Searching for Indicator Species of High Floristic Quality Depressional Wetlands in the US Southern Plains

Floristic Quality Assessment requires compiling a full list of vascular plant species for the wetland. Practitioners may lack the time and taxonomic skills for full-community vegetation surveys, especially when wetlands are large and complex. In this paper we broadly ask whether floristic quality indicator species may exist for wetlands, specifically evaluating indicator species potential for high floristic quality depressional wetlands in the US southern plains. Candidate indicators were identified for a broader context (depressions across Oklahoma prairie ecoregions) and narrower context (depressions in the northern Central Great Plains of Oklahoma) and evaluated based on performance, validity, and robustness criteria. Nine individual species and two species pairs showed exclusivity and ubiquity for high floristic quality, with their value generally improving in the narrower context. However, the overall best indicator (Eleocharis compressa) frequently occurred (> 20 % rate) in lower quality validation sites, and all indicators were lacking in one or more criteria. Combining E. compressa with select other candidates (Ammannia coccinea, Juncus torreyi, Leersia oryzoides) may compensate for weaknesses of individual species but the combinations may rarely be found across the region, suggesting they may not be useful in practice or that high-quality conditions are in fact scarce. Overall, these results offer mixed support for relying on indicator species to rapidly identify or verify high floristic quality depressional wetlands in the US southern plains. We recommend similar studies with larger datasets in other regions and testing other quality levels (low, moderate) before broadly concluding whether floristic quality indicator species may exist for wetlands.


Introduction
Assessments of wetland ecological health often use vegetation-based tools and indicators. Floristic Quality Assessment (FQA) is widely adopted in North America (Spyreas 2019) and for decades has featured strongly in wetland applications including the recent U.S. National Wetland Condition Assessment (USEPA 2002;Magee et al. 2019). Floristic quality is commonly indexed as the product of species richness and mean ecological conservatism (Kutcher and Forrester 2018). Conservatism is estimated by expert opinion and ranges from nonnative and weedy native species at the low end to species of minimally human-disturbed areas at the high end (Taft et al. 1997). These subjective scores assigned to individual species can align with empirically derived scores (Bried et al. 2018) and with a variety of functional traits (Ficken and Rooney 2020), and the index ratings often correspond, to varying degrees, with ecological degradation and human footprint in and around wetlands (DeBerry et al. 2015).
The main requirement and challenge of FQA is compiling a complete species list for the wetland. Although mean conservatism may be robust to some level of This article belongs to the Topical Collection: Applied Wetland Science. misidentifications and failed detections, all extant vascular plants should be detected and accurately identified to species level (Spyreas 2019). The requisite field experience and taxonomic skills may be lacking (Noss 1996;Drew 2011; Spyreas 2019) and many plant species are unidentifiable at certain times of the growing season. Furthermore, compiling a full list takes time, whether setting up plots and transects or meandering throughout the wetland. Wetland researchers have begun to propose shortcuts around full-community FQA, such as ignoring graminoids and non-dominant taxa (Chamberlain and Brooks 2016). Indicator species (Siddig et al. 2016) offer another possibility but have not, to our knowledge, appeared in FQA research or applications. Focusing on a few target species linked to preset levels of floristic quality would ease the skill requirement and improve field work efficiency; in theory a well-supported indicator species could reduce the field visit to minutes or even seconds depending on how fast its detected.
High quality or minimally altered wetlands are sought for regulatory protection and used as baselines for assessing ecological disturbance and management actions Herlihy et al. 2019). Tools for rapid identification of potential reference wetlands are needed, especially when such conditions are rare over large areas containing abundant wetlands. An approach of indicator species, sufficiently calibrated (Bried et al. 2019), is one option worthy of consideration. Such species may be used for example to rapidly screen wetlands for potential regulatory protection and follow-up intensive assessments in the development permitting process (Stapanian et al. 2013).
In this paper we ask if certain species can represent high floristic quality as defined by relatively conservative species-rich plant assemblages. We were specifically interested in finding these species for depressional wetlands in the US southern plains, a region stretching from Texas through Nebraska between the Rockies and Xeric Plains to the west and the Southern Appalachians and Temperate Plains to the east (Kentula and Paulsen 2019). We focus on the prairie ecoregions of Oklahoma where establishing reference standards is a priority of the Wetland Program Plan (www. ok. gov/ wetla nds) and where previous research has suggested floristic quality criteria for potential reference wetlands (Bried et al. 2014;Gallaway et al. 2019). Because Oklahoma's prairie ecoregions and dominant land use extend into neighboring states, developing efficient and reliable assessment tools in Oklahoma is a step towards rapid field identification (or verification) of potential reference-quality wetlands in the southern plains. Our broader question here, beyond the study region, is whether indicator species of wetland floristic quality may exist.

Study Area
Oklahoma is centrally located in the US southern plains and intersected by a dozen Level III ecoregions (Omernik and Griffith 2014), two-thirds of which are largely non-forested and classified as "prairie" (Fig. 1). Historically the prairie ecoregions were dominated by short-grass, mixedgrass, or tall-grass vegetation communities but now they are dominated by pastures and rangeland (Hoagland 2000;Tyrl et al. 2007). The distribution of vegetation types is broadly divided among the ecoregions along a precipitation gradient from driest (< 50 cm annual rainfall) in the High Plains and Southwestern Tablelands to wettest (as much as 150 cm) in the South Central Plains (Hoagland 2000;Gallaway et al. 2019). Wetlands of Oklahoma occur in a diverse topography of hills, ravines, flats, sand dunes, and salt plains, and in four national hydrogeomorphic classes: depressional, lacustrine fringe, riverine, and slope. Depressional wetlands are abundant in Oklahoma and include open (water outlet) and closed (no outlet) shallow surface-water impoundments, open and closed natural contour basins with a confining layer, and groundwater basins in sandy soils where the water table is close to the surface (Dvorett et al. 2012).

Vegetation Surveys
Surveys occurred from mid-May to early August 2012-2018 across several Oklahoma prairie ecoregions, with intensified effort in the depression-rich Central Great Plains north of Interstate 40 (Fig. 1). Depressional study sites (91 total) were located using National Wetlands Inventory data, the National Wetland Condition Assessment probability sample (Kentula and Paulsen 2019), and several Oklahoma-based sources (Bried et al. 2014;. All sites exhibited seasonal to semi-permanent hydroperiods and dominance of scrub or herbaceous communities (i.e., non-forested wetlands).
At 37 sites we used the plot-based methodology of the National Wetland Condition Assessment (USEPA 2011; Kentula and Paulson 2019). Vascular plant species were identified within and overhanging five square plots (100 m 2 each) arranged to representatively assess the wetland and conform with its size or shape (see USEPA 2011 for details). Additional species encountered outside the plots were recorded and used for analysis. At the remaining 54 sites we used meander surveys over the entire accessible area (smaller sites, < 0.5 ha) or in representative vegetation zones until no new species were recorded.

3
The different methods (plot-based vs. meander) produced comparable plant checklists (21.0 vs. 18.3 species on average, t-test P = 0.160) and apparently did not confound floristic quality estimates (Bried et al. 2016). We identified a grand total of 370 species belonging to 222 genera and 87 families.

High Floristic Quality
We calculated site floristic quality using FQI = S is number of species, and C is the species conservatism value (integers 0 to 10) obtained from Ewing and Hoagland (2012) or B. Hoagland (pers. comm.). Sites with FQI ≥ 20.0 were designated as high-quality wetlands; this criterion is empirically supported for Oklahoma wetlands (Bried et al. 2014;Gallaway et al. 2019) and has been used in other wetland assessment programs (Matthews and Endress 2008). Additionally, west of the Interstate-35 corridor ( Fig. 1) we lowered the benchmark (FQI ≥ 13.0 following Gallaway et al. 2019) to account for precipitation stress and naturally lower wetland quality compared to wetter eastern Oklahoma. According to these criteria our sample contained 47 high floristic quality depressions, with 34 located in the northern (north of I-40) Central Great Plains ( Fig. 1). High quality depressions contained significantly more species than lower quality depressions (23.1 vs. 15.3 species on average, t-test P < 0.001) but mean C values were similar (3.35 vs. 3.19, t-test P = 0.202).

Indicator Performance
Indicator species were extracted and valued based on their occurrence fidelity and constancy for high quality sites (FQI ≥ 20.0, or FQI ≥ 13.0 west of Interstate 35) compared to lower quality sites. Indicator fidelity measures exclusivity to the target group and indicator constancy measures ubiquity or frequency of occurrence in that group, such that strongest indicators occur exclusively (max Fidelity = 1) and at all locations (max Constancy = 1) in the target group. In practice, an indicator's fidelity may be viewed as the chance it will correctly identify a target site (positive predictive value) and constancy as the chance of finding the indicator (Dufrêne and Legendre 1997;De Cáceres and Legendre 2009).
We evaluated species individually and in all possible pairs and triplet combinations assuming multispecies indicators cover more environmental space and heterogeneity than any single species (De Cáceres et al. 2012). We used the indicspecies package (De Cáceres and Legendre 2009) to generate the input matrix and calculate performance metrics (Fidelity, Constancy). Candidate indicators were selected at thresholds of 0.6 Fidelity and 0.25 Constancy, reasonable minimum values applied in previous work (De Cáceres et al. 2012;Bachand et al. 2014;Bried et al. 2014). We used 10,000 percentile bootstrap samples to establish 95 % confidence intervals for the Fidelity and Constancy of each indicator (De Cáceres et al. 2012). To limit the candidate pool and avoid selecting upland and weedy species, we analyzed those with facultative, facultative wetland, or obligate regional status (USACE 2010) and at least moderate Oklahoma conservatism value (C ≥ 4) based on C-value availability and distribution between quality levels (Fig. 2). To further reduce the candidate indicators, we removed indicators whose occurrence pattern was nested within that of other indicators or when subsets of indicators provided the same target group coverage as the full set (De Cáceres et al. 2012).
We repeated the analysis for depressional wetlands of Oklahoma prairie ecoregions ("prairie depressions") and for depressional wetlands of the northern Central Great Plains in Oklahoma ("CGP depressions"), expecting stronger indicators for CGP depressions due to increased spatial-environmental context (Bried et al. 2019). For both contexts we used randomly drawn two-thirds samples to find indicator species, giving us 61 prairie depressions and 48 CGP depressions. We drew proportionate numbers of high and lower quality sites and repeated the draws and analysis five times per context to help assess robustness of the indicators.

Indicator Validity
Valid indicator species should occur with relatively high frequency in high floristic quality depressions and low frequency elsewhere. We tested this discriminative ability in candidate indicators using the remaining one-third of study sites, which included 30 prairie depressions and 22 CGP depressions each split evenly into high and lower quality. Indicator validity was measured based on frequency of occurrence in high quality sites (observed true positive rate) and lower quality sites (observed false positive rate) relative to nominal, expected, and predicted rates. Specifically, we expected valid indicators to occur in high-quality sites at rates that equal or exceed Constancy and minimum positive predictive value (Fidelity LCL), and to occur infrequently (< 0.1 preferably, < 0.2 minimally) at lower-quality sites. Additionally, given the importance of false-positive errors in practice, we tested a conservative falsepositive criterion (Observed ≤ 1 -Fidelity UCL) to provide stronger validation and counterbalance using Fidelity and not Fidelity LCL (see De Cáceres et al. 2012) in choosing candidates.

Final Indicators
We judged indicators for practical use based on their performance, validity, and robustness. Performance was measured by Fidelity, Constancy, and the traditional Fidelity × Constancy indicator value (IndVal; Dufrêne and Legendre 1997), which is useful for ranking overall performance and comparing indicators. True positive validity was determined by observed rates in high-quality samples equaling or exceeding predicted rates (Observed ≥ Fidelity LCL) and expected rates (Observed ≥ Constancy). False positive validity was determined by observed rates in lower-quality samples at or below nominal standards (Observed < 0.2, Observed < 0.1) and conservative prediction (Observed ≤ 1 -Fidelity UCL). For each indicator we tallied all instances of meeting these criteria across samples to rank overall validity (V-score) within each context. Robustness was defined by number of samples and contexts in which the indicator appeared.

Indicator Performance
A total of nine species met performance thresholds (0.6 Fidelity, 0.25 Constancy), including six candidate indicators for high floristic quality prairie depressions (Table 1) and 10 for CGP depressions (Table 2). Maximum overall performance (IndVal) improved in CGP depressions for each species appearing in both contexts (Tables 1 and 2). The greatest exclusivities (Fidelity) to high floristic quality were achieved by Ammannia coccinea, Cyperus setigerus, Leersia oryzoides, and the combination Eleocharis compressa + Schoenoplectus pungens, all in the CGP depressions context, with lower confidence limits usually exceeding 0.5 and upper limits all reaching 1 (Supplement Table S2). In both contexts, E. compressa clearly showed the greatest ubiquity (Constancy) and overall performance (Tables 1 and 2, S1, and S2).

True Positive Validity
For prairie depressions, occurrence rates in high-quality validation sites (observed true positives) equaled or exceeded minimum positive predictive value (Fidelity LCL) in Eleocharis compressa and Juncus torreyi (Fig. 3A), with E. compressa doing so in multiple samples (Table S1). Three indicators (E. compressa, J. torreyi, Leersia oryzoides) for prairie depressions satisfied Observed ≥ Constancy, with E. compressa achieving this in multiple samples (Table S1). For CGP depressions, true positives equaled or exceeded predictions in E. compressa and Ammannia coccinea (Fig. 4A), with evidence across all five samples for E. compressa and in one of three samples for A. coccinea (Table S2). Three indicators (A. coccinea, E. compressa, L. oryzoides) for CGP Table 1 Candidate indicators for high floristic quality (≥ 20.0 or ≥ 13.0 index score) depressional wetlands across Oklahoma prairie ecoregions 'C value' -Oklahoma ecological conservatism value; 'Wetland status' -Great Plains obligate (OBL) or facultative wetland (FACW) status; '# samples' -number of samples (5 max, see Supplement Table S1 for sample-level results) meeting 0.6 Fidelity and 0.25 Constancy; 'LCL and UCL' -lower and upper 95 % confidence limits from 10,000 bootstrap iterations; 'IndVal' -Fidelity × Constancy (max IndVal shown for indicators appearing in multiple samples); 'V-score' -tally of achieving true-positive and false-positive validity across samples (  (Table S2).

False Positive Validity
For prairie depressions, occurrence rates in lower-quality validation sites (observed false positives) were below conservative predictions (1 -Fidelity UCL) in Eleocharis compressa and Juncus torreyi (Fig. 3B). Observed falsepositive rates were mostly at or above 0.2 with only J. torreyi detected in < 10 % of lower-quality sites (Table S1). For CGP depressions, false positives were at or below conservative predictions in six indicators (Fig. 4B), each in one sample except for J. torreyi in two samples (Table S2).
Observed false-positive rates for CGP depressions improved compared to prairie depressions, with 70 % of results below 0.2 and five indicators below 0.1 including J. torreyi and Leersia oryzoides in multiple samples (Table S2).

Final Indicators
The most robust indicator of high floristic quality was Eleocharis compressa, the only candidate selected across all five samples in both contexts (Tables S1 and S2). This relatively  Table S1). Species are coded by the first two letters of their genus and specific name given in Table 1 Fig . 4 Indicator validity for Oklahoma northern Central Great Plains depressions, comparing observed frequencies in high-quality sites to minimum positive predictive values (A) and in lower-quality sites to conservative false-positive predictions (B). The best true-positive and false-positive validities are shown for indicators appearing in multiple samples (from Table S2). Species are coded by the first two letters of their genus and specific name given in Table 2 1 3 conservative (C of 6) hydrophyte (FACW) achieved generally strong and consistent overall performance and truepositive validity, but it also revealed limited false-positive validity tied to weaker fidelity. Fidelity of E. compressa improved when paired with Ammannia coccinea or Schoenoplectus pungens, but at a cost to performance, robustness, and true-positive validity (Table 2). Other noteworthy indicators were Ammannia coccinea (OBL), Juncus torreyi (FACW), and Leersia oryzoides (OBL). All three were less prone than E. compressa to commit false positive errors (Tables S1 and S2), but they each contained weaknesses such as lower C value (L. oryzoides) or applicability to CGP depressions only (A. coccinea). Two other species, Cyperus setigerus and Schoenoplectus pungens, showed signs of indicator potential but they lacked robustness and C. setigerus is ambiguously hydrophytic (FAC status).

Discussion
By short-cutting full community Floristic Quality Assessment, indicator species (Dufrêne and Legendre 1997;Siddig et al. 2016) could drastically accelerate field identification of high floristic quality and help to rapidly screen or verify potential reference-quality wetlands, including verification of predictions made remotely (e.g., Host et al. 2005). The present results suggest potential for indicator species but do not unequivocally support using them to discriminate high floristic quality depressional wetlands in the US southern plains. In the same study area Bried et al. (2014) reported plant indicator species of potential reference-quality wetlands defined by multimetric vegetation criteria, including the floristic quality benchmark (≥ 20.0 FQI) used here. They reported encouraging indicator values and false positive predictions but lacked a comparison with observed misclassification rates.
Several challenges may limit the performance, validity, and robustness of floristic quality indicator species. First and foremost, the species are being asked to indicate a level of quality derived from community data. We assumed species combinations (De Cáceres et al. 2012) would mitigate this disproportionality but only two species pairs and no triplet combinations met performance thresholds, and both pairs were outdone by their constituent species alone. Secondly, indicator species are always fixed to a target, but floristic quality is a "moving target" because theoretically many species compositions can result in the same level of quality. It seems indicator species would need strong associations or positive co-occurrence patterns with a large fraction of the assemblage to adequately represent floristic quality. Increased ecogeographic and hydrogeomorphic stratification may reduce compositional variation and strengthen indicators (Brooks et al. 2006;Bried et al. 2019), but effective stratification may be difficult. In our study area wetlands are hard to geomorphically subclassify (Dvorett et al. 2012) and subject to strong climatic gradients and other natural heterogeneity (Hoagland 2000;Gallaway et al. 2019), adding to the challenges.
One candidate, Eleocharis compressa, appeared in every scenario and consistently showed among the highest indicator values and constancy rates. This species also possesses a high C value (6) relative to the low-biased C distribution across the dataset (Fig. 2). In western Oklahoma E. compressa can dominate in small shallow depressions on clay soils (Hoagland 2002), suggesting along with our constancy estimates that it may occur commonly enough for practice. However, E. compressa may be too common, potentially occupying a variety of wetland conditions and incurring an unacceptable risk of indicating a degraded or marginal site as high quality.
Several candidates were less prone than E. compressa to false positive errors, potentially compensating for the weakness. Indeed, pairings of E. compressa with Ammannia coccinea and Schoenoplectus pungens lowered the observed false-positive rate in their respective samples (Table S2). Combining other single-species indicators post hoc could help mutually offset their deficiencies. For example, our results suggest that pairing E. compressa with Juncus torreyi, both relatively conservative, may reduce misclassification error while increasing robustness. Likewise, joining E. compressa and Leersia oryzoides may strengthen indicator fidelity and validity compared to using either species alone. These added species also have the advantage of being definitively recognizable throughout the growing season, unlike E. compressa which flowers early to mid-season and can look similar vegetatively to other Eleocharis species in the region (E. albida, E. montevidensis, E. tenuis).
The question is whether such combinations will sufficiently occur in the target area. This problem is mitigated ad hoc by preset constancy thresholds in the analysis of combinations (De Cáceres et al. 2012 Table S2). Regarding the singlespecies indicators, Eleocharis compressa and J. torreyi co-occurred in only 8.8 % of our 91 study sites, consistent with Hoagland (2002) who did not detect J. torreyi where E. compressa was most abundant. Similarly, E. compressa co-occurred with L. oryzoides in only 5.5 % of sites, and there were only three sites where E. compressa, J. torreyi, and L. oryzoides all co-occurred. These low rates suggest the combinations may create problems in practice, or possibly signal a regional scarcity of high-quality conditions especially considering the low to moderate modal C range in both high and lower quality samples (Fig. 2).
More sample stratification may alleviate misclassifications while strengthening indicator values (Dufrêne and Legendre 1997;Bried et al. 2019). Indicator performance, validity, and robustness improved overall after filtering the samples to a specific prairie ecoregion. Deeper stratification by vegetation types and hydrogeomorphic subclasses (e.g. Dvorett et al. 2012) could further strengthen indicators but requires adequate knowledge and accepting tradeoffs between accuracy and precision. Too much stratification may leave some strata poorly defined and lacking in indicator species or the data needed to extract and validate them. Alternatives to traditional indicator value analysis (Dufrêne and Legendre 1997) might lead to improvements. Stapanian et al. (2013), for example, used classification and regression tree models to find indicator species of wetland vegetation quality in Ohio, USA. Their simplest model containing just two species predicted high-quality wetlands with 13 % overall misclassification rate. A very different predictive approach might exploit the connections between C values and functional traits (Ficken and Rooney 2020) to extract trait-based indicator species of high floristic quality.
The present findings do not preclude the existence of high floristic quality indicators in other regions, especially where wetlands are well classified, training samples are given sufficient spatial-environmental context, and datasets are sufficiently large for validation. Our study also cannot rule out potential for indicators at other levels of floristic quality. Indicators of low quality could be useful in avoiding a costly permitting process (Stapanian et al. 2013) or futile conservation investment, i.e. a site not worth protecting or restoring (a site "beyond repair"). Indicators of moderate quality could help direct protection and management effort to where there is both need and worth. Before broadly concluding whether floristic quality indicator species may exist for wetlands, we recommend exploring indicator potential on larger datasets in other regions and at other quality levels and perhaps trying other statistical approaches (e.g. Stapanian et al. 2013;Sólymos and Azeria 2018).
Author Contributions JTB and SKJ conceived the study idea. JTB designed the study. All authors collected the data. TSF and JTB performed the analysis. JTB wrote the manuscript with reading, input, and approval from TSF and SKJ.
Funding Please see the Acknowledgments section.
Data Availability Available from the corresponding author upon request.
Code Availability Available from the corresponding author upon request.

Declarations
To the best of our knowledge this study was conducted in compliance with all relevant ethical and legal standards, and all necessary permissions (to access field sites, to collect and use data) were obtained.

Conflicts of Interest/Competing interests Not applicable.
Ethics Approval Not applicable.

Consent to Participate
Not applicable.