Development and Validation of Indices of Biological Integrity (IBI) for Depressional Wetlands in the Temperate Prairies Ecoregion

By John A. Genet and Michael Bourdaghs

April 2006

Biological Monitoring Unit Environmental Analysis and Outcomes Division Minnesota Pollution Control Agency 520 Lafayette Road St. Paul, Minnesota (651) 296-6300 www.pca.state.mn.us

Wetland Program Development Grant Section 104(b)3 CWA Part of a Final Report to US EPA Federal Assistance #CD-975768-01

wq-bwm6-01

Table of Contents

List of Tables ...... iv List of Figures...... vi I. Wetland Biological Assessment in a Geographic Context...... 1 INTRODUCTION ...... 1 METHODS ...... 2 Field Sampling...... 2 Disturbance Gradient ...... 3 Macroinvertebrate IBI Evaluation ...... 3 Plant IBI Evaluation...... 4 RESULTS & DISCUSSION – Evaluation of Macroinvertebrate IBI...... 4 Evaluation of NCHF Criteria in Plains Ecoregions ...... 4 Re-examination of NCHF Macroinvertebrate IBI ...... 9 Adjustment of NCHF Macroinvertebrate IBI...... 12 Evaluation of Geographical Classification Frameworks ...... 12 Development of Preliminary Temperate Prairies Macroinvertebrate IBI...... 20 Macroinvertebrate Metrics ...... 25 Macroinvertebrate IBI ...... 29 RESULTS & DISCUSSION – Evaluation of Plant IBI ...... 30 Evaluation of NCHF Criteria in Plains Ecoregions ...... 30 Evaluation of Geographical Classification Frameworks ...... 34 Development of Preliminary Temperate Prairies Plant IBI...... 41 Methods...... 41 Results and Discussion ...... 41 Preliminary TP IBI Metric Descriptions ...... 44 Preliminary Plant Temperate Prairies IBI ...... 48 II. Validation of Preliminary Temperate Prairies Wetland Macroinvertebrate and Plant IBIs and Testing their Applicability in Seasonal Wetlands ...... 53 INTRODUCTION ...... 53 METHODS ...... 54 Field Sampling...... 54 Macroinvertebrate IBI Validation...... 55 Testing the Applicability of Macroinvertebrate IBI in Seasonal Wetlands...... 56 Plant IBI Validation ...... 57 RESULTS & DISCUSSION ...... 58 Macroinvertebrate IBI Validation...... 58 Characteristics of Final Temperate Prairies Macroinvertebrate IBI...... 62 Applicability of Macroinvertebrate IBI in Seasonal Wetlands...... 66 Preliminary Macroinvertebrate Indicators of Condition in Seasonal Wetlands ...... 69 Plant IBI Validation ...... 71 Characteristics of the Final Temperate Prairies Plant IBI ...... 77 Preliminary Impairment Threshold ...... 79 IBI and Metric Precision ...... 81 Plant Sampling Methods Evaluation...... 82

Acknowledgements ...... 84 Literature Cited ...... 84 Appendices...... 92 Appendix A - List of wetland sites sampled in 2002. See Table I-4 for definitions of ecoregion, section, and basin abbreviations...... 93 Appendix B - Macroinvertebrate Sampling Protocols for Depressional Wetlands...... 94 Appendix C - Plant Sampling Protocols for Depressional Wetlands...... 103 Appendix D - Box plots of macroinvertebrate metrics for the 4 least-impacted sites (=reference) vs the 4 most-impacted sites (=impaired), according to ranking by the Human Disturbance Score (HDS), for the 2002 NCHF data set (excluding bogs)...... 115 Appendix E - Tolerant/Intolerant Macroinvertebrate Taxa Designations for Minnesota Depressional Wetlands...... 117 Appendix F - Box plots of macroinvertebrate metrics for the 4 least-impacted sites (=reference) vs the 4 most-impacted sites (=impaired), according to ranking by the Human Disturbance Score (HDS), for the 2002 Temperate Prairies data set...... 124 Appendix G - Macroinvertebrate IBI scores for semi-permanent and permanent depressional wetlands sampled in 2002 and 2003 from the Temperate Prairies ecoregion. Bold text indicates reference sites...... 126 Appendix H - Box plots of potential macroinvertebrate metrics for the 4 least-impacted sites (=reference) vs the 4 most-impacted sites (=disturbed), according to ranking by the HDS, for the 2003 Temperate Prairies seasonal wetlands...... 128

iii

List of Tables

Table I-1. The preliminary NCHF Macroinvertebrate IBI scores for wetlands sampled in 2002...... 6 Table I-2. Pearson correlation coefficients (r) for relationship between IBI scores/metric values and measures of human disturbance. Bold text indicates significant (P < 0.05) correlation coefficients. Chemistry data was Log10 transformed...... 8 Table I-3. Pearson correlation coefficients (r) for relationship between IBI scores/metric values and measures of human disturbance for the 2002 NCHF data set excluding the two wetlands classified as bogs. Bold text indicates significant (P < 0.05) correlation coefficients. Chemistry data were Log10 transformed...... 11 Table I-4. Geographic classification frameworks tested for Minnesota depressional wetlands ...... 14 Table I-5. Strength of geographic classification schemes for Minnesota depressional wetlands based on macroinvertebrate assemblage data. Classification strength (CS) = [B – W]. P values represent the proportion of 10,000 permutations with random assignment of sites into classes having a CS at least as large as the observed CS value for the tested classification...... 17 Table I-6. Component metrics of a preliminary macroinvertebrate IBI for the Temperate Prairies depressional wetlands, indicating which portions of the sampling method are used to derive each, whether it was included in the NCHF IBI, and its observed response to human disturbance...... 21 Table I-7. Pearson correlation coefficients (r) between macroinvertebrate metrics/IBI and measures of human disturbance for Temperate Prairies depressional wetlands. Water and sediment chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10)...... 23 Table I-8. Pearson correlation coefficients (below diagonal) and corresponding p-values (above diagonal) for pairwise relationships among the ten metrics in the Temperate Prairies depressional wetland macroinvertebrate IBI...... 25 Table I-9. Preliminary IBI scores for Temperate Prairies depressional wetlands sampled in 2002...... 29 Table I-10. IBI and metric Pearson correlation coefficients (r) with HDS and selected water and sediment chemistry parameters. All of the chemistry data were Log10 transformed prior to analysis. Results are displayed as: ns = not significant, * = (P < 0.1), and ** = (P < 0.05)...... 33 Table I-11. Strength of geographic classification schemes based on plant assemblages for 5 data sets. Classification strength (CS) = [B – W]. P-values represent the proportion of 10,000 permutations with random assignment of sites into classes having a CS at least as large as the observed CS value for the tested classification...... 36 Table I-12. TP ecoregion plant metric selection criteria...... 42 Table I-13. Preliminary plant TP IBI metrics. Metrics are arranged according to four metric categories. A brief description, inclusion into the NCHF IBI indicator, and the observed response to the anthropogenic disturbance are given...... 43 Table I-14. Pearson correlation coefficients (below diagonal) and corresponding P-values (above diagonal) for all pairwise relationships between preliminary plant TP IBI metrics. 44 Table I-15. Preliminary plant TP IBI scores for the development set...... 49

iv

Table I-16. Preliminary TP IBI and metric Pearson correlation coefficients (r) with selected water and sediment chemistry parameters. All of the chemistry data were Log10 transformed prior to the analysis. Results are displayed as: ns = not significant, * = (P < 0.1), and ** = (P < 0.05)...... 50 Table II-1. Temperate Prairies depressional wetlands sampled in 2003, indicating which sites had also been sampled in 2002 and which were part of the Redwood study...... 59 Table II-2. Pearson correlation coefficients (r) between macroinvertebrate metrics from preliminary TP IBI and measures of human disturbance for 2003 study sites. Water and sediment chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10)...... 61 Table II-3. Pearson correlation coefficients (r) for the relationship between selected metrics and human disturbance measures. Water chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10)...... 62 Table II-4. Component metrics of a macroinvertebrate IBI for the TP depressional wetlands, indicating which portions of the sampling method are used to derive each, whether it was included in the NCHF IBI, and its observed response to human disturbance...... 63 Table II-5. Pearson correlation coefficients (r) and corresponding P-values for the relationship between IBI scores and water chemistry parameters (Log10) in 2002 and 2003...... 64 Table II-6. Pearson correlation coefficients (r) for the relationship between metrics and human disturbance measures. Water chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10)...... 68 Table II-7. Water chemistry characteristics of reference seasonal (C) and semi-permanent (F&G) TP wetlands...... 70 Table II-8. Pearson correlation coefficients (r) for the relationship between metrics and human disturbance measures. Water chemistry data were Log10 transformed. Bold values = significant correlations (P < 0.05); * indicates marginally significant correlations (P < 0.10)...... 70 Table II-9. Final Temperate Prairies plant IBI metric descriptions...... 77 Table II-10. Final plant IBI scores for both the development and validation data sets. Site names appearing in boldface are reference sites...... 78 Table II-11. Final TP IBI and metric Pearson correlation coefficients (r) with selected water and sediment chemistry parameters from validation data. All of the chemistry data were Log10 transformed prior to the analysis. Results are displayed as: ns = not significant, * = (P < 0.1), and ** = (P < 0.05)...... 80 Table II-12. Confidence limits, confidence intervals, and number of condition .categories the TP plant IBI can detect according to the number of replicate samples (N)...... 82 Table II-13. Plant IBI signal:noise and 90% confidence intervals produced from four different sampling scenarios. Bold text indicates the number of standard sample replicates that would be adopted to sample a wetland for a given scenario...... 83

v

List of Figures

Figure I-1. Omernik Level III ecoregions in Minnesota...... 1 Figure I-2. Formulas for determining continuous metric scores...... 4 Figure I-3. Relationships between macroinvertebrate IBI and human disturbance (HDS) for all year/ecoregion combinations...... 7 Figure I-4. Comparison of macroinvertebrate IBI scores at wetlands sampled in 1999 and 2002...... 12 Figure I-5. Relationship between macroinvertebrate IBI (after revisions to Tolerant/Intolerant metrics) and HDS for the 1999 and 2002 NCHF data sets...... 13 Figure I-6. Distribution of 2002 wetland study sites among the various classification frameworks...... 16 Figure I-7. Mean similarity dendrogram based on data from all sites (N = 52) sampled in 2002...... 18 Figure I-8. Comparison of Multi-Dimensional Scaling ordinations illustrating the better grouping of three sites (Carex2, OakGlenEast, OakGlenWest) by the ECS classification framework...... 19 Figure I-9. Relative variance estimates for the Temperate Prairies macroinvertebrate IBI and its component metrics, comparing the within site variance (Noise) to the between site variance (Signal). Signal:noise ratios are presented above the bars for each metric...... 26 Figure I-10. Relationship between macroinvertebrate IBI and human disturbance score (HDS) for Temperate Prairie wetlands sampled in 2002...... 30 Figure I-11. Macroinvertebrate IBI scores plotted against wetland size (Log10 transformed) for Temperate Prairie wetlands sampled in 2002...... 30 Figure I-12. Plant NCHF IBI-HDS relationships for (A) 1999 NCHF IBI development data, (B) 2002 NCHF data, (C) 2002-03 WCBP data, (D) 2002 NGP data, and (E) 2002-03 WCBP & NGP (plains ecoregions) data. The y-intercept, (β0), slope (β1), and coefficient of variation (r2) are included, * = P < 0.1 and ** = P < 0.05...... 31 Figure I-13. Plant NCHF IBI scores from sites that were sampled in both 1999 and 2002...... 34 Figure I-14. Mean dissimilarity dendrograms derived from plant data collected in 1999-2003 in least impacted sites (HDS < 50) and sites that had low abundance of invasive species, for geographic classification schemes that had significant class structure (P < 0.05) according to mean dissimilarity analysis...... 37 Figure I-15. Nonmetric Multi-Dimensional Scaling ordinations of sites that had low abundance of invasive species based on plant assemblages, from data collected in 1999-2003, for (A) Level II and III Ecoregions and (B) ECS Sections. Labeled sites differ in general classification (i.e., prairie vs. forest) between Omernik and ECS classification systems. ... 39 Figure I-16. Preliminary plant TP IBI metric-HDS scatterplots (A-I). The slope (β1) and coefficient of variation (r2) are included, * = P < 0.1 and ** = P < 0.05...... 46 Figure I-17. Preliminary plant TP IBI-HDS scatterplot with y-intercept (β0), slope (β1), and coefficient of variation (r2). ** = P < 0.05...... 49 Figure I-18. Scatterplots of the preliminary plant TP IBI (A) and HDS (B) with wetland area (ha). Wetland area has been square root transformed. * = P < 0.1, ** = P < 0.05...... 51

vi

Figure I-19. Relative between site (signal) and within site (noise) variance estimates for the preliminary Temperate Prairies Plant IBI and component metrics (n = 2). Signal to noise ratios are given above the bars for each metric and the IBI...... 52 Figure II-1. Distribution of human disturbance ratings (HDS) in A) the 2003 Redwood wetlands and in B) all of the 2003 TP wetlands...... 60 Figure II-2. Relationships between TP macroinvertebrate IBI and human disturbance (HDS) for the 2002 and 2003 data sets...... 63 Figure II-3. Comparison of the distribution of IBI scores and HDS among study sites when site selection was targeted in order to obtain a sample of wetlands with conditions spanning the range of human disturbance (2002) versus when site selection was random (2003)...... 65 Figure II-4. Relative variance estimates for the TP macroinvertebrate IBI and its component metrics, comparing the within site variance (Noise) to the between site variance (Signal). Signal:noise ratios are presented above the bars for each metric...... 65 Figure II-5. Nonmetric multidimensional scaling plots based on dip net macroinvertebrate data collected from TP depressional wetlands in 2003...... 67 Figure II-6. Nonmetric multidimensional scaling plot based on dip net macroinvertebrate data collected from TP depressional wetlands in 2003, including only sites with HDS less than 50...... 67 Figure II-7. Comparison of macroinvertebrate community at undisturbed seasonal and intermittently exposed prairie pothole wetlands (Lyon Co., MN) based on data collected from dip net samples...... 68 Figure II-8. Preliminary plant TP IBI-HDS scatterplot with validation data. The y-intercept (β0), slope (β1), and coefficient of variation (r2) are included. ** = P < 0.05...... 71 Figure II-9. Preliminary plant TP IBI metric-HDS scatterplots with validation data. The y- intercept slope (β1), and coefficient of variation (r2) are included. * = P < 0.1 ** = P < 0.05...... 72 Figure II-10. Aquatic Guild Richness-HDS scatterplot with sites classified by water regime (i.e., seasonal vs. semipermanent/permanent)...... 73 Figure II-11. Additional final plant IBI metric-HDS scatterplots derived from both the development (A) and validation dataset (B)...... 75 Figure II-12. Scatterplots of the final TP Plant IBI derived from both the development (A) and validation (B) datasets...... 80 Figure II-13. Relative between site (signal) and within site (noise) variance estimates for the final TP Plant IBI and component metrics (n = 5). Signal to noise ratios are given above the bars for each metric and the IBI...... 82 Figure II-14. Preliminary impairment threshold and application of 90% confidence intervals in depressional wetland ALUS assessment with the TP Plant IBI. 2002-03 data are shown from sites sampled once (N = 1)...... 83

vii

I. Wetland Biological Assessment in a Geographic Context

INTRODUCTION

Utilizing the appropriate framework for classifying wetlands is a necessary component of any monitoring and assessment program. A good classification framework minimizes the natural variability within each class of wetlands so that variability due to anthropogenic impacts can be better distinguished (US EPA 2002). Without such a framework in place, apparent differences in wetland condition based on biological or chemical assessments may reflect natural differences between wetland types (e.g., depressional vs. slope), ecoregions, or landscape position. The Minnesota Pollution Control Agency (MPCA) has adopted a classification scheme for wetlands based on four factors: 1) ecoregions (Level III, Omernik 1987); 2) hydrogeomorphic (HGM) wetland types (Brinson 1993); 3) structural plant communities (e.g., forested vs. emergent); and 4) hydrologic regime (e.g., temporary vs. permanent).

Previous projects in Minnesota (Gernes and Helgen 1999, 2002) focused on the development and validation of indices of biological integrity (IBIs) for semi-permanently to permanently flooded emergent depressional wetlands located in the North Central Hardwood Forest (NCHF) ecoregion (Figure I-1). Because of the predominance of emergent depressional wetlands across the landscape with respect to other wetland types in nearly all regions of the state these wetlands are singularly noteworthy. Thus the MPCA has expressed an interest in establishing biological criteria for these wetlands statewide before expanding to other wetland types. As a first attempt at expanding the geographic coverage of the IBI beyond the NCHF ecoregion, the MPCA proposed to test the applicability of the current depressional wetland biological criteria in the Western Corn Belt Plains (WCBP) and Northern Glaciated Plains (NGP) ecoregions (Figure I-1). In the context of our wetland classification framework, this evaluation represents a test of the suitability of Omernik ecoregions as an appropriate geographical framework. For instance, if NCHF biological criteria are applicable across ecoregions, then Omernik ecoregions may not represent the Figure I-1. Omernik Level III ecoregions in most appropriate geographic classification Minnesota. scheme for wetlands.

Development and Validation of Temperate Prairie Wetland IBIs 1

The previously developed (1999 NCHF) biological criteria will be evaluated in the WCBP and NGP ecoregions by calculating metric and IBI scores for sites within these ecoregions and analyzing their relationship with an independent measure of human disturbance, the Human Disturbance Score (HDS; see Gernes and Helgen 2002). Strong correlation between the IBI/metrics and HDS would indicate applicability of the NCHF wetland IBIs in the WCBP and NGP ecoregions. Poor relationships would indicate the need to develop different biological criteria for these ecoregions, substituting new metrics for any of the original metrics that do not appear to respond to human disturbance in these ecoregions.

METHODS

Field Sampling

Plant and macroinvertebrate data were collected from a total of 47 depressional wetlands during the summer of 2002; with 14 located in the WCBP ecoregion, 14 located in the NGP ecoregion, and 19 located in the NCHF ecoregion (Appendix A). Due to a data recording error, plant data from three sites were lost. These sites were resampled for plants in 2003. Sites were sampled in the NCHF ecoregion in order to evaluate any effects due to annual variability if the NCHF biological criteria failed to exhibit a relationship with human disturbance in the WCBP and NGP ecoregions. The assumption being that if the IBIs displayed strong relationships with disturbance in the NCHF ecoregion but not in the WCBP or NGP ecoregions, that these results would suggest that the IBIs failed because of ecoregional differences rather than yearly differences (i.e., the NCHF IBIs were developed using data collected in 1999). Site selection in the WCBP and NGP ecoregions was targeted, using the expertise of local resource managers to select wetlands that ranged from least-impacted to severely degraded.

Macroinvertebrates were sampled during the seasonal index period of June following standard MPCA sampling protocols (see Appendix B). During the macroinvertebrate sampling visit, water chemistry parameters were measured using dissolved oxygen (DO), pH, and conductivity probes. Surface water grab samples were also collected during each visit for laboratory analysis of the following water chemistry parameters: total Kjeldahl nitrogen (mg/L), total phosphorus (mg/L), total chloride (mg/L), calcium (CaCO3 mg/L), turbidity (NTU), and total sulfate (mg/L). Analysis of these water chemistry parameters was conducted by the Minnesota Department of Health, Environmental Laboratory.

The plant community of each wetland site was characterized during the seasonal index period of July following standard MPCA sampling protocols (see Appendix C). During the plant sampling visit, three sediment cores were collected from the emergent zone of each wetland and the top 5 cm of sediment were extruded and pooled for analysis of nitrogen, Olsen phosphorus, chloride, % moisture, pH, total organic carbon, and heavy metals by the University of Minnesota Soils Analytical Laboratory. In addition, supplementary site information, required for the calculation of the HDS, was noted during each plant visit. This included observations such as the presence of water control structures, man-made berms, vegetation removal, dredging, and vehicle use within the wetland.

Development and Validation of Temperate Prairie Wetland IBIs 2

Disturbance Gradient

Performance of the NCHF IBIs in the plains (NGP & WCBP) ecoregions was evaluated against a measure of anthropogenic disturbance developed by MPCA staff biologists, the HDS (Gernes and Helgen 2002). Determination of an HDS for an individual site requires the semi-quantitative rating of five factors: 1) Buffer landscape disturbance (within 50m of wetland edge); 2) Landscape disturbance (within 500m of wetland edge); 3) Habitat alteration (within wetland and in immediate upland landscape); 4) Hydrologic alteration (within or directly affecting wetland); and 5) Chemical pollution (within wetland). Factors 1 and 2 are rated using low altitude aerial photography, land use data in GIS, and field observations. Factors 3 and 4 are rated primarily based on field observations and low altitude aerial photography. Factor 5 is rated using the water and sediment data collected during the field visits. An additional factor score (up to 4 points) can be added for disturbances not included in any of the other factors (e.g., presence of fish, previous uses of the wetland, etc.). Combining the factor scores results in a total HDS that ranges from 0-100, with 0 representing reference or least-impacted conditions.

Macroinvertebrate IBI Evaluation

For all wetlands sampled in 2002 IBI scores were generated using the metrics selected in the NCHF ecoregion (Gernes and Helgen 2002). Individual metric scores were calculated using a continuous scoring method (Fore 2003b), which converts each metric into unit-less scores by dividing each metric value by its range and multiplying by 10. Adjustments to this general formula resulted in metric scores that range from 0-10 with 10 representing the best biological condition (Figure I-2). For skewed metrics, the natural log of the metric value was calculated before dividing by the range. The minimum, maximum, 95th, and 5th percentile values used in the formulas below were derived from the NCHF biological criteria development data set (1999 & 2001).

The performance of the macroinvertebrate IBI was evaluated by determining pearson correlation coefficients (r) for their relationship with the HDS. Correlation coefficients were determined separately for the NCHF, NGP, WCBP, and the Temperate Prairies (‘plains’= NGP & WCBP) ecoregion. Correlation coefficients determined to be significantly different from zero (P < 0.05) were then compared to the correlation coefficient from the original NCHF IBI development data set to test for significant differences (Zar 1999).

To further test the applicability of the previously developed IBIs in the plains ecoregions, the relationships between individual metrics and measures of disturbance (e.g., HDS, Cl, N, P, etc.) were evaluated for data sets where the IBI was significantly correlated with the HDS. Once again Pearson correlation coefficients were used to test for significant relationships. In the NCHF IBI development project, metrics were significantly correlated to a number of disturbance measures. If such relationships are not present in the plains ecoregions due to inherent differences in the biological community of these wetlands, this would suggest that new metrics need to be substituted for those that are not responding to human disturbance.

Development and Validation of Temperate Prairie Wetland IBIs 3

Metrics that Decrease with Increasing Disturbance:

Score = metric value - minimum value x 10 95th percentile value- minimum value

Metrics that Increase with Increasing Disturbance:

Score = 10 - metric value - 5th percentile value x 10 maximum value- 5th percentile value

Figure I-2. Formulas for determining continuous metric scores.

Plant IBI Evaluation

The plant based NCHF ecoregion IBI was calculated for all of the wetland sites sampled. A combination of continuous (Figure I-2) and discrete metric scoring methods was used to score the IBI. This was due to a number of metrics (5 out of 10) having a curve-linear response to human disturbance. Linear and curve-linear metrics cannot be added together when all metrics are scored continuously as the resulting IBI would have a relationship with disturbance somewhere between linear and curve-linear. Discrete, or categorical, scoring of curve-linear metrics from scatterplots (Gernes and Helgen 2002) dampens the curvature of these metrics and facilitates incorporation with linear metrics in an IBI.

Simple linear regression was used to evaluate the performance of the plant NCHF IBI against HDS in the following ecoregions: NCHF, WCBP, NGP, and the combined plains ecoregions (WCBP & NGP; Figure I-1). The individual metrics were also evaluated against HDS and several chemical measures of disturbance (Cl, N, P, etc.) for all data sets with Pearson correlation. All statistical analyses were performed with SYSTAT® Version 10.2. As with the macroinvertebrate analysis, strong relationships between the IBI/metrics and the various measures of human disturbance would indicate that the NCHF IBI is applicable in the given ecoregions.

RESULTS & DISCUSSION – Evaluation of Macroinvertebrate IBI

Evaluation of NCHF Criteria in Plains Ecoregions

Macroinvertebrate IBI scores were calculated for all the wetlands sampled in 2002 using the continuous scoring criteria (max, min, 95th percentile, 5th percentile) from the 1999/2001 NCHF data set (Table I-1). The sensitivity of the IBI to human disturbance was evaluated separately for each ecoregion sampled in 2002 and the combination of the plains ecoregions (Figure I-3). The

Development and Validation of Temperate Prairie Wetland IBIs 4

IBI was not significantly correlated to human disturbance in the WCBP and NGP ecoregions when analyzed separately (P > 0.05), probably resulting from insufficient sample size in each (N =14). However, when data was pooled for these two ecoregions the relationship between the IBI and disturbance was significant (Figure I-3). A multiple comparison test among the significant 2002 IBI/disturbance relationships and the 1999 NCHF IBI/disturbance relationship yielded no significant differences among any of the pairwise comparisons of correlations coefficients (χ2 = 2.1, v = 2, P > 0.05). These results preliminarily suggest that the NCHF macroinvertebrate IBI may have applicability in the WCBP and NGP ecoregions.

To confirm these preliminary results individual metrics and the IBI were evaluated, testing their sensitivity to various chemical stressors and the HDS. The numerous significant correlations between the IBI/metrics and measures of disturbance present in the 1999 NCHF development data set, were less evident in the 2002 NCHF and WCBP/NGP data sets (Table I-2). In fact, a total of five metrics (% Dom 3 Genera, % Tolerant, Snail Taxa, Leech Taxa, Chiro Taxa) did not exhibit any significant (P > 0.05) correlations with chemical stressors or the HDS in the WCBP/NGP (=‘plains’) data set. The only metric that had more significant correlations in the plains ecoregion was the Corixidae Ratio metric (Table I-2). These results do not corroborate the preliminary finding that the NCHF IBI appears to accurately reflect human disturbance in depressional wetlands of the WCBP and NGP ecoregions. Instead, it appears that some of the metrics have potential in the plains ecoregions and these metrics alone may be accounting for the overall significant correlation between the IBI and HDS (Figure I-3), while inclusion of the other metrics may be diminishing the strength of this relationship.

The 2002 NCHF data set was included in this analysis as a temporal control so that if the 1999 NCHF IBI did not perform well in the plains ecoregions in 2002, the successful performance of the IBI in the NCHF in 2002 would demonstrate that the observed results were not attributed to annual variability. However, given the results from the 2002 NCHF correlations, annual variability can not be discounted as a contributing factor to the poor performance of the NCHF IBI in the plains ecoregions. Even excluding the outlier (Hardscrabble) did not produce results comparable to the 1999 NCHF data set for the number of significant correlations between IBI/metrics and various disturbance measures (Table I-2). In fact, the % Tolerant Taxa metric exhibited a response opposite of that which was developed originally in the 1999 NCHF IBI. So not only do these results not allow the elimination of annual variability as a factor contributing to the poor performance of the IBI in the plains ecoregions, they in fact warrant further examination as to why the IBI did not perform satisfactorily in the NCHF in 2002.

Development and Validation of Temperate Prairie Wetland IBIs 5

Table I-1. The preliminary NCHF Macroinvertebrate IBI scores for wetlands sampled in 2002.

Site Name Rep IBI HDS Site Name Rep IBI HDS

North Central Hardwood Forest: BetShalom 1 31.6 59.0 Lee 1 56.4 54.5 Breen 1 37.9 74.0 LoneTreeWMA 1 21.4 77.0 BushLake 1 50.9 34.0 Malta 1 34.4 58.0 DellRd 1 72.9 48.0 Milan 1 68.4 64.0 Glacial 1 79.5 13.0 New Prairie 1 74.0 66.5 Gleason 1 57.0 61.0 Prairie Marsh 1 67.0 10.0 HardScrab 1 41.7 13.0 Prairie Marsh 2 55.4 10.0 Kipling 1 41.9 57.5 RostWMA 1 52.7 55.5 Lake 21 1 68.4 19.0 TylerWMA 1 34.3 58.5 Legion 1 39.2 81.0 Malardi 1 45.4 83.5 Western Cornbelt Plains: Morraine 1 42.0 45.0 Bryclyn 1 56.7 56.0 Ney 1 50.1 62.5 Carex2 1 52.8 15.5 Prairie 1 68.2 10.0 EastlickMarsh 1 45.3 40.0 RenoRef 1 60.9 35.0 FrancoWMA 1 33.0 79.0 TheoWirth 1 48.4 42.5 FrancoWMA 2 31.6 79.0 Turtle 1 60.8 73.5 GreatOasisWMA 1 78.6 24.0 Westmark 1 67.0 60.0 LakeCharlotte 1 58.8 45.0 Wood 1 52.6 75.0 LakeCharlotte 2 48.2 45.0 LakeElisabeth 1 43.5 21.0 Northern Glaciated Plains: LyonsWMA 1 40.2 66.5 BarryWMA 1 66.1 52.0 Manchester 1 45.5 32.5 FurgameWMA 1 57.3 51.5 OakGlenEast 1 66.5 28.5 GoldenWPA 1 38.6 53.5 OakGlenWest 1 80.1 20.0 Hancock 1 55.2 49.0 OakGlenWest 2 80.3 20.0 Hoffman 1 59.4 59.5 RolhiksWMA 1 62.9 61.0 Hoffman 2 70.9 59.5 WillowLake 1 50.8 65.0 Kerk 1 70.8 16.0 Yohi 1 63.9 37.5

Development and Validation of Temperate Prairie Wetland IBIs 6

NCHF 1999 Development Data Set NCHF 2002 100 100

80 80

60 60 Invert IBI Invert 40 IBI Invert 40 outlier

20 20 w/ out w/o out r = -0.684 r = -0.465 -0.602 p = <0.001 p = 0.045 0.008 0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS WCBP 2002 NGP 2002 100 100

80 80

60 60 Invert IBI Invert Invert IBI Invert 40 40

20 20 r = -0.519 r = -0.307 p = 0.057 p = 0.286 0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS WCBP/NGP 2002 100 Figure I-3. Relationships between macroinvertebrate IBI and human 80 disturbance (HDS) for all year/ecoregion combinations. 60

Invert IBI Invert 40

20 r = -0.478 p = 0.012 0 0 20 40 60 80 100 HDS

Development and Validation of Temperate Prairie Wetland IBIs 7

Table I-2. Pearson correlation coefficients (r) for relationship between IBI scores/metric values and measures of human disturbance. Bold text indicates significant (P < 0.05) correlation coefficients. Chemistry data was Log10 transformed. Data Source Water Chemistry Sediment Chemistry # sig. Ecoregion Year HDS Cl N P Turbidity Cu Ni Pb Zn correlations n

Invert IBI NCHF 1999 -0.684 -0.543 -0.473 -0.562 -0.575 -0.378 -0.306 -0.178 -0.265 7 44 NCHF 2002 -0.602 -0.216 -0.294 -0.491 -0.302 -0.354 0.060 -0.476 -0.421 2 18** Plains 2002 -0.478 -0.220 -0.414 -0.033 -0.020 -0.224 -0.253 0.153 -0.041 2 27 %Dom 3 Genera NCHF 1999 0.392 0.358 0.399 0.485 0.260 0.153 0.216 0.005 0.105 4 44 NCHF 2002 0.239 -0.083 -0.026 -0.161 -0.200 -0.311 -0.377 -0.208 -0.225 0 18** Plains 2002 -0.090 -0.110 -0.279 -0.257 -0.151 -0.341 -0.288 -0.133 -0.143 0 27 %Tolerant Taxa NCHF 1999 0.237 0.161 0.498 0.339 0.412 -0.071 0.015 -0.124 -0.095 3 44 NCHF 2002 -0.081 -0.210 0.035 -0.562 -0.513 -0.399 -0.611 -0.112 -0.297 3* 18** Plains 2002 0.284 0.032 -0.028 -0.204 -0.036 -0.139 -0.040 -0.345 -0.113 0 27 Corixidae Ratio NCHF 1999 0.278 0.171 0.260 0.095 0.288 0.236 0.023 -0.101 0.045 0 44 NCHF 2002 0.408 0.565 -0.187 0.007 0.127 0.100 -0.090 0.342 0.168 1 18** Plains 2002 0.289 0.073 0.334 0.178 -0.133 0.484 0.479 0.090 0.283 2 27 Odonata Taxa NCHF 1999 -0.461 -0.349 -0.299 -0.406 -0.530 -0.276 -0.216 -0.277 -0.278 5 44 NCHF 2002 -0.320 -0.033 0.103 -0.392 -0.071 -0.520 -0.295 -0.516 -0.561 3 18** Plains 2002 -0.423 -0.317 -0.376 -0.307 -0.047 -0.056 -0.052 -0.039 0.003 1 27 Leech Taxa NCHF 1999 -0.319 -0.207 0.061 0.020 -0.173 -0.023 0.086 -0.086 -0.042 1 44 NCHF 2002 -0.131 -0.142 -0.328 -0.012 -0.198 -0.157 0.113 -0.579 -0.226 1 18** Plains 2002 -0.151 -0.091 -0.302 -0.083 -0.304 -0.346 -0.374 0.016 -0.249 0 27 Snail Taxa NCHF 1999 -0.422 -0.368 -0.188 -0.339 -0.437 -0.386 -0.247 -0.307 -0.303 7 44 NCHF 2002 -0.231 0.011 -0.350 -0.377 -0.392 -0.127 -0.055 -0.120 -0.121 0 18** Plains 2002 -0.102 -0.133 -0.182 -0.122 0.098 0.064 0.084 0.211 -0.077 0 27 Total Taxa NCHF 1999 -0.585 -0.514 -0.217 -0.401 -0.513 -0.397 -0.261 -0.340 -0.343 7 44 NCHF 2002 -0.670 -0.399 -0.368 -0.511 -0.439 -0.506 -0.073 -0.582 -0.497 5 18** Plains 2002 -0.382 -0.215 -0.224 0.300 -0.169 -0.051 -0.060 0.196 0.173 1 27 Intolerant Taxa NCHF 1999 -0.665 -0.492 -0.495 -0.544 -0.403 -0.377 -0.259 -0.139 -0.204 6 44 NCHF 2002 -0.577 -0.110 -0.288 -0.661 -0.326 -0.450 -0.179 -0.346 -0.421 2 18** Plains 2002 -0.388 -0.227 -0.529 -0.099 0.001 -0.261 -0.292 0.102 0.001 2 27 ETSD NCHF 1999 -0.499 -0.342 -0.446 -0.467 -0.381 -0.306 -0.294 -0.021 -0.136 6 44 NCHF 2002 -0.333 -0.207 -0.220 -0.702 -0.452 -0.677 -0.494 -0.445 -0.639 4 18** Plains 2002 -0.402 -0.040 -0.455 -0.231 -0.117 -0.518 -0.493 -0.005 -0.183 4 27 Chironomid Taxa NCHF 1999 -0.627 -0.589 -0.274 -0.588 -0.320 -0.325 -0.436 -0.284 -0.346 8 44 NCHF 2002 -0.553 -0.190 -0.160 -0.632 -0.270 -0.333 -0.034 -0.173 -0.338 2 18** Plains 2002 -0.289 -0.113 -0.111 0.040 0.300 -0.097 -0.127 0.132 0.103 0 27 * Correlations are opposite direction than that of the original NCHF development data set. ** Sample size for water chemistry correlations is 16. All correlations do not include data from outlier wetland (=Hardscrabble).

Development and Validation of Temperate Prairie Wetland IBIs 8

Re-examination of NCHF Macroinvertebrate IBI

The macroinvertebrate IBI for the NCHF ecoregion was originally developed using data collected in 1995 from 27 depressional wetlands (Gernes and Helgen 1999). In 1999, this IBI was tested by sampling 44 large depressional wetlands in the NCHF ecoregion in order to determine if the IBI worked for an independent data set (Gernes and Helgen 2002). However, during this validation of the original IBI a number of metrics did not perform as well as some alternative metrics that were eventually used as substitutes in the IBI. Therefore, rather than the entire IBI being validated by the 1999 data set, only eight of the ten metrics were in fact validated, requiring only slight modifications of the old discrete (1,3,5) scoring criteria to accommodate the increased taxa richness found in these large depressional wetlands. The % Erpobdella metric did not exhibit a response to disturbance in the large depressional wetlands and was subsequently replaced with the % Tolerant Taxa metric. The Total Taxa richness metric from the original NCHF IBI was supplemented with the taxa counts of two additional insect orders (Coleoptera & Hemiptera) in the large depressional IBI. Therefore, the large depressional study produced an IBI with eight validated metrics and two in need of further validation.

Overall, the IBI produced from the large depressional wetland study was significantly correlated to HDS for the 2002 NCHF data set (Figure I-3). However, the diminished number of significant correlations between the IBI/metrics and the chemical parameters indicate that the metrics are not responding to disturbance as well as they did in the large depressional study. There are a few possible explanations for these observations: 1) two wetlands included in this data set, Hardscrabble and Bush Lake, are a wetland type different than which this IBI was intended to be used for; 2) the inclusion of an un-validated metric, % Tolerant Taxa, working antagonistically with the IBI; 3) climatic variability (e.g., annual precipitation, temperature patterns) between 1999 and 2002. Each of these possibilities was scrutinized in order to elucidate the cause of the poor response of the metrics to human disturbance in the 2002 NCHF data set.

The macroinvertebrate index developed in the large depressional wetland study (Gernes and Helgen 2002) using data from emergent depressional wetlands with a semi-permanent to permanent water regimes, wetland types 4 and 5 in the Circular 39 classification system (Shaw and Fredine 1956). Two of the wetlands sampled in 2002 had plant communities more characteristic of type 8 wetlands (bogs) with floating sedge communities and significant cover (>50%) of sphagnum moss. The main reason these two sites (Hardscrabble and Bush Lake) were selected for sampling in 2002 was to satisfy obligations of another project, comparing MnRAM and IBI assessments. However, since these two sites as well as seven additional wetlands in Hennepin county area were in the NCHF ecoregion, they were included in the data set to provide a temporal control for the testing of the IBI in the plains ecoregions. Given that there were no wetlands similar to these bogs in the 1999 NCHF IBI development data set and the occurrence of Hardscrabble as an outlier in the 2002 data set (Figure I-3), it appears that the current IBI may not be applicable to these types of wetlands. Therefore, correlations between IBI/metrics and measures of disturbance were re-analyzed after excluding the data from these two wetlands in order to determine if the metrics became more responsive to anthropogenic impacts. Results from these correlations (Table I-3) indicate that excluding these two sites from the data set do not

Development and Validation of Temperate Prairie Wetland IBIs 9

improve the performance of the IBI and metrics to levels comparable with the original 1999 NCHF data set (Table I-2). Therefore, inclusion of these two sites in the analysis does not explain the lack of significant correlations between the metrics and the various measures of human disturbance.

The % Tolerant Taxa metric was identified in 1999 NCHF data set as an indicator of human disturbance that increased with increasing disturbance. Some of the taxa comprising this metric were selected based on empirical analyses, identifying those taxa which tended to increase in the more disturbed sites (Gernes and Helgen 2002). Ultimately, this metric was included in the large depressional wetland macroinvertebrate IBI because it was significantly correlated with nitrogen, phosphorus, turbidity, and chlorophyll, indicating that it was responsive to nutrient enrichment. However, in the 2002 NCHF data set this metric was negatively correlated with phosphorus and turbidity, as opposed to the positive correlation observed in the 1999 NCHF data set. A less stringent test of whether this metric is responding to human disturbance, comparing the box plots of the least-impacted sites to the most-impacted sites, yields further evidence that this metric should not be retained in the NCHF depressional wetland IBI (Appendix D). It does not separate the interquartile ranges of these distributions, an often used criterion for selecting biological metrics (Barbour et al. 1996). In fact, one component of this metric, Hyalella azteca (the predominant amphipod taxa collected), has actually been documented as being intolerant to disturbance in wetlands (Rutherford and Mellow 1994, Graves et al. 1998). Since this was the first attempt at validating this metric, it appears that the % Tolerant Taxa metric (in its current form) has limited utility for detecting anthropogenic impacts outside of the data set it was originally developed with (1999 NCHF). Therefore, replacing this metric with a more robust attribute of the macroinvertebrate community or re-examination of the originally selected tolerant taxa of this metric will be necessary.

To explore the possibility that differences in climatic variables between the 1999 and 2002 seasons were responsible for the poor performance of the metrics in 2002, IBI scores were compared at wetlands that were sampled in both years. With the exception of two wetlands (Legion and Malardi) IBI scores remained relatively consistent between years (Figure I-4). In fact, when including only the 2002 sites that had been sampled previously and used in the development of the NCHF IBI, the IBI becomes very strongly correlated to HDS (r = 0.866, P = 0.001). However, looking at metric/disturbance correlations for just these sites still do not result in the number of significant correlations seen in the large depressional IBI study. So it appears that while the metrics are not as responsive to the measures of disturbance (e.g., HDS, Cl, N, etc.) in the 2002 NCHF data set, the IBI still performs well in terms of its sensitivity to anthropogenic disturbance as measured by the HDS.

Since none of the surmised possibilities adequately explained the apparent decreased sensitivity of the metrics to human disturbance in the 2002 NCHF data set, a less stringent approach to evaluating metrics was performed. Box plots of each metric displaying the distributions of the four least-impacted wetlands and the four most degraded wetlands were examined in order to determine which metrics could vertically separate these distributions (Appendix D). In addition

Development and Validation of Temperate Prairie Wetland IBIs 10

Table I-3. Pearson correlation coefficients (r) for relationship between IBI scores/metric values and measures of human disturbance for the 2002 NCHF data set excluding the two wetlands classified as bogs. Bold text indicates significant (P < 0.05) correlation coefficients. Chemistry data was Log10 transformed.

Water Chemistry* Sediment Chemistry # sig. HDS Cl N P Turbidity Cu Ni Pb Zn correlations n

Invert IBI -0.627 -0.244 -0.298 -0.481 -0.295 -0.361 0.056 -0.477 -0.431 1 17 %Dom 3 Genera 0.237 -0.079 -0.026 -0.172 -0.204 -0.315 -0.380 -0.208 -0.230 0 17 %Tolerant Taxa -0.143 -0.280 0.031 -0.534 -0.507 -0.437 -0.655 -0.118 -0.344 2** 17 Corixidae Ratio 0.402 0.560 -0.189 0.028 0.137 0.095 -0.095 0.343 0.161 1 17 Odonata Taxa -0.388 -0.097 0.102 -0.350 -0.046 -0.557 -0.322 -0.534 -0.613 3 17 Leech Taxa -0.159 -0.171 -0.332 0.015 -0.189 -0.168 0.106 -0.584 -0.243 1 17 Snail Taxa -0.243 0.007 -0.351 -0.383 -0.392 -0.131 -0.058 -0.120 -0.126 0 17 Total Taxa -0.667 -0.392 -0.367 -0.544 -0.450 -0.503 -0.067 -0.584 -0.492 5 17 Intolerant Taxa -0.615 -0.143 -0.294 -0.652 -0.317 -0.463 -0.188 -0.349 -0.440 2 17 ETSD -0.400 -0.274 -0.233 -0.684 -0.445 -0.718 -0.526 -0.460 -0.693 4 17 Chironomid Taxa -0.576 -0.216 -0.163 -0.628 -0.263 -0.338 -0.037 -0.174 -0.347 2 17

* Sample size for water chemistry correlations is 15. ** Correlations are opposite direction than that of the original NCHF development data set.

Development and Validation of Temperate Prairie Wetland IBIs 11

to the % Tolerant Taxa metric, other metrics that didn’t separate the 100 impacted wetlands from the un- YEAR impacted wetlands include: % 1999 2002 Dominant 3 Genera, Leech Taxa 80 richness, and Snail Taxa richness. Unlike the % Tolerant Taxa metric, these results do not necessarily 60 indicate the need to immediately replace these three metrics, because they had been validated in a previous 40 project (Gernes and Helgen 2002).

However, continued monitoring of Macroinvertebrate IBI 20 the performance of these metrics should be accomplished before they are accepted within the group of 0 macroinvertebrate metrics that are n l 1 n e e d e ia 2 io rdi ey ri tl o re c e g la N ai r o B la k e a Tu W able to consistently reflect impaired G La L M Pr conditions in NCHF depressional wetlands. Figure I-4. Comparison of macroinvertebrate IBI scores at wetlands sampled in 1999 and 2002.

Adjustment of NCHF Macroinvertebrate IBI

Given the magnitude of the problem with the %Tolerant Taxa metric in the 2002 NCHF data set, the exhibited opposite response to human disturbance, a more thorough empirical selection of both tolerant and intolerant macroinvertebrate taxa was performed. The entire MPCA wetlands biological database, including data from the NCHF, WCBP, and NGP ecoregions, was utilized for this search. This search yielded a number of new tolerant and intolerant taxa as well as supported some of the previous designations from the large depressional wetland study (Appendix E). The revision of these two metrics (%Tolerant Taxa and Intolerant Taxa Richness) coupled with the exclusion of the two bog-like wetlands significantly improved the performance of IBI in the 2002 NCHF data set (Figure I-5). These revisions also improved the relationship between the macroinvertebrate IBI and HDS in the original 1999 NCHF development data set (Figure I-5). The evaluations of these revised metrics in two partially independent data sets represented their validation and thus were used to replace their previous versions in the NCHF macroinvertebrate IBI. Therefore, the inclusion of these revised metrics constitutes the current valid macroinvertebrate IBI for depressional wetlands in the NCHF ecoregion.

Evaluation of Geographical Classification Frameworks

Given the inconclusive results from the previous analysis, evaluating the performance of the NCHF IBI in the plains ecoregions, further investigation into the most appropriate geographical framework for classifying wetlands was required in order to determine if the NCHF IBI should

Development and Validation of Temperate Prairie Wetland IBIs 12

1999 NCHF Development Data Set 2002 NCHF (exc. Bog sites) 100 100

80 80

60 60

40 40

20 r = -0.724 20 r = -0.687 P < 0.001 P= 0.002 Revised Macroinvertebrate IBI Macroinvertebrate Revised Revised Macroinvertebrate IBI Macroinvertebrate Revised 0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS

Figure I-5. Relationship between macroinvertebrate IBI (after revisions to Tolerant/Intolerant metrics) and HDS for the 1999 and 2002 NCHF data sets. be used to assess wetlands in the WCBP and NGP or if separate IBIs should be developed for each ecoregion (Omernik Level III) or if a single IBI should be developed for the two plains ecoregions combined (Omernik Level I or II). Concurrent with this analysis, an evaluation of alternative geographical frameworks (Ecological Classification System, 4-digit Hydrologic Unit Code (HUC)) was used to gain initial insight on the most appropriate classification scheme for depressional wetlands in Minnesota.

The procedures for evaluating geographic classification schemes followed that of Van Sickle (1997) and Van Sickle and Hughes (2000). These methods focus on the use of similarity/dissimilarity coefficients to compare the faunal assemblages of all pairwise combinations of sites. These coefficients can then be grouped according to a priori classifications as either within-class or between-class. The classification strength (CS) of each scheme can be measured by the difference between mean within-class (W) and mean between- class (B) similarity. The classification scheme with the largest difference has the most potential for a framework which can be used to partition aquatic resources (e.g., wetlands), explaining natural variability between classes. Examination of mean similarity dendrograms was also used to summarize the results from all the analyses allowing direct comparisons of the strength of each classification system.

The first step in this type of analysis was the construction of a taxa x site matrix for all of the sites sampled in 2002. In determining the best classification scheme one is only interested in whether the expectations of the assemblage differ by class (e.g., ecoregion, wetland type, etc.), therefore, initially only least-impacted wetlands (HDS < 50) were included in the analysis in an attempt to limit the confounding influence of human disturbance. However, given the limited sample size (N = 22) resulting from this criterion, a separate analysis which included all the sites

Development and Validation of Temperate Prairie Wetland IBIs 13

sampled in 2002 (N = 52) was also performed. The taxa x site matrix was created using the relative abundances of each taxon collected in the dip net samples.

Bray-Curtis dissimilarity coefficients were calculated for each pairwise combination of sites in the matrix using SYSTAT® Version 10.2. Dissimilarity coefficients range from zero to one, with zero indicating that a pair of sites has exactly the same community composition and structure and one indicating that a pair of sites has no taxa in common. Mean similarity analysis was performed using MEANSIM6 software, available on the EPA, Western Ecology Division web site (http://www.epa.gov/ wed/pages/models.htm). This program computes mean between-class dissimilarity (B), mean within-class dissimilarity (W), and the mean dissimilarity within individual classes (Wi). This methodology was used to test the relative strength of several geographic classification schemes: ecoregions (Level II and III; Omernik 1987), ecological classification system (sections and subsections; Bailey 1995, 1998; Cleland et al. 1997) and major river basins (4-digit HUCs). Given the spatial coverage of the wetlands sampled in 2002 the number of classes within each classification ranged from two (Level II ecoregions and ECS sections) to five (river basins) (Table I-4).

Table I-4. Geographic classification frameworks tested for Minnesota depressional wetlands. Classification Classes

Ecoregions Level II: Mixed Wood Plains (MWP) Temperate Prairies (TP)

Level III: North Central Hardwood Forest (NCHF) Western Cornbelt Plains (WCBP) Northern Glaciated Plains (NGP)

Ecological Sections: North Central Glaciated Plains (251B) Classification Minnesota & NE Iowa Morainal (222 M) System (ECS) Subsections: Coteau Morraines (251Bb) Minnesota River Prairie (251Ba) Oak Savanna (222Me) Big Woods (222Mb) Anoka Sand Plain (222Mc)

River Basins Minnesota River (MN) Upper Mississippi (UM) Lower Mississippi (LM) Cedar River (CE) Des Moines River (DM)

In addition to determining the strength of a classification system, MEANSIM6 also uses a permutation test to determine whether the overall strength of a specific classification scheme is significant in the sense of being greater than would be expected from a random set of sites. The statistic CS was calculated for each of 10,000 randomly chosen reassignments of sites to groups of the same size as used in the tested classification. The resulting P-value gives evidence against the null hypothesis of no class structure and was estimated as the proportion of the 10,000 trials

Development and Validation of Temperate Prairie Wetland IBIs 14

having CS at least as large as the observed CS value for the tested classification.

The wetlands sampled in 2002 were located across the southern portion of the state with good representation in three of the Level III ecoregions, two of the ECS sections, and two of the river basins in Minnesota (Figure I-6). In order to be included in the analysis each class is required to have a minimum sample size of two so that within class dissimilarity can be calculated. There was only one instance where this criterion was not met; the Cedar River basin only had one wetland site which was subsequently dropped from the analysis of the classification strength of the river basin framework. The spatial arrangement of sites sampled in 2002 resulted in there being no difference in between the Level I or Level II ecoregion framework. In other words, the tests of these two frameworks would have yielded identical results given the distribution of sites in 2002 and the hierarchical nature of the two levels. Therefore, the results presented here are assumed to be representative of the Level II framework, with less confidence in applying these results on a broader geographic scale (Level I). This same situation also applies to the ECS provinces/sections; results could represent either framework, but were assumed to be more indicative of the framework at the smaller spatial scale (sections).

This analysis of geographic classification schemes was used to determine whether the IBI that was developed primarily from sites within the NCHF should be applied to other Level III ecoregions throughout the state. In both data sets (least-impacted or all 2002 sites) Level III ecoregions was one of the weakest classification frameworks and in the least-impacted sites had a CS value that was not greater than would be expected by chance alone (Table I-5). This would indicate that Level III ecoregions is not an appropriate classification framework for depressional wetlands and an IBI presumably should work across ecoregion boundaries. However, at the next hierarchical level in the Omernik system, Level II ecoregions offer a stronger classification system than Level III ecoregions (Table I-5). This suggests that in the Level III framework the boundary between the NGP and WCBP ecoregions is a less important distinction for wetland macroinvertebrates than the boundary separating the NGP and WCBP ecoregions from the NCHF ecoregion. Given the relative classification strength of the two Omernik ecoregion frameworks (Level II & III), it appears that if we continue to utilize ecoregions as a geographical framework for depressional wetlands that we should use Level II ecoregions. This suggests that a new macroinvertebrate IBI should be developed that would apply to depressional wetlands in both the NGP and WCBP ecoregions.

Before a geographic framework for depressional wetlands is finalized, however, alternatives to the Omernik system were also evaluated. Regardless of which data set (least-impacted or all 2002 sites) was used to test the classification strength of the various frameworks, ECS sections provided the best geographic classification scheme for depressional wetlands (Table I-5). In the least-impacted data set, only the ECS frameworks had CS values that were greater than those expected by chance alone at the α = 0.05 significance level. When all sites were included in the analysis, all classification schemes had significant CS values according to the permutation test. Therefore, determining the optimal framework relied on the relative classification strength of each. A comparison of the framework we have used in previous projects (Omernik ecoregions) versus alternative schemes indicates that our existing framework may not be the most appropriate.

Development and Validation of Temperate Prairie Wetland IBIs 15

NCHF 222 M NGP

WCBP 251 B

Ecoregions ECS Sections Omernik, Level III

Figure I-6. Distribution of 2002 wetland study sites among the various

UM classification frameworks.

MN LM DM

CE River Basins

Graphical examination of mean similarity dendrograms provides further support for the determination of ECS sections as the best geographic framework for wetland macroinvertebrates (Figure I-7). Graphical representation of the mean similarity results are only provided for complete data set (all sites), as the majority of CS values for the least-impacted data set were not significant (P > 0.05). The dendrogram illustrates the between class dissimilarity (vertical line) and the within class dissimilarity (horizontal line). Therefore, in such a figure the optimal classification scheme will have long horizontal lines to the left of the vertical line, indicating that the dissimilarity within classes is much less than the dissimilarity between classes. The ECS

Development and Validation of Temperate Prairie Wetland IBIs 16

Table I-5. Strength of geographic classification schemes for Minnesota depressional wetlands based on macroinvertebrate assemblage data. Classification strength (CS) = [B – W]. P values represent the proportion of 10,000 permutations with random assignment of sites into classes having a CS at least as large as the observed CS value for the tested classification.

Mean Dissimilarity # of between weighted-within classification Classification classes class (B) class (W) strength (CS) P

Least-impacted, N = 22 Level III Ecoregions 3 0.672 0.653 0.019 0.173 Level II Ecoregions 2 0.701 0.665 0.036 0.097 ECS Sections 2 0.706 0.649 0.057 0.017 ECS Subsections 4 0.694 0.645 0.049 0.036 Basins 4 0.688 0.651 0.037 0.088

All sites, N = 52 Level III Ecoregions 3 0.698 0.665 0.033 0.003 Level II Ecoregions 2 0.712 0.670 0.042 0.003 ECS Sections 2 0.724 0.658 0.066 0.000 ECS Subsections 5 0.706 0.646 0.060 0.000 Basins 4 0.700 0.672 0.028 0.029

subsections appear to have a much stronger classification than ECS sections (Figure I-7), however, the sample size within some of the classes that exhibit long horizontal lines (e.g., 222Mc, N = 2) must be considered. For this reason, classification strength (CS) is calculated using the weighted average of the within class dissimilarities.

The lack of significant CS values in the least-impacted data set elucidates some potential problems with the interpretation of the classification strengths based on the all sites data set. Ideally, an analysis of this nature would only include pristine or minimally disturbed wetlands and would be able to identify the classification framework best able to partition the natural variability among the classes. However, particularly in agricultural southern Minnesota, wetlands that are minimally disturbed by human activity are virtually non-existent. Therefore, these tests of geographic classification frameworks for depressional wetlands are forced to include sites that are affected by some degree of human disturbance. Even the least-impacted data set includes wetlands that are moderately affected by human disturbance. So the results from these tests of classification frameworks could be interpreted various ways. For instance, a strong classification framework may in fact adequately represent the natural variability present between classes in the lack of significant disturbance. Alternatively, a strong classification framework may indicate that human disturbance is affecting wetlands within classes similarly. For example, a strong classification framework is just as likely to result if the majority of sites in one class (e.g., North Central Glaciated Plains) are affected by similar practices in their surrounding adjacent upland (e.g., row cropping). Evidence for this exists in the fact that overall

Development and Validation of Temperate Prairie Wetland IBIs 17

LM DM River Basins UM MN

222M ECS Sections 251B

222Mc 222Mb ECS 222Me Subsections 251Bb 251Ba

Omernik MWP Ecoregions Level II TP

NCHF Omernik Ecoregions WCBP Level III NGP

0.0 0.2 0.4 0.6 0.8 1.0 Bray Curtis Dissimilarity

Figure I-7. Mean similarity dendrogram based on data from all sites (N = 52) sampled in 2002. the stronger classification strengths were observed in the data set that included more sites impacted by human disturbance.

Determination of the driving forces behind the classification strengths of the tested frameworks will not be adequately resolved until more data is collected from sites that are minimally disturbed from multiple regions of the state. Until that time, closer examination of some individual sites may provide further insight into which classification scheme does a better job

Development and Validation of Temperate Prairie Wetland IBIs 18

explaining intrinsic variability between classes. The sites to inspect are those that vary in their class affiliation depending of which classification scheme is used. For instance, Carex2, OakGlenEast, and OakGlenWest all belong to the Temperate Prairies ecoregion in Omernik’s framework whereas in the ECS these sites belong to the broadleaf forests class. Given that these sites are relatively undisturbed wetlands, their relative position in an ordination derived from Multi-Dimensional Scaling (SYSTAT® Version 10.2) may indicate whether one system does better than the other in classifying these wetlands. In the Omernik system, two of the sites (Carex2 & OakGlenEast) are located near the edge of the two clusters of points in ordination space, but OakGlenWest appears to be among other sites of a different ecoregion (Figure I-8). When classified according to the ECS system, OakGlenWest then appears to be among sites within the same geographic class. There are a few other cases where the relative position of least-impacted sites in the ordination indicates that the ECS classification framework is better than Omernik ecoregions.

Omernik Level II Ecoregions ECS Sections 2 2 Temperate Prairies 251B Mixed Woodland Plains 222M 1 1

OakGlenW OakGlenW OakGlenW OakGlenW OakGlenE Carex2 OakGlenE Carex2 0 0 DIM2 DIM2

-1 -1

-2 -2 -2 -1 0 1 2 -2 -1 0 1 2 DIM1 DIM1 Figure I-8. Comparison of Multi-Dimensional Scaling ordinations illustrating the better grouping of three sites (Carex2, OakGlenEast, OakGlenWest) by the ECS classification framework.

There have been numerous attempts to discern the best geographic framework for aquatic macroinvertebrate communities; however, the majority of them have been in streams. A test of geographic classification frameworks for macroinvertebrate assemblages in Missouri streams found that Omernik’s ecoregions and Bailey’s ecological sections (=ECS sections) were equally successful at classifying ‘reference’ streams (Rabeni and Doisy 2000). Waite et al. (2000) determined the CS for Omernik Level III ecoregions to be 0.011 for minimally impacted Mid- Atlantic Highland streams, a value comparable to the low CS (0.019) we found for Level III ecoregions when analyzing the least-impacted wetland data set. However, when ecoregion was nested within stream order CS values for ecoregions increased to 0.033. The alternative

Development and Validation of Temperate Prairie Wetland IBIs 19

geographic framework tested by Waite et al. (2000) was catchments, which also had a low CS value (0.010).

The results presented here represent the initial attempt to identify the most appropriate geographic classification framework for macroinvertebrate assemblages of emergent depressional wetlands in Minnesota. This will be a continual analysis as data from more least- impacted sites, representing additional ecoregions becomes available and will not be completed until a state-wide data set for depressional wetlands has been obtained. Therefore, although the results of this analysis suggest that we should adopt ECS sections as a geographic framework, we will continue to utilize Omernik ecoregions until a statewide coverage is accomplished. Given the relative ease of the transition, however, we will begin to use Level II ecoregions rather than Level III ecoregions as suggested by the direct comparison of these two frameworks (Table I-5). In terms of expanding the IBI to other ecoregions, this suggests that a separate macroinvertebrate IBI for depressional wetlands should be developed for the Temperate Prairies ecoregion, which includes the NGP and WCBP Level III ecoregions.

Development of Preliminary Temperate Prairies Macroinvertebrate IBI

The development of a macroinvertebrate IBI for depressional wetlands in the Temperate Prairies Level II (Omernik 1987) ecoregion followed that of the IBI development procedure for the NCHF large depressional wetland project (Gernes and Helgen 2002). Macroinvertebrate community attributes were examined in order to identify those which were sensitive to human disturbance. In particular, attributes significantly (α = 0.05) correlated to HDS or key water chemistry parameters (N, P, Cl conc.) were identified as metrics. Also, box-and-whisker plots of the attributes were examined in order to evaluate the degree of separation exhibited by the distributions of the least impaired and heavily degraded wetlands (sensu Barbour et al. 1996). To evaluate the redundancy of information provided by the metrics, a correlation analysis of all pairwise combinations of candidate metrics was performed. If two metrics were highly correlated (r > 0.85), the more robust metric was retained. To evaluate the strength of each, box- and-whisker plots were examined to determine which metric had greater separation of the most and least disturbed sites. Other considerations for determining which metric was better included the strength of their relationship with HDS and water chemistry parameters. Once a non- redundant set was obtained, each metric was evaluated for its suitability to be scored using the continuous scoring method. Metrics that have a skewed distribution must be transformed to approach a normal distribution before they can be scored using the continuous method (Fore pers. comm.). However, transformation may not ameliorate skewness for all metrics, resulting in their exclusion from the final set of metrics comprising the IBI.

A total of 134 macroinvertebrate community attributes were evaluated for their sensitivity to human disturbance. The selection criteria outlined above yielded eight metrics, and two additional metrics were included because of their demonstrated performance in the NCHF macroinvertebrate IBI and marginally significant (P < 0.10) correlation found in this data set (Table I-6). For instance, the Corixidae Ratio metric was not significantly correlated with HDS, but was correlated with several water and sediment chemistry parameters indicative of human disturbance. Diptera Taxa Richness was included in this IBI as a surrogate for the Chironomidae Taxa Richness metric used in the NCHF IBI. Its correlation with HDS was marginally

Development and Validation of Temperate Prairie Wetland IBIs 20

significant (P = 0.061) and it demonstrated good separation of the least- and most-impacted wetland sites when boxplots of their distributions were graphically examined. While these two metrics did not meet the selection criterion regarding significant correlations with HDS or key water chemistry parameters, they did however satisfy the other criteria: contribution of non- redundant information and suitability for the continuous scoring procedure. An eleventh metric, %Talitridae, met the selection criteria but was ultimately excluded because of its lack of a relationship with HDS (P = 0.579) and its failure to distinguish least- and most-impacted sites via boxplots.

Table I-6. Component metrics of a preliminary macroinvertebrate IBI for the Temperate Prairies depressional wetlands, indicating which portions of the sampling method are used to derive each, whether it was included in the NCHF IBI, and its observed response to human disturbance. Sampling NCHF Disturbance Metric Method Definition metric Response

dip net & Taxa richness of Ephemeroptera and Trichoptera, ETSD yes decrease activity trap plus presence of Sphaeriidae and Anisoptera dip net & Richness of intolerant taxa (determined empirically; Intolerant Taxa yes decrease activity trap see Appendix E) dip net & Odonata Taxa Taxa richness of Odonata yes decrease activity trap dip net & Diptera Taxa Taxa richness of Diptera no decrease activity trap dip net & Total taxa richness (most groups identified to genus, Total Taxa yes decrease activity trap Hirudinea and Gastropoda identified to species). Abundance of Chironomidae divided by total % Chironomidae dip net no increase abundance of sample Abundance of predators divided by total abundance % Predator dip net no decrease of sample Abundance of tolerant taxa divided by total % Tolerant dip net abundance of sample (determined empirically; see yes increase Appendix E) Abundance of Pleidae divided by abundance of % Pleidae dip net no decrease Hemiptera Corixidae Abundance of Corixidae divided by total abundance activity trap yes increase Proportion of Hemiptera and Coleoptera

Overall, six of the metrics selected in this analysis are also component metrics of the finalized version of the NCHF macroinvertebrate IBI (Table I-6). Thus, it appears that a group of macroinvertebrate community attributes are emerging as a reliable set of metrics for depressional wetlands in Minnesota. Expansion to other regions and other wetland types may further increase their applicability which will be tested in the near future as IBIs continue to be developed in other parts of the state. Four new metrics replaced the NCHF metrics that did not respond strongly to human disturbance in the Temperate Prairies ecoregion. The lack of a response for some of the previous NCHF metrics may be due to a truncated disturbance gradient in the

Development and Validation of Temperate Prairie Wetland IBIs 21

Temperate Prairies ecoregion. The NCHF wetland sites ranged from near pristine to severely impacted by both agricultural and urban stressors. In the Temperate Prairies, there were no near pristine wetland study sites and the severely impacted sites were predominantly in an agricultural setting. Therefore, some of the NCHF metrics may have been more sensitive to disturbances characteristic of urban landscapes (e.g., chloride, hydrologic variability “bounce”, etc.), and thus did not exhibit sensitivity to anthropogenic impacts when urban-impacted wetlands were not included in the sample of study sites.

The Human Disturbance Score (HDS) was used as the primary measure of anthropogenic impacts (x axis) against which the response of macroinvertebrate attributes were gauged in order to select metrics. Therefore, it was not surprising that most of the selected metrics were significantly correlated with HDS (Table I-7). Most metrics were also significantly correlated with concentrations of Kjeldahl nitrogen (ammonium and organic nitrogen) in the water column, indicating sensitivity to the agricultural runoff coming into the wetland from adjacent upland areas. However, none of the selected metrics exhibited sensitivity to the other major contributor to eutrophication: phosphorus. At least anecdotally, this suggests that nitrogen may be the limiting nutrient for primary productivity in these wetlands. In a comprehensive literature review, supplemented with additional empirical analyses, Bedford et al. (1999) concluded that of all the wetland types examined (bog, swamp, marsh, poor fen, rich fen, moderate fen) only marshes appear to be predominantly N limited. A study of wetlands in southern Michigan also demonstrated increased periphyton growth with nitrate additions in 18 of the 22 study sites, suggesting nitrogen limitation in these wetlands (Zheng and Stevenson 2001). Given these findings, it is plausible that at a number of our study sites elevated nitrogen concentrations in the water column is leading to hypereutrophication and thus causing significant deviations in the macroinvertebrate assemblages compared to reference sites.

In temperate climates elevated concentrations of chloride in surface waters are often the result of increased runoff originating from roads receiving deicing compounds (e.g., NaCl, CaCl2, MgCl2 and KCL) during the winter. The lack of significant correlations between the metrics and chloride may be due to the restricted range of chloride conditions captured by the 2002 study sites. In the NCHF large depressional wetland study, sites spanned the range of chloride conditions with some wetlands located in heavily urbanized areas, receiving directed stormwater inputs. The concentration of chloride at those wetlands ranged from 1 to 110 mg/L (Gernes and Helgen 2002). With the exception of two wetlands (69, 62 mg/L), the concentration of chloride within the 2002 study sites did not exceed 40 mg/L. These relatively reduced concentrations may not have been as detrimental to the macroinvertebrate community compared to the overwhelming effects of nutrients and pesticides entering the wetlands in this predominantly agricultural part of the state. If this situation accurately reflects what is occurring in these wetlands, detecting any apparent response of the macroinvertebrates to chloride concentrations would thus be problematic at best.

Due to its solubility in water, sulfate is commonly found at high concentrations in surface waters and aquifers (MPCA 1999). Major sources of sulfate in wetlands include atmospheric deposition (resulting from the combustion of fossil fuels), fertilizers (e.g., Ammonium sulfate, Potassium sulfate, Copper sulfate), and decomposition of organic matter within the wetland itself. Since

Development and Validation of Temperate Prairie Wetland IBIs 22

Table I-7. Pearson correlation coefficients (r) between macroinvertebrate metrics/IBI and measures of human disturbance for Temperate Prairie depressional wetlands. Water and sediment chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10).

a dae ax a mi r t t T Tax e a an a an t rono dato r D ler na l Taxa e ole S o do Pr T Pleid ET Int O Diptera Taxa Tota % Chi % % % Corixidae Prop. IBI

HDS -0.402 -0.517 -0.423 -0.365* -0.401 0.417 -0.425 0.352* -0.444 ns -0.624 Water Chemistry: Calcium Carbonate (mg/L) -0.565 -0.534 ns ns ns 0.495 -0.365* 0.522 -0.588 0.389 -0.637 Chloride (mg/L) ns ns ns ns ns ns ns ns ns ns ns Kjeldahl Nitrogen (mg/L) -0.455 -0.515 -0.376* ns ns 0.475 ns 0.554 -0.571 0.334* -0.593 Sulfate (mg/L) -0.576 -0.618 ns -0.359* ns 0.445 -0.518 0.410 -0.803 0.418 -0.744 Phosphorus (mg/L) ns ns ns ns ns ns ns ns ns ns ns Turbidity (NTU)nsnsnsnsnsnsns0.418nsnsns Conductivity (u S/cm) -0.499 -0.570 ns ns ns 0.451 -0.504 0.374* -0.640 0.380* -0.652 Sediment Chemistry: Aluminum (ug/g) -0.508 -0.352* ns ns ns ns ns 0.427 -0.511 0.467 -0.434 Boron (ug/g) -0.442 -0.451 ns ns ns 0.397 ns 0.432 -0.560 0.507 -0.562 Calcium (mg/g) -0.403 -0.396 ns ns ns 0.424 ns 0.524 ns ns -0.346* Chromium (ug/g) -0.522 -0.368* ns ns ns ns ns 0.399 -0.569 0.527 -0.464 Copper (ug/g) -0.518 -0.353* ns ns ns 0.409 ns 0.429 -0.528 0.484 -0.446 Potassium (mg/g) -0.687 -0.612 ns ns ns ns -0.332* 0.395 -0.626 0.456 -0.532 Magnesium (mg/g) -0.646 -0.586 ns ns ns 0.442 ns 0.516 -0.660 0.427 -0.578 Sodium (mg/g) -0.48 -0.458 ns ns ns 0.427 ns 0.424 -0.506 0.604 -0.544 Nickel (ug/g) -0.493 -0.367* ns ns ns 0.366* ns 0.481 -0.476 0.479 -0.446 Sulfur (mg/g) ns ns ns ns ns 0.391 ns 0.440 -0.454 0.373* -0.392 Selenium (ug/g) ns -0.417 ns ns -0.535 0.638 ns 0.555 -0.371* 0.501 -0.612 Strontium (ug/g) -0.393 -0.391 ns ns ns 0.543 ns 0.589 -0.403 0.374* -0.471 Vanadium (ug/g) -0.538 -0.444 ns ns ns 0.370* ns 0.460 -0.549 0.504 -0.517

Development and Validation of Temperate Prairie Wetland IBIs 23

there appeared to be no spatial pattern of sulfate concentrations among the study sites, it can be presumed that input from atmospheric deposition was similar across the region (southern Minnesota). Therefore, local conditions within the wetland’s catchment basin and characteristics of the wetland itself are likely to be the major determinants of surface water sulfate concentrations in the study sites. In terms of the strength and number of significant correlations, the selected macroinvertebrate metrics exhibited the greatest sensitivity to sulfate concentrations in the water column (Table I-7). Without further investigation, however, the significance of these findings in terms of mechanistic pathways affecting the macroinvertebrate community can not be determined.

A number of metrics were also significantly correlated to the concentration of metals in the sediments. For instance, boron which can be an indicator of septic waste was significantly correlated with six of the metrics and the overall IBI score (Table I-7). Several of the metrics were also significantly correlated with the concentration of selenium in the sediments. Although naturally occurring in the environment as a trace element, anthropogenic activities such as the agricultural drainage of seleniferous soils, disposal of fly ash from coal-fired power plants, and the mining of phosphates and metal ores have elevated levels of selenium in wildlife and selected aquatic ecosystems (e.g., wetlands). These are just a few examples of some of the apparent relationships between the metrics and the chemical properties of the wetland sediments. Closer examination of these correlations to determine which are actually representing cause-and-effect relationships will be essential for the remediation of biologically impaired wetlands. It is expected, however, that this will involve an extensive review of the literature, possibly the collection of additional data, and would likely occur during the stressor identification process. Consequently, elucidation of such stressor-response relationships is beyond the scope of this report.

Nine of the ten selected metrics were able to separate the interquartile ranges of the least impaired and heavily degraded wetlands (Appendix F). The Corixidae Ratio metric did not exhibit good separation of the interquartile ranges, but was still retained in the IBI due to the number of significant correlations with disturbance measures (Table I-7) and its demonstrated utility in the NCHF IBI.

Statistical redundancy among the metrics has been dismissed as irrelevant in the development of an IBI provided that the correlation is not representing biological redundancy as well (Karr and Chu 1999). Therefore, the evaluation of redundant metrics did not focus on identifying statistically significant correlations, rather it sought to identify pairs of metrics that were highly correlated (r > 0.85), contributing virtually the same information to the IBI. Such combinations would then be evaluated further to determine which metric was stronger and thus included in the IBI. The final set of ten metrics is the result of this process with no correlation coefficients exceeding 0.85 (Table I-8).

A number of attributes exhibited characteristics of reliable metrics and met the criteria outlined above, but were ultimately excluded from the final set of metrics because they could not be transformed in order to satisfy the requirements of the continuous scoring method. However, these attributes are still worth mentioning as not all wetland monitoring programs utilize the continuous scoring method and thus may prove to be reliable indicators of wetland condition.

Development and Validation of Temperate Prairie Wetland IBIs 24

Table I-8. Pearson correlation coefficients (below diagonal) and corresponding P-values (above diagonal) for pairwise relationships among the ten metrics in the Temperate Prairies depressional wetland macroinvertebrate IBI.

e . da xa a i op a x a r a x m t P T a t T a or n e T x no t a e n a a o a r a a t a r d e d da r r T i l i D e na e l h e o xi e l t a r i l S o p C P T r P do i o T nt ot E I O D T % % % C %

ETSD XXXX 0.000 0.069 0.761 0.908 0.121 0.200 0.004 0.020 0.000 Intolerant Taxa 0.816 XXXX 0.004 0.022 0.461 0.174 0.048 0.018 0.058 0.000 Odonata Taxa 0.356 0.538 XXXX 0.236 0.109 0.153 0.164 0.129 0.262 0.255 Diptera Taxa 0.061 0.439 0.236 XXXX 0.011 0.565 0.017 0.657 0.903 0.375 Total Taxa -0.023 0.148 0.315 0.484 XXXX 0.002 0.137 0.031 0.384 0.871 % Chironomidae -0.306 -0.270 -0.283 0.116 -0.560 XXXX 0.834 0.001 0.030 0.056 % Predator 0.255 0.383 0.276 0.455 0.294 -0.042 XXXX 0.347 0.840 0.130 % Tolerant -0.532 -0.450 -0.300 -0.090 -0.416 0.590 0.188 XXXX 0.014 0.017 Corixidae Ratio -0.445 -0.369 -0.224 -0.025 -0.174 0.418 0.041 0.468 XXXX 0.004 Pleidae Ratio 0.749 0.644 0.227 0.178 0.033 -0.372 0.299 -0.456 -0.537 XXXX

For instance, in the dip net samples, %Hemiptera, %Lymnaeidae, and %Trichoptera were all good measures of wetland condition, but they could not be transformed to correct for their skewed distributions. Consequently, the preliminary IBI for the Temperate Prairies ecoregion incorporated a total of ten metrics, utilizing the above criteria as a guideline for their selection.

A final evaluation of the ten metrics consisted of determining the signal:noise ratios of each. This process can only be performed when replicate data within a wetland has been collected and it utilizes the analysis of variance (ANOVA) to determine the within site variability, represented by the mean square error (MSE). This nuisance variability diminishes the ability of the IBI to detect differences between sites and is caused by a multitude of factors including: time of day, location of sample collection within the wetland, and consistency of sampling and sample processing (crew error). Signal:noise ratios are a simple way to represent the magnitude of nuisance variability, by comparing within site variability (noise) to between site variability (signal). The Intolerant Taxa and Odonata Taxa metrics and the overall IBI score were very precise with signal:noise ratios exceeding 20 (Figure I-9). Other metrics such as %Predator, Total Taxa, %Tolerant Taxa, and Diptera Taxa had signal:noise ratios < 5, but all exceeded a previously suggested minimum value of 2 (Fore 2003a). Overall, the metrics were relatively precise, however, these results were based on replicate sampling (n = 2) at only five wetlands that were sampled on the same day in approximately the same location. Precision of the metrics will continue to be assessed as more replicate data, incorporating additional sources of nuisance variability (e.g., year-to-year differences, location within the wetland), is obtained.

Macroinvertebrate Metrics The ten metrics comprising the macroinvertebrate IBI for depressional wetlands in the Temperate Prairies ecoregion have been used elsewhere with varying frequencies ranging from those that have consistently proven to be reliable indicators of condition in a variety of aquatic habitats to those that have been used infrequently. For instance, Total Taxa Richness, Intolerant

Development and Validation of Temperate Prairie Wetland IBIs 25

9.8 5.1 4.4 22.1 30.5 4.4 5.0 3.2 2.0 4.1 31.2 100%

80%

60%

40%

% of total% variance 20% Noise Signal 0%

e e a IBI rant id e ida tera Chiro ix ETSD onata le % ol d tal Taxa P Dip Int O o Tolerant Predator Cor T % % Figure I-9. Relative variance estimates for the Temperate Prairies macroinvertebrate IBI and its component metrics, comparing the within site variance (Noise) to the between site variance (Signal). Signal:noise ratios are presented above the bars for each metric.

Taxa Richness, and the Proportion of Tolerant Taxa have been suggested as standard metrics that are applicable in most aquatic habitats (Karr and Chu 1999). Other metrics, such as Odonata taxa richness and % Predator have served as effective indicators of human disturbance in other wetland types (e.g., coastal wetlands) or in depressional wetlands in other ecoregions (see Table I-6). While, to the knowledge of this researcher, Pleidae has only been used as a metric in one other study (Lillie et al. 2002).

The ETSD metric is the total number of mayfly and caddisfly genera plus the presence of fingernail clams and dragonflies. This is one of the strongest metrics in both the Temperate Prairies and NCHF macroinvertebrate IBIs. ETSD decreased significantly with increasing human disturbance (as measured by the HDS), nitrogen, sulfate, conductivity, and calcium carbonate. It was also sensitive to increases in the concentration of pollutants in the sediment (Table I-7). Most wetlands in this study typically had 1 or 2 mayfly (e.g., Caenis and Callibaetis) and caddisfly (e.g., Oecetis and Triaenodes) genera present, so metric values generally did not exceed 5. The largest value for this metric was 10 at the OakGlenEast wetland. This metric is similar to the EPT metric used for assessing the condition of streams and the POET (Apfelbeck 1999) or EPOT (Ludwa 1994, Ludwa and Richter 2000) metric used in other states for assessing the condition of wetlands (where E = Ephemeroptera, T = Trichoptera, O = Odonata, P = Plecoptera). The combined taxa richness of mayflies and caddisflies has also been used as an indicator of human disturbance in Great Lakes coastal wetlands (Burton et al. 1999, 2003, Uzarski et al. 2004).

Development and Validation of Temperate Prairie Wetland IBIs 26

The Odonata Taxa metric is the total number of dragonfly and damselfly genera collected in the dip net and activity trap samples. This metric only exhibited a significant correlation with HDS (Table I-7), but this as well as its applicability in the NCHF ecoregion was enough to warrant its inclusion in the IBI. In the Temperate Prairies ecoregion, a few cosmopolitan odonate genera (e.g., Enallagma, Lestes, Anax) are present in both highly degraded and least-impacted wetlands. The presence of these taxa plus the addition of more sensitive genera (e.g., Libellula, Leurcorrhinia, Erythemis) is indicative of a high quality depressional wetland. Most wetlands in this study had at least 2 genera present, while the most occurring at any one site was 10 (GreatOasisWMA). Ten was also the maximum Odonata taxa richness observed in the NCHF IBI development project (Gernes and Helgen 2002). Since this metric was not significantly correlated with any of the chemical parameters, it may be responding to habitat (e.g., vegetation removal) and landscape disturbances (e.g., fragmentation) which are quantified in the HDS rating. Odonata taxa richness has also been used as an indicator of wetland condition in Great Lakes coastal wetlands (Burton et al. 1999, Uzarski et al. 2004) and in wetlands in Montana (Apfelbeck 1999), and Washington (Ludwa 1994, Ludwa and Richter 2000).

Diptera taxa richness proved to be a stronger metric than Chironomidae taxa richness (NCHF metric), and thus replaced it in the Temperate Prairies macroinvertebrate IBI. Since the majority of dipteran taxa collected in this study were in fact chironomids, the Diptera Taxa metric merely represents an augmentation of the Chironomidae Taxa metric used in the previous NCHF IBI, typically adding 3 - 4 taxa to the total number of chironomid taxa collected at a site. Most dipteran taxa were identified to genera, however, some specimens were only identified to family. In terms of the relationships with various measures of human disturbance, this metric was perhaps the weakest of the IBI with only two marginally significant correlations (Table I-7). It did, however, provide good separation of the interquartile ranges of the most and least disturbed wetlands. While Diptera taxa richness has not been previously documented as an indicator of human disturbance in wetlands, Chironomidae taxa richness has been used as a wetland macroinvertebrate metric by other researchers (e.g., Apfelbeck 1999, Gernes and Helgen 1999, 2002). Diptera taxa richness has been utilized as a metric for assessing the condition of lakes (Blocksom et al. 2002) and streams (Butcher et al. 2003). Since this is a preliminary macroinvertebrate IBI for the Temperate Prairies ecoregion, the final recommendation on whether to include Diptera or Chironomidae taxa richness as a metric will be reserved until the IBI is validated with an independent data set.

The proportion of chironomids in the dip net samples, the % Chironomidae metric, is a robust metric with significant correlations with HDS and numerous water and sediment chemical parameters (Table I-7). Based on these relationships, this metric appears to increase with increasing human disturbance. Similar relationships in wetland habitats have been documented by Graves et al. (1998), Apfelbeck (1999), and Burton et al. (2003). In the Temperate Prairies wetlands sampled in 2002, chironomids generally did not exceed 50% of the total number of macroinvertebrates collected in dip net samples. Wetlands that exceeded this percentage tended to also have very high HDS scores.

The one trophic metric included in the IBI, % Predator, is the proportion of macroinvertebrate taxa classified as predators (Merritt and Cummins 1996) to the total number of macroinvertebrates collected in the dip net samples. This metric was negatively correlated with

Development and Validation of Temperate Prairie Wetland IBIs 27

HDS, sulfate, and conductivity (Table I-7). The relative abundance of predators collected in dip net samples was typically between 5 – 40%, with a maximum value of 68% (Carex2). This macroinvertebrate community attribute has also been proposed as a metric to assess condition in Great Lakes coastal wetlands (Kashian and Burton 2000), herbaceous depressional wetlands in Florida (Lane 2003), and forested depressional wetlands in Michigan (Burton et al. 2003). Although in forested wetlands, the response to disturbance is in the opposite direction of that observed in this study.

One of the less frequently used metrics, % Pleidae, is the proportion of Pleidae (=Neoplea sp.) to the total number of hemipterans collected in the dip net samples. This metric exhibited some of the strongest correlations with measures of human disturbance, decreasing with increasing HDS, nitrogen, sulfate, conductivity, and numerous sediment chemistry parameters (Table I-7). Pleids accounted for as little as 0% and up to 92% of the hemipterans collected in the dip net samples. The abundance of pleids has proven to be a reliable indicator of wetland condition in Wisconsin (Lillie et al. 2002). However, in Wisconsin wetlands Pleidae abundance exhibited a positive relationship with human disturbance, opposite to the results observed in this study. The scientific literature also provides examples of both Pleidae being tolerant as well as being intolerant to various types of disturbance. For instance, Harman (1997) found no response in the frequency of Neoplea sp. occurrence in response to eutrophication in lacustrine wetlands. While populations of Neoplea striola declined in response to another type of disturbance, increasing water level fluctuations, in permanent ponds (Gittelman 1974). Given the inconsistency of observed responses to human disturbance, further supporting evidence for the utility of this metric will be required before it is accepted as a reliable metric for wetlands in the Temperate Prairies ecoregion. It is anticipated that this issue will be resolved during the validation of the Temperate Prairies wetland macroinvertebrate IBI with 2003 data.

The Corixidae Proportion metric represents the total number of corixids (water boatmen) divided by the total number of hemipterans (water bugs) and coleopterans (water beetles) collected in the activity traps. This metric was not significantly correlated with HDS in any of the three data sets in which it was tested: 1999 NCHF (Table I-2), 2002 NCHF (Table I-3), and 2002 Temperate Prairies (Table I-7). In the Temperate Prairies ecoregion it was, however, significantly correlated with calcium carbonate, sulfate, and numerous sediment chemistry parameters (Table I-7). In the 2002 Temperate Prairies study sites, the value of this metric ranged from 0.001 to 0.956. This metric was initially intended to represent a herbivore:predator ratio, however, trophic relationships within this family are genera-specific and the most commonly collected corixid, Trichocorixa sp., in this study has been documented as a predator itself (Merritt and Cummins 1996). Therefore, the rationale for this metric relies on a different interpretation of its relationship with human disturbance. Elevated abundance of corixids in disturbed sites is perhaps the result of two factors: 1) an increase in their food source and 2) a decrease in their invertebrate predators. For instance, many corixids are detritivorous and herbivorous feeding on algae, diatoms, fungi and microbial ooze (Hungerford 1948, Thorpe and Covich 1991), while some corixids may also prey upon microcrustaceans, chironomids, and oligochaetes (Pajunen and Salmi 1991, Reynolds 1975, Reynolds and Scudder 1987). All of these food items are generally tolerant to and often flourish in disturbed conditions. Such conditions, combined with a decrease in their less tolerant invertebrate predators, may result in the proliferation of corixids through cascading trophic interactions.

Development and Validation of Temperate Prairie Wetland IBIs 28

Macroinvertebrate IBI Preliminary IBI scores calculated for Temperate Prairies depressional wetlands ranged from 16 (severely degraded) to 85 (least-impacted) out of a possible range of 0-100 (Table I-9). The total IBI score exhibited a strong relationship with the human disturbance score (r = -0.624, P < 0.001; Figure I-10). Compared to the NCHF macroinvertebrate IBI, the IBI developed specifically for the Temperate Prairies was better able to assess the condition of depressional wetlands in this ecoregion. When applied to wetlands in the Temperate Prairies ecoregion, the relationship between the NCHF IBI score and HDS was not as strong both before revisions to the tolerant/intolerant taxa designations (r = -0.478, P = 0.012; Figure I-3) as well as after these revisions (r = -0.524, P = 0.005) were incorporated into the NCHF IBI. In addition to HDS, the Temperate Prairies macroinvertebrate IBI was significantly correlated to nitrogen, sulfate, conductivity, and the concentration of numerous metals in the sediment (Table I-7).

Table I-9. Preliminary IBI scores for Temperate Prairies depressional wetlands sampled in 2002. Site Name Rep IBI HDS Site Name Rep IBI HDS

BarryWMA 1 59.4 52 Lee 1 51.0 54.5 Bryclyn 1 63.7 56 LoneTreeWMA 1 20.3 77 Carex2 1 56.3 15.5 LyonsWMA 1 26.1 66.5 EastlickMarsh 1 32.0 40 Malta 1 26.4 58 FrancoWMA 1 16.3 79 Manchester 1 69.8 32.5 FrancoWMA 2 23.3 79 Milan 1 67.0 64 FurgameWMA 1 58.4 51.5 OakGlenEast 1 72.0 28.5 GoldenWPA 1 38.1 53.5 OakGlenWest 1 84.6 20 GreatOasisWMA 1 74.0 24 OakGlenWest 2 75.2 20 Hancock 1 71.9 49 Prairie Marsh 1 73.3 10 Hoffman 1 54.2 59.5 Prairie Marsh 2 74.4 10 Hoffman 2 53.1 59.5 RolhiksWMA 1 55.4 61 Kerk 1 75.7 16 RostWMA 1 48.2 55.5 LakeCharlotte 1 58.8 45 TylerWMA 1 37.6 58.5 LakeCharlotte 2 64.7 45 WillowLake 1 61.2 65 LakeElisabeth 1 41.2 21 Yohi 1 64.1 37.5

The Temperate Prairies macroinvertebrate IBI score was not significantly correlated with wetland size (r = -0.235, P = 0.238; Figure I-11). Taxa richness was also not significantly correlated with wetland size (r = -0.185, P = 0.355). These results suggest that the classification system used to reduce the natural variability among wetlands assessed by the IBI is adequate. In other words, it appears we do not need to incorporate a fifth factor (wetland size) into our current wetland classification framework. Recall that this framework currently includes ecoregions, HGM wetland classes, structural plant community, and water regime.

A subset of the sites were sampled twice on the same day in order to obtain estimates of the within-site variability caused by factors such as time of day, location of sample collection, and consistency of sampling and sample processing (crew error). Wetland IBI scores varied on average by 4.9 points within sites, ranging from 1.0 to 9.4 (Table I-9). Additional replicate sampling data from Temperate Prairies wetlands in 2003 and 2004 will supplement this data set

Development and Validation of Temperate Prairie Wetland IBIs 29

and be used to estimate the precision of the IBI and determine its ability to detect differences between sites and changes through time within sites (e.g., power analyses).

100 100

80 80

I I 60 60 B

B

I

I

t

t

r

r

e

e

v v

n

n

I I 40 40

20 20 r = -0.624 p < 0.001

0 0 0 20 40 60 80 100 -1 0 1 2

HDS Wetland Size, Log10 (hectares)

Figure I-10. Relationship between Figure I-11. Macroinvertebrate IBI scores macroinvertebrate IBI and human plotted against wetland size (Log10 disturbance score (HDS) for Temperate transformed) for Temperate Prairie Prairie wetlands sampled in 2002. wetlands sampled in 2002.

RESULTS & DISCUSSION – Evaluation of Plant IBI

Evaluation of NCHF Criteria in Plains Ecoregions

The plant NCHF IBI had negative relationships with HDS in both the WCBP and NGP ecoregions, as well as both plains ecoregions combined (Figure I-12). In general, the NCHF IBI-HDS relationships from the plains ecoregions had greater slopes (β1) and smaller y-intercepts (β0) than the 1999 NCHF development data set (Figure I-12). This indicates that the NCHF IBI values in the plains ecoregions have lower scores in least disturbed sites and that the IBI decreases more rapidly as disturbance increases than in the NCHF ecoregion. The strength of the NCHF IBI-HDS relationship (r2) in the NGP ecoregion was greater than the 1999 NCHF development data set, less than in the WCBP, and comparable when both of the plains ecoregion data are combined (Figure I-12). In addition, the analysis detected an outlier in the WCBP data set (and consequently the combined WCBP & NGP data set) that had a very high IBI score at a lower HDS that could be disproportionately affecting the relationship (Figure I-12C&E). These results suggest that the plant NCHF IBI has applicability in both of the plains ecoregions, but may need to be calibrated to reflect metric ranges of the different geography.

Development and Validation of Temperate Prairie Wetland IBIs 30

A B

100 100

β0 = 70.454 β0 = 49.550

β1 = -0.555 β1 = -0.253 80 r2 = 0.479** 80 r2 = 0.104

I 60 I 60

B B

I I

t t

n n

a a

l l

P 40 P 40

20 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS CDC D

100 100

β0 = 66.385 β0 = 52.645

β1 = -0.709 β1 = -0.580 80 r2 = 0.374** 80 r2 = 0.694**

I 60 I 60

B B

I I

t t

n n

a a

l l

P 40 P 40

20 20

0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS

E

100

β0 = 62.209 Figure I-12. Plant NCHF IBI-HDS β1 = -0.694 relationships for (A) 1999 NCHF IBI 80 r2 = 0.466** development data, (B) 2002 NCHF data, (C) 2002-03 WCBP data, (D) 2002 NGP data, and

I 60 B (E) 2002-03 WCBP & NGP (plains ecoregions)

I

t

n data. The y-intercept, (β0), slope (β1), and

a l 2 P 40 coefficient of variation (r ) are included, * = P < 0.1 and ** = P < 0.05. 20

0 0 20 40 60 80 100 HDS

Development and Validation of Temperate Prairie Wetland IBIs 31

The majority of the NCHF IBI metrics showed strong correlation with many of the various measures of disturbance (HDS and chemical parameters) in the different ecoregions (Table I-10). In the WCBP/NGP (= plains), 7 out of the 10 metrics had strong relationships (P < 0.05) with HDS, 2 metrics (Carex Cover and Dominant 3 Cover) had weak relationships (P < 0.10) with HDS, and 1 metric (Persistent Litter Cover) had no HDS relationship. In the NGP ecoregion, where the IBI has a stronger relationship with HDS (Figure I-12), 6 metrics had strong HDS relationships and 4 had no HDS relationships. Only half of the metrics in the WCBP ecoregion had strong HDS relationships. There were few consistencies between the metrics and the various chemical parameters within or across the geographic regions. The Tolerant Taxa Ratio metric was strongly correlated with all three sediment metal parameters (Cu, Ni, and Zn) in both the WCBP and WCBP/NGP. The Aquatic Guild, Nonvascular Richness, Perennial Richness, and Sensitive Taxa metrics were consistently correlated with Kjeldahl Nitrogen in the water column across the various geographic regions. These results indicate that many of the metrics that make up the NCHF IBI have applicability in the plains ecoregions, but these metrics mask the poor performance of metrics that are not responding to anthropogenic disturbance resulting in an overall IBI relationship to disturbance.

The 2002 NCHF data set did not show a significant relationship with HDS (Figure I-12B). This was primarily due to an outlier that had both high IBI and HDS values. This site (Breen) was somewhat atypical compared to the majority of depressional wetland sites considered in the study. The vegetation community at this site consisted of emergent marsh vegetation grading into an aquatic community with an emergent floating mat towards the center. Floating mat communities have a tendency to support a more diverse assemblage of plants, regardless of the degree of anthropogenic disturbance, than a typical fringing marsh community, and thus have relatively higher IBI scores. Removing Breen from the dataset results in a significant relationship between the IBI and HDS (r2 = 0.278, P = 0.025); however, it is more appropriate to include this site in the data set as the vegetation is not different enough to warrant reclassification to a different vegetation type (i.e., type 8 bog; Shaw and Fredine 1956). This result highlights a possible shortcoming of the sampling protocol where the observer is required to select an area that best represents the entire wetland community to place a single plot. In this case it was impossible to locate the plot on both the floating mat and the fringing marsh and thus capture a representative sample. The observer chose to locate the plot on the floating mat which may have given an artificially high IBI score. Two other sites in the 2002 NCHF data set also had floating mat communities (Bush Lake and Hardscrabble). These sites had high abundances of Sphagnum mosses and were classified as type 8 (bog) wetlands. The relationship between the NCHF IBI and HDS when these two sites are removed was not significant (r2 = 0.042, P = 0.428).

The 2002 NCHF data set was intended for the purpose of controlling the possible confounding factor of annual variability in the analysis. This data set, however, does not fulfill the requirements needed to meet this objective. To guard against time, or annual variability, as a factor that may confound the performance of the IBI, data from the same sites sampled during the specified time frame must be used for the analysis. Less than half (8 out of 17 sites, when Bush Lake and Hardscrabble are removed) of the sites in this dataset were both sampled in 1999 and 2002. Thus, the 2002 NCHF data set as a whole is spatially confounded and has limited use for determining the applicability of the NCHF IBI in other geographic regions of the state. However, focusing on NCHF data gathered at the same sites both in 1999 and 2002, annual

Development and Validation of Temperate Prairie Wetland IBIs 32

Table I-10. IBI and metric Pearson correlation coefficients (r) with HDS and selected water and sediment chemistry parameters. All of the chemistry data were Log10 transformed prior to analysis. Results are displayed as: ns = not significant, * = (P < 0.1), and ** = (P < 0.05). Data Source Water Chemistry Sediment Chemistry † Ecoregion HDS Cl N P Cl N P Cu Ni Zn N WCBP -0.611** ns -0.775** ns ns ns ns ns -0.647** ns 14(11) Plant IBI NGP -0.833** ns -0.535* ns -0.648** ns ns -0.602** ns ns 13 WCBP/NGP -0.683** -0.344* -0.560** ns ns ns ns -0.346* -0.477** ns 27(24) Aquatic WCBP ns ns -0.782** ns ns ns ns ns ns ns 14(11) Guild NGP -0.704** ns -0.778** ns -0.505* ns ns -0.476* ns ns 13 Richness WCBP/NGP -0.485** -0.350* -0.780** ns ns ns ns -0.385** ns ns 27(24) Graminoid WCBP ns ns ns ns ns ns ns ns ns ns 14(11) NGP -0.516* ns ns ns ns -0.496* -0.582** -0.851** -0.830** ns 13 Richness WCBP/NGP -0.462** ns ns ns ns ns ns ns ns ns 27(24) 0.535**‡ Nonvascular WCBP -0.615** -0.790** -0.586* ns ns ns ns ns ns 14(11) NGP -0.625** -0.505* -0.634** ns -0.691** ns ns ns ns ns 13 Richness WCBP/NGP -0.618** -0.616** -0.539** ns ns ns ns ns ns ns 27(24) Perennial WCBP -0.558** ns -0.667** ns ns ns ns ns ns ns 14(11) NGP -0.658** ns -0.595** -0.509* -0.479* ns ns -0.680** -0.570** ns 13 Richness WCBP/NGP -0.568** ns -0.532** ns ns ns ns ns -0.360* ns 27(24) Vascular WCBP -0.568** ns -0.699** ns ns ns ns ns ns ns 14(11) Genera NGP -0.662** ns ns ns ns ns -0.598** -0.747** -0.701** ns 13 Richness WCBP/NGP -0.578** ns -0.496** ns ns ns ns ns -0.351* ns 27(24) WCBP ns ns -0.542* ns ns ns ns ns ns ns 14(11) Carex Cover NGP ns 0.477*‡ ns ns ns ns ns ns -0.550* ns 13 WCBP/NGP-0.374*nsnsnsnsnsnsnsnsns27(24) Dominant 3 WCBP ns ns 0.650** ns ns ns ns ns ns ns 14(11) NGPnsnsnsnsnsnsnsnsnsns13 Cover WCBP/NGP 0.326* ns 0.449** ns ns ns ns ns ns ns 27(24) Persistent WCBP ns ns 0.737** ns ns ns ns ns ns ns 14(11) NGPnsnsns-0.574**‡ ns ns ns ns ns ns 13 Litter Cover WCBP/NGP ns ns ns -0.360* ns ns ns ns ns ns 27(24) Sensitive WCBP -0.546** ns -0.528* ns ns ns ns ns ns ns 14(11) NGP ns ns -0.797** -0.637** ns ns ns ns ns ns 13 Taxa WCBP/NGP -0.495** -0.371* -0.421** ns ns ns ns ns ns ns 27(24) Tolerant WCBP 0.711** ns ns ns ns ns ns 0.579** 0.556** 0.633** 14(11) NGP 0.801** ns ns ns 0.747** ns 0.495* 0.761** 0.664** ns 13 Taxa Ratio WCBP/NGP 0.743** ns ns ns 0.385** ns ns 0.476** 0.481** 0.323* 27(24) † N in parentheses are for water chemistry correlations. ‡ Correlation has opposite response to original NCHF IBI development data set.

Development and Validation of Temperate Prairie Wetland IBIs 33

variability appears to be minimal between the two time periods (Figure I-13). With the exception of Breen, there was no IBI score difference greater than 12. Breen, as mentioned before, is anomalous due to the floating mat present at the site and the limitation of the sampling protocol to capture this variability. The 1999 and 2002 site scores illustrate this where the 1999 score was from data collected along the emergent marsh fringe and the 2002 score was from data collected along the edge of the interior floating mat. Excluding Breen (because the large difference between IBI scores is primarily due to within site sampling location and not time), the average absolute site IBI score difference was 8.5, with no consistent pattern of one year being greater than another between the different sites. A two-tailed paired t-test failed to find a significant difference between 1999 and 2002 IBI scores (P = 0.57). Thus, the effect of annual variability is likely not a factor in the performance of the NCHF IBI in the plains ecoregions.

100 Year 1999 80 2002

60

Plant IBI Plant 40

20

0

e e al 21 ri l gion lardi Ney laci a Turt Breen G ake Le Prai L M ew N Figure I-13. Plant NCHF IBI scores from sites that were sampled in both 1999 and 2002.

Overall, the applicability of the plant NCHF IBI in the plains ecoregions based on the direct performance of the IBI is inconclusive. The IBI shows significant relationships to HDS in the plains ecoregions individually, as well as both combined. The differences in the range, slope, and relationship strength (r2) in both the IBI as well as the individual metrics, however, are perhaps indicative of fundamental differences between the wetland plant communities in these different geographic regions. This direct analysis of IBI performance, though, likely does not have the ability to sufficiently detect and/or test fundamental community differences. Further evaluation of plant community differences between the forested and plains regions is therefore warranted.

Evaluation of Geographical Classification Frameworks

While assessing the direct performance of the plant NCHF IBI in the plains ecoregions does provide important information as to the applicability of the IBI in other areas of the state; it is not

Development and Validation of Temperate Prairie Wetland IBIs 34

the only method that can be used to investigate broader geographic IBI applicability. Classification analyses based on raw plant community data can also be used to assess IBI applicability by determining whether wetland communities are fundamentally different according to a given classification scheme or other broad patterns. The direct IBI comparison can be likened to a “top-down” approach where the IBI is developed in a limited area and community differences in other areas are judged based on the similarity of IBI performance; whereas, multivariate classification is a more “bottom-up” approach where different geographic IBI units are based on groupings of similar sites determined from raw community data. Both approaches will be necessary to evaluate the efficacy of the NCHF IBI in the plains region of the state.

As with the macroinvertebrates, mean dissimilarity analysis (Van Sickle 1997, Van Sickle and Hughes 2000) was used to assess multiple geographic classification frameworks with the plant data. Geographic frameworks included Omernik Level II and III ecoregions; ECS Sections and Subsections; and 4-digit HUC River Basins (Table I-4). Initially, these frameworks were tested with four different data sets. The first was 2002 data from least impacted sites (HDS < 50) which was the least confounded data set in terms of both anthropogenic disturbance as well as annual variability, but was limited in size (N = 17). The 2002 data set from all sites was greater in size (N = 41) but included many highly impacted sites. Given that annual variability was likely small between 1999 and 2002 in the NCHF ecoregion (Figure I-13), geographic frameworks were also assessed with data from a broader time frame (1999-2003). This included separate analyses for least impacted (N = 44) and all sites (N = 112). Taxa x site matrices were created for each of the four datasets using the midpoint percent cover of the observed coverclass for each taxon (Appendix C). SYSTAT® Version 10.2 was then used to compute Bray Curtis dissimilarity coefficients for each pairwise combination of sites. Finally, MEANSIM6 software (http://www.epa.gov/wed/pages/models.htm) was used to perform mean dissimilarity analyses and construct mean dissimilarity dendrograms from the resulting dissimilarity matrices.

There were no classification schemes that consistently had the strongest class structure among the four initial datasets (Table I-11). In the 2002 least impacted data set, none of the classification schemes had significant class structure. The low number of sites in this data set (N = 17) may have contributed to this result. There was a different strongest classification for each of the remaining data sets: ECS Subsections in the 2002 all sites set; ECS Sections in the 1999- 2003 least impacted set; and a tie between Level II and III ecoregions in the 1999-2003 all sites set. Of the different classification schemes, ECS Sections was the only one that showed significant (P < 0.05) class structure in all of these three data sets. River Basins, on the other hand, did not have significant class structure in any of the four data sets. The remaining classification schemes showed inconsistent significant classification structure among the datasets.

Focusing on the 1999-2003 least impacted data set, which was the highest quality data set of the four given that human disturbance was relatively minimized and assuming that annual variability was negligible in the time frame, both ECS Sections and Level III ecoregions had significant class structure, with ECS Sections being slightly stronger (Table I-11). While both of these schemes were statistically significant, their CS values were relatively low and there was little separation between the two, warranting further examination of these relationships. Mean

Development and Validation of Temperate Prairie Wetland IBIs 35

Table I-11. Strength of geographic classification schemes based on plant assemblages for 5 data sets. Classification strength (CS) = [B – W]. P-values represent the proportion of 10,000 permutations with random assignment of sites into classes having a CS at least as large as the observed CS value for the tested classification.

Mean Dissimilarity # of Between Weighted-Within Classification Classification classes Class (B) Class (W) Strength (CS) P

2002 Least-Impacted Sites (N = 17) Level II Ecoregions 2 0.822 0.848 -0.026 0.801 Level III Ecoregions 3 0.830 0.843 -0.013 0.588 ECS Sections 2 0.838 0.837 0.001 0.388 ECS Subsections 4 0.832 0.855 -0.023 0.686 Basins 4 0.840 0.780 0.060 0.083

2002 All Sites (N = 41) Level II Ecoregions 2 0.823 0.792 0.031 0.045 Level III Ecoregions 3 0.806 0.794 0.012 0.169 ECS Sections 2 0.830 0.787 0.043 0.023 ECS Subsections† 4 0.830 0.783 0.047 0.017 Basins 4 0.811 0.783 0.028 0.091

1999-2003 Least Impacted Sites (N = 44) Level II Ecoregions 2 0.878 0.860 0.018 0.055 Level III Ecoregions 3 0.873 0.851 0.022 0.033 ECS Sections 2 0.883 0.857 0.026 0.028 ECS Subsections 5 0.874 0.852 0.022 0.056 Basins† 5 0.864 0.866 -0.002 0.520

1999-2003 All Sites (N = 112) Level II Ecoregions 2 0.862 0.825 0.037 <0.001 Level III Ecoregions 3 0.854 0.817 0.037 <0.001 ECS Sections 2 0.863 0.827 0.036 <0.001 ECS Subsections† 6 0.851 0.822 0.029 <0.001 Basins† 5 0.847 0.833 0.014 0.056

1999-2003 Sites Without Invasive Spp. (N = 36) Level II 2 0.950 0.876 0.074 <0.001 Level III 3 0.939 0.875 0.064 <0.001 ECS Sections 2 0.944 0.884 0.060 <0.001 ECS Subsections† 6 0.926 0.858 0.068 <0.001 Basins‡ 3 0.932 0.892 0.040 0.004 † 1 site removed from analysis due to being uniquely classified into a seperate class. ‡ 2 sites removed from analysis due to being uniquely classified into a sperate class.

Development and Validation of Temperate Prairie Wetland IBIs 36

NCHF (19) Level III Ecoregions WCBP (10) Least Impacted NGP (15) Sites 222M (18) ECS Sections 251B (26)

MWP (22) Level II Ecoregions TP (14)

NCHF (22) Level III Ecoregions WCBP (5) NGP (9)

222M (21) ECS Sections 251B (15) Sites Without Invasive Spp. 222Ma (7) 222Mb (6) 222Md (4) ECS Subsections 222Me (3) 251Ba (6) 251Bb (9)

RD (2) Basins UM (14) MN (18)

00.60.2 0.4 0.8 1

Bray Curtis Dissimilarity

Figure I-14. Mean dissimilarity dendrograms derived from plant data collected in 1999-2003 in least impacted sites (HDS < 50) and sites that had low abundance of invasive species, for geographic classification schemes that had significant class structure (P < 0.05) according to mean dissimilarity analysis.

dissimilarity dendrograms revealed that neither of these classification schemes was particularly strong (Figure I-14). Strong classification systems have classes that are isolated (i.e., high between class dissimilarity) and compact (i.e., low within class dissimilarity). Mean dissimilarity dendrograms display isolation by the position of the vertical line along the x-axis and compactness by the length of the horizontal branches. A strong classification scheme should have a dendrogram with a vertical line with a high value along the x-axis and relatively long and equal negative horizontal branches (Van Sickle 1997). Neither of these two dendrograms had consistent long negative branches and both were located at relatively the same position along the x-axis. The forested classes in both schemes (NCHF Level III ecoregion and 222M ECS Section) in fact had slight positive branching indicating that the forested classes were more

Development and Validation of Temperate Prairie Wetland IBIs 37

dissimilar to themselves than with the plains classes. The plains classes (the WCBP and NGP Level III ecoregions and the 251B ECS Section) had negative branching in both classification schemes. This suggests that there is little or no isolation between the forested and plains classes in either scheme but there is some compactness within the plains classes. To further evaluate the difference between ECS Sections and Level III ecoregions, mean dissimilarity analysis was performed with sites that occurred in the WCBP and NGP Level III ecoregions (i.e., plains area of the state) in the 1999-2003 least impacted data set (N = 25). There was no significant class structure of Level III ecoregions within the plains area of state (CS = 0.029, P = 0.104). These results suggest that there are some slight differences between the plains and forested areas of the state in terms of the wetland plant communities and these differences are better captured by ECS Sections.

Classification analyses of this type should focus on data from minimally impacted sites to insure that natural variation is driving classification and not anthropogenic disturbance. Selecting sites based on HDS is one attempt to do this. However, it should be noted that an HDS = 50 (the criterion used to select least impacted sites) is well above the MPCA’s criteria for selecting reference sites. The MPCA considers sites with an HDS < 30 and with no single factor rated above moderately disturbed, as being reference (Genet et al. 2004). The choice of HDS = 50 represents a compromise to reduce the possibly confounding effect of anthropogenic disturbance and simultaneously maintain sufficient numbers to make the results more robust. Thus, while this is an attempt to minimize disturbance, many of the sites are moderately disturbed. Establishment, and subsequent dominance, of exotic invasive plant species is a very common response of wetlands to anthropogenic disturbance (Wilcox et al. 1985, Kadlec and Bevis 1990, Hudon 1997, Galatowitsch et al. 1999, Rachich and Reader 1999, Parendes and Jones 2000). The presence and abundance of invasive plant species is thus a strong homogenizing factor in even moderately disturbed wetland plant communities and could be masking potential classification groupings based on native plant communities. Considering this strong effect of invasive species, mean dissimilarity analysis was performed with sites from the 1999-2003 data set that had very low abundance of invasive species, regardless of HDS. These sites were selected by having less than a 7 % midpoint percent cover of the following invasive species: Typha angustifolia L., Typha X glauca Godr. (pro sp.), Phalaris arundinacea L., and Lythrum salicaria L. This reduced the 1999-2003 data set from 112 to 36 sites.

In general, classification strengths from the without invasive spp. data set were greater than any of the other data sets (Table I-11). All five classification schemes had significant class structure, with Level II ecoregions having the greatest CS value. This included river basins, which were not expected to show strong classification strength because they tend to cross climatic boundaries that control broad vegetation patterns. However, this data set was largely partitioned into the Upper Mississippi and Minnesota river basins, which by chance correspond closely to the forested and plains areas of the state (Figure I-6). This result is therefore likely an artifact of that correspondence.

Dissimilarity dendrograms point to greater differences between classification schemes than suggested by mean dissimilarity analysis alone (Figure I-14). Both Level II ecoregions and ECS Sections had relatively greater isolation between classes (i.e., greater between class dissimilarity) and relatively long and even horizontal branching, indicating compactness. ECS

Development and Validation of Temperate Prairie Wetland IBIs 38

A

2

1 New London

2

- Manchester Overby

n

o i Carex

s Trappers n 0

e Skarpness m OakGlenWest

i

D -1 Level II Level III MWP NCHF NGP TP WCBP -2 -2 -1 0 1 2 Dimension-1

B

2

1 New London

2

-

n Manchester Overby

o i Carex

s

n 0 Trappers

e Skarpness m OakGlenWest

i

D -1

222M 251B -2 -2 -1 0 1 2 Dimension-1

Figure I-15. Nonmetric Multi-Dimensional Scaling ordinations of sites that had low abundance of invasive species based on plant assemblages, from data collected in 1999-2003, for (A) Level II and III Ecoregions and (B) ECS Sections. Labeled sites differ in general classification (i.e., prairie vs. forest) between Omernik and ECS classification systems.

Development and Validation of Temperate Prairie Wetland IBIs 39

Subsections and Level III ecoregions (which have overall stronger CS than ECS Sections) had relatively less isolation, as well as uneven and, in some cases, positive horizontal branching. The dissimilarity dendrogram revealed that the classification structure for ECS Subsections is being driven by a single class (251Bb, Coteau Morraines) and thus this scheme may not necessarily be as strong as results from mean dissimilarity analysis would indicate. Level III ecoregions had similar uneven branching, implying that there may not be strong differences between all three ecoregions in question. Nonmetric Multidimensional Scaling (MDS) was subsequently performed to further examine the classification strength differences between Level II and III ecoregions in this data set using SYSTAT® Version 10.2 (Figure I-15A). With the exception of a couple of outliers (Oak Glen West and New London) the MDS plot shows that Level II ecoregions are both isolated and compact. At the Level III scale, there appears to be some compactness of the NGP ecoregion but little isolation between the NGP and WCBP ecoregions, further supporting Level II ecoregions as the superior classification scheme for this data set.

As with the macroinvertebrate geographic classification analysis, differences between Level II ecoregions and ECS sections were investigated based on a handful of sites that differed in their general classification (forest vs. plains) between the two schemes. Three sites (Oak Glen West, Manchester, and Carex) were classified as plains in Omernik and forested in ECS. In addition, four sites (New London, Overby, Skarpness, and Trappers) were classified as forested in Omernik and as plains in ECS. MDS ordination plots show that Level II ecoregions, with the exception of two outliers, were relatively isolated and compact (Figure I-15A), whereas the same data classified according to ECS Sections has greater compactness of the forested class but is less compact in the plains class (Figure I-15B). The decreased compactness within the plains class concurrently results in a decrease in isolation between classes, causing the weaker overall CS for ECS sections (Figure I-14, Table I-11).

These results suggest that as the MPCA expands the use of plant based IBIs from the forested region of the state to the plains it would be appropriate to adopt Omernik Level II ecoregions as the geographical framework for wetland IBI assessments. Level II ecoregions was the strongest classification scheme when invasive species (a strong plant community homogenizing factor) was removed (Table I-11, Figures I-14&15). In addition, data from 1999-2003 least impacted sites indicate that there are some differences between wetland plant communities of the plains and communities of the hardwood forest and weak, if any, differences at finer classification scales (Table I-11). Ultimately, however, these results are preliminary. The MPCA will continue to reassess geographic classification schemes based on the plant community, coordinated with macroinvertebrate analyses, as more data is acquired across the state.

Given the apparent grouping of sites according to Level II ecoregions and the direct comparison results of the plant NCHF IBI in the plains area, a separate IBI development process for the Temperate Prairies, Level II ecoregion is warranted. This should encompass independent assessment of a wide range of potential metrics according to standardized selection criteria and subsequent incorporation into an IBI, as opposed to directly adapting the existing NCHF IBI to the region. An IBI resulting from this development process will be specifically tailored for Temperate Prairies depressional wetlands and should outperform the NCHF IBI in the region.

Development and Validation of Temperate Prairie Wetland IBIs 40

Development of Preliminary Temperate Prairies Plant IBI

Methods The approach used to develop the preliminary Temperate Prairies (TP) plant IBI expanded on previous MCPA IBI development efforts (Gernes and Helgen 1999, 2002). This was due to the continuing evolution of the science of ecological indicators (NRC 2000, Jackson et al. 2000, Dale and Beyeler 2001) in general as well as specific IBI metric selection research (Karr and Chu 1999, Fore 2003a). In general, the process consisted of selecting a number of metrics from a large pool of potential metrics based on their ability to meet a number of criteria and outperform other potential metrics in the process. The overall goal was to select a diverse assemblage of metrics that capture multiple attributes of depressional wetland plant communities that are found to indicate anthropogenic impacts. Selection criteria were based on 6 general metric attributes/principles which were adapted to narrative or in some cases numerical criteria for application into the metric selection process (Table I-12). Metrics were selected by evaluating each potential metric according to the 6 criteria in a stepwise fashion according to the order presented in Table I-12; eliminating metrics that failed to meet criteria. Following metric selection, each metric was evaluated for scoring methodology (continuous vs. categorical) based on the type of response to HDS (i.e., linear, curve-linear, or nonlinear) and preliminary scoring criteria were developed. Metric scores were then summed to produce the preliminary TP IBI. The preliminary TP IBI and component metrics were then evaluated against the HDS, as well as selected water and sediment chemical data to investigate potential specific responses to chemical pollution.

Results and Discussion The initial potential metric pool totaled 126 metrics. Seventeen potential metrics were judged to be initially not ecologically meaningful. These metrics, however, were not immediately eliminated and were evaluated for ease of quantification, range, and response (Table I-12) as there continued to be a potential that these metrics could be indicative of anthropogenic impacts after an initial assessment. Sixteen were ultimately rejected because of a lack of response to anthropogenic impacts. The remaining potential metric in this group, Persistent Litter Cover, had a marginal relationship to HDS according to linear regression, but did not have stable variance following standard transformation (arcsine transformation for proportions; Montgomery et al. 2001) and was subsequently tested with Spearman-Rank correlation and interquartile range separation. These analyses failed to show a response and the Persistent Litter Cover was subsequently eliminated. Six potential metrics were initially determined as not being reliably quantifiable according to current sampling protocols. As with the non-ecologically meaningful metrics, these metrics were also assessed based on range and response because they may have potential to indicate anthropogenic impacts if sampling protocols are altered to better quantify these metrics, warranting further study of these metrics. All six of these metrics, however, were ultimately rejected due to insufficient range or not having a response to anthropogenic impacts. Eleven potential metrics were eliminated due to having insufficient range. Sixty seven of the remaining 92 potential metrics at this stage were eliminated by not having a response to anthropogenic disturbance. The remaining 25 metrics that met the first four selection criteria were then evaluated against each other to reduce conceptual redundancy. The potential metric with the strongest response was selected from groups of metrics that measured a different aspect of the same attribute (e.g., Graminoid Richness vs. Graminoid Cover). Sixteen metrics were

Development and Validation of Temperate Prairie Wetland IBIs 41

Table I-12. TP ecoregion plant metric selection criteria. Selection Criteria Attribute/Principle Narrative and/or Numerical Selection Criteria

Metric responses must be explainable based on A metric has ecological meaning if a plausible alternative Ecological sound science. Metrics based on spurious hypothesis based on sound ecological principles can be Meaning results ultimately will not be reliable indicators. established that explains the behavior of the metric.

Metrics must be easily and reliably quantifiable in the field. Metrics that cannot A metric is easily quantifiable if data collection methods Easily be accurately measured will introduce are consistent and accurate, and implementation costs are Quantifiable unwanted variance into the IBI or produce within a reasonable operating budget. misleading results. Metrics that require a great deal of time or money will prohibit broad

The MPCA has a stated goal of detecting 3 condition Metrics must have sufficient range to show levels with IBIs (Genet et al. 2005). A metric has Sufficient graduated differences along anthropogenic sufficient range if it is greater than an estimated minimum Range disturbance gradients. metric range where it would be highly unlikely that the metric could meet the detection goa

A metric shows a response to anthropogenic impacts if it Metrics must respond to anthropogenic has a significant relationship (P < 0.10) with HDS using Disturbance impacts. Metrics that do not respond to linear regression when all statistical assumptions are met. Response impacts will not indicate an effect. Metrics that do not meet assumptions can meet this criteria with a significant S

All metrics that show a response will be evaluated according to their conceptual redundancy (i.e., variation of Low IBIs require a diverse assemblage of metrics. a single attribute). Metrics that show the strongest Redundancy Redundant metrics should be eliminated. response in a particular class or category of metrics will be selected. Metrics that have similarly strong responses and are not Metrics that have high variability due to redundant will be evaluated for their precision based on sampling error or natural variability, will not High Precision signal (between site variance) to noise (within site be able to detect changes due to anthropogenic variance) ratios. Metrics with the highest precision impacts. (signal:noise) will be selected. eliminated as being conceptually redundant. Metric selection stopped at this point as the number of metrics had reached the desired range for assembling an IBI (6-12 metrics: Karr and Chu 1999). Thus, precision did not factor into the metric selection process.

A total of 9 metrics met the selection criteria and thus compose the preliminary plant TP IBI (Table I-13). These metrics are distributed into four general metric categories: taxa richness, community structure, sensitive and tolerant taxa, and diversity. Thus, the IBI is well balanced; incorporating multiple aspects of depressional wetland plant communities. Four of the metrics

Development and Validation of Temperate Prairie Wetland IBIs 42

Table I-13. Preliminary plant TP IBI metrics. Metrics are arranged according to four metric categories. A brief description, inclusion into the NCHF IBI indicator, and the observed response to the anthropogenic disturbance are given.

NCHF Disturbance Metric Definition Metric Response

Taxa Richness Aquatic Guild Richness Number of native aquatic plant species. Yes Decrease

Graminoid Richness† Number of native wetland graminoid species. Yes Decrease

Perennial Richness† Number of native wetland perennial species. Yes Decrease

Vascular Genera Richness† Number of vascular genera. Yes Decrease

Community Structure

Guild Count Number of distinct plant guilds‡. No Decrease Typha Proportional Cover of invasive Typha spp. divided by the No Increase Emergent Cover total emergent cover.

Sensitive & Tolerant Taxa

Sensitive Taxa† Number of taxa sensitive to disturbance. Yes Decrease Number of disturbance tolerant taxa divided by Tolerant Taxa Ratio Yes Increase the total taxa richness.

Diversity Shannon Diversity Shannon Diversity index. No Decrease † Metric is Natural Log transformed ‡ Plant guilds as defined by Galatowitsch and McAdams (1994) were Natural Log transformed to correct the effect of a couple of outliers. These outliers were sites that had unusually high species richness at low HDS (OakGlenWest and Carex). Both sites had floating Carex L. mats which are somewhat atypical in this region of the state. Floating Carex mats are indicative of wetland succession from a marsh to a fen plant community. These sites, however, lacked other fen indicator species supporting the argument that they continue to have a more marsh-like plant community and thus should be included in the dataset. Because the increased richness in these sites is likely more a function of this particular community structure than a relationship with anthropogenic disturbance (HDS) these outliers are most likely not indicating a curve-linear relationship with HDS, where there is a rapid increase of species richness at the low range of anthropogenic disturbance, and are best accounted for with transformation in the y-dimension (i.e., metric transformation as opposed to HDS transformation to linearize the relationship).

Development and Validation of Temperate Prairie Wetland IBIs 43

As expected, many of the metrics are strongly correlated with one another (Table I-14). This statistical redundancy is expected because all of the metrics were selected in large part due to their response to anthropogenic impacts (HDS), and thus will have a certain degree of autocorrelation. Also, metrics based on richness counts essentially consist of different subsets of one another, which also leads to autocorrelation. Metrics that are highly correlated with one another can be thought of contributing the same information to the IBI, at least, quantitatively. The conceptual redundancy, however, has been reduced during the metric selection process. By definition, these metrics are measuring different aspects of the plant community regardless of the numerical correlation observed.

Table I-14. Pearson correlation coefficients (below diagonal) and corresponding P-values (above diagonal) for all pairwise relationships between preliminary plant TP IBI metrics.

a l o r a ti e a ty n r x i ion s e t e a Ra r d d G T a e l i l or v i a r t x i u a Cov ve a no op i D G i ni un r t t T s s n s ul s o n i c s m s s c s P n i a e s C e nt t r r a g ns a a ne ne e ne ne d a r e r no h G h P h V h l e S e n c c c c ph a qu i n i n i n i ui y m n ol h A R L R L R L R G T E L T S Aquatic Guild XXXX 0.483 < 0.001 0.002 0.025 0.088 0.195 0.082 0.054 Richness Ln Graminoid 0.141 XXXX < 0.001 < 0.001 0.034 0.001 < 0.001 < 0.001 0.007 Richness Ln Perennial 0.636 0.712 XXXX < 0.001 < 0.001 0.001 0.001 < 0.001 < 0.001 Richness Ln Vascular Genera 0.572 0.647 0.955 XXXX < 0.001 0.009 0.023 < 0.001 < 0.001 Richness

Guild Count 0.431 0.409 0.774 0.864 XXXX 0.067 0.156 < 0.001 0.001

Typha Proportional -0.335 -0.607 -0.583 -0.493 -0.357 XXXX 0.095 < 0.001 0.096 Emergent Cover

Ln Sensitive Taxa 0.258 0.627 0.595 0.436 0.281 -0.328 XXXX 0.001 0.005

Tolerant Taxa Ratio -0.341 -0.666 -0.7 -0.644 -0.643 0.631 -0.597 XXXX 0.057

Shannon Diversity 0.375 0.508 0.732 0.679 0.6 -0.327 0.52 -0.371 XXXX

Preliminary TP IBI Metric Descriptions The first group or class of metrics is the Taxa Richness metrics. These metrics are simple taxa or species richness counts based on different attributes and describe the composition of the community. Richness metrics rely on the general principle that has been observed in numerous ecosystems and biological assemblages that as anthropogenic impacts increase, richness decreases as more sensitive taxa are excluded by more tolerant taxa. All of these metrics were selected in the NCHF IBI, reflecting their consistent reliability as effective metrics in Minnesota

Development and Validation of Temperate Prairie Wetland IBIs 44

depressional wetlands. Three of the Taxa Richness metrics (Graminoid, Perennial, and Vascular Genera Richness) were Natural Log transformed due to the strong effect of the two previously discussed outlier sites.

The Aquatic Guild Richness metric is the number of species in the sample that are truly aquatic (i.e., having a submergent or floating leaf growth form). The Submersed and Floating plant guilds, as defined by Galatowitsch and McAdams (1994), were used to define aquatic guild species. This metric specifically focuses on the aquatic plant community of the wetland. Aquatic guild species richness decreases with increased anthropogenic disturbance (Figure I- 16A) as turbidity intolerant species are excluded due to decreased light availability and/or as natural hydrologic dynamics are altered beyond the range of sensitive species (Adamus and Brandt 1990, Wilcox and Meeker 1991). Study sites in the TP ecoregion typically had 4-6 aquatic guild species with an absolute range of 0-12. Aquatic Guild Richness had one of the weaker responses of the 9 preliminary metrics; however, it did meet selection criteria and does capture a unique wetland plant community component.

The Graminoid Richness metric is the number of native wetland grasslike species in the sample. This includes members of the Poaceae, Cyperaceae, Juncaceae, Juncaginaceae, and Isoetaceae families that are native to Minnesota and have a Wetland Indicator Status ranging from Facultative-Obligate (sensu Reed 1988). Graminoid Richness decreases with increased anthropogenic disturbance (Figure I-16B). Native Graminoids are a diverse and often dominant portion of the emergent community in minimally impacted wetlands. As wetlands become increasingly impacted, native graminoid richness decreases as native tolerant species become more dominant and shade out less competitive species, or as more often is the case, aggressive introduced invasive species (e.g., Typha angustifolia L., Typha x gluaca Godr. (pro sp.), and Phalaris arundinacea L.) replace less disturbance tolerant native species (Wilcox et al. 1985, Kadlec and Bevis 1990, Hudon 1997). The untransformed range was somewhat narrow when the richest sites were excluded, typically ranging from 0-3. The richest sites ranged up to 13. Graminoid Richness, or variations thereof, has repeatedly been found to be a reliable indicator of anthropogenic impacts in wetlands (Galatowitsch et al. 1999, Simon et al. 2001, Mack 2004).

The Perennial Richness metric is primarily defined by the life history strategy of wetland species. This metric includes all of the perennial species that are native to Minnesota and have a Wetland Indicator Status ranging from Facultative-Obligate (sensu Reed 1988). Perennial Richness has a very strong response to anthropogenic disturbance (Figure I-16C) and the greatest range of the Taxa Richness Metrics (2-43). The typical range extends from 2-16, as the two outlier sites have a substantial effect on this metric. Annual and biennial species tend to be pioneering species, meaning that they are adapted to frequently disturbed communities (regardless if the disturbance is natural or anthropogenic in origin) by having the ability to complete a life cycle quickly. Perennial species tend to occupy more stable habitats where they can persist year after year. The Perennial Richness metric removes the potential confounding factor of increased annual and biennial richness at sites that are repeatedly impacted by human activities, thus improving metric performance compared to a total richness metric. Life history strategy based metrics have been previously applied in Minnesota (Galatowitsch et al. 1999) and northern Indiana (Simon et al. 2001) wetlands.

Development and Validation of Temperate Prairie Wetland IBIs 45

ABC 15 3 4

β1 = -0.066 β1 = -0.021 β1 = -0.023

s

s

2 2 s 2

s

s r = 0.235** r = 0.258** s 3 r = 0.418**

e

e

e

n

n

n

h

h

10 2 h

c

c

c

ii

i

i

R

R

R

l

d

d

ii

l

a

i

i

o 2

u

n

n

ii

n

G

e

m

c

r

i

a

t

e 5 r 1

a

P

G

u

n

q

n 1

L

A

L

0 0 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS DEF 4 15 100

r

e

β = -0.019 β = -0.076 v

s

1 1 o

s

C β = 0.944

e 1

t

n 2 2 80

r = 0.361** r = 0.396** n

h

3 e

c 2

i

g r = 0.180**

r

R

t 10

e

n

a

m

r

u 60

e

E

o

n

l

C

e 2 a

d

n

G

l

i

o

i

r

u t 40

a

r

l

G

o u 5

p

c

o

s

r

a 1

P

V

20

a

n

h

L

p

y 0 0 T 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS G HI 3 100 3

β1 = -0.018 β1 = 0.789 β1 = -0.010 r2 = 0.259** 80 r2 = 0.552** r2 = 0.139*

o

i

y

a

t

t

i

x

a

s

a 2 2

r

R

T

e

a

60 v

i

e

x

v

D

i

a

t

i

T

n

s

t

o

n

n

n

e

a 40

n

r

S

a

e

1 l 1

h

n

o

L

S

T 20

0 0 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS

Figure I-16. Preliminary plant TP IBI metric-HDS scatterplots (A-I). The slope (β1) and coefficient of variation (r2) are included, * = P < 0.1 and ** = P < 0.05.

The Vascular Genera Richness metric is the number of different vascular genera in the sample. This is a modification of total species richness. Because individual species are grouped at the Genus taxonomic level, filters such as nativity and wetland indicator cannot be applied. Vascular Genera Richness in the TP ecoregion typically ranges from 2-17 and has a moderate response to disturbance relative to the other metrics (Figure I-16D).

The Community Structure metric class is focused on capturing the physical structure of the wetland plant community. For example, metrics that measure aspects of the abundance distribution in relation to anthropogenic impacts fall into this class. These metrics rely on the

Development and Validation of Temperate Prairie Wetland IBIs 46

general principle that community structure becomes simplified and abundance becomes less even as ecosystems are increasingly impacted (Magurran 1988).

The Guild Count metric is the number of different plant guilds in the sample. In this context, guilds can be thought of groups of species with similar niche requirements. Guilds are defined by unique combinations of regeneration, morphological, physiological, and phenological traits that represent major differences in potential responses to environmental stress (Galatowitsch and McAdams 1994). Guild Count decreases as anthropogenic impacts increase (Figure I-16E). This indicates that minimally impacted sites support a diverse array of niches and this decreases as sites become more impacted. Guild Count in the TP ecoregion ranges from 2-12. This metric has previously not been identified either in the NCHF ecoregion or other indicator projects as a potential metric.

The Typha Proportional Emergent (TPE) Cover metric is the only metric in the preliminary TP IBI that is computed exclusively with percent cover data. This metric is the total midpoint percent cover of non-native invasive Typha spp. (Typha angustifolia and Typha X glauca) divided by the summed midpoint percent cover of all of the emergent species present. In other words, TPE Cover is the proportion of the emergent abundance occupied by invasive Typha. T. angustifolia and T. x glauca are two of the most widespread aggressive species impacting wetlands in North America (Galatowitsch et al. 1999). T. angustifolia was first collected in brackish Atlantic coastal wetlands in the 1820’s where it may or may not have been native. It easily hybridizes with T. latifolia, which is native to Minnesota, to produce T. X glauca which shows strong hybrid vigor. Both invasive species of Typha have spread rapidly westward across North America since World War I, becoming established in the prairie pothole region during the 1950s (Kantrud 1992). Today these two species likely are the most abundant emergent wetland vegetation in the region. TPE Cover has the second weakest metric response (at least in terms of r2; Figure I-16F); however, there is a great range in the metric. Sites tend to either have 0 or 100% TPE Cover indicating the ability of these two species to completely dominate the emergent community when present.

The Sensitive and Tolerant Taxa metric category focuses directly on the prevalence of sensitive and tolerant taxa. Many of the other metrics are indirectly measuring these aspects as the general model of ecosystem degradation due to anthropogenic impacts is 1) the loss and decreased abundance of sensitive species and 2) the presence and increased abundance of tolerant species as impacts increase. For example, when species richness decreases, the first species to be lost are the species that are sensitive to anthropogenic impacts. Thus, in a sense, species richness is an indirect measure of the decline of sensitive species. Sensitive and Tolerant types of metrics have long been identified as core indicators of ecological health in general (Karr and Chu 1999) and for wetland plant communities (Simon et al. 2001, Mack 2004). Both of the metrics in this category (Sensitive Taxa and Tolerant Taxa Ratio) are included in the NCHF IBI.

The Sensitive Species metric is a simple count of the number of taxa that are sensitive to anthropogenic disturbance. The sensitive species list for Minnesota was determined by literature research and best professional judgment in Gernes and Helgen 2002. This metric was the strongest metric in the NCHF IBI but was one of the weaker metrics in the TP IBI (Figure I- 16G). The typical range was somewhat narrow, ranging from 0-5. However, the outlier sites

Development and Validation of Temperate Prairie Wetland IBIs 47

had an effect on this metric, raising the maximum to 12. Thus the metric was Natural Log transformed.

The Tolerant Taxa Ratio metric was the strongest metric in the preliminary TP IBI. Tolerant Taxa Ratio is the number of anthropogenic disturbance tolerant taxa divided by the total of all taxa in the sample. In other words, it is the proportion of taxa that are tolerant. This metric increases with an increase in anthropogenic impacts (Figure I-16H). Tolerant taxa include introduced species as well as native species that show the ability to tolerate anthropogenic disturbance, and were determined for Minnesota through literature research and best professional judgment in Gernes and Helgen (2002). Tolerant Taxa Ratio ranged from 12-100% in the TP ecoregion.

The final metric (Shannon Diversity) is a diversity index. Diversity indices combine the two facets of diversity into a single index: species richness and evenness of the abundance distribution (Magurran 1988). Because of this property these types of indices are in a category of their own, describing more than just the composition (taxa richness metrics) or the structure (community structure metrics) of a community. A criticism of Shannon Diversity is that it does not provide a clear biological link to community changes due to anthropogenic impacts which could lead to misleading results (Washington 1984). For example, at low to moderate levels of disturbance, additional introduced species may become established in a wetland. If the diversity indices are including all species present the indices will show an increase in diversity which is opposite to the expected response. While this has been an observed outcome and justifiably used as grounds to reject diversity indices as environmental indicators or metrics, diversity indices remain conceptually attractive because of their integrating properties and should continue to be evaluated.

Shannon Diversity (H’) was calculated by multiplying the proportional abundance (pi) by the natural log pi for each species (i) and summed for all species (S) in the sample. S H '= −∑ pi ln pi n=i All vascular taxa were included in calculations. Shannon Diversity passed the metric selection criteria but was the weakest metric in the preliminary TP IBI (Figure I-16I).

Preliminary Plant Temperate Prairies IBI All of the metrics were scored continuously according to the formulas in Figure I-2. Metrics scores ranged from 0-10. A total of 104 values, which represented all of the data collected in TP ecoregion from 2002-04, were used to determine the various scoring cuts (e.g., minimum, maximum, 5th, and 95th percentile) for each metric. This included the entire primary and replicate samples in the IBI development set (2002-03), the IBI validation/Redwood River watershed set (2003), and the precision set (2004). All nine metric scores were then summed together and multiplied by a scaling factor (10/9 = 1.11) to produce the preliminary TP IBI. The additional scaling was performed to increase the possible maximum IBI score from 90 to 100. The MPCA has decided that an IBI that ranges from 0-100 is easier to conceptualize and communicate to others.

Development and Validation of Temperate Prairie Wetland IBIs 48

Preliminary plant TP IBI scores ranged from 14.0-98.2 (Table I-15). By design, the preliminary TP IBI had a strong relationship with HDS (Figure I-17). Compared to the NCHF IBI (Figure I- 12E) the preliminary TP IBI had a greater y-intercept (β0), more negative slope (β1), and greater strength of relationship (r2). The preliminary TP IBI clearly outperformed the NCHF IBI in this ecoregion.

Table I-15. Preliminary plant TP IBI scores for the development set.

SiteName PrelimIBI HDS SiteName PrelimIBI HDS

BarryWMA 41.6 52 LoneTreeWMA 18.9 77 Bryclyn 83.8 56 LyonsWMA 34.0 66.5 Carex 90.8 26.5 Malta 32.0 58 EastlickMarsh 55.4 40 Manchester 80.1 32.5 FrancoWMA 14.0 79 Milan 49.6 64 FurgameWMA 58.2 51.5 OakGlenEast 60.1 28.5 GoldenWPA 26.7 53.5 OakGlenWest 98.2 20 GreatOasisWMA 69.5 24 Prairie Marsh 84.4 10 Hancock 65.6 49 RolhiksWMA 44.1 61 Hoffman 42.1 59.5 RostWMA 40.3 55.5 Kerk 66.0 16 TylerWMA 40.8 58.5 LakeCharlotte 30.5 45 WillowLake 60.8 65 LakeElisabeth 47.8 21 Yohi 68.0 37.5 Lee 42.2 54.5

The preliminary TP IBI was correlated with several water and sediment chemistry parameters 100 β = 92.722 (Table I-16). These included Kjeldahl Nitrogen in 0 β = -0.839 the water column and two metals (Copper and 80 1 r2 = 0.515** Nickel) in the sediments. There were no metric correlations with water and sediment Phosphorous I 60

B

I

as well as sediment Nitrogen and Zinc. t

n

a

Interpretation of these results is complicated by l the general response of wetland plant P 40 communities to anthropogenic stressors where multiple stressors lead to a similar response in the 20 community. For example, invasive Typha are likely the most abundant emergent wetland plants 0 in the TP ecoregion. Invasive Typha can tolerate 0 20 40 60 80 100 moderate to high saline conditions (i.e., chloride; HDS Kantrud 1992) as well as outcompete other species in high nutrient conditions (N and P; Kadlec and Bevis 1990). The TPE Cover metric Figure I-17. Preliminary plant TP IBI- is correlated with chloride in sediments but not HDS scatterplot with y-intercept (β0), with chloride in the water column or any of the slope (β1), and coefficient of variation nutrient parameters. Neither nutrients nor (r2). ** = P < 0.05. chloride can be directly linked to the prevalence

Development and Validation of Temperate Prairie Wetland IBIs 49

Table I-16. Preliminary TP IBI and metric Pearson correlation coefficients (r) with selected water and sediment chemistry parameters. All of the chemistry data were Log10 transformed prior to the analysis. Results are displayed as: ns = not significant, * = (P < 0.1), and ** = (P < 0.05). Water Chemistry (N = 24) Sediment Chemistry (N = 27) Cl N P Cl N P Cu Ni Zn

Plant IBI ns -0.558** ns ns ns ns -0.503** -0.579** ns Aquatic Guild -0.350* -0.780** ns ns ns ns -0.385** ns ns Richness Ln Graminoid ns ns ns ns ns ns ns ns ns Richness Ln Perennial Richness ns -0.652** ns ns ns ns ns -0.333* ns Ln Vascular Genera ns -0.567** ns ns ns ns -0.345* -0.381* ns Richness Guild Count ns -0.512** ns ns ns ns -0.389** -0.489** ns Typha Proportional ns ns ns 0.482** ns ns 0.355* ns ns Emergent Cover Ln Sensitive Taxa -0.498** -0.401* ns -0.356* ns ns ns ns ns

Tolerant Taxa Ratio ns ns ns 0.385** ns ns 0.476** 0.481** ns

Shannon Diversity ns -0.391* ns ns ns ns ns ns ns

of Typha from this dataset. However, invasion and increased abundance has been linked to multiple stressors that not only include salinity and nutrient enrichment but hydrologic disturbance (Wilcox et al. 1985) and physical vegetation removal (Galatowitsch et al. 1999). Thus, it is more likely that a combination of stressors are acting in combination to produce responses in the vegetation community, obscuring the relative roles of individual chemical parameters.

The preliminary TP plant IBI was negatively correlated with wetland area (Figure I-18A). This correlation was not due to the well known species-area relationship, where species richness increases with larger area, as the sampling area was standardized between sites. HDS, however, is positively correlated with wetland area (Figure I-18B), indicating that as wetland size increases it is more likely to be impacted by anthropogenic stressors. This result would be expected in this heavily agricultural landscape where less than 1% of remnant natural habitat exists (MN DNR; http://www.dnr.state.mn.us/prairierestoration/index.html) in relatively small parcels, making it unlikely that a large wetland will not be adjacent to a human altered landuse. Therefore, it is likely that the IBI-area result is due to an autocorrelation effect of the relatively stronger HDS-area correlation and not due to a wetland-area effect on IBI scores.

Precision remains a core component for evaluating IBI performance, though it did not play a role in selecting metrics. During the 2002 field season, a handful of sites were sampled twice to evaluate sampling error according to MPCA QA/QC protocols (Appendix C). Signal to noise

Development and Validation of Temperate Prairie Wetland IBIs 50

AB

100 100 r2 = 0.121* r2 = 0.226** 80 80

I 60 60

B

I

S

t

D

n

a

H

l

P 40 40

20 20

0 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Sqrt Wetland Area (ha) Sqrt Wetland Area (ha)

Figure I-18. Scatterplots of the preliminary plant TP IBI (A) and HDS (B) with wetland area (ha). Wetland area has been square root transformed. * = P < 0.1, ** = P < 0.05. ratios (signal:noise) were computed from variance estimates using ANOVA with the replicate data. A signal:noise is the between site variance (signal) divided by the within site variance (noise). Indicators that maximize signal:noise can detect environmental changes to due anthropogenic stress (signal) in the face of natural variability and sampling error (noise). The preliminary TP IBI and all component metrics had signal:noise greater than 10 (Figure I-19). For comparison, a signal:noise greater than 2 has been identified as a metric precision goal Mid- Atlantic Steam metrics (Fore 2003a). This indicates that sampling error has only a minor effect on the IBI and component metrics, meeting MPCA QA/QC standards. Precision will continue to be evaluated, including the incorporation of sampling location and inter-annual variability, during the validation of the preliminary TP IBI.

The development process presented here is a first step in the overall process for creating a fully developed wetland bioassessment method. Before the IBI can be finalized a validation process will be necessary to assess the consistency of the IBI, where the component metrics will be re- evaluated with an independent data set.

Development and Validation of Temperate Prairie Wetland IBIs 51

11.125.3 78.0 10.0 23.0 181.6 110.1 28.4 33.0 45.1 100

80

60

40

Noise 20

% of total variance total % of Signal 0

s ity ss s s nt IBI nes Count vers la hne h d e Taxa i ichness ic ic v xa Ratio P R siti n D d R l R Guil gent Covern Ta o ia r uil era G en erant ic erenn Ln Se Shann at r G Tol P la onal Eme Aqu Ln GraminoidLn Richne cu as V Ln a Proporti ph Ty

Figure I-19. Relative between site (signal) and within site (noise) variance estimates for the preliminary Temperate Prairies Plant IBI and component metrics (n = 2). Signal to noise ratios are given above the bars for each metric and the IBI.

Development and Validation of Temperate Prairie Wetland IBIs 52

II. Validation of Preliminary Temperate Prairies Wetland Macroinvertebrate and Plant IBIs and Testing their Applicability in Seasonal Wetlands

INTRODUCTION

Based on the 2002-03 data from emergent depressional wetlands in the Temperate Prairies (TP; Omernik Level II) ecoregion, a preliminary macroinvertebrate and plant IBIs were developed. This IBI and its component metrics exhibited responses to anthropogenic disturbance as measured by the Human Disturbance Score (HDS), water column nitrogen and sulfate concentrations, turbidity, and/or sediment heavy metal concentrations. To examine whether these observed relationships are consistent spatially (e.g., in a different set of sites) and temporally (e.g., from year-to-year) and before the ‘preliminary’ status of the index can be removed, a validation process with an independent data set is required.

In 2003 the MPCA initiated a probabilistic survey of depressional wetlands in the Redwood watershed (Genet 2006). The Redwood watershed is located in the TP ecoregion and thus the wetlands sampled in this study could serve as an independent data set for validating the preliminary IBI. However, since site selection in the Redwood was random it may not be the ideal data set for validation because the range of wetland condition (least-impacted to severely degraded) may not be captured by the study sites. Another concern with the Redwood wetland study sites is the inclusion of seasonal wetlands. The preliminary IBIs were developed using data from semi-permanent to permanent depressional wetlands (water regime modifiers F, G, and H sensu Cowardin et al. 1979). A number of wetlands sampled in the Redwood survey were seasonal (water regime C), and thus may not be appropriate to assess using the preliminary IBIs.

Hydrologic regime and associated chemical characteristics are major factors in determining the composition and structure of aquatic invertebrate communities in wetlands (Kantrud et al. 1989, Neckles et al. 1990, Batzer and Wissinger 1996, Euliss and Mushet 2004). Compared to permanently and semi-permanently inundated wetlands, the macroinvertebrate community of temporarily and seasonally flooded wetlands tend to be dominated by invertebrates that have adapted life history strategies that enable them to quickly colonize these habitats once they become inundated. For instance, many wetland invertebrates have desiccation-resistant stages that allow them to overwinter in the dry basin until reflooding occurs. A second strategy for rapid colonization of ephemeral wetlands is adult immigration and oviposition. As a result of the required traits for inhabiting ephemeral wetlands, the invertebrate communities within these habitats are often distinct from those within more permanent wetland types. Therefore, adhering to the rationale of minimizing natural variability in wetlands through proper classification, it may be more appropriate to assess the condition of seasonal wetlands with a separately developed index, rather than using the IBI developed for semi-permanent and permanent wetlands.

Likewise, hydrologic regime is a major determinant of wetland plant communities (Mitsch and Gosselink 2000). Vascular plants, being aerobic organisms, require oxygen for survival. Inundation in large part controls the availability of oxygen in wetland soils by slowing gas

Development and Validation of Temperate Prairie Wetland IBIs 53

exchange with the atmosphere to the point where decomposers (i.e., bacteria and fungus) consume the available oxygen producing a low oxygen or anoxic environment in the rooting zone of the soil. Wetland plant communities that vary in composition and structure have adapted to the soil environments produced by various hydrologic dynamics ranging from seasonally saturated soils (wet meadow) to permanently saturated soils (emergent marsh) to soils with permanent inundation of surface water (shallow open water/aquatic). Thus, plant community variability produced by this natural factor has great potential to effect the ability of a plant based IBI to detect changes due solely to anthropogenic impacts (Wilcox et al. 2002) and needs to be assessed.

The objectives of this component of the project were to: 1) validate the preliminary IBIs developed for depressional wetlands in the TP ecoregion with the 2003 Redwood data set; 2) make any necessary adjustments to the IBI taking into consideration the results of the analysis with the 2003 Redwood data set; 3) finalize the IBIs for depressional wetlands in the TP ecoregion; and 4) assess the applicability of the IBIs in seasonally inundated depressional wetlands located in the TP ecoregion.

METHODS

Field Sampling

In 2003, as part of a probabilistic survey design for assessing the condition of depressional wetlands in the Redwood River watershed, wetland study sites were selected randomly from a sample frame that was generated based on the National Wetlands Inventory (NWI). Another report (see Genet 2006) contains details of sample frame preparation, sample design parameters, site reconnaissance, and site selection. In addition to these randomly selected sites, a number of wetlands previously sampled in 2002, were once again sampled in 2003 as part of MPCA’s wetland trend monitoring efforts. Eight of these sites were located in the TP ecoregion and thus could potentially be used in the IBI/metric validation process. Of the 48 TP wetlands sampled in 2003, 14 were classified as seasonally inundated (water regime C).

Macroinvertebrates were sampled during the seasonal index period of June using the same sampling protocols used in previous projects (see Appendix B). During the macroinvertebrate sampling visit, water chemistry parameters were measured using dissolved oxygen (DO), pH, and conductivity probes. Surface water grab samples were also collected during each visit for laboratory analysis of the following water chemistry parameters: total Kjeldahl nitrogen (mg/L), total phosphorus (mg/L), total chloride (mg/L), calcium (CaCO3 mg/L), turbidity (NTU), and total sulfate (mg/L). Analysis of these water chemistry parameters was conducted by the Minnesota Department of Health, Environmental Laboratory.

The emergent plant community of each wetland site was sampled during the seasonal index period of July using sampling protocols established in previous projects (see Appendix C). During the plant sampling visit, three sediment cores were collected from the emergent zone of each wetland and the top 5 cm of sediment were extruded and pooled for analysis of nitrogen, Olsen phosphorus, chloride, % moisture, pH, total organic carbon, and heavy metals by the University of Minnesota Soils Analytical Laboratory. In addition, supplementary site

Development and Validation of Temperate Prairie Wetland IBIs 54

information, required for the calculation of an HDS, was noted during each plant visit. This included observations such as the presence of water control structures, man-made berms, vegetation removal, dredging, and vehicle use within the wetland.

During the 2004 field season, 6 previously sampled wetlands were intensively sampled to estimate plant IBI and component metric precision. These samples, along with samples from 2002-03 at these sites, facilitated the estimation of the following variance components: crew error, within wetland variability (location), and inter-annual (between year) variability. Three different locations were sampled in each of the 6 sites for both macroinvertebrates and plants. In addition, an alternate plant sampling method was used to explore possible IBI performance improvements using different sampling schemes. This method consisted of deploying a set of 4 small (5m x 5m) plots as a sample, as opposed to a single large (10m x 10m) plot. A set of small plots were sampled at each general location adjacent to the primary large plot samples. For consistency, small plots were located at the emergent marsh/open water interface.

As with the preliminary IBI development dataset, Human Disturbance Scores were generated for all of the wetland sites sampled in 2003. This process utilized several sources of information such as: field observations, water and sediment chemistry data, surrounding land use, and aerial photography. For more details about this site rating system see Gernes and Helgen (2002).

Macroinvertebrate IBI Validation

Metric values were calculated for all of the semi-permanent and permanent depressional wetlands in the TP ecoregion sampled in 2003. These metrics were evaluated against various measures of disturbance such as HDS and water chemistry parameters to determine if the stressor:response relationships identified in the analysis of 2002 data exist within the 2003 data set. The strength of Pearson correlation coefficients (r) and their direction were used as criteria for evaluating whether the stressor:response relationships were replicated in the 2003 data set. Based on the results of this analysis, modifications to the preliminary IBI were conducted in order to create an IBI that was a strong indicator of human disturbance in both data sets, thus allowing the removal of the preliminary status from the TP wetland macroinvertebrate IBI.

Once a final version of the IBI was established, data from the TP depressional wetlands sampled in 2002 and 2003 were used to derive the scoring criteria necessary to generate metric scores according to the continuous scoring procedure developed by Fore (2003b). Scoring criteria for this process includes the minimum, maximum, 5th percentile, and 95th percentile values for each individual metric. These criteria were used to convert metric values into metric scores, allowing the calculation of site IBI scores. Pearson correlation coefficients (r) were used to gauge the sensitivity of the IBI to various types of disturbance in each year separately. Data from sites where replicate samples were collected in 2003 were used to supplement the 2002 data set for evaluating the precision of the final IBI and its component metrics.

Development and Validation of Temperate Prairie Wetland IBIs 55

Testing the Applicability of Macroinvertebrate IBI in Seasonal Wetlands

Two primary techniques were used to determine whether the macroinvertebrate IBI developed for assessing semi-permanent and permanent depressional wetlands in the TP ecoregion could also be used to assess seasonal depressional wetlands within this ecoregion. The first method utilized the same principles and techniques used to evaluate alternative geographic frameworks for minimizing natural variability within wetland classes (see Section I, Evaluation of Geographical Classification Frameworks). This method uses the raw biological data, in the form of a taxa x site matrix, in order to determine whether there are inherent differences in the biological communities between classes. However, instead of using a geographic framework (e.g., ecoregions) to classify wetland sites, the most permanent water regime accounting for at least 5% of the wetland basin (Stewart and Kantrud 1971) was used to create the classes. A combination of field observations and NWI data was used to assign one of three NWI water regimes to each of the 2003 study wetlands: seasonal (C), semi-permanent (F), and intermittently exposed (G).

Bray-Curtis dissimilarity coefficients were calculated for each pairwise combination of sites in the matrix. Dissimilarity coefficients range from zero to one, with zero indicating that a pair of sites has exactly the same community composition and structure and one indicating that a pair of sites has no taxa in common. Nonmetric multidimensional scaling (MDS) was used to compute coordinates that best approximated distances represented by the dissimilarity coefficients for all pairwise comparisons. These coordinates were then plotted in order to evaluate whether wetland sites clustered according to water regime class. Both two-dimensional and three-dimensional plots were examined in this manner. All statistical analyses and graphical plotting was accomplished using SYSTAT® Version 10.2.

The second method for evaluating the applicability of the IBI focused on determining whether component metrics of the IBI are sensitive to human disturbance in seasonal wetlands. This approach would indicate whether any differences in the macro-invertebrate community between seasonal and semi-permanent/permanent wetlands elucidated by the nonmetric MDS analysis hinder the ability of the IBI to detect human disturbance in seasonal wetlands. If the metrics retain their indicator qualities in seasonal wetlands, then further evaluation will examine whether separate scoring criteria is required for seasonal wetlands in order to reflect the intrinsic differences in the macroinvertebrate communities between seasonal and semi- permanent/permanent wetlands (e.g., differences in total taxa richness).

If it is apparent that IBI developed to assess the condition of the semi-permanent/permanent depressional wetlands in the TP ecoregion is not a useful indicator of quality in seasonal wetlands, then an initial attempt will be made to find attributes of the macroinvertebrate community that are indicative of human disturbance in seasonal wetlands. This analysis would utilize data collected from the seasonal wetlands sampled in 2003 and follow the same step-by- step progression used to identify the metrics for the semi-permanent/permanent depressional wetlands macroinvertebrate IBI.

Development and Validation of Temperate Prairie Wetland IBIs 56

Plant IBI Validation

The plant preliminary TP IBI was computed for all sites sampled in the TP ecoregion in 2003. This included all of the Redwood watershed and Trend sites. The preliminary TP IBI and component metrics were then assessed by determining their relationship with HDS using linear regression when statistical assumptions could be met and Spearman-Rank correlation or inter- quartile separation when assumptions could not be met. In addition, any potential confounding effects that could affect metric responses to anthropogenic impacts were also examined. This second component included investigating factors such as water regime (seasonal/temporary vs. semi-permanent/permanent); potential acute impacts due to ditching and subsurface tiling; as well as the distribution of the HDS. Metrics that had a significant HDS relationship (P < 0.10) and were not affected by misclassification or other confounding factors were accepted as final validated metrics. Metrics that were substantially affected by misclassification were re-assessed for applicability for depressional wetland assessment. Metrics that were not affected by misclassification and did not have a significant response to HDS were adjusted or replaced by additional metrics that did show a strong HDS response. Ultimately, the goal of the validation process was to derive a number of metrics that had optimal performance in both the development and validation datasets while maintaining the structural balance required in an IBI.

Following the validation and subsequent adjustment and replacement of metrics, continuous scoring cuts (minimum, maximum, 5th and 95th percentiles; Figure I-2) were determined for all new metrics. These were derived from all 2002-04 data collected in the TP ecoregion which totaled 104 samples. The metrics were summed and scaled from 0-100 to produce the final plant TP IBI. The response of the final IBI to HDS was then assessed using linear regression for both the development and validation datasets.

Additional analyses were performed to facilitate the use of the final TP IBI as a numerical wetland assessment and estimate and evaluate IBI and component metric precision. A preliminary wetland impairment threshold was determined from all available TP ecoregion data. Following previously established MPCA methods for determining wetland IBI assessment standards (Genet et al. 2004); the lowest IBI score among reference sites was used as the impairment threshold. This included averaged IBI scores if a reference site had more than a single sample in a given year. The MPCA defines reference wetlands as sites that have an HDS of 30 or less, with no single HDS category greater than a moderate score. Metric and IBI precision was evaluated with all data from the TP sites sampled in 2004. This included 2002-03 data from those sites. These sites ranged from reference (least-impacted) to heavily impacted. Metric and IBI precision was evaluated with signal to noise ratios (signal:noise) when all variance components were considered simultaneously (sampling error, within site, and inter- annual variation). Signal:noise are the between site variance (signal) divided by the within site variance (noise); the greater the signal:noise the greater the ability of the indicator to detect change due to anthropogenic impacts in the face of variance introduced from sampling error and natural variability. ANOVA was performed in SYSTAT® Version 10.2 to estimate variance. The IBI error variance was also used to compute 90% confidence intervals to incorporate quantified uncertainty into the wetland assessment process.

Development and Validation of Temperate Prairie Wetland IBIs 57

Finally, several alternative sampling schemes were assessed based on the small (5m x 5m) plot data collected in 2004. This was done in response to a conclusion from a statistical consultant that single large plot vegetation sampling was not adequately characterizing vegetation to detect depressional wetland changes due to anthropogenic impacts. Within wetland variability was consistently the greatest variance component in the NCHF IBI and sampling with sets of small plots, as opposed to the single large (10m x 10m) plot sample, approximately doubled NCHF IBI precision (Genet et al. 2005). The signal:noise and size of 90% confidence intervals for 4 sampling schemes were compared when only within site variability was considered: single large plot, pooled set of 4 small plots, single small plot, and averaged set of 4 small plots. These scenarios differ in their number of replicates (i.e., samples or subsamples) and how the IBI is computed (data are combined before computation vs. IBI scores averaged after computation). The single large plot is the existing MPCA plant sampling method, where a single large plot is the primary sample. The pooled set of 4 small plots is where data from a set of 4 small plots are combined prior to IBI calculation and the combined data are the primary sample. The single small plot is just considering one small plot as the primary sample and each set of 4 small plots represents 4 replicate primary samples. For the averaged set of small plots, IBI scores are computed for each small plot and the primary sample is the average of the set. In this scenario, each small plot is a subsample. For the sake of the comparison, it was necessary to derive separate continuous scoring cuts for the pooled set of small plots and for individual small plots as the species-area effect ruled out the possibility of using large plot metric scoring cuts for data collected from a smaller sampling area.

RESULTS & DISCUSSION

Macroinvertebrate IBI Validation

A total of 39 wetlands were sampled for macroinvertebrates in 2003 as part of the probabilistic survey of depressional wetlands in the Redwood watershed (Table II-1). Once the seasonal wetlands were excluded from the validation data set it was apparent that the 2003 Redwood study sites did not provide an adequate human disturbance gradient for testing the performance of the metrics (Figure II-1). While inclusion of the additional eight TP wetlands did not greatly improve the range of human disturbance represented within the data set, it did however allow for the inclusion of the only reference sites that were sampled in 2003 (Prairie Marsh & Kerk). Therefore, while not an entirely independent data set (e.g., eight sites were also part of the 2002 IBI development data set), it was decided that inclusion of the additional sites was required in order to provide a better disturbance gradient for testing the responsiveness of the metrics.

The majority of the preliminary metrics performed as well or better in the 2003 data set as they did in the 2002 development data set (Table II-2). Three metrics, ETSD, % Predator, and % Chironomidae, were not as sensitive to human disturbance in the 2003 data set compared to their stressor:response relationships observed in 2002 (Table I-7). An overall difference between the 2002 and 2003 data sets was the reduced number of significant correlations (P < 0.05) between

Development and Validation of Temperate Prairie Wetland IBIs 58

Table II-1. Temperate Prairies depressional wetlands sampled in 2003, indicating which sites had also been sampled in 2002 and which were part of the Redwood study. NWI Classification 2002 Redwood Site System/Class* Water Regime** Area (ha) HDS Site Site WillowLake PEM/PUB F 63.8 65 X RolhiksWMA PEM/PUB G 56.0 61 X X TylerWMA PEM/PUB G 41.6 58.5 X X Prairie Marsh PEM/PUB G 13.9 10 X BarryWMA PEM/PUB G 7.4 52 X GoldenWPA PEM/PUB F 22.2 53.5 X Hoffman PEM/PUB G 13.9 59.5 X Kerk PEM F 1.6 16 X Lee PEM/PUB F 10.9 54.5 X FrancoWMA PEM/PUB F 24.0 79 X 03linc019 PEM/PUB F 13.9 52.5 X 03linc018 PEM/PUB F 3.4 66 X 03linc004 PEM C 1.2 57.5 X 03lyon082 PEM/PUB F 198.2 45 X 03linc137 PEM C 2.5 44 X 03linc073 PEM F 3.7 63.5 X 03lyon070 PUB G 31.2 48.5 X 03murr028 PEM/PUB F 7.7 57.5 X 03lyon045 PEM C 0.2 63 X 03murr066 PEM/PUB F 28.7 54 X 03lyon080 PEM C 0.8 56.5 X 03lyon099 PUB G 9.7 52 X 03murr101 PEM F 2.6 57 X 03lyon124 PEM C 0.2 41 X 03murr132 PEM/PUB F 9.1 47.5 X 03pipe055 PEM F 2.9 63 X 03lyon146 PUB G 16.0 61 X 03redw008 PEM F 7.8 43.5 X 03redw123 PEM/PUB F 10.1 59.5 X 03redw094 PEM/PUB G 13.6 54.5 X 03redw048 PEM C 0.5 69.5 X 03lyon012 PEM F 2.1 56.5 X 03linc138 PEM F 0.8 75.5 X 03linc122 PEM F 5.4 67 X 03lyon052 PEM C 0.7 62.5 X 03lyon006 PEM C 0.2 17.5 X 03linc093 PEM F 7.5 73 X 03linc007 PEM C 0.6 48 X 03linc003 PEM C 1.4 64.5 X 03lyon110 PEM F 4.1 41 X 03lyon022 PEM C 0.1 32 X 03lyon058 PEM C 0.4 47.5 X 03linc097 PEM/PUB F 14.3 62 X 03linc089 PEM/PUB G 9.2 52 X 03lyon084 PEM/PAB F 3.1 38 X 03lyon142 PEM C 3.5 64 X 03lyon140 PEM C 0.2 23.5 X * PEM = Palustrine Emergent; PUB = Palustrine Unconsolidated Bottom ** C = Seasonally Flooded; F = Semi-permanently Flooded; G = Intermittently Exposed

Development and Validation of Temperate Prairie Wetland IBIs 59

20 20 A B

15 15

10 10 Count

5 5

0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS

Figure II-1. Distribution of human disturbance ratings (HDS) in A) the 2003 Redwood wetlands and in B) all of the 2003 TP wetlands. metrics and sediment chemistry parameters in the 2003 data set. This may be the result of the reduced disturbance gradient represented in the 2003 data set. Also, unlike the 2002 data set, a number of the metrics were significantly correlated with chloride concentrations and turbidity (Table II-2).

As a result of the observed relationships between the preliminary metrics and human disturbance modifications were made to the preliminary TP IBI. Two preliminary metrics were combined into one and one metric was not retained. Thus, the final macroinvertebrate IBI for depressional wetlands in the TP ecoregion was reduced to eight metrics from the original ten that were proposed in the preliminary IBI. Rationale for the modifications is given below.

The ETSD was one of the strongest metrics in the 2002 development data set and in the NCHF data sets where it was originally developed. However, in 2003 this metric did not exhibit many indicator qualities, especially against reliable measures of disturbance such as HDS, chloride, nitrogen, phosphorus, and turbidity (Table II-2). Another metric, Odonata Taxa richness, performed well in 2003 but did not respond to many measures of human disturbance in 2002 (Table I-7). A combination of these two metrics, ETO (combined taxa richness of Ephemeroptera, Trichoptera, & Odonata), exhibited exceptional indicator qualities in both years (Table II-3) and therefore was used as a substitute for the ETSD and Odonata taxa richness metrics in the final version of the IBI.

The lack of significant correlations in 2003 combined with the relatively few significant correlations observed in 2002 resulted in the exclusion of the % Predator metric from the final version of the IBI. Another metric that did not perform well in the 2003 data set was % Chironomidae. However, the strength of this metric in 2002 (Table I-7) combined with a

Development and Validation of Temperate Prairie Wetland IBIs 60

Table II-2. Pearson correlation coefficients (r) between macroinvertebrate metrics from preliminary TP IBI and measures of human disturbance for 2003 study sites. Water and sediment chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10).

. xa a op Ta t Pr t a Taxa Tax ran ae ae ran t a dator e id d D e na r l Taxa e e xi S l ri o do ipte ota Tol Pl o ET Int O D T % Chironomidae % Pr % % C

HDS ns -0.495 -0.411 -0.359 -0.351 0.337* ns 0.422 -0.368 0.307* Water Chemistry: Calcium Carbonate (mg/L) -0.385 -0.383 ns -0.300* ns 0.337* ns 0.402 -0.359 ns Chloride (mg/L) ns -0.521 -0.544 -0.418 -0.304* ns ns ns ns 0.424 Kjeldahl Nitrogen (mg/L) ns -0.408 -0.486 -0.352 ns ns ns ns -0.326* ns Sulfate (mg/L) -0.309* -0.499 ns -0.470 ns ns ns 0.349 -0.575 0.382 Phosphorus (mg/L) ns -0.419 ns ns ns ns ns ns ns ns Turbidity (NTU) ns ns -0.424 ns -0.532 -0.326* 0.344* ns ns ns Conductivity (u S/cm) -0.309* -0.518 ns -0.455 ns ns ns 0.303* -0.453 0.380 Sediment Chemistry: Aluminum (ug/g) ns ns ns ns ns ns ns ns ns ns Boron (ug/g) ns -0.332* ns ns ns 0.403 ns 0.297* -0.365 ns Calcium (mg/g) ns ns ns ns ns ns ns ns ns ns Chromium (ug/g) ns -0.316* ns ns ns ns ns ns ns ns Copper (ug/g) ns -0.389 ns ns ns 0.332* ns ns ns ns Potassium (mg/g) ns -0.352 ns ns ns ns ns ns -0.297* ns Magnesium (mg/g) ns -0.609 -0.421 -0.496 ns ns ns ns ns 0.326* Sodium (mg/g) ns ns ns -0.358 ns ns ns ns ns 0.341* Nickel (ug/g) ns -0.493 ns ns ns ns ns ns ns ns Sulfur (mg/g) ns ns ns ns ns ns ns ns ns ns Selenium (ug/g) ns ns ns ns ns 0.506 ns 0.421 -0.330* ns Strontium (ug/g) ns ns ns ns ns 0.308* ns 0.349 -0.405 ns Vanadium (ug/g) ns -0.369 ns ns ns 0.360 ns ns -0.305* ns

Development and Validation of Temperate Prairie Wetland IBIs 61

marginally significant correlation with HDS in 2003 (Table II-2) warranted the retention of this metric in the final IBI.

Diptera and Chironomidae taxa richness were both good indicators of condition in the 2003 data set. In the 2002 data set, however, Diptera taxa richness was more sensitive to human disturbance. Thus, Diptera taxa richness was selected over Chironomidae taxa richness as a metric for the final version of the TP macroinvertebrate IBI.

The relationships observed in the 2003 data set reinforce the previous findings in 2002 that pleids are sensitive to human disturbance in semi-permanent and permanent depressional wetlands of the TP ecoregion. Similar to the results in 2002, the correlation between % Pleidae and sulfate concentrations in the water column was one of the strongest relationships in the 2003 data set (Tables I-7 & II-2). Given these findings, % Pleidae was retained in the final set of metrics comprising the IBI.

Table II-3. Pearson correlation coefficients (r) for the relationship between selected metrics and human disturbance measures. Water chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10). Kjeldahl Total

Year/Metric HDS CaCO3 Chloride Nitrogen Phosphorus Sulfate Turbidity # sig 2002 ETO -0.490 -0.489 ns -0.484 -0.339* -0.486 ns 5 ETSD -0.402 -0.565 ns -0.455 ns -0.576 ns 4 Odonata Taxa -0.423 ns ns -0.376* ns ns ns 2 2003 ETO -0.366 -0.400 -0.419 ns ns -0.412 ns 4 ETSD ns -0.385 ns ns ns -0.309* ns 2 Odonata Taxa -0.411 ns -0.544 -0.486 ns ns -0.424 4

Characteristics of Final Temperate Prairies Macroinvertebrate IBI

The final macroinvertebrate IBI for assessing the condition of semi-permanent and permanent depressional wetlands in the TP ecoregion is comprised of eight metrics (Table II-4). Four of the eight metrics in this IBI are also components of the NCHF depressional wetland macroinvertebrate IBI. The range of IBI scores was similar between the two years, ranging from 17 to 73 in 2002 and 11 to 69 in 2003 out of a maximum possible range of 0 to 80 (Appendix G). The lowest IBI score among the reference sites, which the MPCA has defined as the impairment threshold for biological assessments (Genet et al. 2004, MPCA 2004), was 55 at the Prairie Marsh site. The final IBI exhibited a significant correlation with HDS in both data sets (Figure II-2). In addition, IBI scores were significantly correlated to a number of water chemistry parameters in both years, exhibiting the strongest relationship with the concentration of sulfate in the water column (Table II-5). IBI scores were not significantly correlated with wetland area (Log10 transformed to correct for non-normality) in neither 2002 (r = -0.205, P = 0.306) nor 2003 (r = -0.257, P = 0.150).

Development and Validation of Temperate Prairie Wetland IBIs 62

Table II-4. Component metrics of a macroinvertebrate IBI for the TP depressional wetlands, indicating which portions of the sampling method are used to derive each, whether it was included in the NCHF IBI, and its observed response to human disturbance. Sampling NCHF Disturbance Metric Method Definition metric Response

dip net & Taxa richness of Ephemeroptera, Trichoptera, and ETO no decrease activity trap Odonata dip net & Richness of intolerant taxa (determined empirically; Intolerant Taxa yes decrease activity trap see Appendix E) dip net & Diptera Taxa Taxa richness of Diptera no decrease activity trap dip net & Total taxa richness (most groups identified to genus, Total Taxa yes decrease activity trap Hirudinea and Gastropoda identified to species). Abundance of Chironomidae divided by total % Chironomidae dip net no increase abundance of sample Abundance of tolerant taxa divided by total % Tolerant dip net abundance of sample (determined empirically; see yes increase Appendix E) Abundance of Pleidae divided by abundance of % Pleidae dip net no decrease Hemiptera Corixidae Abundance of Corixidae divided by total abundance activity trap yes increase Proportion of Hemiptera and Coleoptera

80 80 2002 2003

60 60

40 40 Invert IBI Invert

20 20 r = -0.604 r = -0.540 p = 0.001 p = 0.001

0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS

Figure II-2. Relationships between TP macroinvertebrate IBI and human disturbance (HDS) for the 2002 and 2003 data sets.

Development and Validation of Temperate Prairie Wetland IBIs 63

In 2002 selection of wetland study sites was targeted in order to obtain group wetlands that represented the range of human disturbance, from least-impacted, reference sites to severely degraded. In 2003, site selection for the Redwood study was random. The different manner in which sites were selected in 2002 and 2003 is reflected in the different distributions of both IBI scores and HDS between these two years (Figure II-3). These differences are associated with the goals and objectives for the collection of data in each year. For instance, in order to obtain a data set ideal for developing indicators, an ideal set of study sites would have an equal distribution among each of the bins in the histogram. Site selection in 2002 attempted to do this by targeting wetland sites, but without having HDS calculated prior to the sampling season, it was impossible to determine whether or not we had obtained an equal distribution. While the distribution of HDS in 2002 is close to an equal distribution, it is clear that improvements in the targeted selection process can be made. Therefore, during 2004 site reconnaissance further attempts were made to better characterize the amount of human disturbance affecting a wetland prior to the final selection of sites. For example, a field version of the HDS (minus the chemistry factor) was completed during the initial site reconnaissance visit. These preliminary disturbance ratings provided a preliminary glimpse of the distribution of sites along the human disturbance gradient prior to selecting the final set of study sites, and better insured that an equal distribution was obtained.

Table II-5. Pearson correlation coefficients (r) and corresponding P-values for the relationship between IBI scores and water chemistry parameters (Log10) in 2002 and 2003. 2002 2003 Water Chemistry r P r P

Calcium Carbonate (mg/L) -0.624 0.001 -0.415 0.016 Chloride (mg/L) -0.324 0.100 -0.367 0.035 Kjeldahl Nitrogen (mg/L) -0.597 0.001 -0.340 0.053 Total Phosphorus (mg/L) -0.107 0.597 -0.292 0.099 Sulfate (mg/L) -0.732 0.000 -0.546 0.001 Turbidity (NTU) -0.016 0.937 -0.188 0.339 Conductivity (u S/cm) -0.631 0.000 -0.495 0.003

The 2003 distributions represent a random sample of the population of depressional (semi- permanent & permanent) wetlands in the Redwood watershed. Therefore, an equal distribution was not the goal nor was it expected given the manner in which sites were selected. Both the IBI and HDS exhibit distributions that approach normality (Figure II-3). Since these distributions are the result of a random sample, results (e.g., biological assessments, water chemistry parameters, etc.) can be extrapolated to the entire population of wetlands, a topic which is covered in more detail in Genet (2006).

With the collection of additional replicate data at wetland sites that were also sampled in 2002, further determination of the precision on the IBI and its component metrics was possible. The previous analysis based solely on 2002 replicate data (see Figure I-9) only incorporated variability due to measurement error which could be caused by factors such as time of day, repeatability of the sampling protocol, and microhabitat of sampling locations. The model based on 2002 and 2003 data included measurement error as well as year-to-year differences; thus,

Development and Validation of Temperate Prairie Wetland IBIs 64

20 30

15 Targeted (2002) Targeted (2002) 20 10 Count Count 10 5

0 0

5 10 10

Count Count 20 15 Random (2003) Random (2003) 20 30 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 Invert IBI Score HDS

Figure II-3. Comparison of the distribution of IBI scores and HDS among study sites when site selection was targeted in order to obtain a sample of wetlands with conditions spanning the range of human disturbance (2002) versus when site selection was random (2003).

1.4 1.2 1.1 1.8 1.6 17.8 1.3 0.2 4.4 100%

80%

60%

40% % of total% variance 20% Noise Signal 0%

O a a I T ant ax ant IB idae er Chiro E er T e ler pt % tol al Di ot Pl Corixidae In T %To Figure II-4. Relative variance estimates for the TP macroinvertebrate IBI and its component metrics, comparing the within site variance (Noise) to the between site variance (Signal). Signal:noise ratios are presented above the bars for each metric.

Development and Validation of Temperate Prairie Wetland IBIs 65

annual variability represented an additional source of variability. Given this additional source of variability, it is not surprising that the precision of the metrics and the overall IBI decreased when 2003 replicate data was incorporated into the model (Figure II-4). In other words, the signal:noise ratios decreased, indicating that within-site variability increased relative to between site variability. The only metric that exhibited an increase in precision in this model was % Pleidae (Figure II-4). All of the other metrics did not satisfy the minimum signal:noise ratio criterion of 2.0 recommended by Fore (2003a). In fact, the Diptera taxa richness metric was very imprecise, with a greater proportion of the variability due to within-site factors as opposed to between-site factors (Figure II-4). However, this analysis was still based on a limited data set (5 sites x 3 reps each) and did not include another important source of variability, differences due to sampling location within the wetland. In 2004, the data necessary to evaluate all the major sources of variability (measurement, annual, location) within one model was obtained with the sampling of six sites at three separate locations within the wetland. Therefore, the final evaluation of the precision of this IBI and its metrics will be reserved until this new data set becomes available (e.g., samples processed and identified).

Applicability of Macroinvertebrate IBI in Seasonal Wetlands

The macroinvertebrate communities of TP wetlands sampled in 2003 appeared to be influenced by the hydrologic regime of the wetland basin. Nonmetric multidimensional scaling illustrated this by using water regime as a grouping variable in the plots (Figure II-5). Dimension 3 did not appear to provide any additional separation of the clusters and therefore was not included in subsequent plots. The two dimensional plot suggests a macroinvertebrate community gradient that is related to water regime. Wetlands with semi-permanent to permanent water regimes tended to ordinate on the right side of the two dimensional plot, while seasonal wetlands clustered towards the left side of the plot (Figure II-5). The amount of overlap observed between the water regime clusters and a wide distribution of seasonal wetlands along dimension 2 suggested that other factors were affecting macroinvertebrate community composition and structure in these wetlands. In an attempt to remove the influence of human disturbance, the plot was re-examined, including only sites with an HDS score less than 50. This modification resulted in a more isolated cluster of seasonal wetlands, with one outlying site (Figure II-6). These results corroborate the work of previous investigators that have demonstrated the differences in the macroinvertebrate assemblage between seasonal and semi- permanent/permanent prairie pothole wetlands (Cvancara 1983, Hanson and Swanson 1989, Kantrud et al. 1989, Euliss and Mushet 2004).

The stressor:response relationships observed in the semi-permanent/permanent depressional wetland data set (Tables II-2 & II-3) were not observed when data from the seasonal wetlands were analyzed (Table II-6). In fact, three out of the four significant correlations were in opposite directions than those observed previously in the semi-permanent/permanent wetland data set. The % Chironomidae metric was positively associated with disturbance in semi- permanent/permanent wetlands, but was negatively associated with disturbance (e.g., Cl conc.) in seasonal wetlands. This is not surprising given that midges have adapted several life-history strategies to cope with ephemeral habitats, thus making them one of the most abundant and diverse insects occupying such habitats (Batzer and Wissinger 1996). Therefore, chironomid

Development and Validation of Temperate Prairie Wetland IBIs 66

1

0 DIM2

-1

-2 -2 -1 0 1 2 DIM1 Water Regime C F G

Figure II-5. Nonmetric multidimensional scaling plots based on dip net macro- invertebrate data collected from TP depressional wetlands in 2003.

1 densities are often high in seasonal wetlands even in the absence of human disturbance. The discrepant response exhibited by the % Chironomidae metric, 0 depending on the water regime of the wetland, clearly illustrates the role of classification to reduce variability in DIM2 Water biological assessment. Regime -1 To illustrate some of the naturally C F occurring differences in the G macroinvertebrate community of seasonal and semi-permanent/permanent wetlands, -2 data were summarized from three -2 -1 0 1 2 DIM1 reference wetlands (2 seasonal, 1 intermittently exposed) all located within Figure II-6. Nonmetric multidimensional scaling close proximity to one another. Dipteran plot based on dip net macroinvertebrate data larvae (primarily chironomids) were the collected from Temperate Prairie depressional dominant macroinvertebrate taxon wetlands in 2003, including only sites with HDS less than 50.

Development and Validation of Temperate Prairie Wetland IBIs 67

Table II-6. Pearson correlation coefficients (r) for the relationship between metrics and human disturbance measures. Water chemistry data were Log10 transformed. * indicates marginally significant correlations (P < 0.10), ns = not significant (P > 0.10). Kjeldahl Total Metric HDS Chloride Nitrogen Phosphorus Sulfate Turbidity

ETO ns ns ns ns ns ns Intolerant Taxa ns -0.584 ns ns ns ns Diptera Taxa ns ns ns ns ns ns Total Taxa ns ns ns ns ns ns % Chironomidae ns -0.631 ns ns ns ns % Tolerant Taxa -0.469* -0.566 ns ns ns ns % Pleidae ns ns ns ns ns ns Corixid Proportion ns ns ns ns ns ns

03lyon140 (seasonal)

03lyon006 (seasonal)

Prairie Marsh (intermittently exposed) 0% 20% 40% 60% 80% 100% Relative Abundance (dipnet samples)

Amphipoda Other Crustaceans Gastropoda Bivalvia Diptera ETO Hemiptera Coleoptera

Figure II-7. Comparison of macroinvertebrate community at undisturbed seasonal and intermittently exposed prairie pothole wetlands (Lyon Co., MN) based on data collected from dip net samples. collected with the dip nets in the two seasonal reference wetlands, while Amphipoda was the dominant taxon in the intermittently exposed wetland (Figure II-7). A combination of desiccation-resistant immature stages and the excellent dispersal ability of adults result in chironomids often being the most abundant macroinvertebrate of ephemeral habitats (Delettre 1989). Amphipods were virtually absent from the seasonal wetlands which is likely due to their inability to survive dry periods in these wetlands and their reliance on other organisms (e.g., waterfowl) for dispersal into such habitats once the basin becomes inundated (Swanson 1984). Although not numerically dominant, the macrocrustacean community of the two seasonal

Development and Validation of Temperate Prairie Wetland IBIs 68

reference wetlands was primarily comprised of clam shrimp (Conchostraca) and tadpole shrimp (Notostraca), taxa that were not collected in the intermittently exposed reference site. The above differences are just a few of the examples highlighting the influence of hydrologic regime in structuring macroinvertebrate communities and the need for the development of distinct macroinvertebrate criteria for the assessment of seasonal wetlands. Further evidence supporting this notion can be found throughout the scientific literature, stemming largely from efforts to understand the feeding ecology of breeding North American waterfowl (summarized in Swanson and Duebbert 1989).

With the limited data set that was obtained in 2003, we were also able to gain some insight on the natural differences in the chemical environment of seasonal versus semi-permanent wetlands. Water chemistry data were collected from two seasonal and two semi-permanent reference wetlands in the Temperate Prairies ecoregion in 2003. Chemical parameters such as calcium carbonate, chloride, sulfate, and conductivity were similar between the two wetland types. Turbidity and pH were lower in the seasonal wetlands (Table II-7). The lower pH of the seasonal wetlands may simply be a result of the lower pH of the predominant source of water feeding these two wetlands: precipitation. However, the similarity of the conductivity values of all four reference wetlands would suggest other factors may be contributing as well (e.g., all appear to have similar hydrologic function). Without more information on the major constituents (e.g., algal vs. sediment) affecting water clarity at these sites, it is difficult to account for the observed differences in turbidity. The most striking difference between these two wetland types was the higher nutrient concentrations (N and P) in the seasonal wetlands (Table II-7). LaBaugh et al. (1987) suggests that high N and P concentrations of ephemeral wetlands (seasonal and temporary) are due to the remineralization of nutrients during dry periods that is associated with the decomposition of plant material and their subsequent release from the sediment as basins refill during the growing season. In fact, Detenbeck et al. (2002) found that nutrient dynamics in seasonal prairie pothole wetlands varied more in response to hydrologic regime and vegetation structure than to increased nutrient inputs from surface runoff. Thus, reference seasonal wetlands may often have nutrient concentrations that are characteristic of disturbed semi- permanent wetlands. Therefore, it appears that the distinct hydrologic regime and associated characteristics of seasonal wetlands, resulting in macroinvertebrate communities that are substantially different from those observed in semi-permanent/permanent wetlands, precludes development of a single macroinvertebrate index that would accurately assess both wetland types.

Preliminary Macroinvertebrate Indicators of Condition in Seasonal Wetlands

Given the observed differences in the structure and composition of macroinvertebrate assemblages between seasonal and more permanently flooded wetlands, it is evident that a new suite of indicators needs to be developed in order to accurately assess the condition of seasonal wetlands with macroinvertebrates. A total of 14 seasonally inundated depressional wetlands were sampled in the TP ecoregion in 2003 (Table II-1). While this is an insufficient number of sites with which to develop an IBI with, this data set can provide useful insight into potential metrics for assessing seasonal wetlands which can later be expanded upon

Development and Validation of Temperate Prairie Wetland IBIs 69

Table II-7. Water chemistry characteristics of reference seasonal (C) and semi-permanent (F&G) TP wetlands.

Kjeldahl Total Water Area CaCO3 Chloride Nitrogen Phosphorus Sulfate Turbidity Conductivity pH Site Regime (ha) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (NTU) (uS/cm)

Prairie Marsh G 13.93 74 5.5 1.68 0.057 11 4.8 317 9.00 Kerk F 1.62 64 1.2 1.47 0.087 5 2.8 275 7.74 03lyon006 C 0.21 83 1.5 3.16 0.569 5 1.5 290 6.73 03lyon140 C 0.22 59 1.8 2.65 0.205 5 1.3 175 7.30

Table II-8. Pearson correlation coefficients (r) for the relationship between metrics and human disturbance measures. Water chemistry data were Log10 transformed. Bold values = significant correlations (P < 0.05); * indicates marginally significant correlations (P < 0.10).

Kjeldahl Total † Metric HDS CaCO3 Chloride Nitrogen Phosphorus Sulfate Turbidity Conductivity pH

% Crustacea BT -0.071 -0.675 -0.187 0.559 0.604 -0.56 0.633 -0.597 -0.512* % Crustacea DN 0.301 -0.462* -0.023 0.424 0.525* -0.340 0.750 -0.430 -0.454 % Hemiptera DN 0.546 0.094 0.325 -0.030 0.048 0.203 0.287 0.083 0.241 % Insect BT 0.429 0.881 0.548 -0.539 -0.258 0.798 -0.418 0.933 0.476* % Noninsect BT -0.110 -0.679 -0.167 0.461* 0.477* -0.593 0.594 -0.598 -0.508* % Collector-Filterer DN -0.142 -0.621* -0.313 0.327 0.050 -0.563 0.442 -0.736 -0.357 % Tolerant DN -0.469* -0.505* -0.566 0.311 0.211 -0.363 -0.077 -0.482* -0.027 Intolerant Taxa Richness -0.276 -0.234 -0.584 0.147 0.068 -0.091 -0.030 -0.306 0.113

† BT = relative abundance based on bottle trap data; DN = relative abundance based on dip net data

Development and Validation of Temperate Prairie Wetland IBIs 70

with the collection and analysis of additional data. Visual examination of boxplots comparing the distribution of the five least impaired (according to HDS) and five most impaired wetland sites was used to identify potential candidate metrics (Appendix H). Pearson correlation analysis was also used to identify potential stressor:response relationships in the data, but there were very few statistically significant relationships with HDS. A few attributes were correlated with water chemistry parameters such as chloride, kjeldahl nitrogen, and total phosphorus concentrations. The general pattern observed was an attribute that was positively correlated with chloride, sulfate, and/or conductivity, was typically negatively correlated with nutrient concentrations and turbidity, and vice versa (Table II-8). This pattern is likely due to the generally high nutrient and low chloride and sulfate concentrations observed in the reference seasonal wetlands (Table II-7).

Based on this preliminary analysis, it appears that the chironomid community of undisturbed seasonal wetlands has a higher proportion of the subfamily Orthocladiinae and the tribe Tanytarsini, whereas the disturbed wetlands have a higher proportion of the tribe Chironomini (Appendix H). Total taxa richness, a reliable metric for assessing aquatic habitats, exhibited indicator qualities for assessing seasonal wetlands as well exhibiting separation of the interquartile ranges of least vs most disturbed sites (Appendix H). The metric comprised of the combined relative abundance of tolerant taxa, developed from the semi-permanent/permanent wetland data set, appears to decrease with disturbance in seasonal wetlands rather than increase as it does in more permanent wetlands. However, without a larger data set which includes more reference seasonal wetlands it is difficult to explain the observed response of the % Tolerant taxa metric in these wetlands.

Plant IBI Validation

The preliminary TP IBI performed at a much 100 lower level with the validation dataset compared to the development set (Figure II-8). Both the y- intercept and slope were lower, indicating a less 80 steep response along the HDS range. The IBI-

HDS relationship was significant (P < 0.05), I 60

B

2 I

though the strength of the response (r ) was much t

n

a reduced. IBI scores showed large variation at l P 40 moderate and even high levels of HDS. A number of sites had high IBI scores (IBI > 60) at β = 75.943 high HDS (HDS > 60), which likely contributed 20 0 β1 = -0.458 to the decreased IBI response. r2 = 0.121** 0 The reduced performance of the preliminary TP 0 20 40 60 80 100 IBI was a result of the reduced performance of the HDS component metrics. All 9 preliminary metrics had weaker responses to HDS (Figure II-9). Only 2 Figure II-8. Preliminary plant TP IBI- metrics (Perennial Richness and Tolerant Taxa HDS scatterplot with validation data. Ratio) had strong (P < 0.05) HDS relationships The y-intercept (β0), slope (β1), and coefficient of variation (r2) are included. ** = P < 0.05.

Development and Validation of Temperate Prairie Wetland IBIs 71

ABC

15 3 4 β = -0.014 β = -0.011 1 β1 = -0.012 1

s

s

12 s

s s 2 2 s 3 2 r = 0.10 e e r = 0.113** r = 0.074* e

n

n

n

h

h

2 h

c

c

c

i

i 9 i

R

R

R

l

d

d

i

l

a

i

i

o 2

u

n

n

i

n

G

e 6 m

c

r

i

a

t

e r 1

a

P

G

u

n

q

n 1

L

A 3 L

0 0 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS DEF 4 15 100

r

e

v

s

o

s β1 = -0.008 β = -0.015 C

e 1 β1 = 0.770

t

n 80

n

h 2 3 2 e c r = 0.077* 2

i r = 0.025 g r = 0.073*

r

R

t

10

e

n

a

m

r u 60

e

E

o

n

l

C

e 2 a

d

n

G l

i

o

i

r

u t 40

a

r

l

G

o u 5

p

c

o

s

r

a 1

P

V

20

a

n

h

L

p

y 0 0 T 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS

G HI 3 100 3

β1 = -0.07 β1 = 0.54 β1 = -0.006 80 2 2 2

r = 0.03 o r = 0.137** r = 0.037

i

y

a

t

t

i

x

a

s

a 2 2

r

R

T

e

a 60 v

i

e

x

v

D

i

a

t

i

T

n

s

t

o

n

n

n

e

a 40

n

r

S

a

e

1 l 1

h

n

o

L

S

T 20

0 0 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS

Figure II-9. Preliminary plant TP IBI metric-HDS scatterplots with validation data. The y- intercept slope (β1), and coefficient of variation (r2) are included. * = P < 0.1 ** = P < 0.05.

(Figure II-9C&H), and 3 (Graminoid and Vascular Genera Richness and TPE Cover) had weak (P < 0.10) HDS relationships (Figure II-9B,D,F). The remaining 4 metrics (Aquatic Guild Richness, Guild Count, Sensitive Taxa, and Shannon Diversity) did not have a significant response to HDS (Figure II-9A,E,G,I).

A series of potentially confounding factors were evaluated to determine if misclassification or inaccurate stressor accounting was factoring into the decreased metric performance. The most likely factor was the inclusion of study wetlands that had a seasonally flooded water regime (C) in the validation dataset. The preliminary TP IBI and component metrics were re-evaluated

Development and Validation of Temperate Prairie Wetland IBIs 72

against HDS when the dataset was broken into 9 seasonal and semipermanent/permanent (F, G, H) groups. Including seasonal wetlands had no 8

effect on metric-HDS relationships except for the s

s 7

e

Aquatic Guild Richness metric. When the n

h 6

c seasonal wetlands were removed from the dataset, i

R Aquatic Guild Richness had a significant response 5

d

l

i to HDS (Figure II-10). The seasonal wetlands u 4

G

had a restricted Aquatic Guild Richness range (0- c

i t 3 3) compared to the semipermanent/permanent a

u

q

wetlands (1-8). Including the seasonal wetlands A 2 clearly compromises the response of this metric. 1 Therefore, Aquatic Guild Richness should only be 0 applied in semipermanent/permanent depressional 0 20 40 60 80 100 wetlands. HDS Seasonal Continuing with the evaluation, factors that link Semipermanent/permanent landuses and resulting stressors were then assessed to determine if the HDS is appropriately Figure II-10. Aquatic Guild Richness- capturing anthropogenic impacts. Some HDS scatterplot with sites classified by disturbance factors may have greater potential to water regime (i.e., seasonal vs. impact wetlands than how they are accounted for semipermanent/permanent). in the HDS. For example, subsurface agricultural drainage (tiling) routed to wetlands increases the delivery of excess nutrients compared to sheet runoff from the immediate surroundings by increasing the effective drainage area to the wetland and creating a direct pathway for stressor exposure. Agricultural tiling is factored into the HDS, however, it is unknown whether this is an accurate portrayal of the effect tiling has on depressional wetlands. The majority of the sites in the validation dataset had a moderate HDS (Figure II-3). These sites had a large variation in preliminary TP IBI scores (Figure II-8). Sites that had an HDS of 45-65 were used to test whether there are factors adding excessive variability and clouding potential metric-HDS relationships because of inaccurate HDS scores. This was evaluated by inter-quartile separation of sites according to the following factors: surface water isolation, small size (≤ 2 ha), tiling/ditching, and a combination of all three factors. The reasoning behind this is that small isolated wetlands that may be surrounded by agricultural landuse may not be as impacted as initially thought because of a lack of an efficient delivery system for stressors. If a factor (or combination of factors) is having an unaccounted effect, it was expected that multiple metrics at more or less the same HDS (45-65) would have inter- quartile separation based on the factor. There were very few metric inter-quartile separations and none of these were consistent. Thus, the current classification system and the anthropogenic impacts quantified by the HDS apparently are not introducing systematic bias into the metric evaluation process.

It is likely that the large variation observed in the metric values is due to the limited distribution of sites along the HDS and the nature of the wetland plant community response to anthropogenic impacts. When assessing the response of a factor (y) against another (x), one would expect the greatest variability in y to occur at a moderate range of x, where a moderate exposure to x would

Development and Validation of Temperate Prairie Wetland IBIs 73

have a variable effect on y. There is an indication of this with the preliminary TP IBI and the development dataset where a site with an HDS of 53.5 had a preliminary IBI of 26.7 (Golden WPA) and site with an HDS of 56 had a preliminary TP IBI of 83.8 (Bryclyn). It is likely that it is the extremes of the disturbance gradient that drives the response of the wetland plant community, where reference sites have a high probability of having a diverse native community and highly impacted sites have a high probability of greatly reduced diversity often dominated by introduced species. Sites with moderate exposure to anthropogenic impacts can apparently respond with either extreme. The validation dataset is not evenly distributed along the HDS, with most sites having a moderate score (Figure II-1). The increased metric variability observed with the validation dataset should therefore be accepted with the knowledge that the dataset does not have an ideal distribution for testing metrics.

Once likely confounding factors were eliminated, the remaining 3 non-responsive metrics (Guild Count, Sensitive Taxa, and Shannon Diversity) were re-evaluated to determine if they could be adjusted to increase performance and be accepted as validated metrics. A handful of additional metrics that had previously not been thought of were also evaluated for incorporation into the final TP IBI during this effort.

Sensitive taxa metrics are core metrics in many IBIs, consistently showing strong responses to anthropogenic impacts (Karr and Chu 1999). The relatively weak response of the Sensitive Taxa metric in the development set and the non-response with the validation set was therefore somewhat surprising. Given the history of sensitive metrics in general and the performance of the Sensitive Taxa metric in the NCHF IBI in particular (Sensitive Taxa was the strongest metric in the NCHF IBI; Gernes and Helgen 2002) an attempt was made to improve the metric by redefining the sensitive taxa list for the TP ecoregion. Recall that the sensitive taxa list was originally developed during the NCHF IBI development phase and may be more specifically tailored for species in that region. Coefficient of Conservatism (C) values are numerical rankings of a plant species’ fidelity to remnant natural habitats (Swink and Wilhelm 1994) and thus can be used to identify and define plants that are sensitive to anthropogenic impacts. C- values have been successfully applied in defining sensitive and tolerant species in Ohio wetlands (Simon et al. 2001, Mack 2004). In addition, indices derived from C-values (i.e., mean-C and the Floristic Quality Index) have been found to perform very well as wetland condition indicators (Lopez and Fennessy 2002, Cohen et al. 2004, Bourdaghs et al. in press). C-values have not been developed for the Minnesota flora; however, they have been developed for the floras of Iowa (http://www.public.iastate.edu/~herbarium/coeffici.html) and the Dakotas (NGPFQAP 2001). The TP ecoregion occupies almost the entirety of Iowa and nearly half of the Dakotas. Thus, it is reasonable to assume that the affinity of plants to natural habitats in those states is very similar to the behavior of plants in the TP ecoregion of Minnesota, justifying the use of Iowa and Dakota C-values in redefining the Minnesota TP ecoregion sensitive species list. Species were determined to be sensitive if C ≥ 7 in either Iowa or the Dakotas.

Defining sensitive taxa by Iowa and Dakota C-values greatly improved the performance of the Sensitive Taxa metric (now termed C-value Sensitive Taxa or CST; Figure II-11). The CST metric had the strongest metric response in the validation set and the second strongest in the development data set. Therefore, CST will replace Sensitive Taxa in the final TP IBI.

Development and Validation of Temperate Prairie Wetland IBIs 74

A

4 100 200

β1 = -0.028 β1 = 0.812 β1 = 1.171 2 2 2 r = 0.463** 80 r = 0.52** r = 0.201** 3 150

r

o

e

i

t 60

T

v

a

o

S

R

C

C

2 100

T

n F

T

L 40 S

C

T

1 50 20

0 0 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS

B 4 100 200

β1 = -0.017 β1 = 0.46 β1 = 1.106 2 2 2 r = 0.157** 80 r = 0.085** r = 0.118** 3 150

r

o

i e

t 60

T

v

a

o

S

R

C

C

2 100

T

n F

T

L 40 S

C

T

1 50 20

0 0 0 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS HDS Figure II-11. Additional final plant IBI metric-HDS scatterplots derived from both the development (A) and validation dataset (B).

For consistency, it was decided to use Iowa and Dakota C-values to define the tolerant taxa species list. Like the original sensitive taxa list, the tolerant taxa list was developed during the NCHF IBI development phase and may be more specifically tailored for that ecoregion. Tolerant taxa were defined as species that have an Iowa or Dakota C-value ≤ 3 or are introduced to Minnesota. The metric name was subsequently changed to C-value Tolerant Taxa Ratio or CTT Ratio. Changing the metric decreased the performance of the metric in both the development and validation sets, though the CTT Ratio-HDS relationship was highly significant in both (Figure II-11).

It should be noted that the MPCA is currently developing C-values for the state’s wetland flora (USEPA Assistance Number CD 965454-01). This will facilitate the use of Floristic Quality indices in Minnesota as well provide a scientifically sound method to further define sensitive and tolerant species in the state. Minnesota C-values will be applied to the TP IBI as part of periodic IBI and assessment standard redevelopment efforts in the future.

The Shannon Diversity index not only did not have a significant HDS relationship with validation data (Figure II-9I), but was also the weakest metric with the development data (Figure

Development and Validation of Temperate Prairie Wetland IBIs 75

I-16I). An attempt was made to improve metric performance by only including native species in index calculations. This was done to reduce potential increased variability that could occur when total richness increases at moderate levels of HDS due to initial establishment of introduced species. The response of Native Shannon Diversity was increased with the validation data (r2 = 0.051); however, this result was not statistically significant. In addition, error is introduced into Shannon Diversity when all species in the community are not included in the sample (Peet 1974). Sample plots are located at the emergent/aquatic interface, essentially sampling both communities. Nested plot sampling in the NCHF ecoregion in 2001 showed that species-area curves typically begin to flatten out by 100m2 indicating that the sample plot size is capturing most of the species in both communities (unpublished data). Therefore, the error introduced into index calculations is likely small, though the impact that the error has on metric response to HDS is unknown. Given the weak response with the development data set, the non-response with validation data, and potential systematic error the Shannon Diversity index was eliminated as a metric in the final TP IBI.

Alteration of the Guild Count metric to improve HDS response proved problematic, as there did not appear to be a reasonable way to alter the metric while maintaining its conceptual integrity. For example, including only native guilds seemed inappropriate as the diversity of the physical plant community structure, whether it is produced from native or introduced species, is valuable for providing a diversity of habitats for wildlife. The Guild Count metric is an attractive metric as it adds a structural component to the IBI. Ideally, IBIs are developed in such a way to capture information and responses that incorporates individuals to landscapes (Karr and Chu 1999). This includes not only measures of species richness but measures of community structure. In addition, Guild Count was one of the strongest metrics with development data (Figure I-16E). Ultimately, it was decided that the structural dimension the metric brings to the IBI along with its performance with development data outweighed the non-response with validation data. Guild Count was thus retained in the final TP IBI unaltered.

Finally, the performance of a handful of additional metrics was assessed. This included metrics based on the combined cover of invasive Typha spp. (T. angustifolia and T. x glauca) and small free-floating aquatic spp. (Lemna L., Spirodela Schleid., Wolfia Horkel ex Schleid., Ricciocarpos Corda, and Riccia L. nom. cons.). As the dominance of invasive Typha may be indicative of wetland soil nutrient enrichment (Galatowitsch et al. 1999), the dominance of small floating aquatic species may be indicative of nutrient enrichment of the aquatic environment as they are free floating and uptake nutrients directly from the water column. Small floating aquatic metrics did not meet initial metric selection criteria, but it was often observed during sampling that many of the same sites that had a high abundance of Typha also had a high abundance of free floating aquatics. The sum of the midpoint % cover of invasive Typha and small floating aquatic spp. was found to have a stronger response to HDS than TPE Cover in both datasets (Figure II-11). Therefore, it was decided to replace TPE Cover with Typha/Small Floating Aquatic spp. (TSF) Cover.

Development and Validation of Temperate Prairie Wetland IBIs 76

Table II-9. Final Temperate Prairies plant IBI metric descriptions.

NCHF Disturbance Metric Definition Metric Response

Taxa Richness

Aquatic Guild Richness† Number of native aquatic plant species. Yes Decrease

Graminoid Richness‡ Number of native wetland graminoid species. Yes Decrease

Perennial Richness‡ Number of native wetland perennial species. Yes Decrease

Vascular Genera Richness‡ Number of vascular genera. Yes Decrease

Community Structure Guild Count Number of distinct plant guilds. No Decrease Cover of invasive Typha spp. and small floating Typha/ Small Floating aquatics (Lemna , Spirodela , Wolfia , Riccia , No Increase Aquatic spp. (TSF) Cover Ricciocarpos spp.).

Sensitive & Tolerant Taxa Number of taxa sensitive to disturbance, defined by C -Value Sensitive Taxa Iowa and N. & S. Dakota Coefficient of No Decrease (CST)‡ Conservatism (C ) values (C ≥ 7). Number of disturbance tolerant taxa divided by the C -Value Tolerant Taxa Ratio total taxa richness. Tolerant taxa defined by Iowa No Increase (CTT) and N. & S. Dakota Coefficient of Conservatism (C ) values (C ≤ 3) or is introduced. † Metric not applied to depressional wetlands with a temporary or seasonal water regime ‡ Metric is Natural Log transformed

Characteristics of the Final Temperate Prairies Plant IBI

Unlike the final macroinvertebrate TP IBI, the final plant TP IBI can be applied to at least select seasonal water regime wetlands. Water regime had an effect only on the Aquatic Guild Richness metric (Figure II-10). In other words, with the exception of Aquatic Guild Richness, all of the other metrics responded well to anthropogenic impacts when seasonal wetlands were considered together with semipermenent/permanent wetlands. Thus, Aquatic Guild Richness should only be applied in semipermanent/permanent wetlands and the final TP IBI will have two forms: the full 8 metric IBI applied to semipermanent/permanent depressional wetlands and a 7 metric IBI applied to seasonal depressional wetlands (Table II-9). There is one caveat with applying the IBI to seasonal depressional wetlands and that is the wetlands must have a marsh plant community with at least small open water pockets as opposed to a wet meadow type of community. Applying the IBI to seasonal wetlands without this type of community would represent a break in the standard sampling protocol, where sampling plots are located at the

Development and Validation of Temperate Prairie Wetland IBIs 77

Table II-10. Final plant IBI scores for both the development and validation data sets. Site names appearing in boldface are reference sites.

Development Set SiteName Final IBI HDS SiteName Final IBI HDS

BarryWMA 52.5 52 LoneTreeWMA 27.5 77 Bryclyn 79.4 56 LyonsWMA 21.9 66.5 Carex 89.2 26.5 Malta 20.6 58 EastlickMarsh 61.6 40 Manchester 80.3 32.5 FrancoWMA 13.6 79 Milan 52.9 64 FurgameWMA 70.7 51.5 OakGlenEast 72.0 28.5 GoldenWPA 32.9 53.5 OakGlenWest 98.3 20 GreatOasisWMA 65.5 24 Prairie Marsh 88.0 10 Hancock 77.5 49 RolhiksWMA 39.0 61 Hoffman 37.5 59.5 RostWMA 44.3 55.5 Kerk 72.9 16 TylerWMA 54.0 58.5 LakeCharlotte 23.0 45 WillowLake 62.8 65 LakeElisabeth 54.1 21 Yohi 72.9 37.5 Lee 40.9 54.5

Validation Set SiteName Final IBI HDS SiteName Final IBI HDS 03linc003 70.3 64.5 03lyon110 40.1 41 03linc004 48.7 57.5 03lyon124 97.2 41 03linc007 34.6 48 03lyon140 83.6 23.5 03linc018 54.8 66 03lyon142 30.7 64 03linc019 16.1 52.5 03lyon146 64.0 61 03linc073 50.2 63.5 03murr028 41.7 57.5 03linc089 33.6 52 03murr066 30.7 54 03linc093 48.1 73 03murr101 44.1 57 03linc097 36.4 62 03murr132 59.5 47.5 03linc122 61.9 67 03pipe055 41.2 63 03linc125 40.5 61 03redw008 52.6 43.5 03linc137 78.7 44 03redw048 62.1 69.5 03linc138 82.9 75.5 03redw094 59.9 54.5 03lyon006 80.7 17.5 03redw123 36.4 59.5 03lyon012 41.9 56.5 BarryWMA 51.4 52 03lyon022 37.0 32 FrancoWMA 15.8 79 03lyon045 19.5 63 GoldenWPA 31.5 53.5 03lyon052 73.1 62.5 Hoffman 46.1 59.5 03lyon058 52.8 47.5 Kerk 80.8 16 03lyon070 59.6 48.5 Lee 38.9 54.5 03lyon080 62.6 56.5 Prairie Marsh 78.5 10 03lyon082 76.8 45 RolhiksWMA 29.4 61 03lyon084 81.6 38 TylerWMA 75.0 58.5 03lyon099 20.9 52 WillowLake 63.5 65

Development and Validation of Temperate Prairie Wetland IBIs 78

emergent/aquatic interface, and applying the IBI to a different community type will add a confounding factor that could produce misleading results.

The metrics for both IBI forms were scored continuously according to the formulas in Figure I-2. With the exception of the Aquatic Guild Richness metric, all TP ecoregion data from 2002-04 were used to derive scoring cuts. Aquatic Guild Richness was scored with 2002-04 data when seasonal water regime wetlands were removed. As with the preliminary TP IBI, following summation of component metric scores both IBI forms were scaled from 0-100. The scaling factors were 10/8 = 1.25 for the semipermanent/permanent IBI form and 10/7 = 1.43 for the seasonal IBI respectively. After both IBI forms were scaled there was no discernable difference between seasonal and semipermanent/permanent scores. Therefore, they can be considered to be equivalent.

Overall, the validation process produced a more robust IBI, as metric performance was maximized when both datasets were evaluated simultaneously. The final TP IBI scores ranged from 13.6-98.3 in the development set and 15.8-97.2 in the validation set (Table II-10). There was a slight decrease in final IBI performance with the development set (Figure II-12A) compared to the preliminary IBI (Figure I-17). Conversely, the final IBI showed a performance increase with validation data (Figure II-12B). Thus, there was a slight sacrifice in performance with the development data in order to achieve an acceptable level of performance in the validation set, thereby producing an IBI that has greater ability to detect wetland quality changes in all depressional wetlands in the TP ecoregion

Several water and sediment chemistry parameters were correlated (P < 0.10) with the final TP IBI (Table II-11). These included chloride in the water column and copper and nickel sediment concentrations. There were no significant correlations with nutrient parameters and any of the component metrics, with the exception of a negative correlation between sediment phosphorus and CTT Ratio which was expected to be positive. The significance of the chemistry results is unclear at this time given the amount of available data and other correlated anthropogenic impacts that have similar effects on wetland plant communities.

Preliminary Impairment Threshold The MPCA has developed a process to determine numerical IBI standards for assessing Aquatic Life Use Support (ALUS) of depressional wetlands under the Federal Clean Water Act (Genet et al. 2004). This process was rooted in a 2002 revision of the state’s water quality standards where narrative guidance was developed for using biological data for assessing ALUS for all waters of the state, which includes wetlands (MN Rules Ch. 7050). A key component of this process is determining the IBI score that separates support or non-support of aquatic life. The MPCA uses the lowest reference site score as the impairment threshold. This includes averaged scores if a reference site has more than a single sample in a given year.

The lowest scoring reference site with the TP IBI was GreatOasisWMA at 65.5 (Table II-10). Sites that score below this value can be assessed as not supporting Aquatic Life, or impaired for the ALUS depressional wetland plant quality standard. In 2005 however, it was determined that the set of reference sites in the TP ecoregion was limited and more reference data should be

Development and Validation of Temperate Prairie Wetland IBIs 79

AB 100 100

β0 = 97.574

β1 = -0.895 80 r2 = 0.512** 80

I 60 I 60

B B

I I

t t

n n

a a

l l

P 40 P 40

β0 = 81.997 20 20 β1 = -0.556 r2 = 0.162**

0 0 0 20 40 60 80 100 0 20 40 60 80 100 HDS HDS

Figure II-12. Scatterplots of the final TP Plant IBI derived from both the development (A) and validation (B) datasets.

Table II-11. Final TP IBI and metric Pearson correlation coefficients (r) with selected water and sediment chemistry parameters from validation data. All of the chemistry data were Log10 transformed prior to the analysis. Results are displayed as: ns = not significant, * = (P < 0.1), and ** = (P < 0.05).

Water Chemistry (N = 47) Sediment Chemistry (N = 47) Cl N P Cl N P Cu Ni Zn

Plant IBI -0.491** ns ns ns ns ns -0.282* -0.420** ns Aquatic Guild ns ns ns ns ns ns ns ns ns Richness† Ln Graminoid -0.423** ns ns ns ns ns -0.277* -0.280* -0.300* Richness Ln Perennial Richness -0.330** ns ns ns ns ns -0.382** -0.451** -0.322** Vascular Genera -0.311** ns ns ns ns ns -0.323** -0.372** -0.257* Richness Guild Count ns ns ns ns ns ns ns -0.302** ns

TSF Cover 0.530* ns ns ns ns ns ns 0.244* ns

CST -0.366** ns ns ns ns ns -0.320** -0.471** ns

CTT 0.554** ns ns ns ns -0.372**‡ ns ns ns † Correlations performed with sites with permanent or semipermanent water regimes (N = 33) ‡ Correlation oposite of expected direction

Development and Validation of Temperate Prairie Wetland IBIs 80

collected before finalizing impairment thresholds. Therefore, we are presenting this impairment threshold as a preliminary standard and will continue to evaluate the assessment thresholds for both plant and macroinvertebrate IBIs in 2006-07 with additional reference site data collected in 2005.

IBI and Metric Precision The majority of the metrics, and thus the final TP IBI, had a relatively high level of precision. A commonly accepted guideline for evaluating metric and IBI precision is a signal:noise > 2 (Fore 2003a). Seven of the eight metrics exceeded this guideline (Figure II-13). The IBI had a greater signal:noise than any of the component metrics, indicating a possible integrative performance effect of the IBI, where the IBI performs at a level greater than the sum of its parts. Compared to signal:noise based on 2002 sampling error replicates (Figure I-19), IBI signal:noise presented here was much lower. This was expected given the additional variance components that were considered in the final analysis (within site and inter-annual variability).

The overall IBI variance estimate (when considering sampling error, within site and inter annual variability) was then used to compute 90% confidence intervals and the number of detectable condition categories per number of replicate samples. Standard errors, which were computed from the ANOVA error variance, were multiplied by the area under the normal curve (Z0.10/2) to derive the confidence limits. The 90% confidence intervals were then a doubling of the confidence limits. The 90% confidence limit surrounding final TP IBI scores based on standard sampling effort (N = 1) was +/- 13.85 (Table II-12). In other words, based on the measured variability due to the variance components, we can be 90% confident that the actual IBI score is within +/- 13.85 points of the score returned from a singe plot sample. Confidence limits decrease as replicate samples are gathered at a site. The number of condition categories that the final TP IBI can reasonably detect given the error variance is the overall range of the IBI divided by the confidence interval. For example, based on a single replicate the IBI can detect 100/27.7 = 3.6 condition categories. The MPCA has a condition category detection goal of 3 (Genet et al. 2005). Sampling with a single replicate meets this goal.

The confidence intervals are also incorporated into the depressional wetland ALUS assessment process. This is done to acknowledge the uncertainty inherent in sampling biological communities and how this affects the ALUS assessments. Confidence intervals are incorporated into the assessment process by placing the intervals around the impairment threshold (Figure II- 14). Sites that score above the upper confidence limit are considered to be fully supporting Aquatic Life. Sites that score below the lower confidence limit are considered to be non- supporting of Aquatic Life. Sites that score within the 90% confidence interval are further evaluated in detail by a best professional judgment panel to complete the ALUS assessment (Genet et al. 2004). It should again be clearly noted that the impairment threshold reported here will not be used for depressional wetland ALUS assessment at this time as it is preliminary. The final impairment threshold for the TP plant IBI will be determined in 2006-07.

Development and Validation of Temperate Prairie Wetland IBIs 81

0.3 4.0 3.8 3.8 2.3 4.1 4.0 3.5 6.8 100

80

60

40 % of% total variance 20 Noise Signal 0 s s s t e n BI n t I h CST Ratio n ichness la Ric TT P TSF Cover C Guild Cou

raminoid n G Aquatic GuildL Richness Ln Perennial R

Ln Vascular Genera Richnes

Figure II-13. Relative between site (signal) and within site (noise) variance estimates for the final TP Plant IBI and component metrics (n = 5). Signal to noise ratios are given above the bars for each metric and the IBI.

Table II-12. Confidence limits, confidence intervals, and number of condition categories the TP plant IBI can detect according to the number of replicate samples (N). N = 1 N = 2 N = 3 N = 4 N = 5 90% Confidence 13.85 9.79 8.00 6.93 6.19 Limit 90% Confidence 27.70 19.59 15.99 13.85 12.39 Interval # of Condition 3.6 5.1 6.3 7.2 8.1 Categories

Plant Sampling Methods Evaluation

Small multiple plot sampling only marginally improved the precision of the plant TP IBI (Table II-13). Only one of the small plot scenarios (Single Small Plot) had a smaller confidence interval than the single large plot sampling. The other two small plot scenarios (Pooled and Averaged Small Plots) had smaller signal:noise and larger confidence intervals than the single large plot.

These results are in sharp contrast to the same analysis performed with the NCHF IBI where precision was approximately doubled with small plot sampling (Genet et al. 2005). This may be due to decreased natural within site variability in TP ecoregion wetlands compared to NCHF

Development and Validation of Temperate Prairie Wetland IBIs 82

ecoregion wetlands. Another potential factor is a maturation process in the sampling methodology itself, where observers develop habits over time, locating sample plots more consistently. Regardless of the factors, single large plot sampling apparently is providing adequate data to detect wetland changes due to anthropogenic impacts.

Based on the NCHF IBI results the MPCA aggressively moved to adopt small plot sampling. In 2004-05 primary sampling in the Northern Lakes and Forest (NLF) ecoregion (Figure I-1) for IBI development was done with small plots. Based on the results presented here, the adoption of small plot sampling may have been premature. If small plot sampling does not consistently improve IBI performance it may have been more appropriate to have sampled the NLF ecoregion with the same method as in the NCHF and TP ecoregions to produce a consistent statewide dataset. An increased effort will be given in the future to resolving this issue. The MPCA likely will periodically revisit IBI development to incorporate index improvements and revise assessment standards. It would be appropriate to re-evaluate sampling methods at that time.

100 Impairment Threshold 80 90% CI90%

60

Plant IBI Plant 40

20

0 0 20406080100 HDS

Figure II-14. Preliminary impairment threshold and application of 90% confidence intervals in depressional wetland ALUS assessment with the TP Plant IBI. 2002-03 data are shown from sites sampled once (N = 1).

Table II-13. Plant IBI signal:noise and 90% confidence intervals produced from four different sampling scenarios. Bold text indicates the number of standard sample replicates that would be adopted to sample a wetland for a given scenario.

Signal:Noise 90 % Confidence Interval Sampling Scenario (N = 1) N = 1 N = 2 N = 3 N = 4 N = 5 Single Large Plot 7.5 26.3 18.6 15.2 13.2 11.8 Pooled 4 Small Plots 6.2 34.1 24.1 19.7 17.0 15.2 Single Small Plot 1.8 47.1 33.3 27.2 23.5 21.0 Averaged 4 Small Plots 3.9 32.0 22.6 18.5 16.0 14.3

Development and Validation of Temperate Prairie Wetland IBIs 83

Acknowledgements

Funding for this project was provided by a Wetland Program Development grant, Section 104(b)3 Clean Water Act (Federal Assistance #CD-975768-01). Assistance with field sampling was provided by Joel Chirhart, Harold Wiegner, Bruce Sandstrom, Dan Helwig, Scott Niemela, Mike Feist, Bob Murzyn, Maureen Minister, Adam Hoffman, and Travis Fristed. Scott Milburn of Critical Connections Ecological Services, Inc. provided the plant taxonomic nomenclature crosswalk between the Iowa, N. and S. Dakota, and Minnesota floras to facilitate use of Coefficients of Conservatism from those states. We also appreciate the support and technical assistance provided by our EPA Project Officer, Alicia Hernandez and project technical contact, Sue Elston.

Literature Cited

Adamus, R.R. and K. Brandt. 1990. Impacts on quality of inland wetlands of the United States: a survey of indicators, techniques, and applications of community level biomonitoring data. EPA/600/3-90/073. U.S. Environmental Protection Agency Environmental Research Laboratory, Corvallis, OR, USA.

Apfelbeck, R.S. 1999. Development of biocriteria for wetlands in Montana. Montana Department of Environmental Quality, Helena, MT.

Bailey, R.G. 1995. Description of the ecoregions of the United States. 2nd edition. Misc. Publication. No. 1391, Washington, D.C. U.S. Forest Service. 108 pp with separate map at 1:7,500,000.

. 1998. Ecoregions map of North America: Explanatory Note. Misc. Publication Number 1548, U.S.D.A. Forest Service. 10 pp with map.

Barbour, M.T., J. Gerritsen, G.E. Griffith, R. Frydenborg, E. McCarron, J.S. White, and M.L. Bastian. 1996. A framework for biological criteria for Florida streams using benthic macroinvertebrates. Journal of the North American Benthological Society, 15:185-211.

Batzer, D.P. and S.A. Wissinger. 1996. Ecology of insect communities in nontidal wetlands. Annual Review of Entomology, 41:75-100.

Bedford, B.L., M.R. Walbridge, and A. Aldous. 1999. Patterns in nutrient availability and plant diversity of temperate North American wetlands. Ecology, 80:2151- 2169.

Blocksom, K.A., J.P. Kurtenbach, D.J. Klemm, F.A. Fulk, and S.M. Cormier. 2002. Development and evaluation of the lake macroinvertebrate integrity index (LMII)

Development and Validation of Temperate Prairie Wetland IBIs 84

for New Jersey lakes and reservoirs. Environmental Monitoring and Assessment, 77:311-333.

Bourdaghs, M., C.A. Johnston, and R.R. Regal. In Press. Properties and performance of the floristic quality index in Great Lakes coastal wetlands.

Brinson, M.M. 1993. A hydrogeomorphic classification of wetlands. U.S. Army Corps of Engineers, Wetlands Research Program. Technical Report WRP-DE-4.

Burton, T.M., D.G. Uzarski, J.P. Gathman, J.A. Genet, B.E. Keas, and C.A. Stricker. 1999. Development of a preliminary invertebrate index of biotic integrity for Lake Huron coastal wetlands. Wetlands, 19:869-882.

Burton, T.M., D.G. Uzarski, J.A. Genet, K.S. Genet, and S.K. Riffell. 2003. Baseline reference data collection and development of an index of biotic integrity for forested, depressional wetlands of southern Michigan. Michigan State University, Final Report to Michigan Department of Environmental Quality, Lansing, MI.

Butcher, J.T., P.M. Stewart, and T.P. Simon. 2003. A benthic community index for streams in the northern lakes and forests ecoregion. Ecological Indicators, 3:181- 193.

Cleland, D.T., P.E. Avers, W.H. McNab, M.E. Jensen, R.G. Bailey, T. King, and W.E. Russell. 1997. National hierarchical framework of ecological units In Boyce, M.S. and A. Haney [eds] Ecosystem Management Applications for Sustainable Forest and Wildlife Resources. Yale University Press, New Haven, CT. pp. 181- 200.

Cohen, M.J., S. Carstenn, and C.R. Lane. 2004. Floristic quality indices for biotic assessment of depressional marsh condition in Florida. Ecological Applications, 14:784-794.

Cowardin, L.M., V. Carter, F.C. Golet, and E.T. LaRoe. 1979. Classification of wetlands and deepwater habitats of the United States. U.S. Fish and Wildlife Service, Washington, DC. FWS/OBS 79/31. 103 pp.

Cvancara, A.M. 1983. Aquatic mollusks of North Dakota. North Dakota Geologic Survey Report of Investigation #78. 141 pp.

Dale, V.H. and S.C. Beyeler. 2001. Challenges in the development and use of ecological indicators. Ecological Indicators, 1:3-10.

Delettre, Y.R. 1989. Effect of the duration and intensity of drought on abundance and phenology of adult Chironomidae (Diptera) in a shallow temporary pool. Archive fuer Hydrobiology, 114:383-399.

Development and Validation of Temperate Prairie Wetland IBIs 85

Detenbeck, N. E., C. M. Elonen, D. L. Taylor, A. M. Cotter, F. A. Puglisi, and W. D. Sanville. 2002. Effects of agricultural activities and best management practices on water quality of seasonal prairie pothole wetlands. Wetlands Ecology and Management, 10:335-354.

Euliss, N.H. and D.M. Mushet. 2004. Impacts of water development on aquatic macroinvertebrates, amphibians, and plants in wetlands of a semi-arid landscape. Aquatic Ecosystem Health and Management, 7:73-84.

Fore, L.S. 2003a. Developing biological indicators: lessons learned from Mid-Atlantic streams. Report prepared for EPA under Contract No. 50-CMAA-900065. EPA 903/R-003/003. U.S. Environmental Protection Agency, Office of Environmental Information and Mid-Atlantic Integrated Assessment Program, Region 3, Ft. Meade, MD.

. 2003b. Development and testing of invertebrate biomonitoring tools for Florida streams. Technical Report, Florida Department of Environmental Protection, Tallahassee, FL.

Galatowitsch, S.M., N.O. Anderson, and P.D. Ascher. 1999. Invasiveness in wetland plants in temperate North America. Wetlands, 19:733-755.

Galatowitsch S.M. and T.V. McAdams. 1994. Distribution and requirements of plants on the upper Mississippi River. Unit Cooperative Agreement 14-16-0009-1560, Work Order 36. Iowa Cooperative Fish and Wildlife Research Unit, Ames, IA.

Genet, J.A. 2006. Assessing the quantity and quality of depressional wetlands in the Redwood River watershed utilizing a probabilistic sampling design. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency. Assistance #CD975768-01.

Genet, J.A., M.C. Gernes, and H. Markus. 2004. Defining wetland condition assessment processes. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency. Assistance #CD975938-01.

Genet, J.A., M. Bourdaghs, and M.C. Gernes. 2005. Advancing wetland biomonitoring in Minnesota. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency. Assistance #BG98568800.

Gernes, M.C. and J.C. Helgen. 1999. Indexes of biotic integrity (IBI) for wetlands: Vegetation and Invertebrate IBIs. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency. Assistance #CD995525-01.

. 2002. Indexes of biological integrity (IBI) for large depressional wetlands in Minnesota. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency. Assistance #CD985879-01.

Development and Validation of Temperate Prairie Wetland IBIs 86

Gittelman, S.H. 1974. The habitat preference and immature stages of Neoplea striola (Hemiptera: Pleidae). Journal of the Kansas Entomological Society, 47: 491-503.

Graves, G.A., Strom, D.G. and Robson, B.E. 1998. Stormwater impact to the freshwater Savannas preserve marsh, Florida, USA. Hydrobiologia, 379: 111-122.

Hanson, B.A. and G.A. Swanson. 1989. Coleoptera species inhabiting prairie wetlands of the Cottonwood Lake area, Stutsman County, North Dakota. Prairie Naturalist, 21:49-57.

Harman, W.N. 1997. Otsego lake macrobenthos communities between 1968 and 1993: indicators of decreasing water quality. Journal of Freshwater Ecology, 12: 465- 476.

Hudon, C. 1997. Impact of water level fluctuations on St. Lawrence River aquatic vegetation. Canadian Journal of Fisheries and Aquatic Sciences, 54:2853-2865.

Hungerford, H.B. 1948. The Corixidae of the Western Hemisphere (Hemiptera). University of Kansas Scientific Bulletin, 32:5-827.

Jackson, L.E., J.C. Kurtz, and W.S. Fisher [eds]. 2000. Evaluation guidelines for ecological indicators. U.S. Environmental Protection Agency, Office of Research and Development, Research Triangle Park, NC, USA. EPA 620 R-99 005.

Kadlec, R.H. and F.B. Bevis. 1990. Wetlands and wastewaster: Kinross, Michigan. Wetlands, 10:77-92.

Kantrud, H.A. 1992. History of cattails on the prairies: wildlife impacts. In G.M. Linz [ed.] Cattail Management Symposium. U.S. Dept. of Agriculture, Animal and Plant Health Inspection Service, Animal Damage Control, Denver Wildlife Research Center, U.S. Dept. of Interior, Fish and Wildlife Service. Jamestown, ND.

Kantrud, H.A., G.L. Krapu, and G.A. Swanson. 1989. Prairie basin wetlands of the Dakotas: a community profile. US Fish and Wildlife Service Biological Report 85 (7.28), 116 pp.

Karr, J.R. and E.W. Chu. 1999. Restoring Life in Running Waters-Better Biological Monitoring. Island Press, Washington, DC.

Kashian, D.R. and T.M. Burton. 2000. A comparison of macroinvertebrates of two Great Lakes coastal wetlands: testing potential metrics for an index of ecological integrity. Journal of Great Lakes Research, 26:460-481.

Development and Validation of Temperate Prairie Wetland IBIs 87

Lane, C.R. 2003. Biological indicators of wetland condition for isolated depressional herbaceous wetlands in Florida. Ph.D. Dissertation, University of Florida, Gainesville, FL.

LaBaugh, J.W., T.C. Winter, V.A. Adomaitis, and G.A. Swanson. 1987. Hydrology and chemistry of selected prairie wetlands in the Cottonwood Lake area, Stutsman County, North Dakota, 1979-82. US Geological Survey Professional Paper 1431. US Government Printing Office, Washington, DC.

Lillie, R.A., P. Garrison, S.I. Dodson, R.A. Bautz, and G. LaLiberte. 2002. Refinement and expansion of wetland biological indices for Wisconsin. Wisconsin Department of Natural Resources, Final Report to U.S. Environmental Protection Agency. Assistance #CD975115-01.

Lopez, R.D. and M.S. Fennessy. 2002. Testing the floristic quality assessment index as an indicator of wetland condition. Ecological Applications, 12:487-497.

Ludwa, K.A. 1994. Urbanization effects on palustrine wetlands: empirical water quality models and development of macroinvertebrate community-based biological index. Thesis, University of Washington, Seattle, WA.

Ludwa, K.A. and K. Richter. 2000. Macroinvertebrate distribution, abundance, and habitat use. In R. Homer and A. Azous [eds] Wetlands and Urbanization: Implications for the Future. Lewis Publishers, New York, NY.

Mack, J.J. 2004. Integrated Wetland Assessment Program. Part 9: Field Manual for the Vegetation Index of Biotic Integrity for Wetlands v. 1.3. Ohio EPA Technical Report WET/2004-9. Ohio Environmental Protection Agency, Wetland Ecology Group, Division of Surface Water, Columbus, OH.

Magurran, A.E. 1988. Ecological Diversity and its Measurement. Princeton University Press, Princeton, NJ.

Merritt R.W. and K.W. Cummins [eds]. 1996. An Introduction to the Aquatic Insects of North America, 3rd edition. Kendall/Hunt Publishing Company, Dubuque, IA.

Minnesota Pollution Control Agency (MPCA). 1999. Sulfate in Minnesota’s ground water. Ground Water Monitoring and Assessment Program, Minnesota Pollution Control Agency, St. Paul, MN. http://www.pca.state.mn.us/water/groundwater/ gwmap/sulfate7.pdf

. 2004. Guidance manual for assessing the quality of Minnesota surface waters for the determination of impairment, 305(b) report and 303(d) list. Environmental Outcomes Division, Minnesota Pollution Control Agency, St. Paul, MN. http://www.pca.state.mn.us/publications/ manuals/ tmdlguidancemanual04.pdf

Development and Validation of Temperate Prairie Wetland IBIs 88

Mitsch, W.J. and J.G. Gosselink. 2000. Wetlands, 3rd edition. John Wiley & Sons, Inc., New York, NY.

Montgomery, D.C., E.A. Peck, and G.G. Vining. 2001. Introduction to Linear Regression Analysis, 3rd edition. John Wiley & Sons, Inc., New York, NY.

National Research Council (NRC). 2000. Ecological Indicators for the Nation. National Academy Press, Washington, DC.

Neckles, H.A., H.R. Murkin, and J.A. Cooper. 1990. Influences of seasonal flooding on macroinvertebrate abundance in wetland habitats. Freshwater Biology, 23:311- 322.

Omernik, J.M. 1987. Ecoregions of the conterminous United States. Annals of the Association of American Geographers, 77:118-125.

Pajunen, V.I. and J.Salmi. 1991. The influence of corixids on the bottom fauna of rock- pools. Hydrobiologia, 222:77-84.

Parendes, L.A. and J.A. Jones. 2000. Role of light availability and dispersal in exotic plant invasion along roads and streams in the H.J. Andrews Experimental Forest, Oregon. Conservation Biology, 14:64-75.

Peet, R.K. 1974. The measurement of species diversity. Annual Review of Ecology and Systematics, 5:285-307.

Rabeni, C.F. and K.E. Doisy. 2000. Correspondence of stream benthic invertebrate assemblages to regional classification schemes in Missouri. Journal of the North American Benthological Society, 19:419-428.

Rachich, J. and R.J. Reader. 1999. An experimental study of wetland invasibility by purple loosestrife (Lythrum salicaria). Canadian Journal of Botany, 77:1499- 1503.

Reed, P.B. 1988. National list of plant species that occur in wetlands. Biological Report 88-24, U.S. Fish and Wildlife Service, Washington, DC.

Reynolds, J.D. 1975. Feeding in corixids (Heteroptera) of small alkaline lakes in central B.C. Verh. Internat. Verein. Limnol., 19:3073-3078.

Reynolds, J.D. and G.G.E. Scudder. 1987. Experimental evidence of the fundamental feeding niche in Cenocorixa (Hemiptera: Corixidae). Canadian Journal of Zoology, 65:967-973.

Rutherford, J.E. and R.J. Mellow. 1994. The effects of an abandoned roast yard on the

Development and Validation of Temperate Prairie Wetland IBIs 89

fish and macroinvertebrate communities of surrounding beaver ponds. Hydrobiologia, 294: 219-228.

Shaw, S.P. and C.G. Fredine. 1956. Wetlands of the United States, Their Extent and Their Value for Waterfowl and Other Wildlife. Circular 39, U.S. Fish and Wildlife Service, Washington D.C. 67 pp.

Simon, T.P., P.M. Stewart, and P.E. Rothrock. 2001. Development of multimetric indices of biotic integrity for riverine and palustrine wetland plant communities along Southern Lake Michigan. Aquatic Ecosystem Health and Management, 4:293-309.

Stewart, R.E. and H.A. Kantrud. 1971. Classification of natural ponds and lakes in the glaciated prairie region. US Fish and Wildlife Service Resource Publication. 92. 57 pp.

Swanson, G.A. 1984. Dissemination of amphipods by waterfowl. Journal of Wildlife Management, 48:988-991.

Swanson, G.A. and H.F. Duebbert. 1989. Wetland habitats of waterfowl in the prairie pothole region. In van der Valk, A.G. [ed.] Northern Prairie Wetlands. Iowa State University Press, Ames, IA. pp. 228-267.

Swink, F. and G. Wilhelm. 1994. Plants of the Chicago Region, 4th edition. Indiana Academy of Science, Indianapolis, IN.

Thorpe, J.H. and A.P. Covich. 1991. Ecology and Classification of North American Freshwater Invertebrates. Academic Press, New York, NY.

U.S. Environmental Protection Agency (US EPA). 2002. Methods for evaluating wetland condition: wetlands classification. Office of Water, U.S. Environmental Protection Agency, Washington DC. EPA-822-R-02-017.

Uzarski, D.G., T.M. Burton, and J.A. Genet. 2004. Validation and performance of an invertebrate index of biotic integrity for Lakes Huron and Michigan fringing wetlands during a period of lake level decline. Aquatic Ecosystem Health and Management, 7:269-288.

Van Sickle, J. 1997. Using mean similarity dendrograms to evaluate classifications. Journal of Agricultural, Biological, and Environmental Statistics, 2:370-388.

Van Sickle, J. and R.M. Hughes. 2000. Classification strengths of ecoregions, catchments, and geographic clusters for aquatic vertebrates in Oregon. Journal of the North American Benthological Society, 19:370-384.

Development and Validation of Temperate Prairie Wetland IBIs 90

Waite, I.R., A.T. Herlihy, D.P. Larsen, and D.J. Klemm. 2000. Comparing strengths of geographic and nongeographic classifications of stream benthic macroinvertebrates in the Mid-Atlantic Highlands, USA. Journal of the North American Benthological Society, 19:429-441.

Washington, H.G. 1984. Diversity, biotic, and similarity indices: a review with special relevance to aquatic ecosystems. Water Resources, 18:653-694.

Wilcox, D.A. and J.E. Meeker. 1991. Disturbance effects on aquatic vegetation in regulated and unregulated lakes in northern Minnesota. Canadian Journal of Botany, 69:1542-1551.

Wilcox, D.A., S.I. Apfelbaum, and R.D. Hiebert. 1985. Cattail invasion of sedge meadows following hydrologic disturbance in the Cowles Bog wetland complex, Indiana Dunes National Lakeshore. Wetlands, 4:115-128.

Wilcox, D.A., J.E. Meeker, P.L. Hudson, B.J. Armitage, M.G. Black, and D.G. Uzarski. 2002. Hydrologic variability and the application of index of biotic integrity metrics to wetlands: a Great Lakes evaluation. Wetlands, 22:588-615.

Zar, J.H. 1999. Biostatistical Analysis, 4th edition. Prentice Hall, Inc., Upper Saddle River, NJ.

Zheng, L. and R.J. Stevenson. 2001. Nutrient limitation and algal response on nutrient addition in Michigan wetlands. (abstract) North American Benthological Society Annual Meeting, La Crosse, WI.

Development and Validation of Temperate Prairie Wetland IBIs 91

Appendices

Development and Validation of Temperate Prairie Wetland IBIs 92

Appendix A - List of wetland sites sampled in 2002. See Table I-4 for definitions of ecoregion, section, and basin abbreviations. Area Omernik ECS UTM Coordinates2 1 SiteName ha acres County EcoRegion Section Basin Northing Easting BarryWMA 7.4 18.3 Bigstone NGP 251B MN 5051296.89 220767.64 BetShalom 0.2 0.5 Hennepin NCHF 222M UM 4975726.43 464213.79 Breen 12.1 29.9 LeSueur NCHF 222M LM 4901302.48 435208.53 Bryclyn 1.1 2.6 Freeborn WCBP 251B MN 4836081.79 456893.01 BushLake 1.6 4.0 Hennepin NCHF 222M MN 4964780.28 469293.94 Carex2 0.8 2.0 Freeborn WCBP 222M CE 4840625.06 492966.43 DellRd 0.4 1.1 Hennepin NCHF 222M MN 4966137.33 460282.13 EastlickMarsh 21.1 52.1 Murray WCBP 251B DM 4886954.98 285623.20 FrancoWMA 24.0 59.3 Chippewa WCBP 251B MN 4978867.44 309707.98 FurgameWMA 21.6 53.3 Lyon NGP 251B MN 4923693.25 258339.73 Glacial 71.5 176.8 Pope NCHF 251B MN 5044366.35 304058.98 Gleason 0.5 1.1 Hennepin NCHF 222M UM 4981168.62 460615.85 GoldenWPA 22.2 54.9 Stevens NGP 251B MN 5037654.96 258426.68 GreatOasisWMA 4.8 11.8 Murray WCBP 251B DM 4885969.54 269406.88 Hancock 10.4 25.7 Stevens NGP 251B MN 5042357.70 282432.98 HardScrab 0.8 1.9 Hennepin NCHF 222M UM 4973554.07 446452.04 Hoffman 13.9 34.2 Swift NGP 251B MN 5027827.55 277872.27 Kerk 1.6 4.0 Swift NGP 251B MN 5025728.21 314126.59 Kipling 0.4 0.9 Hennepin NCHF 222M UM 4975001.63 473477.61 Lake 21 11.6 28.5 Kandiyohi NCHF 251B MN 5021366.86 339787.74 LakeCharlotte 22.5 55.7 Kandiyohi WCBP 251B UM 4984799.50 351744.10 LakeElisabeth 8.1 20.0 Kandiyohi WCBP 251B UM 4992852.19 357997.88 Lee 10.9 27.0 Stevens NGP 251B MN 5039138.66 259635.68 Legion 8.9 21.9 Hennepin NCHF 222M UM 4970632.15 479271.30 LoneTreeWMA 30.8 76.2 Lyon NGP 251B MN 4945094.80 285102.11 LyonsWMA 10.5 25.9 Lyon WCBP 251B MN 4910235.09 270339.42 Malardi 59.3 146.5 Wright NCHF 222M UM 4992407.14 429281.14 Malta 17.0 42.0 Bigstone NGP 251B MN 5038456.45 246101.16 Manchester 23.1 57.1 Freeborn WCBP 222M MN 4845191.40 465133.04 Milan 39.9 98.6 Chippewa NGP 251B MN 4999754.07 271331.80 Morraine 1.3 3.3 Hennepin NCHF 222M MN 4966686.10 462491.51 New Prairie 11.4 28.1 Pope NGP 251B MN 5059088.65 293350.50 Ney 3.1 7.7 LeSueur NCHF 222M MN 4932008.91 430090.46 OakGlenEast 18.3 45.3 Steele WCBP 222M LM 4864125.61 495000.59 OakGlenWest 22.7 56.2 Steele WCBP 222M LM 4863743.67 494463.81 Prairie 14.6 36.0 Hennepin NCHF 222M UM 5004365.74 449334.14 Prairie Marsh 13.9 34.4 Lyon NGP 251B MN 4916843.11 263204.91 RenoRef 7.7 19.1 Pope NCHF 251B MN 5062841.42 308282.66 RolhiksWMA 56.0 138.5 Redwood WCBP 251B MN 4925924.48 310675.70 RostWMA 27.3 67.4 Lincoln NGP 251B MN 4929006.30 248341.34 TheoWirth 1.6 4.0 Hennepin NCHF 222M UM 4981980.60 473913.12 Turtle 18.5 45.7 Hennepin NCHF 222M UM 4986949.56 462502.79 TylerWMA 41.6 102.7 Lincoln NGP 251B MN 4907059.82 248485.29 Westmark 0.4 0.9 Hennepin NCHF 222M UM 4976083.26 463300.28 WillowLake 63.8 157.7 Redwood WCBP 251B MN 4909595.84 324345.75 Wood 37.9 93.8 Hennepin NCHF 222M UM 4969275.51 476792.66 Yohi 17.2 42.5 Kandiyohi WCBP 251B UM 4999089.35 351226.00 1 WMA= MNDNR Wildlife Management Area; WPA = USFWS Waterfowl Production Area 2 Source datum for coordinates is WGS 1984, UTM Zone 15N

Development and Validation of Temperate Prairie Wetland IBIs 93

Appendix B - Macroinvertebrate Sampling Protocols for Depressional Wetlands.

Development and Validation of Temperate Prairie Wetland IBIs 94 Minnesota Pollution Control Agency Biological Monitoring Program

MACROINVERTEBRATE COMMUNITY SAMPLING PROTOCOL FOR DEPRESSIONAL WETLAND MONITORING SITES

I. PURPOSE

To describe the methods used by Minnesota Pollution Control Agency’s (MPCA) Biological Monitoring Program to collect macroinvertebrate community information at wetland monitoring sites for the purpose of assessing water quality and developing biological criteria.

II. SCOPE/LIMITATIONS

This procedure applies to all monitoring sites for which an integrated assessment of water quality is to be conducted. An integrated assessment involves the collection of biological (macroinvertebrate and plant) and chemical data to assess wetland condition.

III. GENERAL INFORMATION

Sites may be selected for assessment for a number of reasons including: 1) sites randomly selected for condition monitoring as part of the Environmental Monitoring and Assessment Program (EMAP), 2) sites selected for the development and calibration of biological criteria (e.g., Index of Biological Integrity), and 3) sites selected to evaluate a suspected source of pollution.

IV. ACTION STEPS

A. Field Sampling

For sampling wetland macroinvertebrate assemblages a seasonal index period of June - early July is preferred, this can be earlier if spring temperatures are unusually high that year. In previous wetland work, Minnesota Pollution Control Agency (MPCA) researchers found that some of the invertebrates were too immature to identify when sampled in May, especially the dragonfly nymphs. The sampling window was therefore moved forward to June. In stream invertebrate work, the sampling is done in September to ensure base flow conditions, and to obtain a relatively high percentage of mature larval invertebrates. This approach does not work for wetlands because: a) the wetlands may be dry or unsampleable later in the field season, and b) the wetlands will be heavily colonized by invertebrates which have immigrated into them from other waterbodies. In the latter situation, the invertebrate community in September may be less reflective of the water quality of the wetland itself than the invertebrate community in early summer.

Currently the MPCA has emphasized depressional wetlands in their development of invertebrate indices of biotic integrity (IBI). Depressional wetlands can be stratified into nearshore emergent (shore to 1 m water depth), deep emergent (> 1m water depth), and open water submergent vegetation zones. The MPCA has focused on the nearshore emergent vegetation zone for developing the invertebrate index of biological integrity. In this zone there is a high richness and abundance of invertebrates, including the large predatory insects, due in part to the decomposing vegetation and diverse vegetative microhabitats which occur in this zone. Sampling is conducted in areas that are representative of the wetland emergent zone. However, field partitioning of the wetland for invertebrate sampling as above may need to be modified as the MPCA expands assessment to other wetland types (e.g., riparian, forested).

Sampling of invertebrates by the MPCA Biological Unit is restricted to macroinvertebrates, excluding ostracods and the smaller microinvertebrates which are not retained by a U.S. Standard No. 30 sieve (28 meshes per inch, 0.595 mm openings). Macroinvertebrates are collected in the field using two sampling techniques: dip nets and activity traps. Previous MPCA projects (e.g., Helgen et al. 1993) demonstrated

Development and Validation of Temperate Prairie Wetland IBIs 95

that dip net sampling captures the greatest richness of invertebrates, but the actively swimming or night- active predators may be under-collected by this method. Therefore, activity traps are placed in the wetland for two days to collect the active swimmers (see details below). Previous work by MPCA (Helgen et al. 1993) has shown reduced taxa richness in benthic, or bottom samples taken with core tubes and subsequently this method of sampling is not currently in use.

Dip Net Sampling Two samples are collected from each wetland using a heavy-handled D-frame aquatic dip net with a 600 micron mesh size (Wildlife Supply Company). The two samples are taken in different areas within the same general location of the nearshore emergent vegetation zone and are not intended to be replicates, but rather are done to sample the wetland more widely. Ultimately, the data from the two samples are combined for purposes of calculating IBI metric scores. Each dip net sample consists of two dipnetting efforts composited into one sample. Each effort consists of sweeping the dip net strongly a few times (3 -5 depending on the density of the vegetation), reaching outward and pulling towards the body in a rapid motion. Each sweep should be through the water column and vegetation downwards to near the bottom. If mud is scraped into the net, the sample should be discarded and the sampling effort must be repeated in an area away from the previous netting, after the net has been cleaned out.

A method utilized by MPCA reduces the amount of time associated with separating invertebrates from the vegetation that invariably gets swept into the dip net. This method involves the placement of the entire dip net contents on top of a framed ½ inch hardware cloth screen set over two small pans (Coleman cooler style) containing sieved water (Figure 1). The frame is placed so no open screen area projects beyond the pans of water below. This frame and pan setup is placed into a larger plastic pan (tote tray) which can be floated on the water. Over a period of ten minutes the vegetation is spread apart on the hardware cloth to allow the invertebrates to drop or crawl out into the pans below. After ten minutes a second dipnetting effort is done in a nearby area, the vegetation from the first dip net effort is removed, and the second net's contents are placed on the cleared screen. The spreading process is repeated for about 10 minutes, after which the vegetation is again discarded.

After both sweeping efforts are completed, the contents in the two small pans are poured through a 200 micron nytex nylon net sieve to drain out the water. The sieve is made with 15 cm length of 4” diameter PVC pipe with the net glued on one end with a ring of the PVC. The 200 micron sieve is used to retain the chironomids dislodged from the vegetation. The contents of the sieve is back-flushed with 100% alcohol with a strong squirt-bottle into a sample jar, thus combining the two dip net efforts into one dip net sample. The goal is to end up with 80% alcohol final concentration. Care must be taken to represerve samples containing a large catch of invertebrates, or to divide the sample between two jars (sample #, jar 1 of 2, jar 2 of 2). The jar should have not more than 1/3 volume of invertebrates to alcohol. Sixteen-ounce plastic jars with foam or polypropylene seals are useful for preservation in the field. Labels made with India ink or pencil on 100% cotton paper or other material known to survive the preservatives are placed within the jar. Any label placed on the outside of the jar is only for convenience in managing samples.

Activity Trap Sampling The activity traps work as passive funnel traps to collect organisms that swim into the funnel and pass through the neck into the bottle. Made from clear 2-liter plastic beverage bottles obtained from the manufacturer free of labels or opaque parts, the traps are nearly invisible underwater. The top of the bottle is cut cleanly with a hot wire at the shoulder and inverted into the bottle. The bottle traps used by the MPCA are designed with four two inch grooves cut in to the funnel edge by a hot wire to allow the funnel to snap into the bottle opening without the use of clips or visible straps (Figure 2). The traps are supported on a 4 ft ½" dowel, or a 4 ft fiberglass electric fence post, and attached with a flexible half section of 3" thin wall PVC pipe which allows raising or lowering the activity trap on the dowel (Figures 2 & 3).

Ten activity traps are placed in each wetland for two consecutive nights within the nearshore emergent vegetation zone. The ten activity traps are set out in pairs with each trap in a pair located approximately 3 - 4 meters apart. In the shallowest water (15 cm) the traps are placed just under the surface of the water, but

Development and Validation of Temperate Prairie Wetland IBIs 96

Figure 1. Diagram of hardware cloth and tray apparatus for separating invertebrate specimens from vegetation collected in dip net samples.

should not be resting on the bottom to avoid filling the bottletrap with sediment. In deeper water (> 50 cm) traps are placed horizontally about 15 - 20 cm under the surface. Traps are not placed at the deeper edge of the vegetation in the open water area because capture efficiency goes down as the water gets deeper. The traps are backfilled with water leaving no air bubbles inside in order to reduce predation within the trap. The wingnut should be tightened enough so the trap remains horizontal (see Figure 3a).

After the required two-night period the traps can be collected by slightly loosening the wingnut in order to rotate the trap to a vertical position and slide it up the dowel by slightly compressing the dowel clamp. Then the funnel is removed and the contents of the trap are poured through the 200 micron sieve. The trap is squirted with tap water and the inside is rubbed to dislodge leeches and other invertebrates. Specimens attached to both faces of the funnel opening are also considered part of the sample. These dislodged specimens are then added to the contents of the sieve. The second trap of the pair is collected and its contents are poured into the same sieve. The sieve is back-flushed into a sample jar with 100% alcohol to a final concentration of 80%. Care must be taken to represerve samples having a large catch of invertebrates, or divide the sample between multiple jars (sample #, jar 1 of 2, jar 2 of 2, etc.). The jar should have not more than 1/3 volume of invertebrates to alcohol. Sixteen-ounce plastic jars with foam or polypropylene seals are useful for preservation in the field. Labels with India ink or pencil on 100% cotton paper or other material known to survive the preservatives are placed within the jar. Any label placed on the outside of the jar is only for convenience in managing samples.

Development and Validation of Temperate Prairie Wetland IBIs 97

Figure 2. Activity trap design illustrating adjustable PVC bracket and funnel grooves.

Figure 3. Activity trap design illustrating a) view into funnel and b) lateral view.

Development and Validation of Temperate Prairie Wetland IBIs 98

B. Sample Storage and Maintenance

All preserved samples are kept in a hazardous materials room. They are checked within a week of field sampling and then periodically for adequate preservative volume, and represerved with 80% alcohol as necessary. For samples that require additional alcohol the lids are tightened or replaced in order to prevent further evaporation.

C. Sample Processing

A combination of dissecting microscopes and compound microscopes are used for sorting and identifying macroinvertebrates in the laboratory. At MPCA there is one Olympus SZX microscope, one Olympus SZH, two Olympus SZ40 microscopes, and one Olympus BX40 compound microscope. The procedures for sorting invertebrates from dip net and activity trap samples are outlined in Table1.

Sample Identifications Organisms are identified to the lowest possible taxonomic level. Typically this is to the genus level though often it is to the species level, at a minimum they will be identified to the taxonomic level as designated for each group in Table 2. Once specimens are sorted according to the guidelines listed in Table 2, identifications are then made for each specimen within the sorted groups. Identifying all the specimens in a

Table 1. Sorting protocol for dip net and activity trap macroinvertebrate samples.

Procedure Comments

1) Note start time and site information on to data Check/retain inner label. sheets.

2) Pour the sample into sieve and rinse Collect, cover, and save alcohol for with tap water. re-preservation

3) Backflush sample with water to glass Generally, sample is large, and must be picking tray. Tray should be placed over separated into two or more efforts to grid transparency upon light box. accommodate picking and accuracy.

4) Fill glass tray slightly (1 cm) with water. This Gentle stirring/probing sample also helps helps to separate organisms from debris. dislodge critters from debris.

5) Using forceps, pick entire sample to sorting Pick, count, and visually ID according to trays, jars, and petri dishes according to taxa list taxa List (Table 2). A magnifying lamp (Table 2). Be sure to keep record of separate taxa may be beneficial for small organisms on mechanical counter. Properly fill in lab sheets. and juveniles.

6) Specimens may then be combined in general groups Grouping conserves resources. (dragons/damsels, snails/sphaeriidae, etc.) and placed Label should contain site, date, collector, into vials/jars with proper labeling for later identification. and sample type written in pencil on Preserve in 80% ethanol. cotton stock.

7) Replace original label and sample remnant to Be sure to check prior alcohol for sample jar. Backflush using 80% alcohol and fill strength. using previously saved alcohol.

8) Note end time on datasheet. Calculate total time. Calculate in hourly increments.

Development and Validation of Temperate Prairie Wetland IBIs 99

group (e.g., dragonflies) at the same time facilitates proper designations (e.g., species or genus) by allowing comparisons of closely related taxa. References containing the taxonomic keys for identifications are provided below. Where ambiguity exists, specimens will be set aside for identification by an independent invertebrate taxonomist (also see Reference Collection section).

Table 2. Invertebrate taxa list indicating which groups are counted and identified for each sample type (dip net or activity trap) and the taxonomic resolution for each group.

Activity Trap Dip Net Identify to Group Total Picked Total Picked 3 Amphipoda (Ad) x x x x Genus 3, 4 Amphipoda (Juv, < 3mm) x x Lowest Level Anisoptera (Larvae) x x x x Genus Anostraca x x x x Genus 2Chironomidae (Larvae) x x * 2 Chironomidae (Larvae, < 3mm) x * Coleoptera (Ad) x x x x Genus Coleoptera (Larvae) x x x x Genus 3 Conchostraca x x x Genus 1Corixidae (Ad) x x x x Genus 1, 3, 4 Corixidae (Juv.) x x Family Diptera (Larvae) x x x x Genus Ephemeroptera (Larvae) x x x x Genus Gastropoda (Ad) x x x x Species 3Gastropoda (Juv.) x x Genus Hemiptera (Ad) x x x x Genus Hemiptera (Juv.) x x x x Lowest level Hirudinea (Ad) x x x x Species Hirudinea (Juv.) x x x x Lowest level Isopoda (Ad) x x x x Genus Isopoda (Juv.) x x x Lowest level Lepidoptera (Larvae) x x x x Genus Malacostraca x x x Family Megaloptera (Larvae) x x x x Genus 1, 3 Neoplea (Ad & Juv.) x x Genus 3 Sphaeriidae (Ad & Juv.) x x x Family Trichoptera (Larvae) x x x x Genus Zygoptera (Larvae) x x x x Genus 1 Represents Corixidae & Neoplea which were counted/recorded separately from other Hemiptera. 2 Chironomidae Ids are done on dip net samples only. Estimate abundance of chironomids in activity traps. 3 Represents groups that may be counted within the glass tray. 4 Represents group that may be sub-sampled in high numbers. * Identifications made by Dr. Len Ferrington (University of Minnesota).

Development and Validation of Temperate Prairie Wetland IBIs 100

Reference Collection A macroinvertebrate reference collection is maintained for each project at the MPCA Biomonitoring Laboratory (St Paul Office). This collection consists of specimens of each type of macroinvertebrate that has been collected for individual projects conducted by MPCA staff. A few specimens of each taxon are placed in vials or small jars which are labeled inside for the taxon, date, and collection site. This collection will be reviewed by other biologists to confirm the identifications for each project. Specimens for which the identification is uncertain will be reviewed by other biologists with expertise in the particular group.

D. Quality Assurance

At least ten percent of the sites for each project are sampled twice, either on the same date or within a week in an area that is equally representative of the wetland as was first selected for sampling. At least ten percent of the samples are repicked. If organisms were missed, the entire set of samples is repicked. At least ten percent of the picked samples will be reviewed by a qualified invertebrate biologist to verify identifications. In addition, the reference collection from the project will be reviewed by a qualified invertebrate biologist to verify identifications. Chironomidae will be identified by a specialist in the taxonomy of the group (Dr. Len Ferrington, University of Minnesota).

Data is recorded on standard hard copy lab and field data sheets. These data sheets and field notebooks will be copied and stored in a separate place. In addition, data from each project will be stored and maintained within a Microsoft® Access database that resides on the MPCA network drives and is normally backed up each night.

Following data input all entries are completely proofed before data analysis begins.

V. LITERATURE CITED

Helgen, J.H., K. Thompson, J.P. Gathman, M. Gernes, L.C. Ferrington, and C. Wright. 1993. Developing an Index of Biological Integrity for 33 Depressional Wetlands in Minnesota. Minnesota Pollution Control Agency, St. Paul, MN

Development and Validation of Temperate Prairie Wetland IBIs 101

Taxonomic References for Identifying Invertebrates

Burch, J.B. 1982. North American Freshwater Snails. Museum of Zoology. University of Michigan. Ann Arbor.

Clarke, Arthur H. 1981. The Freshwater Molluscs of Canada. National Museum of Natural Sciences. National Museums of Canada. Ottawa, Canada K1A OM8. 446 pp.

Clarke, Arthur H. 1973. The Freshwater Molluscs of the Canadian Interior Basin. Malacologia Vol 13, No 1-2 (includes snails and fingernail clams and distribution maps).

Edmunds, Jr., George F, S.L. Jensen, L. Berner. 1976. The Mayflies of North and Central America. University of Minnesota Press. Minneapolis.

Jokinen, Eileen H. 1992. The Freshwater Snails (Mollusca: Gastropoda) of New York State. New York State Museum Bulletin 482. New York State Museum Biological Survey. Albany. New York.

Klemm, Donald J. 1982. Leeches (Annelida: Hirudinea) of North America. US EPA Cincinnati, OH. EPA-600/3-82-025.

Hilsenhoff, William L. 1995. Aquatic Insects of Wisconsin. Publication Number 3 of the Natural Museums Council. University of Wisconsin - Madison. 79 pp.

Laursen, Jeffrey R., Gary A. Averbeck, Gary A Conboy. 1989. Preliminary Survey of Pulmonate Snails of Central Minnesota. Veterinary Parasitology, U. Minnesota School of Veterinary Medicine. Final Report to the Minnesota DNR Nongame Division.

Merritt, Richard W. and Kenneth W. Cummins. 1996. An Introduction to the Aquatic Insects of North America, 3rd Edition. Kendall/Hunt Publishing. Dubuque, Iowa. 862 pp.

Needham, James G. and Minter J. Westfall. 1954. The Dragonflies of North America. University of California Press. Berkeley.

Walker, Edmund M. 1953. The Odonata of Canada and Alaska. Volume 1. Part I General, Part II: The Zygoptera -- the Damselflies. University of Toronto Press, Toronto.

Walker, Edmund M. 1958. The Odonata of Canada and Alaska. Volume 2. Part III: The Anisoptera -- Four Families. University of Toronto Press. Toronto.

Walker, Edmund M. and Philip S. Corbet. 1978. The Odonata of Canada and Alaska. Vol 3, Part III: The Anisoptera -- Three Families. University of Toronto Press. Toronto.

Westfall, Minter J. Jr and Michael L. May. Damselflies of North America. 1996. Scientific Publications. Gainesville, FL.

Wiggins, Glenn B. 1996. Larvae of the North American Caddisflies (Trichoptera), 2nd edition. University of Toronto Press. Toronto.

Development and Validation of Temperate Prairie Wetland IBIs 102

Appendix C - Plant Sampling Protocols for Depressional Wetlands.

Development and Validation of Temperate Prairie Wetland IBIs 103 Minnesota Pollution Control Agency Biological Monitoring Program

AQUATIC PLANT COMMUNITY SAMPLING PROCEDURE FOR DEPRESSIONAL WETLAND MONITORING SITES

I. PURPOSE

To describe and document the standard operating procedure (SOP) used by the Minnesota Pollution Control Agency’s (MPCA) Biological Monitoring Program to collect aquatic plant community information at depressional wetland monitoring sites for the purpose of assessing water quality and developing biological assessment criteria.

II. SCOPE/LIMITATIONS

The following SOP applies to all depressional wetland monitoring sites for which an integrated assessment of water quality is to be conducted. An integrated depressional wetland assessment involves the collection of biological (macroinvertebrate and plant) and chemical data to assess wetland condition. The MPCA defines depressional wetlands as wetlands that occur within a shallow depression in the landscape that are not directly associated with streams (i.e., riparian wetland) or lakes (i.e., lacustrine fringe wetland); have a semi-permanent to permanent flooding regime (i.e., not temporarily flooded wetland or vernal pool); and are classified as type 3, 4 or 5 according to U.S. Fish and Wildlife Service Circular 39 (Shaw and Fredine 1956) (i.e., shallow marsh, deep marsh, or shallow water).

III. GENERAL INFORMATION

Sites may be selected for assessment for a number of reasons including: 1) sites randomly selected for ambient condition monitoring, 2) sites selected for the development and calibration of biological criteria, and 3) sites selected to evaluate a suspected source or result of pollution impacts. Although the reasons for monitoring a site may vary, the aquatic plant sampling protocol described in this document applies to all MPCA wetland monitoring sites unless otherwise noted.

IV. PERSONEL REQUIREMENTS

A. Field Crew Leader: The field crew leader must be a professional aquatic biologist with a good knowledge of the Minnesota wetland flora. He or she must have a minimum of a Bachelor of Science degree in aquatic biology, botany, or a closely related field; and have a minimum of six months field experience in wetland plant sampling and plant identification. Field crew leaders should also be proficient with map reading and orienteering; using both Global Positioning System (GPS) and compass. B. Field Assistant/Intern: The field assistant/intern must have at least one year of college education and an interest in aquatic biology. Coursework in environmental, natural resource, and/or biological science is preferred. C. General Qualifications: All personnel conducting this procedure must have the ability to perform rigorous physical activity in an outdoor setting; be capable of

Development and Validation of Temperate Prairie Wetland IBIs 104

lifting up to 50 lbs. of sampling equipment; be able to travel up to four nights per week during the summer months; and maintain a positive attitude within a team setting.

V. RESPONSIBILITIES

A. Field Crew Leader: The field crew leader is responsible for implementing the action steps of the procedure and ensuring that the data generated meets the standards and objectives of the Biological Monitoring Program and the MPCA. In addition, the field crew leader is responsible for planning sampling activities and ensuring that MPCA policies are followed during all sampling activities.

B. Field Assistant/Intern: The field assistant/intern is responsible for implementing the action steps of the procedure; including the maintenance, stocking, and storage of sampling equipment, data collection, and data recording.

VI. TRAINING

All personnel will receive instruction from a trainer designated by the program manager. Major revisions in this protocol require that all personnel that apply this procedure on behalf of the MPCA Biological Monitoring Unit be re-trained in the revised protocol by experienced personnel. The field crew leader will provide additional instruction to the field assistant/intern and will be responsible for monitoring the performance of the field assistant/intern throughout the field season.

VII. ACTION STEPS

A. Equipment Check: Before heading out into the field, check all equipment and supplies necessary to complete this procedure is present and in proper working condition (Table 1).

B. Field Sampling: The wetland vegetation biological assessment techniques employed by the MPCA (i.e., Index of Biological Integrity; Gernes and Helgen 2002) require data on the different kinds of plants growing in a wetland and how abundant those plants are. The vegetation sampling technique described in this procedure is adapted from what is known as releve sampling. Releve sampling was developed by Braun- Blaunquet in Europe and is currently being used by the Minnesota Department of Natural Resources (DNR) County Biological Survey and Natural Heritage Programs (Almendinger 1987). Essentially, releve sampling relies on the observer to select areas within the desired community that are representative of the overall community composition to place a sampling plot where plant data can be quantified.

Since 2001, the MPCA has been collecting field plant data using a hand held Personnel Data Assistant (PDA). This has reduced the amount of time recording data in the field and also reduced the time needed to produce sample results. Field data

Development and Validation of Temperate Prairie Wetland IBIs 105

Table 1. Equipment List-This table identifies all the equipment needed to complete the MPCA wetland vegetation sampling protocol.

Equipment Purpose Operation Check Personal Data Assistant -Field data recording device -Software function (PDA) -Date and time -Associated cords and devices

Global Positioning System -Navigation and sample location recording -Date and time (GPS) -Correct coordinate system and datum -Associated cord

Laptop Computer -Downloading and data storage -Software function -GIS applications -Associated cords and devices -Power inverter

Digital Camera -Photographic site documentation -Memory card(s) -Associated cords -Date and time

Cell Phone -Communication -Associated cord

4-8Rechargeable AA -Spare batteries for GPS and digital camera Batteries & Charger

Site Files and Maps -Site location information

Paper Data Sheets & -Backup in case of PDA failure Clipboard

Field Notebook -recording misc. notes, Backup for recording data

6 Tall Garden Stakes -Sampling plot corner posts, 2-spares

4 50 m Measuring Tapes -For laying out sampling plots, 2-spares

Chest Waders -To keep field workers dry

Raingear -To keep field workers dry

Field Guides -Aid with plant indentification

Hand Lens -Aid with plant indentification

1-2 Gallon Size Plastic -For collecting plant specimens Bags

Development and Validation of Temperate Prairie Wetland IBIs 106

Table 1. Equipment List-Continued. Equipment Purpose Operation Check Cooler with Ice -Short term preservation of water quality samples (see Wetland Water Chemistry SOP) and plant specimens

Plant Press with -For pressing plant specimens Newsprint, Blotters, & Cardboard

Wax Paper -Aid with pressing aquatic plant specimens

Shallow Pan -Aid with pressing aquatic plant specimens

Compass -Navigation & sampling plot layout

Pencils -For recording data

Permanent Marker -For labeling bags & water samples (see Wetland Water Chemistry SOP)

First-Aid Kit -Emergency medical care

sheets continue to be maintained, however, as a backup to the PDA and are included at the end of this SOP.

B.1. Record visit information: Upon arrival at a site begin recording visit information on the Visit Data Sheet (attached at the end of this SOP) or PDA. Record the Site Name, Date, Surveyor Name, and Arrival Time immediately. Also, document weather conditions in the Weather Notes space.

Throughout the remainder of the visit (i.e., during or following vegetation sampling), record other visit or site level data as appropriate. Document any site photographs in the Photo Information section. Record the Camera Make and Model used for the visit and the Photo Number reported from the camera and any associated Photo Notes for each photograph taken. Collect water chemistry measurements and samples (see Wetland Water Chemistry SOP), and record information in the Water Chemistry section. Also during the visit, conduct a site stressor verification assessment. Do this by walking around the margin of the wetland, noting any anthropogenic stressors that may be impacting the wetland. Complete the Habitat Alteration, Hydrologic Modification, and Sedimentation checklists in the Site Stressor Verification section as you proceed. Site stressor information is necessary for developing a Human Disturbance Score (HDS; Gernes and Helgen 2002) for the site. A brief site stressor assessment may have been completed during the initial site reconnaissance (see Wetland Site Reconnaissance SOP). The purpose of the site stressor verification during vegetation sampling is to assess the wetland more thoroughly and add to any

Development and Validation of Temperate Prairie Wetland IBIs 107

information gathered during the site reconnaissance. Finally, record the Leave Time (site departure time) when all of the data have been collected.

B.2. Determine the major plant communities in the wetland: The releve sampling method relies on the observer finding a ‘representative’ location in the wetland that best characterizes the vegetation of the entire wetland to place the sampling plot(s). The first step in this procedure then is to determine what the major plant communities in the wetland are. This can be done by finding an area where the entire wetland can be viewed or by walking around the margin of the wetland.

A

B

Figure 1. Hypothetical lay-out of a 10 m x 10 m (A) and a 5 m x 20 m (B) plot in two wetlands. In wetland A there is a relatively wide and diverse emergent wetland fringe. Wetland B, on the other hand, has a very narrow emergent fringe. In the diagrams on the right the symbols represent different vegetation communities. In both cases the plots are located at the emergent/aquatic vegetation interface to capture as many of the different vegetation types as possible.

Development and Validation of Temperate Prairie Wetland IBIs 108

B.3. Establish the sample location: After the major plant communities have been identified, determine a location where the sampling plot(s) can be placed that would best capture or represent the vegetation types found in the wetland. Typically, this is at the emergent/aquatic vegetation interface (Fig. 1). If the wetland has predominantly emergent vegetation, locate the sample plot(s) in the wettest location of the wetland. If there is not an extensive emergent community present, locate the sampling plot(s) where the emergent community should be.

B.4. Determine the plot size and shape: Over the course of the development of wetland vegetation monitoring at the MPCA, the sampling methods have evolved to better characterize wetland vegetation and increase the performance of the assessment tools. Because of this evolution, a variety of sampling plot sizes and shapes have been, and continue to be, employed with this procedure. Historically, the MPCA used a single large sampling plot to characterize an entire wetland. The size of the plot was standard (100 m2), but the shape was either square (10 m x 10 m) or rectangular (5 m x 20 m). The 10 m x 10 m plot was used when a wide and well developed emergent vegetation fringe was present. The 5 m x 20 m plot was employed when only a narrow emergent vegetation fringe was present to Red Northern better capture the River Minnesota emergent/aquatic Valley Wetlands vegetation interface. More recently, the Northern Lakes & MPCA has Forest investigated alternative sample techniques. During the 2003 field season, North Central a methods comparison Hardwood Forest was undertaken Northern comparing the use of Glaciated the large single plot Plains Driftless versus a four small (5 Area m x 5 m) plot Western Corn Belt sampling technique Plains (Genet et al. 2005). In this scenario the Figure 2. Level III Ecoregions in Minnesota 2 four small plots (Omernik 1987). Use the 100 m plot size in the survey the same area North Central Hardwood Forest, Western Corn (100 m2) and together Belt Plains, and Northern Glaciated Plains 2 are considered to be Ecoregions and multiple 25 m plots in the one wetland Northern Lakes and Forest Ecoregion. IBIs have vegetation sample. not been developed in the remaining Ecoregions. The four small plot

Development and Validation of Temperate Prairie Wetland IBIs 109

technique was found to approximately double IBI precision and it was decided that it will be adopted as the primary MPCA sampling method. However, the single sampling plot technique continues to be used in the areas of the state where the assessment criteria have been developed using the single plot technique for consistency (North Central Hardwood Forest, Western Corn Belt Plains, and Northern Glaciated Plains Ecoregions; Figure 2).

Once a representative plot location has been identified choose either the single sample plot or the four small plot sampling technique based on which Ecoregion in the state the wetland occurs (Figure 2). If the single plot technique is to be used, determine which plot shape (square or rectangular) is appropriate. As a general rule, only use the 5 m x 20 m rectangular plot shape when the emergent vegetation fringe is < 5 m wide from the upland boundary to the aquatic vegetation/open water boundary.

B.5. Lay-out the plot: To lay-out a plot, first pick a point to be corner #1 and plant one tall gardening stake (Table 1) to mark the corner. Using a tape measure (Table 1) mark off the first side of the plot, according to the dimensions of the determined plot shape, holding the tape measure away from your body and walking outside of the plot area to avoid excessive trampling of the vegetation inside the plot. Stake this point (corner #2). Turn 90 degrees using a compass or best visual judgment, and measure out the second side to corner #3. Repeat these steps, establishing corner #4 and enclosing the plot with four sides. Adjust the corners and sides if necessary. The plot should capture the emergent/aquatic vegetation interface (Fig. 1); therefore, a portion of the plot should be in each vegetation type.

B.6. Record releve information: Once a plot, or releve, has been established, begin recording releve level data in the Releve Data Sheet (attached at the end of this SOP) or PDA. If using the Releve Data Sheet, establish the Releve Number (this is done automatically in the PDA). The releve number consists of the date and time of the beginning of the releve and should have the following format: month/day/year-hours:minutes:seconds. Record the Site Name, Surveyor’s Name, and Date. Determine the Releve Result, or use category of the data. A releve is: Reportable if the data in that releve will be used for the primary assessment for the site; Replicate if the data will be used to determine IBI variance (Genet et al. 2005), for QA/QC purposes, or for secondary assessment; and Nonreportable if the data will not be used for any assessment purposes. If the Nonreportable data use category is selected, document the reason the data should not be used for wetland assessment. Record the Releve Shape. If the releve is 5 m x 5 m, also record a Sample letter (beginning with A) and Subsample number (beginning with 1) for the releve. The Sample letter is needed to group multiple 5 m x 5 m plots (i.e., subsamples) together into groups of four. Determine and record the Average, Maximum, and Minimum Water Depth (cm) within the releve. Estimate the percent cover the genus Carex and Open Water occupies in the plot. Open Water is defined as standing water that does not have emergent or floating vegetation shading it. Record the approximate

Development and Validation of Temperate Prairie Wetland IBIs 110

position of the releve with a handheld Global Positioning System (GPS) unit. Save the waypoint in the GPS with a file name that consists of the Site Name and the Sample and Subsample indicators (if 4 3 necessary). If the site was named prior to 2003, use the first six characters of the site name and sample and subsample indicators (if necessary). If the site was named using the year/county/wetland number coding system adopted in 2003 and currently used, record the GPS File Name according to the following format: 2 digit year, first four letters of the county, 2 digit 1 2 wetland number, sample Figure 3. Walking the plot. Begin at corner #1 letter (if necessary), and and follow the arrows until the entire plot has subsample number (if been observed. necessary).

Example: the GPS File Name for the third 5 m x 5 m plot of the second set of plots (sample) in the site named 04CASS011 should be- 04CASS11B3. If any photographs of the releve are taken, record the appropriate Photo Info in the space provided.

B.7. Identify plants within the plot: Next, inventory the plants within the plot. This is done by ‘walking the plot’ (Fig. 3). Begin in corner #1 and walk just inside the plot toward corner #2. Identify and record plants to the lowest taxonomic division possible in the Species Info section as you proceed. Continue around sides 2 and 3. After passing corner #4 go about 1/3 of the way of the remaining side of the plot and cut through to the opposite side to observe the vegetation in the interior. Once on the opposite side, proceed down another 1/3 of that side and cut through the plot again. Return to corner #1. In very dense emergent vegetation it may be necessary to do a third interior path to be able to observe the entire plot. For the 5 m x 20 m plot shape, 4-5 interior paths may be necessary to complete the plant inventory.

Record a Reliability code (Table 2) for each plant encountered to indicate the level of identification confidence. If there are multiple higher level taxonomic identifications in the same plot belonging to the same group, use the tsnGroup space to differentiate individual species (see B.9). If a plant is collected to be identified in the laboratory, mark the Collected box for that plant.

Development and Validation of Temperate Prairie Wetland IBIs 111

B.8. Estimate cover: For each plant taxa encountered in the plot, estimate the percent cover using the Table 2. Identification cover class (CC) scheme given in Table 3. reliability codes. Reliability Description B.9. Unknown plants: All plants encountered in the Code plot should be identified to its lowest taxonomic 7 Unknown division possible. When a plant cannot be 6 cf Genus 5 Genus certain reliably identified to species in the field, the 4 cf species plant should be recorded using a standard 3 species complex naming convention and be collected for 2 species certain identification later in a laboratory. 1 cf var/subsp. The following notation convention should be 0 variety/subsp. certain used to record unknown plants: 1) the scientific name of the lowest known taxonomic division of the plant (e.g., Genus, Family, etc), and 2) a number corresponding to the number of different unknown plants from that taxonomic division encountered in a particular plot. Record the taxonomic division in the Species Name column and the number in the tsnGroup column.

Example: if one were to encounter an unknown species of Carex, Carex should be recorded as the Species Name and a 1 should be recorded in the tsnGroup space. If another unknown species of Carex is encountered in the same plot, the Species Name should be recorded as Carex and a 2 should be recorded in the tsnGroup space. Note that the unknown naming convention corresponds to the plot level; therefore, all unknowns must be named and collected at each plot. Table 3. Cover Classes and corresponding All unknown plants should be collected and pressed ranges of percent cover. for positive identification. In the field, collect as much Cover Class Percent material as necessary, or possible, to facilitate (CC) Cover Range identification of the plant when pressed and dried, and 8 95-100% place in a plastic bag. Label the bag with: 1) the site 7 75-94% name, 2) releve number, 3) plant unknown name (i.e., 6 50-74% Species Name and tsnGroup number recorded), 4) 5 25-49% date, and 5) collector name. Upon returning to the 4 10-24% vehicle, immediately place collection bags into a 3 5-9% cooler with ice and keep specimens cool until they can 2 2-4% 1 1% be pressed. Collected specimens must be pressed 0.5 0.1-0.9% within 24 hours of collection. 0.1 single/few

C. Data and Equipment Security: Immediately after each day of field sampling, the following actions must be taken to secure the data collected during field sampling and maintain sampling equipment for further use:

Development and Validation of Temperate Prairie Wetland IBIs 112

C.1. Download Data: Download any and all field data from the PDA, GPS, and digital camera onto the hard drive of the laptop computer. Make an additional copy of these files onto a portable memory source (e.g., ‘memory stick’, CD) to back up the files. Delete data as necessary on the individual units to reduce duplicate copies of data from downloading the same data multiple times.

C.2. Press Collected Plant Specimens: Specimens must be pressed within 24 hours of collection. Press specimens with a standard plant press that has cardboard ventilators, blotter paper, and newsprint. Each specimen should be placed in an individual piece of newsprint and labeled with the same label as the collecting bag (see B.9). Array the plant so that stems and leaves and any flowering or fruiting material are separated and clearly visible. Aquatic plants may require floating in a tray filled with water and arrangement on wax paper.

C.3. Ship Water Chemistry Samples: See Water Chemistry SOP.

C.4. Equipment Assessment and Maintenance: Assess and maintain sampling equipment as necessary. Clean soiled sediment tubes (See Water Chemistry SOP). Recharge any flat batteries. Organize, update, and maintain site files and maps. Dry and repair waders as necessary. Acquire fresh ice for cooler.

VIII. QUALITY ASSURANCE AND QUALITY CONTROL

Compliance with this procedure will be maintained through annual internal reviews. Technical personnel will conduct periodic self-checks by comparing their results with other trained personnel. Calibration and maintenance of equipment will be conducted according to the guidelines specified in the manufacturer’s manuals.

In addition to adhering to the specific requirements of this sampling protocol and any supplementary site specific procedures, the minimum QA/QC requirements for this activity are as follows:

A. Control of deviations: Deviation shall be sufficiently documented to allow repetition of the activity as performed.

B. QC samples: Ten percent of sites sampled in any given year are re-sampled as a means of determining sampling error and spatial variability.

C. Verification: The field crew leader will conduct periodic reviews of field personnel to ensure that the procedures detailed in this SOP are being followed.

Development and Validation of Temperate Prairie Wetland IBIs 113

IX. LITERATURE CITED

Almendinger, J.C. 1987. A handbook for collecting releve data in Minnesota. Natural Heritage Program, MN Department of Natural Resources, St. Paul, MN.

Genet, J.A., M. Bourdaghs, and M.C. Gernes. 2005. Advancing Wetland Biomonitoring in Minnesota. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency, Assistance #BG98568800.

Gernes, M.C. and J.C. Helgen. 2002. Indexes of biological integrity (IBI) for large depressional wetlands in Minnesota. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency. Assistance #CD-985879-01.

Shaw, S.P., and C.G. Fredine. 1956. Wetlands of the United States. U.S. Fish and Wildlife Service, Circular 39.

Development and Validation of Temperate Prairie Wetland IBIs 114

Appendix D - Box plots of macroinvertebrate metrics for the 4 least-impacted sites (=reference) vs the 4 most-impacted sites (=impaired), according to ranking by the Human Disturbance Score (HDS), for the 2002 NCHF data set (excluding bogs). 1.0 0.9 0.8 0.9 0.7 0.8 0.6 0.5 0.7 0.4

0.6 0.3 % DOM 3 GENERA % DOM % TOLERANT %TAXA TOLERANT 0.2 0.5 0.1 0.4 0.0 impaired reference impaired reference

0.7 10 9 0.6 8

0.5 7

6 0.4 5 0.3 4

0.2 TAXA ODONATA 3 RATIO CORIXIDAE 2 0.1 1 0.0 0 impaired reference impaired reference

7 10 9 6 8 5 7

4 6 5 3 4 SNAIL TAXA LEECH TAXA 2 3 2 1 1 0 0 impaired reference impaired reference

Development and Validation of Temperate Prairie Wetland IBIs 115

Appendix D continued.

80 9

8 70 7 60 6 5 50 4 TAXA TOTAL 40 3 INTOLERANT TAXA 2 30 1 20 0 impaired reference impaired reference

8 20 7 6 15 5

4 10 ETSD 3

2 5

CHIRONOMIDAE TAXA 1 0 0 impaired reference impaired reference

Development and Validation of Temperate Prairie Wetland IBIs 116

Appendix E - Tolerant/Intolerant Macroinvertebrate Taxa Designations for Minnesota Depressional Wetlands.

In the large depressional wetland macroinvertebrate IBI (Gernes and Helgen 2002) taxa were identified as being either tolerant or intolerant based on a very limited data set. Some of these taxa were not selected empirically, instead they were deemed either tolerant or intolerant based on the results of other studies, which may or may not have been conducted in depressional wetlands. As a result, the %Tolerant Taxa metric that was developed in the large depressional wetland project did not respond to disturbance in the same manner when applied to another (partially independent) data set (e.g., 2002 NCHF wetland sites). Therefore, an examination of the complete MPCA wetland macroinvertebrate database was used to identify taxa that consistently increased with disturbance (linear relationships) or exhibited marked increases in abundance at disturbed sites (non-linear relationships). Even though the Intolerant Taxa Richness metric worked well with additional data sets, a search for additional intolerant taxa was conducted since such taxa could easily be identified in the search for tolerant taxa. Taxa that would make good additions to the Intolerant Taxa Richness metric were identified graphically, looking for those that tended to occur at least-impacted sites but were generally absent from disturbed sites.

Tolerant Taxa

Tolerant taxa were identified as those that were significantly correlated with multiple disturbance measures (e.g., HDS, Cl, N, P, turbidity) or those that on a scatterplot with HDS exhibited greatly exaggerated abundances at disturbed sites (e.g, Figure 1). This process was completed for three different data sets: 1999 NCHF, 2002 NCHF, 2002 WCBP/NGP. Taxa that exhibited the characteristics indicative of tolerance to pollution

0.03

a

t

a

t

c

n

u

p 0.02

a

l

l

e

d

b

o

p

r

E

n 0.01

o

i

t

r

o

p

o

r

P

0.00 0 20 40 60 80 100 120 HDS

Figure 1. Relationship between the proportion of Erpobdella punctata in the dip net sample and Human Disturbance Score (HDS).

Development and Validation of Temperate Prairie Wetland IBIs 117

in all three data sets were selected for inclusive into the new version of the %Tolerant Taxa metric. The idea was to begin to build a general list of tolerant macroinvertebrate taxa for wetlands much like the existing lists for streams (e.g., Hilsenhoff 1987, Barbour et al. 1999). Adamus et al. (2001) has led the field in the identification of tolerant/intolerant wetland invertebrate taxa and the understanding of how specific stressors affect various taxa. The National Database of Wetland Invertebrate Sensitivities to Enrichment and Hydrologic Alteration represents a thorough review of the literature, but it includes research from many types of wetlands spanning North America. Therefore, rather than using this database as a source for selecting tolerant taxa, it was used as a reference to compare the empirically selected taxa against.

The search of the MPCA wetland macroinvertebrate database revealed a total of 10 taxa that consistently exhibited increased abundances at disturbed wetlands in the three data sets that were examined (Table 1). Of the original ten taxa selected for the % Tolerant Taxa metric in the large depressional wetland study, six were selected in this analysis, thus representing a validation of these taxa as tolerant wetland macroinvertebrates. The remaining four taxa from the original metric were not selected here because they either didn’t respond consistently among the data sets or exhibited a negative correlation with disturbance (e.g., Dicrotendipes, Paratanytarsus). In general, the National Database

Table 1. Taxa identified as being tolerant to disturbance in Minnesota depressional wetlands in the large depressional wetland study (Original) and in the current analysis (New).

Original New % Tolerant Taxa Metric % Tolerant Taxa Metric

Amphipoda Erpobdella sp. Erpobdella sp. Trichocorixa sp. Trichocorixa sp. Physa/Physella Physa/Physella Enallagma sp. Cricotopus sp. Cricotopus sp. Endochironomus sp. Endochironomus sp. Dicrotendipes sp. Glyptotendipes sp. Glyptotendipes sp. Paratanytarsus sp. Enochrus sp. Helobdella stagnalis Berosus sp. Decapoda

supported the designation of these taxa as ‘tolerant’. Only Decapoda had evidence in the National Database suggesting that it may be relatively intolerant to pollution. For example, Schwartz (1985) observed a decline in Decapoda abundance as a result of raw sewage input into a Lake Champlain wetland.

Development and Validation of Temperate Prairie Wetland IBIs 118

The % Tolerant Taxa metric based on the empirical selection of taxa performed better than the original metric across the three data sets (Table 2). In the 1999 NCHF data set the two metrics performed equally well based on the number of significant correlations with the disturbance factors. It is the performance of these two metrics in the other data sets (2002 NCHF & Plains) that sets them apart. For instance, in the 2002 NCHF data set the original % Tolerant Taxa metric is significantly correlated to only two factors, Phosphorus and Turbidity (Table 2). However, these relationships are in the opposite direction, decreasing with increasing disturbance. The new metric represents an improvement over these results being positively correlated with HDS and sulfate concentrations and having fewer negative correlations (none statistically significant) with the disturbance factors. In the 2002 Plains data set, the original Tolerant metric was not

Table 2. Pearson correlation coefficients (r) and corresponding significance values (p) for relationships between % Tolerant Taxa metrics and various measures of disturbance. Water chemistry data were Log10 transformed. Bold values highlight significant results at the 0.05 significance level.

Original New % Tolerant Taxa % Tolerant Taxa rp rp

1999 NCHF data set (N=46) Human Disturbance Score 0.232 0.121 0.465 0.001 Chloride 0.136 0.366 0.371 0.011 Kjeldahl Nitrogen 0.466 0.001 0.194 0.196 Phosphorus 0.328 0.026 0.184 0.220 Turbidity 0.366 0.012 0.340 0.021 Chlorophyll A 0.293 0.048 0.371 0.011

2002 NCHF data set (N=17) Human Disturbance Score -0.295 0.251 0.523 0.031 Chloride -0.212 0.414 0.282 0.274 Kjeldahl Nitrogen 0.045 0.865 -0.079 0.763 Phosphorus -0.546 0.023 0.223 0.389 Turbidity -0.509 0.037 0.053 0.840 Sulfate 0.098 0.708 0.646 0.005

2002 Plains data set (N=28) Human Disturbance Score 0.247 0.206 0.318 0.099 Chloride 0.059 0.764 0.215 0.272 Kjeldahl Nitrogen -0.037 0.850 0.541 0.003 Phosphorus -0.201 0.305 0.228 0.243 Turbidity -0.009 0.963 0.430 0.022 Sulfate 0.266 0.171 0.425 0.024

significantly correlated with any of the disturbance measures, whereas the new version of the metric was significantly correlated with nitrogen, sulfate, and turbidity (Table 2).

Development and Validation of Temperate Prairie Wetland IBIs 119

Development of this new % Tolerant Taxa metric represents an initial attempt to create a general list of taxa that are tolerant to pollution in depressional wetlands. As more data becomes available, this list may be supplemented with additional taxa that exhibit similar responses to disturbance.

Intolerant Taxa

The Intolerant Taxa Richness metric developed in the large depressional wetland study continued to work well in the additional data sets it was tested in (2002 NCHF & Plains). However, since the taxa that were being included in the % Tolerant Taxa metric were being selected empirically, we wanted to do the same for the Intolerant Taxa metric. This exercise would increase the likelihood of the continued success of this metric as it becomes tested in other regions of the state, and perhaps other wetland types. Unlike the search for tolerant taxa, this analysis focused on the identification of taxa that were frequently present at the least-impacted sites but were rare or absent at the disturbed sites. Since this is a metric where richness, rather than abundance, will decrease with disturbance, correlations were not adequate for identifying potential taxa to include in this metric. Therefore, the main tool for identifying candidate taxa for this metric was graphical examination of the plots of individual taxa against the Human Disturbance Score (e.g., Figure 2). As in the search for tolerant taxa, this process was completed for three different data sets: 1999 NCHF, 2002 NCHF, 2002 WCBP/NGP. Taxa that exhibited the characteristics indicative of intolerance to pollution in all three data sets were selected for inclusive into the new version of the Intolerant Taxa Richness metric.

0.025

.

p

s 0.020

a

i

n

i

h

r

r

o 0.015 c

u

e

L

f

o 0.010

n

o

i

t

r

o

p

o 0.005

r

P

0.000 0 20 40 60 80 100 120 HDS

Figure 2. Relationship between the proportion of Leucorrhinia sp. in the dip net sample and Human Disturbance Score (HDS).

Of the seven taxa that were included in the original version of the Intolerant Taxa Richness metric, five were retained as a result of the current analysis (Table 3). The two taxa that were not retained, Tanytarsus and Oecetis, did not exhibit appear to be intolerant to disturbance in all three data sets. In fact, there have been several studies

Development and Validation of Temperate Prairie Wetland IBIs 120

indicating that Tanytarsus is a relatively tolerant chironomid genera (e.g., King and Brazner 1999, Leslie et al. 1997, Rader and Richardson 1994). In the literature, the caddisfly genera Oecetis has been generally been described as intolerant of pollution (e.g., Patrick and Palavage 1994, Graves et al. 1998), but was not included in this Intolerant Taxa metric because of its relatively uniform distribution across the disturbance gradient in our data set.

Table 3. Taxa identified as being intolerant to disturbance in Minnesota depressional wetlands in the large depressional wetland study (Original) and in the current analysis (New).

Original New Intolerant Taxa Intolerant Taxa

Triaenodes sp. Triaenodes sp. Ceraclea sp. Sphaeriidae Sphaeriidae Hydroptila sp. Libellula sp. Libellula sp. Oxyethira sp. Leucorrhinia sp. Leucorrhinia sp. Labrundia sp. Procladius sp. Procladius sp. Conchapelopia sp. Tanytarsus sp. Thienemannimyia sp. Oecetis sp. additions: Nanocladius sp. Siphlonurus sp. Lauterborniella sp. Macrobdella decora Microtendipes sp. Donacia sp. Zavreliella sp. Erythemis sp.

The original Intolerant Taxa Richness metric was further revised by supplementing the list with 14 additional taxa (Table 3). These taxa demonstrated either a consistent pattern of presence at the undisturbed sites and absence at the disturbed sites across all three data sets or exhibited such a pattern in one of the data sets but did not have sufficient data in the other data sets to distinguish this pattern. The latter situation was a way to acknowledge the fact that individual species ranges may limit its ability to contribute to this metric in all regions of the state, but this should not preclude it as a potential component of this metric. For instance, some taxa may be strong indicators in one ecoregion but may not be common enough, even in the least-impacted wetlands, in another ecoregion to significantly contribute to this metric. The benefit of having a relatively large list of intolerant taxa for this metric is that it increases the likelihood of that this metric will respond to impacts across ecoregions by having a principal set of intolerant taxa that are sensitive regardless of the ecoregion and supplementary intolerant taxa that may by ecoregion-specific. Therefore, it is assumed that supplementary intolerant taxa will be added to this metric as data is collected from depressional wetlands in other regions of the state.

The new version of the Intolerant Taxa Richness metric was more sensitive to disturbance than the original metric, measured in both the strength of the correlations and the number of significant correlations (Table 4). It is believed that the current revisions, with the option of further additions to the list of taxa included in this metric, make this a

Development and Validation of Temperate Prairie Wetland IBIs 121

stronger metric and increase the likelihood that it will continue to perform well as wetland assessments are expanded to other regions of the state.

Table 4. Pearson correlation coefficients (r) and corresponding significance values (P) for relationships between Intolerant Taxa Richness metrics and various measures of disturbance. Water chemistry data were Log10 transformed. Bold values highlight significant results at the 0.05 significance level.

Original New Intolerant Taxa Intolerant Taxa r P r P

1999 NCHF data set (N=46) Human Disturbance Score -0.655 0.000 -0.733 0.000 Chloride -0.444 0.002 -0.582 0.000 Kjeldahl Nitrogen -0.455 0.001 -0.457 0.001 Phosphorus -0.528 0.000 -0.636 0.000 Turbidity -0.347 0.018 -0.413 0.004 Chlorophyll A -0.403 0.005 -0.368 0.012

2002 NCHF data set (N=17) Human Disturbance Score -0.492 0.032 -0.695 0.001 Chloride -0.089 0.734 -0.369 0.144 Kjeldahl Nitrogen -0.314 0.22 -0.313 0.221 Phosphorus -0.668 0.003 -0.746 0.001 Turbidity -0.332 0.193 -0.428 0.087 Sulfate -0.067 0.799 -0.198 0.447

2002 Plains data set (N=28) Human Disturbance Score -0.327 0.090 -0.499 0.007 Chloride -0.259 0.183 -0.284 0.142 Kjeldahl Nitrogen -0.500 0.007 -0.511 0.005 Phosphorus -0.097 0.624 -0.161 0.412 Turbidity -0.038 0.846 0.057 0.771 Sulfate -0.608 0.001 -0.607 0.001

LITERATURE CITED

Adamus, P., T.J. Danielson, and A. Gonyaw. 2001. Indicators for monitoring biological integrity of inland, freshwater wetlands: a survey of North American technical literature (1990-2000). EPA 843-R-01-Fall2001. U.S. Environmental Protection Agency; Office of Water; Washington, DC.

Barbour, M.T., J. Gerritsen, B.D. Snyder, and J.B. Stribling. 1999. Rapid bioassessment protocols for use in streams and wadeable rivers: periphyton, benthic macroinvertebrates and fish, second edition. EPA 841-B-99-002. U.S. Environmental Protection Agency; Office of Water; Washington, DC.

Development and Validation of Temperate Prairie Wetland IBIs 122

Gernes, M.C. and J.C. Helgen. 2002. Indexes of biological integrity (IBI) for large depressional wetlands in Minnesota. Minnesota Pollution Control Agency, Final Report to U.S. Environmental Protection Agency.

Graves, G.A., D.G. Strom, and B.E. Robson. 1998. Stormwater impact to the freshwater Savannas preserve marsh, Florida, USA. Hydrobiologia 379: 111-122.

Hilsenhoff, W.L. 1987. An improved biotic index of organic stream pollution. The Great Lakes Entomologist, 20:31-39.

King, R.S. and J.C. Brazner. 1999. Coastal wetland insect communities along a trophic gradient in Green Bay, Lake Michigan. Wetlands 19: 426-437.

Leslie, A.J., T.L. Crisman, J.P. Prenger, and K.C. Ewel. 1997. Benthic macroinvertebrates of small Florida pondcypress swamps and the influence of dry periods. Wetlands 17: 447-455.

Patrick, R. and D.M. Palavage. 1994. The value of species as indicators of water quality. Proceedings of the Academy of Natural Science of Philidelphia 145: 55-92.

Rader, R.B. and C.J. Richardson. 1994. Response of macroinvertebrates and small fish to nutrient enrichment in the northern Everglades. Wetlands 14:134-146.

Schwartz, L.N. 1985. The effects of sewage on a Lake Champlain wetland. Journal of Freshwater Ecology 3: 35-46.

Development and Validation of Temperate Prairie Wetland IBIs 123

Appendix F - Box plots of macroinvertebrate metrics for the 4 least- impacted sites (=reference) vs the 4 most-impacted sites (=impaired), according to ranking by the Human Disturbance Score (HDS), for the 2002 Temperate Prairies data set. 12 12

10 10 8 8

6 6 ETSD

4 Taxa Intolerant 4

2 2

0 0 impaired reference impaired reference

12 24

10 20

8 16

6 12 4 Diptera Taxa

Odonata Taxa 8 2

0 4 impaired reference impaired reference

80 0.7

0.6 70 0.5

60 0.4

50 0.3 Total Taxa Total 0.2

40 % Chironomidae 0.1 30 0.0 impaired reference impaired reference

Development and Validation of Temperate Prairie Wetland IBIs 124

Appendix F continued.

0.7 0.7

0.6 0.6

0.5 0.5 0.4 0.4 0.3 0.3 % Tolerant

% Predator 0.2 0.2

0.1 0.1 0.0 0.0 impaired reference impaired reference

1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 % Pleidae Corixidae Ratio 0.2 0.2 0.1 0.1 0.0 0.0 impaired reference impaired reference

Development and Validation of Temperate Prairie Wetland IBIs 125

Appendix G - Macroinvertebrate IBI scores for semi-permanent and permanent depressional wetlands sampled in 2002 and 2003 from the Temperate Prairies ecoregion. Bold text indicates reference sites. 2002 2003 Site Area (ha) HDS IBI-rep1 IBI-rep2 IBI-rep1 IBI-rep2 Milan 39.92 64 61 Kerk 1.62 16 67 69 Yohi 17.19 37.5 57 LakeCharlotte 22.53 45 56 61 Hoffman 13.86 59.5 46 49 29 FrancoWMA 24.01 79 17 24 15 Carex2 0.8 15.5 47 Manchester 23.09 32.5 62 Bryclyn 1.05 56 57 EastlickMarsh 21.08 40 31 GreatOasisWMA 4.76 24 64 TylerWMA 41.56 58.5 38 37 41 RostWMA 27.28 55.5 45 FurgameWMA 21.58 51.5 54 Prairie Marsh 13.93 10 67 67 55 LyonsWMA 10.48 66.5 31 LoneTreeWMA 30.84 77 19 RolhiksWMA 56.03 61 51 46 WillowLake 63.82 65 55 37 47 LakeElisabeth 8.08 21 38 OakGlenEast 18.31 28.5 63 OakGlenWest 22.72 20 73 67 BarryWMA 7.4 52 55 38 Malta 16.99 58 32 GoldenWPA 22.23 53.5 39 40 Lee 10.94 54.5 44 30 Hancock 10.38 49 66 03linc019 13.87 52.5 30 03linc018 3.38 66 44 03lyon082 198.2 45 42 03linc073 3.7 63.5 27 24 03lyon070 31.2 48.5 40 03murr028 7.7 57.5 23 03murr066 28.67 54 17 24 03lyon099 9.67 52 25 03murr101 2.6 57 31 03murr132 9.1 47.5 64 03pipe055 2.9 63 29 03lyon146 16 61 11 03redw008 7.8 43.5 39 03redw123 10.1 59.5 45 03redw094 13.6 54.5 34

Development and Validation of Temperate Prairie Wetland IBIs 126

Appendix G continued.

2002 2003 Site Area (ha) HDS IBI-rep1 IBI-rep2 IBI-rep1 IBI-rep2

03lyon012 2.1 56.5 39 03linc138 0.83 75.5 43 03linc122 5.4 67 35 03linc093 7.5 73 55 03lyon110 4.1 41 52 03linc097 14.3 62 30 03linc089 9.2 52 39 03lyon084 3.1 38 69

Development and Validation of Temperate Prairie Wetland IBIs 127

Appendix H - Box plots of potential macroinvertebrate metrics for the 4 least-impacted sites (=reference) vs the 4 most-impacted sites (=disturbed), according to ranking by the HDS, for the 2003 Temperate Prairies seasonal wetlands.

0.5 0.15

0.4 0.10 0.3

0.2

% Hemiptera DN 0.05 % Coleoptera BT 0.1

0.0 0.00 disturbed reference disturbed reference

0.7 0.25

0.6 0.20 0.5

0.15 0.4

0.3 0.10

% Physidae DN 0.2 % Sphaeriidae DN 0.05 0.1

0.0 0.00 disturbed reference disturbed reference

1.0 0.6

0.9 0.5 0.8 0.4 0.7 0.6 0.3

0.5

% Insect BT % Insect 0.2 0.4 0.1 0.3 Chironomini:Chironomidae Prop. 0.2 0.0 disturbed reference disturbed reference

Development and Validation of Temperate Prairie Wetland IBIs 128

Appendix H continued.

1.0 0.3 0.9

0.8 0.2 0.7 0.6 0.1 0.5 Chironomidae Proportion DN % Collector-Gatherer Orthocladiinae+Tanytarsini: 0.4

0.3 0.0 disturbed reference disturbed reference

0.15 0.4

0.3 0.10

0.2

0.05 DN % Predator DN % Herbivore 0.1

0.00 0.0 disturbed reference disturbed reference

0.8 80

0.7

0.6 70

0.5 0.4 60 0.3

0.2 Richness Taxa Total 50 % Tolerant Taxa DN 0.1 0.0 40 disturbed reference disturbed reference

Development and Validation of Temperate Prairie Wetland IBIs 129