Coastal Wetland Management and Restoration: Importance of Abiotic Factors and Vegetation for Healthy Fish Communities in the Laurentian Great Lakes

Abstract

Coastal wetlands in the Laurentian Great Lakes of North America are under increasing stress due to numerous threats. Restoration and management of the remaining wetlands are necessary to ensure that ecosystem functions, critical for fisheries, persist. This study used long-term monitoring datasets for one of the Laurentian Great Lakes, Lake Ontario, including 138 sampling events from 31 different wetlands, to examine the relationship between fish community health and select abiotic and vegetation habitat variables. Eight of 13 habitat variables were found to have significant relationships with fish community health, including total, submerged, and emergent vegetation; submerged aquatic vegetation IBI; water depth; turbidity; conductivity; and water-quality index. Ranges for each significant variable were summarized for each fish community health group to provide guidance when diagnosing impairment or setting restoration goals. An ordination of the fish and environmental data revealed high amounts of variation at sites with poor fish community health relative to excellent health, suggesting a multimetric approach provides valuable insight into community variability. The results from this study provide additional information and alternative methods for assessment of current conditions, target setting, and restoration success assessment for coastal wetland managers.

1. Introduction

Wetlands are one of the most productive ecosystems on Earth, supporting a disproportionately high amount of ecosystem services and species richness, including at-risk species of migratory birds, fishes, amphibians, and plants [1,2,3]. Laurentian Great Lakes fish communities are dependent on coastal wetlands, as the majority of species require wetland habitat for at least a part of their life cycle [2,4,5]. At-risk species, such as Grass Pickerel (scientific names provided in

Table A1. List of all fishes included in this study with family, common, and scientific names. Sensitive or tolerant rating was based on Mandrak and Bouvier [50]. Dependence on vegetation was based on Portt et al. [52]. Dependence was ranked as either H—high, M—medium, L—low, or left blank if no determination was made.

Family	Common Name	Scientific Name	Sensitive/Tolerant	Vegetation
Clupeidae	Alewife	Alosa pseudoharengus	Tolerant
Anguillidae	American eel	Anguilla rostrata	Sensitive
Fundulidae	Banded Killifish	Fundulus diaphanus	Tolerant	H
Centrarchidae	Black Crappie	Pomoxis nigromaculatus	Tolerant	H
Cyprinidae	Blackchin Shiner	Miniellus heterodon	Sensitive	H
Cyprinidae	Blacknose Dace	Rhinichthys atratulus	Tolerant
Cyprinidae	Blacknose Shiner	Notropis heterolepis	Sensitive	H
Centrarchidae	Bluegill	Lepomis macrochirus	Tolerant	H
Cyprinidae	Bluntnose Minnow	Pimephales notatus	Tolerant
Amiidae	Emerald Bowfin	Amia ocellicauda	Tolerant	H
Cyprinidae	Bridle Shiner	Notropis bifrenatus	Sensitive	H
Atherinopsidae	Brook Silverside	Labidesthes sicculus	Tolerant	M
Gasterosteidae	Brook Stickleback	Culaea inconstans	Tolerant	H
Ictaluridae	Brown Bullhead	Ameiurus nebulosus	Tolerant	M
Salmonidae	Brown Trout	Salmo trutta	Sensitive
Umbridae	Central Mudminnow	Umbra limi	Tolerant	H
Salmonidae	Chinook Salmon	Oncorhynchus tshawytscha	Sensitive
Salmonidae	Coho Salmon	Oncorhynchus kisutch	Sensitive
Cyprinidae	Common Carp	Cyprinus carpio	Tolerant
Cyprinidae	Common Shiner	Luxilus cornutus	Tolerant
Cyprinidae	Creek Chub	Semotilus atromaculatus	Tolerant	H
Cyprinidae	Emerald Shiner	Notropis atherinoides	Tolerant
Cyprinidae	Fathead Minnow	Pimephales promelas	Tolerant
Sciaenidae	Freshwater Drum	Aplodinotus grunniens	Tolerant
Clupeidae	Gizzard Shad	Dorosoma cepedianum	Tolerant	L
Cyprinidae	Golden Shiner	Notemigonus crysoleucas	Tolerant	H
Cyprinidae	Goldfish	Carassius auratus	Tolerant	H
Esocidae	Grass Pickerel	Esox americanus vermiculatus	Sensitive	H
Centrarchidae	Green Sunfish	Lepomis cyanellus	Tolerant	H
Percidae	Johnny Darter	Etheostoma nigrum		M
Centrarchidae	Largemouth Bass	Micropterus nigricans	Tolerant	H
Percidae	Logperch	Percina caprodes	Tolerant	H
Cyprinidae	Longnose Dace	Rhinichthys cataractae	Tolerant
Esocidae	Northern Pike	Esox lucius	Tolerant	H
Centrarchidae	Pumpkinseed	Lepomis gibbosus	Tolerant	H
Salmonidae	Rainbow Trout	Oncorhynchus mykiss	Sensitive
Centrarchidae	Rock Bass	Ambloplites rupestris	Tolerant	M
Gobiidae	Round Goby	Neogobius melanostomus	Tolerant
Catostomidae	Silver Redhorse	Moxostoma anisurum	Tolerant
Centrarchidae	Smallmouth Bass	Micropterus dolomieu	Tolerant
Cyprinidae	Spotfin Shiner	Cyprinella spiloptera	Tolerant	H
Cyprinidae	Spottail Shiner	Hudsonius hudsonius	Tolerant	H
Centrarchidae	Sunfish species	Lepomis sp.
Percidae	Tessellated darter	Etheostoma olmstedi	Tolerant
Percidae	Walleye	Sander vitreus	Tolerant	H
Moronidae	White Bass	Morone chrysops	Tolerant
Moronidae	White Perch	Morone americana	Tolerant
Catostomidae	White Sucker	Catostomus commersonii	Tolerant	M
Percidae	Yellow Perch	Perca flavescens	Tolerant	H

) and Pugnose Shiner, are particularly dependent on high-quality wetland habitat [6,7]. Recreational and commercial fishes, such as Largemouth Bass, Walleye, and Yellow Perch, are also dependent on wetland habitat for many parts of their life history [8]. Half of the commercial and approximately 80% of the recreational fish harvest in the Great Lakes is comprised of wetland using species [9]. In addition, wetlands also improve water quality, provide flood control and groundwater recharge, mitigate the effects of climate change, reduce erosion, absorb and provide nutrients, and store carbon [2,10,11].

Despite the importance of their functions, wetlands continue to be lost. The major reason for wetland function loss historically has been conversion to other land uses [1,11,12,13,14,15,16]. Coarse estimates indicate that the total wetland loss in Canada has been 200,000 km² since the 1800s [10]. Mitsch and Gosslink [17] reported that more than 60% of the coastal wetlands in the Great Lakes had been lost. Penfound and Vaz [16] found that there has been an estimated loss of more than 68% of wetlands in southern Ontario since the establishment of European settlers, and Whillians [18] found a 75–100% loss of coastal wetlands along the shore of Lake Ontario with heavily settled adjacent and upstream areas. Currently, a main reason for loss of wetland functions is intensifying anthropogenic pressures on natural ecosystems [12,13,16]. These threats include hydrological alterations, invasive species, adjacent land use, watershed inputs, pollution, and climate change [19,20,21,22,23]. As a result, many coastal wetlands within Lake Ontario, especially within the Greater Toronto Area (GTA), have significantly poorer water quality compared to other less-populated areas in the Great Lakes [20,24].

Given how few coastal wetlands remain in the Great Lakes, management and restoration of those still intact is critical [12,13], and several pieces of Canadian federal and provincial legislation highlight this need [1,23,25]. Numerous agencies are undertaking significant efforts to manage and restore coastal wetlands, but challenges remain when attempting to diagnose causes of impairment, determine current conditions, and set targets for post-restoration monitoring. Several studies have examined species-specific relationships [26,27] and broad, large-scale drivers of Great lakes wetland fish assemblages [28,29,30,31], but few have identified ecologically important ranges of abiotic and vegetation conditions required for healthy fish communities in coastal wetlands. Understanding the impacts of the numerous threats to wetlands can be difficult due to synergistic and cumulative effects and, for that reason, using biological indicators as a proxy for functions or health is often recommended [5,32,33]. Biotic integrity is one type of community health measure that has been defined [34], and it was the basis for the development of the coastal wetland health ratings referred to in this paper. Environment Canada and Central Lake Ontario Conservation Authority [35] provided details on defining health, and background on the development of an index of biotic integrity for Lake Ontario coastal wetlands examined in this study.

This study examines how changes in wetland abiotic and vegetation conditions cause shifts in fish community composition and health using data collected by two integrated monitoring programs in Lake Ontario coastal wetlands for over 10 years. The first goal of the study was to determine which abiotic and vegetation variables were correlated to fish community health, and the ranges of abiotic and vegetation variables within each of the Index of Biotic Integrity (IBI) groupings. A key outcome of this study was to identify abiotic and vegetation conditions required for fish community health, in order to inform management and restoration practices. The second goal of the study was to examine the influence of abiotic and vegetation variables on fish community health using ordination analysis. Previous studies have demonstrated that, as certain abiotic and vegetation variables change, fish communities will respond accordingly [4,28,36,37,38,39]. This can be demonstrated through a health score (e.g., IBI) or as variation in an ordination analysis (e.g., reference condition approach (RCA); refs. [40,41,42]. New approaches combining multimetric and multivariate analyses are becoming more popular, as they draw on the strengths of each [40,43]. This study will provide information for the management and restoration of coastal wetlands in Lake Ontario and elsewhere, which will result in more effective enhancement of wetland functions and fish community health.

2. Materials and Methods

2.1. Study Area

The study area included two regions on the north shore of Lake Ontario (Figure 1,

Table A2. A summary of all wetlands included in this study. Years of sampling was included in the study, as well as decimal degree location information.

Wetland Name	Region	Sampling Years Included in study	Latitude	Longitude
Airport Creek Marsh	Quinte	2010, 2012, 2015	44.176117	−77.096944
Big Island East Marsh	Quinte	2008, 2013	44.132709	−77.191766
Big Island West Marsh	Quinte	2008, 2012, 2015	44.094032	−77.258117
Blessington Creek Marsh	Quinte	2009, 2013	44.164071	−77.321675
Bowmanville Marsh	Durham	2007–2015	43.889873	−78.669071
Carnachan Bay Marsh	Quinte	2009, 2011, 2014	44.074796	−77.021394
Carruthers Creek Marsh	Durham	2009, 2011, 2015	43.828947	−78.986487
Carrying Place Marsh	Quinte	2009, 2014	44.054563	−77.572057
Corbett Creek Marsh	Durham	2007, 2009–2010, 2012–2015	43.854242	−78.889110
Dead Creek Marsh	Quinte	2010, 2012, 2015	44.067101	−77.600465
Duffins Creek Marsh	Durham	2010, 2011, 2014–2014	43.820074	−79.035843
Frenchman’s Bay Marsh	Durham	2007, 2009–2012, 2014–2015	43.815726	−79.090901
Hay Bay Marsh North	Quinte	2008	44.177803	−76.932037
Hay Bay Marsh South	Quinte	2008, 2014	44.160020	−76.885154
Hydro Marsh	Durham	2007, 2011, 2012, 2014, 2015	43.814208	−79.076567
Lower Napanee Marsh	Quinte	2010, 2011, 2014	44.200906	−76.997626
Lower Sucker Marsh	Quinte	2010	44.167426	−77.131410
Lynde Creek Marsh	Durham	2007–2011, 2013–2015	43.848623	−78.954999
McLaughlin Bay	Durham	2008–2011, 2013–2015	43.870368	−78.795776
Oshawa Harbour Wetland Complex	Durham	2009–2010, 2012–2013, 2015	43.868063	−78.825765
Oshawa Second Marsh	Durham	2007–2011, 2013–2015	43.873219	−78.811350
Port Newcastle Marsh	Durham	2007–2012, 2014–2015	43.897094	−78.577324
Pumphouse Marsh	Durham	2011, 2013–2015	43.858643	−78.840447
Robinson Cove	Quinte	2009, 2012, 2015
Rouge River Marsh	Durham	2007, 2009–2012, 2014–2015	43.793963	−79.121589
Sawguin Creek Central Marsh	Quinte	2008, 2012, 2015	44.121685	−77.326227
Sawguin Creek Ditched Marsh	Quinte	2009, 2013	44.086974	−77.335135
Sawguin Creek North Marsh	Quinte	2010, 2011, 2014	44.133841	−77.371001
Westside Marsh	Durham	2007–2010, 2012–2014	43.886768	−78.678787
Whitby Harbour Wetland Complex	Durham	2009–2015	43.854437	−78.936228
Wilmot Creek Marsh	Durham	2007–2008, 2010, 2012, 2014–2015	43.896958	−78.597135

). Sixteen coastal wetlands in the Durham Region are located on the eastern edge of the GTA and are monitored as part of the Durham Region Coastal Wetland Monitoring Program (DRCWMP). Being in close proximity to the large human population of the GTA, these wetlands have been found to have higher amounts of anthropogenic disturbance than other areas in the Lake Ontario basin [24,44]. Fifteen coastal wetlands in the Bay of Quinte (BoQ) are located in northeastern Lake Ontario. The BoQ was listed as an Area of Concern (AOC) by Environment Canada in 1985, and the Bay of Quinte RAP Coordinating Committee (1993) identified four main ecosystem problems: (1) destruction of habitat and ecosystem stability; (2) nutrient enrichment and ecosystem instability; (3) bacterial contamination; and (4) persistent toxic contaminants. Restoration objectives and remedial options have been identified, ranging from educational programs and stewardship initiatives to wastewater treatment and improvements in sanitary sewers [45]. Combined with key species invasions (Zebra Mussel (Dreissena polymorpha) and Quagga Mussel (Dreissena bugensis)), these efforts have resulted in decreased pelagic nutrient loads and increased water clarity [46,47,48]. The BoQ does have large settlement areas (e.g., Belleville) but is primarily considered an agricultural watershed, with significant shoreline alteration and recreational pressures.

In both regions, the wetlands range in area from 7 to 337 hectares and the depth varies considerably, but depths of more than 2 m were not included due to the limited effectiveness of boat electrofishing sampling beyond that depth. All of the wetlands were naturally formed, and there is considerable variation in the type and degree of human alteration. Some wetlands have experienced restoration efforts, whereas various threats continue to intensify in others. Habitat types sampled within each wetland were stratified based on habitat classified as inlet, outlet, nearshore (less than 50 m from shore), and open water (greater than 50 m from shore and less than 2 m in depth, ref. [44]).

2.2. Data Collection

A total of 138 sampling events for the 16 DRCWMP wetlands and 15 BoQ RAP wetlands, recorded in August or September 2005–2015, were included in this study. The DRCWMP began in 2002, and all 16 wetlands have been sampled annually since 2008. Coastal wetland sampling in the BoQ was periodic as early as 2005, but each of the 15 wetlands has been consistently sampled every three years in August or September since 2008.

2.3. Fisheries Sampling

Fisheries sampling was completed by boat electrofishing, as outlined in [42]. The use of a small 4.9 m Jon boat allowed for sampling in low water levels and heavy vegetation. Ethical approval was not required for this study, as the sampling was carried out by CLOCA under appropriate provincial government policies. CLOCA follows best practices to minimize stress on fish, including using bubblers, avoiding extreme heat, ensuring efficient processing, releasing fish at the capture site, and minimizing exposure to electricity.

2.4. Environmental Data

Abiotic and vegetation variables were recorded at the beginning of each transect (see [42] for details). Abiotic variables included turbidity (NTU); pH; conductivity (µS cm); total dissolved solids (mg/L); salinity (g/kg); dissolved oxygen (mg L⁻¹); water depth (measured to nearest cm using a meter stick or weighted string if over 1 m); water temperature (°C); dominant substrate (estimation based on grain size); air temperature (°C); wind strength (Beaufort wind strength scale (0–12); wind direction (azimuthal direction); and cloud cover (tenths scale). Vegetation was estimated visually along the entire length of each transect, covering 2 m on either side. Vegetation variables included submerged, emergent, and floating vegetation cover and total submerged vegetation cover. This was estimated and not summed due to some submerged, emergent, and floating vegetation overlapping. In addition, all types of vegetation were identified to species level, and the percent cover and vegetation height on transect were estimated. For each wetland, abiotic and vegetation data were averaged across all transects, separately for each year.

Annual water quality was sampled in July and a Water Quality Index (WQI, ref. [20]) value was determined based on turbidity, conductivity, water temperature, and pH. In addition, in July and August, additional aquatic vegetation sampling was undertaken separately to calculate the Submerged Aquatic Vegetation (SAV) IBI [35,49].

2.5. Data Analysis

For all 31 wetlands, the DRCWMP coastal wetland fish IBI was used as the measure of fish community health [35]. The fish IBI is composed of six individual metrics with significant relationships to coastal wetland disturbance gradients [35]: number of native species; number of centrarchid species; percent piscivore biomass; number of native individuals; percent non-indigenous biomass; and biomass of Yellow Perch. Scores were categorized into five IBI groups representing different levels of fish community health (

Table 1. Durham Region Coastal Wetland Monitoring Program fish IBI groups, associated score range, and summary of average species richness, average biomass (g/m²), and average abundance (individuals/m²) for each.

IBI Grouping	IBI Score	Abundance (Individuals/m2)	Richness	Biomass (g/m2)
Poor	0–20	0.1841	5.15	22.24
Fair	20–40	0.1511	8.45	15.31
Good	40–60	0.1967	10.79	25.25
Very Good	60–80	0.2479	12.13	17.61
Excellent	80–100	0.3343	11.96	12.05

). To calculate the IBI, fish catch data were pooled from all transects in each wetland for each year. For every year sampled, each wetland was classified into one of five IBI groups based on the resulting IBI score.

Fish community composition and structures were compared between IBI groups. Species lists for all sampling events and for each IBI group were compiled. Means of biomass, abundance, and richness were calculated. Due to the varying number of transects per sampling event, total biomass and abundance were standardized by area (m²). To further interpret species presence and absence within the different fish IBI groups and species locations on CCA, species classifications based on anthropogenic stressors [50] and affinity for aquatic vegetation [51] were assigned to each species. Sites with missing data were removed. Outliers were removed if their z score was greater than 3, to reduce the influence of these points on the subsequent analysis and in order to obtain more accurate relationships between fish IBI and aquatic and vegetation variables [52,53]. Six species with only one catch record were removed from the analysis: Coho Salmon, Longnose Dace, Silver Redhorse, Spotfin Shiner, Tessellated Darter, and White Bass. All calculations were completed using PAST (Version 4.09; ref [54]).

For univariate analysis, linear regression was used to examine the abundance, biomass, and richness relationships between the fish IBI and abiotic and vegetation variables. Box-and-whisker plots were used to determine the range of abiotic and vegetation variables for each IBI group. Differences in abiotic and vegetation variables between IBI groups were evaluated by a one-way analysis of variance (ANOVA), and Tukey’s post hoc pairwise test [55] was used to identify significant pairwise differences. If data were non-parametric, determined by Shapiro–Wilk test, the Kruskal–Wallis one-way analysis of variance was used and pairwise differences were determined using the Mann–Whitney U-test statistic [55].

For multivariate analysis, species data were square-root transformed to reduce the impacts of dominant species [56]. To examine how the fish community responded to the environmental variables, Canonical Correspondence Analysis (CCA) was used [57], as it tolerates skewed species and environmental data [58,59]. Only the abiotic and vegetation variables determined to have a significant relationship with the fish IBI were included in the CCA. Because of the multicollinearity between conductivity, salinity, and total dissolved solids, only conductivity was used in the analysis, as it had fewer missing values. ANOSIM and SIMPER [60] were used to test the significance of groupings in the CCA results. ANOSIM is a non-parametric permutation MANOVA that measures the variation between groups to determine if the distance is significant. Using IBI groups, ANOSIM was run on the Bray–Curtis similarity index based on 9999 permutations. ANOSIM pairwise-relationship significance was based on Bonferroni-corrected p values less than 0.05. SIMPER is based on total dissimilarity between groups and identifies which species most influence the difference between groups.

3. Results

Fish Data

A total of 19,915 fishes, representing 48 species, were caught in 138 sampling events. Of the 138 sampling events, 27 were considered to have an Excellent fish community as measured by the IBI, 19 Very Good, 49 Good, 32 Fair, and 11 Poor. Within each of these groups, Excellent sites were represented by the most species (37) and Poor sites had the fewest (19, Figure 2 and

Table A3. A summary of species caught within each Durham Region Coastal Wetland Fish Index of Biotic Integrity group. “X” indicates the species was detected at the site.

Species List	Excellent	Very Good	Good	Fair	Poor
Alewife	X	X	X	X	X
American eel	X	X		X
Banded Killifish	X	X	X	X	X
Black Crappie	X	X	X	X
Blackchin Shiner	X	X
Blacknose Dace	X		X
Blacknose Shiner	X	X
Bluegill	X	X	X	X
Bluntnose Minnow	X	X	X	X	X
Bowfin	X	X	X	X
Bridle Shiner	X
Brook Silverside	X	X
Brook Stickleback	X
Brown Bullhead	X	X	X	X	X
Brown Trout			X
Central Mudminnow	X	X	X		X
Chinook Salmon		X	X	X
Coho Salmon			X
Common Carp	X	X	X	X	X
Common Shiner	X	X	X	X	X
Creek Chub			X
Emerald Shiner		X	X	X	X
Fathead Minnow	X	X	X	X	X
Freshwater Drum		X	X	X
Gizzard Shad		X	X	X	X
Golden Shiner	X	X	X	X
Goldfish			X	X	X
Grass Pickerel	X
Green Sunfish				X	X
Johnny Darter	X	X	X	X
Largemouth Bass	X	X	X	X	X
Logperch	X	X	X	X	X
Longnose Dace	X
Northern Pike	X	X	X	X
Pumpkinseed	X	X	X	X	X
Rainbow Trout			X	X
Rock Bass	X	X	X	X	X
Round Goby	X	X	X	X	X
Silver Redhorse	X
Smallmouth Bass		X	X	X
Spotfin Shiner	X
Spottail Shiner	X	X	X	X
Tessellated darter	X
Walleye	X	X	X
White Bass	X
White Perch		X	X	X
White Sucker	X	X	X	X	X
Yellow Perch	X	X	X	X	X
Total	37	33	36	32	19

).

Three other measures of fish community were considered: mean abundance (fish/m²); biomass (g of fish/m²); and species richness (Table 1). Richness (r = 0.51, p < 0.001) and abundance (r = 0.36, p < 0.001) were found to have positive relationships with higher IBI scores. When IBI scores were grouped, abundance means between groups were significantly different (F = 6.229, p < 0.001), largely driven by differences between Excellent and Good, Fair, and Poor groups (p < 0.05, Figure 3). Richness was also statistically different between groups (F = 15.16, p < 0.001), largely driven by significant differences between Poor and Excellent, Good, and Fair groups (p < 0.01, Figure 3). Biomass exhibited a negative relationship (r = −0.08, p = 0.30) with improving health groups, but there were no significant differences between groups (F = 1.334, p = 0.26, Figure 3). There was high variability in the biomass data, largely because of the influence of Common Carp. Common Carp accounted for approximately 38% of the biomass and 2.4% of the abundance in the entire dataset.

Eight of the 13 abiotic and vegetation variables had significant relationships with fish IBI scores (p < 0.05, Figure 4,

Table A4. Summary of linear regression analysis comparing Durham Region Coastal Wetland Monitoring Program fish Index of Biotic Integrity groupings with select abiotic and vegetation variables.

Variable	Slope	Error	Intercept	Error	r	p
SAV IBI	0.81027	0.083844	−6.0457	4.9226	0.63807	3.86 × 10−17
WQI	0.026321	0.0022732	−2.2502	0.13346	0.70458	5.29 × 10−22
air temp	−0.0060553	0.010862	23.723	0.63771	−0.047749	0.57811
water temp	−0.0030268	0.010663	22.434	0.62606	−0.024332	0.77696
pH	0.0014812	0.0011919	8.0303	0.069977	0.10596	0.21611
conductivity	−8.839	0.7504	1052	44.057	−0.71063	1.63 × 10−22
DO	−0.017508	0.0093294	10.542	0.54774	−0.15888	0.062707
water depth	0.37116	0.097458	74.499	5.7219	0.31044	0.00021087
turbidity	−0.28075	0.048979	32.486	2.8756	−0.44111	6.13 × 108
Total SAV	0.69776	0.098559	7.3543	5.7865	0.51894	7.01 × 10−11
Emergent	0.061331	0.012358	−1.2546	0.72554	0.39159	2.04 × 10−6
Submergent	0.77046	0.10092	−3.605	5.9254	0.5477	3.60 × 10−12
Floating	−0.0019787	0.044218	10.632	2.5961	−0.0038372	0.96437

). Of the significant variables, SAV IBI, WQI, water depth, total SAV, total emergent vegetation, and total submergent vegetation had significant positive relationships, and conductivity and turbidity had significant negative relationships with fish IBI. Air and water temperature, pH, and total floating vegetation had no significant relationships. Tests on all eight variables strongly rejected the null hypothesis, indicating they were not significantly different among IBI groups (p < 0.001, Figure 5,

Table A5. Tukey’s pairwise results highlighting significance between Durham Region Coastal Wetland Monitoring Program fish Index of Biotic Integrity. Eight abiotic and biotic variables found to be significant through linear regression and ANOVA were included.

Tukey’s Pairwise
SAV IBI	Excellent	Fair	Good	Poor	Very Good
Excellent		1.72 × 10−5	1.72 × 10−5	1.72 × 10−5	2.44 × 10−5
Fair	11.87		0.9996	0.9937	0.004927
Good	11.59	0.284		0.9995	0.009873
Poor	11.28	0.59	0.306		0.01988
Very Good	6.982	4.893	4.609	4.303
WQI	Excellent	Fair	Good	Poor	Very Good
Excellent		1.72 × 10−5	1.72 × 10−5	1.72 × 10−5	4.54 × 10−5
Fair	13		0.9512	0.9861	7.53 × 10−5
Good	11.97	1.023		0.7303	0.001386
Poor	13.72	0.7253	1.748		2.15 × 10−5
Very Good	6.603	6.393	5.37	7.118
Conductivity	Excellent	Fair	Good	Poor	Very Good
Excellent		1.72 × 10−5	3.39 × 10−5	1.72 × 10−5	0.1579
Fair	11.41		0.008695	0.9403	1.73 × 10−5
Good	6.75	4.662		0.0004771	0.08784
Poor	12.5	1.083	5.745		1.72 × 10−5
Very Good	3.197	8.215	3.553	9.298
Water Depth	Excellent	Fair	Good	Poor	Very Good
Excellent		0.1383	0.1104	0.06333	0.99
Fair	3.282		1	0.9977	0.04201
Good	3.42	0.1382		0.9995	0.03171
Poor	3.734	0.4519	0.3136		0.01607
Very Good	0.6645	3.946	4.085	4.398
Turbidity	Excellent	Fair	Good	Poor	Very Good
Excellent		3.35 × 10−5	0.001237	0.009865	0.1915
Fair	6.757		0.8765	0.55	0.06868
Good	5.411	1.346		0.9798	0.4606
Poor	4.609	2.148	0.8023		0.8118
Very Good	3.067	3.69	2.344	1.542
Total SAV	Excellent	Fair	Good	Poor	Very Good
Excellent		1.72 × 10−5	1.72 × 10−5	1.72 × 10−5	7.87 × 10−5
Fair	8.988		0.9408	0.9992	0.3464
Good	10.07	1.081		0.851	0.06826
Poor	8.641	0.347	1.428		0.4962
Very Good	6.376	2.612	3.693	2.265
Emergent	Excellent	Fair	Good	Poor	Very Good
Excellent		0.0003546	0.0001727	0.008058	0.7195
Fair	5.847		0.9998	0.9259	0.03242
Good	6.094	0.2467		0.8599	0.01911
Poor	4.693	1.154	1.4		0.2355
Very Good	1.773	4.074	4.321	2.92
Submergent	Excellent	Fair	Good	Poor	Very Good
Excellent		1.72 × 10−5	1.72 × 10−5	1.73 × 10−5	3.90 × 10−5
Fair	10.48		1	0.6445	0.05587
Good	10.42	0.05162		0.6677	0.0616
Poor	8.533	1.943	1.891		0.6831
Very Good	6.676	3.8	3.748	1.857

). For SAV IBI and WQI, there were significant differences between Excellent and all other groups, and Very Good and all other groups (Figure 5). Differences between Good, Fair, and Poor groups were not significant (Figure 5). For conductivity, all pairwise tests were significant except between Very Good and Good, and Poor and Fair groups (Figure 5). The least number of significant pairings were for water depth and turbidity (Figure 5). Water depth was significantly different between Very Good and Good, Fair, and Poor groups only, and turbidity was significantly different between Excellent and Fair, Good, and Poor groups (Figure 5). Total SAV was significantly different between Excellent and all other groups (Figure 5). Emergent and submergent vegetation variables were only significantly different between the healthy groups (Excellent, Very Good) and most others (Figure 5). There were no significant differences between Poor and Fair groups for any variable, and only one variable, conductivity, had significant difference between Fair and Good groups (Figure 5).

Relationships between fish community and all of the abiotic and vegetation variables were analyzed to determine how each related to fish community health. The first two axes of the CCA, based on the eight abiotic and vegetation variables, explained 77.85% of the variation (Figure 6). IBI group centroids were positively correlated to the first axis from Poor to Excellent (Figure 6. ANOSIM results indicated that IBI groups were significantly different (R = 0.394, p = 0.0001,

Table 2. ANOSIM and SIMPER results examining significant difference between fish community abundance and fish IBI groups, and overall dissimilarity of fish abundance and species with the largest contribution to dissimilarity between fish IBI groups.

Group 1	Group 2	ANOSIM (p-Value)	Overall Dissimilarity	Three Species with the Largest Contribution (Species (% Contribution))
Excellent	Very Good	0.001	51.51	Yellow Perch (13.8)	Bluegill (9.0)	Golden Shiner (7.3)
Excellent	Good	0.001	66.25	Yellow Perch (16.1)	Bluegill (10.8)	Brown Bullhead (7.1)
Excellent	Fair	0.001	75.59	Yellow Perch (18.3)	Bluegill (10.9)	Pumpkinseed (7.6)
Excellent	Poor	0.001	87.06	Yellow Perch (18.9)	Bluegill (10.9)	Pumpkinseed (9.9)
Very Good	Good	0.165	59.48	Brown Bullhead (9.5)	Yellow Perch (9.4)	Bluegill (7.8)
Very Good	Fair	0.002	67.95	Yellow Perch (11.4)	Brown Bullhead (8.9)	Pumpkinseed (8.3)
Very Good	Poor	0.001	81.43	Yellow Perch (12.1)	Pumpkinseed (10.9)	Brown Bullhead (7.8)
Good	Fair	0.003	62.85	Fathead Minnow (10.6)	Brown Bullhead (10.0)	Pumpkinseed (8.8)
Good	Poor	0.001	74.60	Pumpkinseed (11.5)	Brown Bullhead (9.8)	Goldfish (8.9)
Fair	Poor	0.016	72.85	Goldfish (12.77)	Brown Bullhead (11.1)	Fathead Minnow (11.1)
Pooled Groups			67.09	Yellow Perch (11.74)	Brown Bullhead (8.3)	Pumpkinseed (8.1)

). The only pairwise test not significant was between Very Good and Good groups (Table 2). The SAV IBI had the strongest relationship to the Excellent group, whereas conductivity and turbidity had the strongest relationship to the Poor group (Figure 6).

Species relationships to IBI groups were identified using their location on the CCA plot and SIMPER analysis. Blackchin Shiner, Blacknose Shiner, Bluegill, Grass Pickerel, and Yellow Perch had the strongest relationship with the Excellent group, and Brown Bullhead, Common Carp, Fathead Minnow, Goldfish, and White Perch had the strongest relationship with the Poor group. Through SIMPER analysis, Bluegill, Brown Bullhead, Fathead Minnow, Golden Shiner, Goldfish, and Yellow Perch were most commonly one of the top three species contributing to group dissimilarity (Table 2). Yellow Perch and Bluegill contributed the most to distinguishing the Excellent group from the others. Goldfish and Fathead Minnow were in the top three contributing species only when the Poor group was compared to the others. When all groups were pooled, Yellow Perch was the highest contributor to overall dissimilarity, followed by Brown Bullhead, Pumpkinseed, Bluegill, and Fathead Minnow (Table 2). Overall dissimilarity between groups increased as the distance between groups increased; the largest dissimilarity was between the Excellent and Poor groups (87.06%), whereas the smallest dissimilarity value was between the Excellent and Very Good groups (51.51%).

4. Discussion

The results of this study indicate that fish community health and composition are driven by a number of abiotic and vegetation variables, with varying levels of influence. High turbidity and conductivity values had the strongest negative relationships with fish IBI, whereas high WQI and SAV IBI values had the strongest positive relationships. Eight abiotic and vegetation variables had significant differences between fish IBI groups, and ranges for these variables in the Very Good and Excellent groups were generally significantly different from the other groups. When multivariate and multimetric analyses were used in combination, fish community health exhibited a gradient from Poor to Excellent groups, and within-group variability decreased along a Poor to Excellent group gradient. Yellow Perch and Bluegill had the largest changes in population size when comparing the Excellent group to all other groups. Bluegill has been well documented as an indicator of wetland health sensitivity to anthropogenic disturbances [19]. Fish community composition varied considerably when grouped by fish IBI. Overall, a total of 48 fish species were caught, but as few as 19 species were caught in wetlands classified as Poor and as many as 37 species in wetlands classified as Excellent. Trebitz et al. [61] found that anthropogenic disturbance tended to homogenize wetland fish composition but, when variability was found, it was often due to catches of large schooling species, not habitat heterogeneity. Mean richness was highest in the Very Good and Excellent groups. SAV was positively related to various measures of fish community health, which is consistent with previous studies. Fish community relationships with increasing cover and diversity of SAV have been identified through previous studies [19,30,62]. Seilheimer and Chow-Fraser [63] reported a positive relationship between the Wetland Fish Index and the number of submerged aquatic vegetation species. Trebitz and Hoffman [9] found that management actions promoting vegetation structure resulted in a shift from less to more desirable commercial and recreational fish species.

Interestingly, biomass had a negative relationship with fish health. This was largely driven by Common Carp. Common Carp accounted for approximately 38% of the biomass and 2.4% of the abundance across all sampling events. Similarly, Peterson et al. [64] found that Common Carp accounted for less than 1% of abundance but 19% of total biomass in a Lake Superior river mouth. Carp biomass was reported to be as high as 2600 kg/ha in a Lake Ontario coastal wetland (Cootes Paradise Marsh) [65], which causes its influence on this metric to be large. The data in our study suggest total biomass is too variable to be a useful metric when grouped by fish IBI. The annual timing of DRCWMP purposefully targets smaller-bodied fishes, which may make biomass a less important variable for this shallow coastal wetland community, while refinement of the biomass metric, to something more specific than total biomass, could improve its relationship with fish IBI. Minns et al. [62] found a relationship between IBI and piscivore biomass in the littoral zone. Three measures of biomass were found to have a significant relationship to a disturbance gradient when developing the fish IBI: biomass of Yellow Perch, percent piscivore biomass, and percent non-indigenous biomass [35]. This suggests that the usefulness of these biomass metrics varies, making it important to re-evaluate their usefulness and to differentiate between Excellent and Poor groups when applied to new areas.

This study found that tolerant species were commonly found in all IBI groups. One tolerant species, Goldfish, was only found in the Poor or Fair groups. This emphasizes the importance of fish community composition in understanding community health, which has been noted in previous research [66]. Although slight, a decrease in sensitive species was found between Excellent and Poor groups. All four sensitive species in the Good group, and two of the three in the Fair group, were salmonids. Due to timing of sampling, it is likely that these species were migrating through wetlands to access upstream watersheds for spawning. If removed from the analysis, the Fair group is the only group, other than Excellent and Very Good, with a sensitive species (American Eel). This also identifies a potential problem with vague sensitivity definitions. The sensitivity of fishes may vary based on the geographic location of local impacts. Even when linking sensitivity to a specific stressor, determining a score, exclusive of other cumulative and synergistic stressors, can be challenging, especially for infrequently captured species [39].

Four of eight variables with significant linear relationships with fish IBI were measures of vegetation. Floating vegetation was the only vegetation variable not significantly related to fish IBI. It was found to be a poor predictor of fish community health and was slightly more abundant in the Poor group. This is not consistent with Trebitz et al. [30], who reported floating leaf vegetation and submerged vegetation had strong fish-habitat associations. Given floating vegetation is less influenced by adverse abiotic variables (e.g., turbidity), it may be able to survive and potentially thrive when submerged vegetation species and fish communities are degraded [67].

Water depth was also found to have a significant positive relationship with fish IBI. Trebitz et al. [30] and Langer et al. [37] also identified a relationship between fish composition and aspects of physical habitat, including water depth. Depth could provide for increased habitat availability or greater variability in habitat types, which has been identified as a driver of lotic richness [68]. It could provide fishes with the ability to avoid certain adverse abiotic conditions (e.g., low dissolved oxygen at night [69]). There may also be a relationship between water depth and disturbance levels. Harrington and Hoyle [70] found that much of a Lake Ontario coastal wetland (Rattray Marsh) had 300–500 mm sediment deposition from upstream erosion caused by increased land-use change. This caused artificially shallow marshes, decreased habitat heterogeneity, and covered natural organic sediment with mineral silt and sand that is less likely to support aquatic vegetation.

Air and water temperature, pH, and dissolved oxygen did not have significant relationships with the fish IBI. The air and water temperature results are not unexpected, as the data were fairly coarse (single point-in-time measurement per transect averaged for each wetland and year). Both pH and dissolved oxygen were found to be highly variable across fish IBI groups. Although pH and dissolved oxygen are known to be important variables for fishes [14,71,72,73], neither the poorest- or highest-quality sites had values for either variable to a point where they appeared to be a limiting factor. Wetlands are constantly changing, and there may be periods of poor abiotic conditions that influence fish communities but that are not measured [4,5]. For example, Chow-Fraser et al. [69] reported having water-quality stations where dissolved oxygen dropped to as low as 2 mg/L at night due to decaying organic material. These conditions could have a considerable influence on the fish community, but not be detected or measured.

Water-quality measures (WQI, conductivity, turbidity) had the strongest relationship with fish IBI. WQI was expected to be highly correlated with fish IBI, given it combines multiple water-quality variables known to be important to fishes [14,74,75,76]. Conductivity is increasingly becoming a concern in areas with a high amount of impervious cover to which road salt is applied in the winter season [77]. This practice has led to conductivity levels in Lake Ontario coastal wetlands averaging as high as 1800 µS cm. Turbidity was also found to be significantly related to gradients of fish community health or structure, which is consistent with many other studies [19,39,69,78,79,80,81].

A goal of this analysis was to determine ranges of environmental variables for each fish IBI group. Impairment issues can be obvious (e.g., high turbidity) but, in some cases, they can be more difficult to determine. Occasionally, Lake Ontario coastal wetlands will have turbidity values under 4 NTU and large amounts of SAV, yet the IBI score will still be poor. These scenarios can leave managers with difficult decisions regarding setting goals and targets for management or restoration. Having a reference of expected ranges for each fish community health group could provide important information for diagnosis and/or realistic project goal setting. The expected abiotic and vegetation ranges for each fish IBI groups can be used as a diagnostic tool to help determine which variables may be limiting when managing or restoring coastal wetlands in Lake Ontario. For example, if a wetland has a submergent SAV under 80% total cover, WQI below 0, conductivity above 350 mS cm⁻¹, or turbidity above 5 NTU, it is unlikely to score Excellent. Alternatively, if a wetland has an average conductivity of 800 mS cm⁻¹, it is unlikely to score above Poor or Fair in its current condition. Although taking a holistic view to restoration is critical, being able to set specific targets is important for setting management and restoration goals and ensuring future monitoring is directed appropriately, in order to determine the successes and failures of actions taken [82,83].

Multimetric approaches to tracking changes in fish health have become a commonly used tool [31,32,62,66,84,85]. Despite being a valuable tool, limitations of this technique have been identified [86,87,88]. Granados [89] determined that given species richness usually plays an important role in the multimetric approach, re-colonization of formally disturbed sites is typically when an IBI will be most sensitive to changes. If the site is experiencing species composition, rather than richness, changes, the IBI is less likely to detect changes [87]. Trebitz et al. [88] found the multimetric approach was better suited for detecting broad assemblage differences than changes in abundance and variability. The CCA used in this study allowed for an increased understanding of fish community health and its relationship to abiotic and vegetation variables beyond the multimetric assessment. This approach is consistent with other studies analyzing wetland vegetation [13,29,90,91]. Fish IBI groups were added to the CCA to determine if IBI health groups could be defined using ordination. The groups were separated along a gradient, and the Poor group was found to have higher variability in composition (as determined by the size of the 95% ellipse, Figure 6). The Excellent group had the least variation, despite having a larger number of sites than the Poor group. This result is similar to results in other studies, showing that, despite differences in geomorphic wetland conditions, geographic and human influences explain more of the variation [28,30]. The relatively homogenous habitat structure in Lake Ontario wetlands may also be contributing to similar fish communities where anthropogenic influences are minimal [30]. This demonstrates that CCA can be an effective method for tracking changes in coastal wetland fish community health during management or restoration monitoring. Both the location and the amount of variation in CCA plots, as indicated by ellipse size and shape, within the fish community can be measures of impairment. The results of this study suggest that healthy wetland fish communities are predictable, given they have little variability temporally and spatially within the study area. Poor wetland fish communities appear to have large fluctuations in composition that vary annually. The variation observed within IBI groups is important, as species composition shifts can occur for a variety of reasons. Finigan et al. [92] found that increases in temperature caused a shift from a leuciscid to centrarchid community in the littoral zone of 21 Ontario lakes. The ability to detect these changes using a multimetric approach would depend on whether the shifts in species composition were related to at least one metric. Composition shifts generally cannot be predicted and, therefore, changes detected through multimetric analysis would not be known until after they occurred. This highlights the importance of an additional multivariate analysis, which is sensitive to community shifts, to ensure that they are identified as early as possible.

The direction and strength of the relationship between fish community health and significant abiotic and vegetation variables were relatively consistent between the CCA and linear regression results. The CCA results explained more of the variation, due to all variables being considered in the same analysis. With abiotic and vegetation variables and fish species included in the CCA analysis, further interpretation on a site- or species-specific scale increased the value of CCA. The strongest relationships between species and the Excellent group included Blackchin Shiner, Blacknose Shiner, and Grass Pickerel. All require low turbidity and densely vegetated habitats [6,8]. Species with the strongest relationship to the Poor group include Common Carp, Goldfish, and Fathead Minnow. These species are known to tolerant and dominate communities with poor water quality [24,49]. This provides managers with an indication of the species composition required to achieve healthy conditions in Lake Ontario coastal wetlands. This approach emphasizes the importance of considering all fish species when determining fish community health and not a subset based on recreational or commercial significance. Bluegill and Yellow Perch are often viewed as a more tolerant species, but our results indicate that, in unsuitable conditions, Yellow Perch abundance decreases considerably. Given its dependence on wetland vegetation for its early life history [93], if wetland conditions degrade to the point of decreased vegetation cover, Yellow Perch populations will decrease as well.

5. Conclusions

A combination of multimetric and multivariate methods provided a comprehensive method of evaluating fish community health, given the complementary strengths of both approaches. A range of environmental conditions required for fish community health was identified in our study. This will allow specific and ecologically meaningful goals to be identified, a critical component in the restoration process that is rarely implemented, resulting in poor sample design and vague monitoring objectives [83,94,95]. This research will improve the ability of managers to identify impairment factors and establish ecologically relevant restoration goals.

Many of the variables tested had strong relationships with the fish IBI. Abiotic surrogates are valuable for inferring biological health but should not be used in isolation. Minns et al. [62] emphasized the importance of understanding the causality of biotic states and not simply relying on static abiotic surrogate correlations—this still holds true today. This study identified a correlative, not cause-and-effect, relationship, and the relationship between the fish IBI and abiotic and vegetation variables may change over time. New threats, such as biological invasions, emphasize the importance of continuing to monitor biotic communities. Further research investigating the causation of the observed relationship is required. Additional variables and their influence on fish community health should be considered. For example, the degree of connectivity to Lake Ontario should be investigated further; more specifically, the influence of natural (barrier beach) and anthropogenic (water-control structures) barriers on fish movement and how this relates to health and restoration trajectories. Coastal wetlands are at the interface of tributaries and lakes, where important hydrological interactions and nutrient impacts influence the structure and functions of wetland ecosystems [17,96,97].

Restoration efforts, compensation, and offsetting are becoming common policy tools when unavoidable damages cannot be mitigated; therefore, better tracking and understanding of wetland management and restoration trajectory tools are urgently needed [98,99,100,101]. Our study provides management and restoration practitioners with additional methods and supporting data for diagnosing and tracking fish community health in Lake Ontario coastal wetlands. Fish communities in coastal wetlands are dynamic, complex, and subject to many abiotic and vegetation variables. Our approach allows for an increased understanding of current conditions and restoration trajectories using well-defined monitoring protocols, which will aid in a better understanding of Lake Ontario coastal wetland fish community structure and health.