Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 593 2019 NORTHEASTERN NATURALIST 26(3):593–608 Development of an Ecoregional Floristic Quality Assessment Method for the Northeastern United States Don Faber-Langendoen1,*, Don Cameron2, Arthur V. Gilman3, Kenneth J. Metzler4,5, Richard M. Ring6, and Lesley Sneddon1,7 Abstract - Floristic quality assessments (FQA) are widely used to assess ecological condition, based on plant species’ coefficients of conservatism, or C values, but these values are typically assigned at the state level, limiting their regional consistency. We developed ecoregional C values (eC) using standard expert-based team review, partly informed by preexisting state C values for the 5 EPA ecoregions that cover New England and part of New York. We evaluated a total of 3686 taxa, with separate eC values for each ecoregion in which they occurred. We compared the performance of the C and eC values using the response of FQA metrics (mean C, cover-weighted mean C) to a disturbance gradient, based on readily available datasets in Maine and Vermont. The eC values typically performed moderately better than state values and provide a region-wide tool for assessment methods. All eC values are available on the Universal FQA website. Introduction Floristic quality assessment (FQA) is a robust, botanically based method for assessing the quality of species composition of ecological community occurrences and natural areas (Swink and Wilhelm 1979, Taft et al. 1997). Integral to the method is the assignment of a coefficient of conservatism, or C value, to each plant species in a state or region, based on its tolerance to habitat degradation and dependence on pristine natural habitats and processes (Swink and Wilhelm 1994). The C values span from 0 to 10: the most highly conservative native species (C values >7) are typically found under historic, natural, and restricted ecological conditions and are sensitive to anthropogenic disturbances; whereas, the least conservative native species (C values = 1–3) are adapted to or tolerant of a wide range of anthropogenic disturbances; nonnative species are assigned a value of 0. Integrating the C values of all species at a site into 1 or more FQA metrics can provide a valuable indicator of the condition at a site (DeBerry et al. 2015, Miller and Wardrop 2006, Swink and Wilhelm 1979). C-value development has often been based on state boundaries, where a comprehensive flora is often readily available (for a current list of available state FQAs in the US, see DeBerry et al. 2015, Freyman et al. 2016). In the northeast, compilation 1Conservation Science Division, NatureServe, Arlington VA 22202. 2Maine Natural Areas Program, 93 SHS, Augusta, ME 04333. 3Gilman and Briggs Environmental, Barre, VT 05641. 4Environmental Education/Science Education Department, Southern Connecticut State University, New Haven, CT 06515. 5Current address - Willington, CT 06279. 6New York Natural Heritage Program, Albany, NY 12233.7Current address - Waltham, MA 02451. *Corresponding author - don_faber-langendoen@natureserve.org. Manuscript Editor: Douglas DeBerry Northeastern Naturalist 594 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 of C values were completed in 2011 for the flora of each of 7 contiguous states (Connecticut, Maine, Massachusetts, New Hampshire, New York, Rhode Island, and Vermont) under the leadership of the New England Interstate Water Pollution Control Commission (NEIWPCC; Bried et al. 2012). Despite the success of the methods at the state level, Bried et al. (2012) identified several problems with extending these results to the regional level. First, the differences in C-value ratings among states suggests that a floristic quality assessment in one state may have a different meaning from an assessment in another state. Second, the findings imply that state lists should not be combined into a composite C valu e for the region. By using state boundaries, the assignment of C values limits our understanding of how a species responds regionally to stressors. A species may be a generalist in one part of its range and a specialist in another part. For example, Abies balsamea (L.) Mill. (Balsam Fir), a widespread species in native forests of northern New England and northern New York, regenerates well after clearcutting and fires, and can occasionally be found regenerating on abandoned farmlands and old fields (D’Orangeville et al. 2008). However, it becomes less common and increasingly specialized and may be less tolerant of anthropogenic disturbances in forests of southern New England and southern New York. To address these issues, and to provide a regional approach to wetland assessments based on FQA metrics that use these C values, researchers have begun developing C values at an ecoregional level (e.g., Chamberlain and Ingram 2012, Gianopulos 2015). In this study, we sought to improve the use of FQA for northeastern ecologists by addressing the following objectives: (1) develop, and make publicly available, ecoregional plant lists for the 7 states in the northeast, (2) assign ecoregional C values (hereafter eC) values for all species in the northeast using the knowledge already compiled at the state level (Bried et al. 2012), and (3) evaluate the merits of the eC values for use with FQA metrics for wetland assessment, including metric response to an anthropogenic disturbance gradient and to reference conditions for various ecological community types. Methodology Study area EPA Region 1 includes the 6 New England states: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, and Vermont. Some of the 5 EPA level III ecoregions that cover these states (Regions 58, 59, 82, 83, and 84; Griffith et al. 2009) also extend into New York (Fig. 1). For that reason, we included the New York portion of those ecoregions. All 7 states were also part of the initial Northeast project of Bried et al. (2012). A large part of ecoregion 84 occurs in New Jersey, but no C values were available for that state when this project started. Other New York ecoregions extend southward into Pennsylvania and should be evaluated separately with mid-Atlantic partners (Chamberlain and Ingram 2012). Regional species list by ecoregion. We compiled a plant species list for the 7 states using the USDA PLANTS database (hereafter PLANTS; USDA-NRCS 2016), as was done previously by Bried et al. (2012). We used county distributions Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 595 available from PLANTS to create the initial list of taxa by Level III ecoregion (Fig. 1), then supplemented that dataset with county information from individual state floras. If a species was reported from an ecoregion based solely on a county(s) that straddled the ecoregion line, we added a question mark to its ecoregion distribution. We added readily available information from PLANTS on nativity, growth form, and duration (annual, biennial, perennial). We also consulted available floras, such as Flora Novae Angliae (Haines 2011), New Flora of Vermont (Gilman 2015), and a recently revised catalogue of the vascular plants of New York (Werier 2017), to help assess nativity and distribution. Assigning ecoregional C-values. We assigned eC values based on the standard scale of 0 to 10 (Table 1), guided by the definitions of Bried et al. (2012). For all species that are nonnative to the entire region (New England and New York), we assigned a value of 0, whereas for a species that was native somewhere in the ecoregion, we displayed the ecoregion where it is nonnative as 0, where it is native with a value > 0, and where it is absent, with no score. Although using a value of 0 for all nonnatives is sometimes considered arbitrary (DeBerry et al. 2015), it also recognizes that these species typically occupy the most disturbed human habitats Figure 1. Level III ecoregions (Griffith et al. 2009) in EPA Region 1 (6 New England states). The map shows the full extent of the ecoregions in the US, exce pt for Ecoregion 84, which extends into New Jersey’s coastal plain. Also shown are the intersections of county borders with ecoregional boundaries. The large number of counties contained within one ecoregion made it feasible to use USDA PLANTS distribution to generate a first approximation of an ecoregional species list. Ecoregion names are: 58 = Northeastern Highlands; 59 = Northeastern Coastal Zone, 82 = Acadian Plains and Hills, 83 = Eastern Great Lakes Lowlands; and 84 = Atlantic Coastal Pine Barrens. Also shown are several ecoregions found in New York but not in the 6 New England states (60, 61, 62, and 67) Northeastern Naturalist 596 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 (comparable to native species with a rating of 1 or 2), and their long-term ecological behavior in their new location is still uncertain, but not likely to reflect fidelity to native habitats. Still, future research may show the benefits of assigning negative C-values based on their degree of invasiveness (DeBerry et al. 2015). Practically speaking, we assigned a 0 because the 2011 state FQAs used that approach (Bried et al. 2012). We developed an automated process for an initial set of eC values using the state-based coefficients. First, we compiled all state-based C values from Bried et al. (2012) into a single database, except that we replaced the 2011 New York FQA list with the version developed at a 2015 workshop (Ring 2016). We then calculated an eC value for each species, in each ecoregion in which it was reported, based on the average of the C values for each state, weighted by the percentage of the state’s area found in the ecoregion. For example, Balsam Fir occurs in Ecoregion 83, and in the 2 states, New York and Vermont, covered by the ecoregion. In EPA ecoregion 83, NY contains 90.6% of the ecoregion and VT covers 9.4%. Its state C value in NY was 6, and its state C value was 3 in VT; thus, the calculated eC value for ecoregion 83 was (0.906 x 6) + (0.094 x 3) = 5.72, which we rounded to 6 as the initial value. Finally, we assigned 1 team member to lead the eC review in each ecoregion. Team leads flagged all species needing further review or consultations, as well as revised the calculated eC values where they felt appropriate, to produce a preliminary set of eC values for each ecoregion. Our goal was to produce an operational list of vascular plant taxa suitable for an ecoregional FQA methodology, and not to create a definitive catalogue of the flora of each region. For that reason, we developed the following rules to address the issue of what to do when 1 state assigned a C value to a subspecies or variety, but the other state(s) did not. First, when varieties or subspecies within a species had the same C value, we eliminated the varieties to reduce the taxonomic level of expertise expected to conduct FQAs. Second, if 1 state completed C values for varieties or subspecies, but other states did not, we eliminated the variety and used the lower C value for the species-level C value. Table 1. Guiding definitions for coefficients of conservatism, or C values, assigned to the vascular flora of New York and New England (from Bried et al. 2012). CoC Criteria 0 Non-native species with wide range of ecological tolerances. These are often opportunistic species of intact undisturbed habitats. 1–2 Native invasive or widespread native species that is not typica l of (or only marginally typical of) a particular plant community; tolerant of anthropog enic disturbance. 3–5 Native species with an intermediate range of ecological toleran ces and may typify a stable native community, but may also persist under some anthropogenic disturbance. 6–8 Native species with a narrow range of ecological tolerances and typically associated with a stable community. 9–10 Native species with a narrow range of ecological tolerances, hi gh fidelity to particular habitat conditions, and sensitive to anthropogenic disturbance. Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 597 We conducted an eC-value review at a 2.5-d workshop on 18–20 April 2017 at the Albany Pine Bush center in NY. Our 7-state list initially contained 3697 taxa found in the 5 ecoregions, of which 1172 (32%) are nonnative; the eC value of these non-natives was automatically 0 in any ecoregion in which they occurred (Table 1). For the remaining initial list of 2525 native species, we reviewed all species with an initial eC value range of 3 or more (552 taxa) between any of the ecoregions. For example, our spreadsheet showed that preliminary eC values for Balsam Fir had an ecoregional range of 3, whereas Acer saccharinum L. (Silver Maple) and Acer saccharum Marsh. (Sugar Maple) both had an ecoregional range of 1. Thus, we examined the eC values for Balsam Fir at the workshop, but not for Silver Maple or Sugar Maple. Consistent with other ecoregional methods (e.g., Gianopolus 2015), we did not attempt to reconcile the ecoregional range in C values that differed by 1 or 2 (together 1373 species), as this would have been too demanding an effort for the limited amount of additional value it might have added. We conducted expert review for all 552 species to see if there was an ecoregional basis for the distinction. For example, if the differences in eC values between ecoregions could be explained based on differences in species behavior to anthropogenic disturbances (i.e., more tolerant in one ecoregion, less so in another), then we retained the distinct eC. Otherwise, we revised the eC values to be more consistent across ecoregions. For example, expert review supported the different ecoregional scores for Balsam Fir, as it occurs more commonly in both disturbed areas and undisturbed natural areas in ecoregion 82 and was more restricted to undisturbed natural areas in ecoregions 59 and 83. We added all expert-based eC values back into the database. We minimized the use of subspecies and varieties. Applying FQA requires good botanical knowledge, and the need for this expertise can limit its application (DeBerry et al. 2015). This challenge increases when subspecies or varieties of a species have distinct C or eC values. We retained subspecies in our taxa list whenever they had eC values greater than 2, but we also added a species-based eC value for all species with differing subspecies or variety of eC values by taking the lower of the values, except in the case of the native versus a nonnative subspecies or variety, where we took the average. For example, Viburnum opulus L. (High-bush Cranberry), which is treated as a single species in PLANTS and in Flora Novae Angliae, contains a native subspecies (Viburnum opulus ssp. trilobum) and a European subspecies (Viburnum opulus ssp. opulus). The native subspecies has eC values that vary from 0 (exotic in ecoregion 84) to 3 (ecoregion 83) to 4 (ecoregions 58, 59, 83), whereas the nonnative subspecies has an eC value of 0 wherever it occurs. We retained the subspecies, but we also assigned a species level eC value of 0, 1, or 2, based on the average of the 2 subspecies eC values in those ecoregions. We uploaded all decisions for eC values, including a few changes to be consistent with PLANTS, into our ecoregional database and made them available for public review. In particular, we solicited review from the Technical Advisory Committee established for this project by the New England Interstate Water Pollution Control Commission (Faber-Langendoen 2018). Review comments helped clarify the ecoregional distribution of native and nonnative species, and we added those changes to the database. Northeastern Naturalist 598 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 C-value evaluation. To evaluate the merits of our eC values, we compiled 2 datasets made available by the Maine Natural Areas Program and the Vermont Natural Heritage Inventory. Both programs are part of the NatureServe Network and have, as part of their mission, the documentation of exemplary (reference) or best extant locations of natural communities across their respective states. Documentation includes assigning an element occurrence rank (EORANK), which provides an estimate of the ecological integrity and probability of persistence of the occurrence based on an assessment of its on-site condition, size, and landscape context, and which typically reflects changing levels of human disturbance or stress (Faber- Langendoen et al. 2016, NatureServe 2002). We used state-based assignment of condition and landscape-context ratings to determine the “condition gradient”. The gradient varies from A (intact or excellent rating) to D (severely disturbed or poor rating) (Faber-Langendoen et al. 2016:table 1). Both programs collect vegetation plot data within an occurrence. Plots are 400 m2 and contain a full vascular plant species list, with cover estimates, sufficient for calculating C -value metrics. For the Maine data, some locations were sampled using a meandering species list with cover values; we included these data because species richness and mean C values were not significantly different between this method and their plot-data method. The Maine data included 151 wetlands sites, with the bulk of the data concentrated in the 2 best-documented types: large floodplain forests (115) and northern swamps (25) (Schlawin 2018). Maine data collection gave greater emphasis to documenting the full range of swamp and floodplain conditions, so sites spanned the condition (ecological integrity) gradient from excellent (A grade) to fair (C grade). In addition, Maine researchers used a scoring technique that provided a continuous scoring of condition from excellent (4.0) to poor (1.0). Vermont data included 398 plots varying in ecological condition from excellent (A) to fair (C) across 4 wetland types: hardwood swamps (114), softwood swamps (118), open alkaline peatlands (47), and open acidic peatlands (119). Vermont data emphasized the most exemplary sites (i.e., only 23 sites had a fair rating); thus, we used their data to focus on comparing A and B reference conditions. We standardized data from both states to PLANTS taxonomy. We used plot coordinates to determine the ecoregion in which each plot occurred, and used eC values for that ecoregion to calculate both mean C and cover-weighted mean C (C-W mean C) metrics, 2 widely used metrics for wetland condition assessments (DeBerry et al. 2015). Most of Vermont (~90%) is covered by ecoregion 58, with small areas of ecoregion 83 (9%), and 59 (~1%). Maine is covered almost equally by Ecoregions 82 (~55%) and 58 (~42%), with a small sliver of ecoregion 59 (~3%). Maine researchers calculated the FQA metrics using both the old state C values and the new eC values in their database, and then exported the metric scores to the NatureServe team. Vermont data were imported into NatureServe’s EcoObs database (Faber-Langendoen et al. 2016), where we calculated FQA metric scores. Vermont data are also being exported to VegBank (vegbank.org), where they will be made publicly available. We statistically tested whether the responses of the FQA metrics to a condition gradient were improved when based on the eC values as compared to the original Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 599 state C values. Our data sets were all within a state, so our validation test was rather stringent—that is, we tested whether, within a state that has 2 or 3 ecoregions, metrics based on eC values improved our assessment of condition over the state C values. We first conducted regression analyses to evaluate the performance of FQA metrics based on eC values to those based on state C values (Maine data). We assessed the performance of the 2 regressions using several standard statistical measures: (a) higher r2 value in a regression of FQA metrics in relation to a condition gradient; (b) a lower Akaike information criterion (AIC) score, which provides an estimator of the relative quality of the 2 regression models, with lower values indicating a better fit; and (c) a relative likelihood statistic that compares the likelihood that the eC values model is better than the state C values model, with higher values indicating a better model. We conducted a second, categorical comparison of FQA metric scores in relation to A, B, and C condition ratings (Maine and Vermont data). We used 1-way ANOVA, followed by a Tukey HSD post hoc test to test mean scores among condition ratings. We assessed performance of the FQA metric scores for state C-values versus eC values using the spread of means between A and C ratings (greater spread indicates better separation between excellent and fair condition ratings) and the root mean square error (RMSE) (smaller values mean less unexplained variance and a better fit). We conducted all analyses in R version 3.4.2 for statistical analyses (R Development Team 2017). Results Regional FQA species list Our final list of species for the 5 ecoregions combined contained 3411 native and nonnative species, of which 1068 (31.0%) were nonnative and 2343 were native. For the native species, we retained 137 subspecies or varieties that had distinct C values; subspecies or varieties were excluded when only 1 of the subspecies occurred in our region. We also retained 30 native hybrids that had distinct eC values, for a total of 2510 taxa. For the nonnative species, we also retained 15 subspecies or varieties (to track species that had both native and exotic subspecies or varieties), 12 hybrids, and 81 genera whose species were all exotic (a field biologist would only need to know the genus to confidently assign a C value of 0), for a total of 1176 taxa. Thus, our operational final list included 3686 taxa with 1176 exotic taxa (Table 2). Ecoregional distribution and C value Distribution. We initially generated a list of taxa by ecoregion, which we then revised through team-based expert review. Our species-level counts by ecoregion were: ecoregion 58: 3059 species, of which 919 are exotic; ecoregion 59: 2809 species, 989 exotic; ecoregion 82: 2302 species, 1076 exotic; ecoregion 83: 3072 species, 1066 exotic; and ecoregion 84: 2019 species, 481 exotic. The count for ecoregion 84 is low because less than half of the ecoregion was included in our study area. Northeastern Naturalist 600 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 Ecoregional C value. Prior to the workshop, we generated an initial list of 552 native taxa (21.8% of all native taxa) with differences in scores of 3 or more between ecoregions. Based on workshop review, we reduced this number to 143 native taxa (5.7% of all native taxa) with ecoregional ranges of 3 or more (Table 2). Furthermore, prior to the workshop, 109 species had ecoregional ranges of 5 or more, whereas after the workshop only 9 did; we report these 9 species in Table 3. All final eC values are provided in Supplemental File 1 (available online at http:// www.eaglehill.us/SENAonline/suppl-files/26-3-N1693-Faber-Langendoen-s1, and for BioOne subscribers, at https://dx.doi.org/10.1656/N1693.s1) and have been posted, by ecoregion, on the universalFQA.org website. Table 3. The taxa with the largest ecoregional range in C-values (≥5). Nine species are shown, in alphabetical order. See Figure 1 for ecoregion map. The accepted symbol is taken from USDA NRCS (2016). eC Value by ecoregion Symbol Species 58 59 82 83 84 Range CHBE4 Chenopodium berlandieri Moq. (Pitseed Goosefoot) 5 5 2 8 8 6 CUCE Cuscuta cephalanthi Engelm. (Buttonbush Dodder) 6 5 10 6 5 LEMO8 Leymus mollis (Trin.) Pilg. (American Dunegrass) 5 8 6 10 5 POBU2 Polygonum buxiforme Small (Box Knotweed) 4 5 7 1 5 6 POER2 Polygonum erectum S.F. Blake (Leathery Knotweed) 4 5 2 8 6 6 PORA3 Polygonum ramosissimum Michx. (Bushy Knotweed) 4 6 5 1 3 5 PSMI6 Pseudognaphalium micradenium (Weatherby) Nesom 10 5 10 10 5 (Small Rabbit-tobacco) SALY2 Salvia lyrata L. (Lyreleaf Sage) 10 3 10 7 THOC2 Thuja occidentalis L. (Northern White Cedar) 5 8 3 5 7 5 Table 2. Spread of eC value ranges across all native taxa. Ecoregional range represents the maximum difference in eC values across the 5 ecoregions. The 2510 native taxa include 2343 species, 137 subspecies and varieties, and 30 hybrids. There are 1176 nonnative taxa, including 1068 species, 15 subspecies or varieties, 12 hybrids, and 81 genera. See text fo r details. Total taxa Ecoregional range of C values Pre Workshop Post Workshop 0 836 986 1 688 805 2 466 576 3 298 112 4 145 22 5 54 5 6 33 3 7 18 1 8 4 0 Total native taxa 2542 2510 Total nonnative taxa 1172 1176 Total taxa 3714 3686 Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 601 Ecoregional C-value validation Evaluation of ecoregional C-value metrics in response to condition gradient. In Maine, the C-W mean C metric showed a positive correlation to a condition gradient using both old state and new eC values, affirming the overall usefulness of this FQA metric in tracking wetland condition (Fig. 2a). The relationship was much stronger for swamps than for floodplain forests (Fig. 2b, c), indicating that floodplain forests tended to have more generalist species in reference sites. In addition, for all types evaluated collectively and for large-river floodplains, the metrics based on the new eC values typically showed a better fit to the data model, based on the AIC statistic, as compared to the metrics based on state C-values (Table 4). There was a slight, but not significant, improvement in the responsiveness of eC values to the gradient, as shown in the slightly steeper slope of the relationship, especially for the “all wetland types” regression (Fig. 2a). A steeper slope is an improvement because it means that there is a stronger metric response to the condition gradient. The eC value metric scores also tended to be higher than the state C-value metric scores. A categorical view of the FQA metrics response to the condition gradient was also informative for both Maine and Vermont data. The C-W mean C scores were significantly different between either A versus B and C, or A and B versus C for both eC and state C-values (Figs. 3a, b). However, for Maine data, the mean C metric (which does not include abundance of species, only their presence/absence) showed improved ability to discriminate between for A, B, and C conditions, as compared to C-W mean C (Fig. 3c). The 2 sets of C-values produced comparable results, depending on the metric and state data being used. Thus for Maine, using C-W mean C (Fig. 3a), eC performed better, based on a greater spread of mean values (eC = 1.219 vs. state = C 1.051) and less unexplained variance (RMSE eC 0.828 vs state C 0.910), whereas using mean C (Fig. 3c), eC performed better based on a greater spread of means (eC = 1.158 vs. state C = 0.906) but had more unexplained variance (RMSE eC = 0.569 vs state C = 0.510). For Vermont, using C-W mean C (Fig. 3b), Table 4. Response of the cover-weighted mean C and mean C metrics to a condition gradient when based on state C values or eC values (see also Fig. 2). For AIC (Akaike information criterion), lower values indicate a better fit of the regression model. For the relative likelihood statistic, higher values indicate a better model. An asterisk(*) denotes P < 0.001. All data are from Maine. r2 AIC Relative likelihood State C eC State C eC State C eC All wetlands Mean C 0.28* 0.31* 264.4 236.3 less than 0.001 0.99 C-W mean C 0.19* 0.26* 411.4 379.2 less than 0.001 0.99 Large floodplain Mean C 0.12* 0.19* 169.8 159.3 less than 0.01 0.99 C-W mean C 0.11* 0.17* 315.2 287.4 less than 0.001 0.99 Northern swamp Mean C 0.60* 0.55* 29.3 37.8 0.99 0.01 C-W mean C 0.52* 0.49* 46.8 50.8 0.88 0.12 Northeastern Naturalist 602 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 Figure 2. Maine data. Linear regression, showing response of the coverweighted mean C metric to a condition gradient, based on the ecological integrity assessment (EIA) score for condition (1.0 = poor condition–4.0 = excellent condition). (a) all wetland types, (b) large floodplain forest, and (c) northern swamp. Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 603 Figure 3. Maine and Vermont data, comparing categorical response of FQA metric scores to condition ratings across all types. (a) C-W mean C for Maine,( b) C-W mean C for Vermont, and (c) mean C for Maine. Box plots summarize the FQA metric scores; the heavy line crossing the box is the median, the bottom and top of the box are the lower and upper quartiles, and the whiskers are the minimum and maximum values. Clusters with the same letter code are not significantly different (Tukey HSF: P < 0.05). Northeastern Naturalist 604 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 performance varied; eC performed less well based on a smaller spread of means (eC = 0.765 vs state C = 1.015), but it performed better based on less unexplained variance (RMSE eC = 0.990 vs state C = 1.136). The results appear to be affected by sample size because there were comparatively fewer Maine data points for Aranked sites and comparatively fewer Vermont data points for C-ranked sites. Discussion Ecoregional C values Completion of the eC values for the Northeast furthers the objective of having an FQA methodology that can be applied regionally. We have posted our list of species and their eC values by ecoregion on the universal FQA website (www. universalFQA.org), making them publicly accessible for users across the region (see also Supplemental File 1, available online at http://www.eaglehill.us/SENAonline/ suppl-files/26-3-N1693-Faber-Langendoen-s1, and for BioOne subscribers, at https://dx.doi.org/10.1656/N1693.s1). We also upgraded NatureServe’s EcoObs database to serve as a regional database tool, one that readily provides FQA calculations. The database already includes the capability to manage species (including synonymy between different floristic standards, such as PLANTS and Flora Novae Angliae) and their FQA values. Thus, it provides a suitable data management tool for FQA data for the Northeast. There are challenges to implementing eC values. The first hurdle is in creating an ecoregional species list because plant distribution data are not comprehensively available at the county level. Further, after the list is established, it can be challenging to maintain because plant taxonomic updates and new distribution data will need to be tracked. Still, our current ecoregional lists, with their emphasis on species-level taxonomy, appear to be comprehensive enough to meet most FQA needs. Some researchers have adapted the method to make it even less taxonomically demanding, by relying only on dominant species or on readily recognizable species (Bourdaghs 2012). The number of native ecoregional variants with a range of 3 or more is relatively low as a percentage of the regional (or state) flora (5.7%). A future consideration in developing ecoregional species lists would be to coordinate them with state-based species lists, where a state C-value is amended to include ecoregional variants, where needed. Users could then choose either a statewide C value or an eC value. Our eC values for New York did not include Ecoregion 60 (High Allegheny Plateau) and our assessment of Ecoregion 84 (Coastal Plain) did not include the New Jersey portion. A completed state FQA list was recently completed for New Jersey (Kathleen Walz, New Jersey Department of Environemental Protection, Trenton, NJ, 2017 pers. comm.). By completing the New York ecoregions and integrating New Jersey FQA results into a Northeast and Mid-Atlantic ecoregional product, EPA and wetland assessment partners would have access to a comprehensive set of ecoregional scores from Maine to the Carolinas (Chamberlain and Ingram 2012, Gianopolous 2015). Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 605 Evaluation of C-values and FQA metrics in response to condition gradient The mean C and C-W mean C metrics based on either state or ecoregional eC values showed a positive response to a wetland disturbance gradient, validating these metrics as valuable tools for assessing wetland condition (Fig. 2). This finding is consistent with many other studies that have used these metrics to assess wetland condition (e.g., Miller and Wardrop 2006). When based on eC values, these metrics often showed a minor to moderate improvement in tracking changes in condition, suggesting eC values better reflect changes in a species behavior across its range, even within a state. Further investigation is needed to determine why mean C performed better than the C-W mean C metric in distinguishing between A, B, and C condition ranks (Fig. 3). Although the response of FQA metrics to the condition gradient were not substantial, they may also reflect the degree to which many northeastern states are relatively small and ecologically homogeneous, such that ecological behavior did not differ strongly within any given state. They may also reflect our improved knowledge of species behavior across the region now that many years have passed since the release of the initial state-based C values. This improved knowledge may also account for why eC values tended to be higher, on average, than state C values in Figures 2 and 3; i.e., we are more confident that certain species reflect minimally disturbed conditions. There have been few statistical comparisons of eC or state C values, including when revisions are made to those values (Ring 2016, Swink and Wilhelm 1994). We provide only a few limited tests here. There is a need to encourage development of a priori datasets and statistical tests to guide proposals for making such revisions, including datasets that span a diverse set of ecosystems across the region and the full disturbance gradient. But evaluation and interpretations of revisions will be challenging, given that thousands of species are being re-evalu ated. Regional FQA metrics and assessments of ecological condition Multiple studies have demonstrated the effectiveness of various FQA metrics in helping inform our understanding of ecological condition, particularly for wetlands (Bourdaghs et al. 2006, Bried et al. 2013, DeBerry et al. 2015, Matthews 2003, USEPA 2002, Wardrop et al. 2007), though validation is an ongoing process (DeBerry et al. 2015, Matthews et al. 2015). The US Environmental Protection Agency (USEPA) recognized the merits of the approach and has encouraged state and regional development of the method for state wetland assessments (Medley and Scozzafava 2009, USEPA 2002). FQA is also used in combination with other wetland assessment metrics, such as the multi-metric approach of NatureServe’s Ecological Integrity Assessment method (Faber-Langendoen et al. 2016, 2019), and EPA’s Vegetation Multi-Metric Index (VMMI) for the National Wetland Condition Assessment (Serenbetz 2016). Our completion of this product provides the possibility of providing consistent FQA metric scores for scoring ecological condition of wetland reference datasets that span multiple states. These datasets can serve as benchmark sites for diverse projects, including restoration and mitigation evaluations and statewide wetland assessments (Brooks et al. 2016). Northeastern Naturalist 606 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 It is now widely recommended that the interpretation of ecological condition using FQA metrics (such as mean C or C-W mean C) be assessed within relatively similar ecological community types (Bourdaghs 2012, Ervin et al. 2006, Johnston et al. 2009) because types differ in the degree to which generalist native species are part of the characteristic species combination. For example, because floodplain forests and salt marshes regularly experience flooding, they tend to have more generalist species even under natural disturbance regimes, and thus, high-quality examples of floodplain forests have lower mean C-values than bogs and fens (Bourdaghs 2012). Groupings of similar types can be done using standardized regional or national classifications, such as the US National Vegetation Classification (usnvc. org). With the availability of eC values, the use of FQA metrics can be reliably extended across the multiple geographic regions in which the community types are found. The combination of ecoregional C values and regionally or nationally based classifications provides users with strong tools for assessing and monitoring the ecological condition of ecosystems at multiple geographic scales. Acknowledgments We are grateful for financial assistance from the New England Interstate Water Pollution Control Commission, through EPA project funds (EPA CODE: RFA17017), and for the leadership and project support provided by Kimberly Roth, NEIWPCC Environmental Analyst. She facilitated the process of setting up the Technical Advisory Committee (TAC), hosting calls, and providing opportunities for us to present our results at various meetings, particularly for the New England Biological Assessment of Wetlands Work Group (NEBAWWG). The TAC provided helpful guidance throughout the duration of the project. We especially appreciated Charlie Hohn’s review of the ecoregional C values and Sandy Crystall’s editorial review. Developing the overall methodology and data management for the Northeast Floristic Quality Assessment (FQA) tools has been a team effort. We especially thank Kristin Snow and Mary Harkness for their assistance with compiling coefficients of conservatism (C values), and for improving the EcoObs database to manage all FQA information. Michael Lee also assisted with EcoObs database re visions. We thank Everett Marshall and Eric Sorenson of the Vermont Natural Heritage Program and Justin Schlawin and Kristen Puryear of the Maine Natural Heritage Program for sharing their plot data for use in the validation process. Patrick McIntyre, NatureServe ecologist, stepped in to help complete the statistical analyses. Finally, Will Freyman was receptive to the idea of improving the Universal FQA Calculator website (universalfqa.org), and provided his own insights and guidance on how we might make the improvements. We thank Michele Bottiaux, at NatureServe, for all the enhancements to the Universal FQA Calculator website. Literature Cited Bourdaghs, M. 2012. Development of a Rapid Floristic Quality Assessment. Minnesota Pollution Control Agency, St. Paul, MN. Bried, J.T., K.M. Strout, and T. Portante. 2012. Coefficients of conservatism for the vascular flora of New York and New England: Inter-state comparisons and expert opinion bias. Northeastern Naturalist Special Issue 19 (Special Issue 6 ):101–114. Bried, J.T., S.K. Jog, and J.W. Matthews. 2013. Floristic quality assessment signals human disturbance over natural variability in a wetland system. Ecological Indicators 34:260–267. Northeastern Naturalist Vol. 26, No. 3 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 607 Brooks, R.P., D. Faber-Langendoen, G. Serenbetz, J. Rocchio, E.D. Stein, and K. Walz. 2016. Toward creating a national Reference Wetlands Registry. National Wetlands Newsletter 38(3):7–11. Chamberlain, S.J., and H.M. Ingram. 2012. Developing coefficients of conservatism to advance floristic quality assessment in the Mid-Atlantic region. Journal of the Torrey Botanical Society 139:416–427. DeBerry, D.A., S.J. Chamberlain, and J.W. Matthews. 2015. Trends in floristic quality assessment for wetland evaluation. Wetland Science and Practice (June 2015):12–22. D’Orangeville, L., A. Bouchard, and A. Cogliastro. 2008. Post-agricultural forests: Landscape patterns add to stand-scale factors in causing insufficient hardwood regeneration. Forest Ecology and Management 255:1637–1646. Ervin, G.N., B.D. Herman, J.T. Bried, and D.C. Holly. 2006. Evaluating non-native species and wetland-indicator status as components of wetland floristic assessment. Wetlands 26:1114–1129. Faber-Langendoen, D. 2018. Northeast regional floristic quality assessment tools for wetland assessments. Contributing authors: D. Cameron, A.V. Gilman, K.J. Metzler, R.M. Ring, M. Bottiaux, K. Snow, and L. Sneddon. NatureServe, Arlington VA. Available online at http://neiwpcc.org/ our–programs/wetlands–aquatic–species/nebawwg/nqa/. Accessed 5 July 2019. Faber-Langendoen, D., W. Nichols, J. Rocchio, K. Walz, J. Lemly, R. Smyth, and K. Snow. 2016. Rating the condition of reference wetlands across states: NatureServe’s ecological integrity assessment method. National Wetlands Newsletter 38:12–16. Faber-Langendoen, D., J. Lemly, W. Nichols, J. Rocchio, K. Walz, and R. Smyth. 2019. Development and evaluation of NatureServe’s multi-metric Ecological Integrity Assessment method for wetland ecosystems. Ecological Indicators 104:7 64–775. Freyman, W.A., L.A. Masters, and S. Packard. 2016. The Universal Floristic Quality Assessment (FQA) Calculator: An online tool for ecological assessment and monitoring. Methods in Ecology and Evolution 7:380–383. Gianopolus, K, 2015. Coefficient of conservatism database development for wetland plants occurring in the southeastern United States. Report to the EPA, Region 4. North Carolina Department of Environment and Natural Resources, Division of Water Resources, Wetlands Branch. Raleigh, NC. 33 pp. Gilman, A.V. 2015. New Flora of Vermont Memoirs of the New York Botanical Garden, Volume 110). The New York Botanical Garden Press, Bronx, NY. 608 pp. Griffith, G.E., J.M. Omernik, S.A. Bryce, J. Royle, W.D. Hoar, J.W. Homer, D. Keirstead, K.J. Metzler, and G. Hellyer. 2009. Ecoregions of New England (color poster with map [map scale 1:1,325,000], descriptive text, summary tables, and photographs). US Geological Survey. Reston, VA. Haines, A. 2011. New England Wild Flower Society’s Flora Novae Angliae: A Manual for the Identification of Native and Naturalized Higher Vascular Plants of New England. Yale University Press, New Haven, CT. 973 pp. Johnston, C.A., J.B. Zedler, M.G. Tulbure, C.B. Frieswyk, B.L. Bedford, and L. Vaccaro. 2009. A unifying approach for evaluating the condition of wetland plant communities and identifying related stressors. Ecological Applications 19:1739–1757. Matthews, J.W. 2003. Assessment of the floristic quality index for use in Illinois, USA, wetlands. Natural Areas Journal 23:53–60. Matthews, J.W., G. Spyreas, and C.M. Long. 2015. A null-model test of floristic quality assessment: Are plant species’ coefficients of conservatism valid? Ecological Indicators 52:1–7. Northeastern Naturalist 608 D. Faber-Langendoen, D. Cameron, Arthur V. Gilman, K.J. Metzler, R.M. Ring, and L. Sneddon 2019 Vol. 26, No. 3 Medley, L., and M. Scozzafava. 2009. Moving toward a national floristic quality assessment: Considerations for the EPA National Wetland Condition Assessment. National Wetland Newsletter 31:6–9. Miller, S.J., and D.H. Wardrop. 2006. Adapting the floristic quality assessment index to indicate anthropogenic disturbance in central Pennsylvania wetlands. Ecological Indicators 6:313–326. NatureServe. 2002. Element occurrence data standard. Available online at http://www. natureserve.org/conservation-tools/standards-methods/element-occurrence–data-standard. Accessed 5 Jully 2019. R Core Team. 2017. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Ring, R. 2016. Developing coefficients of conservatism values for New York’s native flora. New York Flora Association Newsletter (Spring 2016):27(2). Schlawin, J. 2018. The status of floodplain forests along Maine’s Western Rivers. MOHF Project #152–03–02, Final Report, January 2018 Maine Natural Areas Program, Maine Department of Agriculture, Conservation, and Forestry, Augusta, ME. Serenbetz, Gregg. 2016. National Wetlands Condition Assessment 2011–2016: Lessons learned and moving forward. National Wetlands Newsletter 38(3):17–20. Swink, F., and G. Wilhelm. 1979. Plants of the Chicago Region. Revised and expanded edition with keys. The Morton Arboretum, Lisle, IL. 922 pp. Swink, F. and G. Wilhelm. 1994. Plants of the Chicago Region. 4th Edition. Morton Arboretum, Lisle, IL. 1390 pp. Taft, J.B., G.S. Wilhelm, D.M. Ladd, and L.A. Masters. 1997. Floristic quality as sessment for vegetation in Illinois: A method for assessing vegetation integrity. Erigenia 15:3–95. US Department of Agriculture, NRCS (USDA-NRCS). 2016. The PLANTS Database. Natural Resources Conservation Service, National Plant Data Team, Greensboro, NC. Available online at www.plants.usda.gov. Accessed 18 April 2017. US Environmental Protection Agency (USEPA). 2002. Methods for evaluating wetland condition: Using vegetation to assess environmental conditions in wetlands. EPA- 822-R-02-020, Office of Water, US Environmental Protection Agency, Washington, DC. Wardrop, D.H., M.E. Kentula, D.L. Stevens, S.F. Jensen, and R.P. Brooks. 2007. Assessment of wetland condition: An example from the Upper Juniata watershed in PA, USA. Wetlands 27:416–431. Werier, D. 2017. Catalogue of the Vascular Plants of New York State. Memoirs of the Torrey Botanical Society, Volume 27. The New York Botanical Garden, Bronx, NY. 542 pp.