Relations between macroinvertebrates, nutrients, and water quality ...

Environ Monit Assess (2014) 186:1167–1182 DOI 10.1007/s10661-013-3447-1

Relations between macroinvertebrates, nutrients, and water quality criteria in wadeable streams of Maryland, USA Matthew J. Ashton & Raymond P. Morgan II & Scott Stranko

Received: 1 May 2013 / Accepted: 14 September 2013 / Published online: 11 October 2013 # Springer Science+Business Media Dordrecht 2013

Abstract In an ongoing effort to propose biologically protective nutrient criteria, we examined how total nitrogen (TN) and its forms were associated with macroinvertebrate communities in wadeable streams of Maryland. Taxonomic and functional metrics of an index of biological integrity (IBI) were significantly associated with multiple nutrient measures; however, the highest correlations with nutrients were for ammoniaN and nitrite-N and among macroinvertebrate measures were for Beck’s Biotic Index and its metrics. Since IBI metrics showed comparatively less association, we evaluated how macroinvertebrate taxa related to proposed nutrient criteria previously derived for those same streams instead of developing nutrient–biology thresholds. We identified one tolerant and three intolerant taxa whose occurrence appeared related to a TN benchmark. Individually, these taxa poorly indicated whether streams exceeded the benchmark, but combining taxa notably improved classification rates. We then extracted major physiochemical gradients using principal components analysis to develop models that assessed their influence on nutrient indicator taxa. The response of intolerant taxa was predominantly influenced by a M. J. Ashton (*) : S. Stranko Maryland Department of Natural Resources, Monitoring and Non-Tidal Assessment Division, 580 Taylor Avenue, C-2, Annapolis, MD 21401, USA e-mail: [email protected] R. P. Morgan II Appalachian Laboratory, University of Maryland Center for Environmental Science, 301 Braddock Road, Frostburg, MD 21532-2307, USA

nutrient-forest cover gradient. In contrast, habitat quality had a greater effect on tolerant taxa. When taxa were aggregated into a nutrient sensitive index, the response was primarily influenced by the nutrient-forest gradient. Multiple lines of evidence highlight the effects of excessive nutrients in streams on macroinvertebrate communities and taxa in Maryland, whose loss may not be reflected in metrics that form the basis of biological criteria. Refinement of indicator taxa and a nutrientsensitive index is warranted before thresholds in aquatic life to water quality are quantified. Keywords Stream nutrients . Nutrient criteria . MBSS . Macroinvertebrates

Introduction Eutrophication of aquatic ecosystems is a major cause of water quality impairment (USEPA 1999). The effects of excessive nutrients derived from agriculturally dominated landscapes and delivered from the headwaters to estuaries are pervasive across all levels of trophic complexity (Cooper 1993; Smith et al. 1999; Castro et al. 2003; Galloway et al. 2003; Kemp et al. 2005). Eutrophication may also result in costly impacts to fisheries and human health (Heisler et al. 2008; Dodds et al. 2009). In response to this pressing environmental problem, the USEPA published nutrient criteria while allowing the States and authorities under the Clean Water Act flexibility to either adopt or to develop regionally specific, scientifically defensible criteria (Dodds and Welch 2000; USEPA

1168

2000a, b). However, determining nutrient criteria that are protective of aquatic ecosystem health and biological integrity is difficult. Eutrophication directly alters benthic macroinvertebrate communities by increasing low-level producers and consumers (Peterson et al. 1993; Cross et al. 2006; Bowman et al. 2007). Although relationships between nutrients and eutrophication in streams are well documented (Cole 1973; Smith et al. 1999), relationships between nutrients and benthic macroinvertebrate assemblages in wadeable streams have been explored in just a handful of studies (e.g., Miltner and Rankin 1998; Smith et al. 2007; Wang et al. 2007; Evans-White et al. 2009; Miltner 2010). The difficulty in describing the nutrient–eutrophication relationship is because macroinvertebrate communities are influenced by physical habitat (Poff and Ward 1990), flow and thermal regimes (Ward and Stanford 1992; Poff and Ward 1989), geomorphology (Richards et al. 1996), land use (Richards et al. 1993), and factors across multiple scales (Poff 1997). Consequently, metrics based on taxonomy may not respond to stressor gradients because they often lack a mechanistic link (Poff et al. 2006). Further complicating this task is that agricultural practices, which are most often linked to excessive nutrients and eutrophication, may also alter the aforementioned factors in streams (Cooper 1993; Richards et al. 1993; Stone et al. 2005; Zheng et al. 2008). In a prior study, Morgan and Kline (2011) used the extensive MBSS water quality database to describe stream nutrient concentrations in relation to landcover characteristics across multiple spatial scales in first- to fourth-order non-tidal streams of Maryland. Using various methods, multiple criteria for stream nutrients were proposed to protect biological integrity. Our first objective was to examine and describe the relationships between measures of benthic macroinvertebrate communities and nutrient concentrations. Based on the patterns we initially observed in this analysis, we opted not to continue investigation into nutrient–biology thresholds (e.g., Evans-White et al. 2009; Miltner 2010; King et al. 2011) until we obtained a better understanding of the relations between benthic macroinvertebrates and environmental gradients. Therefore, our second objective was to define macroinvertebrate taxa that were indicators (e.g., tolerant and intolerant) of chronic total nitrogen (TN)

Environ Monit Assess (2014) 186:1167–1182

concentrations in relation to empirically derived water quality criteria for nutrients in Maryland streams (e.g., Morgan and Kline 2011). We evaluated this list of macroinvertebrate taxa seemingly responsive to stream nutrients against an independent set of macroinvertebrate and nutrient data and compared to published tolerance values (Barbour et al. 1999; Southerland et al. 2007). To understand the utility of this potential indicator, we examined the ability of individual and combinations of indicator taxa to differentiate between streams that were above or below Morgan and Kline’s (2011) proposed TN benchmark of 1.68 mg/L. Ultimately, a useful indicator should maximize correct classifications, minimize misclassifications, and occur at monitoring sites with some regularity (Karr and Chu 1997). Knowing that bivariate relationships often obscure abiotic influences associated with multiple stressors, we also sought to elucidate how major physiochemical and landscape gradients influenced nutrient–biological community relationships as the patterns could help further refine biologically protective water quality criteria and indicators of nutrient impairment.

Methods Nutrient–macroinvertebrate relationships The sampling design and applications of the Maryland Biological Stream Survey (MBSS) to water quality standards and biocriteria are described exhaustively by others (Klauda et al. 1998; Stranko et al. 2005; Southerland et al. 2007; Southerland et al. 2009). In short, water quality samples are collected during a spring index period (March 1–April 30) during baseflow conditions (Kazyak 2000). Sample collection and analysis follows USEPA-approved methods (USEPA 1987; APHA 1998) and is rigorously discussed as they relate to water quality analysis and nutrient concentrations found throughout Maryland’s first- to fourthorder wadeable streams by Morgan and Kline (2011). Although phosphorus is often a factor limiting instream production (King and Richardson 2003; Paul and McDonald 2005; Miltner 2010), the extent of potential environmental problems from excessive nitrogen concentrations are more pervasive in Maryland than from phosphorus (Morgan and Kline 2011). These excessive and increasing nitrogen loads account for a


1169

majority of the mass-balance input of nutrients into Chesapeake Bay and has degraded estuary health (Kemp et al. 2005). Therefore, we focused our analyses on TN and its common forms including, nitrite (NO2N), nitrate (NO3-N), and ammonia-N (NH4-N). Minimum detection levels in water quality analysis for each parameter were 0.0561, 0.0019, 0.0019, and 0.002 mg/L, respectively. Samples of TN and it forms displayed a wide range of concentrations within two statewide sampling periods MBSS Round 2 (2000– 2004) and Round 3 (2007–2009) highlighting the heterogeneity of landscape and nutrient loading regimes found in Maryland (Table 1). Independent variables to assess the relationship between stream community and nutrients included the MBSS benthic macroinvertebrate index of biotic integrity and its taxonomic and functional metrics. We also calculated other community metrics with demonstrated response to nutrients including, Beck’s Biotic Index, percent shredders and percent filters based on the hypothesis that stream eutrophication could alter their relative abundance in the benthic community (Southerland et al. 2007). We used Spearman (rank) correlations to determine the level of association between highly correlated (p1.68 mg/L, out of the total number of sites. Sites with indicator taxa having TN>1.68 mg/L and sites without indicator taxa having TN≤1.68 mg/L make up the remainder of sites. These misclassifications represent Type I and Type II error, respectively. Influence of physiochemical and landscape gradients We employed a pair of analyses in an attempt to distinguish the effects of nutrients from other factors (e.g., water quality or physical habitat) on macroinvertebrate communities. First, we used principal component analysis (PCA) to synthesize abiotic variables into major environmental gradients and describe the variability in conditions across MBSS Round 2 sample sites (McCune and Grace 2002). Chemical variables included TN, NO3-N, NH4-N, conductivity, acid neutralizing capacity (ANC), and chloride (CL). We excluded NO2N from these analyses due to the high percentage of sites where values were below the MDL and because a prior analysis including NO2-N indicated it explained little additional variability in environmental conditions. Physical habitat variables were visually estimated at sites and include riparian buffer width (meter), epifaunal substrate quality (rank, 0–20), riffle-run extent (meter), and riffle embeddedness (percent) (Kazyak 2000). Landscape variables included the amounts of urban, agricultural, and forested lands (percent) in upstream catchments, which were determined for each site by hand delineating catchments from USGS 7.5 minute quarter quad topographic maps in ArcGIS version 9.3 (ESRI, Red Lands, CA) and intersecting them to 30×30 m resolution satellite derived landcover data (2001 NCLD; Homer et al. 2007). Principal component axes were retained by inspection of broken-stick eigenvalues (Jackson 1993). Principal component plots were generated to illustrate patterns along environmental gradients between sites classified as below or above the TN benchmark (1.68 mg/L). We then tested whether the presence of nutrient indicator taxa defined in prior analyses differed within and across major environmental gradients by constructing generalized linear models (MacCullagh and Nelder 1989) using standardized PC scores as explanatory variables. We also created a nutrient sensitivity index by scoring sites in relation to the presence of nutrient indicator taxa as: 0 (tolerant taxa present), 1


(sensitive taxa absent), 2 (tolerant and intolerant taxa present), and 3 (only intolerant taxa present). We only evaluated models with main effects because we observed no significant interactions (p>0.02) between PC axes after adjusting our level of significance for sequential tests. Because we were interested in the overall influence of major abiotic gradients on macroinvertebrate taxa, we did not construct alternative models to evaluate their performance or the relative influence of individual variables amongst one another (e.g., Burnham and Anderson 2002). Simple linear regression models were then fit to predicted values of macroinvertebrate taxa occurrence to estimate the effect size of PC axes. Statistical analyses Prior to analyzing, we examined a variety of high and low order transformations of response (e.g., biological metrics) and explanatory variables (e.g., nutrient parameters) and their effects on variables. We observed marginal improvement in the homogeneity of variance in transformed variables after examining quantile–quantile plots (Easton and McCulloch 1990), but also considerable distortion and little support that assumptions of statistical normality had improved (Kolmogorov–Smirnov tests). Accordingly, we opted to use untransformed data because (1) we were not establishing nutrient benchmarks (Dodds and Welch 2000), (2) the proposed benchmark was derived from transformed data (Morgan and Kline 2011), (3) unlike pH, macroinvertebrates experience nutrients as un-transformed, equal interval values, (4) cumulative frequencies were unaltered by logtransformations (Trebitz 2012), (5) a PCA of transformed data explained little additional variance than a PCA of raw data, suggesting that linear relationships among variables were not improved (McCune and Grace 2002), and (6) the difficulty in translating results from transformed stressor-response data to policy (Trebitz 2012). Stream nutrient concentrations below their respective MDL’s were censored from analyses. All statistical analyses were performed in R v2.15.2 (R Core Development Team 2012).

Results Nutrient–macroinvertebrate relationships All but one of the benthic macroinvertebrate assemblage metrics (% climbers) were significantly correlated with at least


one measure of nutrient concentration (Table 2). The variables with the highest correlation coefficients were generally metrics that measure richness of pollution intolerant macroinvertebrate taxa (e.g., EPT richness). Percentages of Ephemeroptera, clingers, swimmers, and urban intolerants also displayed strong correlation with stream nutrients, as did an index of organic pollution intolerance (Beck’s Biotic Index) and its metrics. Overall, TN concentrations exhibited weak to moderate correlation with measures of benthic macroinvertebrate communities. Nitrite and NH4-N concentration were correlated with most measure of macroinvertebrates, including all correlations where r≥|0.30|. Concentrations of NO3-N were the nutrient measures least associated with benthic community metrics and indexes. The benthic macroinvertebrate assemblage metrics with the highest correlation coefficients consistently displayed a negative response to increasing NO2-N and NH4-N concentration (Fig. 1). At low NO2-N and NH4-N concentrations, macroinvertebrate metrics spanned a wide range of values, but at high concentrations, metric values were overwhelmingly low (i.e., poor condition). Percent of chironimids and dipterans were the only metrics that illustrated positive relationships to increasing nutrient concentrations. Taxa Table 2 Spearman correlation coefficients between benthic macroinvertebrate metrics and nutrient concentrations

Metric

richness and % scrapers illustrated positive relationships with TN and NO3-N, but negative relationships with NO2-N and NH4-N suggesting that TN composition may differentially influence macroinvertebrate communities. Nutrient indicator taxa Five benthic macroinvertebrate taxa identified from samples collected during MBSS Round 2 were classified as intolerant of chronic TN concentrations (Fig. 2a–e). That is, a majority of their occurrences at stream monitoring sites had TN concentrations ≤1.68 mg/L. These taxa are primarily stoneflies and caddisflies that are shredders and predators (Table 3). Two taxa were classified as being tolerant of chronic TN concentrations; a majority of their occurrences were at monitoring sites with TN>1.68 mg/L (Fig. 2f, g). From the MBSS Round 3 macroinvertebrate samples, four taxa were classified as being intolerant of chronic TN concentrations (Fig. 2h–k). This list of nutrient-intolerant taxa shared taxonomic and functional characteristics with the list of taxa developed from the MBSS Round 2 samples (Table 3). A single taxon was classified as being tolerant of chronic TN concentrations (Fig. 2l). The nutrient-tolerant taxa identified from MBSS Rounds 2 and 3 were representative of

TN

NO2-N

NO3-N

NH4-N

0.14

−0.08

0.16

−0.15

EPT taxa richness

−0.04

−0.23

0.03

−0.45

% Ephemeroptera

0.03

−0.13

0.07

−0.38

% Climbers

0.05

0.03

0.04

0.08

% Chironimdae

0.24

0.14

0.22

0.21

−0.07

−0.13

0.01

−0.38

0.13

0.09

0.17

−0.01

Taxa richness

% Clingers % Tanytarsini midges

0.21

−0.01

0.20

−0.20

−0.04

−0.15

−0.01

−0.31

0.15

0.05

0.13

0.18

−0.29

−0.29

−0.21

−0.34

0.02

−0.12

0.05

−0.18

−0.06

−0.01

−0.05

−0.09

0.17

0.01

0.20

−0.16

Class 1 taxa richness

−0.14

−0.32

−0.07

−0.47

Class 2 taxa richness

−0.01

−0.22

0.05

−0.44

% Class 1 taxa

−0.22

−0.35

−0.15

−0.47

% Class 2 taxa

−0.13

−0.25

−0.05

−0.48

Beck’s biotic index

−0.09

−0.28

0.01

−0.46

% Scrapers % Swimmers % Diptera % Intolerant to urban Benthic macroinvertebrate IBI % Shredders % Filterers

Highly significant correlations (p