Food web restructuring across an urban estuarine gradient

Abstract

Food webs in urban estuaries support valuable ecosystem services that are subject to a wide range of stressors that can degrade the structure of trophic networks. Multiple trophic pathways stabilize food webs by providing complementary diet resources for consumers but the consequences of urbanization on estuarine food webs are relatively unknown. In estuarine creeks across an urban-to-suburban gradient, we demonstrate trophic decoupling of benthic and pelagic pathways, trophic niche contraction, and increasing human health risk arising with the same factors that are associated with ecological degradation. This suggests an urban estuarine paradox—human activities often create larger volumes of deep water habitat, yet human activities also render much of this area unproductive with measurable opportunity costs to food webs. Our findings emphasize the shared consequences of environmental degradation for the ecological integrity of urban estuaries and the health of urban communities that rely on estuaries for sustenance.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13280-021-01610-1.

Introduction

Anthropogenic alteration of estuaries is a global concern due to their high economic and ecological value (Barbier et al. 2011). Anthropogenic effects converge most acutely in urban tidal waters, which are expanding rapidly as a growing human population also becomes more coastal (Todd et al. 2019; Sterzel et al. 2020). Urban estuaries frequently suffer from intensive and long-term systemic alterations to bathymetry (e.g., dredging, channelization), shoreline armoring, changes in water residence time, and nutrient amendments arising from surface runoff, groundwater inputs, or point sources (e.g., waste water treatment facilities) (Chapman et al. 2018). Previous studies have clearly documented negative ecological impacts of urbanization on estuarine biota through comparative analyses of urban and non-urban ecosystems. Reported impacts of urbanization include reductions in biodiversity, altered assemblage composition, and changes in vital rates and predator diet composition (Able et al. 1999; Morley et al. 2012; Hall-Scharf et al. 2016). These changes can lead to taxonomic and functional (i.e., trait-based; Carlucci et al. 2020) shifts in biotic communities that, along with degradation of physical and chemical conditions, fundamentally affect the ability of urban estuaries to fulfill key ecosystem services such as providing habitat, supporting biodiversity, enhancing fisheries production, acting as biogeochemical reactors, and buffering against physical and environmental disturbances (Boerema and Meire 2017). While numerous studies have demonstrated ecological differences between urban and non-urban estuarine areas, very few have focused on how specific estuarine processes change across specific habitat and watershed gradients nested within an urban landscape.

This nesting of dendritic estuarine channels, tributaries, and sub-tributaries within heterogeneous urban landscapes can result in unique combinations of ecological pressures within a single urban estuary. In commercial and industrial centers, adjacent landscapes are often dominated by impervious surfaces, vertically built structures such as seawalls and bulkheads have replaced natural shorelines, and in-water reaches are dredged to accommodate shipping (Chapman et al. 2018). In less intensively developed areas, the amount of impervious surface often declines while parks and manicured green spaces increase, lower-relief artificial shorelines (e.g., rip-rap) or non-hardened shorelines become more common, and dredging is typically less extensive. These land cover differences can influence the overland flow of water and associated dissolved and particulate nutrient, sediment, and ecotoxicant loads to the proximal estuarine reach (Kennish 2002). Nutrient and sediment amendments from urban runoff can degrade bottom water habitats by facilitating hypoxia and altering benthic sediment grain-size composition, respectively. Ecotoxicant concentrations are often high in urban estuaries, with characteristic combinations of heavy metals, industrial organic compounds, microplastics, and pesticides that derive from current inputs and industrial legacies. The living intertidal and shallow subtidal areas that provide important habitat for estuarine fauna (a key ecosystem service of estuaries, Gittman et al. 2016) are often absent or degraded in urban areas, leading to changes in the composition of faunal assemblages and habitat function (Bilkovic and Roggero 2008; Munsch et al. 2016). Given the spatial diversity of environmental stressors in urban estuaries, the potential consequences of those pressures on estuarine conditions, and the ever-expanding urban footprint in coastal areas, there is a critical need to develop a synthetic understanding of how the ecological properties and processes of estuary ecosystems change within urban land use and seascape gradients. Efforts to develop conceptual models in other urban ecosystems such as freshwater streams (i.e., urban stream syndrome, Walsh et al. 2005) have provided an invaluable framework for developing and informing research on the functioning of these ecosystems. Such synthetic models are currently lacking for urban estuaries but are needed to inform future coastal development strategies and help target restoration and mitigation of existing urban coastlines to maximize ecological and human health goals (e.g., Gittman et al. 2016; Chapman et al. 2018).

Estuarine food webs, composed of reticulated trophic pathways through which biomass and energy are transferred from basal to tertiary trophic levels, offer a powerful lens through which to evaluate ecosystem change. These food webs support very high levels of secondary production among many estuaries, an important ecosystem service that is critically important to subsistence, commercial, and recreational fisheries in urban areas (Taylor and Suthers 2021). At the base of food webs, the loss of foundational basal groups (e.g., microphytobenthos, submerged aquatic vegetation, fringing mangroves) or over-stimulation of alternative basal groups such as phytoplankton or macroalgae can have profound effects on herbivore community composition and productivity, effects that can propagate to higher trophic levels (McClelland and Valiela 1998; Bernardino et al. 2018; Burkholder et al. 2018). In addition to shifting prey populations, changes in the availability of appropriate resources for higher trophic levels can also arise from lost access to foraging habitat. Reduced access to foraging habitat can include exclusion from deep waters during hypoxic events or from shallow (or surface) waters due to increased water temperatures, as well as habitat loss due to physical destruction or die-offs of biogenic habitat. Such fundamental changes to the identity or availability of trophic resources can precipitate the narrowing or collapse of trophic niche space and (or) a fundamental alteration of trophic pathways (Layman et al. 2007; Warry et al. 2016; Bernardino et al. 2018).

Here, we used stable isotope-based approaches to examine patterns in trophic niche characteristics of a demersal mesopredator fish species (white perch, Morone americana) across an urban-to-suburban regional gradient within a large estuary. White perch are widely distributed in estuaries along the US Atlantic coast and have been introduced to freshwater systems throughout the Eastern and Midwestern United States. White perch are often highly abundant and their generalist and benthic-heavy diet make them an apposite species for investigating benthic–pelagic coupling in urban aquatic systems. The Patapsco River Estuary, which borders the City of Baltimore, MD, was the focal system for this study (Fig. 1a). The central urban areas of the estuary have been foci of intense public interest in their potential for recreational waters, as well as their ecosystem function.

Fig. 1.

Map of the Patapsco River study area (left) with individual creek systems, delineated by sub-watershed, shows Rock Creek (RC), Stoney Creek (SC), Curtis Creek (CC), Middle Branch (MB), Inner Harbor (IH), and Bear Creek (BC). For each creek system, the proportion of the sub-watershed covered by impervious surface (Developed LU Index [LU_Index]), proportion of shallow water (≤ 3 m depth; Bathymetric Index [Bathy_Index]), integrated total water column chlorophyll-a (chl-a [µg]; mean ± SD), and average summer % dissolved oxygen saturation profiles (Supplementary Data S1)

We hypothesized that the relative contribution of benthic trophic pathways to predatory fish would decline as water quality in bottom habitats became increasingly degraded (Trophic Connectivity Hypothesis). As a corollary, we hypothesized that predator niche width and physiological condition would decline across the same gradient of bottom habitat conditions due to losses in foraging opportunities (Predator Niche and Physiological Condition Hypotheses). Finally, we measured mercury body burden of white perch to determine if fish Hg contamination correlated with changes in trophic connectivity with bottom-associated trophic pathways or the intensity of sub-watershed urbanization (i.e., impervious cover; Contaminant Analysis). Our hypotheses and associated tests are intended to directly assess the extent to which a network of estuaries across an urban landscape support important ecological properties linked to food web function (i.e., connectivity among trophic pathways, maintenance of niche space, health of higher trophic levels) and whether losses in food web function are correlated with ecotoxicant risks.

Materials and methods

Field and laboratory methods

Sampling was conducted in the Patapsco River, MD, USA, an urbanized tributary of Chesapeake Bay with a lower watershed dominated by the city and county of Baltimore, MD. The focal predator in this study was white perch Morone americana (Moronidae), an abundant estuarine mesopredator occurring in salinities ranging from fresh to mesohaline in Chesapeake Bay and other temperate estuaries of the North American Atlantic coast (Stanley and Danie 1983). White perch were sampled by trawl from each of six unique sub-estuaries of the Patapsco River during middle (August) and late summer (September) 2018, and during early (May), middle (July), and late summer (September) 2019. These sub-estuaries included four creek systems (Rock Creek, Stoney Creek, Curtis Creek, Bear Creek) and two branches of the Patapsco River (Inner Harbor, Middle Branch). A bottom trawl (3.2-m footrope, 2.5-cm stretch body mesh, and 1-cm cod end liner) was towed for 5 min at ~ 2–2.5 knots. White perch from the trawl were measured for total length (TL, mm), then up to ten individuals per region per season were euthanized in a bath of MS-222 and preserved on ice. A total of 3–4 replicate bottom trawls were conducted each season in approximately the same area of each region (Fig. 1).

Prior to trawling, vertical profiles of water quality were collected using either a Manta plus 3.5 (Eureka, TX) or a YSI EXO multiprobe sonde (Xylem Inc., NY). Salinity, water temperature, chl-a (fluorescence), and pH were recorded at 0.5–1-m intervals from surface to bottom proximal to trawling locations. At the same locations, suspended particulate matter from the surface (0.5 m depth) and bottom (0.5 m above bottom) water was pumped onto the research vessel and a known volume filtered onto pre-ashed (6 h at 500 °C) glass fiber filters (GFFs). GFFs were collected upstream (1.5–2.6 km), within, and downstream (1.3–3.3 km) of trawling areas. GFFs were wrapped in foil and placed on ice. Benthic fauna and surficial sediment were collected from trawl areas by deploying and retrieving duplicate Ponar grabs. Approximately 1 cm³ of surface sediment was collected from each grab (0–5 mm deep) and stored in Whirlpaks® to provide a local baseline for the benthic trophic pathway. The remaining sediment grab was sieved via elutriation to 500 µm, then the sample was fixed in buffered formalin with a vital stain (Rose Bengal). The specimens of the bivalve Mytilopsis leucophaeata incidentally captured in the trawls were set aside. The structure-associated filter-feeding mussel, M. leucophaeta, was used to supplement isotope baseline estimates for trophic pathways. All specimen samples were frozen and held at -50 °C in the laboratory until processing.

In the laboratory, whole fish were thawed, blotted dry, and weighed to obtain wet weight (g). Stomachs and livers were dissected from each individual, blotted dry, and weighed (g). A sample of white muscle tissue was collected from the anterior dorsal region of M. americana specimens for stable isotope analysis (SIA). A subsample of M. americana carcasses were set aside for heavy metals analysis. Abductor muscle from bivalves was excised and aggregated into groups of 2–3 individuals (M. leucophaeata) for SIA. All tissue samples collected for SIA were examined to ensure they were free of bone, skin, scales, or shell, triple rinsed in DI water, and oven dried at 60 °C until completely dry (minimum 48 h).

Subsamples of fish tissue were freeze dried. Approximately 0.15 g of sample was then digested using a Milestone EOTHO-EZ microwave using concentrated ultrapure 4.5 ml Nitric Acid (HNO₃) and 0.5 ml of concentrated ultrapure Hydrochloric Acid (HCl). Samples were heated to 180 °C, allowed to reflux for 15 min, then the diluted samples were analyzed for total mercury (THg) following EPA 1631 using a Tekran 2600 (Supplemental Data S2 and S3). The standard reference materials DORM II and Torte 3 were used to assess accuracy of ± 10%. A reporting detection limit for THg was 0.3 ng g⁻¹.

Benthic samples were transferred to ethanol, then identified and assigned to one of five taxonomic groups: Amphipoda, Bivalvia, Isopoda, Decapoda, and Polychaeta. Individuals were aggregated, dried at 60 °C, and weighed to obtain dry wt (g) for each taxonomic group. Sediment samples and GFFs were similarly dried, but then dry sediment samples were split in half. During 2018, one half was acidified to remove inorganic carbonates prior to carbon (C) isotope analysis by directly applying 1 N HCl to sediment (Kennedy et al. 2005). Acidified sediment was then repeatedly rinsed with DI via centrifugation to remove Cl residue, redried, and packed in silver capsules for stable isotope measurement. The unacidified sediment sample was packed in tin capsules and analyzed for nitrogen (N) stable isotope composition. Comparison of acidified and unacidified δ¹³C values from 2018 samples revealed no difference between samples (paired t test, df = 2, t-statistic = 0.69, p = 0.28). Therefore, sediment samples in 2019 were not acidified prior to analyzing δ¹³C values. GFFs and muscle tissues were packed into tin capsules (muscle was pulverized first) for C and N stable isotope analysis. Samples were run on a Thermo Fisher Delta V+ isotope ratio mass spectrometer coupled to a Carlo Erba NC 2500 Elemental Analyzer at the Central Appalachians Stable Isotope Facility at the University of Maryland Center for Environmental Science’s Appalachian Laboratory in Frostburg, MD. Stable isotope values are expressed relative to international standards (Vienna PeeDee Belemnite for C, and air for N). Long-term analytical precision (SD) of quality controlled standards (USGS 40, USGS 41) at CASIF is 0.1 ‰ for both δ¹³C and δ¹⁵N.

Data analysis

Predictor variables

Watersheds for each region were delineated from county and city watersheds and supplemented with the USGS streamstats tool (CoNED, AACO 2011; USGS 2016; BCPD 2019; O’Neill, C. 2019. BaltimoreSubBasins. Baltimore City Department of Planning, Personal Communication). Land cover characteristics for each sub-watershed were extracted from the National Land Cover Database, a raster dataset of land cover in the United States allocated to 16 land use classes at 30 m resolution (https://www.mrlc.gov/). The proportion of each sub-watershed area designated as light-, medium-, or high-intensity developed land use, corresponding to pixel % impervious surface values ranging 20–100%, was calculated to yield an index of urban land use intensity (LU index). A shallow water bathymetry index (Bathy index), calculated as the proportion of the basin area for each region ≤ 3 m deep, was calculated from the USGS CoNED Topobathy for the Chesapeake Bay (NOAA Station 8574680, Baltimore, MD; CoNED). The richness of benthic prey was calculated as the number of taxonomic groups present in each creek during each season. Biomass of benthic prey was calculated as the sum of the dry weights of all taxonomic groups. The Bivalvia group was excluded from both of these calculations because bivalves represent a very minor component of M. americana diet relative to the other taxonomic groups (Rudershausen and Loesch 2000; St-Hilaire et al. 2002; Weis 2005; Buchheister and Latour 2015).

Water quality data collected during this study and by the Baltimore Water Watch monitoring survey (Blue Water Baltimore [BWB], https://baltimorewaterwatch.org/) were analyzed to provide additional contextual conditions present in each region. The chl-a concentration at each depth interval was averaged across seasons, then the depth-specific concentrations were summed and converted to µg to provide a measure of the average total water column chl-a for each creek (Fig. 1). Five years (2015–2019) of monthly dissolved oxygen saturation data collected at co-located BWB monitoring stations from June to September were averaged at 1-m intervals from surface to bottom to provide a time-integrated measure of oxygen profiles for each creek (Fig. 1).

Contribution of trophic pathways

Baseline stable isotope values for benthic and pelagic trophic pathway endmembers were estimated using multifactor analysis of variance. Predictor variables included in the model were type (BOM, SPOM, M. leucophaeata), region, season, and location in the water column. Two interaction terms were considered (type × region, region × season) but the region × season interaction was not significant and was excluded from the final models. Model fit was assessed through visual examination of the residuals and verification of the assumptions of homogeneity of variance and residual normality (e.g., quantile–quantile plots). The fitted δ¹³C (N = 164, F-statistic_{(degrees of freedom [numerator, denominator] = 38, 125)} = 10.86, p < 0.0001; r² = 0.77) and δ¹⁵N (N = 164, F_{(df = 38, 125)} = 10.91, p < 0.0001; r² = 0.77) models were used to predict mean and standard deviation of BOM and SPOM values for each combination of season and region.

Predator and trophic baseline stable isotope data were analyzed using MixSiar, a Bayesian multi-endmember linear modeling framework implemented in r (Stock et al. 2018). Trophic position of M. americana was assumed to be 3.4 (Stanley and Danie 1983; Weis 2005). Stable isotope trophic enrichment factors (i.e., the serial enrichment of stable isotope ratios across trophic transfers) were based on best available estimates from the literature M. americana (δ¹³C = 1.3 ± 0.3; δ¹⁵N = 3.4 ± 0.3; Vander Zanden and Rasmussen 2001; McCutchan et al. 2003). All predator δ¹³C and δ¹⁵N values from each region were examined to ensure they fell between that region’s benthic and pelagic endmembers (i.e., mixing polygon in isospace; Supplementary Fig. S1). We included tributary as a class variable and TL as a continuous covariate in the mixing model. Means and standard deviations from the posterior distributions were used to calculate the estimated contribution of the benthic trophic pathway (BP) to white perch. An alternative mixing model structure, obtained by calculating individual-level BP estimates was examined and results were found to be qualitatively identical (Supplementary Text S1).

Stable isotope ellipse area (SEA) in δ¹³C and δ¹⁵N space was used as a proxy for predator trophic niche area in each region. Bayesian estimates of SEA (SEA_B) and associated uncertainty, which are unbiased relative to sample size and robust to small sample size, were modeled for each region using the SIBER package in r (Jackson et al. 2011). Posterior probability distributions of region SEA_B values were compared using likelihood ratios. Mean SEA_B values were calculated for each region and analyzed with respect to predictor variables.

Modeled BP and SEA_B values were analyzed using weighted least squares (WLS) regression analysis in which the response variables were treated as estimated dependent variables (EDV models; Lewis and Linzer 2005). Dependent variable weights were calculated as the reciprocal of the standard errors associated with SEA_B estimates for each region. The WLS approach can yield more efficient parameter estimates at the risk of underestimating regression standard errors (Lewis and Linzer 2005); therefore, we compared EDV results using WLS and ordinary least squares (OLS). We also compared results from the EDV models with a non-parametric measure of correlation (Kendall’s tau) between response and predictor variables. Individual-level response variables including HSI, stomach fullness, and THg values were analyzed using general linear models. Residuals were examined and the parametric assumptions of residual homoscedasticity (Levene’s test) and normality (quantile–quantile plots) were satisfied.

Results

The Patapsco River Estuary is highly urbanized, with a dendritic network of branching creeks that drain sub-watersheds that vary greatly in dominant land use characteristics, particularly in the proportion covered by impervious surface (Fig. 1b; Supplementary Data S1). Of the 6 creek systems included in this study, the percent of the sub-watersheds classified as lightly to highly developed land cover (i.e., 20 to 100% impervious surface classifications; National Land Cover Database [2016], Dewitz 2019) ranged from 35 to 90% of the total area of each watershed. The tidal portion of the creeks themselves also differ in the extent to which shallow habitat has been lost through artificial channel widening and deepening (Fig. 1c). Creek basins range from relatively shallow to deep, with 8 to 65% of each basin area less than 3 m deep.

In our evaluation of the Trophic Connectivity Hypothesis, we found that the contribution of benthic trophic pathways (BP) to the diet of white perch increased linearly with the availability of shallow water habitat (≤ 3 m deep; Fig. 2, Table 1). In the shallowest region (Bear Creek), 65% of the bottom is ≤ 3 m deep and BP contributed an estimated 76 ± 7% to the assimilated diet of white perch. The contribution of BP to white perch assimilated diet declined to 41 ± 11% in the deepest and most highly modified creek system (Inner Harbor, 8% shallow water habitat). This decoupling of white perch from the dominant trophic pathway supporting it in shallow systems represents a 54% reduction in the contribution of BP to predator assimilated diet. We did not find a relationship between contribution of BP and impervious surface cover in the adjacent sub-watershed but we did observe a relationship between a proxy for trophic niche area, stable isotope niche area (calculated as bivariate ellipses in δ¹³C and δ¹⁵N space [SEA_B, ‰²]; Jackson et al. 2011), and impervious surface (Predator Niche Hypothesis). Across creeks, SEA_B declined linearly as the amount of impervious surface in the sub-watershed increased (Fig. 2, Table 1). White perch niche area in the least developed sub-watershed, Rock Creek, was among the largest at 6.98 ± 1.65 ‰². White perch trophic niche was reduced by 46% to 3.54 ± 0.99 ‰² in the Inner Harbor, the creek system with the most impervious surface in the sub-watershed. Unlike BP contribution, we did not see a relationship between SEA_B and the availability of shallow water habitat (Fig. 2).

Fig. 2.

Modeled standard ellipse area corrected for small sample size in stable isotope niche space (SEA_B; mean ± SE) and proportional contribution of benthic trophic pathways (mean ± SE) to white perch from each creek system plotted against the proportion of the sub-watershed covered by impervious surface (Developed LU Index), proportion of shallow water (≤ 3 m depth; Bathymetric Index), and benthic invertebrate forage functional group richness (Functional group richness). Linear regression lines and shaded 95% confidence intervals from EDV models are shown where significant at α = 0.05

Table 1.

Results from Estimated Dependent Variable (EDV) regression modeling, including root mean square error (RMSE), F-statistic and associated p-value, and adjusted R². Response variables include the proportional contribution of benthic trophic pathways (BP) and standard ellipse area corrected for small sample size in stable isotope niche space (SEA_B) for white perch from each creek system. Predictor variables included proportion of shallow water (≤ 3 m depth; Bathy_Index), proportion of the sub-watershed covered by impervious surface (LU_Index), and benthic invertebrate forage biomass (Benthos_Bio) and functional group richness (Benthos_Rich). β-parameter estimates (and SE) shown for significant EDV models, fitted using both weighted- and ordinary least squares (WLS, OLS). Non-parametric correlation coefficient between pairwise response and predictor variables (Kendall’s τ) provided with associated p-value

Model	Predictor	EDV model performance				Estimate ± SE		Kendall's τ
		RMSE	F	p	Adj-R2	βWLS	βOLS
BP	BathyIndex	0.155	41.88	0.003	0.89	0.69 (0.11)	0.70 (0.11)	0.73 (p = 0.04)
	LUIndex	0.500	0.44	0.54	− 0.13			− 0.47 (p = 0.19)
	BenthosBio	0.157	0.46	0.53	0.10			0.20 (p = 0.57)
	BenthosRich	0.076	15.11	0.02	0.79	0.17 (0.04)	0.16 (0.04)	0.69 (p = 0.06)
SEAB	BathyIndex	0.895	0.90	0.40	0.18			0.20 (p = 0.57)
	LUIndex	0.442	16.03	0.02	0.80	− 5.92 (1.48)	− 6.67 (1.53)	− 0.73 (p = 0.04)
	BenthosBio	1.535	0.56	0.50	0.12			0.20 (p = 0.57)
	BenthosRich	0.879	9.90	0.03	0.71	1.71 (0.54)	1.86 (0.47)	0.83 (p = 0.02)

We analyzed the biomass (kg per m²) and functional diversity of known benthic prey of white perch to determine whether the observed patterns in BP contribution and trophic niche area could be explained by characteristics of the benthic forage base. Benthic biomass declined continuously from spring to late summer but did not differ consistently among creeks (General linear model [GLM], N = 17, F-statistic_{(degrees of freedom [numerator, denominator] = 6, 10)} = 6.41, p = 0.005; r² = 0.79). Seasonally, benthic biomass declined (F_{1, 10} = 36.40, p = 0.0001) at a rate of 0.27 ± 0.04 (SE) kg m⁻² per month, from a high of 1.08 ± 0.30 kg m⁻² during the spring to 0.01 ± 0.004 kg m⁻² by the late summer. Unlike seasonality, there were not consistent, creek-level differences in benthic biomass (F_{5, 10} = 0.42, p = 0.83) or significant relationships of benthic biomass with white perch trophic indices (Table 1). In contrast, taxonomic functional richness of benthic prey showed both seasonal and spatial differences (GLM, N = 16, F_{6, 10} = 4.61, p = 0.02; r² = 0.73), with richness declining seasonally (F_{1, 10} = 7.78, p = 0.02) from 2.67 ± 0.31 taxonomic groups to 1.50 ± 0.31 taxonomic groups from spring to late summer, respectively. Spatial differences were present among creeks (F_{5, 10} = 3.97, p = 0.03), although pairwise comparisons were not significant (T-test, df = 10, |t-statistic|≤ 3.38, p ≥ 0.06). When analyzed in regressions with BP and SEA_B, benthic taxonomic richness was significantly and positively related to both trophic indices (Fig. 2, Table 1).

A measure of white perch physiological condition, the ratio of the liver weight relative to total body weight (minus stomach contents; i.e., hepatosomatic index, HSI), was negatively related to impervious surface after accounting for the effect of body size and season (Physiological Condition Hypothesis; Table 2). HSI can vary depending on a range of factors but is often used as an index of physiological condition because many fishes use the liver as an energy storage depot (Lambert and Dutil 1997; Schloesser and Fabrizio 2017). A qualitative negative relationship between impervious surface and stomach fullness (ratio of stomach contents weight to body weight, represented as a %) suggested a similar pattern but was not significant at α = 0.05 (p = 0.07; Table 2). Neither HSI or stomach fullness was related to the amount of shallow water habitat (p ≥ 0.38).

Table 2.

Output from generalized linear models including F-statistic and associated p-values for each variable. Both parameter estimates and standardized parameter coefficients (with SEs) are provided for each continuous variable in the models. Total sample size, optimal response variable distribution (Dist.), and Pearson’s product-moment correlation coefficient (R_P) between predicted and observed values provided for each model. Response variables include hepatosomatic index (HSI), stomach fullness (Fullness), and ln-transformed total mercury tissue concentration (Ln-THg) of white perch from each creek system. Predictor variables included sampling season, total length of individual perch (TL), proportion of the sub-watershed covered by impervious surface (LU_Index), the proportion of shallow water (≤ 3 m depth; Bathy_Index), and interaction terms

Model	Predictor*	F	p	Estimate	SE	StdCoeff	SE	N	Dist	RP
HSI	Season	4.75	0.01					125	Gamma	0.37
	TL	21.95	< 0.001	− 0.005	0.001	− 2.74	0.58
	LUIndex	9.13	0.003	− 0.75	0.25	− 1.8	0.59
	BathyIndex	0.78	0.38	0.29	0.33	0.55	0.62
Fullness	Season	6.3	0.003					131	LogN	0.35
	TL	2.89	0.09	− 0.0015	0.0009	− 0.84	0.5
	LUIndex	3.36	0.07	− 0.38	0.21	− 0.93	0.51
	BathyIndex	0.07	0.78	− 0.07	0.26	− 0.14	0.51
Ln-THg	TL	6.4	0.02	− 0.011	0.004	− 3.21	1.27	30	Gamma	0.87
	LUIndex	5.08	0.03	− 1.44	0.64	1.71	0.76
	BathyIndex	6.48	0.004	1.25	0.49	1.27	0.5
	LUIndex x TL	10.17	0.02	0.0186	0.006	3.15	0.99
	BathyIndex x TL	11.52	0.002	− 0.0179	0.005	− 2.51	0.74

^*Interactions not shown in the table were not significant during preliminary model development and were excluded from the final model variant

Total mercury (THg) content of white perch muscle tissue (ng THg g⁻¹ dry tissue) increased as the amount of shallow water habitat declined and as the amount of sub-watershed impervious surface increased (Contaminant Analysis; Table 2). Interestingly, there was an interaction between body size and both of these predictors, such that the effects of these predictors on THg tissue content were greater for larger white perch than juveniles. This was explored by predicting THg tissue content for two size classes of white perch (50 mm and 150 mm) in systems with high (60%) and low (10%) amounts of shallow water habitat across a range of impervious surface values (Fig. 3). Model results indicated THg tissue content in large white perch ranged from approximately 2 to 8 times higher in predominantly deep water systems (i.e., those dominated by pelagic trophic pathways), and increased by approximately twofold as the amount of impervious surface increased in the sub-watershed (regardless of shallow water habitat availability). Fish in the urban Patapsco River system had higher THg concentrations (14.7–242.4 ng g⁻¹) than specimens collected from offshore waters in Chesapeake Bay, away from urban systems (18.1–49.6 ng g⁻¹; Supplementary Data S2 and S3).

Fig. 3.

Model predicted ln-transformed total mercury tissue content (Ln-THg, ηg g⁻¹ [± SE]) of a 50-mm (orange symbols) and 150-mm (blue symbols) white perch Morone americana sampled from a creek with abundant shallow water habitat (left panel, 60% bathymetry ≤ 3 m), and a creek with little shallow water habitat available (right panel, 10% area ≤ 3 m). White perch image from IAN Image Library (ian.umces.edu)

Discussion

This study provides direct evidence of bottom-up changes to food web architecture in an urban estuary. Our findings suggest that habitat alterations associated with deepening bathymetry in urban estuaries can decouple the flow of biomass from benthic trophic pathways to higher trophic levels. At the same time, increasing impervious surface in adjacent sub-watersheds was associated with trophic niche contraction. Further, both altered bathymetry and impervious surface were related to size-specific total mercury content of white perch tissues. As such, our study clearly demonstrates the following: (1) simplification of trophic structures in an estuarine food web in response to habitat alteration, (2) connectivity of sub-watersheds with the adjacent estuary and ecological consequences of increased impervious surface cover, and (3) the potential for human health risk increases with the same factors that are associated with ecological degradation. These findings also suggest an urban estuarine paradox, while human activities in these systems can result in a larger volume of potential habitat (deeper waters), surrounding land use and legacy contaminants ultimately result in measurable opportunity costs to the food web.

The loss of trophic connectivity between consumers and different energy pathways can reduce the stability of food webs and affect the resilience of food webs (Rooney et al. 2006; Telsnig et al. 2019). In aquatic systems, benthic pathways exhibit slower production rates that typically lag periods of peak production in pelagic pathways (Rooney et al. 2006; McMeans et al. 2015). Under typical conditions, mobile predators can integrate these different trophic pathways, exploiting the pulsed availability of pelagic resources while relying on the slower yet more stable availability of benthic resources (McMeans et al. 2015). The availability of multiple trophic pathways and, importantly, accessibility to those pathways by predators reflect a measure of the adaptive capacity of an ecosystem to respond to natural and anthropogenic perturbations (McMeans et al. 2016). In this study, the reduced contribution of benthic trophic pathways was associated with increasingly poor water quality in artificially deepened areas and reduced functional diversity of benthic prey groups. In urban estuaries, the pulsed nature of estuarine pelagic production can be amplified due to the vulnerability of these systems to large nutrient loading events arising from runoff and sewage overflows (Anderson et al. 2002; Yang and Lusk 2018). Amplification of pelagic primary production and the reduction in benthic pathway contribution suggest urban food webs are particularly vulnerable to destabilizing conditions such as hypoxia, characteristics associated with lower secondary production or biodiversity such as that documented here for benthic invertebrate functional groups.

Contractions in predator niche area with increasing watershed urbanization could arise from a homogenization of diet, a less diverse forage base, or a combination of these two factors (Layman et al. 2007). Benthic and epibenthic community composition in urban areas is often depauperate and dominated by a small number of highly tolerant, ruderal species that are able to persist or rapidly recolonize areas despite sediment disturbance, toxicity, and low dissolved oxygen levels in pore water and the overlying near-bottom water. Runoff events or the composition of surface runoff appears to be particularly important given the strong association between proximal watershed impervious surface cover and predator niche area (Fig. 2). The loss of biodiversity in urban benthic food webs is problematic because species loss in slow, stable trophic pathways (e.g., benthic pathways) can lead to food web instability (Rooney and McCann 2012). This is in addition to any effects that environmental variability may play on habitat conditions for higher trophic levels (Schrandt and MacDonald 2020). Based on our findings, remediation efforts targeting the regeneration or creation of shallow water habitat may serve as an ecological buffer against the loss of trophic connectivity between pelagic and benthic pathways and an environmental refuge for poor water quality in artificially deepened areas.

Contaminant loads in urban runoff, including metals (Tiefenthaler et al. 2008), are typically high and contribute to substantial legacy stores in estuarine sediments adjacent to industrialized areas (Mason et al. 2004). Mercury concentrations in urban surface runoff are similar to concentrations in precipitation (Eckley and Branfireun 2008), leading to large pulses of Hg (as well as other contaminants) into local water bodies. Ions transported in freshwater runoff are rapidly salted out in brackish coastal waters, resulting in continual and historic sediment loading by many contaminants. Urban sediment contaminants are often above toxicity thresholds as a whole (but more questionable for metals alone Garcia et al. 2011; Hartzell et al. 2017), thereby attracting attention as potentially being the major source of Hg to estuarine food webs. However, for the most effective accumulation into the food web, Hg must first be methylated and while sediment Hg concentrations are high in the Baltimore Harbor region, the proportion of Hg that occurs as MeHg is relatively low compared to other contaminated sites (Heyes et al. 2004) with anoxia, resulting in high levels of sulfide retarding Hg methylation (Benoit et al. 1999). Furthermore, sediments in the study area are unlikely to provide a food source to mobile benthic predators, due to mortality of benthic prey arising from pore water anoxia or restricted access to benthic habitats by foraging predators due to coincident hypoxia in overlying waters. Coupling food web metal accumulations to sediment sources requires substantive sediment–water flux or resuspension to remobilize contaminants into the pelagic food web. Alternatively, metals are bioavailable during the post-runoff period of contaminant introduction from the adjacent watershed. Runoff-associated nutrients stimulate algal production (Anderson et al. 2008), and provide a means of incorporating contaminants into the pelagic food web. Direct measurements of pulsed trophic assimilation events are difficult to trace, requiring intense sampling of runoff and food web compartments. Here, we used the structure of the food web itself to identify pelagic trophic pathways as the important mercury accumulation pathway. Using large-scale watershed land use classifications provides some indication of the relative importance of the watershed relative to internal sources (Hurley et al. 1998; Seelen et al. 2020) but the complexity of the urban estuary requires a finer-scale approach that considers the ecology of the system, especially when considering cost effective restoration activities.

To date, urban ecology has largely focused on freshwater and terrestrial ecosystems. For example, only 6% of articles published in the journal Urban Ecosystems (Springer, 1997-present) were flagged during a broad search using the terms ‘estuary’ or ‘estuarine’ (81 out of 1251 articles, link.springer.com, accessed 12/8/2020). While estuaries have received some attention, there is a clear need to continue expanding this focus given the continued pressures of coastal development on these ecosystems. Here, we have identified measurable characteristics of physiological, food web, and habitat impairment that are likely to occur in other urban areas and might be diagnostic of an estuarine equivalent of urban stream syndrome (Walsh et al. 2005), i.e., an ‘urban estuary syndrome.’ These characteristics are listed in Fig. 4’s table and provide an opportunity to customize restoration outcomes and monitoring programs to capture key structural and functional features of estuaries that are likely to describe other urban estuarine systems as well.

Fig. 4.

Left: Conceptual diagram depicting shift in predator trophic niche and accumulated mass-specific total mercury body burden in an urban estuary associated with changes in proximal land use characteristics (impervious surface) and availability of shallow water habitat (≤ 3 m depth). White perch and housing images from IAN Image Library (ian.umces.edu). Right: Generalized abiotic conditions and biotic responses associated with urbanization of estuaries based on: A—the current study, B—previously published literature, C—hypothesized response based on findings from this and previous studies (see Supplementary Data S4 for references)

Conclusion

These ecological pressures (i.e., loss of shallow habitats, poor deep water quality, high connectivity with watershed impervious surfaces) on the food webs of urban estuaries are particularly important given the significant role these ecosystems play in supporting local economies and food security for residents through artisanal sustenance harvest. Fishing offers a low-cost, local source of protein for sustenance fishers that is often socially and culturally important (Quimby et al. 2020); yet, the quality and availability of fisheries species is modulated by anthropogenic pressures ranging from sub-lethal contamination [e.g., heavy metals (Figs. 3, 4), organic compounds] to large-scale die-offs (e.g., HABs, anoxia). Even within catch-and-release fisheries, losses in the productivity of urban estuaries can lead to lost revenue and changes in the socio-ecological system of urban fisheries (sensu Taylor and Suthers 2021). Here, we have documented evidence that the architecture of urban food webs might be particularly prone to instability due to the twin processes of in-water habitat alteration through artificial deepening of waterways and landscape development to impervious surface in proximal sub-watersheds. Development strategies or remediation efforts that seek to maintain or increase the availability of shallow water habitat in urban estuaries could serve multiple ecological benefits by providing surfaces and interstitial space for epifauna and small fish that is removed from deep water hypoxia while providing supplemental foraging opportunities for predators. Further, we show that the conditions that lead to trophic decoupling are also associated with heavy metals contamination in fish, a convergence between ecological degradation of estuarine food webs and increased consumption-based human health risks through increased heavy metal exposure. Given that these processes are likely to occur in urban areas, it is most likely that urban population segments that use and rely on artisanal harvest from proximal estuary reaches will be most affected by these losses in ecosystem integrity and health risks.

Supplementary Information

Below is the link to the electronic supplementary material.

Biographies

Ryan J. Woodland

is an Associate Professor at the University of Maryland Center for Environmental Science (UMCES) Chesapeake Biological Laboratory (CBL). His research interests include estuarine ecology, trophic ecology, and human–environment interactions.

Lora Harris

is an Associate Professor at UMCES/CBL. Her research interests include systems ecology, theoretical ecology, and diversity and inclusion in the geosciences.

Erin Reilly

is a Staff Scientist at the James River Association (formerly at UMCES/CBL). Her research interests include estuarine ecology, wetland ecology, and coastal processes.

Alexandra Fireman

is a Research Technician at the University of Florida (formerly at UMCES/CBL). Her research interests include sea turtle ecology, conservation biology, and stable isotope ecology.

Eric Schott

is an Associate Professor at the UMCES/Institute of Marine & Environmental Technology. His research interests include marine pathogens, urban aquatic health, and marine invertebrate health and disease.

Andrew Heyes

is an Associate Research Professor UMCES/CBL. His research interests include aquatic toxicology, geochemistry, and analytical chemistry.