales, and 198 (38%) exhibited some pressure at both scales. Importantly, adjacent-scale pressure metrics were uncorrelated with catchment-scale metrics (Table 3); their spatial scales also differ considerably with adjacent pressure assessed over a 0.062 km² area vs catchments having a median area of 6.3 km² (range 0.02 to 7984.1 km²). These findings confirm the validity of a primary assumption behind our analysis, namely that anthropogenic pressure to wetlands can play out in disparate ways across disparate spatial scales.

Table 3.

Pearson correlation matrix among anthropogenic pressure metrics across the 525 inland wetland sites. CALT_NUT, CALT_SED, and CALT_SAL are wetland-adjacent pressures likely to elevate nutrients, suspended sediments, and salts (ions) respectively, and percent of catchment in agriculture or urban landuse together constitute catchment-scale pressure. Correlations ≥0.5 are in bold font, and are all significant with Bonferroni-corrected p-values <0.001.

	CALT_NUT	CALT_SED	CALT_SAL	Agric. (%)	Urban (%)
CALT_SED	0.65
CALT_SAL	0.56	0.55
Agriculture (%)	0.05	0.06	0.17
Urban (%)	-0.10	-0.02	0.02	0.01
Catchment-scale pressure (%)	-0.01	0.04	0.16	0.88	0.47

The NWCA design allows for inference from the sampled probability-site wetlands to the wetland areal extent represented by these sites across the conterminous USA. Cumulative distribution plots that make this inference show some substantial differences among geographic regions and HGM types in the estimated distribution of anthropogenic pressure. The PLN region stands out as having much more right-shifted catchment-scale pressure than any other region (Figure 2, top left). Only ~10% of the conterminous-USA wetland area in the PLN was estimated to have minimal catchment pressure and >50% had high catchment pressure, whereas the other 4 regions all had >40% with minimal catchment-scale pressure and <30% with high pressure. At the adjacent scale, the EMU and ICP region were estimated to have substantially more left-shifted distributions (lower pressure) than the PLN and WMT and XER region. For example, the cumulative distributions show >40% of EMU and ICP with minimal adjacent pressure but only 10–20% of the other three regions with minimal adjacent pressure (Figure 2, bottom left). Among HGM types, Flats and Slopes were estimated to have substantially less catchment-scale pressure than Riverine and Dep/Lac-closed wetlands, while Dep/Lac-open wetlands were intermediate (Figure 2, top right). Adjacent-scale pressure was estimated to be lowest in Flats (>70% of Flats area having CALT_SUM=0) and highest in Dep/Lac-closed and Slope wetlands, with Riverine and Dep/Lac-open wetlands intermediate (Figure 2, bottom right).

Figure 2:

Cumulative distribution plots comparing catchment-scale (top) and adjacent-scale (bottom) pressure among biogeographic regions (left) and HGM types (right). Data are weighted such that estimates are for the USA wetland area represented by 2016 NWCA inland sites that yielded a water sample (hand-picked sites not included). Vertical dashed lines show trisection of pressure scores: CALT_SUM of 0 =minimal, <79 =low, 79 to 157 =medium, >157 =high; and CAT_PAGURB of <10 =minimal, 10 to 40 =low, 40 to 70 =medium, and >70 =high.

Further differences in anthropogenic pressure distributions are evident among combinations of geographic region and Riverine and Dep/Lac-closed sites (the two HGM x region categories with sufficient sample size to evaluate; per Table 2). Percentagewise, the largest adjacent-scale pressure category for all HGM x region combinations was “low”; most wetlands had non-zero adjacent-scale pressure but only a small percentage were estimated to have adjacent-scale pressure medium or high (Figure 3, left). Minimal adjacent-scale pressure occurred most often in ICP-Riverine, ICP-Dep/Lac-closed, and EMU-Dep/Lac-closed sites (~40% each), while adjacent pressure was most often medium or high in Dep/Lac-closed sites in the PLN (29%) and WMT region (46%). The pattern of catchment-scale pressure was bimodal (Figure 3, right). Minimal-pressure made up the largest estimated percentage in WMT-riverine sites, WMT-Dep/Lac-closed sites, and ICP-Dep/Lac-closed sites but medium or high pressure made up the largest percentage of EMU-riverine, PLN-riverine, and XER-Dep/Lac-closed sites. Notably, catchment-scale pressure was estimated to be high at 45% of WMT-Dep/Lac-closed sites and 62% of PLN-Riverine sites.

Figure 3:

Bar graphs showing distribution of anthropogenic pressure scores by biogeographic region for riverine wetlands (top) and Dep/Lac-closed wetlands (bottom) at the adjacent scale (left) and catchment scale (right). Data are weighted such that estimates are for the USA wetland area represented by 2016 NWCA inland sites that yielded a water sample (hand-picked sites not included). The largest segment in each stacked bar is labelled with its percentage.

Seven eastern-USA sites, all within the Appalachia coal-mining region, had field-crew comments indicating recent or historic mining impacts and SO₄⁼ was elevated above 200 mg/L at six of them (Figure 4). SO₄⁼ was also elevated in three sites in that region that lacked such comments (Figure 4) but for which mining impacts were supported by the river being on the ‘impaired waters’ 303(d) list for acid mine drainage (site with 227 mg/L SO₄⁼) or strip-mines visible a few kilometers away in Google Earth imagery (sites with 490 mg/L and 849 mg/L SO₄⁼). 200 mg/L SO₄⁼ translates to about 4200 in μeq/L units, well above the 1000 μeq/L level indicating strong mining impacts in rivers and streams (Herlihy et al. 1990) and the 300 μeq/L level used to differentiate mining runoff from atmospheric deposition (Baker et al. 1991). Five of the mining-impacted wetlands were Riverine while the other four were the Dep/Lac-closed HGM type (Figure 4). These mining impacts were not captured by the adjacent or catchment-scale anthropogenic pressure metrics which often scored minimal or low; out of the 9 sites, 3 had “minimal” and 4 had “low” adjacent-scale pressure, and 5 had “minimal” and 2 had “low” catchment-scale pressure.

Figure 4:

Comparison of sulfate levels among sites in the Appalachia coal-mining region with mining-related field-crew comments, in that region but without such comments, or outside that region but in the states containing that region (Indiana, Kentucky, Ohio, Pennsylvania, Tennessee, or West Virginia). Box-plots show median and 25^th to 75^th percentile range; circles show individual sites and are red for sites considered mining-impacted, letters indicate HGM type (R=riverine, DLC=Dep/Lac-closed).

General wetland water quality patterns

The 525 inland wetlands spanned a wide range of water quality. Total nitrogen (TN) and the TN:TP ratio spanned 3 orders of magnitude, conductivity (COND) and dissolved organic carbon (DOC) spanned 4 orders of magnitude, chloride (Cl-) and total phosphorus (TP) and planktonic chlorophyll-a (CHLA) and turbidity (TURB) spanned 5 orders of magnitude, and sulfate (SO₄⁼) spanned 6 orders of magnitude (Table 1). Below detection rates were <1% for all analytes except CHLA and SO₄⁼; Cl^- and SO₄⁼ are missing from one site, CHLA is missing from 2 sites, and DOC is missing from 14 sites (Table 1).

The anions Cl- and SO₄⁼ were positively intercorrelated (Figure 5, top row). SO₄⁼ levels were generally higher and rose more quickly relative to Cl- in the XER and PLN region than the WMT and EMU and ICP region, with higher correlations for XER than PLN due to tighter point clustering around the line. The lowest individual Cl- values occurred in the EMU and ICP region but median Cl- was lowest in the WMT region. Slope wetlands had the lowest median and smallest range in Cl- values and were the only HGM type where SO₄⁼ did not increase with increasing Cl- (flat-line relationship). Dep/Lac-open wetlands had the highest median SO₄⁼, the steepest increase in SO₄⁼ relative to Cl-, and the tightest clustering around the SO₄⁼ vs. Cl- line as indicated by the high correlation coefficient.

Figure 5:

Cross-analyte relationships for all inland wetlands for SO₄⁼ vs Cl^- (top row), TN vs TP (second row), CHLA vs TURB (third row), and DOC vs pH (bottom row). Left and right graphs are same data (grey polygons show extent) but with median values (circles) and regression lines colored by biogeographic region on left vs HGM type on right. Printed numbers are correlation coefficients; significant at Bonferroni-corrected p≤0.01 unless marked “(NS)”. Outlier at DOC=9.0 and pH=3.02 (“X” in bottom panels) omitted from polygons and fitted lines.

The nutrient metrics TN and TP were also positively correlated (Figure 5, second row). Median TN and TP were lowest in the WMT and highest in the PLN region, and lowest in the Slope and highest in the Dep/Lac-closed HGM type. The ICP region had a substantially flatter TN vs. TP regression line than the other 4 regions, with the tightest point cluster (highest correlation) in the PLN and EMU. Among HGM types, the tightest TN vs. TP point cluster was for Dep/Lac-closed and Dep/Lac-open wetlands; Flats had a flatter TN vs. TP regression line than the other 4 HGM types (correlation not significant). Across all sites, the TN:TP ratio was higher in the EMU and ICP (median 31 and 36) than in the PLN and WMT and XER region (medians ~17) and was higher in Flats (median = 56) than the other 4 HGM types (medians 20 to 25). The range in TN:TP ratios was primarily driven by TP, as the correlation to log10 TP was strongly negative (r=0.59), whereas the correlation to log10 TN was only weakly positive (r=0.28).

Planktonic CHLA was positively intercorrelated with TURB (Figure 5, third row), which measures suspended particles in general. The two western regions (WMT and XER) had lower median TURB and CHLA than the other 3 regions and CHLA made the least contribution to TURB in the ICP region (flattest line), but the difference among regions was not large. Dep/Lac-closed and Dep/Lac-open wetlands had an order of magnitude higher median CHLA values than the other three HGM types, consistent with being morphologically most lake-like and most likely to support planktonic algae. However, the five HGM types differed little in median TURB or in the fitted TURB vs. CHLA line (the Flats correlation was not significant but the slope was similar to the others). The tightest point cluster (highest correlation) around the TURB vs CHLA line was for the XER region and for Dep/Lac-open and Slope HGM types.

DOC and pH were negatively correlated (Figure 5, bottom row), but the correlation was generally weaker than for the above analyte pairs. The spread in DOC across all sites (grey polygon) was narrow at low pH but wide at high pH, and DOC values extended higher in the EMU and ICP than the other three regions. Fifty-five sites (10% of the total) had pH <6.0, all in the eastern USA and all but two in the EMU and ICP region which also had the lowest median pH among regions. One site with extremely low pH (3.02; “x” in Figure 5, bottom row) is an outlier for its magnitude (lowest pH in entire dataset), for its region (next lowest PLN site has pH=5.8), and for lack of concurrently high DOC. This might be a result of acid mine drainage; this is one of the nine mining-impacted sites (in Figure 4) but pH exceeded 6.0 in the other eight such sites. With or without that outlier, the PLN was the only region where the DOC vs. pH line was flat. The DOC vs pH lines in the WMT and XER were also not significant (Figure 5) but likely because of little range in either analyte. The only HGM group lacking sites with pH <6.0 was Dep/Lac-open. Median DOC was higher and the regression of DOC vs pH was less steep in Flats and Dep/Lac-closed wetlands than the other HGM types, with Flats having the highest correlation because of the least spread in DOC at high pH.

General water quality -- anthropogenic pressure associations

Box plots (Figure 6 and 7) and cross-analyte correlations (Table 4) make it is clear that both analyte concentrations and inter-analyte relationships tend to change in association with anthropogenic pressure. Low levels of pressure at either the adjacent or catchment scale were sufficient to elevate the distribution of COND and SO₄⁼ significantly above (to the right of) the minimal-pressure sites (Figure 6, top row). Medium/high levels of pressure were associated with further right-shifts in COND and SO₄⁼ distributions, with the magnitude greater under adjacent or combined adjacent and catchment pressure than under catchment pressure alone. Cl- had a different pattern, with low levels of pressure not associated with a right-shifted distribution and medium/high pressure increasing median Cl- (center of notch) most strongly at the catchment scale. Correlations indicate that SO₄⁼ became a much stronger contributor to COND at medium or high anthropogenic pressure than at minimal pressure (r increase from 0.39 to 0.74; Table 4). Mining was associated with dramatically elevated SO₄⁼ and COND but did not have an outsized effect on Cl- (Figure 6).

Figure 6:

Notched box plots comparing water quality distributions for sites with anthropogenic pressure classified as minimal at both scales (N=92, green), no more than low (N=259, tan), medium/high adjacent but lower catchment scale (N=22, grey), medium/high catchment but lower adjacent scale (N=116, blue), medium/high both scales (N=27, black), or mining in conjunction with any adjacent or catchment-scale score (N=9, red). To aid in seeing significant differences (non-overlapping notches), vertical lines mark the right-hand notch edge for minimal-pressure sites (left-hand edge for DOC).

Figure 7:

Notched box plots comparing water quality distributions for sites with anthropogenic pressure classified as minimal at both scales (N=92, green), no more than low (N=259, tan), medium/high adjacent but lower catchment scale (N=22, grey), medium/high catchment but lower adjacent scale (N=116, blue), medium/high both scales (N=27, black), or mining in conjunction with any adjacent or catchment-scale score (N=9, red). To aid in seeing significant differences, vertical lines mark the right-hand notch edge for minimal-pressure sites; for TN:TP vertical lines instead mark thresholds for nitrogen limitation (<20, 1.3 in log10 units) and phosphorus limitation (>50, 1.7 in log10 units).

Table 4.

Pairwise pearson correlations among water quality analytes, all of them log10 transformed except pH (see Table 1 for analyte names and units). Top block is sites with minimal adjacent and catchment-scale pressure (N=88 to 92, depending on analyte pairing), bottom block is sites having medium/high pressure at one or both of these scales (N=163 to 175); sites with identified mining pressure are excluded from both sets of correlations. Bold font indicates correlations ≥0.5 (all having Bonferroni-corrected p-values <0.001), and correlation coefficients that change by ≥0.2 units between the two blocks are followed by upward or downward arrows.

	log10 COND	log10 Cl-	log10 SO4=	log10 TN	log10 TP	log10 CHLA	log10 TURB	log10 DOC
a) pressure minimal
log10 Cl-	0.53
log10 SO4=	0.39↑	0.49
log10 TN	0.03	0.02	-0.34↓
log10 TP	0.11	-0.25	-0.23↓	0.53
log10 CHLA	0.07	0.05	-0.05	0.55↓	0.47
log10 TURB	0.05	0.07	-0.04	0.62↓	0.57	0.65
log10 DOC	0.13	0.09	-0.36↓	0.77↓	0.36	0.49	0.54↓
pH	0.43	-0.01	0.25↑	-0.34↓	0.18	-0.15	-0.11	-0.46↓
b) pressure medium or high
log10 Cl-	0.64
log10 SO4=	0.74↓	0.44
log10 TN	0.19	0.18	-0.00↑
log10 TP	0.15	0.07	-0.01↑	0.64
log10 CHLA	-0.03	0.03	-0.10	0.30↑	0.42
log10 TURB	-0.13	0.01	-0.13	0.32↑	0.42	0.51
log10 DOC	0.15	0.16	-0. 04↑	0.30↑	0.47	0.34	0.18↑
pH	0.55	0.07	0.45↓	0.02↑	-0.01	-0.06	-0.11	-0.15↑

Anthropogenic pressure was generally associated with DOC distribution shifts to lower values but pH distribution shifts to higher values (Figure 6, bottom row). The negative correlation between DOC and pH in minimal pressure sites became substantially weaker under medium/high pressures (r= −0.46 vs −0.15, Table 4). This appears to result from sites with pH<6.5 not occuring in combination with medium or high pressure at the adjacent scale or both scales combined (compare presence vs absence of left-hand tail in Figure 6 boxplots). Also reflecting this right-shift of pH under medium/high adjacent pressure, the correlation of pH to SO₄⁼ became much stronger than at no-pressure sites (Table 4). Mining was the only anthropogenic pressure that did not shift median pH upward relative to its minimal-pressure value but neither did it shift pH downward (Figure 6). With minimal anthropogenic pressure, DOC was also positively correlated with TN and TURB, but this correlation became much weaker with medium or high pressure (Table 4).

Low levels of anthropogenic pressure were not enough to elevate the nutrients TN and TP, and at medium or high pressure it was the catchment-scale that was most clearly associated with elevated nutrients (Figure 7, top row). However, low levels of either catchment or adjacent scale pressure were enough to shift TN:TP ratio distributions leftward towards N-limitation (ratios <20), and cases of P-limitation (ratios >50) only occurred within the central box (the 25^th to 75^th percentile range) for sites having minimal anthropogenic pressure. Unexpectedly, there was no indication that CHLA or TURB distributions shifted upwards with medium or high catchment-scale pressure (Figure 7, bottom row). This apparent disconnect between nutrient increases and algae/particulate responses is confirmed by the correlation analysis, which showed that TN, TP, CHLA, and TURB were positively intercorrelated in sites with minimal anthropogenic pressure but that correlation weakened under medium/high pressure (Table 4). Mining was associated with substantially elevated TURB (Figure 7, bottom row) but not CHLA, so we infer that TURB at mining-impacted sites reflects suspended sediments rather than phytoplankton.

Specific water quality responses by wetland class and anthropogenic pressure type

The pattern for pH was strongly geographically organized. Sites with circumneutral pH (6 to 8) occurred across the continent, but acidic sites (pH<6) were found primarily in the EMU and ICP regions along the Atlantic and Gulf coasts and northern Great Lakes, whereas alkaline sites (pH>8) were found primarily away from coasts in the PLN and WMT and XER region (Figure 8, top). On a percentage basis, Slopes and Flats were most likely to have pH<6 and least likely to have pH>8, whereas Riverine and Dep/Lac-open sites were least likely to have pH<6 and most likely to have pH>8 (Figure 8, bottom). The actual number of sites with pH<6 was highest in Dep/Lac-closed sites (N=21) because of large sample size in the EMU and ICP region, with Flats ranked second for number of pH<6 sites (N=18). The pattern for pH was not strongly organized by anthropogenic pressure. In the PLN and WMT and XER, there was no trend in pH in association with either adjacent or catchment pressure. In the EMU and ICP, most cases of pH<6 occurred with <20% agriculture + urban landuse, but adjacent-scale pressure vs pH could not be analyzed because high-pressure cases were lacking.

Figure 8:

Patterns in wetland pH depicted geographically across all sites (top) and as distributions in HGM categories for sites in EMU/ICP vs. PLN/WMT/XER geographic regions (bottom). Site numbers are printed at the top of the bars.

Cl- patterns in association with catchment-scale pressure were weak across the dataset as a whole but strong within road-salt using states, where ~90% of sites had Cl- <10 mg/L when catchment pressure was minimal and declined to <40% when catchment pressure was high (Figure 9, left). By HGM type, the linear correlation of Cl- to catchment pressure was substantial for Dep/Lac-closed and Riverine and Dep/Lac-open wetlands (r= 0.59, 0.54, and 0.48 respectively) but not in Flats and Slope wetlands (r<0.2) which had very few sites with high catchment-scale pressure to evaluate (Figure 2, upper right). Cl- levels were not strongly associated with adjacent-scale landuse; there was an increase in the percentage of sites having >100 mg/L Cl- across CALT_SAL pressure categories but no consistent trend in the other Cl- logarithmic categories (Figure 9, right).

Figure 9:

Association of wetland Cl- to anthropogenic pressure in states classified as road-salt using. Left panel shows catchment-scale pressure when adjacent pressure is minimal or low (N=263); right panel shows adjacent-scale pressure when catchment pressure is minimal or low (N=245). Cl- is depicted in logarithmic categories.

TP levels were clearly associated with catchment-scale pressure and not with adjacent-scale pressure. Across sites with minimal or low adjacent-scale pressure, the percentage of sites with TP <100 ug/L was ~60% when catchment pressure was minimal and declined below 20% when catchment pressure was high (Figure 10, left). By HGM type, the association of TP to catchment pressure was steeper in Dep/Lac-closed and Riverine and Dep/Lac-open wetlands (r= 0.46, 0.37, and 0.38 respectively) than in Slope and Flats wetlands (r=0.24 and 0.19). Across sites with minimal or low catchment-scale pressure, there was no trend in TP distributions in association with adjacent pressure (Figure 10, right). The primary geographic pattern in TP with minimal anthropogenic pressure was higher values in the PLN (median 146 μg/L) than the other four regions (median 48 to 88 μg/L).

Figure 10:

Association of wetland TP to anthropogenic pressure for all sites. Left panel shows catchment-scale pressure when adjacent pressure is minimal or low (N=422); right panel shows adjacent-scale pressure when catchment pressure is minimal or low (N=393). TP is depicted in logarithmic categories.

Discussion:

The National Wetland Condition Assessment samples numerous wetlands of multiple hydrogeomorphic types across the continental USA, thereby enabling a large-scale picture of wetland conditions and relationships to anthropogenic pressures to be formed. The 2016 NWCA wetland water quality data set is, to our knowledge, the most extensive ever collected for the USA. More extensive than the 2011 NWCA survey because of additional analytes (DOC, TURB, Cl^-, SO₄⁼), expanded sampling frame (adding National Wetland Inventory to US FWS Status and Trends for sampling frame), and notably better water quality data coverage in the southern EMU region, western and southern PLN region, and the pacific-northwest. And much broader spatially than the only other large-scale USA wetland survey we are aware of, the ongoing Great Lakes coastal wetland survey which recently reported on water quality of 500+ sites (Gentine et al. 2022; Harrison et al. 2020). However, the breadth of wetland types and geographic locations covered by the 2016 NWCA, combined with the logistical constraints that such an extensive survey raise, means that the general patterns that emerge are underlain by considerable among-site variability (Herlihy et al. 2019). Both of these statements – the broad generality of patterns and the underlying variability – are major themes in our discussion.

Anthropogenic pressure types and effects:

We examined presumptive anthropogenic pressures at two contrasting spatial scales: physical disturbances directly within or in close proximity to the wetland, and landuse in the surrounding upland likely to load nutrients, sediments, and salts into water that eventually reaches the wetlands. Which of these spatial scales of pressure yield observable water quality responses depended on the analyte; for example, ionic-strength metrics (Cl-, SO₄⁼, COND) were elevated in association with both catchment and adjacent-scale pressure, whereas nutrients (TN and TP) were elevated primarily in association with catchment scale pressure. These differences may reflect the biochemical processing ability of the wetlands (e.g., plants and bacteria take up nutrients but not chloride) or varying adjacent vs. catchment delivery methods, but likely also reflect the differing prevalence of observed types of pressure. Our analyses focus on partitioning the two spatial scales of pressure, but in fact they can interact. For example, a protective buffer may be most important for wetland biota when landscape urbanization is high (e.g., Houlahan and Findlay 2004, Trebitz et al. 2009). Regionalization was not necessary to extract general associations to anthropogenic pressure but was helpful in understanding specific patterns, as evident with pH where acidic sites occurred in the eastern USA only, with sulfate where many (although not all) elevated cases were associated with the Appalachia coal-mining region, and with chloride where associations to landuse were most evident in northern-tier road-salt using states.

Variation in the strength of the relationship among wetland water chemistry analytes and to anthropogenic pressure arise geographically but also due to characteristics of the wetland and terrestrial landscape beyond those captured by HGM types and NLCD-based pressure metrics. On the landscape side, factors contributing to variability include watershed size, surficial geology, and intensity and distance away of the landuse (e.g., Craft and Casey 2000, Dillon and Molot 1997; Thornton and Elledge 2021). Also, while agricultural and urban landuse categories are frequently combined they are not entirely interchangeable, as illustrated by different effects on soil and plant tissue nutrient levels in Virginia riparian wetlands (Hogan and Walbridge 2009) and in water chemistry ordinations of Prarie Pothole ponds (Wanker et al. 2012). Our data hint at this difference also, as in preliminary analyses where we separated the two categories of catchment-scale pressure, chloride had a stronger relationship to urban landuse, whereas TP had a stronger relationship to agricultural landuse. Data to further resolve the role of such landscape factors are available in various geospatial datasets but beyond the scope of our current analyses. On the wetland side, contributing factors include varying inundation cycles and biochemical processing (e.g., Heagle et al. 2013, Johnston 2001, Vanderhoof et al. 2017) that are measurable in site-specific studies but beyond the capacity of a synoptic survey such as the NWCA. In this regard, the predictability of wetland water chemistry is analogous to the predictability of lake water chemistry, where clear patterns emerge from general catchment-scale predictors but lake-specific characteristics that are more difficult to measure are the key to resolving further variability (Read et al. 2015).

Given the recognition that wetlands worldwide have been extensively destroyed and heavily anthropogenically impacted (see introduction) we were surprised by how many of the 2016 NWCA sites were scored as having minimal pressure at one or both spatial scales. Our standards for this scoring seem stringent enough, requiring absence of physical alteration checkmarks and >90% of catchment in natural landuse categories which is empirically supported (Online supplement Figure 1) and exceeds the Environment Canada guidelines (summarized in Croft-White et al. 2017) of <10% urbanized and at least 40% forested and wetland area. Factors that might bias the observed wetland pressure rates downward include that anthropogenic actions often eliminate wetlands entirely so that they are no longer sampleable whereas even heavily impacted lakes and streams tend to physically still exist; that NWCA does not sample wetlands in use for crops or aquaculture which might keep heavily disturbed sites from the dataset; and that NWCA checklist protocols would miss adjacent pressures if the 100m ring around the AA falls entirely within wetland rather than upland areas. Anthropogenic pressure may in fact be less prevalent on wetland shores than on stream and lake shores simply because wetlands lie in saturated sediments that are less conducive to human development. Likewise, catchment scale agriculture and urbanization may be less likely in wetland-rich areas due to a paucity of higher-elevation land, as suggested by our finding that low pH – high DOC wetlands do not spatially overlap with high catchment-scale pressure.

On ionic strength metrics

Treatment of roadways with salt to prevent winter ice buildup is recognized as a major source of chloride to lakes and streams (Dugan et al. 2017; Hintz et al. 2022); Cl- is not naturally present in any significant quantities in inland waters (Hem 1985). Our results suggest that Cl- runoff also impacts a broad swath of wetlands across the northern USA and that Cl- remains elevated well into summer (the NWCA sampling season), probably due to a combination of rainfall re-mobilizing salts deposited onto roadways and drains and Cl- being biologically inactive and thus not readily removed (Corsi et al. 2015, Helmuller et al. 2019). Even larger Cl- values than our NWCA data (exceeding 700 mg/L) have been found in urban wetlands by Hill and Sadowski (2016). The increased ionic strength of waters resulting from Cl- inputs can negatively influence other aspects of wetland water chemistry such as decreasing phosphorus sorption and nitrogen removal capacity and increasing the potential for sulfate to be reduced to more toxic sulfide (reviewed in Herbert et al. 2015).

Chloride is one of the few NWCA water chemistry analytes having toxicity-based regulatory standards in the USA (860 mg/L acute, 230 mg/L chronic; USEPA 1988). The chronic standard corresponds to a level affecting 10% of tested laboratory species (Evans and Frick 2001) but some wetland-dependent species may be much more sensitive; for example, wood frogs are absent from Canadian plains wetlands at Cl- above 40 mg/L (Donald 2021). The 2016 NWCA data had no sites above this acute threshold but eleven above the chronic threshold. Three of them probably receive salt runoff from adjacent highways (sites near Chicago IL, Grand Rapids MI, and Milwaukee WI) and one site in North Carolina probably receives seawater influence from an estuary ~1km away; the cause for elevated Cl- at the other seven sites is not so obvious (all in rural terrain) although one in Montana looks like a natural salt flat from aerial imagery and field-crew photographs.

Besides acid mine runoff – which is implicated for sulfate elevation at several of the 2016 NWCA sites -- the primary anthropogenic source of sulfate to the environment is thought to be fossil-fuel-related pollution especially near busy roadways and heavy industry (Chapra et al. 2012). Sulfate is also a component of some fertilizers (Powlson and Dawson 2021) but is not normally a component of the crystals or brine used to salt roads. Elevated sulfate does occur naturally in some wetland-rich regions of North America (e.g., prairie plains; LeBaugh 1989); anthropogenic pressure was minimal for the Montana prairie site with the single highest sulfate concentration in the data set (and also very high Cl-, per previous paragraph).

Both Cl- and SO₄⁼ contribute to electrical conductivity (COND). One state, Nebraska, has implemented a COND standard that applies to wetlands, namely <2000 μS/cm for wetlands feeding an agricultural water supply. The 2016 NWCA data set has fourteen sites exceeding 2000 μS/cm COND, twelve of them in PLN-region states but none actually in Nebraska, and one each in Grand Rapids MI and rural Arizona already noted for elevated Cl-. We might have expected high COND at some sites due to high CSTD_SAL and substantial catchment-scale pressure but five sites with high COND had very low adjacent pressure (score <5) and only modest agriculture and urban landuse levels (<35%). Southwestern desert wetlands and wetlands in the Great Plains may have significant natural causes of elevated COND due to high-salt groundwater inputs and high evaporation rates that serve to concentrate them (Williams 1999, Piña & Lougheed 2022).

Wetland nutrient response patterns

Consistent with many non-wetland studies, our data show a strong correlation between TN and TP and a notable increase in both when catchment-scale pressure is high. Our data also show a shift in nutrient ratios away from P-limitation and towards N-limitation with increasing anthropogenic pressure. The strong positive relationship between wetland CHLA and TURB confirms that planktonic algae are a substantial component of particulates in wetland water and the likewise positive relationship between CHLA and both TN and TP in minimal-pressure wetlands confirms that planktonic algae increase as ‘background’ nutrient levels increase. So why in the 2016 NWCA dataset do CHLA and TURB not track the nutrient increase that comes with increasing landscape disturbance, as evidenced by declining strength of correlation to TN and TP (Table 4) and lack of shifts in the CHLA and TURB distribution (Figure 7)? Great Lakes coastal wetlands are the wetland type most studied for nutrient responses, and CHLA and TURB clearly do increase in concert with increased nutrients there (e.g., Crosbie and Chow-Fraser 1999; Gentine et al. 2022; Lougheed et al. 2001; Morrice et al. 2008). Cheruvelil et al. (2022) speak of the limited understanding of nutrient responses in very shallow lakes (maximum depth <5m) and find the CHLA vs TP relationship to be less steep in shallow than non-shallow lakes, confirming our finding in wetlands of even more muted response of CHLA to nutrients. We can speculate that across the breadth of NWCA sites, nutrients are generally channeled to vascular plants and periphyton and thus not reflected in planktonic CHLA. Shading by vascular plants or DOC-stained waters would also impair the ability of phytoplankton to respond to nutrient loads in many wetlands (Myrstener et al. 2022).

Given the well-recognized value of wetlands for intercepting nutrients and thereby ameliorating conditions further downstream (e.g., Golden et al. 2019, Cheng et al. 2020), there is a dearth of literature on wetland nutrient response to watershed pressure (beyond the already mentioned Great Lakes coastal wetlands) and water quality criteria for nutrients in most USA states do not encompass wetlands. The situation is quite different in lakes and streams where nutrient responses are widely published on and nutrient standards are widely promulgated. The only states having wetland-applicable nutrient standards are Minnesota and Nebraska (per state-by-state standards summarized at the USEPA Wetland Water Quality Standards Resources website). In Minnesota, the highest “shallow lake” TP criteria (for the southwestern region) is 90 μg/L; 78% of the NWCA 2016 wetlands sampled across Minnesota exceeded this level. In Nebraska, the nutrient criteria for eastern waters including wetlands are TP<50 μg/L and TN<1000 μg/L while western waters criteria are TP<40 μg/L and TN<800 μg/L; six of the seven NWCA wetlands sampled in Nebraska exceeded 50 μg/L TP and five of them exceed 1000 μg/L TN.

On pH and DOC

Water pH is a major factor in structuring wetland biogeochemistry and vegetation (Mitsch and Gosselink 2015) and thus the distribution of pH across NWCA sites is of interest. None of the sites with acidic waters (pH<6) exhibited high adjacent or high catchment pressure. Yet it seems unlikely that anthropogenic impacts have physically destroyed all acidic wetlands or sufficiently raised their pH that water ceases to be acidic, and all of the NWCA sites with pH<5.5 (excepting the previously discussed outlier) had DOC over 10 mg/L suggesting the low pH results from natural organic acids. Which leaves the conclusion that places where the low pH high DOC wetlands occur are places unlikely to have high anthropogenic pressures. Is the geographic realm of the two different because boggy areas are unattractive for agriculture and human settlement? Geographically, the high DOC low pH wetlands in the southeastern USA are the same region having tannin-stained ‘blackwater’ rivers (Flotemersch 2022), and the ones in the northeast USA and upper Great Lakes coincide with areas of naturally acidic lakes and streams (Baker et al. 1991). Many states do have pH standards (typically between 6.5 and 9.0 unless “naturally outside that”) but they are not usually wetland specific. North Carolina allows “swamp waters” to have pH as low as 4.3 if “the result of natural conditions” (per USEPA Wetland Water Quality Standards Resources website); two 2016 Flats sites in North Carolina had pH even lower (pH=3.3 and 4.1) with high DOC (> 45 mg/L) and minimal anthropogenic pressure, suggesting that the allowable pH for natural could be even lower.

No evidence for wetland ‘isolation’

The topic of wetland connectivity has featured prominently in wetland science and policy discussion recently (e.g., Calhoun et al. 2017, Fritz et al. 2018), motivated by the notion that some wetlands might not need protection from adverse landuse because of being isolated from larger hydrologic networks. We focused on HGM types in our analyses and further split depressions and lacustrine sites into ‘closed’ vs. ‘open’ specifically to assess whether water source organizes the pattern of association of water quality to adjacent or catchment pressure. The two HGM types most directly connected to the landscape via stream inputs are Riverine and Dep/Lac-open. Flats wetlands also receive surface runoff but episodically and over a smaller spatial scale, not via permanent stream channels. Slope wetlands are dominated by shallow groundwater inputs, which can vertically exchange with surface stream networks (Fritz et al. 2018). Dep/Lac-closed wetlands have both precipitation inputs and groundwater connections, the latter varying with topographic position such that wetlands high in the landscape tend to recharge to groundwater whereas wetlands low in the landscape tend to receive discharge from groundwater (e.g., Yuan et al. 2021). For all three of the HGM types without inflowing streams, our examination of water chemistry vs StreamCat-mapped (Hill et al. 2015) catchment-scale anthropogenic pressure amounts to asking whether that association makes sense.

The 2016 NWCA dataset does reveal some differences in water chemistry patterns among the HGM types. For example, acidic conditions (pH<6) were most common percentagewise in HGM types lacking surface inflows (Flats, Slope, Dep/Lac-closed), and TN:TP ratios indicative of P-limitation (>50 per Guildford and Hecky 2000) were observed only in Flats consistent with their water source being direct precipitation (generally low in phosphorus). However, the dataset does not support any differentiation among HGM types in the degree to which water chemistry was associated to catchment-scale pressure. For example, mining-associated SO₄⁼ elevation occurred in Dep/Lac-closed as well as Riverine sites, Cl- increased just as steeply in association with catchment-scale pressure in Dep/Lac-closed sites as in Riverine and Dep/Lac-open sites, and the correlation of TP to catchment-scale pressure was actually steeper in Slope and Dep/Lac-closed sites than Riverine and Dep/Lac-open sites. In short, all of the wetland HGM types that the 2016 NWCA sampled exhibited water chemistry associations to anthropogenic pressure at the catchment scale; they were not isolated from the surrounding upland even in the absence of inflowing streams. Which is consistent with the few published studies that have examined ‘isolated’ wetland water chemistry for sensitivity to landuse: Craft and Casey (2000) found no inherent difference between floodplain (i.e., Riverine) and depression (i.e., Dep/Lac-closed) wetlands in carbon and nutrient accumulation in response to watershed loading, and Whigham and Jordan (2003) concluded that the variable water chemistry of Slope and Flats wetlands was explained by a combination of water source and watershed landuse.

Summary:

Data from the 2016 National Wetland Condition Assessment, spanning more than 500 inland wetlands, provide the basis for examining wetland water quality patterns on an ecoregion and HGM type basis at a conterminous-USA scale. Our analyses found associations between water quality and anthropogenic pressure characterized at the adjacent scale (from field checklists) and the catchment scale (from NLCD landuse classification), thereby demonstrating water quality patterns consistent with those known from much more extensively studied lakes and streams. Chloride was elevated in association with both adjacent and catchment-scale pressure (particularly in road-salt using states) with some sites exceeding the chloride EPA chronic standard, and sulfate was elevated in association with mining at a small number of sites. Nutrients were elevated and nutrient ratios shifted towards N-limitation with increasingly agricultural and urban landuse but a concomitant chlorophyll-a response was not evident, suggesting that nutrients were not generally channeled towards planktonic algae. Sites having acidic pH and high DOC did not co-occur with high adjacent or catchment-scale pressure, which we think reflects unsuitability of boggy regions for human development rather than anthropogenic pressure directly changing wetland pH and DOC. We found no evidence to support the notion that wetland HGM types lacking surface stream inputs (i.e., Flats, Slopes, Dep/Lac-closed) were more isolated from landuse pressures than were wetlands of the Riverine and Dep/Lac-open HGM type. These general patterns in the NWCA 2016 dataset were accompanied by substantial among-wetland variability; more detailed biological status and anthropogenic impacts data will likely be necessary to delve into causes for elevated water chemistry analytes on a regional and site-by-site basis.

Data Availability:

Data generated by the NWCA are publicly available at the US EPA National Aquatic Resources Surveys website, https://www.epa.gov/waterdata/national-aquatic-resource-surveys-nars.