Long-Term Benthic Macroinvertebrate Community ... - Springer Link

Environmental Management (2011) 47:1077–1095 DOI 10.1007/s00267-010-9610-3

Long-Term Benthic Macroinvertebrate Community Monitoring to Assess Pollution Abatement Effectiveness John G. Smith • Craig C. Brandt Sigurd W. Christensen

•

Received: 30 September 2008 / Accepted: 29 December 2010 / Published online: 28 January 2011 Ó Springer Science+Business Media, LLC (outside the USA) 2011

Abstract The benthic macroinvertebrate community of East Fork Poplar Creek (EFPC) in East Tennessee was monitored for 18 years to evaluate the effectiveness of a water pollution control program implemented at a major United States (U.S.) Department of Energy facility. Several actions were implemented to reduce and control releases of pollutants into the headwaters of the stream. Four of the most significant actions were implemented during different time periods, which allowed assessment of each action. Macroinvertebrate samples were collected annually in April from three locations in EFPC (EFK24, EFK23, and EFK14) and two nearby reference streams from 1986 through 2003. Significant improvements occurred in the macroinvertebrate community at the headwater sites (EFK24 and EFK23) after implementation of each action, while changes detected 9 km further downstream (EFK14) could not be clearly attributed to any of the actions.

The submitted manuscript has been authored by a contractor of the U.S. Government under contract DE-AC05-00OR22725. Accordingly, the U.S. Government retains a nonexclusive, royaltyfree license to publish or reproduce the published form of this contribution, or allow others to do so, for U.S. Government purpose. J. G. Smith (&) Environmental Sciences Division, Oak Ridge National Laboratory, Building 1504, MS-6351, P.O. Box 2008, Oak Ridge, TN 37831, USA e-mail: [email protected] C. C. Brandt Biosciences Division, Oak Ridge National Laboratory, Building 1505, MS-6038, P.O. Box 2008, Oak Ridge, TN 37831, USA S. W. Christensen Environmental Sciences Division, Oak Ridge National Laboratory, Building 2040, MS-6290, P.O. Box 2008, Oak Ridge, TN 37831, USA

Because the stream was impacted at its origin, invertebrate recolonization was primarily limited to aerial immigration, thus, recovery has been slow. As recovery progressed, abundances of small pollution-tolerant taxa (e.g., Orthocladiinae chironomids) decreased and longer lived taxa colonized (e.g., hydropsychid caddisflies, riffle beetles, Baetis). While assessments lasting three to four years may be long enough to detect a response to new pollution controls at highly impacted locations, more time may be needed to understand the full effects. Studies on the effectiveness of pollution controls can be improved if impacted and reference sites are selected to maximize spatial and temporal trending, and if a multidisciplinary approach is used to broadly assess environmental responses (e.g., water quality trends, invertebrate and fish community assessments, toxicity testing, etc.). Keywords Recovery Benthic macroinvertebrates Community structure Biomonitoring Water pollution abatement Long-term study

Introduction Since enactment of the Clean Water Act in the United States (U.S.) in the early 1970s, advancements in pollution control technologies have led to wide-scale improvements in the quality of many waterbodies (Cairns 1990; Karr 1995). Initially it was assumed that these improvements would also improve biological conditions, thus, there was little to no follow-up biomonitoring. However, it was soon realized that these early engineering solutions to improve water quality were not necessarily protective of biological conditions, thus, biological degradation continued. The realization that biological degradation was still occurring

123

1078

eventually led to greater emphasis on biological monitoring, particularly after the 1970s (e.g., Barbour and others 1999; Davis and Simon 1995; Yoder and Rankin 1998). Even with extensive documentation of the importance and benefits of biological monitoring (e.g., Davis and Simon 1995), few studies have specifically assessed the effectiveness of pollution control technologies to protect biological conditions (Bash and Ryan 2002; Bernhart and others 2005; Chapman 1999). Instead, pollution control has typically focused on engineering goals and meeting regulatory requirements for water quality (Bash and Ryan 2002). A majority of the studies published on the biological responses to improvements in environmental conditions lasted less than three years (Yount and Niemi 1990). Studies that track changes for multiple years may experience changes in sampling schedule or methods (e.g., Beard and others 1999; Chadwick and others 1986; Stone and Wallace 1998; Townsend and others 1987). Such inconsistencies can limit comparability of data or inadequately characterize the inherent temporal variability in populations and communities, making it difficult to distinguish between recovery and natural variation (Power 1999; Thrush and others 1994). This increases the likelihood of drawing the wrong conclusions about biological responses to pollution controls (Hewitt and others 2001) and may limit the value of such assessments. When the consequences of pollution controls are not adequately known, effective management of our aquatic resources is not possible (Gore and others 1990; Karr 1995). In 1985, the U.S. Department of Energy’s (DOE’s) Y-12 National Security Complex (referred to hereafter as the Y-12 Complex) in Oak Ridge, Tennessee, was issued a National Pollutant Discharge Elimination System (NPDES) permit that required monitoring of both water quality and biological conditions of the receiving stream, East Fork Poplar Creek (EFPC; Loar and others 1989). Over the next decade and a half, the Y-12 Complex implemented several pollution control measures as part of their Water Pollution Control Program to improve water quality in the receiving stream (Loar and others 2011). In 1985, a Biological Monitoring and Abatement Program (BMAP) was developed to help satisfy the requirements for biological monitoring and evaluate the effectiveness of implementation of the Y-12 Complex’s Water Pollution Control Program. A multidisciplinary strategy was implemented that included monitoring of contaminant exposure, instream and effluent toxicity testing, assessments of various bioindicators of fish health, and monitoring the periphyton, benthic macroinvertebrate, and fish communities. This paper focuses on benthic macroinvertebrate community monitoring studies for evaluating the effectiveness of the Y-12 Complex’s pollution control efforts; results of other BMAP studies have either been published (Hill and others 2010) or they

123

Environmental Management (2011) 47:1077–1095

are included in this special series (Adams and Ham 2011; Greeley and others 2011; Loar and others 2011; Ryon 2011; Southworth and others 2011; Stewart and others 2011). Although several measures were taken to reduce and control releases of pollutants during the study (Loar and others 2011), the volume of the waste streams with each of these measures was usually low relative to the volume of water in EFPC (i.e.,\1%). There were, however, three distinct periods when at least one significant pollution abatement action was completed that either involved the entire stream at the point of the action or had a significant effect on toxicant loading to the stream (Loar and others 2011; Stewart and others 2011). The primary objective of this study was to evaluate the effectiveness of these actions by assessing the response of the benthic macroinvertebrate community in EFPC.

Methods Study Area A detailed description of the study area has been presented elsewhere (Loar and others 2011), therefore, only a brief summary follows. EFPC is in the southern Appalachian Ridge and Valley Physiographic Province of eastern Tennessee (Fig. 1). The stream originates on the DOE’s Oak Ridge Reservation from several small springs within the boundaries of the Y-12 Complex. At their origin, the springs flow into a network of culverts before emerging from a single large culvert (referred to as the North–South Pipe) into a straightened channel that is *5 m wide (Loar and others 2011). The banks along much of first 2 km of stream are steeply incised and stabilized with limestone rip-rap. The substrate in EFPC is typical of streams in the Ridge and Valley ecoregion, consisting predominantly of a mixture of cobble, gravel, and sand, with occasional outcrops of bedrock (ftp://ftp.epa.gov/wed/ecoregions/tn/tn_back.pdf). Until July 1998, all stream flow passed through a settling basin *1.5 km downstream of the North–South Pipe. In 1985 there were more than 200 outfalls discharging from the Y-12 Complex into EFPC upstream of the basin, and fewer than five storm drains from the facility discharging downstream of the basin. EFPC exits the Y-12 Complex’s boundaries about 1 km downstream of the basin, and then flows through the city of Oak Ridge for *15 km before reentering the Oak Ridge Reservation. Approximately 25 km downstream of the North–South Pipe, EFPC flows into Poplar Creek, a tributary to the lower Clinch River. In 1982, it was estimated that *52% of the land in the headwaters of EFPC encompassed by the Y-12 Complex was impervious surfaces such as buildings, concrete, and asphalt (Pritz and Sanders 1982). The limited vegetation consists primarily of scattered plots of maintained grasses


1079

Fig. 1 Map of East Fork Poplar Creek and reference sampling sites. The ‘‘Settling Basin’’ represents the approximate location of New Hope Pond and its replacement, Lake Reality, which was located *30 m north of New Hope Pond

and some trees, and the riparian vegetation consists of narrow strips (generally \5 m) of mixed shrubs, vines, and small trees. Land use downstream of the Y-12 Complex includes extensive commercial and residential development, light industry, small plots of pastureland, and variously-sized stands of secondary growth forests largely restricted to nearby ridge-tops and immediate slopes. The benthic macroinvertebrate community was monitored at three sites in EFPC (Fig. 1). The two most upstream sites in EFPC (EFK24 and EFK23; EFK = EFPC kilometer from the mouth of the stream) were second-order reaches located immediately adjacent to the Y-12 Complex. Site EFK24 was approximately 500 m upstream of the settling basin (New Hope Pond/Lake Reality), and downstream of all but 15 of the effluent outfalls present in the mid-1980s. Site EFK23 was approximately 300 m downstream of settling basin, as well as, all effluent discharges. The third site, EFK14, was a fourth order reach in the city of Oak Ridge about 9 km downstream of EFK23. Some unique characteristics of EFPC made selection of a suitable reference site a challenge. An undisturbed area upstream of an impacted zone is often used as a reference (Whittier and others 2007), but EFPC is impacted at its origin making it necessary to select reference sites on other streams (Loar and other 2011; Stewart and others 2011). The large number of effluents discharged into the headwaters of EFPC increased flow by a factor of almost 5x, making the discharge in EFPC more similar to streams with

much larger watersheds (Loar and others 2011; Stewart and others 2011). With the volume of water added to EFPC from Y-12 Plant operations, the discharge in EFPC was more similar to streams with much larger watersheds. With limitations on the number of reference sites that could be included in the study, a decision had to be made on selecting sites that more closely matched watershed area (e.g., *3 km2 at the headwater sites) or stream discharge and other physical characteristics (e.g., similar channel width, depths, etc.). Another factor in the selection of reference sites was the desire to maximize integration of results among as many of the other BMAP studies as possible, thus, a common location was preferred. Because the long-term strategy for controlling inputs of pollutants to EFPC did not include actions that would significantly reduce the discharge, reference sites were selected on two nearby streams that received no industrial effluents, were comparable in physical size (e.g., channel width and depth), and had relatively comparable discharges (Table 1; Loar and others 2011). Initially, a single reference site on a fourth order reach in nearby Brushy Fork (BFK7) was used for all studies (Fig. 1). A fourth order reach on second stream, Hinds Creek (HCK20), was later included as a reference site for two reasons (Fig. 1). First, polychlorinated biphenyls were detected in fish from Brushy Fork making it unsuitable for contaminant monitoring (Loar and others 1992). Second, it was felt that having two reference sites would provide a better estimate of natural variation in

123

1080


Table 1 Chemical and physical characteristics for sampling sites on East Fork Poplar Creek, Brushy Fork and Hinds Creek Parameter

Site EFK24

EFK23

EFK14

BFK7

HCK20

Riffle depth (cm)

17.2 (9–34)

17.1 (7–29)

18.2 (4–40)

20.4 (3–47)

17.8 (6–38)

Width (m)

4.9

5.1

7.7

8.9

8.0

1989–1992 (PAP1)

20.6 (14.7–26.4)

19.8 (12.1–27.5)

15.8 (5.8–24.6)

14.1 (5.0–21.1)

–c

1993–1996 (PAP2)

19.6 (13.3–25.6)

18.5 (10.0–27.4)

14.2 (6.6–25.1)

13.7 (7.2–21.5)

13.8 (5.4–22.2)

1997–2003 (PAP3)

16.9 (11.4–21.4)

17.2 (10.8–23.0)

15.8 (7.2–23.9)

13.9 (4.7–20.9)

14.0 (5.5–22.0)

9.6 (7.1–12.5) 8.1 (7.7–9.1)

10.1 (6.7–15.2) 8.2 (7.0–9.1)

10.0 (7.7–13.8) 8.0 (7.0–9.2)

9.5 (6.6–13.2) 7.9 (7.5–8.8)

10.0 (7.5–13.2) 8.2 (6.9–8.9)

Temperature (°C)a

Dissolved oxygen (mg/L) PH (SU) Conductivity (lS/cm) Study average

370 (165–880)

361 (110–730)

304 (120–520)

211 (90–332)

257 (160–376)

1986–1988 (PAP0)

361

347

299

226

244

1989–1992 (PAP1)

441

401

310

194

271

1993–1996 (PAP2)

403

433

335

213

267

1997–2003 (PAP3)

265

270

274

201

247

EFK East Fork Poplar Creek kilometer, BFK Brushy Fork kilometer, HCK Hinds Creek kilometer Unless otherwise noted values are means based on data collected concurrently with invertebrate samples from 1986 to 2003. Values in parentheses are ranges a

Average annual temperatures collected using Peabody Ryan Model J90 thermographs (MG Ryon and WK Roy, ORNL, unpublished data); range of average monthly temperatures in parentheses. Temperature data were not available during the baseline period. (PAP0) Data were not available for HCK20 in PAP2

area streams (Whittier and others 2007). Neither Brushy Fork, nor Hinds Creek was located on the Oak Ridge Reservation, and neither stream received industrial discharges. Both streams were in rural watersheds of the Ridge and Valley Physiographic Province, and like EFPC they were in the lower Clinch River drainage. The Brushy Fork site was approximately 6 km north of the Y-12 Complex, and the Hinds Creek site (HCK20) was about 28 km northeast of the Y-12 Complex. Land use in both watersheds was predominately a mixture of low-density housing, agriculture (primarily pastureland and hay production), and mixed second-growth hardwood forests. The substrate at both sites was predominantly a mixture of cobble, gravel, sand and silt, and sedimentation was evident at both sites. Given their location in rural watersheds, the reference sites did not represent pristine conditions. Rather they represented conditions that might have been expected in EFPC had the Y-12 Complex not been present. All EFPC and reference sites were well oxygenated and had moderately basic pH (Table 1). Conductivity was generally highest at the EFPC sites, particularly from 1985 through 1996, but after 1996, conductivity at these sites was usually only slightly higher than at the reference sites. From 1989 to 1996, temperatures at EFKs 24 and 23 were elevated by *5°C relative to the reference sites, but the difference dropped to *3°C after 1996 (see Loar and others 2011 for further detail). Temperature at EFK14

123

exhibited no major changes during the study, and was generally comparable to or only slightly higher than the temperatures at the reference sites. Pollution Abatement Actions Only a brief description of the actions assessed in this study is provided; a more detailed account of these actions and other pollution abatement actions taken by the Y-12 Complex is given in Loar and others (2011). Only four of the most significant actions were evaluated because it was felt that they had the greatest potential for causing broad-scale changes in water quality (e.g., settling basin replacement, flow management) or reducing in-stream toxicity (e.g., effluent dechlorination). Several other pollution abatement actions were completed during the study but not assessed. These actions were either completed gradually over an extended period of time (e.g., consolidation and elimination of multiple effluent discharges), or their contribution to the total stream flow was small (i.e., generally B1% of the total stream flow); thus, any effects from these individual actions would have been difficult to isolate and detect. The evaluated abatement actions were completed in three distinct time periods separated by at least 4 yr each. Hereafter, these periods will be referred to a pollution abatement periods or PAPs (Table 2). The period from 1985 through 1988, when only a few minor actions were


1081

Table 2 Pollution abatement periods assessed with benthic macroinvertebrate community structure Pollution abatement period identifier

Period

Action

Completion date

Purpose

PAP0

1986–1988

None

–

Baseline period

PAP1

1989–1992

Replace New Hope Pond settling basin with Lake Reality

November 1988

Eliminate historical sources of contaminants

PAP2

1993–1996

Effluent dechlorination

November 1992

Reduced chlorinated inputs (biocide) to stream

PAP3

1997–2003

(1) Flow management and (2) settling basin bypass

(1) January 1997 and (2) July 1998

(1) Toxicity reduction and maintenance of minimum flows, and (2) eliminate basin as source of methyl mercury

implemented, served as the baseline period (i.e., PAP0). The first abatement period (PAP1) included 1989 through 1992, which followed replacement of the original settling basin (New Hope Pond) with a new basin (Lake Reality) in November 1988. The second period (PAP2) followed dechlorination of several major effluent discharges upstream of EFK24. This action was completed in November 1992, and the period of assessment was 1993 through 1996. Two separate actions were assessed in the third period (1997–2003, PAP3) including flow management and closure of Lake Reality (i.e., flow bypass). Flow management began intermittently in July 1996 before being fully implemented in January 1997. This action involved pumping uncontaminated water from the nearby Melton Hill Reservoir into EFPC through a rip-rap lined channel *1 km upstream of EFK24. The added water increased the discharge in EFPC by *60%. The diversion of flow around Lake Reality began on an intermittent basis in mid-1996, before becoming permanent in July 1998, at which time, the use of a settling basin ceased.

Data Analysis To account for changes in the taxonomy of some taxa (e.g., Baetidae) and changes in the level of identification of chironomids, abundances for these taxa were pooled into lower taxonomic levels where standardization was needed across all years (e.g., taxa richness metrics, multivariate analyses). Additionally, while the initiation, rate, and magnitude in the response of the invertebrate community to abatement actions showed distinct seasonal differences, similar conclusions concerning the effectiveness of each action were reached regardless of season. Therefore, only the results from April sampling periods are included. For all multivariate analyses, unless otherwise noted, the BrayCurtis similarity measure was used on transformed abundances (log10 X ? 1). The Bray-Curtis similarity measure is one of the most robust resemblance measures for data sets with zeros, and is therefore generally used with community data (Legendre and Legendre 1998). General Spatial and Temporal Trends

Sample Collection and Processing Benthic macroinvertebrate sampling began in 1985. Initially, samples were collected monthly and then reduced to biannual in 1989 (i.e., April and October). A permanent riffle at each site was delineated into a numbered grid representing sampling cells, and on each sampling date, five cells (replicates) were selected randomly for sampling. Samples were collected with a Hess sampler (0.086 m2, 368 lm mesh net) to a depth of *10 cm. Excess water was drained from the samples, placed into glass jars, and preserved with 95% ethanol in glass jars. In the laboratory, samples were sorted, identified to the lowest practical taxon, and enumerated. Most insects, crustaceans, and mollusks were identified to genus, but some taxa were identified to class or order only (e.g., Oligochaeta, Nematoda). Chironomids were initially identified to genus, but in the early 1990s they were identified to only subfamily or tribe to expedite sample processing and to reduce costs.

Nonmetric multidimensional scaling (NMDS) ordination was used to visualize spatial and temporal trends in assemblage structure among sites and pollution abatement periods. Transformed abundances were averaged by site and sampling date, and then fifty runs of the NMDS algorithm were used to derive the final solution (PRIMER software, version 5.2.9, Clarke and Warwick 2001). Community convergence (i.e., increased similarity between two communities through time) of each EFPC site with the Brushy Fork reference site was assessed with a Spearman rank correlation analysis across time (all years) on temporally-paired Bray-Curtis similarity values (Philippi and others 1998). HCK20 was not included in comparisons with EFPC sites because of insufficient data in the baseline period (i.e., PAP0). To better gauge whether trends were associated with changes at an EFPC site or BFK7, convergence also was assessed between the reference sites with data from 1988 through 2003. Because similarity values are

123

1082

not independent, probability values were calculated by a randomization procedure (Legendre and Legendre 1998). The PROC FREQ procedure in SAS (SAS Institute Incorporated, Cary, North Carolina, PC SAS Version 9.1) was used for the correlation analysis and calculation of probability values with a two-sided Monte Carlo randomization test (10,000 permutations). A significant (P \ 0.05) positive correlation signified convergence (i.e., increasing similarity) and implied recovery, while a significant negative correlation provided evidence of divergence (i.e., decreasing similarity). An insignificant correlation indicated no change, and was the anticipated response for the reference sites. As a site recovers from a disturbance, taxonomic changes occur and community structure may become increasingly dissimilar from the initial state (Philippi and others 1998). An analysis of this change (i.e., progressive change) at an impacted site can be used as evidence of recovery. Progressive change within each EFPC and reference site was evaluated as change from the initial sampling period (1986 for EFPC sites and BFK7, and 1988 for HCK20). Similarity values for each site were calculated between the initial sampling period and each of the other sampling periods. The correlation between time and the resulting similarity values was then tested with a Spearman’s rank correlation analysis. A significant negative correlation (P \ 0.05) would indicate that the community structure at a site was becoming increasingly different from the initial sampling period, and therefore, could signify that conditions were either improving or becoming worse. If species composition and community structure remained relatively stable, as anticipated at a reference location or a study site unaffected by pollution abatement, then a significant correlation would not be expected. The SAS PROC FREQ procedure was used for the correlation analysis and to calculate P-values with a two-tailed Monte Carlo test and 10,000 permutations. Analysis of Pollution Abatement Actions A before-after control-impact (BACI) paired-site approach (Smith 2002; Stewart-Oaten and others 1986) was used to test for changes in total taxonomic richness (i.e., average number of taxa/sample), Ephemeroptera, Plecoptera, and Trichoptera (EPT) taxonomic richness, and EPT density during each of the three major abatement periods. For this analysis, the differences between mean reference- and impact-site values were paired in time, and the resulting differences for the baseline period (PAP0) were compared with those in each pollution abatement period with a t-test (PROC TTEST, SAS Institute Incorporated, Cary, North Carolina). HCK20 was excluded from the analyses because samples were collected just once during the baseline period. Data from the baseline period for all sites were tested for

123


conformance with statistical assumptions of independence, additivity, normality, and homogeneity of the variance (Quinn and Keough 2002). When assumptions were not met, a suitable transformation was identified and applied. A one-way analysis of similarity (ANOSIM) was used to test for within-site changes in community similarity between the baseline and abatement periods (PRIMER software, Clarke and Warwick 2001); PAP1 was used as the baseline period for HCK20. ANOSIM is a non-parametric multivariate permutation procedure that tests the null hypothesis of no difference in community composition between, for example, sites or events within a site. The analysis uses a similarity matrix to compare ranks of between-group similarities to within-group similarities. The resulting test statistic, R (range = -1 to 1), provides an estimate of the strength of group separation. When R = 1, then all within-group samples are more similar than any between-group samples, and when R = 0, then withinand between-group samples are the same on average. The R value can be used to compare the magnitude of separation among test groups, and sometimes is a better estimate of group separation than the probability value because as the number of replicates increases, the chances of producing a significant effect increase (Clarke 1993). The overall ANOSIM test for period differences (i.e., global test) for each site was followed by pair-wise tests to identify the abatement periods that differed from the baseline. A Monte Carlo test with 1000 permutations was used to calculate P-values for the global test, and the maximum number of permutations possible was used (i.e., 35–120) for pair-wise tests among abatement periods. Taxonomic Changes A similarity percentage analysis (SIMPER) was used to help identify important taxonomic trends in each abatement period (PRIMER software, Clarke and Warwick 2001). SIMPER identifies the taxa most responsible for changes in community similarity, and thus, provided information on chronological changes in taxonomic composition as recovery progressed. In this analysis, taxa were considered good discriminators between comparison groups (e.g., PAP0 vs. PAP1) when they consistently contributed the most to between-group differences in dissimilarity.

Results General Temporal and Spatial Trends in Community Structure In PAP0, total and EPT richness were lower at EFKs 24 and 23 than at EFK14 and the reference sites (Fig. 2). With


1083

Number of taxa/sample

50

PAP0

PAP1

PAP2

PAP3

40

30

20

10

Number of EPT taxa/sample

0 25

20

15

10

5

0

EPT density (no./0.1 m2)

500 EFK24 EFK23 EFK14 Reference 95% CI

400

300

200

100

0 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03

Year Fig. 2 Temporal trends in means for taxonomic richness, EPT richness, and EPT density. Vertical bars for EFPC sites are upper 95% confidence limits; gray shading is 95% confidence intervals for reference sites calculated from all sample units from both reference sites except for 1986 and 1987 which are based on Brushy Fork data only

few exceptions, the upper 95% confidence limits of both richness metrics for EFKs 24 and 23 did not overlap with the lower confidence limits for the reference sites. Total and EPT richness varied considerably at EFK14, but mean values were usually within or just outside of the reference site confidence intervals. From 1986 through 1994, EPT density was virtually zero at EFK24 (Fig. 2). Even after 1994, EPT density at EFK24 was always less than the lower 95% confidence limits for the reference sites, and with few exceptions the upper 95% confidence limits did not overlap with the lower confidence limits for the reference sites. Temporal trends in EPT density at EFK23 and the magnitude of difference from the reference sites were similar to those at EFK24. Mean EPT density at EFK14 in contrast, was usually within the 95% confidence interval for the reference sites

and even exceeded the upper confidence limit on some sampling dates (Fig. 2). Spatial differences and temporal changes in assemblage structure were clearly demonstrated with the NMDS ordination (Fig. 3). EFKs 23 and 24 were distinctly different from the reference sites, while the difference between EFK14 and the reference sites was much less. Additionally, EFKs 23 and 24 exhibited a general unidirectional shift towards EFK14 on the plot (i.e., left to right along Axis 1), indicating that their assemblage structures were becoming increasingly dissimilar with time. This shift also indicated that their assemblages were becoming more similar to the assemblage at EFK14. Analysis of convergence indicated that the community at all three EFPC sites became more similar to the community at BFK7 during the study (Fig. 4). Convergence was greatest at EFKs 24 and 23 (P B 0.01, rs = 0.84 and 0.70 respectively), although maximum similarity between these two sites and BFK7 never exceeded the minimum estimate of similarity between BFK7 and HCK20. Convergence between EFK14 and BFK7 was significant (P = 0.02) but less pronounced (rs = 0.54) than at the other two EFPC sites. BFK7 and HCK20 showed no persistent changes in similarity (P [ 0.9, rs = 0.03). Analysis of progressive change demonstrated that community composition and structure at EFKs 24 and 23 changed significantly through time (P \ 0.001, rs = -0.84 and -0.83 respectively; Fig. 5). Although similarity with the baseline year declined noticeably at EFK14, the trend was not significant (P [ 0.17, rs = -0.34), while no discernable trends were evident for either reference site. Period-Specific Trends in Community Structure Changes in community metrics were detected at some sites after abatement periods (Table 3; Fig. 2). Total richness increased significantly in all pollution abatement periods at EFKs 24 and 23. Except for PAP1 at EFK24, this same trend was also observed for EPT richness and EPT density. Fewer significant changes were detected at EFK14 during the study (Table 3; Fig. 2). Total richness increased significantly in PAP1 and PAP2 but not in PAP3. Changes in EPT richness were similar to those for total richness except that the difference was significant in PAP1 only. No significant change was detected in EPT density in any abatement period. Comparisons of similarities (ANOSIM) between the baseline period and each abatement period within each site showed that with few exceptions, community composition at EFKs 24 and 23 became more dissimilar in each consecutive abatement period (Table 4). At EFK24, the decrease in similarity was significant in all abatement periods. At EFKs 23 and 14 the decrease in similarity was

123

1084


Stress=0.12

Axis 2

EFK24

EFK24 EFK23 EFK14

PAP0 PAP1 PAP2 PAP3

BFK7 HCK20

EFK14

Axis 2

EFK23

BFK7

Axis 2

HCK20

Axis 1

Axis 1

Fig. 3 Non-metric multidimensional scaling plots of benthic macroinvertebrate community data for EFPC and reference sites. The upper left graph includes all sites and the estimate of stress for the analysis (a estimate of how well the analysis describes patterns from

the data set), and the other five graphs are site-specific plots from the same analysis to better show temporal trends. The key to symbols used in the site-specific plots is shown in the graph for EFK24. All graphs are plotted on the same scale

not significant in PAP1, but strong changes were still evident (R = 0.5 and 0.54 for EFKs 23 and 14 respectively). A significant decrease in similarity was detected at BFK7 in PAP3, but no significant changes were detected at HCK20.

The community at EFK24 shifted from almost exclusively chironomids (primarily Orthocladiinae) to a more diverse assemblage (Table 5). Planarians contributed to change in similarity at EFK24 in PAP2, while in PAP3 several taxa were responsible for changes in similarity including net spinning caddisflies (Cheumatopsyche and Hydropsyche), beetles (Psephenus and Stenelmis), and the chironomid tribes Chironomini and Tanytarsini. The significant change in similarity at EFK23 in PAP1 was primarily associated with an increase in the abundance of the limpet, Ferrissia, and a decrease in the density of Tanypodinae chironomids (Table 5). After PAP1, the density of Ferrissia decreased and remained lower for the rest of the study (Table 5). In PAP2, there were several taxa that contributed to changes in similarity at EFK23, including planarians, Asiatic clams (Corbicula),

Taxonomic Changes The taxa contributing the most to within-site changes in similarity between the baseline period (PAP0) and each abatement period at the EFPC sites varied considerably (Table 5). At EFKs 24 and 23, there were eight and ten taxa, respectively, that contributed significantly to change in community similarity. In most instances, the contributions by these taxa reflected increases in their abundances, but abundances of a few taxa decreased.

123


1085

80

100

60

EFK24

EFK24 Similarity value

Similarity value

rs = 0.84

p < 0.01

40

20

80

r = -0.84 s p < 0.01

60

40

20 0

100

EFK23

Similarity value

rs = 0.70

EFK23

p = 0.01 60

40

Similarity value

80

20

80

r = -0.83 s p < 0.01

60

40

20 100

EFK14

Similarity value

rs = 0.54

EFK14

p = 0.02 60

40

Similarity value

0 80

80

r = -0.34 s p = 0.17

60

40

20 20

100

60

HCK20

40

80

60

r = 0.19 s p = 0.47

40

20

20

100

rs = 0.03

HCK20

p = 0.93

0 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03

Year Fig. 4 Convergence in benthic macroinvertebrate community similarity of EFPC sites and the Hinds Creek reference control site with Brushy Fork. rs = Spearman Rank Correlation coefficient. n = 18

Similarity value

Similarity value

80

Similarity value

BFK7 0

80

60

r = -0.29 s p = 0.29

40

20 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03

Year

Chironomini and Tanytarsini. Most of these same taxa also were major contributors to the changes in PAP3, but higher densities of Cheumatopsyche, Hydropsyche, and Nemertea also contributed to the change in similarity. Several taxa contributed to changes in similarity at EFK14 during the study, but they were more subtle and complex than at the upstream sites (Table 5). Oligochaetes contributed considerably to differences in all abatement periods. In PAP1 and PAP2 oligochaete density increased dramatically but dropped 89% in PAP3, although it was

Fig. 5 Progressive change in benthic macroinvertebrate community similarity within each site using the first sampling period as the baseline. rs = Spearman Rank Correlation coefficient. n = 18 for EFPC sites and BFK7; n = 15 for HCK20

still higher than in PAP0. Blackflies (Simulium) also were important contributors to change in PAP1, but their abundances returned to baseline values after PAP1. Declines in density of Hydropsyche and Baetis after PAP1 contributed to the change in similarity in PAP2 and PAP3, and a large

123

1086 Table 3 Results of the beforeafter-control-impact analysis for total richness, EPT richness, and EPT density


Site/period

df

Total richness t-value

P-value

df

EPT richness

df

t-value

P-value

EPT density t-value

P-value

EFK24

Comparisons were with the Brushy Fork reference control site. Significant results (P \ 0.05) are in bold df, Degrees of freedom a

X0.4 transformation; b Log10(X ? 1) transformation; c Square root (X) transformation; d No transformation; e X0.25 transformation Table 4 Analysis of similarity (ANOSIM) statistics for withinsite comparisons of the baseline period (PAP0) and each pollution abatement period

PAP1

5a

5.28