Addressing Wetland Conservation Issues by Combining Remote Assessment with Intensive Field Sampling

Home , Malus coronaria

CLARKSON UNIVERSITY

Addressing Wetland Conservation Issues by Combining Remote Assessment with Intensive Field Sampling.

A Dissertation

Kinga M. Stryszowska

Institute for a Sustainable Environment

Submitted in partial fulfillment of the requirements

for the degree of

Doctor of Philosophy, Environmental Science and Engineering

April 14, 2016

Accepted by the Graduate School

______, ______Date Dean of Graduate School The undersigned have examined the thesis/dissertation entitled “Addressing wetland conservation issues by combining remote assessment with intensive, field sampling” presented by Kinga M. Stryszowska, a candidate for the degree of Doctor of Philosophy (Environmental Science and Engineering), and hereby certify that it is worthy of acceptance.

______Date Tom Langen, Ph.D.

______Date Michael Twiss, Ph.D.

______Date Michelle Crimi, Ph.D.

______Date Glenn Johnson, Ph.D.

______Date James Schulte, Ph.D.

ABSTRACT

Wetland ecosystems are delicate and unique systems providing services that are important to the ecological landscape. Wetlands in the United States have been experiencing a gradual degradation over the last two centuries and today a growing body of research is directed and understanding, protecting, and restoring these systems. The primary goal of this was to explore two methods of studying freshwater wetlands: a remote landscape level method and an intensive field sampling method to gain understanding of how each method can enhance the other. Field sampling is a traditional way of studying ecosystems and there is a considerable body of knowledge on various techniques, metrics, and indicators as they apply to wetlands. Remote sensing is a relatively new method of studying ecosystems that grew with the advancement of aerial photography and software capable of processing large data sets. In this dissertation

Chapter I details a study comparing anuran, bird, fish, plant, and water quality metrics between

17 wetlands located in an environmentally degraded Great Lakes Area of Concern (AOC) and 10 natural reference wetlands. GIS landscape analysis was performed on all field sampled wetlands to discern some of the drivers of diversity differences between the two areas. The results indicated that whereas the AOC was not significantly different from reference sites in terms of ecological indicators, some landscape level components were significantly difference between the two areas. In Chapter II, a species distribution model (SDM) for the threatened Blanding’s turtle (Emydoidea blandingii) was performed using a combination of 14 years of field capture records and surrounding landscape variables. Using the mapping and analysis software ArcGIS, two types of models were constructed and validated: a generalized linear model (GLM) using presence/absence records and Maxent model using presence/background records. The results indicated that GLM was not as successful as Maxent at predicting habitat suitability for E.

iii blandingii and that the range of the species is limited by factors related to elevation. In Chapter

III correlation analyses were performed on three levels of wetland bioassessment methods; remote sensing, rapid assessment, and exhaustive field sampling, to determine the level of overlap between these methods. Results indicated that some higher effort (Level Three) metric correlated strongly and significantly with lower effort (Level One) metrics suggesting that one

Level could replace another if needed. When studying wetlands, it is evident that a combination of traditional field sampling methods and modern remote sensing methods provides the best assessment of the integrity of the wetland ecosystem.

ACKNOWLEDGMENTS

Over the past five years I have received support and encouragement from a great number of individuals. I would like to express my deep appreciation and gratitude to my advisor Dr. Tom

A. Langen. He has been my sounding board, my cheerleader, my mentor, and ultimately a great friend. I would like to thank my graduate committee of Dr. Michael Twiss, Dr. Glenn Johnson,

Dr. Michelle Crimi, and Dr. James Schulte for their insightful comments and direction as I moved through the process of developing my ideas.

Thank you to all my field technicians without whom this research would have stood still:,

Jon Podoliak, Jeremy Ozolins, Kate Gilpin, Lorianny Rivera, Mitchell Laughlin, Amy Hait, and

Stephen Kelso. I would also like to acknowledge the St. Lawrence River Research and Education

Fund, Northern New York Audubon, the Joseph and Joan Cullman Conservation Foundation, the

New York State Wetland Forum and the Clarkson University Department of Biology for providing funding in support of my research and education.

A special thank you is due to my friends Catherine Benson, Stefanie Kring, Angelena

Ross, Brendan Carberry, and Reshica Baral. Their company, conversation, laughter, and encouragement have made my graduate career a true pleasure.

Last but not least, I would like to express deep gratitude to my family: my fiancé

Nathaniel Hill for his patience, unwavering support, and unconditional love, my daughter Lila

Hazel Hill for giving me the best reason to strive to be a female role model, to my parents Dana and Stanley Stryszowski, and brothers Lukas and Tom Stryszowski for always believing that I can achieve anything I set out to do, and to my sisters Martyna Zmijewska and Melissa Hazlett for being there with a smile and a cheer every step of the long way.

TABLE OF CONTENTS

ABSTRACT ...... iii

ACKNOWLEDGMENTS ...... v

LIST OF TABLES ...... x

LIST OF FIGURES ...... xi

LIST OF APPENDICES ...... xii

INTRODUCTION...... 1

References ...... 5

CHAPTER I: Evaluating Beneficial Use Impairments in Wetlands of the Massena Area of

Concern Using Biotic, Water Quality, and Landscape Indicators...... 7

Abstract ...... 7

Keywords ...... 8

Introduction ...... 8

Methods ...... 10

Site Description ...... 10

Indicators of Wetland Habitat Quality and Wildlife Population Status ...... 12

Biotic Indicators...... 13

Indices of biotic integrity (IBI) ...... 16

Water quality indicators ...... 17

Landscape analysis ...... 19

Data analysis ...... 19

Results and Discussion ...... 20

Biotic indicators ...... 21

Birds ...... 21

Fish ...... 22

Anurans ...... 23

Vascular Plants ...... 23

Water Quality ...... 24

Landscape ...... 24

Selection of Reference Sites ...... 26

Management recommendations ...... 26

Conclusions ...... 33

Acknowledgements ...... 34

References: ...... 35

CHAPTER II: Species distribution modeling of the threatened Blanding’s Turtle’s

(Emydoidea blandingii) range edge as a tool for conservation planning ...... 40

Abstract ...... 40

Key words ...... 41

Introduction ...... 41

Materials and Methods ...... 43

Study Area...... 43

Data Sources...... 43

Model Building ...... 46

Model Evaluation ...... 48

Projection ...... 48

Results ...... 48

vii

Variable Contributions ...... 49

Habitat Suitability Predictions within the Study Region ...... 50

Habitat Suitability Predictions Projected outside the Study Region ...... 51

Discussion ...... 52

Conclusions and Management Implications ...... 58

Acknowledgments ...... 59

References ...... 60

CHAPTER III: Effectiveness of three-tier wetland assessments at evaluating the ecological integrity of wetlands in the St. Lawrence Valley, New York...... 64

Introduction ...... 64

Materials and Methods ...... 69

Study area ...... 69

Level One: Landscape Assessment ...... 70

Level Two: Rapid Assessment ...... 72

Level Three: Intensive Site Assessment ...... 73

Statistical analysis ...... 77

Results ...... 78

Level One metrics ...... 79

ORAM ...... 80

Level Three metrics ...... 81

Composite PCs ...... 81

Discussion ...... 82

Applications ...... 89

viii

References ...... 92

CONCLUSION ...... 97

References ...... 101

APPENDIX A ...... 102

APPENDIX B ...... 113

APPENDIX C ...... 117

APPENDIX D ...... 121

LIST OF TABLES

Table 1.1 — Summary of statistical tests comparing individual biotic, water quality, and landscape metrics between wetlands in St. Lawrence River at Massena AOC and reference wetlands in Louisville. Significant results (Bonferroni corrected biotic: p<0.004, water quality: p<0.002, landscape: p<0.007) are indicated in bold. Square root and natural log data transformations are noted in parentheses. Superscripts indicate the statistical test applied: aStudent’s t-test, bWelch’s test, cMann-Whitney test, dnested mixed factor ANOVA, eANCOVA...... 30

Table 2.1 — Eleven predictor variables used to build the SDMs for E. blandingii in New York………………………………………………………………………………………...……46

Table 2.2 — Performance of GLM and Maxent models at 250-m and 8,000-m scales. The mean + SD AUC is reported for model fit to 75% training and 25% independent validation data (n = 3)……………………………………………………………………………………...………….49

Table 3.1 — The NLCD land use class emergy coefficients equated to the most equivalent Florida land use classes. The coefficients were used to calculate the LDI for each wetland in this study………………………………………………………………………………...……………72

Table 3.2 — Sixteen wetland assessment metrics, grouped into three levels of intensity, used to rank 71 wetlands on a gradient of ecological integrity. Each metric is followed by the possible range of scores for that metric where the higher score indicates higher ecological integrity, except for LDI where a higher score indicates lower ecological integrity. For some metrics, the maximum obtainable metric value is dictated by the local species pool. …………………………...……….79

Table 3.3 — Pearson and Spearman Rank correlations between pairs of 16 wetland assessment metrics. Bonferroni corrected significant results (p < 0.0004) are in bold………….……………83

LIST OF FIGURES

Figure 1.1 — Surveyed wetlands in the St. Lawrence River at Massena Area of Concern and reference area in Louisville, NY…………………………………………..…………………...…13

Fig. 2.1 — Regional limits for SDM in New York’s St. Lawrence River Valley, USA, and E. blandingii survey locations………………………………………………………………………44

Figure 2.2 — Mean importance of predictor variables for (A) GLM and (B) Maxent models at 250-m and 8,000-m scales. Error bars = SD, polarity symbols (+/-) indicate direction of effect on habitat suitability. Missing bars indicate variables not included in the models. Importance in GLM models is expressed as the regression coefficient after z-value standardization; higher coefficients are more important. Importance in Maxent models is expressed as percent contribution to the model……………….………………………………………………………………………..…..50

Figure 2.3 — Probability of occurrence of E. blandingii within the modeled region of northeastern New York using GLM or Maxent with buffer distances of 250 m or 8,000 m. Gray areas indicate high probability of occurrence. Circled area is a predicted gap between two areas of high probability of occurrence…………………………………………………………………………51

Figure 2.4 — Projected areas of high habitat suitability for E. blandingii outside of the model building extent using Maxent (A) and GLM (B) with a buffer distance of 250 m. The dotted line delineates the current known distribution of E. blandingii in New York………………………...54

Figure 2.5 — Projected areas of high habitat suitability for E. blandingii outside of the model building extent using Maxent (A) and GLM (B) with a buffer distance of 8,000 m. The dotted line delineates the current known distribution of E. blandingii in New York…………….…………...55

Figure 3.1 — Seventy-one surveyed palustrine wetlands sites in the St. Lawrence River Valley, New York……………………………………………………………………………...…………70

Figure 3.2 — The gradients of possible conditions of wetland ecological integrity as captured by the 16 metrics used in this study. For all metrics, except LDI, low metric scores indicate poor ecological integrity……………………………………………………..………………………...80

Figure 3.3 — Scatterplots of Composite Level 1 and 3 scores and ORAM. Level 1 Composite and ORAM (Spearman r=0.61, p<0.00001), Level 3 Composite and ORAM (Pearson r=0.41, p<0.00001), and Level 1 Composite and Level 3 Composite (Spearman r=0.43, p<0.00001)….85

LIST OF APPENDICES

Table A1.1 — Species of Greatest Conservation Need (SGCN) found in Massena AOC and the reference site in Louisville, NY…………………………………………………………………103

Table A1.2 — Scientific and common names of all bird species identified in the Massena AOC and Louisville study wetlands. This list includes birds observed outside of the designated 100 meter survey radius. Species of Greatest Conservation Need are denoted by an *………….…...104

Table A1.3 — Scientific and common names of the fish species identified in the study wetlands……………………………………………………………………………………..….106

Table A1.4 — Scientific and common names of all frog species identified in the study wetlands. Highlighted are SGCN species. Species of Greatest Conservation Need are denoted by an *………………………………………………………………………………………………....107

Table A1.5 — Scientific and common names of the vascular plant species identified in the Massena AOC and Louisville study wetlands. Species classified as non-native by the USDA Plants Database are designated by an *. Species classified by the New York State Department of Environmental Conservation as aggressive invasive are designated by a †. The rank represents whether or not the species is classified as a wetland dependent (OBL), facultative wetland (FACW), facultative (FAC), or upland (UPL) species…………………………………….……108

Table A2.1 — Checklist used to score 71 wetland sites for the Minnesota Disturbance Gradient (MDG), a level 1 metric (Gernes and Helgen, 2002). The maximum obtainable score is 75 points……………………………………………………………………………..……………114

Table A2.2 — Checklist used to score 71 wetland sites for the Ohio Disturbance Gradient (ODG), a level 1 metric (Lopez and Fennessy, 2002)……………………………….………………….115

Table A2.3 — Sample Ohio Rapid Assessment Method (ORAM) worksheet (Mack, 2001b), a level 2 metric, used to score 71 wetland sites. Wetland class was either Forested, Emergent, or Shrub/Scrub as classified by Cowardin and others (1979). Metric 5 received an automatic score of ten if: two out of four GIS based significant habitat areas were present (Bird Conservation Region (BCR13) waterbird, shorebird, landbird focus areas, NYS significant coastal habitat). Metrics completed in the laboratory are designated by an * and metrics completed in the field are designated by a †……………………………………………………………….…………..…...116

Table A3.1 — Site scores for 71 wetlands using 16 assessment metrics of three levels of intensity………………………………………………………………………………………... 118

xii

Table A4.1 — Eigenvalues for the composite Principal Component Analysis of Level One metrics………………………………………………………………………………………...…12 2 Table A4.2 — Importance of components and eigenvector loadings for the composite Principal Component Analysis of Level One metrics………………………………………...………..….122

Table A4.3 — Site scores for 71 wetlands based on the first principal component of the composite Level One Principal Component Analysis……………………………………..……123

Table A4.4 — Eigenvalues for the composite Principal Component Analysis of Level Three metrics. Metric WQI was excluded from this analysis………………………………..…..125

Table A4.5 — Importance of components and eigenvector loadings for the composite Principal Component Analysis of Level Three metrics……………………………………………..…..…125

Table A4.6 — Site scores for 71 wetlands based on the first principal component of the composite Level Three Principal Component Analysis……………………………………….…………....126

Figure A4.1 — Principal component axes for the composite Level One analysis………..……..122

Figure A4.2 — Principal component axes for the composite Level Three analysis……..…..…125

xiii

INTRODUCTION

The U.S. Fish and Wildlife service estimates that over a period of 200 years, the conterminous

Unites States lost approximately 53% of their original wetlands (Dahl, 1990). In the 19th and early 20th centuries wetland conversion, destruction, and drainage were common-place and even encouraged by government agencies. Attitudes towards wetlands began to change during the

1960’s and 1970’s as the recognition of wetland ecosystem services increased. Today, wetlands are protected under the federal Clean Water Act, the 1989 federal mandate of “no net wetland loss”, and state legislation.

Wetlands are a unique ecosystem, structured by a combination of geology, biochemistry, biota, and hydrology. There is a great variety of wetlands in the world, each one composed of a different combination of these elements. As a semi-aquatic system, wetlands at times create a transition between deep water and upland ecosystems and other times exist as isolated patches in the landscape, recharged by groundwater or precipitation. Despite this variety of structures, all wetlands perform a series of similar and important ecosystem functions. These functions are specific to wetlands themselves and are performed by the wetlands whether or not humans put any value on them. In the past several decades however humans have developed an increasing appreciation for, and attribute an increasing value to, these naturally performed functions. The most general functions performed by wetlands are those of primary production, food web diversity, litter decomposition, and nutrient cycling (Van der Valk, 2012). Additionally, there are more specific functions such as sediment removal, groundwater recharge, and attenuation of water (Lewis, 2001). In recognition of these functions, society has started a movement towards the restoration of degraded wetlands and preservation of the remaining ones.

The position of a wetland in the landscape has much influence on the structure and the ecosystem functions that the wetland can perform. Water quantity, velocity and direction, nutrient inputs and exports, sediment transport, and all biota occupying a wetland, are a reflection of the landscape setting (Fretwell et al., 1996; Bedford, 1999). The types of land use adjacent to a wetland and in the watershed can affect wetland biodiversity (DeLuca et al., 2004;

Houlahan et al., 2006; Mensing et al., 1998) and water quality (Lenat and Crawford, 1994;

Houlahan and Findlay, 2004). The type, size, quantity, and distribution of the natural habitat surrounding a wetland can also have an effect on a wetland’s biological assemblages (Tozer et al., 2010; Tsai et al., 2012). Finally the proximity of other wetlands, the connectivity between wetlands, and the distribution of wetland types in the landscape influences both wetland structure and function (Attum et. al., 2007; Johnson, 2005; Cosentino et al., 2010; Quesnelle et al., 2013).

While many landscape features naturally govern the structure and function of wetlands, the landscape can also be a significant source of wetland impairment. Degradation from the landscape can come in many forms and degrees of severity. Some of the most commonly documented landscape impacts are nutrient enrichment from agricultural land uses, sedimentation and chemical runoff from impervious surfaces, and an exacerbation of both these impacts through the removal of vegetative buffers.

Since the implementation of wetland protection legislation, the fields of wetland ecology, restoration, conservation, mitigation, monitoring, and management have experienced a surge in research and associated knowledge. In addition, with the advance of computer software capable of processing large datasets along with the development of high resolution aerial photography, remote sensing and spatial model building became a new way to analyze the landscape. Spatial analysis using Geographic Information System (GIS) software has revealed that land use types

2 do in fact have an effect on wetland water quality and biodiversity and that even landuse at the watershed level can have implications on localized wetland quality (Tiner, 2004; Brooks et al.,

2004). Wetland engineers now attempt to understand the landscape setting before restoring wetlands (Mitsch and Gosselink, 2007; Bedford, 1999). Spatial analysis has given researchers a glimpse at the macro scale and also allowed for a rapid analysis with a greater spatial and temporal extent than was possible before. Today, wetland research often falls into two camps; intensive on site sampling and remote spatial modeling. These two research directions have both greatly enhanced our current wetland knowledge, however they rarely come together. Landscape level models do not have the data resolution and level of detail obtained through the traditional methods of intensive sampling while filed sampling does not capture the landscape setting and the large scale patterns that may be drivers of wetland attributes. It is when these two approaches come together however that the assessment of a wetland really comes into focus. Some of the best examples of such endeavors have come from research done on coastal wetlands of the Great

Lakes basin. The integration of field based methods and remote sensing have allowed multiple stakeholders and agencies to develop GIS wetland assessment models (Lopez et al., 2006) and ecological and water quality indices of biological integrity (Chow-Fraser, 2006; Niemi et al.,

2009; Smith-Cartwright and Chow-Fraser, 2011). The result has been a basin-wide characterization of the condition of coastal wetlands with a better understanding of the large scale spatial and temporal drivers of changes in their condition.

The principal goal of this PhD thesis research is to empirically combine freshwater wetland research in the St. Lawrence River Valley, NY at the landscape level with intensive field sampling to gain understanding of how each direction of research can enhance the other. Chapter

I details a study comparing anuran, bird, fish, plant, and water quality metrics (e.g., species

3 richness, abundance, diversity) between 17 wetlands located in an environmentally degraded

Great Lakes Area of Concern (AOC) and 10 natural reference wetlands. GIS landscape analysis is performed on all field sampled wetlands to discern some of the drivers of diversity differences between the two areas. In Chapter II, a species distribution model (SDM) for the threatened

Blanding’s turtle (Emydoidea blandingii) is performed using a combination of 14 years of field capture records and surrounding landscape variables. This combination of information is meant to understand the drivers of range limits of the turtle at the edge of its eastern distribution. In

Chapter III, exploratory correlation analyses are performed on three levels of wetland bioassessment methods; remote sensing, rapid assessment, and exhaustive field sampling, to determine the level of overlap between these methods.

References Attum, O., Lee, M., Roe, J.H., Kingsbury, B.A. 2007. Upland-wetland linkages: relationship of upland and wetland characteristics with watersnake abundance. Journal of Zoology. (271): 134-139. Bedford, B.L. 1999. Cumulative Effects on Wetlands Landscapes: Links to Wetland Restoration in the United States and Southern Canada. Wetlands. (19): 775-788. Brooks, R.P., Wardrop, D.H., Bishop, J.A. 2004. Assessing Wetland Condition on a Watershed Basis in the Mid-Atlantic Region Using Synoptic Land-Cover Maps. Environmental Monitoring and Assessment (94): 9-22. Chow-Fraser, P. 2006. Development of the Water Quality Index (WQI) to Assess Effects of Basin-wide Land-use Alteration on Coastal Marshes of the Laurentian Great Lakes. In Eds. Simon TP and Stewart PM. 2006. Coastal Wetlands of the Laurentian Great Lakes: health, habitat and indicators. AuthorHouse: Bloomington IN, 1st edition. Cosentino, B.J., Schooley, R.L., Phillips, C.A. 2010. Wetland hydrology, area, and isolation influence occupancy and spatial turnover of the painted turtle, Chrysemys picta. Landscape Ecology. (25): 1589-1600. Crewe, T.L., Timmermans, S.T.A., Jones, K.E. 2006. The Marsh Monitoring Program 1995 to 2004: A Decade of Marsh Monitoring in the Great Lakes Region. Published by Bird Studies Canada in cooperation with Environment Canada. Dahl, T.E. 1990. Wetland Losses in the United States 1780's to 1988's. U.S. Department of the Interior, Fish and Wildlife Service, Washington n. D.C. Dahl, T.E. 2011. Status and trends of wetlands in the conterminous United States 2004 to 2009. U.S. Department of the Interior; Fish and Wildlife Service, Washington, D.C. DeLuca, W.V., Studds, C.E., Rockwood, L.L., Marra, P.P. 2004. Influence of Land Use on the Integrity of Marsh Bird Communities of Chesapeake Bay, USA. Wetlands. (24):837-847. Fretwell, J.D., Williams, J.S., Redman, P.J. 1996. National Water Summary on Wetland Resources. U.S. Geological Survey. Water-Supply Paper 2425. Houlahan, J.E., Findlay, S.C. 2004. Estimating the ‘critical’ distance at which adjacent land-use degrades wetland water and sediment quality. Landscape Ecology. (19): 677-690. Houlahan, J.E., Keddy, P.A., Makkay, K., Findlay, C.S. 2006. The effects of adjacent land use on wetland species richness and community composition. Wetlands. (26):79-96. Johnson, J.B. 2005. Hydrogeomorphic Wetland Profiling: An Approach to Landscape and Cumulative Impact Analysis. EPA-620-R-05-001. U.S. Environmental Protection Agency, Washington DC. Lenat, D.R., Crawford, J.K. 1994. Effects of Land Use on Water Quality and Aquatic Biota of Three North Carolina Piedmont Streams. Hydrobiologia. (294):185-199. Lewis, Jr. W.M. 2001. Wetlands Explained; Wetland Science, Policy, and Politics in America. Oxford University Press, New York. Lopez, R.D., Heggem, D.T., Sutton, D., Ehli, T., Remortel, R.V., Evanson, E., Bice, L. 2006. Using Landscape Metrics to Develop Indicators of Great Lakes Coastal Wetland Condition. U.S. Environmental Protection Agency, Office of Research and Development, Washington DC. EPA/600/X-06/002. Mensing, D.M., Galatowitsch, S.M., Tester, J.R. 1998. Anthropogenic effects on the biodiversity of riparian wetlands of a northern temperate landscape. Journal of Environmental Management. (53): 349-377.

Mitsch, W.J., Gosselink, J.G. 2007. Wetlands. Fourth Edition. New Jersey: John Wiley & Sons, Inc. Niemi, G.J., Brady, V.J., Brown, T.N., Ciborowski, J.J.H., Danz, N.P., Ghioca, D.M., Hanowski, J.M., Hollenhorst, T.P., Howe, R.W., Johnson, L.B., Johnston, C.A., Reavie, E.D. 2009. Development of ecological indicators for the U.S. Great Lakes coastal region- A summary of applications in Lake Huron. Aquatic Ecosystem Health and Management. (12): 77-89. Quesnelle, P.E., Fahrig, L., Lindsay, K. 2013. Effects of habitat loss, habitat configuration and matrix composition on declining wetland species. Biological Conservation. (160): 200- 208. Smith-Cartwright, L.A., Chow-Fraser, P. 2011. Application of the Index of Marsh Bird Community Integrity to Coastal Wetlands of Georgia Bay and Lake Ontario, Canada. Ecological Indicators. (11): 1482-1486. Tiner, R.W. 2004. Remotely-sensed indicators for monitoring the general condition of “natural habitat” in watersheds: an application for Delaware’s Nanticoke River watershed. Ecological Indicators. (4): 227-243. Tozer, D.C., Nol, E., Abraham, K.F. 2010. Effects of local and landscape-scale habitat variables on abundance and reproductive success of wetland birds. Wetlands Ecology and Management. (18):679-693. Tsai, J., Venne, L.S., Smith, L.M., McMurry, S.T., Haukos, D.A. 2012. Influence of local and landscape characteristics on avian richness and density in wet playas of the southern Great Plains, USA. Wetlands. (32):605-618. Van der Valk, A.G. 2012. The Biology of Freshwater wetlands. Second Edition. Oxford University Press, New York.

CHAPTER I: Evaluating Beneficial Use Impairments in Wetlands of the Massena Area of Concern Using Biotic, Water Quality, and Landscape Indicators.a

Kinga M. Stryszowska1, Michael R. Twiss2, Tom A. Langen2

1Institute for a Sustainable Environment, Clarkson University, Potsdam, NY 13699, USA. 2Department of Biology, Clarkson University, Potsdam, NY 13699, USA.

Abstract

Wetlands along the St. Lawrence River were severely impacted by habitat alteration and contamination as a consequence of construction of the St. Lawrence Seaway and the Moses-

Saunders power dam, and associated industrial development. Due to environmental degradation, the St. Lawrence River at Massena, New York has been designated as an Area of Concern

(AOC) in the Laurentian Great Lakes. Within this AOC, there is an information gap on the current status of two Beneficial Use Impairments (BUIs): (1) loss of fish and wildlife habitat and

(2) degradation of fish and wildlife populations. Both BUIs have the same evaluation endpoint: no difference between the AOC and comparable reference areas outside of the AOC. We evaluated coastal and palustrine wetland habitats by surveying biological, water quality, and landscape environmental quality indicators within a sample of 17 wetlands in the AOC and 10 reference wetlands outside the AOC to establish georeferenced indices of biotic integrity and water quality. We did not detect a difference between the AOC and reference wetlands in any of the 14 biotic and 16 water quality metrics, but did find a difference in landscape setting. AOC wetlands were smaller and fewer, especially for woody wetlands. These results suggest that wildlife habitat quality and communities are not impaired in AOC wetlands yet it would be

______aStryszowska K.M., Twiss M.R., Langen T.A. 2016. Evaluating Beneficial Use Impairments in wetlands of the Massena Area of Concern using biotic, water quality, and landscape indicators. Journal of Great Lakes Research. In Press.

7 beneficial for additional key fish and wildlife assemblages and habitat types to be surveyed, multi-year monitoring of key biotic indicators implemented, and more specific redesignation criteria for wetland habitat and matrix landscape composition defined and met.

Keywords

Biotic indicators, ecological integrity, ecosystem health, environmental management

Introduction

Many aquatic habitats within the Laurentian Great Lakes watershed have suffered degradation due to population growth and industrialization since the 19th century (Dahl, 1990; Environment

Canada and the USEPA, 2005). As a result, in 1987, the International Joint Commission identified 43 Great Lakes Areas of Concern (AOC) throughout the United States and Canada pursuant the Great Lakes Water Quality Agreement. Within each AOC, fourteen Beneficial Use

Impairments (BUIs) were assessed in the initial process (Stage I) to identify degradation. The St.

Lawrence River (SLR) at Massena (NY) and Cornwall (ON) bi-national AOC was originally listed because of elevated levels of heavy metals and polychlorinated biphenyls (PCB) in the waters, fish tissue, and sediments of the SLR and local tributaries (NYSDEC, 1990). The SLR at

Massena AOC currently has two BUIs and five likely and unknown BUIs; it also has a unique

15th BUI addressing trans-boundary impacts (NYSDEC, 2006b).

Following the drafting and implementation of the Stage II Remedial Action Plan (RAP), the SLR at Massena AOC Remedial Action Committee (RAC) concluded that data remained insufficient on the present status of the confirmed BUI No. 14, loss of fish and wildlife habitat

8 and the suspected BUI No. 3, degradation of fish and wildlife populations (NYSDEC, 2006a).

These two related BUIs are known or suspected to be affected by the physical disturbance associated with the construction and use of the St. Lawrence Seaway and the Moses-Saunders hydroelectric power dam, water quality degradation, and sediment contamination originating from nearby industrial sites. Affected habitats include shorelines and riparian zones, fish spawning beds, fish passages between the main-stem SLR and tributaries, nesting and feeding areas for birds, and inland wetlands. Affected wildlife populations include any that bioaccumulate contaminants such as PCB and mercury, and those known or suspected to be affected by physical and biotic changes to the river, its shoreline, and its coastal and nearby inland wetlands (NYSDEC, 2006a).

Previously, various entities have been monitoring water quality and animal populations within the limnetic and littoral zones of the SLR at Massena AOC and at nearby locations (Bode et al., 2004; NYSDEC, 2013; Twiss et al., 2010). However, much less attention has been given to habitat quality and biological diversity of freshwater wetlands within the AOC. The coastal riverine and palustrine wetlands of the SLR provide essential wintering, migratory stopover, and breeding habitat for waterfowl and other birds, spawning and foraging areas for fish, and critical habitat for many threatened species of plants and animals (NYSDEC, 1990; NYSDEC, 2005).

The purpose of this study is to aggregate data on biodiversity, water quality, and landscape features to determine whether BUI Nos. 3 and No.14 are presently impaired in the wetland component of the SLR at Massena AOC. For the Massena AOC, the redesignation endpoints for both BUI No. 3 and BUI No. 14 are specified by the RAC as no significant differences in fish and wildlife habitat or fish and wildlife communities between the AOC and appropriate reference areas outside the AOC (NYSDEC, 2012). However, little guidance is

9 provided as to what data are valid, necessary, and sufficient to make a determination as to whether the redesignation endpoints have been achieved. Our study is one of very few evaluating

Great Lakes AOCs to focus on BUI Nos. 3 and 14, and the first to use a combination of environmental indicator and GIS-based landscape analyses. Our assessment is intended to provide the information needed by the Massena AOC RAC to evaluate whether these two BUIs can now be classified as unimpaired, or alternatively what forms of environmental mitigation and management are needed to bring about recovery, as the State of New York is working to satisfy the national obligation to the Great Lakes Water Quality Agreement at this AOC. While our study was intended to address information needs for the Massena AOC RAP, our approach and methodology may provide a useful model for other AOCs striving to assess and meet redesignation requirements for these BUIs.

Methods

Site Description

Unless otherwise stated, in this paper the term “AOC” refers specifically to the SLR at

Massena AOC. The wetlands surveyed in our study were located in St. Lawrence County, New

York State, USA in the towns of Massena (pop. dens. 89/km2) and Louisville (pop. dens. 19/km2;

Figure 1.1). Both towns are located on the shore of the SLR and together encompass 310 km2.

The boundary of the AOC includes a 25 km section of the SLR that encompasses the Moses-

Saunders hydroelectric power dam, Long Sault Dam, the Eisenhower and Snell lock system, and sections of three tributaries: the Raquette, St. Regis, and Grasse rivers. The natural land cover of the area is dominated by northern hardwood forest (Gawler, 2008). Massena (including areas of the town outside the AOC) has a higher proportion of developed land cover (15%) than

Louisville (4%), but both towns have equal proportions of agricultural land cover (11%).

Forested and emergent wetlands make up approximately 12% of Louisville and 7% of Massena

(Homer et al., 2015).

The redesignation endpoints for both BUI No. 3 and BUI No. 14 are defined as no significant difference in fish and wildlife habitat and populations between the AOC and reference areas outside the AOC (NYSDEC, 2012). Our study is designed to specifically address these redesignation targets for freshwater wetlands. We selected seventeen wetlands (eight SLR coastal wetlands and nine palustrine wetlands) within the AOC and ten reference sites (five coastal and five palustrine) from outside of the AOC in Louisville, NY (Figure 1.1). Due to access issues, we did not include wetlands located in St. Regis Mohawk Territory or on islands on the SLR. The unbalanced design was the result of focusing efforts toward wetlands in the

AOC; fewer reference sites were assessed due to time and cost limitations.

The choice of a reference site can profoundly affect the outcome of an environmental assessment (White and Walker, 1997). We used three criteria for selecting the reference sites: (1) proximity to the AOC, (2) similarity of attributes of the reference wetlands to AOC wetlands in terms of (a) distance and hydrological context in relation to the SLR, (b) site ownership and stewardship (e.g. private land owner, public natural area), and (c) surrounding landscape land cover and land use, and (3) location of reference sites upstream and upwind of the AOC, to minimize the spillover effects from the AOC on nearby wetlands beyond its borders. We chose the town of Louisville as a location to select reference wetlands because it is adjacent to the

AOC and is located along the SLR. Although upstream and upwind of the AOC, Louisville coastal wetlands will have been similarly impacted as Massena by altered water-level

11 fluctuations associated with historical hydrological modification of the river channel (Wilcox et al., 2007).

We attempted to include a fully representative range of wetland sizes, disturbance histories, and surrounding land uses. Wetlands were located using New York State Department of Conservation (NYSDEC) regulatory freshwater wetland maps (http://gis.ny.gov/, 15,

November 2012), U.S. Fish and Wildlife Service National Wetland Inventory maps

(http://www.fws.gov/wetlands/, 1, November 2012), 2011 USGS National Land Cover Database raster file (Homer et al., 2015), and digital orthoimagery (aerial photos; https://gdg.sc.egov.usda.gov/, 1, November 2012). Site selection was finalized after site visits.

Sites were selected based on wetland type (emergent and scrub shrub; Cowardin et al., 1979), distance from the SLR, and accessibility (i.e., landowner permission, access by land). Preference was given to sites located directly on or near to the SLR (Figure 1.1).

Indicators of Wetland Habitat Quality and Wildlife Population Status

We assessed a total of 14 biotic metrics, 16 water quality metrics measured in each of two seasons, and eight landscape metrics to evaluate the impairment status of wetland habitat and associated fish and wildlife populations. The three categories of metrics offer a comprehensive evaluation of wetland habitat from a biotic, physical-chemical, and landscape perspective; we include such a large number of metrics (n = 54) under the principal that a comprehensive survey is most likely to detect signals of degradation and provide clues as to the sources of impairment. The biotic indicators were chosen based on their known response to wetland habitat impairment (Niemi et al., 2006; USEPA, 2002) and because they represent assemblages with many wetland-associated species. Water quality metrics were chosen based on their widespread conventional use in habitat assessment (Chow-Fraser, 2006; Trebitz et al.,

2007). Landscape metrics were chosen to represent the influence of the availability of certain habitat types on wildlife populations and wetland habitat quality. All three categories of metrics are interrelated in the wetland ecosystem, and because BUIs No. 3 and No. 14 are considered likely to be closely linked (NYSDEC, 2006a), we used the results from all 54 metrics to make inferences about the status of both BUIs.

Figure 1.1: Surveyed wetlands in the St. Lawrence River at Massena Area of Concern and reference area in Louisville, NY.

Biotic Indicators

We surveyed each wetland for birds on a total of three mornings (05:30 - 09:00) during

(a) the last two weeks of May, (b) the middle two weeks of June, and (c) late June to early July,

13 in 2012; surveys were a minimum of ten days apart (Marsh Monitoring Program, 2009b). On each sampling day, wetlands from both the AOC and reference locations were surveyed. Marsh- nesting birds, aerial foragers, and other birds were surveyed via a 10-minute, 100 m circular radius point count (Bibby et al., 2000), and some marsh-nesting birds were detected using a call broadcast survey. The call broadcast survey consisted of three replicate playbacks of each focal species’ vocalization, one minute silence between each playback bout, and five additional minutes vigilance after the last broadcast (call broadcast methods after Conway, 2011). The speaker volume was standardized by ear to a sound pressure level of 80-90 dB at 1 m in front of the speaker. Focal species (in order of vocalization broadcast) included least bittern (Ixobrychus exilis), sora (Porzana carolina), Virginia rail (Rallus limicola), common gallinule (Gallinula galeata), American coot (Fulica americana), pied-billed grebe (Podilymbus podiceps), king rail

(Rallus elegans), yellow rail (Coturnicops noveboracensis), American bittern (Botaurus lentiginosus), sedge wren (Cistothorus platensis), marsh wren (Cistothorus palustris), and golden-winged warbler (Vermivora chrysoptera). We surveyed from a single point located on the open water - emergent vegetation interface. Detected birds were classified as wetland-associated

(i.e., ‘waterbirds’) based on Brooks and Croonquist (1990). We define a wetland site’s relative bird species richness as the sum of species detected within the 100 m radius and all focal species detected during the call broadcast across all three survey periods.

We surveyed each wetland for calling anurans (frog and toad) on a total of three nights

(20:00 – 23:30) during (a) the fourth week of April, (b) the fourth week of May, and (c) the third week of June, in 2012 or 2013; surveys were done minimally 15 days apart and coinciding with three air temperature ranges (5-9°C, 10-16°C, and >16 °C). We surveyed from a single 100 m, semi-circular, fixed radius point located on the open water - emergent vegetation interface of the

14 wetland. Each survey initiated with a one minute pause, followed by a three minute aural point count; species were identified by vocalization (Marsh Monitoring Program, 2009a; USGS,

2012).

We surveyed small fish, including small species and young of larger species, during the last week of July to the second week of August, 2012, using baited cylindrical funnel traps (42 cm long, 22 cm in diameter, 3 cm diameter entry opening, 0.48 cm2 mesh size, baited with 40 cat food pellets in 30 mL, perforated plastic containers). Three traps were placed amid submerged macrophytes in the littoral zone along the perimeter of each wetland at 1 m depth. Traps were checked daily on four consecutive days (12 trap-nights per wetland); all fish were identified to species and released. In addition to richness and abundance, we calculated the Shannon-Wiener diversity index for each site.

We surveyed submerged, emergent, and wetland-bordering upland vascular plants during the second week of August 2012, using a transect quadrat method. Each transect ran along the elevation gradient; the first transect passed through the bird and anuran survey point and two additional transects were placed in the same orientation at 50 m and 100 m distance along the water - emergent vegetation interface. At each transect, we placed a 1 m2 quadrat at three elevations (+20 cm, 0 cm, -20 cm), where the 0 cm elevation was estimated by observing field indicators of the maximum spring water level line such as debris and sediment accumulation (US

ACE, 1987), and the other two elevations determined using a sounding pole with a string attached at 20 cm. The three quadrats provided a good representation of the shift from wholly aquatic to transitional upland plants (Tiner, 1999). At each quadrat, all vascular plants were identified to the lowest taxonomic level, and percent cover was estimated (1%, 5%, 10%, and after at 10% increments) for each plant taxon. The USDA Plants Database

(http://plants.usda.gov, 29, September 2015) was used to classify plants as non-native and the

NYS DEC website was used to categorize noxious plants (NYSDEC, 2014). Plant identifications were validated by collecting vouchers and having them checked by a local botanical expert (Ann

Johnson; Eldblom and Johnson, 2010). The Shannon-Wiener diversity index was calculated using richness and plant cover data.

Indices of biotic integrity (IBI)

Indices of Biotic Integrity (IBI) are combinations of biotic metrics that covary with the range of conditions across a disturbance gradient (Cvetkovic and Chow-Fraser, 2011; Seilheimer et al., 2009). We used the validated amphibian IBI and the vegetation IBI developed for the

Great Lakes Coastal Wetland Monitoring Plan (Burton et al., 2008). We also used the index of marsh bird community integrity (IMBCI) developed for Chesapeake Bay wetlands (DeLuca et al., 2004). For the amphibian IBI, three metrics were assessed: relative total species richness, relative woodland species richness, and probability of detection of woodland-associated species.

The IMBCI combines guild-based community structure with species attributes. Bird species were determined to belong to the marsh obligate guild if they ranked a five on Croonquist and Brook’s

(1991) wetland dependence list. Species attributes described foraging, nesting, migration, and breeding range, and ranged from a score of 1 (generalist) to 4 (specialist) for each attribute.

Scores for each attribute were determined by using rankings developed by Croonquist and

Brooks (1991) and the bird species guides developed by the Cornell Lab of Ornithology (see

Table 2 in DeLuca et al., 2004).

To assess the herbaceous vegetation community at each wetland, we used the mean conservatism ratio (mCR). The mCR is a metric of the Floristic Quality Index and is centered on each species’ Coefficient of Conservatism (C) value (Swink and Wilhelm, 1979), which

16 expresses the propensity of plants to occupy least-altered or least-stressed wetland habitat. Each plant species was assigned a C value using Bried et al. (2012). The mCR is a ratio sensitive to the presence and dominance of non-native plant species, and is derived by dividing the mean C of all species by the mean C of all non-native species (Burton et al., 2008).

Water quality indicators

We collected water samples, using clean sampling techniques based on Turk (2001) and other protocols, during the third week of July 2012, and the second and third week of May 2013.

Water was collected at two locations per wetland at 1 m depth and within 3 m of the emergent vegetation, using 1 L acid washed polyethylene bottles. We measured temperature (ºC) and conductivity (µS/cm) at each sampling location using a calibrated multi sensor probe (YSI

Model 600XL). Water samples were returned to the laboratory and processed on the same day.

Chlorophyll-a, total phosphorus, dissolved (<0.2 µm) silica, and nitrate samples were stored in a refrigerator (at 4°C) and analyzed within a few days after collection.

We filtered chlorophyll-a (mg/L) samples onto 0.2-μm pore-sized polycarbonate filters, extracted with 90% acetone, and measured fluorimetrically using the Welschemeyer (1994) technique with a calibrated fluorimeter (TD-700, Turner Designs). Phytoplankton composition

(µg/L) was characterized on unfiltered samples by photosynthetic pigment fluorescence using a multi-spectral submersible fluorometer (FluoroProbe with Workstation 25; bbe Moldaenke,

GmbH) that was corrected for background water color at each site; the concentration of chlorophyll-a was partitioned among four categories of phytoplankton: Chlorophyta &

Euglenophyta, Phycocyanin-rich Cyanobacteria, Heterokontophyta & Dinophyta, and

Phycoerythrin-rich Cyanobacteria & Cryptophyta. We quantified CDOM (colored dissolved

17 organic matter; mg/L) by fluorometry (TD-700, Turner Designs); CDOM was expressed as equivalents of Suwannee River fulvic acid.

We measured total phosphorus (µg/L) and dissolved silica (µg/L) concentrations using colorimetric analysis (Wetzel and Likens, 2000) in a spectrophotometer (Genesys 20, Thermo

Scientific). For total phosphorus, unfiltered water samples and standards underwent persulfate digestion. Filtered (a 0.2-μm pore sized syringe filter) dissolved silica samples and standards were treated with 0.25N HCl, 5% ammonium molybdate, 1% disodium EDTA, and 17% sodium sulfite (Wetzel and Likens, 2000).

We measured turbidity (absorbance at 750 nm; A750) in unfiltered water by light spectroscopy using a spectrophotometer at 750 nm. Nitrate (mg/L), chloride (mg/L), and sulfate

(mg/L) were measured with ion chromatography (CARES analytical laboratory, Clarkson

University). Alkalinity (CaCO3; mg/L) and pH were measured in the laboratory using a Mettler

Toledo S220 SevenCompact™ pH/Ion meter. We calculated alkalinity using the Gran titration method by adding known concentrations of 0.1 M HCl to 50 mL of unfiltered sample water; each sample was titrated beyond the equivalence point (pH 3.5).

We used a combination of our water quality metrics to calculate a Water Quality Index

(WQI; Chow-Fraser, 2006). WQIs were developed by Chow-Fraser (2006) based on 12 water quality parameters that are significantly related to Great Lakes basin-wide land use stressors. We used a subset model (Equation #3 in Chow-Fraser, 2006) that best incorporated our water quality metrics: WQI = 10.753047 – 0.946098 x log TURB – 0.837294 x log COND – 1.319621 x log

TEMP – 4.604864 x log pH – 0.387189 x log TP – 0.353713 x log TN – 0.337888 x log CHL.

Landscape analysis

Landscape analysis was done using ArcGIS 10.1. We digitized wetland sites into polygons by outlining the wetland boundary using high resolution orthoimages, and used the polygons to calculate the area of each wetland. A buffer of 1 km was delineated around the edge of each wetland polygon; this scale is commonly used in landscape analyses (Brooks et al., 2004;

Mack, 2006) and provides an intermediate measure of landscape pattern (smaller than watershed, larger than site level). We used the 2011 USGS National Land Cover Database (30 m raster;

Homer et al., 2015) to calculate the proportion of each land cover type within the buffer around each sample wetland site. We grouped some very similar land cover types together for a total of six classes: open water, developed (residential and commercial) land, upland forest, agricultural

(row crop, hay, and pasture) land, woody-vegetated wetlands, and emergent-vegetated wetlands.

Additionally, road density (road-km/km2) was calculated for each wetland buffer. To evaluate the number of wetlands, distribution of wetland sizes, and overall wetland coverage for all wetlands within a region, we used the most recent U.S. Fish and Wildlife Service National

Wetland Inventory (USFWS, 1983) for mapped emergent and forested/shrub wetlands within a 2 km distance of the open water of the SLR (including islands).

Data analysis

To statistically test whether AOC and reference wetlands differed for each of the metrics, we used a Student’s two sample t-test, assuming normal distribution and equal variances, or else

Welch’s test when variances were unequal; the nonparametric Mann-Whitney test was used for non-normal distributions. A nested mixed factor ANOVA was used to compare the effect of location on individual water quality metrics, where each wetland site had multiple water samples nested within it. To maintain an experiment-wise Type I error rate of 5%, we used a Bonferroni

19 correction on each suite of metrics (biotic, water quality, and landscape). We used Bartlett’s test to confirm homogeneity of variance, and the Anderson-Darling test to detect deviations from normality. Some variables were natural log or square root transformed prior to statistical analysis. Because sampled wetland area averaged larger in Louisville than the AOC, and some biotic indicator metrics may covary with wetland size (e.g. wetland bird species richness), we used ANCOVA with wetland size as a covariate on metrics related to species richness. Statistical analyses were completed using SigmaPlot 11.0 (Systat Software, San Jose, CA) and R version

2.15.1 (R Core Team, 2012).

Results and Discussion

We found no differences in aggregate biotic metrics or water quality metrics between wetlands in the AOC and the reference location in Louisville (Table 1.1). The AOC wetlands, including the smallest ones, provided habitat for regionally-characteristic wetland-associated and wetland-restricted flora and fauna. The wetland-associated species we detected included those that are restricted to qualitatively different kinds of wetlands, including forested, herbaceous emergent, or shrub wetlands, coastal or inland wetlands, and large or small wetlands. We did detect, however, a difference in landscape metrics between the AOC and reference wetlands at the scale of 1 km around each site (Table 1.1). The AOC wetlands were smaller, had a lower representation of woody (shrub swamp, swamp forest) wetlands, and had a greater proportion of surrounding upland forest than the reference wetlands. Based on the aggregate indicators evaluated in our study, little evidence of significant beneficial use impairments on wetland habitat quality and wetland-associated biota was detected. However, the quantity of wetland

20 habitat, both in terms of the sizes of individual wetlands and overall landscape wetland land cover, is lower in the AOC than comparable reference sites.

Biotic indicators

Birds

We found that AOC wetlands provide habitat to bird communities similar to reference wetlands including wetland-associated Species of Greatest Conservation Need (SGCN;

NYSDEC, 2005). A total of 74 bird species were detected among the 27 wetland survey sites: 65 species in the AOC (range: 13-31 per site) and 55 species in reference wetlands (range: 13-26;

Table 1.1). Five wetland-dependent SGCN species were detected: American bittern, black- crowned night heron (Nycticorax nycticorax), and common tern (Sterna hirundo) in the AOC,

American bittern, common tern, least bittern, and pied-billed grebe in Louisville, and American bittern and common tern in both. We did not detect differences in any bird metrics between the

AOC and reference wetlands (Table 1.1). Neither total bird species richness nor waterbird species richness covaried with wetland size (Table 1.1). Bird communities in the area likely respond to large-scale landscape factors, such as proximity to the SLR and watershed land use patterns, rather than local wetland characteristics alone.

Bird conservation and management efforts within and nearby the AOC likely also contribute to maintaining waterbirds in the landscape, including within the AOC. For example, the NYSDEC (2005) Comprehensive Wildlife Conservation Strategy identifies the tail-waters of the Moses-Saunders dam, Lake St. Lawrence, and the Grasse River as important areas for tern colonies, overwintering waterfowl, and avian SGCN, and discusses ongoing and recommended management practices for these birds (NYSDEC, 2005). The North American Bird Conservation

Initiative identifies sections of the AOC as focus areas for both landbirds and waterbirds

(USFWS, 2007). Projects intended to improve habitat for wetland and grassland birds and to improve nesting success of some waterbird SGCN within the Massena AOC and adjacent areas

(including Louisville) are required actions under the 2003 Federal Energy Regulatory

Commission license to the New York Power Authority for the St. Lawrence-FDR Power Project

(NYPA, 2003). Bird habitat improvement is also mandated as part of a 2013 settlement between the multi-agency Natural Resource Trustees of the St. Lawrence River Environment and three heavy industries within the AOC (Trustee Council, 2013). These projects are currently underway and are showing successes at increasing numbers of the targeted bird species (Michael Morgan,

NYSDEC, personal communication). Overall, the species richness and biotic integrity of surveyed birds, including SGCN, indicate that AOC wetlands likely contribute to sustaining wetland-associated breeding bird populations in the upper SLR Valley.

Fish

We collected a total of 600 fish of 15 species (but no SGCN) among the 19 wetlands surveyed: 12 species in the AOC (range: 0-7 per site) and 10 species in Louisville (range: 1-5 per site; Table 1.1). We did not detect differences in any fish metrics between the AOC and reference wetlands (Table 1). However, the minnow traps used during our surveys captured only small fish; thus we only evaluated two components of the wetland fish fauna, small species and immatures of larger species. Nevertheless, many juveniles of larger fish are dependent on wetlands for foraging and safe cover, and small wetland fish species provide food for larger fish and waterbirds. Thus the diversity and abundance of small fish in a wetland can be an indicator of its quality for fish and wildlife generally.

Anurans

Eight anuran species were detected among all surveyed wetlands: seven species in the

AOC (range: 0-6 per site) and seven species in Louisville (range: 1-6; Table 1.1). We did not detect a difference in anuran metrics between the AOC and reference wetlands (Table 1.1).

Importantly, boreal chorus frog (Pseudacris maculata; Lemmon et al., 2007), an SGCN species, was detected in half of the AOC wetlands we surveyed; none were detected in reference sites.

Boreal chorus frogs breed in fish-free graminaceous ephemeral wetlands associated with meadows, pastures, and hayfields (Gibbs et al., 2007); many AOC wetlands included areas that had these habitat attributes. Habitat destruction and dominance of wetlands by invasive wetland plant species have been responsible for the decline of amphibians in New York State (Gibbs et al., 2005; Gibbs et al., 2007; Mifsud, 2014), but the AOC wetlands sustained anuran assemblages similar to the Louisville reference wetlands.

Vascular Plants

A total of 159 vascular plant species, including 29 non-native species and six noxious species, were identified among the 27 wetland sites: 128 species including 25 non-native and six noxious at AOC wetlands and 104 species including 15 non-native and five noxious in Louisville

(Table 1.1). We did not detect a difference in any plant metrics between the AOC and reference wetlands (Table 1.1). The plant detected at the most AOC sites was reed canary grass (Phalaris arundinacea) and in Louisville was purple loosestrife (Lythrum salicaria), both noxious species.

High introduced and noxious species vegetation cover is indicative of anthropogenic disturbance, including nutrient enrichment, impaired hydrology, and excessive sedimentation (Galatowitsch et al., 1999; Zedler and Kercher, 2004). In turn, introduced plant dominance results in lower species diversity of native wetland-associated biota and impaired habitat quality (Craft et al.,

2007; Lavoie et al., 2003; Wilcox et al., 2008). Although non-native and noxious plant species diversity and coverage did not differ between AOC and reference wetlands, habitat management plans to preserve wetland biodiversity for AOC wetlands should include surveillance and control of the spread of invasive plant species.

Water Quality

We did not detect a difference in any of the water quality metrics between wetlands in the

AOC and Louisville (Table 1.1). Our results indicate that water quality in AOC wetlands is comparable to that of reference wetlands, and suggests AOC wetlands do not have impairments that would affect habitat quality or the species richness, community composition, and abundance of wetland associated biota, beyond what is typical in the human-dominated landscape of New

York’s SLR Valley. However, we did not analyze water quality or sediment for contaminants such as PCBs, mercury, or microbes associated with fecal contamination; these were and are still being investigated by other researchers, and could affect fish and wildlife populations (Kauss et al., 1989; NYSDEC, 2009).

Landscape

The town of Louisville had more mapped wetlands and a higher percentage coverage of wetlands within 2 km of the SLR than the town of Massena (Louisville: 17.4% wetlands, 3.5 wetlands / km2; Massena: 6.3% wetlands, 3.0 wetlands / km2). Louisville wetlands greater area on average than Massena, and the largest Louisville wetland was six times the area of Massena’s largest (Wilcoxon rank sum test: w = 11461, p = 0.014; Louisville: median = 0. 7 ha, maximum

= 314.5 ha, n = 171 wetlands; Massena: median = 0. 4 ha, maximum = 54.9 ha, n = 159 wetlands).

Reflecting the differences in wetland size distributions between the two towns, the mean surface area of wetland sites we surveyed within the AOC was smaller than the reference sites in

Louisville (Table 1.1). In general, species richness increases with wetland habitat size, because many wetland-associated plants and animals are area sensitive, such that they only inhabit wetlands larger than some threshold size (Findlay and Houlahan, 1997). This is well documented, for example, in wetland-associated birds (Riffell et al., 2001). Given this, we were surprised that Louisville and the AOC were so similar for the biotic indicators we sampled, and there was no indication that species richness covaried with wetland size for any taxon we surveyed (Table 1.1). However, our results did indicate that AOC wetlands are surrounded by fewer woody wetlands and more upland forest within 1 km of the wetland edge (Table 1.1). Land cover and land use of the landscape matrix can affect species distributions and abundances

(Fahrig, 1997; Findlay and Houlahan, 1997). Because forest is a matrix land cover that facilitates connectivity between wetlands for many species, wetland-associated herpetofauna are more prevalent where there is a matrix consisting of high forest coverage (Attum et al., 2008;

Quesnelle et al., 2013). Moreover, wetland-associated birds are more frequent at wetlands that have other wetlands nearby (Alsfeld et al., 2010; Quesnelle et al., 2012).

The high forest cover in the AOC is a consequence of large areas of public parkland and restricted-use buffers having been created around St. Lawrence Seaway and Moses-Saunders hydropower dam infrastructure. The extent of forest cover within the AOC likely has contributed positively to the quality of its wetlands. It is unclear why in a well-forested landscape that woody wetlands are underrepresented, but it may be related to altered hydrology; woody wetlands develop where there are water-level fluctuations that result in periodic drying-out (Trettin et al.,

1997). Though the lower cover of woody-wetlands in the AOC seemingly has had no negative

25 impact on our biotic indicators, woody wetlands provide an important wildlife habitat indicator that may warrant attention in the AOC.

Selection of Reference Sites

We used a common field experimental design in that we compared sites that had experienced an impact history of concern (i.e. wetlands in the AOC) to suitable reference sites.

As such, our conclusions are sensitive to the appropriateness of the reference sites to serve as

‘controls’ that are comparable to impacted sites in every way except the impact history. The delineation of a suitable reference condition for wetland assessment is a vexing problem for wetland ecologists, and requires careful consideration of the aims of the study (Granados et al.,

2014). Historical datasets suitable for before-after comparisons are usually not available, and a suitable nearby reference area may not be present in a generally degraded region. For our study, the principal aim was to evaluate the AOC wetlands in relation to wetlands elsewhere in New

York’s SLR Valley, a primarily agricultural landscape. Though we tried to match reference wetlands to AOC wetlands by size, we were unsuccessful. The reference wetlands averaged over

9-times larger than AOC wetlands (Table 1.1); this difference was not due to haphazard site selection, but rather due to real differences in the sizes of wetlands in the two regions and the limitations of site access. While differences in wetlands size and coverage equates to less wetland-associated fish and wildlife habitat in the AOC than the reference region, our biotic indicators signaled no difference in wetland quality, and indeed no effect of wetland size on wetland wildlife and vegetation composition within the range of wetland sizes we surveyed.

Management recommendations

We evaluated two BUIs in the SLR at Massena AOC: the confirmed BUI No. 14 (loss of fish and wildlife habitat) and suspected BUI No. 3 (degradation of fish and wildlife populations).

The redesignation endpoints for BUIs No. 3 and No.14 are specified by the RAC as no significant differences in habitat and fish and wildlife communities, respectively, between the

AOC and appropriate reference areas outside the AOC (NYSDEC, 2012). To date, four US and three Canadian AOCs have successfully redesignated all of their BUIs including the two addressed here. However the redesignation criteria for BUIs Nos. 3 and 14 are not standardized and can vary from AOC to AOC.

The Oswego River AOC on Lake Ontario in New York redesignated BUIs Nos. 3 and 14 following dam relicensing activities, which restored river flow and fish access to river habitats

(NYSDEC, 2006b). The Severn Sound AOC on Lake Huron in Ontario satisfied a long list of redesignation criteria that included specific objectives for nearshore and tributary fish communities, colonial waterbirds, exotic species, waterfowl, forest birds, reptiles and amphibians, wildlife toxicity, as well as upland, riparian, and wetland habitats (Sherman, 2002).

The SLR at Cornwall RAP has satisfied all of the BUI No. 3 criteria, as supported by data from the Marsh Monitoring Program, bird surveys, osprey reproduction monitoring, egg contaminant level testing, and fish community monitoring (Environment Canada, 2012). BUIs Nos. 3 and 14 have been addressed in coastal wetlands of the Bay of Quinte AOC by using three IBIs and the

WQI (Macecek and Grabas, 2011); biotic communities and water quality were in better condition in the AOC than outside it.

Our study provides an assessment-based process to assist the Massena RAC in redesignating BUIs Nos. 3 and 14. In terms of wildlife populations within wetlands (BUI No. 3), the biotic assemblages we surveyed, each of which represents an important component of wetland-associated fish or wildlife, indicate no difference between the AOC and reference region. However, for suspected BUI No. 3 we do not recommend redesignating at this time. An

27 important caveat is that we did not survey all relevant wetland-associated wildlife assemblages; assemblages we did not survey included mammals, reptiles, or larger-sized fish. Some of these, including turtles and wetland-associated mammalian predators such as the American mink

(Neovison vison), may be affected by PCB and other bioaccumulating and biomagnifying contaminants present in riverine sediments within the AOC. Before BUI No. 3 is redesignated, a determination on the health of populations of fish and wildlife species is recommended to be completed, especially those vulnerable to bioaccumulating contaminants.

Our results indicate that there is less wetland habitat in the AOC than the adjacent reference area of Louisville, in terms of overall coverage and average surface area per wetland

(Table 1.1). Thus, for BUI No.14, it appears that it remains impaired in terms of the amount of wetland fish and wildlife habitat, particularly woody wetland habitat. If wetland coverage and composition similar to the landscape outside the AOC are required for redesignating BUI No. 14, wetland restoration and enhancement projects, particularly restoration projects that implement or restore water-level fluctuations and other hydrological features conducive to development of woody wetland vegetation, would be beneficial. For example, a woody wetland within the AOC, restored on industrial property as an environmental mitigation obligation, had the best representation of wetland-associated birds among the wetlands we surveyed.

Our study focused on only one habitat component albeit a very important one: wetlands.

We chose this habitat type because of the important ecosystem services that wetlands provide, the exceptional importance of wetlands for fish and wildlife populations, and because wetlands are known to be directly affected by the kinds of disturbance history that occurred in the AOC.

However, for a more complete assessment of BUI Nos. 3 and 14, other habitat types and wildlife assemblages would need to be evaluated, particularly riparian forest habitat, river benthos, and

28 submerged aquatic vegetation beds. For example, the SLR at Cornwall AOC used remote sensing assessments of forest cover and riparian habitat for delineating redesignation criteria for the loss of fish and wildlife habitat BUI (Environment Canada, 2012; Environment Canada,

2013; Gillespie, 1998; Hickey, 2002). The SLR at Massena AOC may consider habitat redesignation criteria from the SLR at Cornwall AOC and other AOCs in developing appropriate redesignation criteria for other habitats in addition to wetlands.

We included landscape variables in our study because the landscape matrix setting of a wetland plays a critical role in whether the wetland can support species that require both wetland and upland habitat, whether it can maintain connectivity for species that make overland movements between wetlands, and how its water quality and quantity are impacted by the watershed. In this regard the AOC appears to be in good condition relative to the reference area, given the higher forest coverage and similar emergent wetland coverage in the AOC. We recommend that the redesignation process for BUI Nos. 3 and 14 in the AOC include the use of defined landscape endpoint criteria both at the local and watershed levels. For example,

Environment Canada (2013) issued a habitat guidance document, based on a review of a large body of scientific literature on the Great Lakes, which recommends minimum wetland coverage of 10% and a minimum forest cover of 30% of the watershed.

Our results provide only a snapshot of the current wetland habitat quality and the state of fish and wildlife communities. Long-term survey efforts, such as provided by the Marsh

Monitoring Program (Tozer, 2013), the North America Amphibian Monitoring Program (USGS,

2012), and various invasive plant species monitoring programs would provide the best indicators of the overall quality of wetland habitats in the AOC and detect changes in state over time. Such surveys could be done as organized citizen-science programs, under the aegis of a public

Table 1.1: Summary of statistical tests comparing individual biotic, water quality, and landscape metrics between wetlands in St. Lawrence River at Massena AOC and reference wetlands in Louisville. Significant results (Bonferroni corrected biotic: p<0.004, water quality: p<0.002, landscape: p<0.007) are indicated in bold. Square root and natural log data transformations are noted in parentheses. Superscripts indicate the statistical test applied: aStudent’s t-test, bWelch’s test, cMann-Whitney test, dnested mixed factor ANOVA, eANCOVA

Massena AOC Louisville (Reference) test Metric p-value Mean SD N Mean SD N statistic Biotic Metrics Bird species richness 21.7 4.4 17 21 4.5 10 F=0.14e 0.72 Waterbird species richness 3.3 1.8 17 3.5 1.9 10 F=0.08e 0.78 IMBCI 6.2 2.1 17 6.2 2.2 10 F=0.001e 0.98 Fish species richness 3.1 2.4 12 3.4 1.4 7 F=0.11e 0.74 Fish abundance 19.0 37.1 12 53.1 50.9 7 W=73c 0.01 Fish Shannon-Wiener diversity 1.1 0.3 8 0.7 0.3 6 W=10c 0.08 Anuran species richness 3.2 1.8 17 3.9 1.6 10 F=1.36e 0.26 Amphibian IBI 77.6 22.6 17 84.6 14.1 10 F=0.81e 0.38 Plant total species richness 22.4 7.4 17 21.5 6.3 10 F=0.10e 0.75 Plant total Shannon-Wiener diversity 2.2 0.5 17 2.3 0.4 10 t=-0.42a 0.68 Native plant species richness 16.9 5.6 17 17.8 5.9 10 t=0.38a 0.71 Non-native plant species richness 5.5 2.8 17 3.7 1.6 10 F=3.23e 0.09 Non-native plant percent cover 43.6 19.6 17 22.5 12.8 10 t=-3.03a 0.006 mCR 0.7 0.1 17 0.8 0.1 10 F=3.35e 0.08

Water Quality Metrics Nitrate (mg/L) (2012) (ln) 0.10 0.17 12 0.01 0.02 8 F=5.05d 0.04 Nitrate (mg/L) (2013) (ln) 0.4 0.5 14 0.04 0.03 7 F=5.66d 0.03 Chloride (mg/L) (2013) 26.3 19.5 14 30.7 47.2 7 F=0.76d 0.39 Sulfate (mg/L) (2013) (ln) 22.0 23.2 14 13.5 10.2 7 F=0.52d 0.49 CDOM (mg/L) (2012) 14.2 12.8 12 23.5 11.9 8 F=3.13d 0.09

Massena AOC Louisville (Reference) test Metric p-value Mean SD N Mean SD N statistic CDOM (mg/L) (2013) 12.9 9.6 14 28.5 14.5 9 F=9.07d 0.007 Temperature (Cº) (2012) 24.5 2.9 12 25.2 3.2 8 F=0.27d 0.61 Temperature (Cº) (2013) 14.5 4.8 14 13.3 2.9 9 F=0.44d 0.52 Conductivity (µS/cm) (2012) 379.6 162.0 12 262.2 64.2 8 F=4.02d 0.06 Conductivity (µS/cm) (2013) (ln) 415.1 242.7 14 369.3 305.0 9 F=0.38d 0.54 d Turbidity (Absorbance750nm) (2012) (sqrt) 0.06 0.11 12 0.02 0.02 8 F=0.52 0.49 d Turbidity (Absorbance750nm) (2013) (sqrt) 0.02 0.03 14 0.02 0.02 9 F=0.02 0.91 Total Phosphorus (µg/L) (2012) (ln) 59.1 67.5 12 36.6 17.7 8 F=0.004d 0.95 Total Phosphorus (µg/L) (2013) (ln) 34.5 25.3 14 59.0 51.5 9 F=2.83d 0.12 Dissolved Silica (µg/L) (2012) (ln) 0.3 0.4 12 0.2 0.3 8 F=0.90d 0.36 Dissolved Silica (µg/L) (2013) (ln) 0.3 0.4 14 0.3 0.4 9 F=0.17d 0.69 pH (2012) 8.3 0.5 12 8.2 0.8 8 F=0.21d 0.66 pH (2013) 7.7 0.5 14 7.4 0.5 9 F=2.28d 0.15 d Alkalinity (CaCO3; mg/L) (2012) 96.5 16.4 7 102.6 35.0 7 F=0.22 0.65 d Alkalinity (CaCO3; mg/L) (2013) 149.9 95.5 14 110.3 82.2 9 F=1.04 0.32 Chlorophyll-a (µg/L) (2012) (ln) 11.8 16.7 12 5.1 4.7 8 F=0.07d 0.80 Chlorophyll-a (µg/L) (2013) (ln) 5.6 6.1 14 6.7 5.9 9 F=0.68d 0.42 Phytoplankton (µg/L) Chlorophyta & Euglenophyta (2012) 5.3 6.3 8 3.5 4.2 6 F=0.28d 0.60 Chlorophyta & Euglenophyta (2013) (sqrt) 2.5 5.1 14 0.7 0.7 9 F=1.24d 0.29 PC-rich Cyanobacteria (2012) (sqrt) 3.6 3.7 8 0.54 0.62 6 F=7.47d 0.02 PC-rich Cyanobacteria (2013) (sqrt) 0.2 0.3 14 0.06 0.09 9 F=4.59d 0.04 Heterokontophyta & Dinophyta (2012) (ln) 5.4 6.7 8 1.2 1.1 6 F=7.69d 0.02 Heterokontophyta & Dinophyta (2013) (sqrt) 1.5 1.7 14 1.9 2.3 9 F=0.02d 0.88 Cryptophyta & PE-rich Cyanobacteria (2012) (sqrt) 1.7 5.6 8 1.6 2.6 6 F=0.58d 0.46 Cryptophyta & PE-rich Cyanobacteria (2013) (sqrt) 0.04 0.05 14 0.5 0.8 9 F=6.10d 0.02

Massena AOC Louisville (Reference) test Metric p-value Mean SD N Mean SD N statistic WQI (2012) 0.66 0.84 12 0.94 0.33 8 t=0.83a 0.42 WQI (2013) 0.69 0.80 14 0.88 0.57 7 W=57c 0.59

Wetland area (ha) 2.1 2.9 17 19.5 32.4 10 W=134c 0.01

Landscape Metrics Open water (%) 26.0 9.8 17 36.7 25.1 10 t=1.28b 0.23 Developed land (%) 3.7 2.1 17 2.5 2.3 10 t=-1.43a 0.17 Upland forest (%) 45.5 6.1 17 29.4 13.5 10 t=-3.55b 0.004 Agricultural land (%) (sqrt) 4.2 5.9 17 11.7 9.8 10 t=2.18a 0.04 Woody wetlands (%) (sqrt) 5.0 4.7 17 13.3 5.1 10 t=4.70a 0.00008 Emergent wetlands (%) (sqrt) 0.6 0.6 17 2.3 2.5 10 t=2.73b 0.02 Road density (km/km2) 1.9 0.7 17 1.3 0.4 10 t=-2.23a 0.04

32 resource agency. Additionally, we recommend that long term habitat monitoring programs integrate the use of IBIs and the WQI to include the AOC within the regional context of the

Great Lakes. The ease of use and interpretation of IBIs makes them good candidates for long- term habitat and biotic monitoring; they can even be used to refine redesignation criteria. The

Lakewide Management Plans for each of the Great Lakes contain the same BUIs as the AOCs

(see Annex 1 and 2 of the 2012 Great Lakes Water Quality Agreement). Therefore, the approach employed here, i.e. use of georeferenced IBI and WQI, may prove a useful method for evaluating lakewide wetland habitat in addition to wildlife communities. We also recommend that indicator data from our reference set be used by the RAC as a baseline for future long term wetland monitoring and potential mitigation with the aim to redesignate BUI No. 3 and BUI No. 14.

Conclusions

We conclude that in the St. Lawrence River at Massena AOC, wetland habitat quality for fish and wildlife is not impaired, at least for the suite of indicators we surveyed, but wetland habitat quantity remains lower in the AOC than in comparable regions nearby but outside the

AOC. Though major data gaps for wetlands in the AOC have been filled by our study, a final redesignation decision for BUI Nos. 3 and 14 in terms of wetlands needs to include additional biotic indicators (e.g., wetland-associated mammals, large fish), multi-year indicator monitoring, and achievement of specific and defensible landscape endpoints.

Our study applied a quantitative methodology that incorporated biotic, water quality, and landscape indicators for evaluating two BUIs that have for the most part been addressed elsewhere qualitatively. Our approach addressed the delisting criteria specified by the SLR at

Massena AOC RAP, and provides specific conclusions and recommendations related to

33 redesignating BUI No. 3 and BUI No. 14. The framework and methodology used here can be applied to evaluate and monitor wetland fish and wildlife populations and wetland habitat even if redesignation criteria are adjusted in the future, for example by implementing more explicit and quantified benchmarks for recovery. Our framework and methodology may be helpful to other

AOCs struggling with how to evaluate the current state of their AOC in regards to these BUIs, and what necessary management actions are required to eventually meet redesignation criteria for these BUIs.

Acknowledgements

Biotic surveys were done under permit of NYSDEC, and protocols were approved by the

Clarkson University IACUC. We greatly appreciate the funding support of the St. Lawrence

River Research and Education Fund, Northern New York Audubon, and the Joseph and Joan

Cullman Conservation Foundation. For assistance in collecting the survey data, we are grateful to J. Podoliak, J. Ozolins, and K. Gilpin, and thank S. Kring for her assistance with water sample analysis. We thank G. Johnson and A. Johnson for significant help in species identification and sampling methodology. For wetland access and logistical support, we thank many private landowners, the New York Department of Environmental Conservation, New York State Office of Parks, Recreation and Historic Preservation, Saint Lawrence Seaway Development

Corporation, New York Power Authority, and Aluminum Company of America. We appreciate the reviewers’ and associate editor’s constructive comments on an earlier draft of the paper.

References: Alsfeld, A.J., Bowman, J.L., Deller-Jacobs, A., 2010. The influence of landscape composition on the biotic community of constructed depressional wetlands. Restor. Ecol. 18, 370-378. Attum, O., Lee, Y.M., Roe, J.H., Kingsbury, B.A., 2008. Wetland complexes and upland- wetland linkages: landscape effects of the distribution of rare and common wetland reptiles. J. Zool. 275, 245-251. Bibby, C.J., Burgess, N.D., Hill, D.A., Mustoe, S., 2000. Bird Census Techniques, 2nd Edition. Academic Press, California. Bode, R.W., Novak, M.A., Abele, L.E., Heitzman, D.L., Smith, A.J., 2004. 30 Year trends in water quality of rivers and streams in New York State, based on macroinvertebrate data 1972-2002. New York State Department of Environmental Conservation, Stream Biomonitoring Unit, Division of Water, Albany, NY. (http://www.dec.ny.gov/docs/water_pdf/sbu30yrintro.pdf, 11, February 2016) Bried, J.T., Strout, K.L., Portante, T., 2012. Coefficients of conservatism for the vascular flora of New York and New England: Inter-state comparisons and expert opinion bias. Northeast. Nat. 19, 101-114. Brooks, R.P., Croonquist, M. J., 1990. Wetland, habitat and trophic response guilds for wildlife species in Pennsylvania. J. Pa. Acad. of Sci. 64, 93-102. Brooks, R.P, Wardrop, D.H., Bishop, J.A., 2004. Assessing wetland condition on a watershed basis in the Mid-Atlantic region using synoptic land-cover maps. Environ. Monit. Assess. 94, 9-22. Burton, T.M., Brazner, J.C., Ciborowski, J.J., Grabas, G.P., Hummer, J., Schneider, J., Uzarski, D.G., Eds., 2008. Great Lakes coastal wetlands monitoring plan. Great Lakes Coastal Wetlands Consortium. Report to US EPA’s Great Lakes National Program Office, IL. (http://glc.org/files/docs/Great-Lakes-Coastal-Wetlands-Monitoring-Plan-FINAL-March- 2008.pdf, 11, February 2016) Chow-Fraser, P., 2006. Development of the water quality index (WQI) to assess effects of basin- wide land-use alteration on coastal marshes of the Laurentian Great Lakes, in: Simon, T.P., Stewart, P.M. (Eds.), Coastal Wetlands of the Laurentian Great Lakes: Health, Habitat and Indicators. Authorhouse, Indiana, pp. 137–166. Conway, C.J., 2011. Standardized North American marsh bird monitoring protocol. Waterbirds. 34, 319-346. Cowardin, L. M., Carter, V., Golet, F.C., LeRoe, E.T., 1979. Classification of wetlands and deepwater habitats of the United States. U.S. Department of the Interior, Fish and Wildlife Service. FWS/OBS-79/31 Craft, C., Krull, K., Graham, S., 2007. Ecological indicators of nutrient enrichment, freshwater wetlands, Midwestern United States (U.S.). Ecol. Indic. 7, 733-750. Croonquist, M. J., Brooks, R. P., 1991. Use of avian and mammalian guilds as indicators of cumulative impacts in riparian-wetland areas. Environ. Manag. 15, 701-714. Cvetkovic, M., Chow-Fraser, P., 2011. Use of ecological indicators to assess the quality of Great Lakes coastal wetlands. Ecol. Indic. 11, 1609-1622. Dahl, T.E., 1990. Wetlands losses in the United States 1780’s to 1980’s. Report to Congress. U.S. (No. PB-91-169284/XAB). Department of the Interior, Fish and Wildlife Service, Washington, D.C. (http://www.fws.gov/wetlands/Documents/Wetlands-Losses-in-the- United-States-1780s-to-1980s.pdf, 11, February 2016)

DeLuca, W.V., Studds, C.E., Rockwood, L.L., Marra, P.P., 2004. Influence of land use on the integrity of marsh bird communities of Chesapeake Bay, USA. Wetl. 4, 837-847. Eldblom, N.C., Johnson, A.M., 2010. Plants of St. Lawrence County, NY: An Annotated Checklist of Vascular Flora. Bloated Toe Publishing, New York. Environment Canada, 2012. St. Lawrence River (Cornwall) Area of Concern. Stage 3 Remedial Action Plan (RAP) Report. Environment Canada, Ontario. Transition Coordinator, Karen Douglass Cooper, Tel. 613-936-6620 (ext.229), e-mail [email protected]. Environment Canada, 2013. How much habitat is enough? Third Edition. Environment Canada, Ontario. (http://www.ec.gc.ca/nature/E33B007C-5C69-4980-8F7B- 3AD02B030D8C/894_How_much_habitat_is_enough_E_WEB_05.pdf, 17, February 2016) Environment Canada and the USEPA, 2005. State of the Great Lakes 2005. EPA 905-R-06-001. Fahrig, L., 1997. Effects of habitat loss and fragmentation on population extinction. J. Wildl. Manag. 6, 603-610. Findlay, C.S., Houlahan, J., 1997. Anthropogenic correlates of species richness in southeastern Ontario wetlands. Conserv. Biol. 11, 1000-1009. Galatowitsch, S.M., Anderson, N.O., Ascher, P.D., 1999. Invasiveness in wetland plants in temperate North America. Wetl. 19, 733-755. Gawler, S.C., 2008. The Northeast habitat classification and mapping project. Report to the Virginia Department of Game and Inland Fisheries on behalf of the Northeast Association of Fish and Wildlife Agencies and the National Fish and Wildlife Foundation. NatureServe, Massachusetts. (http://rcngrants.org/content/northeastern- terrestrial-wildlife-habitat-classification, 16, February 2016) Gibbs, J.P., Whiteleather, K.K., Schueler, F.W., 2005. Changes in frog and toad populations over 30 years in New York State. Ecol. Appl. 25, 1148-1157. Gibbs, J.P., Breisch, A.R., Ducey, P.K., Johnson, G., Behler, J.L., Bothner, R.C., 2007. The Amphibians and Reptiles of New York State: Identification, Natural History, and Conservation. Oxford University Press, New York. Gillespie, I., 1998. Methodologies for defining wildlife habitat within the St. Lawrence River RAP Area of Concern. Final Report. Atmospheric Environment Branch-Geomatics Unit, Environment Canada-Ontario Region. (https://www.rrca.on.ca/slrrc/browse.php, 16, February 2016) Granados, M., Mandrak, N.E., Jackson, D.A., 2014. Synthesizing reference conditions for highly degraded areas through best professional judgment. J. Great Lakes Res. 40, 37-42. Hickey, M.B.C., 2002. Analysis of forest habitat distribution in the St. Lawrence River (Cornwall) Area of Concern: an assessment of progress toward delisting. St. Lawrence River Restoration Council, Ontario Ministry of Environment, Environment Canada. (https://www.rrca.on.ca/slrrc/browse.php, 16, February 2016) Homer, C.G., Dewitz, J.A., Yang, L., Jin, S., Danielson, P., Xian, G., Coulston, J., Herold, N.D., Wickham, J.D., Megown, K., 2015. Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogramm. Eng. Remote Sens. 81, 345-354. Kauss, P.B., Hamdy, Y.S., Hamma, B.S., 1989. St. Lawrence River environmental investigations Volume 1, background: assessment of water, sediment, and biota in the Cornwall, Ontario and Massena, New York section of the St. Lawrence River 1979-1982. Queen’s Printer for Ontario.

Lavoie, C., Jean, M., Delisle, F., Letourneau, G., 2003. Exotic plant species of the St. Lawrence River wetlands: a spatial and historical analysis. J. Biogeogr. 30, 537-549. Lemmon, E.M., Lemmon, A.R., Collins, J.T., Lee-Yaw, J.A., Cannatella, D.C., 2007. Phylogeny-based delimitation of species boundaries and contact zones in the trilling chorus frogs (Pseudacris). Mol. Phylogeny Evol. 44, 1068-1082. Macecek, D., Grabas, G.P., 2011. Applying a regional coastal wetland monitoring framework to refine and report on wildlife and habitat delisting criteria in the Bay of Quinte Area of Concern. Aquat. Ecosyst. Health Management. 14, 94-103. Mack, J.J., 2006. Landscape as a predictor of wetland condition: An evaluation of the landscape development index (LDI) with a large reference wetland dataset from Ohio. Environ. Monit. Assess. 120, 221-241. Marsh Monitoring Program, 2009a. Participant's handbook for surveying amphibians. Published by Bird Studies Canada in cooperation with Environment Canada and the U.S. Environmental Protection Agency. Kathy Jones, Ontario Volunteer Coordinator, Bird Studies Canada, P.O. Box 160, Port Rowan, ON, N0E 1M0, [email protected]. Marsh Monitoring Program, 2009b. Participant's handbook for surveying marsh birds. Published by Bird Studies Canada in cooperation with Environment Canada and the U.S. Environmental Protection Agency. Kathy Jones, Ontario Volunteer Coordinator, Bird Studies Canada, P.O. Box 160, Port Rowan, ON, N0E 1M0, [email protected]. Mifsud, D.A., 2014. A status assessment and review of the herpetofauna within the Saginaw Bay of Lake Huron. J. Great Lakes Res. 40, 183-191. Niemi, G.J., Axler, R., Brady, V., Brazner, J., Brown, T., Ciborowski, J. H., Danz, N., Hanowski, J.M., Hollenhorst, T., Howe, R., Johnson, L.B., Johnston, C.A., Reavie, E., Simcik, M., Swackhamer, D., 2006. Environmental indicators of the U.S. Great Lakes coastal region. Report NRRI/TR-2006/11 to the U.S. Environmental Protection Agency STAR Program, ver. 1. Agreement R82-8675, Washington DC. Prepared by Great Lakes Environmental Indicators Collaboration, Natural Resources Research Institute, University of Minnesota Duluth. NYPA, 2003. St. Lawrence-FDR Power Project Land Management Plan. (http://www.stl.nypa.gov/Land%20Management%20Plan/St.%20Lawrence%20LMP%20 final.102604.htm, 28, November 2015) NYSDEC, 1990. St. Lawrence River at Massena Remedial Action Plan, Stage 1. New York State Department of Environmental Conservation, New York. (http://www.dec.ny.gov/docs/water_pdf/massenarapstagei1990.pdf, 16, February 2016) NYSDEC, 2005. A strategy for conserving New York’s fish and wildlife resources. New York State Department of Environmental Conservation, Fish, Wildlife & Marine Resources, New York. (http://www.dec.ny.gov/docs/wildlife_pdf/cwcs2005.pdf, 16, February 2016) NYSDEC, 2006a. St. Lawrence River at Massena, New York Remedial Action Plan Status Report. New York State Department of Environmental Conservation, Division of Water, Bureau of Water Assessment and Management, New York. (http://www.dec.ny.gov/docs/water_pdf/massenarapupdate06.pdf, 16, February 2016) NYSDEC, 2006b. Oswego River Remedial Action Plan Stage 3 – Delisting. New York State Department of Environmental Conservation, Division of Water, New York. (http://www.epa.gov/sites/production/files/2013-12/documents/oswego_stage3a.pdf, 16, February 2016)

NYSDEC, 2009. The Saint Lawrence River basin waterbody inventory and priority waterbodies list. New York State Department of Environmental Conservation, Division of Water, Bureau of Water Assessment and Management, New York. (http://www.dec.ny.gov/docs/water_pdf/pwlstlw09.pdf, 16, February 2016) NYSDEC, 2012. Addendum to Stage 2 Remedial Action Plan St Lawrence River at Massena, Area of Concern. New York State Department of Environmental Conservation, Division of Water, Bureau of Water Assessment and Management, New York. (http://www.dec.ny.gov/docs/water_pdf/stlrrapfnl.pdf, 16, February 2016) NYSDEC, 2013. 2012 Annual Report. Lake Ontario Unit and St. Lawrence River Unit to the Great Lakes Fishery Commission’s Lake Ontario Committee. New York State Department of Environmental Conservation, Bureau of Fishes, New York. (http://www.dec.ny.gov/docs/fish_marine_pdf/lorpt12.pdf, 16, February 2016) NYSDEC, 2014. New York State Prohibited and Regulated Invasive Plants. (http://www.dec.ny.gov/docs/lands_forests_pdf/isprohibitedplants2.pdf, 26, October 2015). Quesnelle, P.E., Fahrig, L., Lindsay, K.E., 2013. Effects of habitat loss, habitat configuration and matrix composition on declining wetland species. Biol. Conserv. 160, 200-208. R Core Team, 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. Riffell, S.K., Keas, B.E., Burton, T.M., 2001. Area and habitat relationships of birds in Great Lakes coastal wet meadows. Wetl. 4, 492-501. Seilheimer, T.S., Mahoney, T.P., Chow-Fraser, P., 2009. Comparative study of ecological indices for assessing human-induced disturbance in coastal wetlands of the Laurentian Great Lakes. Ecol. Indic. 9, 81-91. Sherman, R.K., 2002. Severn sound remedial action plan stage 3 report: the status of restoration and delisting of Severn Sound as an area of concern. Environment Canada and Ontario Ministry of the Environment. (http://www.severnsound.ca/Shared%20Documents/Reports/SSRAP_Stage_3_Report_20 02.pdf, 16, February 2016) Swink, F., Wilhelm, G., 1979. Plants of the Chicago Region: a Checklist of the Vascular Flora of the Chicago Region, with Keys, Notes on Local Distribution, Ecology, and Taxonomy, and a System for Evaluation of Plant Communities. Morton Arboretum, Lisle, IL. Tiner, R.W., 1999. Wetland Indicators, a Guide to Wetland Identification, Delineation, Classification and Mapping. Lewis Publishers, New York. Tozer, D.C., 2013. The Great Lakes Marsh Monitoring Program 1995-2012, 18 years of surveying birds and frogs as indicators of ecosystem health. Published by Bird Studies Canada, Ontario. (http,//www.birdscanada.org/download/GLMMPreport.pdf, 12, November 2015). Trebitz, A.S., Brazner, J.C., Cotter, A.M., Knuth, M.L., Morrice, J.A., Peterson, G.S., Sierszen, M.E., Thompson, J.A., Kelly, J.R., 2007. Water quality in Great Lakes coastal wetlands: Basin-wide patterns and responses to an anthropogenic disturbance gradient. J. Great Lakes Res. 33, 67-85. Trettin, C.C., Jurgensen, M.F., Grigal, D.F., Gale, M.R., Jeglum, J.K. (Eds.), 1997. Northern Forested Wetlands Ecology and Management. CRS Press, Inc., Florida. Trustee Council, 2013. St. Lawrence River Environment Natural Resource Damage Assessment: Restoration and Compensation Determination Plan and Environmental Assessment, Final

Revised, May 2013. (https://casedocuments.darrp.noaa.gov/greatlakes/lawrence/pdf/ENV_ENFORCEMENT- 2371812-v1-Alcoa__Final_May_2013_RCDP__FILED.pdf, 28, November 2015). Twiss, M.R., Ulrich, C., Kring, S.A., Harold, J., Williams, M.R., 2010. Plankton dynamics along a 180 km reach of the Saint Lawrence River from its headwaters in Lake Ontario. Hydrobiol. 647, 7-20. Turk, J.T., 2001. Field guide for surface water sample and data collection. USDA Forest Service. Air program. (http://www.fs.fed.us/waterdata/PDFfiles/FieldGuide_Turk.pdf, 16, February 2016) USACE, 1987. Wetland delineation manual. U.S. Army Engineer Waterways Experiment Station. Vicksburg, MS, USA. Technical Report Y-87-1. USEPA, 2002. Methods for evaluating wetland condition: Introduction to wetland biological assessment. Office of Water, U.S. Environmental Protection Agency, Washington, DC. EPA-822-R-02-014. USFWS, 2007. Bird conservation plan for the Lower Great Lakes/St. Lawrence Plan Bird Conservation Region (BCR 13). U.S. Fish and Wildlife Service, Division of Bird Habitat Conservation, Virginia. (http://acjv.org/BCR_13/BCR13_Final_Plan_July07.pdf, 16, February 2016) USGS, 2012. North American Amphibian Monitoring Program (https://www.pwrc.usgs.gov/naamp/, 29, September 2015). Patuxent Wildlife Research Center. Welschemeyer, N.A., 1994. Fluorometric analysis of chlorophyll a in the presence of chlorophyll b and phaeopigments. Limnol. Oceanogr. 39, 1985-1992. Wetzel, R.G. and Likens, G.E., 2000. Limnological Analyses, 3rd edition. Springer Science+Business Media, Inc., New York. White, P.S., Walker, J.L., 1997. Approximating nature’s variation: selecting and using reference information in restoration ecology. Restor. Ecol. 4, 338-349. Wilcox, D.A., Thompson, T.A., Booth, R.K., Nicholas, J.R., 2007. Lake-level variability and water availability in the Great Lakes. U.S. Geological Survey Circular 1311, Virginia. Wilcox, D.A., Kowalski, K.P., Hoare, H.L., Carlson, M.L., Morgan, H.N., 2008. Cattail invasion of sedge/grass meadows in Lake Ontario: photointerpretation analysis of sixteen wetlands over five decades. J. Great Lakes Res. 34, 301-323. Zedler, J.B., Kercher, S., 2004. Causes and consequences of invasive plants in wetlands: Opportunities, opportunists, and outcomes. Crit. Rev. Plant Sci. 23, 431-452.

CHAPTER II: Species distribution modeling of the threatened Blanding’s Turtle’s (Emydoidea blandingii) range edge as a tool for conservation planning.b

Kinga M. Stryszowska1, Glenn Johnson2, Lorianny Rivera Mendoza3, Tom A. Langen4

1Institute for a Sustainable Environment, Clarkson University, Potsdam, New York, USA 2Department of Biology, The State University of New York at Potsdam, Potsdam, New York, USA 3Department of Environmental Science, University of Puerto Rico at Río Piedras, San Juan, Puerto Rico 4Department of Biology, Clarkson University, Potsdam, New York, USA

Abstract

The delineation of a species range is challenging because of the number of interacting factors at multiple spatial scales that affect a species’ distribution. Species distribution models

(SDM) can be used to identify factors that are most associated with a species’ presence and thus potentially define a range edge. We evaluated the utility of two popular SDM approaches, maximum entropy models (Maxent) and generalized linear models (GLM), for determining the range edge for the threatened Blanding’s Turtle, Emydoidea blandingii, in northeastern New

York, USA. Using the mapping and analysis software ArcGIS, we constructed and validated

SDMs using presence/absence records (GLM) and presence/background records (Maxent) with

11 environmental predictor variables. We found that because of the limits imposed by the low number of absences, GLM was not as successful as Maxent at predicting habitat suitability for rare and cryptic species like E. blandingii. Our results also indicated that a distinct environmentally-induced range edge is associated with factors related to elevation. Both GLM and Maxent models also projected the presence of suitable habitat outside of the current range,

______bStryszowska K.M., Johnson, G., Rivera Mendoza, L., Langen, T.A. 2016. Species distribution modeling of the threatened Blanding’s Turtle’s (Emydoidea blandingii) range edge as a tool for conservation planning. Journal of Herpetology. In Press.

40 including locations with known disjunct populations. We conclude that a presence/background

SDM approach like Maxent is valid when accurate data on locational absences are lacking, as is typical for rare, cryptic species. Using SDM to understand the factors shaping the range edge can aid at planning habitat conservation and management of threatened species such as E. blandingii.

Key words

Biogeography; Freshwater turtle; GLM; Maxent; New York; Rare species

Introduction

Range edge dynamics of rare species are challenging to study because these species typically have low detection probabilities, so determining presence versus absence at localities is very difficult (Engler et al., 2004; Marini et al., 2010).

Nevertheless, rare species have a greater need for range edge delineation than common species because knowing the physical and environmental limits to population persistence is requisite for successful conservation planning. Species distribution models (SDMs), which use species habitat tolerances and selection to predict their geographical distributions, can be used to delineate range edges, but are sensitive to the type of occurrence data provided (Arntzen and Espregueira

Themudo, 2008; Seabrook et al., 2014). Species occurrence locality data used to build SDMs are typically presence-only since true absences are difficult to determine, especially for low-density and cryptic species (Segurado and Araujo, 2004; Elith et al., 2006). Repeated surveys are required for confidence about absences, but such survey effort is frequently not feasible

(MacKenzie et al., 2002). Even though SDMs have been used as tools for species management in

41 recent decades, it is still not certain whether absence records produce more accurate models than presence-only (Brotons et al., 2004; Rupprecht et al., 2011).

Turtles are undergoing steep population declines globally mostly as a result of habitat degradation and loss, but also due to unsustainable harvest, introduction of invasive species, and global climate change (Gibbons et al., 2000). Emydoidea blandingii (Blanding’s Turtles)

(Holbrook, 1838) are semiaquatic turtles of the northern United States and southeastern Canada that are of conservation concern across their range; they are listed as Threatened in New York

(Ross and Johnson, 2013) and Globally Endangered on the IUCN Red List (for the full range see van Dijk and Rhodin, 2013). The species is notable for long seasonal overland movements among permanent and ephemeral wetlands (Congdon et al., 2011; Millar and Blouin-Demers,

2011). Degradation and loss of wetlands, and possibly habitat fragmentation caused by roads, have contributed to the historical decrease in distribution and abundance of E. blandingii (Ernst and Lovich, 2009; Ross and Johnson, 2013). Because of uncertainties about its historical distribution, the extent and stability of the current range of E. blandingii is unclear. The species is at the eastern limits of its contiguous range in northeastern New York, but small disjunct populations occur elsewhere in New York, New England, and Nova Scotia. Although there is evidence that the edge populations are increasingly isolated genetically from the core

Midwestern U.S. population (Mockford et al. 2007; McCluskey et al., in press), the stability of the turtle’s range is presently unknown.

Our objectives were to evaluate the effectiveness of two popular SDM methods at predicting occurrences of a rare species at the edge of its range when using presence-only versus presence-absence records, and to use SDM to determine which environmental and landscape factors affect the distribution of E. blandingii, especially those contributing to its range limit in

42 northeastern New York. We hypothesized that E. blandingii in the St. Lawrence River (SLR)

Valley of northeastern New York is constrained by the abrupt elevation rise at the Adirondack

Mountains and by the occurrence of habitat fragmentation from agriculture, residential development, and roads within the valley. We also hypothesized that an SDM algorithm that incorporates true absence records (GLM) would perform better at predicting species occurrences at a range edge than one that does not (Maxent).

Materials and Methods

Study Area

We modeled the distribution of E. blandingii within the SLR Valley of northeastern New

York, bordered to the southeast by the Adirondack Mountains (Figure 2.1). Based on the 2011

National Land Cover Data, the region is predominately a mix of agricultural uses and northern hardwood-conifer forest; emergent and shrub wetlands compose 3.5% of the landscape (Jin et al.,

2013). The 585,000 ha region selected for SDM was delineated by using the longest distance between two adjacent E. blandingii occurrence records (21.1 km) to buffer around all occurrence records (after Aitken et al., 2007). The northern edge was defined by the New York border.

Data Sources

We obtained E. blandingii occurrence records from a 14-yr (1999–2013) regional survey, performed by setting baited hoop nets in wetlands for multiple nights (15–2,170 trap-nights per site) (Johnson and Conrad, 2012). We selected survey sites to provide thorough geographic coverage and targeted accessible woody (forested or shrub) wetlands, a habitat preferred by E. blandingii. Wetlands near ad hoc turtle detections (e.g. road crossings or road-kill records, informant reports) were also surveyed. Among the 228 surveyed sites, 87 sites had at least one

Figure 2.1: Regional limits for SDM in New York’s St. Lawrence River Valley, USA, and E. blandingii survey locations.

detection. For SDM, we supplemented survey results with other regional detection records. To reduce spatial autocorrelation, we only included occurrences > 100 m apart. For the coarser spatial scale SDMs, only one occurrence record was retained in each 800 m x 800 m raster cell.

Of the final 211 occurrence records, 99 were road crossings, 66 were hand captures at survey sites, and 46 were trap records. We constructed models by randomly selecting 75% of the occurrence records and leaving 25% for independent validation. To calculate the probability of a true absence if there was no detection at a survey site, we first estimated the proportional likelihood of detection per trap-night (Psite) for each of eight sites known to have E. blandingii present and for which survey effort was at least 300 trap-nights (after Kery, 2002). The number of turtles captured (n) was divided by the total trap-nights (t) at a site: Psite= n / t, and the eight

44 sites were averaged (Pmean = 0.0169 ± SD 0.0110); we then calculated the proportional likelihood

0 of a true absence at a surveyed site where no turtles were trapped (q): q = 1 - (Pmean * (1 -

t Pmean )). To retain a suitable number of sampled sites for SDM, we used a cut-off of 0.3 (70% chance of occurrence despite no detections) for the proportional probability of a true absence, which resulted in retaining 113 out of 131 no-detection sites.

We used eleven putative predictor variables associated with climate, land cover, and topography (Table 2.1); these were selected based on our knowledge of E. blandingii ecology and comparable turtle distribution models (Rizkalla and Swihart, 2006; Attum et al., 2008; Millar and Blouin-Demers, 2012; Quesnelle et al., 2013). We verified that variance inflation factors were low (<5) to reduce multicollinearity. Because of the coarse resolution of the two climate variables, they were not used at the 250-m scale of modeling.

To evaluate the effects of different landscape scales on the occurrence of E. blandingii, we extracted environmental data within two circular buffers around each presence-absence point:

250 m (raster resolution of 30 m) and 8,000 m (rasters re-sampled using bilinear interpolation to

800 m). The 250-m radius corresponded to a previously used buffer length intended to match E. blandingii mean daily movement distances, and thus turtles’ direct interactions with the landscape (Millar and Blouin-Demers, 2012). The 8,000-m buffer was intended to capture indirect landscape-scale factors such as climate, biotic interactions, and metapopulation dynamics (Marchand and Litvaitis, 2004; Austin and Van Niel, 2011). For both GLM and

Maxent, and at each buffer scale, we ran three replicate models using different, random combinations (75% train-25% test) of occurrence records. All spatial analyses were conducted using ArcGIS Desktop 10.2.1 (ESRI).

Table 2.1: Eleven predictor variables used to build the SDMs for E. blandingii in New York. Variables (units) Source 250 m 8,000 m Range Range Mean monthly precipitation (mm) WorldClim a NA 76.7–88.6 Mean monthly maximal temperature (ºC) WorldClim NA 19.7–20.3 Elevation (m) NED b 60.6–191.9 59.0–224.5 Road density (km/km2) TIGER c 0–8.4 0.5–3.2 Stream density (km/km2) NAH d 0–4.9 0.1–1.1 Land cover (%) Forested/shrub wetland NWI e 0–84.4 0.3–25.5 Emergent wetland NWI 0–88.1 0.4–4.0 Canopy density NLCD f 0–82.1 3.8–71.7 Hardwood forest NTWHC g 0–77.5 1.4–53.1 Open water NLCD, NWI 0–61.5 0.5–75.5 Alfalfa/corn Cropland h 0–50.5 0.2–20.0 a 1960–1990 WorldClim dataset at 800 m resolution, averaged for the E. Blandingii active period (April–October) (Hijmans et al., 2005). b 2009 USGS National Elevation Dataset at 1/9 arc second (30 m) resolution (Gesch, 2007). c 2013 TIGER roads polyline file (U.S Census Bureau, 2013). d Northeast Aquatic Habitat stream polyline file (Olivero and Anderson, 2008). e National Wetland Inventory file (NWI; USFS, 1983). f National Land Cover Data at 30 m resolution (Jin et al., 2013). g Northeastern Terrestrial Wildlife Habitat Classification map at 30 m resolution (Ferree and Anderson, 2013). h 2010 USDA Cropland dataset at 30 m resolution (USDA National Agricultural Statistics Service Cropland Data Layer, 2010).

Model Building

We compared the two most popular methods of SDM: GLM and Maxent (Elith et al.,

2006; Pearson et al., 2007). GLM is an extension of linear regression that can model binomial data distributions (Guisan et al., 2002); for this reason, GLMs are used when both presence and absence records are available (Franklin, 2009; Rupprecht et al., 2011). Maxent is a machine learning model that uses presence records compared to a random sample of background locations to find the probability distribution of maximum entropy (i.e., closest to uniform) without over- fitting the model (Phillips et al., 2006).

All GLMs were run in R version 3.0.3 (R Core Team) using the function glm with a binomial distribution and a logit link function. For both spatial scales, we generated all model

46 subsets and selected the model with the lowest Akaike Information Criterion (AIC) (Akaike,

1974). We determined variable contributions by standardizing all predictor variables (z-scores); the highest slope coefficients were judged the most influential. The log odds ratio was converted into probability of occurrence (y) from 0 (low) to 1 (high) and imported into ArcMap using the equation: y = 1 / (1 + exp - (a + Σ x*b)), where (a) is the intercept, (x) is the regression coefficient for each model variable and (b) is the variable raster.

We used Maxent version 3.3.3 to build an SDM on 159 presence records. We used the default parameters, with a few modifications (Phillips et al., 2006; Phillips and Dudik, 2008).

Maxent was set to uniformly sample 10,000 background locations across the study region, intended to characterize the distribution of environmental parameter values. Because trapping surveys were not a random sample of localities within the region, but instead targeted woody wetlands and tended to be near roads, we implemented a wetland-road bias to select background points for the model. The bias file was created by buffering all roads and all forested and shrub wetlands by the mean distance of an occurrence record to these features. We increased the number of iterations over the Maxent default to 5,000 to allow adequate time for convergence.

We also adjusted the default prevalence value of 0.5 to 0.382 to better represent the prevalence of this rare species (Elith et al., 2011; Guillera-Arroita et al., 2015). Prevalence equals the number of surveyed sites with detections (87) divided by total surveyed sites (228). Variables’ relative contributions to the SDM were inferred by the increase in model gain when added.

Contribution was also judged by inspecting ‘jackknife’ contribution plots. Model predictions were imported into ArcMap in the logistic format, providing a predicted spatial probability of occurrence from 0 (low) to 1 (high).

Model Evaluation

We used area-under-the-curve (AUC) of the receiver-operating-characteristic (ROC) curve, which is plotted using sensitivity (proportion of presences correctly predicted) and specificity (proportion of absences correctly predicted), as the primary method of model evaluation. AUC ranges from 0.5 (models no better than random) to 1 (perfect discrimination), and a rule-of-thumb is AUC values above 0.75 are considered informative (Swets, 1988;

Eskildsen et al., 2013). AUC was used to evaluate the fit of the final models to the building points and to evaluate the models’ successes in predicting the validation points. For the purpose of additional model evaluation and interpretation via habitat suitability maps, we set thresholds for both GLM and Maxent models to convert continuous probability data into a binary format.

For both models we chose a threshold that maximized the sum of sensitivity and specificity (Liu et al., 2005; Jimenez-Valverde and Lobo, 2007). Replicate binary map rasters were averaged for a single display map.

Projection

We projected our SDM results to the rest of New York (outside the modeled region) to evaluate whether suitable habitat for E. blandingii, as indicated by our models, existed outside of the modeled range. As a validation, we compared our model projections to the known distribution of E. blandingii throughout the state.

Results

Both SDM methods performed well at both scales, according to our acceptance criterion of

AUC = 0.75, and closely-fit the training points (Table 2.2). The mean training AUC value among models was 0.959 ± SD 0.004 (Maxent) and 0.855 ± 0.018 (GLM). The mean validation AUC

48 value was 0.911 ± 0.027 (Maxent) and 0.661 ± 0.159 (GLM). GLM training and validation AUC values were highest at the 8,000-m scale, whereas for Maxent, training AUC values were the same at both scales and validation AUC was highest at the 250-m scale (Table 2.2). Overall,

Maxent models, using presence locations with background samples and a bias file, performed better both in model fit and validation than the presence/absence-based GLM models (Table 2.2).

Table 2.2: Performance of GLM and Maxent models at 250-m and 8,000-m scales. The mean + SD AUC is reported for model fit to 75% training and 25% independent validation data (n = 3). Model Scale Mean AUC Training Validation GLM 250 m 0.846 + 0.008 0.612 + 0.215 8000 m 0.864 + 0.022 0.710 + 0.098

Maxent 250 m 0.959 + 0.004 0.913 + 0.020 8000 m 0.959 + 0.005 0.909 + 0.036

Variable Contributions

When using GLM, the most important variable averaged across both scales was elevation, followed by mean monthly maximal temperature and mean monthly precipitation

(Figure 2.2A); the latter two were only used at the 8,000-m scale because of their coarse resolution. Habitat suitability for E. blandingii was lower at higher elevations, higher mean temperatures, and lower precipitation. Forested/shrub wetland cover and alfalfa/corn cover were not included in any GLM models at any scale (Figure 2.2A).

When using Maxent, the most important explanatory variable averaged across both scales was elevation, followed by road density; the latter was a very important predictor at the 250-m scale but trivial at the 8,000-m scale. Habitat suitability was lower at higher elevations and, at the

250-m scale, decreased with increasing road density (Figure 2.2B). The least important variables were forest canopy and alfalfa/corn cover. The most marked difference between Maxent and

GLM models was that temperature was unimportant in the Maxent models, whereas it was the second highest contributing variable in the GLM models (Figure 2.2).

Figure 2.2: Mean importance of predictor variables for (A) GLM and (B) Maxent models at 250-m and 8,000-m scales. Error bars = SD, polarity symbols (+/-) indicate direction of effect on habitat suitability. Missing bars indicate variables not included in the models. Importance in GLM models is expressed as the regression coefficient after z-value standardization; higher coefficients are more important. Importance in Maxent models is expressed as percent contribution to the model.

Habitat Suitability Predictions within the Study Region

At the 250-m scale, both algorithms predicted small patches of high occurrence probability across the study area (Figures 2.3A, 2.3C). At the 8,000-m scale both algorithms predicted high probability of occurrence in the same locations and an area of low probability of

50 occurrence in the middle of the region (Figures 2.3B, 2.3D). The most notable difference between Maxent and GLM model predictions was evident at the 8,000-m scale, where Maxent predicted small patches of high probability of occurrence while GLM predicted large areas, especially in the northeast of the region. At all scales for both Maxent and GLM predictions, probability of E. blandingii occurrence decreased when moving in the southeastern direction away from the SLR Valley and toward the Adirondack Mountains.

Figure 2.3: Probability of occurrence of E. blandingii within the modeled region of northeastern New York using GLM or Maxent with buffer distances of 250 m or 8,000 m. Gray areas indicate high probability of occurrence. Circled area is a predicted gap between two areas of high probability of occurrence.

Habitat Suitability Predictions Projected outside the Study Region

When projecting outside of the range, GLM at the 250-m scale performed differently from the 8,000-m scale. At the 250-m scale, high habitat suitability was predicted along Lake

Ontario and down the Hudson River Valley. This predicted distribution encompassed three known, disjunct populations of E. blandingii (Ross and Johnson, 2013) (Figure 2.4B). At the

8,000-m scale, high habitat suitability was predicted for few sites outside the current range, and no longer classified the region of the Hudson River Valley populations as suitable (Figure 2.5B).

At the 250-m scale, the Maxent projection was very similar to the GLM (Figure 2.4A). At the

8,000-m scale, unlike GLM, large patches of high suitability habitat were predicted for almost the entire Hudson River Valley, encompassing two regions with documented extant populations

(Figure 2.5A).

Discussion

Conservation activities for rare species are often hindered by lack of information on their local and regional distribution. Habitat suitability models can, therefore, benefit conservationists by predicting distributions of rare species from minimal records, and improving understanding of factors shaping those distributions. This is the first study comparing two SDM algorithms using a large sample size of presence/absence records and presence/background records in a small geographic area at the range edge of a rare species. We have demonstrated the utility of SDMs for the conservation of rare and cryptic species by providing important information about the distribution and habitat associations of the threatened Blanding’s turtle in New York.

Statistical evaluation of model fit and independent validation, in combination with visual assessment of predictive maps, indicates that the presence-only, machine learning method

Maxent is better at characterizing habitat suitability for rare and cryptic species such as E. blandingii both within the modeled region and projected outside of the region (Table 2.2); at least two other studies have had similar results (Khatchikian et al., 2011; Rupprecht et al., 2011).

We had hypothesized that in a small geographic area at the range edge, where the gradients of environmental variables may be limited, the inclusion of absences would provide an important level of discrimination, but our results do not support this conjecture.

Small study regions pose the challenge of incompletely representing the entire breadth of a species’ environmental niche. Not sampling the entire gradient of the tolerances of a species can seriously bias model predictions (Hortal et al., 2008; Jimenez-Valverde et al., 2009).

Populations of species of conservation concern are often located in very small geographic regions that only contain truncated ranges of environmental gradients. Our results indicate that when the environmental gradient is truncated within a study region, Maxent is the better- performing algorithm (Table 2.2). Maxent, by default, selects 10,000 background points to characterize the entire study region, while in our study GLM was limited to the 113 absences we provided. These 113 locations were not a random sample of the region, but prospectively selected because they had habitat indicators associated with suitable E. blandingii habitat. GLM may have been over-constrained by the very low cutoff value applied to the selection of our absence records (30% probability of absence). This cutoff value may have decreased the precision of the GLM models by unreliably classifying unsuitable habitat (Gu and Swihart,

2004). High confidence of true absence can require impractically high sample effort for cryptic, rare species such as E. blandingii; for species that are more detectable, GLM with presence- absence records may perform better.

We found that the elevation variable was the most important overall predictor of E. blandingii habitat suitability at both scales (Figure 2.2) and is primarily responsible for the distinct range edge of this species in northeastern New York (Figure 2.3). Elevation is not a

Figure 2.4: Projected areas of high habitat suitability for E. blandingii outside of the model building extent using Maxent (A) and GLM (B) with a buffer distance of 250 m. The dotted line delineates the current known distribution of E. blandingii in New York.

Figure 2.5: Projected areas of high habitat suitability for E. blandingii outside of the model building extent using Maxent (A) and GLM (B) with a buffer distance of 8,000 m. The dotted line delineates the current known distribution of E. blandingii in New York.

55 variable commonly used in modeling the habitat preferences of freshwater turtles (Millar and

Blouin-Demers, 2012; Quesnelle et al., 2013), but in our case was important because of the proximity of the Adirondack Mountains. There is a pronounced elevation gradient increasing from the shore of the SLR southeast into the Adirondack Mountains, and E. blandingii is restricted to lower elevations. We hypothesize that the response of E. blandingii to elevation may be a proxy for more ecologically meaningful factors such as topographic relief and underlying geology and their effect on wetland hydrology and distribution. Elevation is also associated with a suite of other relevant factors including microclimate, soils, land cover and use, food availability, and microhabitat characteristics (Guisan and Zimmermann, 2000). Other wetland- associated species in the region may also limited by this topographic barrier.

Model projection is the application of SDMs to extrapolate suitability predictions into geographic areas or time periods not included in the original model construction. Generally projection is discouraged because (1) model algorithms may continue a fitted trend beyond the range of parameter values (Elith and Graham, 2009) and (2) the model is trained under a combination of variables that may not be ecologically relevant to the species in distant sectors of its range (Guisan and Zimmermann, 2000). Nevertheless, we chose to project our SDMs to the rest of New York to evaluate their performance in relation to several disjunct known populations of E. blandingii, and to evaluate whether there may be suitable habitat elsewhere that potentially could be of conservation value for this species. Many applications of SDM as a conservation tool require projection outside of a modeled region, so there is a need to evaluate the performance of

SDM projection using species for which distribution in a projected area is known.

At the 250-m scale both GLM and Maxent models predicted regions of high habitat suitability that encompassed known small, disjunct populations, but also included some

56 extensive regions where there are no records of the species (Figure 2.4). Suitable but unoccupied habitat outside of the current range may indicate that (1) model projections are flawed, (2) undetected populations of turtles exist in these areas, (3) the species has not yet reached these suitable areas but may do so as a result of range expansion, or (4) populations have been extirpated from these areas because of historical habitat alteration and fragmentation. We hypothesize that the regions of predicted extensive suitable habitat were once important components of the eastern E. blandingii range, but the historical land cover change and hydrological modification of the region for agriculture and industry resulted in region-wide extirpation, leaving as residuals the current small, disjunct populations.

The projections of the GLM models at the 8,000-m scale were restricted to northern New

York because of the limits imposed by the inclusion of the temperature variable (Figure 2.5B).

Maximum mean temperature in our training region spanned less than 1 ºC (Table 2.1). Because highest suitability was found at the lowest temperatures, GLM models predicted areas at higher temperatures to be unsuitable. This response to temperature is likely very local and is opposite that found by Millar and Blouin-Demers (2012) in Ontario. For this reason, including temperature in projections outside of a modeled range may be problematic (Randin et al., 2006).

Unlike GLM, Maxent projections at the 8,000-m scale were not limited by the temperature variable because temperature did not strongly contribute to the final model (Figure 2.2). Maxent predicted high habitat suitability at locations of known populations, and also in some other regions for which there are no occurrence records but it is plausible that E. blandingii once existed (Figure 2.5A). The fact that our SDM projections encompass known populations of E. blandingii indicates that model projection can be used as a conservation tool to locate promising sites for population surveys or suitable habitat for population restoration.

Conclusions and Management Implications

Our results indicate that the range edge of E. blandingii in northeastern New York is limited by factors related to higher elevation, so efforts to conserve E. blandingii should focus on understanding local population dynamics and managing habitat within the current SLR Valley range; conservation efforts to extend the range edge boundary via habitat management or population translocations should strongly consider the natural topographic and geologic barrier.

Our model projections do indicate that potentially suitable habitat may exist in corridor-like patches outside of the current E. blandingii range, which suggests that the species may have once occupied a much larger region of New York but has suffered range collapse because of habitat loss. Because SDM projections are inherently uncertain, one must be cautious making conservation decisions based on their forecasts. Nevertheless, our results indicate that projections can provide clues to the historical species distribution and potential for species range expansion, and the environmental factors that currently limit the distribution. Areas identified as suitable by projections can be targeted for future surveys and even evaluated as candidates for habitat management and population translocations to connect disjunct populations.

The best SDM method for rare species remains controversial (Engler et al., 2004;

Franklin et al., 2009). Our results indicate that whereas both Maxent and GLM are very good at predicting habitat suitability and range limits, Maxent is better suited at making predictions for rare and cryptic species when obtaining high confidence of absences is problematic. Maxent has been consistently shown to be a robust algorithm (Elith et al., 2006; Phillips et al., 2006; Pearson et al., 2007), but it has rarely been compared to presence-absence models (Khatchikian et al.,

2011; Rupprecht et al., 2011). Our results suggest that the background selection method used in

Maxent models is effective enough to replace true-absence data. Since the determination of

58 absences requires a much more intensive sampling strategy, especially for rare and cryptic species, eliminating the need for absence data from SDMs can greatly increase the efficiency of building occurrence record databases (i.e. surveying more sites with less intensity per site), ultimately resulting in better presence-only models.

Acknowledgments

The St. Lawrence River Research and Education Fund supported this research. For assistance in collecting the survey data, we are grateful to T. Crockett, A. Breisch, J. Ozard, A. Ross, E.

McCluskey, and many undergraduate students from SUNY Potsdam. We thank many private landowners for wetland access, and the New York Department of Environmental Conservation and SUNY Potsdam Research Foundation for logistical and financial support for the surveys.

This research was conducted in accordance with Institutional Animal Care and Use Committee protocol numbers 08-S-012, 10-F-017, and 11-S-019.

References Aitken, M., D. W. Roberts, and L. M. Shultz. 2007. Modeling distributions of rare plants in the great basin, western North America. Western North American Naturalist 67:26–38. Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control 19:716-723. Arntzen, J. W., and G. Espregueira Themudo. 2008. Environmental parameters that determine species geographical range limits as a matter of time and space. Journal of Biogeography 35:1177–1186. Attum, O., M. Lee, J. H. Roe, and B. A. Kingsbury. 2008. Wetland complexes and upland- wetland linkages: landscape effects on the distribution of rare and common wetland reptiles. Journal of Zoology 275:245–251. Austin, M. P., and K. P. Van Niel. 2011. Improving species distribution models for climate change studies: variable selection and scale. Journal of Biogeography 38:1–8. Brotons, L., W. Thuiller, M. B. Araujo, and A. H. Hirzel. 2004. Presence-absence versus presence-only modelling methods for predicting habitat suitability. Ecography 4:437– 448. Congdon, J. D., O. M. Kinney, and R. D. Nagle. 2011. Spatial ecology and core-area protection of Blanding’s Turtle (Emydoidea blandingii). National Research Council 89:1098–1106. Elith, J., C. H. Graham, R. P. Anderson, M. Dudik, S. Ferrier, A. Guisan, R. J. Hijmans, F. Huettmann, J. R. Leathwick, A. Lehmann, J. Li, L. G. Lohmann, B. A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J. M. Overton, A.T. Peterson, S. J. Phillips, K. Richardson, R. Scachetti-Pereira, R. E. Schapire, J. Soberon, S. Williams, M. S. Wisz, and N. E. Zimmermann. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29:129–151. Elith, J., and C. H. Graham. 2009. Do they? How do they? Why do they differ? On finding reasons for differing performance of species distribution models. Ecography 32:66–77. Elith, J., S. J. Phillips, T. Hastie, M. Dudik, Y. E. Chee, and C. J. Yates. 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17:43–57. Engler, R., A. Guisan, and L. Rechsteiner. 2004. An improved approach for predicting the distribution of rare and endangered species from occurrence and pseudo-absence data. Journal of Applied Ecology 41:263–274. Ernst, C. H., and J. E. Lovich. 2009. Turtles of the United States and Canada, 2nd edn. Johns Hopkins University Press, USA. Eskildsen, A., P. C. le Roux, R. K. Heikkinen, T. T. Hoye, W. D. Kissling, J. Poyry, M. S. Wisz, and M. Luoto. 2013. Testing species distribution models across space and time: high latitude butterflies and recent warming. Global Ecology and Biogeography 22:1293– 1303. Ferree, C., and M. G. Anderson. 2013. A Map of Terrestrial Habitats of the Northeastern United States: Methods and Approach. The Nature Conservancy, Eastern Conservation Science, Eastern Regional Office, USA. Available at http://www.webcitation.org/6XLaM1ijQ Franklin, J. 2009. Mapping Species Distributions; Spatial Inference and Prediction. Cambridge University Press, USA. Franklin, J., K. E. Wejnert, S. A. Hathaway, C. J. Rochester, and R. N. Fisher. 2009. Effect of species rarity on the accuracy of species distribution models for reptiles and amphibians in southern California. Diversity and Distributions 15:167–177.

Gesch, D. B. 2007 The National Elevation Dataset. Digital Elevation Model Technologies and Applications: The DEM User’s Manual (ed. By Maune, D.), 2nd edn., pp. 99–118. Bethesda, Maryland, American Society for Photogrammetry and Remote Sensing, USA. Gibbons, J. W., D. E. Scott, T. J. Ryan, K. A. Buhlmann, T. D. Tuberville, B. S. Metts, J. L. Greene, T. Mills, Y. Leiden, S. Poppy, and C. T. Winne. 2000. The global decline of reptiles, deja vu amphibians. Bioscience 50:653–66. Gu, W., and R. K. Swihart. 2004. Absent or undetected? Effects of non-detection of species occurrence on wildlife-habitat models. Biological Conservation 116:195–203. Guillera-Arroita, G., J. J. Lahoz-Monfort, J. Elith, A. Gordon, H. Kujala, P. E. Lentini, M. A. McCarthy, R. Tingley, and B.A. Wintle. 2015. Is my species distribution model fit for purpose? Matching data models to applications. Global Ecology and Biogeography. In press. Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147–186. Guisan, A., T. C. Edwards, and T. J. Hastie. 2002. Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecological Modelling 157:89–100. Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones, and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965–1978. Holbrook, J. E. 1838. North American Herpetology; or, a Description of the Reptiles Inhabiting the United States. Ed. 1, Vol. 3, USA. Hortal, J., A. Jimenez-Valverde, J. F. Gomez, J. M. Lobo, and A. Baselga. 2008. Historical bias in biodiversity inventories affects the observed environmental niche of species. Oikos 117:847–858. Jimenez-Valverde, A., and J. M. Lobo. 2007. Threshold criteria for conversion of probability of species presence to either-or presence-absence. Acta Oecologica 31:361–369. Jimenez-Valverde, J. M., A. Lobo, and J. Hortal. 2009. The effect of prevalence and its interaction with sample size on the reliability of species distribution models. Community Ecology 10:196–205. Jin, S., L. Yang, P. Danielson, C. Homer, J. Fry, and G. Xian. 2013. A comprehensive change detection method for updating the National Land Cover Database to circa 2011. Remote Sensing of Environment 132:59–175. Johnson, G., and L. Conrad. 2012. Distribution of wetland-dependent amphibians and reptiles of greatest conservation need in northern New York- an application of the New York State Herps Atlas. State University of New York at Potsdam. Final report to the New York State Department of Environmental Conservation, USA. Kery, M. 2002. Inferring the absence of a species: a case study of snakes. The Journal of Wildlife Management 66:330–338. Khatchikian, C., F. Sangermano, and D. Kendell. 2011. Evaluation of species distribution model algorithms for fine-scale container-breeding mosquito risk prediction. Medical and Veterinary Entomology 25:268–275. Liu, C., P. M. Berry, T. P. Dawson, and R. G. Pearson. 2005. Selecting threshold of occurrence in the prediction of species distributions. Ecography 28:385–393.

MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droedge, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248–2255. Marchand, M. N., and J. A. Litvaitis. 2004. Effects of habitat features and landscape composition on the population structure of a common aquatic turtle in a region undergoing rapid development. Conservation Biology 18:758–676. Marini, M. A., M. Barbet-Massin, L. E. Lopes, and F. Jiguet. 2010. Predicting the occurrence of rare Brazilian birds with species distribution models. Journal of Ornithology 151:857– 866. McCluskey, E. M., S. W. Mockford, K. Sands, T. B. Herman, G. Johnson, and R. A. Gonser. 2016. Population structure of Blanding’s turtle (Emydoidea blandingii) in New York. Journal of Herpetology. In press. Millar, C. S., and G. Blouin-Demers. 2011. Spatial ecology and seasonal activity of Blanding’s Turtles (Emydoidea blandingii) in Ontario, Canada. Journal of Herpetology 45:370–378. Millar, C. S., and G. Blouin-Demers. 2012. Habitat suitability modelling for species at risk is sensitive to algorithm and scale: a case study of Blanding’s turtle, Emydoidea blandingii, in Ontario, Canada. Journal of Nature Conservation 20:18–29. Mockford, S. W., T. B. Herman, M. Snyder, and J. M. Wright. 2007. Conservation genetics of Blanding’s turtle and its application in the identification of evolutionarily significant units. Conservation Genetics 8:209–219. Olivero A. P., and M. G. Anderson. 2008. Northeast aquatic habitat classification system. The Nature Conservancy, Eastern Regional Office, USA. Available at http://www.webcitation.org/6XLaWXJGp Pearson, R. G., C. J. Raxworthy, M. Nakamura, and A. Townsend Peterson. 2007. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34:102–117. Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190:231–259. Phillips, S. J., and M. Dudik. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31:161–175. Quesnelle, P. E., L. Fahrig, and K. E. Lindsay. 2013. Effects of habitat loss, habitat configuration and matrix composition on declining wetland species. Biological Conservation 160:200– 208. Randin, C. F., T. Dirnbock, S. Dullinger, N. E. Zimmermann, M. Zappa, and A. Guisan. 2006. Are niche-based species distribution models transferable in space? Journal of Biogeography 33:1689–1703. Rizkalla, C. E., and R. K. Swihart. 2006. Community structure and differential responses of aquatic turtles to agriculturally induced habitat fragmentation. Landscape Ecology 21:1361–1375. Ross, A. M., and G. Johnson. 2013. DRAFT Recovery Plan for New York State Populations of the Blanding’s turtle (Emydoidea blandingii). New York State Department of Environmental Conservation, USA. Rupprecht, F., J. Oldeland, and M. Finckh. 2011. Modelling potential distribution of the threatened tree species Juniperus oxycedrus: how to evaluate the predictions of different modelling approaches? Journal of Vegetation Science 22:647–659.

Seabrook, L., C. McAlpine, J. Rhodes, G. Baxter, A. Bradley, and D. Lunney. 2014. Determining range edges: habitat quality, climate or climate extremes? Diversity and Distributions 20:95–106. Segurado, P., and M. B. Araujo. 2004. An evaluation of methods for evaluation species distributions. Journal of Biogeography 31:1555–1568. Swets, J. A. 1988. Measuring the accuracy of diagnostic systems. American Association for the Advancement of Science 240:1285–1293. USDA National Agricultural Statistics Service Cropland Data Layer. 2010 Published crop- specific data layer [Online]. USDA-NASS, USA. Available at http://www.webcitation.org/6XLae0S5e U.S Census Bureau. 2011. TIGER/Line Shapefiles [machine-readable data files]. Geography Division. Available at https://www.census.gov/geo/maps-data/data/tiger-line.html U. S. Fish and Wildlife Service. 1983 National Wetlands Inventory website. U.S. Department of the Interior, Fish and Wildlife Service, USA. Available at http://www.webcitation.org/6XLamGHOd Van Dijk, P. P. and A. G. J. Rhodin. 2013. Emydoidea blandingii. The IUCN Red List of Threatened Species. Version 2015.2. Available at http://www.iucnredlist.org

CHAPTER III: Effectiveness of three-tier wetland assessments at evaluating the ecological integrity of wetlands in the St. Lawrence Valley, New York.

Introduction

The Lake Ontario-St. Lawrence River (SLR) Basin is the second largest in New York

State (NY), and the SLR is an important waterway in North America, connecting the Great

Lakes to the Atlantic Ocean. The river and adjacent valley have been used by humans for at least

10,000 years (Thompson et al., 2002). The aquatic, shore, and wetland habitats along the SLR have been identified as important spawning areas for warm water fish; wintering, nesting, and feeding habitat for waterfowl; and significant habitat for many other species including the threatened Blanding’s Turtle (NYSDEC, 1990). Since the SLR Valley has become industrialized and extensively farmed, the natural wetlands, and associated wildlife communities, have been impacted by major habitat alteration such as drainage, fragmentation, industrial pollution, agricultural runoff, and invasive species. Overall, NY is estimated to have lost approximately

60% of its wetlands since the late 18th century. (Dahl, 1990, Dahl, 2011).

The United States (US) Clean Water Act 40 CFR, sections 303(d) and 305(b) require that states and Native American tribal governments report on the quality of waters within their borders, with the principal objective to restore and maintain the chemical, physical, and biological integrity of U.S. waters. As a response, and under the guidance of the US

Environmental Protection Agency (EPA), most states have developed surface water monitoring programs and regularly report the results to a national database. These programs have resulted in interstate water quality reporting consistency, more focused water quality objectives, improved assessment methodologies, more informed standards, and an overall better understanding of the state of waterbodies both in the short and long term; however they have consistently omitted the

64 assessment of wetlands (USEPA, 2003). In the most recent (2004) National Water Quality

Inventory Report to Congress, only 10 states reported on wetland water quality; the total acreage of assessed wetlands covered less than 1% of the nation’s wetland resources (USEPA, 2009).

Recognizing this significant monitoring gap, the EPA has made it a national priority to guide states and tribes in developing comprehensive wetland monitoring strategies (USEPA, 2006).

In 2011 the EPA undertook a nationwide wetland assessment project, the National

Wetland Condition Assessment (NWCA), which was intended to jumpstart the development of assessment methods and metrics for use in individual states (USEPA, 2011). As done for the

NWCA, the EPA suggested that when developing or adapting their wetland monitoring protocols, states and tribes use a hierarchical three-tiered approach (USEPA, 2006). This three- tier framework is an assessment method that uses three intensities of wetland assessment to generate a thorough evaluation of wetland ecological integrity (Fennessy et al., 2007). Ecological integrity is the combination of the biological, chemical, and physical components of a wetland.

The EPA recommends that each level build upon the previous one for the best assessment outcome. As one would expect, the degree of confidence in the output of an assessment increases with greater effort. Level One is landscape level assessment, typically done on a very coarse scale using remotely-sensed data with a Geographic Information System (GIS). A large variety of Level One methods has been developed by states, ranging from very simple (e.g. percent forest cover, Brooks et al. 2004) to complex, multi-metric models (Comer and Hak, 2012). A site visit to a wetland is rarely necessary to complete this level. The outcome of a Level One assessment is an estimation of the ecological integrity of a wetland as it responds to landscape level factors. The entire wetland assemblage of a state can be rapidly assessed using this approach, since a site visit is not necessary. The results of a Level One assessment can be used to

65 target higher (Level Two and Three) assessments, identify priority or critical areas for conservation and management, and characterize habitat quality changes over space and time

(Faber-Langendoen et al., 2008).

A Level Two wetland assessment is a medium intensity evaluation characterized by the

Rapid Assessment Method (RAM). A RAM is a field-based method that employs a suite of field indicators spanning the biological, hydrological, chemical, and functional components of a wetland that are associated with overall wetland integrity; this is often in the form of a checklist to use when making a site visit (Fennessy et al., 2007). Though the final field survey instrument is intended to be easy to use, quick, and reliable, the development of a valid RAM is not a simple process. It requires expert knowledge of regional wetland ecosystems and involves a selection of metrics that respond to a disturbance gradient. Hundreds of metrics may need to be evaluated, calibrated, and independently validated before the final suite is assembled (Faber-Langendoen et al., 2008; Faber-Langendoen et al., 2012a; Sutula et al., 2006). The outcome of a Level Two assessment is a numerical score representing the relative quality of a wetland on a scale from degraded to pristine. Once a RAM is developed, it is practical to conduct a Tier Two assessment at a large number of wetlands.

A Level Three wetland assessment is the most intensive evaluation method, requiring a rigorous data collection process that typically employs a standardized sampling design. Level

Three assessments are most often made by applying an Index of Biotic Integrity (IBI). IBIs provide a quantitative score of ecosystem integrity, and can be created for representative wetland biotic assemblages including birds (Smith-Cartwright and Chow-Fraser, 2011), vegetation

(Lopez and Fennessy, 2002; Mack, 2001a), macroinvertebrates (Burton et al., 1999), fish (Minns et al., 1994), and amphibians (Micacchion, 2004). The underlying assumption of an IBI is that a

66 chosen assemblage successfully integrates a variety of physical, chemical, biological, and hydrological information over time and space and summarizes it as a function of ecological integrity (Faber-Langendoen et al., 2008). IBIs are constructed by characterizing multiple indicator metrics for each chosen assemblage (e.g. species richness, relative abundance, presence/absence of specific species, etc.) and then scoring and weighting the metrics to produce a single ecological integrity score for a wetland (USEPA, 2002a; USEPA, 2002b; Faber-

Langendoen et al., 2012a; Faber-Langendoen et al., 2008). IBIs are not only designed for a specific biological assemblage but also typically for a specific wetland class and a specific geographical region. Because of this specificity, IBIs rarely have a national scope; each state must adapt and validate its own set (USEPA, 2002b). Additionally, IBIs are expensive and time consuming and are rarely used to evaluate large numbers of wetlands. Instead, Level Three assessments are used to refine the outcomes of Level One and Two assessments (USEPA, 2002a;

USEPA, 2002b).

Recognizing the need for a comprehensive statewide wetland evaluation methodology, the NY Natural Heritage Program is currently in the process of developing a New York State- specific monitoring protocol involving each of the three tiers of wetland assessment (Feldmann et al., 2012). The protocol is in the final stages of development; its utility has yet to be determined. The Mid Atlantic Wetland Workgroup (MAWWG) is an EPA funded regional research group working to facilitate the development of monitoring and assessment strategies for wetlands (Chamberlain and Brooks, 2016). Regulators and scientists are working on developing and implementing all levels of the three-tier framework for Mid - Atlantic States, including New

York, although the majority of the research is focused on Delaware tidal wetlands. At the 1998

State of the Lakes Ecosystem Conference (SOLEC), an outcome was the recognition that there

67 was a great need for wetland monitoring and the delineation of valid wetland indicators, particularly for Great Lakes coastal wetlands. Since then, increased research effort has been applied toward developing standard wetland assessment methods (Branzer et al., 2007; Niemi et al., 2006; Niemi et al., 2009; Uzarski et al., 2005). However, this effort has resulted in a substantial range and number of assessment methods. To unify the regional efforts, the Great

Lakes Coastal Wetland Consortium assembled a regional guide that includes a series of protocols to evaluate coastal wetland quality (Burton et al., 2008). The protocols span a variety of biological assemblages and meet the criteria for EPA Level Three wetland assessments. The

Consortium does not, however, provide Level One or Two monitoring protocols for the Great

Lakes basin. Most of the proposed protocols have been validated for the entire Great Lakes basin and therefore can be applied to multiple states, including New York. In the eight years since the conception of the guide, little published literature has been generated using the proposed wetland assessment methods.

The EPA suggests that to fully understand the complexity of a wetland ecosystem, all levels of the three-tier framework should be performed, each one identifying and addressing different management questions, and each increasing level of intensity informing the previous methods (USEPA, 2006). Currently it is widely accepted that a higher level of effort and sampling intensity yields a more precise, accurate, and effective wetland evaluation. This assumption however has rarely been empirically tested. The objectives of our study was to use the three-tier framework to survey 71 freshwater wetlands in the SLR Valley, NY to determine

(1) whether there are significant differences in the way the three approaches assess wetland ecological integrity, and (2) whether lower intensity assessment methods (Level One, Two) could replace the more effort-laden Level Three metrics. We hypothesized that Level Three

68 assessments evaluate wetlands more accurately than lower tier methods because of the incorporation of ecological indicators, but nevertheless lower tier assessment methods can be effective at evaluating wetland ecological integrity, and at a lower level of effort.

Materials and Methods

Study area

The St. Lawrence River Valley (SLV) in NY is approximately 8,200 km2 and encompasses Franklin, Jefferson, and St. Lawrence Counties. The area is defined by the St.

Lawrence River to the north, the Adirondack Mountains to the south and east, and Lake Ontario to the west. The valley is the US section of the St. Lawrence/Great Lakes lowlands/Champlain

Valley, bordered to the north and south by the ancient rocks of the Canadian Shield (Figure 3.1).

The dominant land cover in the SLV is deciduous forest (30%) and the dominant land use is for pasture and hay (23%). Wetlands make up 17% of the valley landscape. We surveyed 71 palustrine wetlands belonging to the emergent, forested, or scrub-shrub class (Cowardin et al.,

1979) ranging in size from 0.2 ha to 104 ha. Thirty-one of the surveyed wetlands were restored under two major federal habitat restoration programs, and 40 were natural wetlands. We attempted to survey wetlands that spanned a gradient of disturbance therefore wetland sites were located in a variety of landscape settings, from agricultural and urban, to forest and wildlife preserve. The landscape setting of surveyed wetlands reflected a variety of stress sources including agricultural runoff, buffer impairment, road salt, road fragmentation, hydrological modification, grazing, and invasive species.

Level One: Landscape Assessment

We used four Level One methods of characterizing each wetland based on the surrounding landscape. Two of the methods used basic checklists and diagrams to determine the disturbance gradient of the wetlands. The Minnesota Disturbance Gradient (MDG) is a score adapted from a checklist developed for depressional wetlands in Minnesota (Gernes and Helgen,

2002). The original score is based on five factors: disturbance within a 50 m buffer, disturbance within a 500 m buffer, habitat, hydrology, and chemical pollution, and ranges from 0 (no evidence of disturbance) to 100 (high level of disturbance). We chose to omit the metric associated with chemical measurements of the water and sediments because those were not within our scope of our work. Our final MDG scores ranged from 0 (no evidence of disturbance) to 75 (high level of disturbance).

Figure 3.1: Seventy-one surveyed palustrine wetlands sites in the St. Lawrence River Valley, New York.

The Ohio Disturbance Gradient (ODG) is a rule-based diagram developed to objectively quantify relative levels of disturbance based on the dominant local landscape characteristics

(Lopez and Fennessy, 2002). The classification system scores each wetland based on the surrounding land cover, the land cover of a 100 m buffer, and hydrological modification in the wetland, and ranges from 1 (relatively high impact) to 24 (relatively low impact). Each wetland was scored using the MDG and the ODG by a single person with the aid of ArcGIS digital orthoimagery and digitally drawn buffers.

Two additional Level One assessments were performed using GIS exclusively. The

Landscape Development Intensity (LDI) index is a procedure developed by Brown and Vivas

(2005) for watersheds and wetlands in Florida. The index is a function of non-renewable energy

(i.e., electricity, fuels, fertilizers, pesticides, irrigation water) use per unit area of land use, from which an emergy coefficient is developed for each land use type. We used the 2011 USGS

NLCD raster file at 30 m resolution to adapt Florida’s LDI coefficients and calculate the LDI index. Brown and Vivas (2005) report emergy coefficients for 27 classes of land use; our NLCD raster used only 16. The Florida LDI coefficients were equated to the most equivalent NLCD land use, guided by Mack (2006) (Table 3.1). For each land use class, we calculated percent cover within a 100 m buffer from the edge of each wetland polygon. The equation for calculating the LDI index is, LDITotal = Σ(%LUi * LDIi) where, LDITotal = the LDI score, %LUi = percent of total area of land use i within a 100 m buffer, and LDIi = landscape development intensity coefficient for land use i (Brown and Vivas, 2005). The LDI index ranges from 1 (landscape with low-intensity uses) to 10 (landscape with high energy-intensive uses).

The second remote assessment was a calculation of % forest cover within a 1 km radius from the center point of each wetland (Forest). Forest cover consisted of the aggregate upland

71 deciduous, evergreen, and mixed NLCD forest classes. This simple metric has been found to be a good predictor of landscape condition (Brooks et al. 2004; Wardrop et al., 2007) where high percent cover of forest represents a less degraded landscape.

Table 3.1: The NLCD land use class emergy coefficients equated to the most equivalent Florida land use classes. The coefficients were used to calculate the LDI for each wetland in this study. Emergy FL land use class NLCD land use class coefficient Natural open water Open water 1.00 Natural system Palustrine forested wetland 1.00 Natural system Palustrine scrub/shrub wetland 1.00 Natural system Palustrine emergent wetland 1.00 Natural system Deciduous forest 1.00 Natural system Evergreen forest 1.00 Natural system Mixed forest 1.00 Natural system Scrub/shrub 1.00 Recreational / open space – low-intensity Bare land 1.83 Improved pasture (without livestock) Grassland/herbaceous 2.77 Improved pasture – low intensity (with livestock) Pasture/Hay 3.41 Single family residential – low intensity Developed, open space 6.92 Agriculture – high intensity Cultivated crops 7.00 Single family residential – high density Developed, low intensity 7.55 Low – intensity commercial Developed, medium intensity 8.00 Central business district (average 2 stories) Developed, high intensity 9.42

Level Two: Rapid Assessment

We used the Ohio Rapid Assessment Method (ORAM) Version 5.0 for wetlands as our sole Level Two assessment (Mack, 2001b). The method was developed for freshwater wetlands in the U.S. state of Ohio for regulatory purposes. Although ORAM was never intended to be used for determining the ecological value of wetlands or quantifying biodiversity, ORAM scores have been found to correlate well with more intensive wetland assessment methods, even outside of Ohio (Peterson and Niemi, 2007; Stapanian, 2004; Mack, 2007). ORAM is comprised of several sections including background information about the size, location, class of wetland,

72 narrative questions about critical and significant wetland habitat, and a quantitative rating. For our study, the narrative rating was omitted because of its specificity to Ohio. The quantitative portion consists of six general metrics representing wetland area, buffer land cover / use, hydrology, habitat types, special wetland classes, and plant communities. Each metric is subdivided into multiple sub-metrics. For our study, we omitted two metrics because of unavailability of information and adjusted the final score to reflect this. The total ORAM score is the sum of all the sub-metric scores and ranges from 0 (very poor condition) to 100 (reference condition). Detailed ORAM field methods and protocols are described elsewhere (Mack, 2001b).

Each of the 71 wetlands was evaluated using ORAM by a single individual in July and August

2013. Out of the 16 metrics and sub-metrics scored, four were evaluated in the laboratory using

ArcGIS and digital orthoimagery and 12 were assessed in the field during a one hour site visit.

Level Three: Intensive Site Assessment

Intensive site assessments involve on-site collection of detailed information on wetland characteristics. The wetland biological assessment data used in this study was collected between

2009 and 2015. We intensively sampled each wetland for birds, anurans, vascular plants, and water quality. Survey protocols for vertebrates were approved by the Clarkson University

Institutional Animal Care and Use Committee and done under permit from the New York State

Department of Environmental Conservation. Bird 10 minute point counts followed by vocalization playback surveys were completed on two mornings per site and were scheduled to coincide with bird breeding. Survey points in each wetland were located on the open water - emergent vegetation interface. Monitoring methods were based on the Standardized North

American Marsh Bird Monitoring Protocol (Bibby et al., 2000; Conway, 2011). We conducted nighttime anuran (frog and toad) calling surveys on three nights scheduled to overlap peak

73 breeding of different anuran species, coinciding with three air temperature ranges (5-9°C, 10-

16°C, and >16 °C). We based monitoring techniques on those described by Heyer et al. (1994) and the Marsh Monitoring Program (2009). We surveyed submerged, emergent, and upland vascular wetland plants using a transect plot method. A transect was located at the vantage point used for bird and anuran surveys and two more transects were located at 50 m intervals away from the first. We placed one meter squared quadrats at three elevations (+20 cm, 0 cm, -20 cm) along each transect; the 0 cm elevation was estimated by observing field indicators of the maximum spring water level line (U.S. Army Corps of Engineers, 1987). At each quadrat, all vascular plants were identified to the lowest taxonomic level and percent cover was recorded for each plant taxon. We collected water samples using clean sampling techniques modeled after

Turk (2001) and other protocols. Each wetland was sampled in two locations using 1 L acid washed polyethylene bottles, at approximately 1 m depth, and within 3 m of the emergent vegetation. We measured temperature and conductivity on site using a YSI Model 600XL probe.

All samples were stored on ice in a cooler and were processed on the same day. The final list of measured water quality metrics included nitrate, temperature, conductivity, turbidity, total phosphorus, pH, and chlorophyll-a. Detailed methods on the biological surveys and water quality measurement are provided in Stryszowska et al (2016).

The species richness of wetland indicators has been shown to be associated with wetland habitat quality (Jorgensen, 2010). We used the species richness per wetland of anurans (ASR), birds (BSR), and vegetation (VSR) as basic Level Three metrics. Recent interest in Level Three assessment has led to the development of more complex multi-metric indices of biological integrity (IBI), a quantitative value of community condition which predictably correlates with the range of conditions across a disturbance gradient (Bried et al., 2013; Stapanian et al., 2013). For

74 our Level Three assessment, we mainly used IBIs developed for and validated in the Great Lakes region, including amphibians (Burton et al., 2008), birds (DeLuca et al., 2004), vegetation

(Burton et al. 2008; Swink and Wilhelm, 1979) and water quality (Chow-Fraser, 2006). The amphibian and bird IBIs are based on classifying the species assemblages into guilds. For the amphibian IBI (AIBI), frogs and toads were classified into woodland species (Hyla versicolor,

Pseudacris crucifer, Pseudacris maculata, Rana sylvatica) and total species. Three metrics were derived from these guilds: relative total species richness, relative woodland species richness, and probability of detection of woodland-associated species. For relative species richness calculations, the total possible frog and toad species richness and the total possible woodland species richness for the region were determined by examining state amphibian distribution maps.

We set a conservative value of 1 for the probability of detection of woodland-associated species because we only visited each wetland site in a single location. The sum of the three metrics was the final AIBI score for each wetland and ranged from 0 (poor frog community structure) to 100

(excellent frog community structure).

For the bird IBI, we used the index of marsh bird community integrity (IMBCI) (DeLuca et al., 2004). The IMBCI combines guild-based community structure with species attributes. Bird species were determined to belong to the wetland obligate guild if they rated a five on

Croonquist and Brook’s (1991) wetland dependence list. Species attributes indicating foraging, nesting, migration, and breeding range were scored and ranged from a score of 1 (generalist) to 4

(specialist) for each attribute. Scores for each attribute were determined by using rankings developed by Croonquist and Brooks (1991) and the bird species guides developed by the

Cornell Lab of Ornithology (see Table 2 in DeLuca et al., 2004). The final IMBCI score is a sum of all the species attributes for all bird species found in a wetland and the number of wetland

75 obligate species. The score starts at 0 (only generalist species present) and goes up indefinitely depending on the number of specialized species that can potentially be found in a wetland.

We used metrics of the Floristic Quality Index to assess the herbaceous vegetation community at each wetland. All metrics were centered on each species’ Coefficient of

Conservatism (C) value (Swink and Wilhelm, 1979), which expresses the propensity of plants to occupy least-altered habitat. Each plant species was assigned a C value using Bried et al. (2012).

Taxa identified only to the genus level were excluded from analysis. All non-native taxa (USDA and NRCS, 2015) were given a C value of zero and were included in all metric calculations.

Using C, the mean C (mC) was calculated for each wetland site by dividing the sum of all C values by the number of plant species. The mC ranges from 0 (wetland occupied by generalist or non-native plant species) to 10 (wetland occupied by native plant species that require unaltered habitat). The mC is then multiplied by the square root of the total species number to yield the floristic quality index (FQI) (Burton et al., 2008; Herman et al., 2001; Feldmann et al., 2012;

Swink and Wilhelm, 1979). The FQI starts at 0 (wetland occupied by generalist or non-native plant species) and increases indefinitely depending on the number of species requiring unaltered wetland habitat in the region. These basic floristic metrics have been found to correlate well with a wetland disturbance gradient (Bried et al. 2013; Lopez and Fennessy, 2002). Following Miller and Wardrop (2006) we calculated I, which adjusts the FQI to be less sensitive to species richness and ranges from a score of 1 to 100. Following Milburn et al. (2007) we calculated the weighed FQI (wFQI) to incorporate the percent cover of each plant species. The wFQI ranges from 0 (wetland occupied in high proportion by generalist or non-native plants) to 1 (wetland occupied in low proportion by generalist or non-native plants). The mean conservatism ratio

(mCR) was calculated as per Burton et al. (2008) by dividing the mC of all species by the mC of

76 only native species. The mCR is a measure of the prevalence of non-native species and ranges from 0 (wetland occupied by non-native species) to 1 (wetland occupied by native species only).

We used a combination of our water quality metrics to calculate a Water Quality Index

(WQI; Chow-Fraser, 2006). WQIs were developed by Chow-Fraser (2006) based on 12 water quality parameters that are significantly related to Great Lakes basin-wide land use stressors and sensitive to road density (deCantanzaro et al., 2009). We used a subset of the full 12 parameter model (Equation #3 in Chow-Fraser, 2006) that best incorporated our water quality metrics: WQI

= 10.753047 – 0.946098 x log TURB – 0.837294 x log COND – 1.319621 x log TEMP –

4.604864 x log pH – 0.387189 x log TP – 0.353713 x log TN – 0.337888 x log CHL. The final

WQI score ranges from -3 (highly degraded) to +3 (excellent).

Statistical analysis

All metrics were adjusted so that the polarity of impairment was the same for all. A

Principal Component Analysis (PCA) was performed on the 16 z-score standardized metrics to determine if Level One, Two, and Three assessments would group together along principal components. We created composite indices by conducting individual PCAs on Level One and

Level Three metrics, and retaining the eigenvectors from the first PC. For the composite Level

Three index, the WQI metric was excluded because of its low correlation with other metrics and its lower sample size. We used a varimax rotation to simplify the interpretation of the composite

PCs. To test how each metric related to the others we used pairwise correlation analyses.

Additionally, we used paired correlation analysis between the composite PCs for Level One and

Three metrics, and ORAM (Level Two). Pearson correlation was calculated when variables were normally distributed and else the Spearman Rank correlation was used. To maintain an experiment-wise Type I error rate of 5%, we used a Bonferroni correction on all paired

77 correlations which results in a p-value of 0.0004. The statistical software R version 2.15.1 (R

Core Team, 2012) was used for all analyses.

Results

Out of the 16 metrics used, five captured the full gradient of possible conditions (Forest,

MDG, ODG, ASR, and wFQI), four had good distributions given a ceiling based on the local species pool (IMBCI, BSR, FQI, and VSR), and the remaining metrics were missing values from either the lowest or highest ends of the gradient (Figure 3.2, Table 3.2). Out of the Level One metrics, the LDI had the worst gradient distribution, indicating that all wetlands were surrounded by low energy-intense land use. The Level Two metric, ORAM, had a good distribution but failed to capture the lowest limits of the scale representing the most degraded wetlands. Out of the 11 Level Three metrics, four were related to species richness and had the upper end of the gradient based on the local species pool, which was difficult to determine. Nevertheless, these four metrics seemed to have a good distribution of disturbance intensity. Out of the remaining seven, the wFQI and the ASR captured the full gradient of disturbance. The other five either failed to capture the high limits of the scale (high ecological integrity; mC and I), low limits of the scale (highly degraded wetlands; mCR and AIBI), or either end (WQI) (Figure 3.2).

We observed some pronounced differences in how metrics scored the same wetland sites.

Level One metrics most frequently indicated high wetland quality for sites with poor species richness of birds and vegetation. When looked at in aggregate, the vegetation IBIs scored wetlands for high habitat quality when ORAM scored them for low. The species richness metrics in aggregate scored wetlands for high habitat quality when the MDG metrics scored them for low. Significant correlations were found between all three levels of assessment methods

(between-level) (Table 3.3). Strong, significant correlations were also found among metrics of the same level (Levels One and Three) (within-level).

Table 3.2: Sixteen wetland assessment metrics, grouped into three levels of intensity, used to rank 71 wetlands on a gradient of ecological integrity. Each metric is followed by the possible range of scores for that metric where the higher score indicates higher ecological integrity, except for LDI where a higher score indicates lower ecological integrity. For some metrics, the maximum obtainable metric value is dictated by the local species pool.

Metric N Mean ± SD Range Level 1 Minnesota Disturbance Gradient (MDG) (0 - 75) 71 49.4 ± 14.2 7 - 72 Ohio Disturbance Gradient (ODG) (0 - 24) 71 15.5 ± 6.1 6 - 23 Landscape Development Index (LDI) (1 - 10) 71 1.7 ± 0.7 1 - 3.8 Percent Forest Cover (Forest) (0 - 100) 71 44.4 ± 18.9 6.2 - 93.2 Level 2 Ohio Rapid Assessment Method (ORAM) (0 - 100) 71 60 ± 13.6 28 - 88 Level 3 Index of Marsh Bird Community Integrity (IMBCI) (0 - Pool) 71 6.9 ± 2.9 2.7 - 15.4 Bird Species Richness (BSR) (0 - Pool) 71 22.5 ± 6.8 8 - 41 Amphibian Index of Biotic Integrity (AIBI) (0 - 100) 71 89.3 ± 16.1 33.3 - 100 Anuran Species Richness (ASR) (0 - 10) 71 4.9 ± 2.0 0 - 9 Mean Coefficient of Conservatism (mC) (0 - 10) 68 2.8 ± 0.4 1.6 - 3.9 Floristic Quality Index (FQI) (0 - Pool) 68 13.9 ± 3.5 6 - 23.4 Weighed Floristic Quality Index (wFQI) (0 - 1) 68 0.44 ± 0.13 0.1 - 0.82 Adjusted Floristic Quality Index (I) (0 - 100) 68 31 ± 3.8 20.5 - 41.4 Mean Conservatism Ratio (mCR) (0 - 1) 68 0.79 ± 0.09 0.53 - 0.94 Vegetation Species Richness (VSR) (0 - Pool) 68 25.8 ± 9.3 8 - 57 Water Quality Index (WQI) (-3 - 3) 63 0.65 ± 0.65 -1.52 - 1.82

Level One metrics

Level One metrics correlated significantly with other Level One metrics (Table 3.3). The strongest within-level correlation was between ODG and LDI. Level One metrics also correlated significantly with ORAM and Level Three metrics. Metrics MDG, ODG, and LDI were significantly correlated with ORAM. The strongest between-level correlation was between

ORAM and MDG. Metric ODG and LDI correlated with five out of the possible 15 metrics, including Level Three metrics, whereas Forest correlated with only one other metric; LDI.

Figure 3.2: The gradients of possible conditions of wetland ecological integrity as captured by the 16 metrics used in this study. For all metrics, except LDI, low metric scores indicate poor ecological integrity.

ORAM

ORAM was significantly correlated with three out of the four Level One metrics and one out of the 11 Level Three metrics; mCR (Table 3.3). The strongest correlation was between

ORAM and MDG. ORAM correlated much more strongly with Level One metrics than Level

Three.

Level Three metrics

Level Three metrics correlated significantly with other Level Three metrics (Table 3.3).

The six plant metrics were significantly and strongly correlated amongst themselves; the strongest correlation being between I and mC. Amphibian species richness (ASR) and the AIBI were also strongly correlated. Metrics mC, FQI, and mCR were the only ones that correlated with Level One metrics. Neither the species richness metrics nor the bird and anuran IBIs correlated outside of Level Three. The metric mCR was the only one that correlated with

ORAM. The strongest between-level correlation was between mCR and ODG. The metric mC correlated significantly with six of the 15 variables, the most out of any other metric. The WQI did not correlate with any other metric (Table 3.3).

Composite PCs

The first PC of the composite Level One metrics explained 58% of the variance among the four metrics and could be represented by equation: y = (0.59 * ODG) + (0.56 * LDI) + (0.50

* MDG) + (0.36 * Forest). The first PC of the composite Level Three metrics explained 43% of the variance among the eleven metrics and could be explained by the equation: y = (0.41 * mC)

+ (0.40 * FQI) + (0.38 * mCR) + (0.38 * I) + (0.30 * ASR) + (0.29 * AIBI) + (0.28 * wFQI) +

(0.27 * BSR) + (0.20 * VSR) + (0.11 * WIMBCI). The Level One composite correlated strongly with both ORAM and the Level Three composite and the Level Three composite correlated strongly with ORAM, however the composites were not as strongly correlated as the majority of the individual metrics (Table 3.3, Figure 3.3).

Discussion

The EPA requires that states monitor and report on the condition of their wetlands as is being currently done for surface water quality. It further suggests that wetland assessment be performed in a buildable process deemed the 1-2-3 framework, which uses three levels of assessment effort. Our study used 16 previously tested wetland assessment metrics of various levels of effort to evaluate the habitat quality of palustrine wetlands in the St. Lawrence River

Valley, New York. We detected significant correlations across multiple wetland condition assessment Tiers, and metrics within Tiers, including Level One, ORAM, and Level Three. This indicates that these metrics are measuring the same attributes of wetlands. The implication of this finding is that there are strong links between the landscape setting (as assessed by Level One) and the biotic wetland community (as assessed by Level Three), which can be used to predict the ecological condition of a wetland using even the most basic of assessment methods. We also found that some of the most commonly used wetland assessment methods, such as surrounding forest cover and water quality, may not be as effective as previously thought at determining ecological condition. There may be much more informative alternatives to these metrics which are not more difficult, costly, or time consuming.

The four Level One metrics performed well at characterizing various disturbance levels.

The metrics ODG, MDG, and Forest captured full gradients of disturbance (Figure 3.2). The metric LDI however represented only low levels of disturbance and did not capture wetlands at higher disturbance intensities. Other studies successfully used LDI to characterize wetland resources (Mack, 2006; Reiss and Brown, 2007). However, the landscape setting for the wetlands in our study may have been too rural to represent some of the high intensity land uses incorporated into the LDI. Forest cover may be a more appropriate landscape level metric to use

Table 3.3: Pearson and Spearman Rank correlations between pairs of 16 wetland assessment metrics. Bonferroni corrected significant results (p < 0.0004) are in bold.

MDG ODG LDI Forest ORAM IMBCI BSR AIBI ASR mC FQI wFQI I mCR VSR WQI MDG ODG 0.63 LDI 0.38 0.72 Forest 0.01 0.35 0.47 ORAM 0.72 0.60 0.51 0.07 IMBCI 0.18 0.06 -0.05 -0.02 0.23 BSR 0.26 0.18 0.03 0.18 0.25 0.54 AIBI 0.09 0.11 0.08 0.15 0.10 0.21 0.62 ASR 0.17 0.09 0.03 0.06 0.18 0.28 0.64 0.89 mC 0.22 0.44 0.42 0.21 0.36 0.06 0.23 0.21 0.26 FQI 0.26 0.45 0.36 0.28 0.37 0.13 0.38 0.34 0.37 0.66 wFQI 0.11 0.12 0.12 -0.03 0.23 -0.06 0.08 0.23 0.29 0.58 0.31 I 0.18 0.35 0.35 0.16 0.28 0.09 0.21 0.16 0.20 0.95 0.55 0.56 mCR 0.23 0.49 0.42 0.22 0.42 -0.09 0.21 0.15 0.24 0.81 0.72 0.42 0.61 VSR 0.23 0.29 0.20 0.19 0.19 0.06 0.29 0.26 0.26 0.09 0.78 -0.01 0.01 0.29 WQI -0.10 0.07 0.24 0.15 0.03 -0.06 -0.14 -0.03 -0.02 0.14 0.20 -0.08 0.19 0.05 0.21

83 when evaluating rural areas rapidly and remotely. The ease of use of the Level One metrics varied slightly and was based mostly on the knowledge of mapping software and familiarity with the wetland sites. The ODG and MDG did not require the use of GIS but did require some knowledge of hydrological modifications to a wetland, which are difficult to ascertain from viewing aerial photos and would, for best results, require a site visit. In addition, scoring wetlands using MDG and ODG requires familiarity with wetland ecology and hydrology. MDG and ODG correlated strongly and significantly with the higher Level Two metric ORAM (Table

3.3); a finding which speaks to the importance of site level knowledge required to perform these

Level One assessments. Whereas ORAM requires a more rigorous wetland visit, background information, and aerial photo interpretation, the correlation of ORAM with ODG and MDG indicates that ORAM scores can potentially be forecast using a much faster assessment.

The metrics Forest and LDI are based on the use of a GIS but, unlike MDG and ODG, can be done fully without visiting a wetland site. Additionally, these assessments can be performed without any knowledge of wetland ecology or hydrology. LDI correlated with only two higher Tier metrics and Forest did not correlate with any (Table 3.3). This observation indicates that forest cover may not be a good indicator of the complex ecology of a wetland, and that overall landscape level analyses may be poor substitutes for the professional judgement of a wetland ecologist. The strengths of the Level One metrics are their ease and speed of use, which reduces the need for time and financial resources. We found that while some Level One metrics performed poorly at assessing wetland ecological integrity, others are strong candidates for substituting higher level metrics.

The rapid assessment method ORAM was the single Level Two assessment used in this study. This method requires a multi-hour visit to a wetland site, familiarity with the surrounding

Figure 3.3: Scatterplots of Composite Level 1 and 3 scores and ORAM. Level 1 Composite and ORAM (Spearman r=0.61, p<0.00001), Level 3 Composite and ORAM (Pearson r=0.41, p<0.00001), and Level 1 Composite and Level 3 Composite (Spearman r=0.43, p<0.00001).

landscape setting, and knowledge of wetland ecology. The strength of this method is the speed with which it can be performed compared to intensive field sampling, which can take days or weeks. The ORAM check list is detailed and takes into consideration a multitude of wetland characteristics such as hydrology, habitat, and landscape setting, which is significantly more information than provided by the Level One assessments. We found strong and significant correlations between ORAM and three out of the four Level One metrics, which suggests that perhaps the extra effort invested in performing ORAM may not have added much information on

85 the condition of a wetland site. If time and resources are available however, it is always prudent to use an assessment method that provides more information on which to evaluate a site. Detailed information about wetland ecology, hydrology, vegetation etc. can provide more direction to wetland management decisions and more useful information for adaptive management applications. Many states have successfully created and adapted their own rapid assessment methods (Fennessy et al., 2007; Sutula et al., 2006) and it would serve New York State to use a rapid assessment method of their own.

Level Three assessments are the most intensive, requiring significant time, labor, and money. Level Three assessments also require a level of expertise, sometimes very specific, much beyond that required by the lower level assessment methods. The rationale behind performing such intensive assessments is the concept of ecological indicators as providers of information on wetland quality. Organisms in a wetland integrate and provide information about the state of a wetland both spatially and temporally. Ecological indicators can represent a complex system in a way that is easier to interpret. Because ecological indicators are an integral living component of the ecosystem, they are thought to offer a much better representation of wetland habitat quality than Level One or Two assessment methods, which rarely include them. In our study, we used 11

Level Three metrics spanning vegetation, birds, anurans, and water quality. Not all of our metrics represented the full disturbance gradient. Many captured either the low or the high end of the gradient scale. The vegetation metrics, in particular, were sensitive to the presence of invasive species and frequently stuck to the lower end of the gradient, which represented disturbed habitat

(ex. I and mC). The majority of the vegetation metrics were also dependent on the coefficient of conservatism (C) values for the plants. The generally low C values for the plant species identified at all of the sites, reduced the ability of these indices to discriminate between wetland

86 conditions. When plant proportional cover was included in the analysis however, the gradient was much better represented. Disturbance gradients were much better represented by the species richness metrics.

As expected, there was considerable collinearity among many of the Level Three metrics, particularly the various vegetation metrics (Table 3.3). This is a useful finding because practitioners wondering which vegetation metric to use, since there is a growing number of metrics being developed (Chamberlain and Brooks, 2016; Milburn et al., 2007; Miller and

Wardrop, 2006), can potentially rest assured that all vegetation metrics indicate a similar condition of a wetland site. Species richness metrics also correlated well with each other, reducing the need to use multiple species assemblages to represent wetland condition.

When correlating Level Three metrics with the lower levels, only three vegetation metrics correlated with Level One or Level Two metrics. From a vegetation perspective, this is an interesting finding because some of the simplest Level One metrics did not take vegetation into consideration at all, but rather looked at landscape setting. Landscape setting, land use patterns, and wetland buffer size, composition, and condition may thus have a strong influence on the vegetative makeup of a wetland. Using lower Level assessments may be a sufficient method to describe at least the vegetative component of a wetland without having to field sample. This is a promising concept as collecting vegetation data in the field is one of the most time consuming methods of evaluating wetlands. The metric ORAM does incorporate a vegetation component, and the fact that one vegetation Level Three metric correlated well with

ORAM suggests that ORAM may be sufficient to evaluate the condition of a wetland.

None of the vertebrate metrics correlated with either of the lower levels suggesting that anurans and birds may be responding to components of a wetland or landscape that are not

87 captured by the lower level assessments. The frog metrics correlated well with the bird metrics suggesting a relationship, albeit an expected one, between the two. Vertebrate metrics also did not correlate with the Level Three plant metrics. Using ecological indicators such as the species richness of vertebrates may be a method that cannot be easily replaced by lower level methods and one that provides important information about the ecological integrity of wetlands.

Alternately, it is possible that neither species richness nor IBIs for vertebrate indicators are good indicators of wetland ecological integrity. Vertebrate ecological indicators seem to be associated with components of wetlands that are not captured by examining the landscape setting or the plant assemblage. For example, frogs and toads may be responding to the presence of fish in a wetland and birds may be responding to community dynamics and movement patterns on a much larger scale than what was captured by the landscape metrics. More research might be needed to better understand what variables drive the variation in vertebrate diversity metrics.

The WQI is the only metric of all 16 used that did not correlate with any other metric either within or between levels. It is surprising that the WQI does not correlate with Level One or

Two metrics since these metrics strongly incorporate the surrounding landscape in their assessment, and landscape is known to affect water quality (deCatanzaro and Chow-Fraser,

2011; Trebitz et al., 2007). It may be that water quality may not be a good representative of wetland habitat quality. Water quality samples offer just a brief snapshot in time of a complex and variable chemical system. For example, throughout a diurnal cycle, the temperature has peaks and lows in response to the sun. Similarly dissolved oxygen peaks during the day when plants are photosynthesizing. Finally, because wetlands tend to be very shallow and have deep sediment deposits, frequent sediment resuspension alters turbidity, temperature, and phosphorus

88 retention (Wang and Mitsch, 2000). The variability in water quality through time and space may be just too great to say anything about long term wetland condition.

We created composite Level One and Level Three metrics by combining all of the metrics from each Level. We hypothesized that a composite of multiple metrics would have a stronger relationship between levels than the individual metrics. We however did not find this relationship to be true; the correlations between the composite metrics were weaker than the majority of individual correlations (Figure 3.3, Table 3.3). Whereas the availability of a composite could alleviate the confusion of choosing a single metric out of the many being developed, our data indicate that single metrics have a stronger relationship between levels. It does not seem necessary to collect and combine multiple metrics at the same level of intensity to successfully evaluate the ecological integrity of a wetland.

Applications

Wetland monitoring and assessment has become a research focus for states and regions as they pursue compliance with the CWA and strive to manage their wetland resources. The collaboration between scientists and regulatory personnel in groups such as MAWWG and the

Great Lakes Coastal Wetland Consortium has generated important literature on individual assessment methods (Burton et al., 2008; Mack, 2007; Sutula et al., 2007; Wardrop et al., 2007) but research to explore and apply the 1-2-3 framework is still scarce. This project aims to understand the relationships between all three levels of the 1-2-3 framework so that practical application of the framework could become more realistic. A correlation between higher level and lower level assessments indicates that the metrics of the three levels vary together. When such a relationship exists, a lower level assessment can potentially be used to evaluate the

89 ecological integrity of a wetland in place of a higher level metric, when time and resources are limited.

The EPA is encouraging states and tribes to regularly report on the condition of their wetlands. Having the flexibility to choose from a variety of metrics, of both low and high intensities, and knowing that the final assessment result will have validity can help move states in the direction of reporting to the EPA. Instead of spending considerable amounts of time developing their own assessment methods, states can use existing metrics and indices to start regularly evaluating their wetland resources. In addition, having the option of using lower level metrics can help governments report on a larger number of wetlands rather than spending significant effort on evaluating just a few wetlands using Level Three methods. We do not suggest that simplistic Level One methods replace excellent programs such as the North

American Amphibian Monitoring Program or the Marsh Monitoring Program. Such programs have been in place for a long time and offer long term, often high resolution, datasets on the community structure of various ecological indicators. Additionally, these programs involve citizen scientists in getting interested in wetland conservation. When governments do not have the resources to carry out or continue such programs, using lower intensity methods can be a feasible solution to wetland monitoring and conservation.

The application of the 1-2-3 framework is not exclusive to the regulatory realm. Wetland conservation initiatives can use the framework and the results of this study to improve their research strategies. The Great Lakes Restoration Initiative (GLRI) for example aims to restore, protect, and enhance 60,000 acres of Great Lakes coastal wetlands in the next five years. The success of this objective will be based on sound wetland assessment procedures, which can benefit from considering the 1-2-3 framework. The application of the framework can extend

90 globally as well. The Ramsar Convention on Wetlands is an international treaty promoting the conservation of wetlands among 168 participating countries. The Convention recognizes the importance of wetland inventory, assessment, and monitoring. Their scientific review panel is continuously working on creating and updating guidance documents to keep participating countries informed about the best way to meet the objectives. Implementing the 1-2-3 framework and the relationships between the Levels observed in this study into the guidance documents can help countries and regions work on meeting their inventory, assessment, and monitoring goals.

References Bibby, C.J., Burgess, N.D., Hill, D.A., Mustoe, S. 2000. Bird census techniques. 2nd Edition. Academic Press, San Diego, CA. Branzer, L.C., Danz, N.P., Niemi, G.J., Regal, R.R., Trebitz, A.S., Howe, R.W., Hanowski, J.M., Johnson, L.B., Ciborowski, J.J.H., Johnston, C.A., Reavie, E.D., Brady, V.J., Sgro, G.V. 2007. Evaluation of geographic, geomorphic and human influences on Great Lakes wetland indicators: A multi-assemblage approach. Ecological Indicators (7):610-635. Bried, J.T., Strout, K,L., Portante, T. 2012. Coefficients of conservatism for vascular flora of New York and New England: inter-state comparisons and expert opinion bias. Northeastern Naturalist (Special Issue 6):101-114. Bried, J.T., Jog, S.K., Matthews, J.W. 2013. Floristic quality assessment signals human disturbance over natural variability in wetland system. Ecological Indicators (34):260- 267. Brooks, R.P., Wardrop, D.H., Bishop, J.A. 2004. Assessing Wetland Condition on a Watershed Basis in the Mid-Atlantic Region Using Synoptic Land-Cover Maps. Environmental Monitoring and Assessment. 94(1-3):9-22. Brown, M.T., Vivas, M.B. 2005. Landscape development intensity index. Environmental Monitoring and Assessment (101):289-309. Burton, T.M., Uzarski, D.G., Gathman, J.P., Genet, J.A., Keas, B.E., Stricker C.A. 1999. Development of a preliminary invertebrate index of biotic integrity for Lake Huron coastal wetlands. Wetlands (19): 869-882. Burton, T.M., Branzer, J.C., Ciborowski, J.J., Grabas, G.P., Hummer, J., Schneider, J., Uzarski, D.G. Editors. 2008. Great Lakes Coastal Wetlands Monitoring Plan. Great Lakes Coastal Wetlands Consortium. Chamberlain, S.J., Brooks, R.P. 2016. Testing a rapid Floristic Quality Index on headwater wetlands in central Pennsylvania. Ecological Indicators (60): 1142-1149. Chow-Fraser, P. 2006. Development of the water quality index (WQI) to assess effects of basin- wide land-use alteration on coastal marshes of the Laurentian Great Lakes, in: Simon, T.P., Stewart, P.M. (Eds.), Coastal Wetlands of the Laurentian Great Lakes: Health, Habitat and Indicators. Authorhouse, Bloomington, IN, pp. 137–166. Comer, P.J., Hak, J. 2012. Landscape Condition in the Conterminous United States. Spatial Model Summary. NatureServe, Boulder, CO. Conway, C.J. 2011. Standardized North American marsh bird monitoring protocol. Waterbirds (3):319-346. Cowardin, L.M., Carter, V., Golet, F.C. 1979. Classification of wetlands and deepwater habitats of the United States. U.S. Department of the Interior, Fish and Wildlife Service. FWS/OBS-79/31 Croonquist, M.J., Brooks, R.P. 1991. Use of avian and mammalian guilds as indicators of cumulative impacts in riparian-wetland areas. Environmental Management (15):701-714. Dahl, T.E. 1990. Wetlands losses in the United States 1780’s to 1980’s. U.S. Department of the Interior, Fish and Wildlife Service, Washington, DC. Dahl, T.E. 2011. Status and trends of wetlands in the conterminous United States 2004 to 2009. U.S. Fish and Wildlife Service, Fisheries and Habitat Conservation, Washington, DC. deCatanzaro, R., Cvetkovic, M., Chow-Fraser, P. 2009. The relative importance of road density and physical watershed features in determining coastal marsh water quality in Georgian Bay. Environmental Management (44):456-467.

92 deCatanzaro, R., Chow-Fraser, P. 2011. Effects of landscape variables and season on reference water chemistry of coastal marshes in eastern Georgian Bay. Canadian Journal of Fisheries and Aquatic Sciences (68):1009-1023. DeLuca, W.V., Studds, C.E., Rockwood, L.L., Marra, P.P. 2004. Influence of land use on the integrity of marsh bird communities in Chesapeake Bay, USA. Wetlands (24):837-847. Feldmann, A.L., Howard, T.G., Spencer, E.A. 2012. Pilot wetland condition assessment of palustrine emergent marsh in the Upper Hudson River watershed. A report prepared for the NYSDEC Division of Water by the New York Natural Heritage Program. Albany, NY. Fennessy, M.S., Jacobs, A.D., Kentula, M.E. 2007. An evaluation of rapid methods for assessing the ecological condition of wetlands. Wetlands (27):543-560. Faber-Langendoen, D.G., Kudray, G., Nordman, C., Sneddon, L., Vance, L., Byers, E., Rocchio, J., Gawler, S., Kittel, G., Menard, S., Comer, P., Muldavin, E., Schafale, M., Foti, T., Josse, C., Christy, J. 2008. Ecological Performance Standards for Wetland Mitigation: An Approach Based on Ecological Integrity Assessments. NatureServe, Arlington, VA. Faber-Langendoen, D., Hedge, C., Kost, M., Thomas, S., Smart, L., Smyth, R., Drake, J., Menard, S. 2012. Assessment of wetland ecosystem condition across landscape regions: A multi-metric approach. Part A. Ecological Integrity Assessment overview and field study in Michigan and Indiana. EPA/600/R-12/021a. U.S. Environmental Protection Agency Office of Research and Development, Washington, DC. Gernes, M.C., Helgen, J.C. 2002. Indexes of biological integrity (IBI) for large depressional wetlands in Minnesota. Minnesota Pollution Control Agency, Biological Monitoring Program, Environmental Outcomes Division, MN. Herman, K.D., Masters, L.A., Penskar, M.R., Reznicek, A.A., Wilhelm, G.S., Brodovich, W.W., Gardiner, K.P. 2001. Floristic quality assessment with wetland categories and examples of computer applications for the state of Michigan. Revised 2nd Edition. Michigan Department of Natural Resources, Wildlife Division, Natural Heritage Program. Heyer, W.R., Donnelly, M.A., McDiarmid, R.W., Hayek, L.C., Foster, M.S., 1994. Measuring and monitoring biological diversity; standard methods for amphibians. London and Washington: Smithsonian Institution Press. Homer, C.G., Dewitz, J.A., Yang, L., Jin, S., Danielson, P., Xian, G., Coulston, J., Herold, N.D., Wickham, J.D., Megown, K. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogrammetric Engineering and Remote Sensing 81:345-354. Jorgensen, S.E. 2010. Application of ecological indicators for the assessment of wetland ecosystem health. In: Eds. Jorgensen SE, Xu FL, Costanza R. Handbook of ecological indicators for assessment of ecosystem health. CRC Press, USA. Lopez, R.D., Fennessy, S.M. 2002. Testing the Floristic Quality Assessment Index as an Indicator of Wetland Condition. Ecological Applications. 12(2): 487-497. Mack, J.J. 2001a. Vegetation Index of Biotic Integrity (VIBI) for Wetlands: ecoregional, hydrogeomorphic, and plant community comparisons with preliminary wetland aquatic life use designations. Final Report to U.S. EPA Grant No. CD985875-01. Ohio Environmental Protection Agency, Division of Surface Water, Wetland Ecology Unit, Columbus, Ohio. 99 pp. Mack, J.J. 2001b. Ohio Rapid Assessment Method for wetlands. Manual for using version 5.0. Ohio EPA Technical Bulletin Wetland/2001-1-1. Ohio

Environmental Protection Agency, Division of Surface Water, 401 Ecology Unit, Columbus, OH. Mack, J.J. 2006. Landscape as a predictor of wetland condition: an evaluation of the landscape development index (LDI) with a large reference wetland dataset from Ohio. Environmental Monitoring and Assessment (120):221-241. Mack, J.J. 2007. Developing a wetland IBI with statewide application after multiple testing iterations. Ecological Indicators (7):864-881. Marsh Monitoring Program, 2009. Participant's handbook for surveying marsh birds. Published by Bird Studies Canada in cooperation with Environment Canada and the U.S. Environmental Protection Agency. Micacchion, M. 2002. Amphibian Index of Biotic Integrity (AmphIBI) for Wetlands. Final Report to U.S. EPA Grant No. CD985875-01. Ohio Environmental Protection Agency, Division of Surface Water, Wetlands Ecology Group, Columbus, OH. Milburn, S.A., Bourdaghs, M., Husveth, J.J. 2007. Floristic Quality Assessment for Minnesota Wetlands. Minnesota Pollution Control Agency, St. Paul, MN. Miller, S.J, Wardrop, D.H. 2006. Adapting the floristic quality assessment index to indicate anthropogenic disturbance in central Pennsylvania wetlands. Ecological Indicators (6):313-326. Minns, C.K., Cairns, V.W., Randall, R.G., Moore, J.E. 1994. An Index of Biotic Integrity (IBI) for fish assemblages in the littoral zone of Great Lakes’ Areas of Concern. Canadian Journal of Fisheries and Aquatic Sciences (51): 1804-1822. Niemi, G.J., Axler, R., Brady, V., Brazner, J., Brown, T., Ciborowski, J.H., Danz, N., Hanowski, J.M., Hollenhorst, T., Howe, R., Johnson, L.B., Johnston, C.A., Reavie, E., Simcik, M., Swackhamer, D. 2006. Environmental indicators of the U.S. Great Lakes coastal region. Report NRRI/TR-2006/11 to the U.S. Environmental Protection Agency STAR Program, ver.1. Agreement R82-8675, Washington DC. Prepared by Great Lakes Environmental Indicators Collaboration, Natural Resources Research Institute, University of Minnesota Duluth. Niemi. G.J., Brady, V.J., Brown, T.N., Ciborowski, J.J.H., Danz, N.P., Ghioca, D.M., Hanowski, J.M., Hollenhorst, T.P., Howe, R.W., Johnson, L.B., Johnston, C.A., Reavie, E.D. 2009. Development of ecological indicators for the U.S. Great Lakes coastal region – A summary of applications in Lake Huron. Aquatic Ecosystem Health & Management (12):77-89. NYSDEC. 1990. St. Lawrence River at Massena Remedial Action Plan, Stage 1. New York State Department of Environmental Conservation, Albany, NY. NYSDEC. 1999. New York State regulatory freshwater wetlands for St. Lawrence County outside of the Adirondack Park (ARC Export: 1999). 1st edn. New York State Department of Environmental Conservation, Albany, NY. Peterson, A.C., Niemi, G.J. 2007. Evaluation of the Ohio rapid assessment method for wetlands in the western Great Lakes: an analysis using bird communities. Journal of Great Lakes Research (Special Issue 3):280-291. R Core Team. 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. Reiss, K.C., Brown, M.T. 2007. Evaluation of Florida palustrine wetlands: application of USEPA levels 1, 2, and 3 assessment methods. EcoHealth (4):206-218.

Smith-Cartwright, L.A., Chow-Fraser, P. 2011. Application of the Index of Marsh Bird Community Integrity to Coastal Wetlands of Georgia Bay and Lake Ontario, Canada. Ecological Indicators. 11(5): 1482-1486. Stapanian, M.A., Waite, T.A., Krzys, G., Mack, J.J., Micacchion, M. 2004. Rapid assessment indicator of wetland integrity as an unintended predictor of avian diversity. Hydrobiologia (520):119-126. Stapanian, M.A., Mack, J., Adams, J.V., Gara, B., Micacchion, M. 2013. Disturbance metrics predict a wetland Vegetation Index of Biotic Integrity. Ecological Indicators (24):129- 126. Stryszowska, K.M., Twiss, M.R., Langen, T.A. 2016. Evaluating Beneficial Use Impairments in wetlands of the Massena Area of Concern using biotic, water quality, and landscape indicators. Journal of Great Lakes Research, In press. Sutula, M.A., Stein, E.D., Collins, J.N., Fetscher, E., Clark, R. 2006. A practical guide for the development of a wetland assessment method: the California experience. Journal of the American Water Resources Association (42):175-175. Swink, F., G. Wilhelm. 1979. Plants of the Chicago region: a checklist of the vascular flora of the Chicago region, with keys, notes on local distribution, ecology, and taxonomy, and a system for evaluation of plant communities. Morton Arboretum, Lisle, IL. Thompson, E., Moss, K., Hunt, D., Novak, P., Sorenson, E., Ruesink, A., Anderson, M., Olivero, A., Ferree, C., Khanna, S. 2002. St. Lawrence – Champlain Valley Ecoregion biodiversity conservation plan, first iteration. The Nature Conservancy. Arlington, VA. Trebitz, A.S., Branzer, J.C., Cotter, A.M., Knuth, M.L., Morrice, J.A., Peterson, G.S., Sierszen, M.E., Thompson, J.A., Kelly, J.R. 2007. Water quality in Great Lakes coastal wetlands: basin-wide patterns and responses to an anthropogenic disturbance gradient. Journal of Great Lakes Research (33):67-85. Turk, J.T. 2001. Field guide for surface water sample and data collection. USDA Forest Service. Air program. US Army Corps of Engineers. 1987. Wetland delineation manual. U.S. Army Engineer Waterways Experiment Station. Vicksburg, MS, USA. Technical Report Y-87-1. USDA FSA APFO. 2011. Digital Ortho Mosaic, National Agriculture Imagery Program (NAIP). USDA FSA APFO Aerial Photography Field Office, Salt Lake City, UT. USDA, NRCS. 2015. The PLANTS Database (http://plants.usda.gov, 13 August 2015). National Plant Data Team, Greensboro, NC 27401-4901 USA. USEPA. 2002a. Methods for Evaluating Wetland Condition: Introduction to Wetland Biological Assessment. Office of Water, U.S. Environmental Protection Agency, Washington, DC. EPA-822-R-02-014. USEPA. 2002b. Methods for Evaluating Wetland Condition: Developing Metrics and Indexes of Biological Integrity. Office of Water, U.S. Environmental Protection Agency, Washington, DC. EPA-822-R-02-016. USEPA. 2003. Elements of a State Water Monitoring and Assessment Program. Office of Wetlands, Oceans, and Watersheds, U.S. Environmental Protection Agency, Washington, DC. EPA-842-B-03-003. USEPA. 2006. Application of Elements of a State Water Monitoring and Assessment Program for Wetlands. Office of Wetlands, Oceans, and Watersheds, U.S. Environmental Protection Agency, Washington, DC. Available from: http://www.epa.gov/owow/wetlands/monitor/

USEPA. 2009. National Water Quality Inventory: Report to Congress. 2004 Reporting Cycle. Office of Water, U.S. Environmental Protection Agency, Washington, DC. EPA-841-R- 08-001. USEPA. 2011. National Wetland Condition Assessment 2011 Draft Technical Report. Office of Water, U.S. Environmental Protection Agency, Washington, DC. EPA-843-R-15-006. USFWS, 1983. National wetlands inventory website. U.S. Department of the Interior, Fish and Wildlife Service, Washington, DC. Uzarski, D.G., Burton, T.M., Cooper, M.J., Ingram, J.W., Timmermans, S.T.A. 2005. Fish habitat use within and across wetland classes in coastal wetlands of the five Great Lakes: Development of a fish-based Index of Biotic Integrity. Journal of Great Lakes Research (31 Supplement):171-187. Wang, N., Mitsch, W.J. 2000. A detailed ecosystem model of phosphorus dynamics in created riparian wetlands. Ecological Modelling (126):101-130. Wardrop, D.H., Kentula, M.E., Stevens, Jr. D.L., Jensen, S.F., Brooks, R.P. 2007. Assessment of wetland condition: an example from the Upper Juanita watershed in Pennsylvania, USA. Wetlands (27):416-431.

CONCLUSION

Wetlands provide important ecosystem services and are one of the most productive ecosystems on earth. The growing body of knowledge about wetland biology, hydrology, chemistry, structure, and function has increased the public appreciation of the indispensable wetland ecosystem services and improved wetland protection laws. The fields of wetland restoration, wetland assessment, and wetland ecology are thriving. Research on the methodologies and success of wetland creation and restoration (Mitsch et al., 1998; Zedler,

2000) is discovering ways to improve how we can increase wetland cover. Research about the ecosystem services that wetlands provide (Banjerjee et al., 2013; Woodward and Wui, 2001;

Zedler, 2003) is increasing awareness about the importance and benefits of wetlands. The EPA conducts new and ongoing wetland research projects (USEPA 2002; USEPA, 2015) and funds regional groups such as MAWWG and GLRI to carry our research that has more localized applications. Some wetland research is done from a landscape perspective (Findlay and

Houlahan, 1997, McElfish et al., 2008) while other projects are focused on either a single species

(Lelong et al., 2007), habitat type (Alsfeld et al., 2010), or location (Markle and Chow-Fraser,

2016). Advances in remote sensing technology are being incorporated into wetland research to observe new relationships between the landscape and biotic assemblages (Quesnelle et al., 2013) or to simply map the distribution of wetland resources (Rebelo et al., 2009). Very rarely however has there been specific insight into how remote sensing and the more traditional wetland assessment approaches could be integrated to produce an improved approach to studying wetlands. This dissertation looked at techniques of combining remote, landscape level wetland

97 assessment, with high intensity field sampling methods with application to local wetland conservation issues.

In chapter I, I looked at how combining both land use/land cover assessment with intensive field sampling can inform critical decisions about the state of a Great Lakes Area of

Concern (AOC), an environmentally degraded area. I found that while the field data indicated no difference between a focal wetland set and a reference wetland set, the landscape data suggested that study wetlands may be lacking in size and surrounding wetland habitat. I concluded that for optimal insight into the state of the AOC, decision makers should consider a combination of ecological indicators and landscape level assessment.

In chapter II, I used a high resolution and long term turtle capture data set, in combination with landscape variables to hypothesize about the habitat requirements of the threatened

Blanding’s turtle. The species distribution model that I created shed light on the current limits of the turtle’s range and even on the potential historical movement patterns of this species. This study is an example of how neither field data nor landscape analysis can stand on their own to fully capture the conservation history and future of a focal species. Field data, on the one hand, can take a great amount of time to collect (in this study 14 years) which can mean very slow progress toward conservation. In addition, the local specificity of turtle trap records limits the spatial extent at which populations can be documented, monitored, and managed. Landscape data, on the other hand, cannot capture the specific habitat requirements of a species that is rare, cryptic, and specialized. It is when these two data sets come together that a clear pattern of habitat use on the small and large scale can be discerned.

In chapter III, I looked for patterns and overlap between three intensities of wetland assessment methods. I found that low intensity methods, such as spatial analysis using remote

98 sensing analysis software, correlate well with high intensity methods, such as field sampling, but not always. It is evident that whereas low intensity methods can sometimes be used to replace high intensity ones, especially when resources are lacking, there is still a need for both of these types of methods. For the most complete picture of the ecological integrity of a wetland, a combination of low, medium, and high intensity methods is best.

Remote sensing technology will likely continue to improve in resolution, spatial coverage, and derived products. We can already observe the surface of the earth at 1m resolution

(IKONOS) and also obtain information about the shape of the surface in three dimension

(LiDAR). Either one of these products could greatly enhance the results of all three chapters of this dissertation. As next steps to the research reported in this dissertation, I would like to use current, high resolution remotely sensed images such as LiDAR, IKONOS, or Landsat to (1) inventory wetland coverage of permanent and seasonal wetlands on the watershed and local scale of the Massena AOC, (2) fine-tune the species distribution model for the Blanding’s Turtle, specifically in understanding what proxy variables are affecting the turtle’s response to elevation, and (3) develop a Level One assessment method that would be a better representative of wetland ecological integrity as assessed by Level Three methods.

Field sampling methods are some of the most fundamental ways in which humans have studied the world around them. Many ecological field methods have gone unchanged for decades, if not centuries, and for good reason, as they are effective. With the advent of technology however, we are starting to see ecosystems from a new, remote sensing perspective.

Remote sensing has allowed us to see patterns in landscapes over large spatial scales, as well as track temporal changes. This dissertation looked at how combining field sampling and remote sensing can be used to study wetland ecosystems. It is evident that neither method alone is as

99 effective as a combination of the two. As remote sensing technology evolves, we must not forget the effectiveness of standing face-to-face with the wetland ecosystem.

100

References Alsfeld, A.J., Bowman, J.L., Deller-Jacobs, A. 2010. The influence of landscape composition on the biotic community of constructed depressional wetlands. Restoration Ecology (18): 370-378. Banerjee, S., SEcchi, S., Fargione, J., Polasky, S., Kraft, S. 2013. How to see ecosystem services: a guide for designing new markets. Frontiers in Ecology and the Environment (11): 297- 304. Findlay, S., Houlahan, J. 1997. Anthropogenic correlates of species richness in Southeastern Ontario Wetlands. Conservation Biology (11):1000-1009. Lelong, B., Lavoie, C., Jodoin, Y., Bellzile, F. 2007. Expansion pathways of the exotic common reed (Phragmites australis) a historical and genetic analysis. Diversity and Distributions (13):430-437. Marle, C.E., Chow-Fraser, P. 2016. Habitat selection by the Blanding’s Turtle (Emydoidea blandingii) on a protected island in Georgian Bay, Lake Huron. Chelonian Conservation and Biology (13): 216-226. McElfish, J.M. Jr., Kihslinger, R.L., Nichols, S. 2008. Setting buffer sizes for wetlands. National Wetlands Newsletter (30): 6-17. Mitsch, W.K., Wu, X., Nairn, R.W., Weihe, P.E., Wang, N., Deal, R., Boucher, C.E. 1998. Creating and Restoring Wetlands. BioScience (48): 1019-1027. Quesnelle, P.E., Fahrig, L., Lindsay, K. 2013. Effects of habitat loss, habitat configuration and matrix composition on declining wetland species. Biological Conservation. (160): 200- 208. Rebelo, L.M., Finlaysonbhatla, N. 2009. Remote sensing and GIS for wetland inventory mapping and changes. Journal of Environmental Management (90): 2144-2153. USEPA. 2002. Methods for evaluating wetland condition, introduction to wetland biological assessment. Office of Water, U.S. Environmental Protection Agency, Washington, DC. EPA-822-R-02-014. USEPA. 2015. Connectivity of Streams & Wetlands to Downstream Waters: A Review & Synthesis of the Scientific Evidence. Office of Research and Development, U.S. Environmental Protection Agency, Washington, DC. EPA-600-R-14-475F Woodward, R.T., Wui, Y. 2001. The economic value of wetland services: a meta-analysis. Ecological Economics (37)-257-270. Zedler, J.B. 2000. Progress in wetland restoration ecology. TREE (15): 402-407. Zedler, J.B. 2003. Wetlands at your service: Reducing impacts of agriculture at the watershed scale. Frontiers in Ecology and the Environment (1): 65-72. .

101

APPENDIX A

Species Lists

102

Table A1.1: Species of Greatest Conservation Need (SGCN) found in Massena AOC and the reference site in Louisville, NY. Massena AOC Massena AOC Louisville Louisville American Bittern BlackOutside-billed of Cuckoo 100 m American Bittern CaspianOutside Ternof 100 m Black Crowned Night Greater YellowlegsRange Bobolink Cooper’sRange Hawk BrownHeron Thrasher Common Loon Brown Thrasher Common Loon Common Tern Pied-billed Grebe Common Tern Northern Harrier Osprey Least Bittern Osprey Scarlet Tanager Pied-billed Grebe Scarlet Tanager Willow Flycatcher Willow Flycatcher Wood Thrush Wood Thrush

103

Table A1.2: Scientific and common names of all bird species identified in the Massena AOC and Louisville study wetlands. This list includes birds observed outside of the designated 100 meter survey radius. Species of Greatest Conservation Need are denoted by an *. Common Name Scientific Name 1 Alder Flycatcher Empidonax alnorum 2 American Bittern* Botaurus lentiginosus 3 American Coot Fulica americana 4 American Crow Corvus brachyrhynchos 5 American Goldfinch Carduelis tristis 6 American Kestrel Falco sparverius 7 American Redstart Setophaga ruticilla 8 American Robin Turdus migratorius 9 American Tree Sparrow Spizella arborea 65 Baltimore Oriole Icterus galbula 10 Barn Swallow Hirundo rustica 11 Belted Kingfisher Ceryle alcyon 12 Black-and-white Warbler Mniotilta varia 13 Black-billed Cuckoo* Coccyzus erythropthalmus 14 Black-capped Chickadee Poecile atricapilla 15 Black-crowned Night Heron* Nycticorax nycticorax 16 Blackpoll Warbler Dendroica striata 17 Black-throated Green Warbler Dendroica virens 18 Blue Jay Cyanocitta cristata 19 Bobolink* Dolichonyx oryzivorus 20 Brown Creeper Certhia americana 22 Brown Thrasher* Toxostoma rufum 21 Brown-headed Cowbird Molothrus ater 23 Canada Goose Branta canadensis 24 Caspian Tern* Sterna caspia 25 Cedar Waxwing Bombycilla cedrorum 26 Chestnut-sided Warbler Dendroica pensylvanica 27 Chipping Sparrow Spizella passerina 28 Cliff Swallow Petrochelidon pyrrhonota 30 Common Grackle Quiscalus quiscula 31 Common Loon* Gavia immer 32 Common Merganser Mergus merganser 33 Common Moorhen Gallinula chloropus 34 Common Raven Corvus corax 95 Common Snipe Gallinago gallinago 35 Common Tern* Sterna hirundo 36 Common Yellowthroat Geothlypis trichas 37 Coopers Hawk* Accipiter cooperii 38 Double-crested Cormorant Phalacrocorax auritus 39 Downy Woodpecker Picoides pubescens 40 Eastern Kingbird Tyrannus tyrannus

104

Common Name Scientific Name 41 Eastern Meadowlark Sturnella magna 42 Eastern Phoebe Sayornis phoebe 81 Eastern Towhee Pipilo erythrophthalmus 43 Eastern Wood-Peewee Contopus virens 44 European Starling Sturnus vulgaris 45 Field Sparrow Spizella pusilla 46 Gray Catbird Dumetella carolinensis 47 Great Blue Heron Ardea herodias 48 Great-crested Flycatcher Myiarchus tyrannulus 49 Greater Yellowlegs* Tringa melanoleuca 50 Green Heron Butorides virescens 51 Green-winged Teal Anas crecca 52 Hairy Woodpecker Picoides villosus 53 Herring Gull Larus argentatus 54 House Sparrow Passer domesticus 55 House Wren Troglodytes aedon 56 Indigo Bunting Passerina cyanea 57 Killdeer Charadirus vociferus 58 Least Bittern* Ixobrychus exilis 59 Least Flycatcher Empidonax minimus 60 Mallard Anas platyrhynchos 61 Marsh Wren Cistothorus palustris 62 Mourning Dove Zenaida macroura 63 Northern Cardinal Cardinalis cardinalis 29 Northern Flicker Colaptes auratus 64 Northern Harrier* Circus cyaneus 80 Northern Rough-winged Swallow Stelgidopteryx serripennis 66 Osprey* Pandion haliaetus 67 Ovenbird Seiurus aurocapillus 68 Pied-billed Grebe* Podilymbus podiceps 69 Pileated Woodpecker Dryocopus pileatus 70 Pine Warbler Dendroica pinus 72 Purple Finch Carpodacus purpureus 73 Purple Martin Progne subis 74 Red-eyed Vireo Vireo olivaceus 76 Red-tailed Hawk Buteo jamaicensis 75 Red-winged Blackbird Agelaius phoeniceus 77 Ring-billed Gull Larus delawarensis 78 Rock Dove Columba livia 79 Rose-breasted Grosbeak Pheucticus ludovicianus 82 Savannah Sparrow Passerculus sandwichensis 83 Scarlet Tanager* Piranga olivacea 84 Song Sparrow Melospiza melodia 71 Spotted Sandpiper Actitis macularia

105

Common Name Scientific Name 85 Swamp Sparrow Melospiza georgiana 86 Tree Swallow Tachycienta bicolor 87 Turkey Vulture Cathartes aura 88 Veery Catharus fuscescens 89 Virginia Rail Rallus limicola 90 Warbling Vireo Vireo gilvus 91 White-breasted Nuthatch Sitta carolinensis 92 White-throated Sparrow Zonotrichia albicollis 93 Wild Turkey Meleagris gallopavo 94 Willow Flycatcher* Empidonax traillii 96 Wood Duck Aix sponsa 97 Wood Thrush* Hylocichla mustelina 101 Yellow Warbler Dendroica petechia 98 Yellow-bellied Sapsucker Sphyrapicus varius 99 Yellow-billed Cuckoo Coccyzus americanus 102 Yellow-rumped Warbler Dendroica coronata 100 Yellow-throated Vireo Vireo flavifrons

Table A1.3: Scientific and common names of the fish species identified in the study wetlands. Common Name Scientific Name 1 Bluegill Lepomis macrochirus 2 Bluntnose Minnow Pimephales notatus 3 Brook Stickleback Culaea inconstans 4 Brown Bullhead Ictalurus nebulosus 5 Central Mudminnow Umbra limi 6 Channel Darter Percina copelandi 7 Emerald Shiner Notropis atherinoides 8 Fathead Minnow Pimephales promelas 9 Largemouth Bass Micropterussalmoides 10 Longnose Dace Rhinichthys cataractae 11 Northern Redbelly Dace Phoxinus eos 12 Pumpkinseed Lepomis gibbosus 13 Rock Bass Ambloplites rupestris 14 Smallmouth Bass Micropterus dolomieui 15 Yellow Perch Perca flavescens

106

Table A1.4: Scientific and common names of all frog species identified in the study wetlands. Highlighted are SGCN species. Species of Greatest Conservation Need are denoted by an *. Common Name Scientific Name 1 American toad Bufo americanus 2 Boreal chorus frog* Pseudacris maculata 3 Bullfrog Rana catesbeiana 4 Green frog Rana clamitans 5 Grey treefrog Hyla versicolor 6 Mink frog Rana septentrionalis 7 Northern leopard frog Rana pipiens 8 Spring peeper Pseudacris crucifer

107

Table A1.5: Scientific and common names of the vascular plant species identified in the Massena AOC and Louisville study wetlands. Species classified as non-native by the USDA Plants Database are designated by an *. Species classified by the New York State Department of Environmental Conservation as aggressive invasive are designated by a †. The rank represents whether or not the species is classified as a wetland dependent (OBL), facultative wetland (FACW), facultative (FAC), or upland (UPL) species.

Common name Scientific name Rank 1 Agrimony sp. Agrimonia sp. 2 Alfalfa* Medicago sativa 3 Allegheny monkeyflower Mimulus ringens OBL 4 Amerian hogpeanut Amphicarpaea bracteata FAC 5 Amerian water horehound Lycopus americanus OBL 6 Amerian white waterlily Nymphaea odorata OBL 7 American basswood Tilia americana FACU 8 American beech Fagus grandifolia FACU 9 American eelgrass Vallisneria americana OBL 10 American hazelnut Corylus americana FACU 11 American hornbeam Carpinus caroliniana FAC 12 American red raspberry Rubus idaeus FACU 13 Arrow leaf tear thumb Polygonum sagittatum OBL 14 Aster horsetail Equisetum fluviatile OBL 15 Bald spikerush Eleocharis erythropoda 16 Balsma groundsel Packera paupercula FAC 17 Bebb willow Salix bebbiana FACW 18 Bebb's sedge Carex bebbii OBL 19 Bird vetch* Vicia cracca 20 Bird's foot trefoil* Lotus corniculatus FACU 21 Blacksnakeroot sp. Sanicula sp. 22 Blue skullcap Scutellaria lateriflora OBL 23 Blunt broom sedge Carex tribuloides FACW 24 Blunt spikerush Eleocharis obtusa OBL 25 Bog goldenrod Solidago uliginosa OBL 26 Bristly dewberry Rubus hispidus FACW 27 Brittlestem hempnettle* Galeopsis tetrahit FACU 28 Broadfruit bur-reed Sparganium eurycarpum OBL 29 Broadleaf arrowhead Sagittaria latifolia OBL 30 Broadleaf cattail Typha latifolia OBL 31 Broom sedge Carex scoparia FACW 32 Bulblet bearing water hemlock Cicuta bulbifera OBL 33 Calico aster Symphyotrichum lateriflorum FAC 34 Canada goldenrod Solidago altissima FACU 35 Canada goldenrod Solidago canadensis FACU 36 Canadian horseweed Conyza canadensis FACU

108

Common name Scientific name Rank 37 Canadian rush Juncus canadensis OBL 38 Canadian waterweed Elodea canadensis OBL 39 Climbing nightshade* Solanum dulcamara FAC 40 Codlins and cream* Epilobium hirsutum FACW 41 Common boneset Eupatorium perfoliatum FACW 42 Common buckthorn* Rhamnus cathartica FAC 43 Common cinquefoil Potentilla simplex FACU 44 Common dandelion* Taraxacum officinale FACU 45 Common duckmeat Spirodela polyrhiza OBL 46 Common duckweed Lemna minor OBL 47 Common frogbit*† Hydrocharis morsus-ranae OBL 48 Common milkweed Asclepias syriaca UPL 49 Common plantain* Plantago major FACU 50 Common reed*† Phragmites australis FACW 51 Common selfheal* Prunella vulgaris FAC 52 Common spikerush Eleocharis palustris OBL 53 Coon's Tail Ceratophyllum demersum OBL 54 Crack willow* Salix fragilis FAC 55 Crested sedge Carex cristatella FACW 56 Curly dock* Rumex crispus FAC 57 Curly pondweed*† Potamogeton crispus OBL 58 Curlytop knotweed* Polygonum lapathifolium FACW 59 Cypress-like sedge Carex pseudocyperus OBL 60 Devil's beggatick Bidens frondosa FACW 61 Dotted howthorn Crataegus punctata 62 Dudley's rush Juncus dudleyi FACW 63 Earth loosestrife Lysimachia terrestris OBL 64 Eastern daisy fleabane Erigeron annuus FACU 65 Eastern marsh fern Thelypteris palustris FACW 66 Eastern poison ivy Toxicodendron radicans FAC 67 Eurasian water milfoil*† Myriophyllum spicatum OBL 68 European bur-reed Sparganium emersum OBL 69 False baby's breath* Galium mollugo 70 Field bindweed* Convolvulus arvensis UPL 71 Field horsetail Equisetum arvense FAC 72 Flatstem pondweed Potamogeton zosteriformis OBL 73 Flat-top goldenrod Euthamia graminifolia FACW 74 Floating pondweed Potamogeton natans OBL 75 Flowering rush*† Butomus umbellatus OBL 76 Fowl bluegrass Poa palustris FACW 77 Fox sedge Carex vulpinoidea OBL 78 Fragrant bedstraw Galium triflorum FACU

109

Common name Scientific name Rank 79 Fringed loosestrife Lysimachia ciliata FAC 80 Giant goldenrod Solidago gigantea FACW 81 Golden zizia Zizia aurea FAC 82 Gray dogwood Cornus racemosa FAC 83 Green alder Alnus viridis FAC 84 Green ash Fraxinus pennsylvanica FACW 85 Green bulrush Scirpus atrovirens OBL 86 Groundnut Apios americana FACW 87 Hairy goldenrod Solidago hispida 88 Hedge false bindweed* Calystegia sepium FAC 89 Hyssop sp. Agastache sp. 90 Intermediate woodfern Dryopteris intermedia FAC 91 Interrupted fern Osmunda claytoniana FAC 92 Jewelweed Impatiens capensis FACW 93 Jointleaf rush Juncus articulatus OBL 94 Knotted rush Juncus nodosus. OBL 95 Large St. Johnswort Hypericum majus FACW 96 Leafy pondweed Potamogeton foliosus OBL 97 Little green sedge Carex viridula OBL 98 Longhair sedge Carex comosa OBL 99 Mannagrass sp. Glyceria sp. 100 Marsh seedbox Ludwigia palustris OBL 101 Marsh skullcap Scutellaria galericulata OBL 102 Meadow willow Salix petiolaris FACW 103 Missouri river willow Salix eriocephala FACW 104 Narrowleaf cattail* Typha angustifolia OBL 105 Needle spikerush Eleocharis acicularis OBL 106 Northern green rush Juncus alpinoarticulatus OBL 107 Northern water plantain Alisma triviale OBL 108 Northern wildrice Zizania palustris OBL 109 Poverty oatgrass Danthonia spicata 110 Purple loosestrife*† Lythrum salicaria OBL 111 Pussy willow Salix discolor FACW 112 Queen Anne's lace* Daucus carota 113 Red clover* Trifolium pratense FACU 114 Red maple Acer rubrum FAC 115 Redosier dogwood Cornus sericea 116 Reed canarygrass*† Phalaris arundinacea FACW 117 Rice cutgrass Leersia oryzoides OBL 118 Riverbank grape Vitis riparia FAC 119 Rough avens Geum laciniatum FACW 120 Roundleaf dogwood Cornus rugosa

110

Common name Scientific name Rank 121 Sensitive fern Onoclea sensibilis FACW 122 Shortspike watermilfoil Myriophyllum sibiricum OBL 123 Showy goldenrod Solidago erecta UPL 124 Silky dowgood Cornus amomum FACW 125 Silverweed cinquefoil Argentina anserina FACW 126 Slenderleaf false foxglove Agalinis tenuifolia FACW 127 Smallflower false foxglove Agalinis paupercula OBL 128 Smallspike false nettle Boehmeria cylindrica OBL 129 Smooth blackberry Rubus canadensis 130 Softstem bulrush Scirpus tabernaemontani 131 Spotted joe pye weed Eutrochium maculatum OBL 132 Spotted lady's thumb* Polygonum persicaria FAC 133 Steeplebush Spiraea tomentosa FACW 134 Strawcolored flatsedge Cyperus strigosus FACW 135 Sulphur cinquefoil* Potentilla recta 136 Swamp fly honesuckle Lonicera oblongifolia OBL 137 Swamp milkweed Asclepias incantata OBL 138 Swamp thistle Cirsium muticum OBL 139 Swamp verbena Verbena hastata FACW 140 Swamp white oak Quercus bicolor FAC 141 Sweet crab apple Malus coronaria 142 Sweetclover sp. Melilotus sp. FACU? 143 Threelobe beggaticks Bidens tripartita FACW 144 Threepetal bedstraw Galium trifidum FACW 145 Timothy* Phleum pratense FACU 146 Trumpet creeper Campsis radicans FAC 147 Variableleaf pondweed Potamogeton gramineus OBL 148 Variegated scouringrush Equisetum variegatum FACW 149 Virginia marsh St. Johnswort Triadenum virginicum OBL 150 Virginia strawberry Fragaria virginiana FACU 151 Virginia threeseed mercury Acalypha Virginica FACU 152 Virginia water horehound Lycopus virginicus OBL 153 Water knotweed Polygonum amphibium OBL 154 Water shield Brasenia schreberi OBL 155 Watermilfoil Myriophyllum sp. OBL 156 Whire rod Viburnum nudum FACU 157 White meadowsweet Spiraea alba FACW 158 White panicle aster Symphyotrichum lanceolatum FACW 159 Wild parsnip* Pastinaca sativa 160 Willowleaf aster Symphyotrichum praealtum FACW 161 Woodbine Parthenocissus vitacea FACU 162 Wool grass Scirpus cyperinus OBL

111

Common name Scientific name Rank 163 Woolly sedge Carex pellita OBL 164 Wrinkle leaf goldenrod Solidago rugosa FAC 165 Yellowfruit sedge Carex annectens FACW

112

APPENDIX B Worksheets for Level One and Two wetland bioassessment methods

113

Table A2.1: Checklist used to score 71 wetland sites for the Minnesota Disturbance Gradient (MDG), a level 1 metric (Gernes and Helgen, 2002). The maximum obtainable score is 75 points. Factor 1. Buffer landscape disturbance Best – as expected for reference site, no evidence of disturbance (0) Mod.- predominately undisturbed, some human use influence (6) Fair – significant human influence, buffer area nearly filled with human use (12) Poor–nearly all or all of the buffer human use, intensive landuse surrounding wetland (18) Factor 2. Landscape (immediate) Influence Best – landscape natural, as expected for reference site, no evidence of disturbance (0) Mod.- predominately undisturbed, some human use influence (6) Fair – significant human influence, landscape area nearly filled with human use (12) Poor – nearly all or all of the landscape in human use, isolating the wetland (18) Factor 3. Habitat alteration—immediate landscape Best – as expected for reference, no evidence of disturbance (0) Mod. –low intensity alteration or past alteration that is not currently affecting wetland (6) Fair – highly altered, but some recovery if previously altered (12) Poor – almost no natural habitat present, highly altered habitat (18) Factor 4. Hydrologic alteration Best – as expected for reference, no evidence of disturbance (0) Mod. –low intensity alteration or past alteration that is not currently affecting wetland (7) Fair – less intense than “poor”, but current or active alteration. (14) Poor – currently active and major disturbance to natural hydrology (21)

114

Table A2.2: Checklist used to score 71 wetland sites for the Ohio Disturbance Gradient (ODG), a level 1 metric (Lopez and Fennessy, 2002).

Forested and/or Natural Grassland Land Cover Surrounding Site Forested Buffer Little Hydrologic Modification 1 Human modified hydrology 2 Grass Buffer Little Hydrologic Modification 3 Human modified hydrology 4 No Buffer Little Hydrologic Modification 5 Human modified hydrology 6 Fallow Crop Land or Pasture Land Cover Surrounding Site Forested Buffer Little Hydrologic Modification 7 Human modified hydrology 8 Grass Buffer Little Hydrologic Modification 9 Human modified hydrology 10 No Buffer Little Hydrologic Modification 11 Human modified hydrology 12 Row Crop Agriculture Land Cover Surrounding Site Forested Buffer Little Hydrologic Modification 13 Human modified hydrology 14 Grass Buffer Little Hydrologic Modification 15 Human modified hydrology 16 No Buffer Little Hydrologic Modification 17 Human modified hydrology 18 Urban Land Cover Surrounding Site Forested Buffer Little Hydrologic Modification 19 Human modified hydrology 20 Grass Buffer Little Hydrologic Modification 21 Human modified hydrology 22 No Buffer Little Hydrologic Modification 23 Human modified hydrology 24

115

Table A2.3: Sample Ohio Rapid Assessment Method (ORAM) worksheet (Mack, 2001a), a level 2 metric, used to score 71 wetland sites. Wetland class was either Forested, Emergent, or Shrub/Scrub as classified by Cowardin and others (1979). Metric 5 received an automatic score of ten if: two out of four GIS based significant habitat areas were present (Bird Conservation Region (BCR13) waterbird, shorebird, landbird focus areas, NYS significant coastal habitat). Metrics completed in the laboratory are designated by an * and metrics completed in the field are designated by a †. Date Name of Wetland Vegetation Communities Wetland Class Wetland Size Lat/Long of sampling site USGS Quad name County Comments:

Quantitative: Score Metric 1: Wetland area (Max 6 pts)* Metric 2: Upland buffers (Max 14 pts) 2a (Average buffer width)* 2b (Intensity if land uses)† Metric 3: Hydrology (Max 27 pts) 3a (Source of water)† 3b (Connectivity)* 3c (Max water depth)† 3d (Duration of inundation/saturation)† 3e (Modification of natural hydrologic regime)† Metric 4: Habitat Alteration (Max 20 pts) 4a (Substrate/Soil Disturbance)† 4b (Habitat development)† 4c (Habitat alteration)† Metric 5: Special wetland communities (Max 10 pts)* Metric 6: Vegetation (Max 23 pts) 6a (Vegetation communities)† 6b (Horizontal interspersion)† 6c (Coverage of invasive plant species)† 6d (Microtopography)†

116

APPENDIX C

Wetland scores for 71 sites using 16 bioassessment metrics

117

Table A3.1: Site scores for 71 wetlands using 16 assessment metrics of three levels of intensity.

Site Name Restored MDG ODG Forest LDI ORAM IMBCI BSR mC FQI I wFQI mCR VSR WQI AIBI ASR Beeler R 55 22 0.87 1.10 67 5.10 20 3.38 23.38 35.66 0.49 0.90 48 0.78 74 3 Brasher N 49 22 0.52 1.00 64 6.17 21 2.76 18.73 30.78 0.36 0.80 46 1.21 100 6 BridgesL R 71 15 0.45 2.77 79 9.07 23 2.28 11.40 26.87 0.44 0.72 25 0.69 100 9 BridgesU R 41 20 0.52 1.64 46 10.03 24 2.26 17.09 26.36 0.32 0.74 57 0.66 100 6 Cornett N 71 23 0.41 1.07 77 4.48 18 3.17 17.08 32.88 0.48 0.93 29 0.72 74 3 HMP R 50 22 0.62 1.06 58 4.89 18 2.51 15.69 28.18 0.31 0.79 39 1.05 100 6 Jackson R 34 8 0.11 2.57 50 3.36 19 2.15 11.16 27.07 0.41 0.63 27 0.58 100 5 JoboN R 34 8 0.33 3.75 30 11.46 22 1.58 6.88 21.76 0.17 0.53 19 0.01 74 3 Lewis R 57 8 0.13 2.76 50 10.22 20 2.67 14.61 29.21 0.36 0.83 30 0.94 83 5 Montroy R 36 16 0.51 1.72 51 7.76 18 3.15 16.08 34.96 0.39 0.88 26 0.63 83 4 Pagel2 R 48 8 0.29 2.98 60 6.39 23 2.11 11.15 27.04 0.28 0.61 28 1.03 83 4 Poore R 46 22 0.26 1.00 51 4.43 18 3.00 15.87 35.86 0.47 0.82 28 1.26 74 3 Rieksts R 63 22 0.52 1.04 75 5.63 19 2.74 15.95 30.69 0.38 0.79 34 0.85 66 2 Robinson R 71 23 0.35 1.23 88 5.58 16 3.08 18.98 32.08 0.40 0.92 38 1.59 74 3 SevenSprings N 56 23 0.93 1.04 61 3.64 16 3.21 17.01 34.02 0.51 0.89 28 0.73 100 5 Slack N 63 23 0.54 1.13 73 7.59 26 2.94 16.34 34.45 0.44 0.87 31 1.08 100 7 Zufall R 38 8 0.21 2.73 46 6.63 21 2.74 16.87 30.30 0.36 0.82 38 0.70 100 6 Barkley R 61 22 0.68 1.00 66 8.95 21 NA NA NA NA NA NA 1.10 99 4 Bartlette1 R 42 8 0.27 2.69 49 14.54 27 2.90 13.30 33.30 0.38 0.76 21 -0.38 100 7 Beattie1 N 47 23 0.79 1.00 65 13.94 41 3.50 17.90 37.20 0.41 0.88 26 0.78 100 5 Beattie3 R 62 16 0.58 1.77 67 11.52 30 3.00 13.10 35.00 0.39 0.74 19 0.68 100 7 Brennan R 54 10 0.13 3.60 61 8.72 27 2.80 12.90 32.20 0.65 0.76 21 0.29 100 7 Buckley R 45 14 0.48 2.64 55 3.10 25 2.60 14.00 28.60 0.54 0.86 28 0.74 100 7 Clarkson N 55 11 0.46 1.14 66 9.96 39 3.40 17.80 34.80 0.53 0.93 28 1.13 100 8 CrookedCreek N 58 23 0.66 1.00 76 6.62 26 3.50 19.60 36.40 0.46 0.94 31 1.06 100 8 Cutway R 50 10 0.20 1.04 77.5 2.67 23 2.70 13.60 29.00 0.36 0.88 25 0.22 100 8 Emrich N 32 14 0.58 1.75 51 8.96 25 3.10 15.70 32.70 0.51 0.89 26 0.56 100 7 FishCreek1 N 62 21 0.60 1.34 67 8.03 30 2.40 13.70 27.40 0.38 0.74 34 0.93 100 6

118

Site Name Restored MDG ODG Forest LDI ORAM IMBCI BSR mC FQI I wFQI mCR VSR WQI AIBI ASR FishCreek2 N 60 21 0.38 2.01 58 9.69 24 2.80 15.60 31.90 0.48 0.77 31 0.95 100 5 Gould R 32 10 0.06 2.96 54 6.14 33 2.70 13.60 30.40 0.40 0.80 25 0.52 100 6 Hebb N 66 23 0.45 1.21 77 12.42 32 NA NA NA NA NA NA NA 100 6 Jasikoff R 30 8 0.32 2.67 42.5 3.45 21 3.40 19.70 35.40 0.58 0.91 34 1.09 100 7 Jewett R 61 10 0.51 1.37 66 4.35 33 3.40 20.10 35.00 0.49 0.94 35 0.50 99 4 Kogut R 44 8 0.37 2.55 46 11.12 29 2.90 13.40 33.60 0.52 0.73 22 1.73 100 8 Kring N 51 21 0.32 2.35 72 15.41 34 3.00 13.10 31.70 0.48 0.90 19 -0.39 100 7 LonesomeBay N 67 23 0.72 1.00 59 4.85 27 2.80 14.90 29.90 0.49 0.89 28 0.61 100 7 McCormick1 R 56 20 0.76 1.76 57 5.57 37 2.70 14.10 30.70 0.46 0.78 27 -0.02 100 6 McCormick2 N 65 23 0.69 1.00 79 5.58 32 3.00 16.10 32.10 0.50 0.89 28 0.98 100 7 Peters R 49 22 0.78 1.18 57 5.55 28 NA NA NA NA NA NA 0.40 100 6 Rausch R 32 8 0.30 3.07 32 3.50 15 2.70 14.60 31.10 0.57 0.73 30 0.70 100 6 Scarlett R 50 8 0.66 1.76 63 6.90 25 2.30 10.30 28.50 0.46 0.65 20 1.42 100 6 Snyder R 51 8 0.36 2.18 71 9.50 38 2.80 16.10 32.20 0.48 0.78 32 0.65 100 5 Spencer N 47 23 0.66 1.00 74 9.09 28 3.60 19.90 37.60 0.46 0.93 30 1.08 99 4 Ward R 61 20 0.25 1.87 64 8.98 28 2.80 18.90 31.00 0.44 0.84 44 0.95 100 7 1 N 45 15 0.54 1.15 42 3.55 22 2.80 10.00 31.60 0.16 0.77 13 NA 100 5 9 N 44 23 0.44 1.11 69 4.81 18 3.90 11.00 41.40 0.82 0.88 8 NA 93 5 60 N 31 8 0.41 1.29 62 6.04 13 2.40 12.50 27.10 0.43 0.76 27 0.98 42 1 61 N 49 8 0.46 2.35 44 5.25 18 2.50 9.40 31.20 0.30 0.64 14 NA 66 2 139 N 61 17 0.43 1.29 82 9.88 21 3.10 16.30 32.50 0.72 0.89 28 -0.03 100 6 176 N 31 14 0.45 1.55 62 6.16 16 2.50 9.40 29.40 0.28 0.68 14 1.26 83 2 641 N 63 15 0.41 1.28 62 4.71 14 2.70 10.30 31.10 0.55 0.73 15 0.49 83 4 741 N 61 15 0.40 1.52 70.5 7.07 21 2.60 12.90 29.50 0.35 0.79 24 0.66 50 2 902 N 41 14 0.41 2.03 41 6.64 21 2.40 11.80 27.80 0.57 0.72 25 -0.56 50 2 646 N 63 15 0.49 1.33 62 5.18 20 2.20 13.20 28.00 0.33 0.63 36 0.09 99 4 910 N 45 15 0.42 2.58 41 3.72 16 2.30 10.10 26.10 0.44 0.79 19 NA 83 2 1038 N 33 6 0.60 2.16 54 5.53 18 2.50 11.70 26.90 0.55 0.86 22 -1.52 83 4 1058 N 38 10 0.50 2.00 48 3.18 20 3.00 9.90 33.20 0.25 0.82 11 NA 83 2

119

Site Name Restored MDG ODG Forest LDI ORAM IMBCI BSR mC FQI I wFQI mCR VSR WQI AIBI ASR 1170 N 72 23 0.07 1.42 76.5 3.83 18 2.70 13.10 29.20 0.36 0.83 24 0.09 83 4 1521 N 66 22 0.29 1.00 73.5 10.10 20 3.10 10.70 35.60 0.66 0.75 12 0.54 83 4 1618 N 60 17 0.28 1.17 67.5 6.16 16 2.60 11.40 28.50 0.39 0.80 20 1.01 83 4 1624 N 51 9 0.37 1.36 69.5 9.03 20 2.90 12.60 34.30 0.48 0.72 19 1.40 100 6 1699 N 66 22 0.07 1.49 76.5 4.32 17 3.10 14.70 33.70 0.70 0.86 22 0.03 99 4 1716 N 7 8 0.46 3.11 28 2.86 11 2.40 10.70 29.80 0.47 0.65 20 NA 58 1 D1 R 49 22 0.51 1.01 77 8.25 24 2.70 15.20 29.80 0.41 0.81 32 -0.84 83 5 D2 N 27 10 0.56 2.05 48 4.25 14 2.40 10.60 28.20 0.26 0.74 19 1.82 33 0 D3 N 15 8 0.42 1.87 39 8.88 17 1.80 6.00 20.50 0.10 0.71 11 -0.01 66 2 D4 N 34 8 0.39 2.80 64 4.94 8 2.60 10.80 29.00 0.36 0.78 18 1.00 66 2 D5 N 63 17 0.13 1.56 55 5.90 20 2.90 10.10 33.70 0.46 0.75 12 0.80 91 3 NLDC1 N 30 10 0.51 1.00 51.5 4.89 18 2.30 11.10 26.40 0.48 0.71 23 1.80 100 5 NLDC4 R 25 8 0.51 1.24 31 3.96 13 3.20 13.10 35.00 0.61 0.83 17 NA 100 6 NLDC5 N 63 17 0.40 1.41 65.5 6.39 22 1.90 7.80 24.20 0.29 0.60 16 -1.17 100 6

120

APPENDIX D

PCA results for the composite Level One and Level Three wetland bioassessment methods

121

Level One composite PCA:

Table A4.1: Eigenvalues for the composite Principal Component Analysis of Level One metrics.

PC1 PC2 PC3 PC4 Standard Deviation 1.526 1.004 0.623 0.524 Proportion of Variance 0.582 0.251 0.097 0.069 Cumulative Proportion 0.582 0.834 0.931 1.000

Table A4.2: Importance of components and eigenvector loadings for the composite Principal Component Analysis of Level One metrics.

PC1 PC2 PC3 PC4

MDG 0.460 0.606 -0.443 -0.475

Forest 0.359 -0.767 -0.508 -0.157

ODG 0.588 0.143 -0.047 0.795

LDI 0.561 -0.156 0.737 -0.344

-10 -5 0 5

Forest 63 0.2 15 70 120 6556 66 39 2769 Minnesota 5 41 43 37 50 47 0.1 10 57 6 45 4 25 18 67 3638 642 32 DG 46 532348

28 24 40 0 0.0 21 33 8 61 55 34 1613 54

PC2 42 19 LDI100 49 17 125152 11 31 71 26 7 35 -0.1 29 30 5 60 3

59 -5 14 44 9

-0.2 22 68

-0.3 58 -10

-0.3 -0.2 -0.1 0.0 0.1 0.2

PC1 Figure A4.1: Principal component axes for the composite Level One analysis.

122

Table A4.3: Site scores for 71 wetlands based on the first principal component of the composite Level One Principal Component Analysis.

Site Name PC1 Score 1 0.4451729 9 1.0335986 60 -1.0459526 61 -1.1703871 139 0.8404089 176 -0.5736451 641 0.6813798 741 0.4229095 902 -0.7002896 907 0.8085231 910 -0.8690107 1038 -1.4595969 1058 -0.9818075 1170 0.9966957 1521 1.4430817 1618 0.6210849 1624 -0.421792 1699 0.6636825 1716 -3.1016921 Barkley 2.0172349 Bartlette1 -2.0102067 Beattie1 1.8737517 Beattie3 0.6962322 Beeler 2.1158859 Brasher 1.325346 Brennan -2.3825303 BridgesL -0.116712 BridgesU 0.3934545 Buckley -0.8947024 Clarkson 0.2443043 Cornett 1.8699951 CrookedCreek 1.9884657 Cutway -0.438176 D1 1.3084212 D2 -1.2556592 D3 -1.9729409 D4 -2.139175 D5 0.1352598 Emrich -0.452529

123

Site Name PC1 Score FishCreek1 1.5363369 FishCreek2 0.5634973 Gould -2.7431767 Hebb 1.6825786 HMP 1.5017946 Jackson -2.4845665 Jasikoff -2.2870595 Jewett 0.2618739 JoboN -2.9676325 Kogut -1.6622814 Kring -0.1100871 Lewis -1.8407916 LonesomeBay 2.3891983 McCormick1 1.2374762 McCormick2 2.263893 Montroy -0.2349809 NLDC1 -0.4687633 NLDC4 -1.0070373 NLDC5 0.7554305 Pagel2 -2.0083973 Peters 1.6863518 Poore 0.741188 Rausch -2.575816 Rieksts 1.7467581 Robinson 1.6302132 Scarlett -0.3084319 SevenSprings 2.408246 Slack 1.8230839 Snyder -1.1665371 Spencer 1.6227652 Ward 0.353324 Zufall -2.2765342

124

Level Three Composite PCA

Table A4.4: Eigenvalues for the composite Principal Component Analysis of Level Three metrics. Metric WQI was excluded from this analysis.

PC1 PC2 PC3 PC4 PC5 PC6 PC7 PC8 PC9 PC10 Standard Deviation 2.060 1.416 1.189 0.984 0.736 0.589 0.565 0.364 0.143 0.093 Proportion of Variance 0.425 0.201 0.141 0.097 0.054 0.035 0.032 0.013 0.002 0.001 Cumulative Proportion 0.435 0.625 0.766 0.863 0.917 0.952 0.984 0.997 0.999 1.000

Table A4.5: Importance of components and eigenvector loadings for the first five principal components of the composite Principal Component Analysis of Level Three metrics.

PC1 PC2 PC3 PC4 PC5 IMBCI 0.113 0.419 -0.180 -0.644 0.395 mC 0.414 -0.303 -0.161 -0.155 -0.171 FQI 0.397 -0.066 0.453 -0.074 0.065 I 0.378 -0.283 -0.253 -0.164 -0.111 wFQI 0.281 -0.231 -0.315 0.279 0.744 mCR 0.380 -0.258 0.117 -0.082 -0.269 AIBI 0.294 0.379 -0.153 0.463 -0.206 VSR 0.201 0.121 0.720 0.066 0.287 ASR 0.302 0.426 -0.123 0.373 -0.007 BSR 0.274 0.432 -0.060 -0.296 -0.217

-5 0 5

71 3 4 0.2 28

35 5 IMBCI19 BSRASR 34 AIBI 41 21 42 11 37 44 66 30 20 0.1 54 24 7 22 29 6 17 2 69 9 VSR38 646126 36 1627

45 23

0 0.0 49 25 50 60 40 FQIi PC2 56 48 58 43 55 68 59 33 wFQIi 5352 51 mCR -0.1 13 10 32 ImCi 57 70 15 4767 1462 63 12 -5

65 5 1 -0.2

46 -0.3

-0.3 -0.2 -0.1 0.0 0.1 0.2

PC1 Figure A4.2: Principal component axes for the composite Level Three analysis.

125

Table A4.6: Site scores for 71 wetlands based on the first principal component of the composite Level Three Principal Component Analysis.

Site Name PC1 Scores 1 -1.250008 9 2.4450638 60 -2.8451763 61 -2.9835404 139 2.2447349 176 -2.7930327 641 -1.3634356 741 -1.8361319 902 -2.1142057 907 -1.6812736 910 -2.4255285 1038 -0.9426458 1058 -1.4007326 1170 -0.9218735 1521 0.2268273 1618 -1.4116229 1624 0.3249366 1699 1.1990149 1716 -3.195987 Barkley NA Bartlette1 0.9446916 Beattie1 3.3376665 Beattie3 1.0847064 Beeler 2.4854927 Brasher 1.129728 Brennan 1.0492582 BridgesL -0.5294869 BridgesU 0.2007384 Buckley 0.6418861 Clarkson 3.7098969 Cornett 0.865018 CrookedCreek 3.4651523 Cutway 0.4016977 D1 0.4642241 D2 -3.9047168 D3 -5.2303241 D4 -2.5765684

126

Site Name PC1 Scores D5 -0.8229753 Emrich 1.9700912 FishCreek1 -0.1639253 FishCreek2 0.8391114 Gould 0.5442559 Hebb NA HMP -0.1991433 Jackson -2.0214137 Jasikoff 3.0155694 Jewett 3.025808 JoboN -5.0203366 Kogut 1.2850349 Kring 1.9102137 Lewis -0.1480578 LonesomeBay 1.2306041 McCormick1 0.8690174 McCormick2 2.0311542 Montroy 0.894075 NLDC1 -1.5483678 NLDC4 0.9482118 NLDC5 -3.1742374 Pagel2 -2.594568 Peters NA Poore 0.3700597 Rausch 0.1068126 Rieksts -0.7809033 Robinson 0.8535279 Scarlett -1.2785906 SevenSprings 1.5349401 Slack 1.9288927 Snyder 1.5494132 Spencer 3.2261785 Ward 2.0389982 Zufall 0.7661059

127