Floristic quality indices are used to monitor and assess wetland condition by measuring a plant community's tolerance to environmental stress. The aim of our research was to evaluate whether stormwater and reclamation marshes supported wet meadow plant communities that were similar in floristic quality to reference wetlands. Coefficients of conservatism were assigned to a comprehensive list of marsh plant species by nine expert botanists. Various metrics such as the floristic quality index (FQI) were tested for a linear relationship to regional environmental stress gradients in 78 sites in the northern prairies (Aspen Parkland) and 66 sites in the Boreal Plains ecoregions in Alberta, Canada. Sensitivity of floristic quality metrics to the stress gradients was higher when rare species were excluded (species with <5% site-level cover). An adjusted FQI that eliminated bias towards sites with higher species richness yielded the strongest relationship to the stress gradient in both ecoregions (r2 = 0.55 in northern prairies; r2 = 0.46 in Boreal Plains). The adjusted FQI also yielded more consistent scores than richness-weighted metrics in a subset of 47 sites where sampling was replicated in dry and wet years (r = 0.75). Including exotic species in floristic quality metrics was not beneficial in the northern prairies where reference sites were located near areas of high urban and agricultural development. This study demonstrates that floristic quality assessments are reasonably good predictors of plant community condition in relation to environmental stress. Results of this study also highlight that existing stormwater ponds and reclamation marshes are not successfully restoring plant community habitat.
Human activities and land-use changes have converted or degraded over half of wetland ecosystems in North America (Moreno-Mateos et al., 2012). Historical agricultural conversion as well as ongoing urban and industrial development has led to 70% loss of wetlands in the Canadian prairies (Kennedy and Mayer, 2002). In the Boreal Plains, which lie north of the more populated southern areas of Alberta, landscape transformations in the oil sands have led to large-scale loss of peatlands. Oil company closure plans project a 67% reduction of wetland area compared to the pre-mining landscape (Rooney et al., 2012). Since there is no existing way to restore peatlands on a large scale, post-mining landscapes are being reclaimed instead by extensive upland interspersed with sub-saline marshes, which are naturally uncommon in the Boreal Plains (Rooney and Bayley, 2010).
Mitigation policies and strategies have developed out of the necessity to offset wetland loss and to conserve the functions and services of wetlands on the landscape. Nevertheless, a recent meta-analysis found that the rate and degree of ecosystem recovery is slow in wetlands that are small, hydrologically isolated, and located in colder climates (Moreno-Mateos et al., 2012). All three of these factors apply to shallow depressional marshes in northern Canada and could negatively influence the success of ecosystem recovery. Recently, a suite of plant-based assessment tools was developed to evaluate the success of wetland construction in the northern prairies’ Aspen Parkland region (Wilson and Bayley, 2012) and reclamation marshes in the Boreal Plains (Raab and Bayley, 2012; Rooney and Bayley, 2011a, 2011b).
Biotic metrics that display an empirical relationship with anthropogenic disturbance are used to indicate biological integrity or impairment of an ecosystem (Karr and Chu, 1999). Metrics are tested against a disturbance gradient representing varying degrees of stress between least-disturbed and most-disturbed endpoints (e.g. Rooney and Bayley, 2010). Plant communities are commonly used as indicators of ecosystem condition, and floristic quality metrics describe the habitat quality of a plant community (Miller and Wardrop, 2006) in relation to environmental stress. The main component of the floristic quality index (FQI) is the coefficient of conservatism (CC), which ranks species on an ordinal scale between 0 and 10 based on its fidelity to a particular habitat or tolerance to disturbance (Lopez and Fennessy, 2002). Low CC values indicate less conservative species likely to occur in disturbed habitats whereas high CC values are given to the most sensitive species with specific autecological requirements. Floristic quality assessments have been found to be better at predicting stress and estimating habitat quality than simple measurements such as species richness and diversity (Miller and Wardrop, 2006). Floristic quality metrics can be used alone or incorporated into more comprehensive tools such as indices of biological integrity to evaluate ecosystem condition (Mack et al., 2008; Raab and Bayley, 2012; Wilson and Bayley, 2012).
The original FQI devised by Swink and Wilhelm (1994) is calculated as the sum of a site's CC values divided by the square root of native species. Although Swink and Wilhelm reasoned that exotic species should be excluded in floristic quality calculations due to their absence during the evolution of local plant communities, others have stated that there is no empirical validation for their exclusion (Cohen et al., 2004; Ervin et al., 2006). The original FQI has also received criticism for its bias towards larger sites with greater species richness (Cohen et al., 2004; Miller and Wardrop, 2006). Species richness does not necessarily have a linear relationship with disturbance, and often exhibits non-linear patterns, as exemplified by the intermediate disturbance hypothesis. Several variants of floristic quality have subsequently been proposed that address these issues by including exotic species (e.g. Cohen et al., 2004) and eliminating weighting the FQI by species richness (e.g. Miller and Wardrop, 2006).
The main goals of this study were to (1) develop coefficients of conservatism for marsh species found in the Aspen Parkland and Boreal Plains; (2) examine the sensitivity of floristic quality and richness metrics to regional stress gradients; (3) evaluate whether floristic quality at constructed stormwater ponds and reclamation marshes is similar to reference wetlands in the two ecoregions.
Between 2007 and 2011, 144 semi-permanent and permanent shallow open waters and marshes (Stewart and Kantrud, 1971) were sampled in two contiguous ecoregions: the Aspen Parkland, characterized by mixed-wood forests, grasslands, and depressional wetlands; and the Boreal Plains, characterized by mixed-wood forests, extensive peatlands and some marshes. Of the 144 sites, 78 were sampled in the Aspen Parkland. Site selection in this region was restricted to shallow open water wetlands and marshes formed in post-glacial depressions that ranged in salinity from fresh to moderately brackish. The suite of wetlands included natural, restored and constructed marshes. Natural wetlands were grouped into reference (n = 28) and agricultural (n = 17) categories. Agricultural wetlands were surrounded by less than 50% forest within a 500 m buffer whereas reference wetlands were surrounded by more than 50% forest and represented the least-disturbed condition. Restored wetlands (n = 9) were located on former agricultural land and had undergone restoration at least three years prior to sampling. Constructed sites were also grouped into two a priori categories: naturalized stormwater ponds (n = 16) and standard stormwater ponds (n = 11). Naturalized stormwater ponds have specific design and construction features, such as salvaged wetland soils, undulating basins, and planted vegetation that are meant to emulate the structure and function of natural wetlands. Stormwater ponds were all >3 years old with an average age of 17 years. Natural and constructed marshes were relatively small (1 to 14 ha), at least 1 km from adjacent wetlands, and within a 60 km study area. A subset of natural and constructed sites (n = 47) in the Aspen Parkland was resampled in 2010 and 2011 to evaluate the consistency of floristic quality assessments under inter-annual variability. Precipitation between May and September was low in 2008 (240 mm) and 2009 (198 mm) compared to 2010 (324 mm) and 2011 (329 mm), which facilitated the comparison of floristic quality scores in dry and wet years.
The remaining 66 marshes located in the Boreal Plains were sampled between 2007 and 2009. Natural marshes are scarce in this region and were consequently scattered across the region, whereas test sites were constrained to reclaimed land on Suncor Energy Inc. and Syncrude Canada Ltd. leases. Test sites were grouped into two a priori categories: oil sands process affected sites, which were exposed to oil sands tailing contamination (n = 13), and oil sands reference sites, which were free of tailing contamination (n = 12). Natural marshes were subdivided into agricultural (n = 12) and reference sites (n = 29). The marshes sampled ranged from fresh to sub-saline and had depths <2 m.
Environmental stress gradients were developed for the Aspen Parkland (Wilson and Bayley, 2012) and Boreal Plains (Rooney and Bayley, 2010) by measuring a suite of environmental variables. In the Aspen Parkland, 41 variables from three abiotic categories, i.e. physical structure, water chemistry, and sediment chemistry, were analyzed. In the Boreal Plains, 52 variables were analyzed from the same three categories stated above as well as oil sands contaminants. Environmental variables were predicted to reflect variation in human disturbance and influence plant condition. Principal Components Analysis (PCA) was used to select environmental variables within each abiotic category with the largest eigenvector on each significant axis. Significant axes were chosen using the heuristic broken-stick model (McCune and Grace, 2002) or by visually identifying cut-offs of unexplained variance in scree plots (Rooney and Bayley, 2010). Eight final variables were chosen to make up the stress gradients in each region. The variables comprising the stress gradient in the Aspen Parkland were NO2 + NO3, total N, and conductivity in the water; % P, % N and % water in the sediment; shoreline slope, and turbidity (Wilson and Bayley, 2012). The stress gradient variables used in the Boreal Plains were total cations, total N, and Cl− in the water; % water and % oil in the emergent zone sediment; proportion Secchi depth, maximum depth, and water amplitude (Rooney and Bayley, 2010). Variables were standardized by percentile binning and then averaged within each abiotic category such that each category was weighted equally. The scores from each abiotic category were then summed to give a final stress score.
Marsh vegetation in the wet meadow zone was sampled during peak biomass in July and August. Boundaries of the wet meadow zone were delineated by vegetation, i.e. sedges, marsh grasses and herbs, and position between the upland transition and emergent zones. In natural wetlands in Alberta, the wet meadow zone is typically the widest vegetative zone. Only the wet meadow zone was sampled because Wilson and Bayley (2012) found that additional zones did not improve wetland condition assessments. Vegetation in six quadrats were deployed along three equidistant transects around the wetland in the centre of the wet meadow zone, which a power analysis showed was sufficient to detect a difference in species richness between reference and constructed sites (beta = 0.95, Wilson and Bayley, 2012). Herbaceous species found in the quadrat were identified to species-level if possible. Species names were updated according to the Integrated Taxonomic Information System Online Database (Accessed 12 May 2012, http://www.itis.gov).
Floristic quality assessment system
Nine expert botanists from university, provincial and federal governments, professional consultancies, and independent contractors, were contracted to assign coefficients of conservatism (CC) to 407 marsh species found in Alberta. Botanists worked independently to provide additional information about variability in opinion (Cohen et al., 2004) as well as to simplify the logistics of assembling botanists dispersed around the province. CC values were assigned to species separately for the two study regions because species tolerance can vary depending on region and wetland type. As a hypothetical example, a species with high fidelity to pristine marshes in one region might be an indicator of altered hydrology in another region. Each botanist assigned CC values based on the following key adapted from previous methodologies (Andreas and Lichvar, 1995): 0: non-native species and ruderal species growing on waste ground; 1–3: species commonly found in a wide variety of conditions with a high tolerance to disturbance; 4–6: species usually found within a specific plant community, but tolerant of moderate disturbance; 7–8: species found in advanced stages of succession that tolerate minor disturbance; 9–10: species with very low tolerance to disturbance. Final CC values represent the median value assigned by the nine botanists. Taking the sample median moderated the influence of outliers and is more appropriate for ordinal data than the mean (Zar, 1999). Medians that were half-integers were rounded up to the nearest whole number. CC values of species found in the wet meadow zone are provided in Table 1.
The simplest floristic quality metric is the mean CC value (Swink and Wilhelm, 1994):Equation (2) is weighted by species richness, whereas Equation (3) is unbiased by species and weighted by the proportion of native species relative to the total species (N + E). Equations (1), (2), and (3) were calculated both with and without exotic species. All floristic quality metrics excluded rare species with less than average of 5% cover. This process was performed to minimize the influence of accidental and rare species. Examples of accidental species are intruders from neighboring environments or species at the edge of their range. Any species that did not have a CC value was excluded from FQI calculations.
Floristic quality metrics were tested against the regional stress gradients using linear regression in SYSTAT 13 (Systat Software Inc., 2011). A fixed model analysis of variance (ANOVA) with a Tukey's Honest Significant Differences test was performed to assess differences in floristic quality among the a priori designated marsh type categories. The consistency of floristic quality metrics was examined under climatic variability between dry (2008 and 2009) and wet (2010 and 2011) years by performing correlation tests using the subset of replicated sites in the Aspen Parkland.
One hundred and thirty-eight species were encountered in the wet meadow zone during vegetation sampling. An average of 19 species were found per site in the Aspen Parkland and 16 species in marshes in the Boreal Plains. A total of 29 exotic species were encountered in the Aspen Parkland, whereas only 9 exotic species were found in the Boreal Plains.
An average of 5.6 and 7.5 botanists assigned a CC value to species in the Aspen Parkland and Boreal Plains, respectively. Pairwise agreement (Spearman's r) in assigned CC values among botanists was 0.79 (range = 0.70 - 0.90) for species in the Aspen Parkland and 0.70 (range = 0.54 - 0.81) for species in the Boreal Plains.
Floristic quality metrics had better correlations with the regional stress gradients than simple measures of species richness or exotic species (Table 2). The adjusted FQI′, which eliminated bias towards higher species richness (Equation (3)), was a reasonably good predictor of stress scores in both ecoregions (Aspen Parkland: r2 = 0.55 without exotics, p < 0.001; Boreal Plains: r2 = 0.46 with exotics, p < 0.001, Table 2). Inclusion of exotic species improved metric sensitivity with the stress gradient in the Boreal Plains (Table 2), as the number of exotic species was also correlated with stress scores (r2 = 0.26, p-value < 0.001, Table 2). Although the FQI was only slightly weaker than the adjusted FQI′ in the Aspen Parkland (Table 2), it produced inconsistent scores between dry and wet years (Table 3). In contrast, the adjusted FQI′ yielded consistent scores between dry and wet years (Pearson's r = 0.75, Table 3).
A fixed factor ANOVA revealed that variance in floristic quality differed among site types in both regions (Aspen Parkland: F-value = 21.5, p < 0.001; Boreal Plains: F-value = 14.2, p < 0.001). Natural sites (i.e. reference and agricultural) had higher floristic quality than stormwater ponds in the Aspen Parkland (Figure 1). Likewise, natural sites in the Boreal Plains had higher floristic quality than reclaimed sites (Figure 2).
Floristic quality metrics were developed for shallow open water wetlands and marshes in two ecoregions in Alberta that were exposed to urban and agricultural development in the Aspen Parkland and oils sands mining operations in the Boreal Plains. This study demonstrates that floristic quality metrics provide a reasonably good estimate of plant condition and habitat quality in relation to environmental stress in these ecoregions. In the Aspen Parkland, floristic quality was significantly lower at stormwater ponds than reference sites (Tables 2 and 3, Figure 1). Similarly, both categories of reclamation sites had lower floristic quality than natural marshes in the Boreal Plains (Tables 2 and 3, Figure 2).
There are several reasons why mitigation wetlands in these ecoregions seem to be unable to support healthy wet meadow marsh plant communities. In stormwater ponds, altered hydrology, water and soil chemistry, as well as steep-graded shorelines, and impermeable clay substrates with low organic material may impair biological structure and function (Wilson and Bayley, 2012). Reclamation marshes are affected by oil sands mining disturbances such as leaching of contaminants and salts from oil extraction processes, and recruitment of salt-tolerant plants to saline reclamation sites may be impeded by the relative isolation and scarcity of natural saline marshes on the landscape. In addition, cold climates, isolated hydrology, and small wetland sizes that hinder wetland recovery (Moreno-Mateos et al., 2012) are all factors influencing wetlands in this study.
The amount of variance in environmental stress scores represented by FQI metrics depended on how floristic quality was calculated. The adjusted FQI, or FQI′, was sensitive to both regional stress gradients (Table 2) and yielded consistent scores between dry and wet years (Table 3), whereas the original FQI (Equation (2)) was a poorer indicator in the Boreal Plains and yielded inconsistent scores in the Aspen Parkland. Inter-annual climatic variation can alter vegetation community composition (Van der Valk and Davis, 1978; Van der Valk, 2005) and consequently influence metrics such as species richness. Whereas the original FQI was biased by inter-annual differences in species richness arising from natural climatic variation, the FQI′ was a consistent indicator of plant condition since the relative proportion of native species remained fairly consistent between wet and dry years (Table 3).
In the Boreal Plains, the FQI′ calculated with and without exotic species were both sensitive to the stress gradient (Table 2). In contrast, including exotic species slightly reduced metric sensitivity of the FQI′ in the Aspen Parkland (Table 2). Nevertheless, we agree with Cohen et al. (2004) and Ervin et al. (2006) that there is no empirical evidence to ignore exotic species. Many exotic species were introduced in North America a relatively long time ago and have consequently become established species in these altered ecosystems. Land-use disturbances such as channelization and degradation of buffer zones provide pathways for propagule dispersal of exotic species (Facon et al., 2006; Galatowitsch et al., 1998). Some exotic species are extremely detrimental and can degrade entire ecosystems (Ervin et al., 2006) through mechanisms such as competition and displacement, deterioration of environmental conditions, and weakening of ecosystem resilience.
Once a floristic quality assessment tool has been developed in a region, measuring floristic quality can be done rapidly in the field by a biologist familiar with the regional flora. The two floristic quality indices developed for ecoregions in Alberta provide a site-level evaluation of plant condition based on sampling the wet meadow zone, which expedites field sampling time in comparison to conventional field sampling of the entire marsh. Both Wilson and Bayley (2012) and Murray-Hudson et al. (2012) found that wetland condition assessments based on vegetation sampling in the wet meadow zone were as effective as sampling vegetation along an elevation gradient (e.g. from open water to emergent to wet meadow).
In summary, this study demonstrates that floristic quality indices can provide a good estimate of plant habitat quality in relation to environmental stress and can be used to assess and monitor the condition of constructed and reclaimed marshes. The best overall floristic quality indicator at measuring anthropogenic stressors was the adjusted FQI, or FQI′, which highlighted that existing stormwater ponds and reclamation marshes are not providing acceptable replacement of wetland habitat. We conclude that compensation wetlands need to be designed to replace biological structure and function to mitigate the effect of ongoing loss of wetlands in Alberta. This study demonstrates that current efforts to recover wetland habitat through mitigation practices have not been successful.
We gratefully acknowledge that this research was supported by Alberta Innovates Technology Future (Alberta Ingenuity Fund) and by the North American Waterfowl Management Plan Grant. We also appreciate the time contributed by P. Achuff, B. Cornish, P. Cotterill, D. Fabijan, J. Gould, D. Griffiths, D. Johnson, K. Ottenbreit, K. Timoney and L. Kershaw. We would not have been able to conduct this study without their expertise in Alberta flora and plant ecology.