Stable isotopes of algae and macroinvertebrates in streams respond to watershed urbanization, inform management goals, and indicate food web relationships

Abstract

Watershed development and anthropogenic sources of nitrogen are among leading causes of negative impacts to aquatic ecosystems around the world. The δ¹⁵N of aquatic biota can be used as indicators of anthropogenic sources of nitrogen enriched in ¹⁵N, but this mostly has been done at small spatial extents or to document effects of point sources. In this study, we sampled 77 sites along a forest to urban land cover gradient to examine food webs and the use of δ¹⁵N of periphyton and macroinvertebrate functional feeding groups (FFGs) as indicators of watershed development and nitrogen effects on streams. Functional feeding groups had low δ¹⁵N variability among taxa within sites. Mean absolute differences between individual taxa and their respective site FFG means were < 0.55‰, whereas site means of δ¹⁵N of FFGs had ranges of approximately 7–12‰ among sites. The δ¹⁵N of periphyton and macroinvertebrate FFGs distinguished least disturbed streams from those with greater watershed urbanization, and they were strongly correlated with increasing nitrogen concentrations and watershed impervious cover. Nonmetric multidimensional scaling, using δ¹⁵N of taxa, showed that changes in macroinvertebrate assemblages as a whole were associated with forest-to-urban and increasing nitrogen gradients. Assuming an average +3.4‰ per trophic level increase, δ¹⁵N of biota indicated that detrital pathways likely were important to food web structure, even in streams with highly developed watersheds. We used periphyton and macroinvertebrate FFG δ¹⁵N to identify possible management goals that can inform decisions affecting nutrients and watershed land use. Overall, the δ¹⁵N of periphyton and macroinvertebrates were strong indicators of watershed urban development effects on stream ecosystems, and thus, also could make them useful for quantifying the effectiveness of nitrogen, stream, and watershed management efforts.

1. Introduction

Anthropogenic sources of nutrients continue to be among leading causes of negative impacts to stream ecosystems in the United States and in many parts of the world (Carpenter et al., 1998; Vörösmarty et al., 2010; Stoddard et al., 2016). In developed areas, human activities, growing populations, and impervious cover in watersheds increase nitrogen loads to downstream water bodies from wastewater, sewers, septic systems, fertilizers, and stormwater runoff (Kaushal et al., 2011; Kaushal et al., 2014; Pennino et al., 2016). Urbanization alters flow paths, hydrology, and geomorphology, and decreases vegetation, which further changes the transport and cycling of nutrients in watersheds and streams (Walsh et al., 2005; Wollheim et al., 2005; Kaushal and Belt, 2012). These changes in nutrient dynamics, habitat, and hydrology affect biogeochemistry, rates of primary production, and food webs in streams (Meyer et al., 2005; Miltner, 2010; Klose et al., 2012; Kautza and Sullivan, 2016), which can have negative downstream consequences for recreational opportunities, tourism, and property values, along with possible increased health risks (Dodds et al., 2009; Sobota et al., 2015). Nutrient pollution in watersheds also creates problems for estuarine and coastal ecosystems that are commonly nitrogen limited (Howarth et al., 2002; Seitzinger et al., 2005; Boyer et al., 2006). As a result, developing indicators based on ecological responses to nutrients and watershed land cover can assist with setting effects-based goals for stream ecosystem protection, management, and restoration, and with informing future decisions affecting land use and aquatic resources.

Given that human wastewater is enriched with the ¹⁵N isotope and is an abundant source of nitrogen in urban and suburban watersheds (Heaton, 1986; Castro et al., 2003; Carey et al., 2013), measuring nitrogen stable isotope ratios (δ¹⁵N) in organisms has become a useful approach to identifying effects of different nitrogen sources on aquatic ecosystems (McClelland et al., 1997; Lake et al., 2001; Vander Zanden et al., 2005). In addition to point source discharges of wastewater, septic systems and leaking sewers are substantial sources of nitrogen entering streams (Groffman et al., 2004; Steffy and Kilham, 2004). Atmospheric deposition and inorganic lawn fertilizers could lower δ¹⁵N in systems affected by wastewater because they typically range from 2–8‰ and −3 to 3‰, respectively, whereas human and animal wastes are typically >10‰ (Heaton, 1986; Valiela et al., 2000). However, wastewater sources often dominate nitrogen loads in developed watersheds (Carey et al., 2013; Divers et al., 2014), and microbial denitrification, which benefits water quality by removing N, has preferential uptake of ¹⁴N and increases δ¹⁵N in downstream waters affected by inorganic fertilizers (Kellman and Hillaire-Marcel, 1998; Vander Zanden et al., 2005). Some evidence even suggests that nitrogen inputs to suburban streams from pet waste can exceed contributions from lawn fertilizers (Groffman et al., 2004; Carey et al., 2013). Despite differences in δ¹⁵N from multiple sources, δ¹⁵N in algae, macroinvertebrates, and fish generally increase in response to human and animal sources of nitrogen (Ulseth and Hershey, 2005; di Lascio et al., 2013; Hicks et al., 2017).

Examining stable isotopes of multiple trophic levels is useful because urbanization affects basal resources and alters the biomass and abundance of macroinvertebrate functional feeding groups in streams, such as shredders, scrapers, collectors, and predators (Stepenuck et al., 2002; Sterling et al., 2016; García et al., 2017). Measuring δ¹⁵N in aquatic organisms among different trophic levels (e.g., primary producers, consumers, and predators) provides information regarding how food webs change, because of trophic fractionation in biota (Fry, 1991; Vander Zanden and Fetzer, 2007; Layman et al., 2012). The δ¹⁵N incorporated into the bodies of aquatic organisms reflects the isotope ratio of their nitrogen source or diet, with an average enrichment of +3.4‰ (typical range of 2–4‰) per higher trophic position (Post, 2002). The δ¹³C of consumers and higher trophic levels, which has little to no fractionation (<0.5‰), can provide additional insight into basal resources, energy transfer, and sources of carbon in food webs (France, 1995; Post, 2002; García et al., 2017). Stable isotopes of biota also integrate the exposure to and effects of urbanization and nitrogen from human sources reaching streams over time. Temporal variability in δ¹⁵N of macroinvertebrates and primary producers still exists due to their internal turnover of nutrients, with long-lived organisms and higher trophic positions typically having less variability (Cabana and Rasmussen, 1996; Post, 2002; Jardine et al., 2014). However, within season variation is relatively low, whereas among season variation can be greater (Peipoch et al., 2012; Woodland et al., 2012; Pastor et al., 2014).

Studies of biota δ¹⁵N in streams with developed watersheds mostly have focused on a small number of sites (typically < 20) with limited spatial extents (<100 km² watersheds) or on direct upstream-to-downstream effects of wastewater treatment plants (e.g., Morrissey et al., 2013; Hicks et al., 2017). In this study, we used a spatially-balanced sampling design in a >4400 km² coastal watershed, which targeted stream sites along a forest to urban land cover gradient, to determine if δ¹⁵N of periphyton and macroinvertebrate consumers and predators were effective indicators of nitrogen and urban effects on stream ecosystems. We (1) documented within and among site variability of δ¹⁵N for functional feeding groups, (2) identified responses of biota δ¹⁵N to watershed urbanization and nitrogen concentrations, (3) used results to inform possible nitrogen management goals, and (4) examined if watershed urbanization affected food webs.

2. Materials and methods

2.1. Site selection and sampling

Wastewaters from point sources, sewers, and septic systems associated with watershed development and high population density are the leading contributors of nitrogen in the Narragansett Bay watershed (4421 km²), which is located in northeastern USA, has 34.5% developed land cover, and 380 people km⁻² (Castro et al., 2003; US EPA, 2007). We used ArcGIS 9.3 (Environmental Systems Research Institute, Redlands, California) and the 2006 National Land Cover Database (NLCD) to randomly select 105 possible 2nd to 4th order stream sites within categories of watershed impervious cover (<1%, 1–5.5%, 5.5%–10%, 10–20%, 20–30%, and >30%). For each of these categories, we used GIS to overlay a tessellated hexagon grid on the watershed and randomly selected one site within each hexagon. This survey design increased the likelihood of sampling sites along a continuous gradient of watershed development, while ensuring a representative spatial distribution of sites within each impervious cover category throughout the watershed (Smucker et al. 2016). We used NHDPlus Basin Delineator Software (www.horizon-systems.com) to delineate watersheds, which were checked for accuracy using US Geological Survey (USGS) 7.5-min quadrangles (1:24,000). We used NLCD 2006 for site selection because of available GIS tools that facilitate quick estimates of land cover, but once sites were selected we used photo-interpreted aerial imagery to quantify impervious cover and land cover in upstream watersheds for each site. Rhode Island land cover used 2003–2004 aerial imagery with 0.6 m resolution, and Massachusetts land cover used 2005 aerial imagery with 0.5 m resolution (RIGIS www.edc.uri.edu/rigis; MassGIS www.mass.gov/anf/research-and-tech/it-serv-and-support/application-serv/office-of-geographic-information-massgis).

We sampled 77 sites with watersheds <200 km² between July and October 2012 (Fig. 1). Other sites were not sampled because they were dry, too deep, or presented safety concerns. We collected water (77 sites), periphyton (75 sites), and macroinvertebrate (69 sites) samples during base flow conditions. Following rain events, we waited until nearby United States Geological Survey stream gauges indicated a return to normal flows so effects of atypical stormflow chemistry and hydrology were minimized. Samples for chemical analyses were collected in 1-L Nalgene^® bottles from the water column at each site. Six stones evenly dispersed throughout a 50 m stream reach encompassing a riffle were collected in a zigzag pattern, and periphyton was removed from their surfaces using a firm bristled brush. Benthic macroinvertebrates were qualitatively collected using kick nets, removal from woody debris and leaves, and from scraping substrata within the same reach. Water, periphyton, and macroinvertebrate samples were stored in the dark on ice until returning to the lab where they were processed within 24 h.

Fig. 1.

Map showing the distribution of 77 sample sites in the Narragansett Bay watershed. Urban, suburban, and natural (forests and wetlands) land cover are shown. MA = Massachusetts, RI = Rhode Island, CT = Connecticut, NLCD = National Land Cover Database.

2.2. Lab procedures

Filtered water (0.45 μm pore size membranes) was used for measuring PO₄–P, NO₃–N (as NO₃+NO₂–N), NH₄–N, dissolved organic carbon (DOC), SO₄²⁻, Cl⁻, Ca²⁺, Mg²⁺, Na⁺, and K⁺ concentrations, and unfiltered water was used for measuring total P and total N concentrations. All analyses followed EPA approved protocols (US EPA, 2017). Unfiltered nutrient samples underwent persulfate digestion, and all nutrient samples were analyzed using a Lachat flow-injection analyzer (Lachat Instruments, Milwaukee, WI). TN and NO₃–N were measured using the cadmium reduction method and NH₄–N was measured using the phenolate method. Given the objectives of this study, we focused on nitrogen results. Details on other chemistry results can be found in Smucker et al. (2016).

Periphyton was homogenized and a 50 ml subsample was concentrated into a pellet using a centrifuge and then dried, weighed, and analyzed for δ¹⁵N and δ¹³C stable isotope ratios. Macroinvertebrates were sorted to their lowest taxonomic level, highest being family, dried, pulverized, and analyzed for δ¹⁵N and δ¹³C stable isotope ratios. Stable isotope ratios were determined using a continuous flow isotope ratio mass spectrometer (Isoprime 100 Mass Spectrometer, Elementar Americas, Mt. Laurel, NJ, USA) and reported as per mil differences (‰) between samples and reference materials (δ¹⁵N and δ¹³C). Atmospheric N₂ and Vienna Pee Dee Belemnite were reference standards for nitrogen and carbon stable isotopes, respectively.

2.3. Data analysis

To summarize the variability of δ¹⁵N in biota, we reported means and ranges of periphyton and macroinvertebrate functional feeding groups (FFGs) among sites. We used the term detritivore to be more inclusive and representative of the dietary source, rather than just shredder taxa. Scrapers likely depended mostly on periphyton, and collectors likely had detrital and algal food sources suspended in, or settled from, the water column. Although omnivore is not a true FFG, we grouped three taxa under this term because publications found they can be omnivorous and often have variable diets (e.g., Merritt and Cummins, 1996; Lancaster et al., 2005; Table S1). Within site variability of each FFG was examined (1) by calculating mean absolute differences between each pair of taxa within each FFG at sites and (2) by calculating the mean absolute difference of each taxon to its corresponding FFG mean for each site (e.g., if four predators were present at a site, the mean of all predators was calculated and then the mean absolute difference between each of the four predator taxa and that site mean was determined).

To characterize relationships of macroinvertebrate δ¹⁵N with watershed urbanization and nitrogen concentrations, we conducted a nonmetric multidimensional scaling ordination (Bray-Curtis coefficient) using δ¹⁵N ratios of all taxa to describe assemblage change. We calculated Pearson correlations of axis scores with land cover, water chemistry, and δ¹⁵N ratios of FFGs. Pearson correlations also were used to identify relationships of δ¹⁵N ratios of periphyton, FFGs, and macroinvertebrate taxa with land cover, population density, and nitrogen for examining support for their use as indicators of watershed urbanization effects. To determine each indicator’s ability to discern between least disturbed sites and sites with more heavily developed watersheds, we used t-tests to examine if significant differences in δ¹⁵N ratios of biota existed between sites with <5% and sites with >5% impervious cover, a value at which substantial effects on biota frequently have been documented (Utz et al. 2009, King et al. 2011, Fitzgerald et al. 2012, Smucker et al. 2013b). We used locally weighted scatter plot smoothing (LOWESS) with a 0.5 sampling proportion to descriptively examine if possible threshold responses of periphyton and FFG δ¹⁵N to watershed impervious cover and nitrate existed. If multiple taxa within a FFG were present at a site, the mean of their δ¹⁵N was used to produce one value per FFG per site to be used in the aforementioned correlations, t-tests, and LOWESS plots.

We examined two independent and one combined criteria approach to identifying possible nitrogen management goals and to classifying streams. The two independent approaches included (1) using the 75th percentile of nitrate concentrations from 20 sites with ≤5% impervious cover and (2) identifying points along the nitrate concentration gradient at which LOWESS curves indicated a change in the response of periphyton or macroinvertebrate FFGs. Using the 75th percentile of values from least disturbed streams historically has been used as a basic way to identify potential nutrient targets, but criteria based on ecological responses are preferred (US EPA, 2000; Smith and Tran, 2010; Herlihy et al., 2013; Smucker et al. 2013a). The 75th percentile approach also recognizes that pristine systems are rare and reduces the possible influence of atypically high values that may occur due to uncommon local features. We used a combined criteria approach to classify sites based on exceeding or meeting the goals identified by the 75th percentile of nitrate concentrations and the 75th percentiles of biota δ¹⁵N at least disturbed sites. Nitrate and δ¹⁵N of biota both being less than 75th percentiles is evidence of higher quality waters, nitrate and δ¹⁵N of biota both being greater than 75th percentiles is evidence of nitrogen pollution associated with urban land cover, and disagreements in which only one is higher or only some of the response variables are greater than 75th percentiles could indicate possible or moderate nitrogen enrichment from human sources.

Food webs were examined by comparing (1) δ¹⁵N of predators to δ¹⁵N of consumers at sites where they co-occurred and (2) δ¹⁵N of consumers to δ¹⁵N periphyton within sites. Given that periphyton likely did not comprise a substantial portion of detritivore diets, we assumed detritivores likely were representative of unmeasured detrital basal resources when making comparisons to scrapers and collectors. We used linear regression to examine possible changes in trophic relationships across watershed impervious cover gradients (i.e., differences between regression lines at the most forested and the most urban sites). To further examine differences in δ¹⁵N within FFGs and food webs, we used paired t-tests to examine significant differences between δ¹⁵N and δ¹³C ratios of (1) the two most commonly co-occurring predator families at sites (Corydalidae and Aeshnidae) and (2) their possible prey (detritivores and collectors).

3. Results

3.1. Summary of δ¹⁵N variability in biota

The most commonly collected macroinvertebrates were predators (n = 60 sites) in the Aeshnidae, Corydalidae, and Perlidae families, detritivores (n = 53 sites) in the Tipulidae and Limnephilidae families, collectors (n = 34 sites) in the Hydropsychidae and Philopotamidae families, and scrapers (n = 11 sites) in the Heptageniidae and Psephenidae families (Table S1). The ranges of δ¹⁵N values among sites were 13.67‰ for periphyton, 12.07‰ for detritivores, 10.35‰ for predators, 9.02‰ for omnivores, 8.24‰ for collectors, and 6.97‰ for scrapers (Table 1). For sites where multiple taxa within FFGs were observed, mean absolute differences ± SE of δ¹⁵N among taxa within each FFG were 0.94 ± 0.07‰ for predators, 1.06 ± 0.34‰ for detritivores when excluding two rare taxa observed from only one site, and 1.08 ± 0.28‰ for collectors. Mean absolute differences ± SE between individual taxa δ¹⁵N and their corresponding site means were 0.54 ± 0.04‰ for predators, 0.54 ± 0.13‰ for collectors, and 0.37 ± 0.21‰ for detritivores. Scrapers were too sparsely collected to make meaningful within site comparisons. The much greater among site differences in δ¹⁵N of FFGs relative to the low within site variability of δ¹⁵N among taxa within each FFG indicates a high signal to noise ratio.

Table 1.

Descriptive statistics showing the ranges, means, and standard errors of functional feeding groups and periphyton δ15N and δ13C among sites.

	(‰)	Minimum	Maximum	Mean	Standard error
Predators	δ15N	4.88	15.23	9.62	±0.28
	δ13C	−35.01	−26.00	−28.92	±0.54
Omnivores	δ15N	5.31	14.33	9.67	±0.43
	δ13C	−34.75	−25.14	−28.92	±0.54
Detritivores	δ15N	1.56	13.63	6.91	±0.41
	δ13C	−35.68	−23.59	−29.28	±0.36
Collectors	δ15N	4.32	12.56	8.75	±0.39
	δ13C	−37.64	−26.25	−30.58	±0.43
Scrapers	δ15N	5.92	12.89	9.85	±0.72
	δ13C	−38.76	−26.91	−32.56	±1.25
Periphyton	δ15N	2.94	16.61	8.02	±0.30
	δ13C	−33.94	−22.94	−29.58	±0.21

3.2. Responses of biota δ¹⁵N to watershed urbanization and nitrogen concentrations

Increased δ¹⁵N of periphyton, FFGs, and the most commonly observed macroinvertebrate taxa were strongly correlated with decreased watershed percent forest and increased watershed percent impervious cover, population density, and nitrogen concentrations (Table 2). The NMDS ordination showed that macroinvertebrate assemblage structure was associated with impervious cover and nitrogen concentrations (Fig. 2). NMDS axis 1 scores were positively correlated with increasing concentrations of NH₄–N, NO₃–N, and TN, increasing watershed percent impervious cover, and increasing human population densities (p < 0.05). The δ¹⁵N of predators, detritivores, collectors, scrapers, and periphyton were positively correlated with axis 1 scores (p < 0.05), and the percent watershed forest was negatively correlated with axis 1 scores (p < 0.05).

Table 2.

Pearson correlations of watershed characteristics and total and inorganic nitrogen concentrations with δ15N ‰ of periphyton, macroinvertebrate functional feeding groups, and commonly sampled macroinvertebrate families, and with nonmetric multidimensional scaling axes (NMDS) from an ordination of macroinvertebrate assemblages.

	Impervious cover	Population density	Forest	TN	NO3	NH4
Periphyton δ15N	0.64***	0.59***	−0.63***	0.57***	0.57***	0.30**
NMDS axis 1	0.45***	0.4	−0.49***	0.29*	0.2	0.48***
NMDS axis 2	−0.22	−0.17	0.17	−0.27*	−0.31*	−0.22
Predator δ15N	0.68***	0.66***	−0.60***	0.55***	0.60***	−0.01
Omnivore δ15N	0.56**	0.48*	−0.43*	0.54**	0.57**	0.09
Collector δ15N	0.60***	0.56***	−0.55***	0.52**	0.65***	0.15
Detritivore δ15N	0.47***	0.44***	−0.58***	0.56***	0.48***	0.34*
Scraper δ15N	0.74**	0.78**	−0.67*	0.61*	0.69*	0.4
Perlidae δ15N	0.78***	0.69**	−0.66**	0.77***	0.75***	0.55*
Tipulidae δ15N	0.41**	0.38*	−0.52***	0.60***	0.45**	0.44**
Aeshnidae δ15N	0.65***	0.62***	−0.59***	0.49***	0.58***	−0.02
Corydalidae δ15N	0.71***	0.66***	−0.57***	0.58***	0.65***	0.24
Hydropsychidae δ15N	0.53**	0.42*	−0.47**	0.48**	0.54**	0.1
Limnephilidae δ15N	0.61***	0.64***	−0.73***	0.37	0.28	0.06

^***P ≤ 0.001,

^**P ≤ 0.01,

^*P ≤ 0.05.

Fig. 2.

Nonmetric multidimensional scaling (NMDS) ordination of sites using δ¹⁵N values of macroinvertebrate taxa (stress = 0.17). Significant correlations (p < 0.05) of site axis scores with land cover, human population density in watersheds, inorganic and total nitrogen, and δ¹⁵N of functional feeding groups are shown. Text in the gray box labels lines ending within the gray circle from right to left (NH₄ to NO₃). Dark gray circles denote sites with <5% watershed impervious cover. For better visual interpretation, vectors representing correlations, which range from −1 to 1, were scaled in the ordination from −1.5 to 1.5.

The NMDS axis 1 scores and δ¹⁵N of all macroinvertebrate FFGs and periphyton were significantly greater in sites with >5% watershed impervious cover than in the least disturbed sites with <5% watershed impervious cover (Fig. 3). LOWESS curves and scatter plots indicated that periphyton, scrapers, collectors, and predators had the greatest increases in δ¹⁵N from 0 to 5% watershed impervious cover, after which δ¹⁵N still increased, but to a slightly lesser degree (Fig. 4). Detritivores and omnivores had smaller increases in δ¹⁵N from 0 to 5% watershed impervious cover and larger increases from 5 to 10%. As watershed impervious cover increased beyond 10–12%, the δ¹⁵N of periphyton and all FFGs had smaller increases and greater variability (Fig. 4).

Fig. 3.

Box plots showing significant differences (two sample t-tests) of δ¹⁵N ‰ of periphyton and macroinvertebrate functional feeding groups and of raw NMDS axis 1 scores between sites with less than (white) or greater than (gray) 5% watershed impervious cover. Detrit. = detritivores, NMDS = nonmetric multidimensional scaling. ^***P ≤ 0.001, ^**P ≤ 0.01, ^*P ≤ 0.05. Boxes are 25th–75th interquartile range, lines are medians, whiskers denote 5th and 95th percentiles.

Fig. 4.

Locally weighted scatter plot smoothing, with 0.5 sampling proportion, showing relationships between watershed percent impervious cover and δ¹⁵N of macroinvertebrate functional guilds and periphyton (A–D, F). Due to sparse sampling of scrapers (E), a plot of the running mean, with 0.2 sampling proportion, was used to show the relationship with watershed impervious cover. Vertical gray lines highlight 5% and 10% watershed impervious cover.

LOWESS curves and scatter plots indicated that predators, periphyton, collectors, and detritivores had the greatest increases in δ¹⁵N as nitrate concentrations increased to 170, 190, 245, and 260 μg/l, respectively (Fig. 5). Above these concentrations, δ¹⁵N still increased until mostly plateauing beyond 725, 510, 980, and 970 μg nitrate/l for predators, periphyton, collectors, and detritivores, respectively. Some additional increase in δ¹⁵N occurred beyond 1500 μg nitrate/l. The δ¹⁵N responses changed at 180 and 475 μg nitrate/l for omnivores and at 255 and 825 μg nitrate/l for scrapers (Fig. 5).

Fig. 5.

Locally weighted scatter plot smoothing (LOWESS), with 0.5 sampling proportion, showing relationships between nitrate concentrations and δ¹⁵N of macroinvertebrate functional feeding groups (A, C, D) and periphyton (F). Due to sparse sampling of (B) omnivores above 500 μg NO₃/l and scrapers in general (E), a plot of the running mean, with 0.2 sampling proportion, was used. Vertical dashed gray lines highlight points at which LOWESS curves notably change. Concentrations at these points are approximately 170 and 725 μg NO₃/l for predators, 180 and 475 μg NO₃/l for omnivores, 245 and 980 μg NO₃/l for collectors, 260 and 970 μg NO₃/l for detritivores, 255 and 825 μg NO₃/l for scrapers, and 190 and 510 μg NO₃/l for periphyton. An outlier of 5327 μg nitrate/l was excluded.

3.3. Identifying nitrogen management goals

Two independent approaches to identifying possible management goals for nitrogen concentrations in streams were examined. The first identified a concentration of 223 μg nitrate/l based solely on using the 75th percentile of concentrations at least disturbed streams (i.e., those with <5% watershed impervious cover). The second could generate possible management goals ranging from 180–300 μg nitrate/l based on the lowest values at which nutrient–δ¹⁵N relationships of FFGs and periphyton changed in the LOWESS plots. The mean of these nitrate concentrations using all five FFGs and periphyton was 217 μg nitrate/l. Similarly, higher nitrate concentrations at which responses changed could be used to distinguish moderately and highly enriched sites and serve as an incremental goal for streams with the highest concentrations of nitrate. The mean of these higher concentrations for FFGs and periphyton was 748 μg nitrate/l.

The combined criteria approach was used to classify sites (Table 3, Fig. S1). All periphyton and FFG δ¹⁵N values were <75th percentiles at 14.7% of sites. Half or fewer of periphyton and FFG δ¹⁵N values were >75th percentiles at 13.3% of sites. More than half, but not all, periphyton and FFG δ¹⁵N values were >75th percentiles at 12% of sites. All periphyton and FFG δ¹⁵N were >75th percentiles at 60% of sites. For 60.9% of all observations (periphyton and FFGs), nitrate and δ¹⁵N of biota were above their 75th percentiles, and 15.2% of observations were below both 75th percentiles (Table 3). Nitrate concentrations were below while δ¹⁵N of biota were above their 75th percentiles for 14.8% of observations, and nitrate concentrations were above while δ¹⁵N were below 75th percentiles for 9% of observations.

Table 3.

Number of observations above and/or below 75th percentile values of NO3–N (223 μg/l) and δ15N ‰ of periphyton and macroinvertebrate functional feeding groups (FFG) from sites with <5% watershed impervious cover. Number of sites with FFG and periphyton observations are in parentheses. The % agree is the proportion of observations in which δ15N‰ of biota and NO3–N concentrations were both below or both above 75th percentiles. The “% of all” row is the sum of each scenario (e.g., <NO3 and <δ15N‰) divided by the sum of observations in all four scenarios for all FFGs and periphyton.

	δ15N ‰	<NO3	>NO3	<NO3	>NO3	% agree
	75th percentile	<δ15N	<δ15N	>δ15N	>δ15N
Predators (60)	8	10	3	12	35	75
Omnivores (23)	7.6	2	1	4	16	78.3
Detritivores (53)	4.5	9	5	5	34	81.1
Collectors (34)	7.2	5	4	3	22	79.4
Scrapers (11)	7.6	2	0	1	8	90.9
Periphyton (75)	6.5	11	10	13	41	69.3
% of all		15.2	9	14.8	60.9	76.2

3.4. Food webs and effects of watershed urbanization

Using data from sites where both predators and primary consumers were collected, predator δ¹⁵N was +3.07 ± 0.23‰ (mean ± SE), +0.88 ± 0.14‰, and +0.66 ± 0.48‰ than that found in detritivores (n = 46), collectors (n = 32), and scrapers (n = 11), respectively. Based on regression lines, δ¹⁵N of predators changed from being +2.6 to +3.2‰ than that of detritivores and from +0.1 to +1.7 than that of collectors in the most forested and most urban streams, respectively (Fig. 6). When making within site comparisons, collector δ¹⁵N was +0.90 ± 0.21‰ (mean ± SE), detritivores were −1.35 ± 0.25‰, and scrapers were +1.38 ± 0.44‰ than that of periphyton. Collectors δ¹⁵N changed from being +1.8‰ to +0.1‰ than that of periphyton and from +2.4‰ to +1.5‰ than that of detritivores in the most forested and most urban streams, respectively (Fig. 6). Scrapers δ¹⁵N changed from being +1.2‰ to +2.1‰ than that of periphyton in the most forested and most urban streams (Fig. 6). The difference in δ¹⁵N between detritivores and periphyton remained mostly consistent, only changing from −1.0‰ to −1.3‰ in the most forested and most urban streams, respectively (Fig. 6). The δ¹⁵N of detritivores on average increased from 3.5‰ in the most forested streams to 9.5‰ in the most urban streams (Fig. 6).

Fig. 6.

Linear regressions showing relationships of δ¹⁵N (‰) of periphyton and functional feeding groups with % watershed impervious cover (A) and nitrate concentrations (B). R² values for impervious cover and nitrate, respectively: predators = 0.46, 0.36; omnivores = 0.32, 0.43; scrapers = 0.56, 0.48; collectors = 0.31, 0.42; detritivores = 0.23, 0.30; periphyton = 0.39, 0.37. An outlier of 5327 μg nitrate/l, which was atypically high and nearly double the next highest concentration, was excluded based on regression diagnostics.

When the two most commonly co-occurring predators were present at the same sites (n = 28), Aeshnidae had significantly higher δ¹⁵N than did Corydalidae (9.40 ± 0.41‰ and 8.32 ± 0.38‰, respectively [means ± standard error]; paired t-tests, p < 0.01) and significantly lower δ¹³C (−29.39 ± 0.36‰ and −28.68 ± 0.36‰, respectively; paired t-tests, p < 0.01). These results may indicate Aeshnidae consumed more collectors and fewer detritivores than did Corydalidae, because collectors were more enriched in δ¹⁵N (9.08 ± 0.38‰) and less enriched in δ¹³C (−30.64 ± 0.49‰) than detritivores (6.83 ± 0.45‰ δ¹⁵N; −29.13 ± 0.49‰ δ¹³C) within sites where they co-occurred with these two predators (paired t-tests, p < 0.01). When collectors and shredders were collected at the same sites regardless of either predator being present (n = 30), collectors had significantly higher δ¹⁵N than did detritivores (9.12 ± 0.39‰ and 6.94 ± 0.47‰, respectively; paired t-tests, p < 0.01) and significantly lower δ¹³C (−30.66 ± 0.48‰ and −29.06 ± 0.47‰, respectively; paired t-tests, p < 0.01).

4. Discussion

4.1. Responses of biota δ¹⁵N to watershed urbanization and nitrogen

The δ¹⁵N of periphyton and macroinvertebrate FFGs and assemblages (NMDS) were effective indicators associated with increasing concentrations of nitrogen, watershed percent impervious cover, and population density. These relationships (1) provide context for understanding how differences in concentrations of nitrogen and extents of watershed development affect stream ecosystems and (2) provide support for their use as responsive indicators in monitoring and management efforts at watershed and regional scales. Our results showed that increased δ¹⁵N of biota likely were caused by larger amounts of nitrogen delivered to streams from human-related sources. One source of increased nitrogen includes the prevalence of septic systems in suburban watersheds throughout the region, which are of variable performance and condition, substantially contribute nitrate to groundwater and streams (Castro et al., 2003; Williams et al., 2005; Carey et al., 2013), and have been linked to increased δ¹⁵N of aquatic biota (Steffy and Kilham, 2004). Leaky sewer infrastructure also contributes to increased nitrate concentrations in streams and groundwater (Aravena et al., 1993; Groffman et al., 2004; Hatt et al., 2004; Kaushal and Belt, 2012). Speculatively, the leveling of LOWESS curves and greater variability of δ¹⁵N in biota at sites with >10–12% impervious cover could be due to a combination of (1) an increase in the extents of commercial and industrial land uses, which might also contribute to δ¹⁵N being lower than expected based on impervious cover estimates or (2) a transition from homeowner septic systems to centralized waste water collection that would reduce exposure of biota to elevated ¹⁵N, but also possibly increase variability in δ¹⁵N of biota depending on the condition of sewers.

Finer taxonomic resolution and quantitative sampling for biomass and abundance likely would provide additional understanding of how macroinvertebrate and periphyton communities respond to nitrate and watershed development, given that they substantially change in response to urban stressors and affect ecosystem functions, such as organic matter processing and energy transfer through food webs (Smucker and Detenbeck, 2014; García et al., 2017; Tant et al., 2015; Sterling et al., 2016). However, functional feeding groups likely were effective indicators in our study because they are diet based, so even if taxa differ among streams due to biogeography or stressors, they still incorporate ¹⁵N into their bodies in response to that of their food source (Cabana and Rasmussen, 1996; Lake et al., 2001; Post, 2002). Differences in growth rates among taxa, organism size, and associated dietary changes during their life histories could contribute to δ¹⁵N variability within FFGs (Cabana and Rasmussen, 1996; Jardine et al., 2014). However, we found that mean within FFG δ¹⁵N differences between all pairwise comparisons among taxa at each site was <1.1‰ for all FFGs and mean absolute differences to their respective site FFG means were <0.55‰. For comparison, within site variation ≤1‰ is considered small, particularly given wide ranges of δ¹⁵N among sites (Post, 2002; Anderson and Cabana, 2005).

4.2. Nitrogen management goals

LOWESS curves and the 75th percentile of nitrate concentrations at least disturbed streams each provided support for identifying approximately 220 μg nitrate/l as a possible goal for stream protection and management. Higher values based on nutrient-response relationships could be used to distinguish moderately and highly enriched sites and to identify incremental targets for improving streams with high nitrate concentrations. Sites also could be classified as experiencing nitrogen enrichment based on (1) any individual FFG or periphyton exceeding their 75th percentiles of δ¹⁵N at least disturbed sites, which would be the most protective, (2) all FFGs and periphyton exceeding 75th percentiles, or (3) a proportion of them being above 75th percentiles given that 25.3% of sites had disagreements in classification among FFGs and periphyton. When disagreements in classifications exist, this could be a sign of potential or intermediate nitrogen enrichment from human sources. Another option for when disagreements exist could be to place more emphasis on results from higher trophic positions, because variability in primary producers can be notably higher than in consumers and higher trophic levels, which integrate δ¹⁵N variability in their diets over longer timespans (Post, 2002).

A combined approach that jointly considers causal (e.g., nitrate concentrations) and response (e.g., biota δ¹⁵N) parameters can benefit stream assessments by classifying sites and informing decisions. Streams with δ¹⁵N of biota and nitrate concentrations below both 75th percentile values from least disturbed streams likely represent higher quality low nutrient systems and could be used to inform protection efforts and management goals for other streams. Streams that exceed both values could be identified as likely being more highly polluted by nitrogen. Situations in which δ¹⁵N of biota and nitrate concentrations are below and above 75th percentiles, respectively, could suggest that while these streams had elevated nitrate concentrations, ¹⁵N was possibly low or δ¹⁵N of biota was slow to respond. Streams in which δ¹⁵N of biota are above while nitrate concentrations are below 75th percentiles could indicate a need for additional monitoring or identifying these streams as moderately affected by anthropogenic nitrate. This disconnect could be (1) due to biological uptake reducing water column nitrate concentrations, a result of nitrate effects even if concentrations are below a targeted value, or (2) due to human-related sources having ¹⁵N comprising a large proportion of nitrogen available to biota even if concentrations are low. For either situation in which biota δ¹⁵N and nitrate concentrations disagree, additional examination of watersheds, nutrients, and local conditions would help determine if unique habitats or another source of nitrogen with a different isotopic signature contributed to the disagreement (Tucker et al., 1999; Valiela et al., 2000).

4.3. Food webs and urbanization

If assuming an average of +3.4‰ per increasing trophic level, or even +2‰ at the lower end of estimates (Cabana and Rasmussen, 1996; Post, 2002), detritivores appeared to comprise the majority of predators’ diets, because on average δ¹⁵N of predators ranged from +2.6 to +3.2‰ than that of detritivores from forested to more urban streams, respectively, and predator δ¹³C was most similar to detritivores (+0.36‰). Predators likely consumed macroinvertebrates in other FFGs, and possibly even their own, but their δ¹⁵N averaged only +0.88‰ and +0.66‰ than that of collectors and scrapers, respectively, at sites where they co-occurred. The δ¹⁵N and δ¹³C of the two most common predators, Aeshnidae and Corydalidae, indicated minor differentiation in the amount of collectors comprising their diets, though δ¹⁵N still indicated detritivore heavy diets for both. Collectors possibly comprised a growing proportion of predator diets as watershed impervious cover increased, given that predators were +0.1‰ and +1.7‰ than that of collectors in the most forested and most urban streams, respectively. This was even more likely, because (1) of the scarcity of scrapers as prey in general, (2) predator δ¹⁵N was +3.2‰ than that of detritivores in the most urban streams versus being only +2.6‰ in the most forested streams, implying that predators were consuming increased proportions of more ¹⁵N-enriched prey, and (3) this increase to +3.2‰ occurred even with collector δ¹⁵N becoming more similar to that of detritivores in more urban streams (i.e., collectors were consuming more detritus). Predator, shredder, and scraper biomass can decrease as urbanization increases, which can lead to an increased proportion of collectors being present (Stepenuck et al., 2002; Compin and Céréghino, 2007; García et al., 2017), even if collector biomass does not increase (Sterling et al., 2016).

The evidence of detrital pathways was further supported by results from a meta-analysis, which found that mean ± SD of δ¹⁵N of riparian leaf litter was 0.5 ± 1.9‰, 2.3 ± 2.6‰, and 3.2 ± 3.4‰ in forested, suburban, and urban watersheds, respectively, and as high as 4.2‰ in areas affected by septic systems (Hall et al., 2015). The elevated δ¹⁵N likely was due to leaky sewers, subsurface septic effluent, pet wastes, and vehicle emissions increasing the ¹⁵N accessible to riparian plants via ground and surface water (Hall et al., 2015). The mean value of detritivores in our study was 6.9‰ and on average ranged from ~3.5‰ in the most forested streams to ~9.5‰ in the most developed, which corresponds with being approximately +3‰ than that associated with changes in leaf litter (inclusive of standard deviations) reported in Hall et al. (2015). Some of this δ¹⁵N enrichment in both studies also could result from increased algal and microbial growth often observed on leaf litter (Danger et al., 2013; Kuehn et al., 2014; Guo et al., 2016).

Within sites, δ¹⁵N of collectors was on average +0.90‰ than that of periphyton and ranged from being +2.4‰ to +1.5‰ than that of detritivores in the most forested and most urban streams, respectively. These findings indicated that collectors (1) had diets consisting less of detritus and more of periphyton or unmeasured food sources that were more enriched in ¹⁵N, and (2) their dietary sources possibly became more similar to detritivores as urbanization increased. Scrapers likely depend heavily on periphyton as a food source and were on average +1.38‰ than that of periphyton and shifted from being +1.2‰ to +2.1‰ than that of periphyton in the most forested and urban streams, respectively. One study found that detrital material can comprise 30% of periphyton and that scrapers likely selectively incorporate nutrients from more nutritious and more easily digested algal components of periphyton (Rasmussen, 2010). This could have contributed to greater δ¹⁵N of scrapers in more urban streams if algal biomass of periphyton increased, but while maintaining similar ratios to detrital biomass. More complete and quantitative characterizations of habitat and δ¹⁵N of basal resources likely would provide more robust insights into food webs and effects of urbanization in the future.

4.4. Conclusions

The responsiveness of biota δ¹⁵N to urban development in watersheds and anthropogenic increases in nitrogen support their use as indicators in monitoring programs, particularly at larger watershed scales. These responses of periphyton and macroinvertebrate δ¹⁵N can inform management and protection goals and be used as one way to quantify the effectiveness of future nitrogen, stream, and watershed management efforts. While our study showed that these are effective indicators and provide insights into food webs, additional research would benefit their development and use in other regions and countries given the differences in development practices, infrastructure, and climate (Booth et al., 2016; Hale et al., 2016; Parr et al., 2016). The ongoing development and application of indicators responsive to anthropogenic nitrogen will continue to be important as populations grow and development expands, further impacting aquatic ecosystems while increasing demands on freshwater resources.