EVFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter 6ce, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 800/521-0600

STRUCTURAL AND FUNCTIONAL BOUNDARY STUDIES ON AN UPLAND-TO-WETLAND TRANSITION IN BETSCH FEN PRESERVE, OHIO : AN EVALUATION OF METHODS TO DETERMINE LANDSCAPE BOUNDARIES

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

By

Devi Nandita Choesin, M.S.

*****

The Ohio State University 1997

Dissertation Committee: Approved by

Ralph E.J. Boemer, adviser

William A. Jensen Adviser Departmèm of Plant Biology William J. Mitsch

Mohan K. Wali mil Number: 9801667

UMI Microform 9801667 Copyright 1997, by UMI Company. All rights reserved.

This microform edition is protected against unauthorized copying under Title 17, United States Code.

UMI 300 North Zeeb Road Ann Arbor, MI 48103 ABSTRACT

Landscape boundary locations are important not only because of their ecological

significance, but also because of their potential regulatory implications. This study

examined and compared different approaches to boundary determination suggested from

both traditional vegetation analysis and landscape ecological literature by applying them to

field data collected fi-om Betsch Fen, an alkaline wetland in Ohio. This basic ecological

objective was then related to practical applications in regulatory wetland delineation.

Betsch Fen is a high quality fen exhibiting typical vegetation zonation and many

plant species representative of Ohio fens. Two years’ monitoring of water chemistry suggested a significant relationship between water alkalinity levels and plant community distribution. However, spatial patterns were more complex than previously assumed, and functional boundaries of water alkalinity did not coincide entirely with structural boundaries of vegetation. Soil chemistry was not as important in determining vegetation establishment.

Three boundary detection methods were examined: (1) gradient analysis by detrended correspondence analysis (DCA), (2) the moving split-window (MSW) technique, and (3) the federal method of wetland delineation as outlined by the US Army

Corps of Engineers technical manual. DCA and MSW were used to test hypothesized boundary locations determined through field observations. DCA was more successful in

detecting vegetation changes at the community level, but it was often difficult to

extrapolate this information to a landscape level. In contrast, MSW detected changes at a

landscape level which overestimated minor shifts in species composition at the community

level. Although results fi'om MSW were more easily interpretable, neither method completely confirmed hypothesized boundary locations. While DCA and MSW should ideally be used in conjunction to provide maximum information on boundary location and ecological significance, this would be unrealistic in practical applications. A method must be selected which satisfies the specific objectives most, taking into consideration the compromise between practicality and reliability. As relates to wetland delineation, it is also important to consider issues of time, manpower, expenses and practicality. The objective should be to minimize time and effort while employing an ecologically sound quantitative method which considers the importance of vegetation, water and soil.

Ill with love, admiration, respect and gratitude to My Father and the loving memory of My Mother

IV ACKNOWLEDGMENTS

First and foremost, I owe a debt of gratitude to Ralph Boemer for his constant and

patient guidance during this project and all my years at Ohio State. As an adviser, Ralph

gave me enough freedom to pursue my own ideas, but would always be there whenever

things went wrong. His inputs of thought, time and energy into this project have been

invaluable.

I thank the many faculty members who have contributed their time and suggestions

to this project and throughout my studies at Ohio State. Drs. Jon Bart, Roger Hangarter,

William Mitsch and Fred Sack provided comments and suggestions during the proposal stage of this project. During the writing of this dissertation, I received valuable input from

Drs. Peter Curtis, William Jensen, William Mitsch, Deanna Stouder and Mohan Wali.

Funding for this project was provided by grants from The Ohio State University

Janice Beatley Herbarium Awards, the Nature Conservancy research grants program, and part of a fellowship from the the World Bank/Indonesian Second University Development

Project.

I am grateful to Jeff Knoop and Steve Sutherland from the Ohio Chapter of The

Nature Conservancy for introducing me to Betsch Fen, and allowing me to carry out my research plans. I was very fortunate to have found great student helpers who helped with my field

work for three field seasons, and who shared my interest in and in this project:

Meryl Hattenbach helped me set-up the project in 1994, while Megan Gahl was a

tremendous help during 1995 and part of 1996. In addition, Rick Shamblen also came out

a few times to the field.

To past and present students of the Boemer Lab whom I have gotten to know over the years: thank you for your fiiendship and sharing of knowledge, experiences and ideas. Thank you also to Jennifer Brinkman who was always willing to help out in the lab.

I am also grateful to the Department of Biology at Bandung Institute of

Technology in Indonesia for giving me the opportunity to pursue this higher degree, and for tolerating my very long absence.

Many other individuals have contributed their time and expertise to this project, and I would like to especially thank and mention a few of them here. Jennifer Windus gave me the initial introduction to Ohio fens, and thus helped me formulate my research questions. Jennifer and the staff* at the Division of Natural Areas and Preserves of the

Ohio Department of Natural Resources issued permits for my field work and also helped with the identification of plant species. Bob Klips not only came out to the field with me several times, but also spent many hours helping me identify plant species. Mike Feher of

Henderson Aerial Surveys in Columbus generously provided the most recent aerial photograph of Betsch Fen, taken especially for this study. 1 thank the above and all others too numerous to mention.

VI I dedicate this work with much love, admiration, respect and gratitude to my parents who always taught me the importance of the pursuit of knowledge. My mother may not be here to share this little accomplishment, but I’m sure she’s watching over me somewhere.

Finally, of course, I will forever be indebted to my husband, Donny, for all his love and understanding while sharing the vicissitudes of this endeavor. I owe him not just for his continued moral support and encouragement, but for so much more, including the long hours he sacrificed to help me out in the field and on the computer. Throughout all of this, he has always kept things in perspective by reminding me of life's priorities and the importance of faith.

vn VITA

April 6, 1961 ...... Bom - Jakarta, Indonesia.

1987 ...... B.S. (Sarjana) in Biology, Bandung Institute of Technology, Bandung, Indonesia.

1988-present ...... Faculty Member, Department of Biology, Bandung Institute of Tech­ nology, Indonesia.

1990 ...... M.S. in Plant Biology, The Ohio State University, USA.

1992-1997 ...... Graduate Teaching Associate, De­ partment of Plant Biology, The Ohio State University, USA.

PUBLICATION

Choesin, D.N. and R.E.J. Boemer. 1991. Allyl isothiocyanate release and the allelopathic potential of Brassica napus L. (Brassicaceae). American Joumal of Botany. 78(8): 1083-1090.

FIELDS OF STUDY

Major Field: Plant Biology Studies in Plant Ecology

v n i TABLE OF CONTENTS

Page

Abstract ...... ii

Dedication ...... iv

Acknowledgments ...... v

Vita ...... viii

List of Tables ...... xi

List of Figures ...... xiii

Chapters :

1. Introduction ...... 1 1.1 Background ...... 1 1.2 Research objectives ...... 3

2. Literature Review...... 5

2.1 Landscape boundaries: concepts and definitions ...... 5 2.2 Approaches to boundary determination ...... 15 2.3 Delineation of wetlands...... 32 2.4 Fen ecology ...... 37

3. Description of study area ...... 41

4. Vegetation mapping and description ...... 47

4.1 Background and objectives ...... 47 4.2 Methods ...... 48 4.3 Results...... 50 4.4 Discussion ...... 57

ix 5. Water chemistry ...... 62

5.1 Background and objectives ...... 62 5.2 Methods...... 65 5.3 Results ...... 70 5.4 Discussion ...... 84

6. Soils...... 92

6.1 Background and objectives ...... 92 6.2 Methods ...... 95 6.3 Results ...... 97 6.4 Discussion ...... Ill

7. Boundary Determination...... 115

7.1 Background and objectives ...... 115 7.2 Methods ...... 117 Method I : Gradient analysis by Detrended Correspondence Analysis (DCA) 117 Method 2: Moving split-window ( MSW) analysis ...... 120 Method 3: Federal method of wetland delineation ...... 122 7.3 Results...... 129 Method 1: Gradient analysis by Detrended Correspondence Analysis (DCA) 129 Method 2; Moving split-window (MSW) analysis ...... 144 Method 3: Federal method of wetland delineation ...... 154 7.4 Summary ...... 156 Method 1: Gradient analysis by Detrended Correspondence Analysis (DCA) 156 Method 2: Moving split-window (MSW) analysis ...... 157 Method 3; Federal method of wetland delineation ...... 157

8. General Discussion and Comparison of Methods ...... 158

9. Recommendations and Summary ...... 177 9.1 Recommendations ...... 177 9.2 Summary ...... 183

Literature Cited...... 187

APPENDICES ...... 202 Appendix A; Table of plant species recorded in Betsch F e n ...... 203 Appendix B: Map of water-sampling well locations in the field ...... 208 Appendix C: Map of the placement of transects established for vegetation analysis in the south fen ...... 210 Appendix D: Field notes on observations during wetland delineation ...... 212 X LIST OF TABLES

Table Page

2.1 Explanations and descriptions of boundary-related terms selected from the ecological literature (arranged chronologically by publication date ...... 12

2.2 Quantitative methods for studying landscape boundaries (summarized from Johnston et al. 1992) ...... 24

4.1 Estimate o f fen area (i.e. central marl, sedge meadow and shrub meadow) by image analysis of aerial photographs. Areas reported as means and standard errors from ten repeated measures ...... 56

5.1 Repeated measures analysis of variance results for water alkalinity levels in both the south and north fens ...... 75

5.2 Range of mean water pH recorded during the study. Values are means from sampling points within a community type for a particular sampling period ...... 77

5.3 Semivariogram model parameters for water chemistry based on complete data sets and selected subsets. Relative structural variance, as a measure of spatial dependence, is expressed as the proportion of total model variance (C+Co) represented by structural variance (C) ...... 81

6.1 Results of one-way analysis of variance for soil organic carbon, carbonate content and pH in the south fen ...... 103

6.2 Results of one-way analysis of variance for soil organic carbon, carbonate content and pH in the north fen ...... 107

6.3 Qualitative soil color characteristics along water-sampling transects from both south and north Betsch Fen, as

xi determined through use of the Munsell Soil Color Charts (Kolimorgen Corporation 1975) ...... 110

7.1 Braun-Blanquet scale used for field data collection, and conversion values used for data analysis in this study. Conversion values were chosen arbitrarily for ratings + and 1, and by taking the midpoint of the percentage range for ratings 2 to 5 ...... 119

7.2 Plant indicator status categories, fi’om the 1987 US Army Corps of Engineers Wetland Delineation Manual (Environmental Laboratory 1987) ...... 126

7.3 Results of boundary detection by detrended correspondence analysis (DCA) and moving split-window technique (MSW) as compared to major vegetation zone boundary lines subjectively determined through field measurements and qualitative observations ...... 135

7.4 Canonical coefiBcients and intraset correlations of water alkalinity, soil carbonate content and soil pH with the first three axes of CCA for transect SB ...... 142

7.5 Results of wetland determination along three transects characterized according to the routine onsite determination method of the US Army Corps of Engineers wetland delineation manual (Environmental Laboratory 1987) ...... 155

8.1 Summary of strengths and weaknesses of detrended correspondence analysis (DCA), moving split-window analysis (MSW) and wetland delineation (WD). A "+" sign indicates a strength or desirable feature of the method; a sign indicates a weakness or undesirable feature of the method ...... 173

A. 1 List of vascular plant species recorded in Betsch Fen (denoted by "x" sign) fi’om vegetation analyses and surveys conducted during this study (1994-1996). Observations mainly conducted in fen area south of Blackwater Creek, unless otherwise noten by N (north). Nomenclature follows Weishaupt (1971) ...... 203

Xll LIST OF FIGURES

Figure Page

2.1 Concept of boundary adapted from Forman (1995) ...... 6

2.2 Characteristics of boundaries along environmental gradients: (A) abrupt, high-magnitude change, (B) abrupt, low-magnitude change, (C) gradual change where boundaries are determined arbitrarily (reproduced from Johnston et al. 1992, based on Burrough 1986) ...... 9

2.3 Approaches to gradient analysis in plant ecology (adapted from Kent and Coker 1992, and ter Braak1995) ...... 19

2.4 Use of the moving-window approach to analyse two- dimensional data. A summary value is assigned to the central pixel based on values for each of the nine pixels in the scan window. Summary value can be derived using various statistical metrics, e.g. mean, mode, median etc. The window is then moved until the entire area has been covered (reproduced from Johnston et al. 1992) ...... 30

3.1 Location of Betsch Fen in Ohio (arrow). Hatched area indicates the Glaciated Allegeny Plateau physiographic and phytogeographic region (from McCance and Bums 1984 and Andreas 1989) ...... 42

4.1 Vegetation map showing the five major plant communities in Betsch Fen...... 54

5.1 Principle of ground water sampling well (adapted from Kadlec 1989) ...... 67

5.2 (A-G): Temporal fluctuations in water alkalinity (closed circles) and conductivity (open circles) in the south fen

xiii according to plant community. Vertical bars represent standard errors of the mean. In G, data points are missing for observation periods in which well was empty, or insufiBcient sample volume was collected (see text for explanation) ...... 71

5.3 (A-C); Temporal fluctuations in water alkalinity (closed circles) and conductivity (open circles) in the north fen according to plant community. Vertical bars represent standard errors of the mean ...... 73

5.4 Short-term monitoring of water level fluctuations in a few representative wells. Asterisks (*•) indicate maximum depth of monitoring well; actual water level is below this depth ...... 78

5.5 Mean water alkalinity (closed circles) and conductivity (open circles) of water samples collected along a transect in the south fen which traverses the major plant communities ...... 79

5.6 Annual spatial patterns in water alkalinity levels (in mg CaCOs/L) in a 190 X 105 m^ area in the south fen based on (A) 1994 data only; (B) 1995 data only; and (C) 1994 and 1995 data combined. Maps generated by kriging using the best-fit semivariogram model 82

5.7 Seasonal spatial patterns in water alkalinity levels (in mg CaCOs/L) in a 190 X 105 m^ area in the south fen based on (A) spring data only; (B) summer data only; and (C) autumn data only. Maps generated by kriging using the best-fit semivariogram model ...... 83

5.8 Spatial patterns in water conductivity (in pS/cm) in a 190 x 105 m^ area in the south fen based on complete data set collected during the study. Map generated by kriging using the best-fit semivariogram model.. 85

5.9 Spatial patterns in water pH in a 190 x 105 m^ area in the south fen based on complete data set collected during the study. Map generated by kriging using the best-fit semivariogram model ...... 85 6.1 Mean readings for (A) soil organic carbon, (B) soil carbonate content, and (C) soil pH of samples collected along a water-sampling transect in the south fen which traverses the major plant communities. Vertical bars represent standard errors of the mean ...... 98

6.2 Readings for (A) soil organic carbon, (B) soil carbonate content, and (C) soil pH fi’om samples collected at every two meters

xiv along a vegetation transect in the south fen ...... 100

6.3 Mean readings for (A) soil organic carbon, (B) soil carbonate content, and (C) soil pH of samples collected along a water-sampling transect in the north fen. Vertical bars represent standard errors of the mean ...... 105

6.4 General boundaries of major soil types in Betsch Fen based on draft maps of the unpublished soil survey of Ross County, and field observations during the present study. Aa=Adrian muck, Cd=Carlisle muck, KaB=Kendallville silt loam 2-6% slopes, KeC2=Kendallville-Eldean complex 6-12% slopes, RdE2=Rodman gravelly loam, 20-35% slopes ...... 108

7.1 The moving split-window technique for analysis of one-dimensional data, using a window width of eight sample points. The dissimilarity between window halves is calculated based on mean sample values in each half. The window is then moved sequentially by one plot along the transect, and calculation of dissimilarity is repeated until the entire transect is covered (fi-om Johnston et al. 1992) ...... 121

7.2 Placement of transects (A, B and C) to conduct wetland delineation ...... 124

7.3 Example of Data Form 1 from the 1987 US Army Corps of Engineers Manual, completed during field observations for wetland determination ...... 125

7.4 (A-F): DCA plot ordinations of vegetation data (percent cover by species) from six belt transects of contiguous 1.0 X 0.5 m^ plots in the south fen. Numbers correspond to sequential plot numbers along the transect, beginning from plot 1 at the riparian woods boundary zone, and increasing as transect moves towards upland woods boundary zone ...... 130

7.5 (A and B): DCA plot ordinations of vegetation data (percent cover by species) from two belt transects of contiguous 1.0 X 0.5 m^ plots in the north fen. Numbers correspond to sequential plot numbers along the transect, beginning from plot I at the riparian woods boundary

XV zone, and increasing as transect moves towards upland woods boundary zone ...... 141

7.6 CCA plot ordinations of vegetation data from transect SB, with environmental variables represented by arrows. Soil carbonate effects were not significant in the first two ordination axes ...... 143

7.7 (A-F): Peaks of squared euclidean distance (SED) determined by moving split-window analysis (window width=8) of percent vegetation cover data from six belt transects of contiguous 1.0 x 0.5 m^ plots in the south fen. Transects begin from plot 1 at the riparian woods boundary zone and moves towards upland woods boundary zone. Number above peaks indicate points which satisfy the minimum SED value used to determine a significant peak or boundary line location ...... 145

7.8 (A and B): Peaks of squared euclidean distance (SED) determined by moving split-window analysis (window width=8) of percent vegetation cover data from two belt transects of contiguous 1.0 x 0.5 m^ plots in the north fen. Transects begin from plot 1 at the riparian woods boundary zone and moves towards upland woods boundary zone. Number above peaks indicate points which satisfy the minimum SED value used to determine a significant peak or boundary line location ...... 149

8.1 DCA plot ordinations of vegetation data from (A) a subset of 40 plots from transect SB, and (B) a subset of 20 plots from transect SB. Inset = DCA plot ordination basedon all plots in transect S B ...... 160

8.2 Comparison of results from moving split-window analysis on transect SB using four different window widths...... 163

8.3 Comparisons of boundary line positions detected by detrended correspondence analysis (DCA) and moving split-window analysis (MSW) in six transects in the south fen ...... 166

XVI 8.4 Comparisons of boundary line positions detected by detrended correspondence analysis (DCA) and moving split-window analysis (MSW) in two transects in the north fen ...... 167

8.5 Direct substitutions of MSW results into DCA for (A) transect SB, and (B) transect NB. Different symbols indicate different plant zones as separated by M SW ...... 169

B. I Map of water-sampling well locations (denoted by "w") established in the field. Map shows mainly the south fen area ...... 208

C. 1 Map of the placement of transects (SA to SF) established for vegetation analysis in the south fen ...... 210

xvii CHAPTER 1

INTRODUCTION

1.1 Background

There has been a resurgence of interest in recent years in the ecological importance of landscape boundaries. These transition zones between ecosystems are now recognized as being important in controlling the flow of energy, material, and organisms between ecosystems. The understanding of ecological processes within these boundaries has thus become essential to the understanding of ecosystem functioning and for effective management.

The concept of boundary, and the related terms and edge have actually been in the literature for some time, particularly as they are related to wildlife habitat, tree lines, and transitions between major biomes (Risser 1995a). However, recent developments in the field of landscape ecology, paralleled by advances in technological tools, have allowed researchers to examine boundaries from a different perspective.

Transitional areas are no longer viewed as static components of vegetation, but rather as dynamic elements of the landscape that are important to studies of global change, conservation of biodiversity, and sustainable development (Holland and Risser 1991, di Castri and Hansen 1992, Risser 1993, 1995a and b). The present concept of boundary

clearly recognizes its functional role in the landscape (Risser 1993, Forman 1995).

Landscape boundaries are identifiable and meaningful only relative to specific

questions and specific points of reference (Gosz 1991). Recognizing their importance in

the management and restoration of changing environments (Holland and Risser 1991)

raises two important issues; boundary detection and boundary management. Obviously if boundaries are to be considered important, then we should be able to first establish where they are located.

The detection of boundaries requires the ability to determine spatial change, with or without a temporal component (Johnston et al. 1992). As this change may be either abrupt or gradual, the determination of a distinct boundary line may in some cases be dfficult. Despite this, however, strict delimitation of a natural system with diffuse boundaries is often necessary for regulatory and management purposes.

The importance of landscape boundary determination in management applications is clearly illustrated in the currently controversial topic of wetland delineation. In this time of concern over wetlands, there is a compelling need to identify and delineate wetlands, i.e. to state where a wetland’s structure and function end, regardless of whether its boundaries are abrupt or gradual. However, accurate wetland identification and delineation is a very difficult process that involves knowledge of wetland and upland plants, soils and hydrology (Environmental Laboratory 1987, Lyon 1993, NRC 1995).

The difficulty in defining wetlands has come about because less science and more politics have often been used in the definition (Mitsch and Gosselink 1993). The debate over an appropriate federal regulatory definition of wetlands has become so politicized that there is a potential risk of adopting delineation criteria which may be scientifically unsound, and which may consequently reduce protection for remaining wetland resources

(Kusler 1992). Science may not be able to resolve legal problems concerning wetlands on private lands, but it can support the development of objective and consistent means for identifying wetlands and their boundaries (NRC 1995).

The present study examines the concept of landscape boundary determination by reviewing the state of the science, and by comparing different approaches for boundary determination suggested from both traditional vegetation analysis and landscape ecological literature. This basic ecological objective will then be related to practical applications in regulatory wetland delineation.

To meet the general objectives of this research, a specific wetland site, i.e., Betsch

Fen Preserve in Ohio, has been studied as a model system. This site, which is currently under conservation management, represents a specific type of natural wetland known to have unique and extreme ecological conditions.

1.2 Research objectives

The general objective of this research project is to compare and evaluate different approaches to landscape boundary determination by applying them to field data collected from a wetland ecosystem. Field data were obtained by quantitative sampling of vegetation, water and soil. The study will then assess how ecological principles and methods can contribute to regulatory policies in wetland identification. Specifically, this study will; (I) describe, characterize and map both the internal and external structural boundaries of the wetland, (2) quantify the nature, shape, spatial position, and temporal variability of the functional boundaries of this wetland, as defined by groundwater chemistry, and (3) relate structural and functional boundary analyses, with an application to wetland delineation in mind.

In addition to the above mentioned objectives, this study will also provide baseline ecological data for the particular study area. As ecological studies on Betsch Fen are lacking, this study should provide preliminary information which may be useful for the purposes of management. In this case, the study will: (I) describe, quantify and map vegetation patterns and boundaries in Betsch Fen, (2) provide baseline data on temporal fluctuations and spatial distributions in water chemistry, (3) provide baseline data on soil characteristics, and (4) relate vegetation patterns to environmental conditions.

In this dissertation, results from studies on vegetation, water chemistry and soil will first be examined separately in chapters 4, 5 and 6, respectively, before being analysed within the broader context of boundary determination in chapter 7. Finally, chapters 8 and 9 will discuss the comparison of methods and suggest recommendations. CHAPTER 2

LITERATURE REVIEW

2.1 Landscape boundaries: concepts and definitions

The terms boundary, ecotone and edge have all been used in describing the spatial transition between adjacent landscape elements. Although often used inter­ changeably, however, these terms have been defined as distinct concepts. In one of the most recent treatments of the subject, Forman (1995) defined a boundary (or boundary zone) as a zone composed of the edges o f adjacent ecosystems. He thus distinguishes the term from edge, which is defined as the portion of an ecosystem near its perimeter, where influences of the surroundings prevent development o f interior environmental conditions. Forman then referred to the line separating the edges of landscape elements as the border (Figure 2.1).

According to those definitions, an ecotone is found in the case where species distributions within the boundary zone change progressively or evenly, analogous to a compressed gradient (Forman 1995). In most of the literature, however, distinctions between the usage of the terms boundary and ecotone have not been clear. For example. border

J I______I I______I L interior 1 j edge ______1 ed g e 2 [ interior 2

boundary

Figure 2.1 : Concept of boundary adapted from Forman (1995). Gosz (1991) saw no compelling arguments to differentiate between the terms ecotone, landscape boundary and transition zone, and thus used them interchangeably

Research in recent years has revealed the importance of as dynamic components of the landscape. They are now recognized as being important in relation to biological diversity, productivity, material flow, wildlife habitat, and as indicators of global change (Risser 1995a). This resurgence of interest has prompted the publication of a number of comprehensive volumes on ecotones, including those edited by Naiman and Decamps (1990), Holland et al. (1991), Hansen and di Castri (1992) and Risser

(1995b). Use of the term ecotone in these volumes has generally followed the definition suggested by the Scientific Committee on Problems of the Environment (SCOPE)

(Holland 1988 cited in Holland et al. 1991), i.e., a zone of transition between adjacent ecological systems, having a set o f characteristics uniquely defined by space and time scales and by the strength of the interactions between adjacent ecological systems. In this case, ecological system is analogous to landscape patch (Holland et al. 1991), i.e., a non-linear surface area differing in appearance from its surroundings (Forman and

Godron 1986, Forman 1995).

Within the broad definition above, an ecotone is a system that can be envisaged at any hierarchical level (i.e., fi-om population to the biosphere), and at any spatial scale

(i.e., from a few centimeters to thousands of kilometers), according to the particular research perspective and working hypothesis (di Castri and Hansen 1992). An ecotone thus might occur between ecological systems at a broad spatial scale (e.g., boundaries between continental biomes) or at a fine scale, such as the transition between two specific biological communities (Risser 1990). Historically, the ecotone concept was originally used for large-scale boundaries and transition zones, but later became largely studied on the community level (van der Maarel 1990).

As the ecotone concept is used to emphasize a relatively sharp change in species distributions, it is often contrasted with the gradient concept. From a landscape perspective, an area may be spatially heterogeneous by exhibiting either a gradient or a mosaic. Gradients, i.e., gradual changes as a function of distance, produce a heterogeneous system that is not patchy. On the other hand, a landscape mosaic contains patches with abrupt discontinuities or boundaries. Since boundaries are not found in gradients, boundary and gradient become mutually exclusive concepts (Forman 1995).

In natural systems, ecotones are most commonly perceived as involving a change in vegetation (Johnston et al. 1992). Therefore, relevant to the issue of spatial heterogeneity is the divided opinion over the nature of plant communities, i.e., between the individualistic theory and the community unit theory. In the former perspective, the community is perceived as simply the assemblage of species that happen to coexist at some point in time and at some point along a continuum of gradual change over distance.

The latter theory recognizes the existence of primary community types separated by ecotones (see Forman and Godron 1986, Glavac et al. 1992).

The distinctness of a boundary depends on the abruptness of change over a given distance, and the magnitude of this change (Johnston et al. 1992) (Figure 2.2).

Terminology is thus further complicated by the distinction between the terms ecotone and ecocline (van der Maarel 1990). Based on agents and patterns of change, spatial

8 A

Value of attribute

B

Environmental gradient

Figure 2.2: Characteristics of boundaries along environmental gradients: (A) abrupt, high-magnitude change, (B) abrupt, low-magnitude change, (C) gradual change where boundaries are determined arbitrarily (reproduced from Johnston et al. 1992, based on Burrough 1986). patterns may be abrupt or gradual. An abrupt change implies nonlinear behavior, and if

present, this nonlinearity defines the boundary as an ecotone in the strict sense. If the

change is gradual, and the fiinctional and structural trends of the system under study

approach a certain linearity, it is preferable to refer to ecoclines or to true gradients (di

Castri and Hansen 1992). The original concept of ecotone in a strict sense is an

environmentally stochastic stress zone, while an ecocline is a gradient zone which is

relatively heterogeneous but environmentally more stable (van der Maarel 1990).

Gradients in environmental factors have been important in the study of vegetation

patterns. However, landscapes are generally viewed as a mosaic of patches because

boundary distinctness is scale dependent (Johnston et al. 1992), and resultant vegetation

gradients are generally undetected when viewed at a landscape level (Causton 1988). It

is important to note, however, that although it may be informative to view ecosystems as

a collection of resource patches separated by boundaries, the approach is not proposed

as a replacement for continuum or gradient concepts. It should rather be viewed as

complementary to them, providing a perspective that operates over different spatial and

temporal scales (Naiman and Decamps 1990).

Gosz (1991) has also cautioned against putting too much emphasis on the

ecotone concept because the recent focus on the ecology of boundaries may represent the danger of going fi'om one extreme to another. While concentrating on homogeneous regions can misrepresent landscapes, concentrating on boundaries will also present a problem. Gradients are ever present, and insufScient information will be obtained when contrasting only between the flattest and steepest measures (i.e., between patch interior

10 and edge); relationships must be developed for changes in structural and functional

characteristics across all degrees of gradient change (Gosz 1991).

Other explanations for boundary-related concepts have been suggested in the

literature (Table 2.1). A review of the diversity of interpretations on the concept of

boundary can be summarized as follows:

1. As there are rarely sharp lines of demarcation in natural systems, most authors

perceive boundaries as a zone or area of transition. Therefore, usage of the term is

not analogous to the general notion of man-made boundaries which is represented by

one-dimensional lines. In this respect, natural boundaries have been compared to

membranes in organismal or physical systems (Wiens et al. 1985, Forman and Moore

1992), i.e., layers exhibiting thickness and specific function.

2. The concept of boundary or ecotone is based on the landscape concept of patches,

not on gradient change.

3. More recent formulations of boundaries have recognized their dynamic nature, and

thus include dimensions of both space and time. In addition, current concepts

strongly emphasize the strong functional connotation in the concept (e.g., Wiens et

al. 1985, Forman and Moore 1992, Forman 1995).

The above discussion illustrates how the study of boundaries is complicated by differences in how the concept is interpreted. In particular, studies which attempt to define boundary locations must take into consideration the following points:

II Term Explanation/description Author

ecotone a tension zone where principal species from adjacent Clements 1905 communities meet their limits. (cited by Gosz 1991)

edge a zone of transition particularly rich in the number of species. Leopold 1933

ecotone the place where two major commimities meet and blend together. Smith 1966

limes convergeas boundary type characterized by spatial concentration van Leeuwen 1966 and instability in time. and Westhoff 1971 limes divergeas boundary type characterized by spatial dispersion and stability in time.

ecotone zone of intergradation as two communities come into contact Daubenmire 1968

ecotone a transition between two or more diverse communities; Odum 1971 a junction zone or tension belt which may have considerable linear extent but is narrower than the adjoining coirununity areas themselves.

boundary location where the rates or magnitudes of ecological transfers Wiens et al. 1985 (e.g. energy flow, nitrogen exchange) change abruptly in relation to those within the patches. ecotone a relatively narrow overlap zone between two communities. Forman and Godron 1986 edge a discontinuity, where one type of vegetation ends and Orloci and Orloci 1990 another begins; sharp transition; narrow ecotone. ecotone a habitat created by the juxtaposition of distinctly Ricklefs 1990 different habitats; an edge habitat; a zone of transition between habitat types. ecotone a narrow and fairly sharply defined transition zone between Allaby 1994 two or more different commimities. ecotone a set of spatially contiguous locations, creating long and Fortin 1994 relatively narrow boundaries, where the majority of the variables show the highest rate of change.

Table 2.1: Explanations and descriptions of boundary-related terms selected from the ecological literature (arranged chronologically by publication date).

12 1. It is important to define the concept according to the intended usage or research

objective. This includes clarification of whether the definition will be used in a strict

or general sense.

2. It is important for comparison purposes to recognize contradictions between existing

interpretations in the literature. For example, while most perceive a boundary as a

zone of transition, McCoy et al. (1986) contend that "....a boundary is an absolute

barrier, not a transition zone”. They are thus perhaps using the term in the context

of border according to Forman (1995). To avoid confusion, perhaps the terms

boundary zone and boundary line can be used when clarification is necessary, as in

studies of edge detection which are aimed at locating distinct lines.

In this dissertation, the term boundary zone will be used to clarify usage of boundary according to Forman (1995) above. The terms boundary line {border sensu

Forman 1995) will be used to denote the imaginary line between adjacent communities or systems.

Definition of terms becomes particularly important when discussing wetland systems and their delineation from a regulatory perspective. In this case, delineating a wetland implies locating a boundary line between the wetland patch and the surrounding upland area, based on characteristics of the ecological boundary zone. The aim of wetland delineation is thus to define the limits of where the wetland function ends, regardless of whether changes are gradual or abrupt. Whether or not this limit corresponds to the system border (sensu Forman 1995) is still open to question.

13 In order to understand the structure and function of boundaries, three boundary

dimensions need to be considered: (1) the width between the boundary line (border) and the interior of the patch, (2) the vertical dimension of height and stratification, and (3) the form or length of the boundary which includes its overall curvilinearity (Forman and

Moore 1992, Forman 1995). When considering some wetlands, the temporal changes in boundary width as a result of high temporal variability in water level presents a complicating factor in boundary determination. While the structural boundaries of the wetland (as indicated by vegetation, for example) may not exhibit significant shifts within a short period of time, its functional boundary locations (as determined by hydrology or soil saturation) may change. Thus the question of boundary dynamics becomes particularly relevant in wetland systems.

Boundary width, verticality and length describe the structural attributes which control or determine the functional roles of boundaries. These functions can be classified into five categories, i.e. as (1) conduit, (2) filter or barrier, (3) source, (4) sink, and (5) habitat (Forman and Moore 1992, Forman 1995). While boundaries are identified as having specific functional roles, the understanding of boundary function in wetlands should not be confused with the specific concept of wetland function which considers ecological processes or functions within the wetland system. This term is often used in conjunction with wetland value, which may include an estimate of the economic worth of a wetland based on its ecological function (Brooks 1989, Richardson 1994).

The concept of wetland boundary should also not be confused with the idea of the wetland as an ecotone. Because wetland systems are transitional zones between

14 uplands and aquatic systems, they are often considered to be ecotones in and of

themselves (Mitsch and Gosselink 1993). Indeed a boundary may exhibit distinctive

characteristics which differ from the adjacent ecosystems, and thus be considered a

unique system (Forman 1995).

2.2 Approaches to boundary determination

Studies on ecological boundaries have used methods which can be broadly

classified into two approaches: (1) traditional vegetation analysis, and (2) landscape

ecological analysis. The former approach is based on the perceived notion that natural

boundaries involve a structural change in vegetation. The latter approach includes

methods which are not restricted to the detection of only vegetation boundaries, and are

thus more widely applicable. I have distinguished these two approaches because although

the field of landscape ecology acknowledges its historical roots, it has developed into a discipline which differs significantly from conventional vegetation analysis (Turner et al.

1991).

Two basic objectives of vegetation analysis are: (1) description and mapping, and (2) ecological analysis, i.e., investigation of the relationships between species and the environment (Causton 1988). In many quantitative studies of vegetation, boundaries have been ignored or omitted because sampling strategies were designed to identify relatively uniform areas and to measure characteristics of the representative species composition (Risser 1993). Consequently, some authors have questioned the effectiveness of vegetation analysis techniques, e.g., multivariate ordination, to identify

15 boundaries. Although it can be argued that classification methods which identify

ecologically homogeneous areas indirectly identify boundaries as the limits of the

homogeneous area, these techniques focus on contents rather than on boundaries and

have not been very successful when applied to boundaries (Johnston et al. 1992).

Despite the aforementioned problem, the importance of edges in vegetation has

actually been recognized in vegetation analysis literature (e.g., Orloci and Orloci 1990,

Glavac et al. 1992). Furthermore, the use of vegetation analysis techniques has

nevertheless added insight into the nature of boundaries. For example, ordination

techniques have been used to study ecotone dynamics and determine boundaries in the

Great Dismal Swamp (Carter et al. 1994), and in forests of the Rocky Mountains (Gosz

1992, Stohlgren and Bachand 1997).

While vegetation analysis and landscape ecology may have developed into

separate fields, certain analytical procedures related to boundaries share common theoretical bases, or require similar data collection methods. Therefore, in some respects it may be difficult to clearly differentiate between the two approaches. In the following discussion, however, I will attempt to review these two approaches separately.

Vegetation analysis

Patterns in vegetation are created through the delimitation of communities by edges; ecological theory suggests that a vegetation edge should be expected where an environmental edge is detected (Orloci and Orloci 1990). A higher degree of homogeneity in species composition is found within a community than between

16 communities, and this difference in level of homogeneity should be sufficient to allow the delineation of communities (Goodall 1954, cited in Greig-Smith 1964).

Vegetation ecology is then concerned with the identification of plant communities and how they are related to each other and to the environment (Mueller-

Dombois and Ellenberg 1974). In order to accomplish this, quantitative methods have been developed to sample, describe and analyse vegetation (e.g., as reviewed by Greig-

Smith 1964 and 1983, Mueller-Dombois and Ellenberg 1974, Kershaw and Looney

1985, Causton 1988, Kent and Coker 1992, Jongman et al. 1995).

Data collected in both community and landscape ecology are mostly multivariate, i.e. each statistical unit of sampling contains many attributes (Jongman 1995). For example, data collected on vegetation typically include information on presence, frequency and of a suite of many species. The complexity, interrelatedness and bulkiness of this type of data make interpretations extremely difficult (Gauch 1982), and call for special multivariate methods (Jongman 1995).

Multivariate methods to analyse vegetation can be classified into three groups:

(1) direct gradient analysis, or regression analysis, (2) indirect gradient analysis, or ordination, and (3) classification, or cluster analysis (Kent and Coker 1992, Jongman

1995). Despite this classification, however, the term ordination (i.e., “to set in order” ) is often used interchangeably with gradient analysis (groups 1 and 2 above) as a collective term for multivariate techniques that aim to arrange vegetation samples based on their similarity in species composition and/or their asssociated environmental controls

(Causton 1988, ter Braak 1995). These descriptive techniques reduce and present data to

17 allow the formulation of ideas about plant community structure and their possible causal

relationships with the environment.

Direct gradient analysis uses environmental data as a basis to order the vegetation

samples, and thus directly displays the variation of vegetation in relation to

environmental factors. In contrast, indirect gradient analysis examines variation within

vegetation independent of environmental data. Thus any environmental interpretation is

indirect: possible abiotic gradients can be inferred by comparing environmental data to

the summarised vegetation data (Whittaker 1967, Kent and Coker 1992) (Figure 2.3).

The use of indirect gradient analysis provides certain advantages over direct

gradient analysis. While species composition is clearly distiguishable and easy to sample,

environmental data are more difiBcult to characterize exhaustively. Furthermore, it may

be difficult to determine beforehand which environmental variable most strongly affects

. Therefore, the resultant species composition may indirectly provide

more information about the environment than any given set of measured environmental variables (ter Braak 1995). In addition, relationships between species and the environment can be better understood by examining general patterns of how combinations of species cooccur, as the occurrence of any particular species may be too unpredictable to directly relate to environmental conditions (ter Braak 1995).

Results of gradient analysis or ordination are presented by plotting species or sample points against two or three axes which correspond to a dimension in space. The gradient of greatest variation will be depicted by the ‘first’ axis, the second most important gradient will be depicted by the ‘second’ axis and so on (Causton 1988). A

18 DATA SAMPLING Vegetation data Environmental & MEASUREMENT : (e.g. species presence data 1 & abundance) DATA ANALYSIS : INDIRECT DIRECT GRADIENT Synthesis of GRADIENT ANALYSIS ANALYSIS environmental factors (vegetation ordination) (environmental ordination)

ORDINATION GRAPHS

INTERPRETATION : Vegetation pattern w/o Vegetation pattern Environmental factors environmental data related to environ, data w/o vegetation data \ HYPOTHESIS GENERATION : I FURTHER RESEARCH

Figure 2.3: Approaches to gradient analysis in plant ecology (adapted from Kent and Coker 1992, and ter Braak 1995).

19 graphical summary of the data will show that points representing sites that are similar in

species composition will be grouped closely to each other, while points that are

dissimilar in species composition will appear farther apart (Figure 2.3). The success of

ordination methods in generating graphical summaries is very variable, and to some

extent depends on the structure of the data set (Causton 1988).

Various techniques of indirect ordination have evolved since polar ordination

(PO), the earliest method, was developed by Bray and Curtis (1957). The most popular

techniques have been principal components analysis (PCA) (Orloci 1966, Gittins 1969),

reciprocal averaging (RA) or correspondence analysis (CA) (Hill 1973), and detrended

correspondence analysis (DCA) (Hill 1979, Hill and Gauch 1980). Other methods which

have not been extensively used include non-metric multidimensional scaling (Fasham

1977, Prentice 1977), a potentially powerful method for data that are nonnormal or are on arbitrary, discontinuous, or questionable scales (McCune and Meflford 1995).

Although techniques differ in their analytical approach, they all produce graphical summaries in the form of ordination diagrams as described above.

Developments in analytical techniques have largely paralleled advances in computing technologies. While PO was first calculated and drawn manually using compass construction, subsequent techniques required more sophisticated computer analysis. PC A has been widely used since 1966, but is now not recommended due to certain distortion effects in its outcome (e.g., Kent and Ballard 1988). RA/CA was used extensively between 1973 and 1985, but has now been improved and replaced by DCA

(Kent and Ballard 1988, Kent and Coker 1992). Finally, canonical correspondence

2 0 analysis (CCA) (ter Braak 1986, 1988) is another technique which is becoming widely used. This technique is not a strictly indirect ordination method because its axes are constrained by multiple regression with environmental factors (Kent and Coker 1992).

The history of ordination analysis, as well as the strengths and weaknesses of each technique, are discussed by Causton (1988), Kent and Coker (1992), and ter Braak

(1995).

Besides gradient analysis, classification or cluster analysis is another method that summarizes species data by (1) providing information on the concurrence of species, (2) establishing community types for descriptive syntaxonomy and mapping purposes, and

(3) detecting relations between communities and the environment (van Tongeren 1995).

Among the several types of cluster analyses, a major distinction can be made between agglomerative and divisive methods. The former type of analysis agglomerates single objects (i.e., samples) into larger clusters. In the latter approach, all samples are first divided into two smaller groups, and this division is repeated subsequently for all previously formed groups, until some kind of ‘stopping rule’ is satisfied. Another way of distinguishing different methods is by whether or not they are based on hierarchical concepts. Hierarchical methods are based on the assumption that a certain difference is more important than another, and therefore should be expressed at a higher hierarchical level. This treatment of data is not imposed in non-hierarchical methods which are usually preferred for data reduction (van Tongeren 1995).

Results of cluster analysis can be presented in various ways; most commonly by a species-by-sites table, or by a dendrogram which expresses the hierarchical structure of

21 the site groups. However, because neither of these can present more than one dimension,

it is useful to present results in an ordination diagram of the same set of sites so that

more complex relations with the environment can be elucidated (van Tongeren 1995).

Cluster analysis attempts to group sites in such a way that community composition of sites varies least within groups, and most between groups. It is thus particularly useful for defining vegetation mapping units based on floristic data (Jongman

1995).

Methodology in vegetation analysis continues to advance; many studies have developed and applied vegetation analysis techniques to measure compositional changes along gradients and examine boundaries along these gradients (e.g., Werger et al. 1983,

Wilson and Mohler 1983). While these studies mostly provide information at a community level, their significance in describing landscape level processes need to be explored. Orloci and Orloci (1990), for example, used a combination of a deviations profile, an angles profile, and an ordination map to detect vegetation patterns with focus on locating vegetation edges.

Landscape ecological analysis

Landscapes are perceived in terms of patches and their characteristics; landscape structure identifies the spatial relationships among distinctive patches, i.e., the distribution of energy, materials and species in relation to patch configurations (Forman and Godron 1986, Turner et al. 1991). In order to understand landscape patterns, it is first important to recognize the nature of spatial data and the importance of scale.

2 2 Turner et al. (1991) presented a review of methods to detect pattern and scale in landscapes. They classified different techniques according to their applicability to (1) landscapes that exhibit repeated pattern, and (2) landscapes that exhibit irregular pattern.

The first set of techniques is based on variance measures which are generally univariate, while the second set is mostly multivariate and applicable to the detection of edges.

Ecological edges result from discontinuities in species abundances. However, discontinuities may occur in subtle ways; although a discontinuity in dominant vegetation is obvious, distinct boundaries or edges are demonstrated mathematically in groups of subdominant species (Brunt and Conley 1990). Therefore, landscape theory must consider the responses of a majority of species, and techniques to detect ecological edges must accomodate the multivariate character of ecological data (Brunt and Conley 1990).

Different techniques may be required to quantify different boundary types. As the choice of analytical technique depends on the type and spatial distribution of the source data, Johnston et al. (1992) presented quantification methods that are applicable to various types of boundaries at a wide range of scales (Table 2.2).

Use of the moving split-window technique for one-dimensional transect data was proposed by Webster and Wong (1969) to detect soil boundaries along regularly sampled transects, although Whittaker (1960) had applied a similar approach to analyse vegetational change. This approach has received particular attention in the ecological literature as it can be applied to a wide array of both biotic and abiotic data (Ludwig and

Cornelius 1987, Wierenga et al. 1987, Brunt and Conley 1990, Gosz 1991, Turner et al.

1991, Johnston et al. 1992). Although it has mainly been suggested for gradsects

23 Data dimension Data source Data analysis

One-dimensional Field transect Moving split-window

Two-dimensional Field sampling/mapping Spatial statistics Aerial photography Image analysis Satellite remote sensing Geographic Information Systems (GIS) analysis

Table 2.2: Quantitative methods for studying landscape boundaries (summarized from Johnston et al. 1992).

(gradient oriented transects) (Ludwig and Cornelius 1987), the technique is equally applicable to non-gradient oriented transects (Brunt and Conley 1990).

The technique involves viewing the collected data through a window placed over a fixed, even number of adjacent samples on the transect. This window is divided into two halves, and a statistical metric is used to calculate the dissimilarity between attribute values in each half. The window is then moved sequentially by one sample plot along the transect until statistical comparisons are computed for the entire transect length.

Boundary locations will be detected at positions where the dissimilarity metric reaches maximum values, indicating maximum rates of attribute change. These maximum values will appear as recognizable peaks when the metric is plotted against transect length.

Various dissimilarity metrics have been used to calculate the difference between adjacent window halves when using the moving split-window technique (Brunt and

Conley 1990, Johnston et al. 1992). These include the Student’s t-test (Webster and

24 Wong 1969), coefficients of dissimilarity (Beals 1969), discriminant functions and

Mahalanobis distances (Webster 1973), Hotelling-Lawley trace F-values (Wierenga et al.

1987) and Wilk’s lambda (Nwadialo and Hole 1988). However, the squared Euclidean

distance (SED) is perhaps the most commonly used .

SED is computed as the square of the difference between means of each variable

in adjacent windows, summed across all variables measured (described later in Chapter

7). Besides being effective for various ecological data, this metric is computationally

simple and has provided results that tended to agree with field observations (Wierenga et

al. 1987, Turner et al. 1991). However, Brunt and Conley (1990) cautioned that SED

can present several mathematically undesirable properties, and examined the behavior of

this metric on data with known properties. They found that SED is effective for simple

edges and complex edges with certain characteristics, but the recognition of boundaries

becomes more difBcult as boundaries become smaller relative to background

heterogeneity.

Similar problems are encountered when using other statistical metrics. For

example, window width will also affect the results of Student t-test analyses; narrow

windows might detect too many boundaries as they tend to record short-range changes,

while wider windows will incorporate too much smoothing (Burrough 1986). The

Mahalanobis distance, defined through discriminate analysis of covariance, avoids the

assumption in SED that attributes are independent. However, problems will arise as this lack of independence is accounted for by covariance, which can only be used as an efficient descriptor for linear relationships (Turner et al. 1991). In general, problems

25 associated with the moving split-window technique can be attributed to the nature of

ecological field data; they often show strong nonlinear relationships, which in turn

produce strong nonlinear correlations among responses (Turner et al. 1991).

Furthermore, ecological data are highly variable and tend to exhibit non-Gaussian

distributions (Brunt and Conley 1990).

Although differing in detail, the basic principle of the moving split-window

technique is similar to other boundary analysis techniques described in the literature

(Ludwig and Cornelius 1987). These techniques aim specifically at locating dis­

continuities along gradsects using multivariate data, and thus differ in computational

detail from related methods which aim to: (1) search for repetitive patterns of response

in one variable along a one-dimensional transect, e.g., spectral analysis, semivariograms,

correlograms, periodgrams etc., (2) predict, smooth or interpolate the response pattern

of one variable across a two-dimensional grid, e.g., Laplacian smoothing splines, kriging,

etc., or (3) classify community data that repeat as a result of multivariate sampling along

random transects, e.g., cluster analysis, association analysis etc. (Ludwig and Cornelius

1987).

In addition to the moving split-window technique. Turner et al. (1991) cite global zonation, a technique developed in geology and hydrology by Gill (1970), as having potential application in landscape ecology. The procedure involves subdividing a transect into a series of segments; edges are identified through statistical comparisons to locate points which divide the transect into segments that are most internally homogeneous and most distinct fi'om each other. This technique is particularly applicable to landscapes with

26 discrete, homogeneous patches, but may be less useful for transects exhibiting gradual change with high variability within segments. Furthermore, this approach may be computationally burdensome when applied to long transects with many attributes

(Turner et al. 1991).

While the use of boundary detection methods on one-dimensional source data provides information on the location and width of a boundary, the shape and length of that boundary can only be quantified when two-dimensional data are available. However, field data are usually not collected on a fine two-dimensional grid, and researchers have preferred to use interpolation techniques which assume a model of continuous, rather than abrupt change (Burrough 1986). Due to assumptions inherent in many interpolation and contouring techniques, however, there is often still a need to look for discontinuities in two-dimensional spatial data (Burrough 1986).

Sources of two-dimensional data include field data collected from numerous points within the landscape, aerial photographs and remotely sensed images. In addition, maps and GIS (geographic information systems) data bases provide a secondary data source derived from the afore-mentioned sources (Johnston et al. 1992).

The differences among analytical techniques for two-dimensional data (Table 2.2) are becoming less distinct; while data sources may differ, the techniques for analyzing the data are generally interchangeable, and data from one source is often transferred to another (Johnston et al. 1992). For example, the two-dimensional moving window approach used in image analysis is a capability found in raster based GIS. Similarly, the

27 use of spatial statistics to analyse field data is actually applicable also to image analysis

and to GIS.

Spatial statistics, aimed at characterizing mapped data (Ripley 1981), differs from

traditional statistical techniques as it takes into consideration both the attribute and

location of data points. Analysis takes into account the spatial autocorrelation among

data points, i.e., the degree at which an observed value of a variable at one locality is

dependent on values of the same variable at other localities. The assumption is that any

response is not uniformly distributed in space, and that variation for a given attribute is a function of distance (Clarke 1990, Burrough 1995).

Although tests for spatial autocorrelation may indirectly identify patch boundaries by determining the extent of homogeneous areas, it is inadequate for long distance extrapolations because it does not account for regionalized variables that are too irregular to be modelled by a smooth mathematical function (Johnston et al. 1992). This problem can be remedied, however, by using kriging, a method of optimally interpolating two-dimensional data based on regionalized variable theory. Kriging expresses the spatial variation o f a particular variable as the sum of three components: (1) a structural component, associated with a constant mean value or trend, (2) a random, spatially correlated component, and (3) a random noise or residual error term (Matheron 1971, cited in Johnston et al. 1992).

Image analysis is used to identify boundaries by extracting information from images of entire landscapes obtained through remote sensing. For this purpose, the principle of the moving-window technique can be used to scan images with a two-

28 dimensional window of specified size; the value of the central pixel (picture element) of each window is assigned based on analysis of all pixel values in the window (Johnston et al. 1992) (Figure 2.4). Similarly, textural analysis using a moving window can measure landscape pattern and quantify ecotone contrast by comparing the reflectance values of picture elements (Musick and Grover 1991).

The two-dimensional moving-window technique is applicable to any digital data surface such as aerial photography scanned with a video digitizer or scanning camera

(Johnston et al. 1992). It has been used, for example, to analyse ecotones using GIS on

Landsat Thematic Mapper (TM) images (Johnston and Bonde 1989, Metzger and Muller

1996).

Various algorithms for detecting edges from aerial and remotely sensed image data are currently available (Burrough 1986). However, these algorithms cannot be directly related to sampled ecological data that would require an overly intensive and laborious sampling effort (Fortin 1994). Thus, boundary detection on two-dimensional sampled data require either modified or new algorithms that are able to handle sampling resolutions usually encountered in ecology. Fortin (1994) presented and investigated the reliability of two edge detection algorithms for regularly and irregularly sampled two- dimensional data. Based on simulated data, she found that these algorithms can be used effectively to identify boundaries where abundance data for several plant species are available.

Landscape patterns and boundaries can also be quantified using fractal models

(Milne 1991, Forman 1995); boundary lengths obtained through two-dimensional data

29 Scan Window

Central Pixel

Analysis Area

Figure 2.4: Use of the moving-window approach to analyse two- dimensional data. A summary value is assigned to the central pixel based on values for each of the nine pixels in the scan window. Summary value can be derived using various statistical metrics, e.g., mean, mode, median etc. The window is then moved until the entire area has been covered (reproduced from Johnston et al. 1992).

30 analysis can be measured as a fractal dimension, i.e., a scaling parameter used to relate a quantity of interest to a resolution or length scale. A fractal dimension of I.O denotes a straight line boundary, while a fractal dimension approaching 2.0 denotes an enormously convoluted boundary where nearly the whole area is boundary (Forman 1995).

Finally, computerized geographic information systems (GIS), which incorporate powerful and sophisticated quantitative techniques, are rapidly expanding in use. These systems can digitally store layers of spatially explicit information about a landscape, and present this information visually or graphically for efficient comparisons and correlations

(Burrough 1986, Clarke 1990, Forman 1995). A GIS can be used to quantify ecotone structure by measuring its length, area, density or fractal dimension (Johnston et al.

1992).

Quantification of boundaries through conventional data collection can now be supported by new technologies which enable remote monitoring of gaseous fluxes along transects, thus providing important data on ecosystem processes associated with boundaries (e.g., Gosz et al. 1988). While boundaries are most easily perceived in terms of changes in vegetation or other land surface features, there are subsurface discontinuities (e.g., in edaphic features) which may not coincide with surface conditions, and for which special detection methods are required. Advanced technologies are also enabling the detection of underground and underwater boundaries (in studies cited by

Johnston et al. 1992).

Boundary locations are important because of their perceived ecological significance, and for regulatory applications, as in the case of determining legal

31 boundaries to natural wetlands. The regulatory procedure for determining these boun­

daries is discussed in the next section.

2.3 Delineation of wetlands

The issue of boundary determination becomes more complicated when we

consider wetland ecosystems because of the relatively high spatial variability in their boundaries as a result of temporal changes in water saturation or inundation. As wetlands form part of a continuous gradient between open water and uplands, defining both the lower and upper limits of the system, i.e. toward wetter or drier conditions, can be problematical (Mitsch and Gosselink 1993). Boundaries between wetlands and upland areas often lie within broad transition zones, and are particularly difficult to determine where there are gentle gradients, or where microtopography causes wetlands to be interspersed with uplands (NRC 1995).

Wetland boundaries should ideally be drawn at the point where critical functions diminish rapidly as one moves fi’om the wetter to the drier parts of the ecosystem.

However, as scientific data on functional capacity are difficult to obtain, structural attributes which can be examined over shorter periods of time are often used as surrogate measures (Holland 1996). For this purpose, hydrologie indicators, soil type, and species composition have all been useful indicators of wetland functioning (Holland

1996).

Specifically, wetlands are characterized and distinguished from other ecosystems by having the following common features: (1) the presence of water at either the soil

32 surface or within the root zone, (2) unique soil conditions that differ from adjacent

uplands, and (3) the presence of vegetation adapted to wet conditions, conversely

accompanied by the absence of flood-intolerant vegetation (Mitsch and Gosselink 1993).

Currently accepted definitions of a wetland (Cowardin et al. 1979, Environmental

Laboratory 1987, NRC 1995) indicate that these three factors of hydrology, soils, and vegetation strongly affect each other.

Wetlands are the only ecosystem type in the US which is comprehensively regulated across all public and private lands by the federal government (NRC 1995).

Consequently, the question of defining wetlands has become important not only for wetland scientists, but also for a variety of groups, including regulators, developers, engineers, and property owners in general.

Prior to the 1970s, policies of the US federal government encouraged the conversion of wetlands to filled or drained lands that could be used for agriculture or other purposes. However, by the 1980s, there was increasing concern over the large losses in wetland area. While the conversion of wetlands increased productive agricultural lands in the US, and eliminated some of the socioeconomic nuisances associated with wetlands, it also reduced valuable attributes of wetlands, such as water quality maintenance and support of waterfowl (NRC 1995).

Federal regulation of wetlands began to take effect on a broad scale in the 1970s as a result of political support for the protection of wetlands. In 1990, an executive order by President Bush expanded the scope of permitting activities with the goal o f “no net loss” in wetlands (National Wetland Policy Forum 1988). In order to determine whether

33 a piece of land was a wetland, and therefore whether it was necessary to obtain a federal permit to dredge or fill that wetland, federal agencies began to develop guidelines to delimit wetland boundaries in a process that became known as wetland delineation

(Mitsch and Gosselink 1993).

A variety of groups require knowledge of methods to identify the presence of wetlands. Local and state governments need wetland assessments to manage and plan existing properties, or to evaluate properties for future acquisition and development

(Lyon 1993). Property owners also require wetland delineation manuals in order to know which parts of their land could be within the regulatory jurisdiction of one or more federal statutes. For this purpose, a delineation manual gives details about what constitutes a wetland, and what must be confirmed during delineation. It is thus not meant to define a wetland, but rather to aid the delineator in applying a definition of wetland (NRC 1995).

Technical manuals that provide guidance on wetland delineation are a relatively recent arrival in federal wetlands programs (NRC 1995). The US Army Corps of

Engineers (USACE) published a technical manual for wetland delineation in 1987. This was followed by the development of separate documents by the Environmental

Protection Agency, the Soil Conservation Service, and the US Fish and Wildlife Service.

Finally, after a period of negotiations, these four federal agencies jointly published the government’s approach to wetlands in a single Federal Manual for Identifying and

Delineating Jurisdictional Wetlands in January 1989 (Mitsch and Gosselink 1993, NRC

34 1995). This manual stipulated the three mandatory criteria required for a parcel of land to be declared a wetland, i.e., hydrology, soils, and vegetation.

Enforcement of strict wetland definitions was met by heavy lobbying from developers, agriculturalists, and industrialists who sought for a relaxing of the definitions, supposedly to lessen the regulatory burden on the private sector (Mitsch and

Gosselink 1993). In response to this, modifications to the manual were proposed, and a new manual was published for public comment in August 1991. This proposed manual, however, became heavily criticized for its unworkability and lack of scientific credibility

(e.g., see Bedford et al. 1992, Environmental Defense Fund and World Wildlife Fund

1992, Kusler 1992), and was eventually abandoned in 1992. At present, the 1987

USAGE technical manual is being used until a more acceptable version is adopted

(Mitsch and Gosselink 1993). The history and development of wetland delineation procedures are discussed in detail by Mitsch and Gosselink (1993), and the National

Research Council (1995).

Several methods have been used for wetland identification; ( 1 ) vegetation-based methods, (2) soil-based methods, (3) three-parameter methods, and (4) primary indicators method (Tiner 1996). However, major shortcomings and limitations have been found in the first two methods, and federal agencies have mainly developed the three- parameter approach which requires making observations of vegetation, soils and hydrology. For example, the 1987 USAGE manual requires finding positive indicators of all three parameters (hydrophytic vegetation, hydric soils and wetland hydrology) in

35 order for an area to be identified as wetland, although hydrology can be assumed when vegetation and soil indicators ares strong (Environmental Laboratory 1987).

Lists of positive wetland indicators are also provided in the USAGE manual. For example, the best vegetative indicators of wetlands are plants classified as “obligate hydrophytes”, i.e., plant species that grow only in wetlands. However, given that the afiBnity for wetlands varies considerably among species, plant species have been classified into four “wetland indicator categories” that reflect different expected fi'equencies of occurrence in wetlands: (1) obligate wetland (OBL), (2) facultative wetland (FACW),

(3) facultative (FAC), and (4) facultative upland (FACU). Other plants are classified as upland (UPL) (Distinctions between these categories are clarified later in Chapter 7).

Based on plant indicator lists (Environmental Laboratory 1987, Reed 1988), an area is considered wetland if (1) 50% of the dominant plants found growing on the site are those commonly found in wetlands (2) the soils in the area are considered hydric or waterlogged, and (3) the soils show demonstrable evidence of hydrological conditions associated with flooding or ponding of water (Environmental Laboratory 1987).

In addition to this approach, the primary indicators method (PRIMET) has been proposed as an outgrowth of traditional methods which attempts to use vegetation patterns, soil properties and other features unique to wetlands as diagnostic for wetland identification and delineation (Tiner 1993, 1996). It is based on the notion that in the absence of significant hydrologie modification, these unique features can be reliably used to make wetland determinations. Wetlands and their boundaries are then determined by the presence of any one of numerous primary indicators, e.g., the presence of more than

36 50% OBL plant species, or the presence of significant patches of peat moss {Sphagnum

spp.) (Tiner 1993, 1996).

Improved mapping techniques and advanced technologies such as the use of global positioning systems (Hook et al. 1995) and geographic information systems (GIS)

(Lyon and McCarthy 1995) have allowed more precise wetland boundary determinations.However, major disadvantages of GIS use are the high costs in time, expenses and manpower (NRC 1995). Furthermore, GIS data bases are only as good as the source from which they were derived (NRC 1995).

While wetland determination and delineation will remain important for wetland managers, the exact rules and procedures on how to conduct them will remain uncertain and subject to political change (Mitsch and Gosselink 1993). From an ecological perspective, it must be realized that an understanding of environmental gradients

(physical, chemical, hydrological and biological) is a prerequisite to delineating wetland boundaries (Mulamoottil et al. 1996), and that wetland functions are a product of all components of the wetland ecosystem (Holland 1996). The need for better understanding of wetland characteristics and boundaries has prompted the publication of several comprehensive volumes on the subject, including those compiled by Naiman and

Decamps (1990), the National Research Council (1995), and Mulamoottil et al. (1996).

2.4 Ohio fen ecology

The present study will examine Betsch Fen Preserve as a model system. This site represents a specific type of wetland called fens, which are generally defined as peat-

37 accumulating wetlands that receive some drainage from surrounding mineral soils and usually support marshlike (i.e., herbaceous) vegetation (Mitsch and Gosselink 1993).

Fens are thus a particular type of peatland, a generic term used to indicate any wetland that accumulates partially decayed plant matter. They are distinguished from bogs, which are acidic, peat-accumulating wetlands that do not have significant water inflows or outflows, except for precipitation.

Terminology used to classify wetland types in general has been confusing and contradictory (Mitsch and Gosselink 1993). For example, the term peatland is actually synonymous to the European terms mire and moor found in the literature. Understanding of peatlands is also complicated by the fact that different ecological bases of classification have been used by various authors (e.g., Moore and Bellamy 1974,

Cowardin et al. 1979, Heathwaite et al. 1993). Peatlands have historically been classified according to their topography, ontogeny, hydrology, hydrochemistry, and/or plant community composition (Bridgham et al. 1996).

Because the term fe n may have slightly different meanings for researchers in different geographical locations, it is important to define the term as it relates to the present study. One of the most comprehensive definitions of fens, as they occur in Ohio, is presented by Andreas (1985):

A fen is characterized by having (1) relatively clear water coming from an artesian source which surfaces as springs or seeps, (2) a wet, springy calcareous substrate which supports minerotrophic species of Sphagnum and other bryophytes which do not accumulate to form a continuous mat, (3) vegetation dominated by members of the Cyperaceae, Compositae, Rosaceae and Gramineae with approximately 20% of the vegetation made up of shrubs, usually including Potentilla fruticosa, and (4) water pH between 5.5 and 8.0.

38 In addition, fens exhibit more species richness than their ombrotrophic bog counterparts

(Bryan and Andreas 1988).

Fens are found within glaciated regions and are typically located around esker- kame complexes associated with major end moraines (Stuckey and Denny 1981). In

Ohio, glacier-created fens and bogs are located at the southern edge of their North

American range. Thus, representative communities are small in terms of area covered, and are not represented by classical examples such as those found in Michigan,

Wisconsin or Minnesota (Andreas 1985).

Fens are highly alkaline because the glacial deposits through which water percolates contain large amounts of limestone-rich gravel or dolomite. Groundwater moves through these highly permeable deposits and dissolves limy materials containing calcium and magnesium bicarbonates which may precipitate as marl, a gray substrate rich in lime (Denny 1979, 1994). High water alkalinity, in association with the oxygen deficiency of ground water, significantly reduces the number of decomposer species and results in the accumulation of peat which may consequently cause nitrogen deficiency in the waters (Denny 1979). Furthermore, water fi’om springs tend to remain relatively cold throughout the growing season; temperatures at root level can reach 4.5° C cooler than surface temperatures, thus reducing the ability for roots to absorb water (Denny 1994).

A limited number of species can survive the extremes in wetness, nutrient levels, temperatures and alkalinity in fen environments; limiting factors favor species which can tolerate cold, moist soils and a short growing season, and can minimize water loss

39 through évapotranspiration (Denny 1994). Some of the best examples of fen indicator

species are Potentilla fruticosa (shrubby cinquefoil), Pamassia glauca (grass-of-

pamassus). Lobelia kalmii (Kalm’s lobelia), Gentiana procera (fringed gentian), and

Solidago ohioensis (Ohio goldenrod) (Stuckey and Denny 1981). Other common fen

taxa are listed by Gordon (1969), Stuckey and Denny (1981), and Andreas (1985).

A well-developed fen is characterized by three successional stages or zones of

development: (1) the open marl zone, (2) the sedge-meadow zone, and (3) the shrub-

meadow zone (Stuckey and Denny 1981). The marl meadow, often centrally located, is

characterized by sparse growth of low-growing sedges, rushes and grasses. Unlike in

bogs, peat in fens is mostly derived from sedges and other higher plants, rather than from

Sphagnum (Pringle 1980). Farther away from the center, thin deposits of sedge peat

accumulate over the marl and improve substrate drainage, thus enabling establishment of

other species, and forming the sedge meadow zone. Finally, woody species in the shrub

meadow zone is established as mineral soils begin to develop (Stuckey and Denny 1981).

Stuckey and Denny (1981) have classsified Ohio fens into prairie fens and bog fens, based on their floristic composition and geographical affinities. Prairie fens contain species which typically occur in wet prairies of western Ohio, and are generally found in the west-central part of the state. Bog fens, found in the northeast, are more typically associated with acid bogs of northern Ohio.

40 CHAPTERS

DESCRIPTION OF STUDY AREA

Betsch Fen is located along Blackwater Creek in Green Township, northern Ross

County, Ohio (Figure 3.1). It lies east of US 23, and 2.4 km (1.5 miles) south of the

Ross-Pickaway county line (at approx. 39°27’30” N, 82°57’30” W). This 14 hectare (35

acre) wetland was formerly known as Blackwater Fen or Goodman Bog. It was

dedicated as Betsch Fen, a state nature preserve, after Charles and Dorothy Betsch of

Chillicothe donated the site to the Ohio Chapter of The Nature Conservancy (TNC) in

1984.

The preserve is considered to be one of the largest remaining alkaline fens in

Ohio. It has statewide significance because of its southerly location and unusually high quality; there appears to have been little disturbance to the plant communities and natural features of the land over thousands of years. This relict wetland is located within the

Glaciated Allegheny Plateau physiographic region of Ohio (Figure 3.1) (McCance and

Bums 1984, Andreas 1989), and was formed more than 10,000 years ago by the

Wisconsinan glacier. It occurs on a kame-esker complex (Andreas 1985), in a

41 OHIO

20 40 60 m iles

20 40

Figure 3.1 : Location of Betsch Fen in Ohio (arrow). Hatched area indicates the Glaciated Allegheny Plateau physiographic and phytogeographic region (from McCance and Bums 1984 and Andreas 1989).

42 depression between two end moraines along the southernmost boundary of the glacier

(Hillmer 1991).

Blackwater Creek runs through the preserve and divides the fen into a northern and southern section. The glacial moraine deposits on both sides of this creek provide the source of clear, cold flowing water for the fen through a series of permanent springs.

There has been, however, no known study on the hydrology or chemical characteristics of water at this site.

Surficial geology in the study area is dominated by Wisconsin glacial till. In the published soil survey of the county, Petro et al. (1967) mapped two main soil series in the fen area proper, i.e., the Willette (Terric Medisaprists) and Wallkill (Thapto Histic

Fluvaquents) series, with Carlisle muck (Typic Medisaprists) at the more wooded northwestern comer of the preserve. However, the Ross County soil survey is currently being updated; at this time field work and mapping has been completed, but the final report has not been published (Soil Conservation Service, personal communication).

Advance copies of maps from the new survey, which are still subject to change, show the same area as being dominated by Carlisle muck, Adrian muck (Terric

Medisaprists) and a Kendalville-Eldean (Typic Hapludalfs) complex. Both Carlisle muck and Adrian muck are very poorly drained hydric soils with a seasonal high water table depth at near or above the surface, and a depth to bedrock of more than 150 cm.

Kendall ville and Eldean are well drained soils with moderately slow permeability (ODNR

DSWC 1993). Soil in the wooded upland bordering the south of the fen proper is

43 mapped as Rodman (Typic Hapludolls) gravelly loam, an excessively drained soil, with

moderately rapid to very rapid permeability.

Ross County has a daily mean temperature of 18.2° C, a mean annual

precipitation of about 96.5 cm (mean annual snowfall of 54.9 cm), and a growing season

of 176 days (Andreas 1989). A combination of these climatic factors, i.e. higher daily

mean temperature, lower snowfall, and longer growing season compared to most

counties in the physiographic region, has affected the differences observed in the overall

vegetation of Ross County and the other extreme southern counties within the Glaciated

Allegheny Plateau (Andreas 1989).

Due to the lack of alien species and presence of many fen indicator species,

Betsch Fen represents an outstanding example of a fen community. There are several

plant species classified as potentially threatened with extinction in Ohio, e.g. Carex

tenera, Filipertdula nibra, Gentiana procera, Pamassia glauca, Potentilla fruticosa,

Sanguisorba canadensis, and Solidago ohioensis (unpublished TNC files, Roberts and

Cooperrider 1982, Stuckey and Roberts 1982). Although TNC lists Carex trichocarpa

as a threatened species, it was not listed as such by McCance and Bums (1984), and is

listed as being potentially threatened by Stuckey and Roberts (1982). TNC has

conducted a number of vegetation surveys, but most have only been qualitative, and very

little of the information has been published.

Stuckey and Denny (1981) included Betsch Fen in their study on the floristic

composition and geographic afiflnities of fens in Ohio, and thus provided some information on plant species found in the area. They have classified Betsch Fen as a

44 prairie fen because species found there typically occur in wet prairies of western Ohio.

From vascular plant species considered distinctive of fens and analysed in Stuckey and

Denny’s study, Betsch Fen contained 11 of 24 wet prairie species, and 18 of 30 species characteristic o f both prairie and bog fens. In a more recent study, Schneider (1992)

identified 15 sedge species in Betsch Fen; Rhynchospora capillacea, Scleria verticillata, and 13 species of Carex.

Although fens can be distinguished from other communities by their requirement for constant supplies of calcareous waters and formation of marl, these characteristics are not always obvious in the field (Anderson 1982). This is indeed the case in the northern section of Betsch Fen where there is no clear open marl flat or distinct pattern of vegetation zonation. In contrast, the fen section south of Blackwater Creek clearly shows the typical zones of vegetation development as described by Stuckey and Denny

(1981): (1) a somewhat centrally located open marl zone, (2) a sedge meadow zone, and

(3) areas of shrub meadow, bordered by upland woods. In addition to the above zones, the south fen is also characterized by distinct and almost pure stands of sweet flag

{Acorus calamus).

The A. calamus stands seem to be preferred by deer {Odocoiletis virginiamts) which use the fen quite extensively as a feeding and bedding site. Deer trails are found throughout the area, and some places become heavily trampled by late summer. Betsch

Fen is also home to the uncommon spotted turtle (Clemmys guttata) and the Baltimore checkerspot butterfly (Euphydriasphaeton) (Hillmer 1991).

45 In order to maintain the delicate balance of the fen community, TNC has been

managing the site through volunteer stewardship efforts. These efforts include

prescribing bums and cutting back woody plants such as sandbar willow {Salix exigiia) and dogwood (Comus sp.) to prevent their spread into the open fen area (Hillmer 1991).

46 CHAPTER 4

VEGETATION MAPPING AND DESCRIPTION

4.1 Background and objectives

Betsch Fen has been under The Nature Conservancy (TNC) management since

1984. Although prior to this time the site had been in private ownership, it has long been recognized as an ecologically important wetland. For example, Herrick (1974), Cusick and Troutman (1978), and Stuckey and Denny (1981) referred to plant species found in the site, known at the time as Blackwater Fen or Goodman Bog. Under its management,

TNC has conducted qualitative vegetation surveys of the area (unpublished TNC files).

However, few ecological studies have been conducted or published, and thus ecological information is generally lacking. As the present research project aimed to characterize boundary locations within the site, it was important to first obtain a base map and recent vegetation description of this site.

This chapter aims to present a more detailed description of vegetation patterns in the study area. This part of the study was conducted (1) as reconnaissance to characterize and stratify the site into areas of similar vegetation, and (2) to meet one of the specific objectives mentioned in Chapter 1, i.e., to provide an updated vegetation map of the area. The map, produced through aerial photo analysis and field surveys, was

47 later used to determine the most appropriate sampling locations and methodology for subsequent stages of the research. It also provided approximate vegetation boundary locations which were used to formulate hypotheses to be tested by boundary detection methods described later in chapters 7 and 8. In addition, the understanding of vegetation patterns in this part of the study is also supported by examining changes shown in historical aerial photographs.

4.2 Methods

Vegetation mapping

Vegetation mapping followed procedures based partly on Kiichler’s

Comprehensive Method (Küchler 1988a), using a combination of aerial photo analysis and field surveys. The first step was to study the area to be mapped based on available literature and maps (e.g., topographical and soil survey maps). Because the area to be mapped is relatively small, emphasis was placed on the next step, which was to conduct careful study of aerial photographs. Observations were primarily based on a 1981 aerial photograph, with a scale of 1:4,800 (1 inch = 400 feet). During the course of this study, in October 1994, a new aerial photograph of the site was taken by Henderson Aerial

Surveys, Columbus, Ohio. In general, there were no significant differences between vegetation patterns in the 1981 and 1994 photographs, and the latter was used to support earlier observations. Contrasts between different types of vegetation were noted.

48 and every area seen as different from neighboring areas were bounded by a marker line traced on an acetate sheet placed directly on the photograph.

In the field, every individual area or patch outlined from the aerial photographs were visited, and all relevant data entered in the phytocenological record (Küchler

1988a). The main focus was to survey and identify dominant species in each patch.

Specimens of unidentified species were collected for later identification. Considering the objectives of this preliminary stage of the study, comprehensive transect sampling was not conducted at this stage. The important objective of this exercise was to conduct ground truth measurements of distances and areas covered by each vegetation patch, thus noting how boundary lines may need to be shifted, added or removed after field inspections. Records for each patch included pertinent physiognomic and floristic information, as suggested by Zonneveld (1988) and Küchler (1988b). Additional comments on soils, water conditions etc. were also noted.

After adequate field surveys, the base vegetation map was then prepared based on field records. This map showed the outline of each vegetation type, i.e. each area shown separately on the aerial photographs, coded by numbers which corresponded to field observation records. Classification for the final vegetation map was partly based on physiognomy (whether dominant species was herb, shrub or tree), but labeled floristically according to dominant species. A minimum number of categories was designated in stratifying the site to show major structural stages of the fen vegetation

49 The final map was designed using Freehand™ graphics software (Altsys

Corporation 1994), and scale measurements confirmed using SigmaScan™ image measurement software (Jandel Corporation 1995).

Changes in historical aerial photographs

Aerial photographs of Betsch Fen available for study were taken in 1938, 1957,

1964, 1973, 1981 and 1994. These photographs were scanned using a TWAIN compatible scanner and converted into Joint Photographic Experts Group (JPEG) file format, to be retrieved in SigmaScan™ (Jandel Corporation 1995). Within this program, dimensions of the preserve were spatially calibrated according to the legal description of the property (unpublished TNC files) then measurements of open area in both the south and north fens were taken. The open area, visually determined according to intensity contrasts in the aerial photographs, included all plant associations encompassed by, but not including, woodland canopy. To reduce errors in calibration and/or measurement, each estimation was conducted ten times, and mean measurements presented.

4.3 Results

Field observations

As mentioned in the previous chapter, Blackwater Creek runs through Betsch

Fen preserve and divides it into a northern and southern section. While observations were conducted in both sections of the fen, the present study has focused primarily on the southern section as it is a more well developed fen which exhibits distinct vegetation

50 zonations typical of fen systems (Stuckey and Denny 1981). Furthermore, this southern section is almost entirely enclosed within the preserve, while parts of the northern section are directly bordered by private property, and would thus have been unsuitable for natural boundary studies.

The meadow plant community in the south fen is characterized by three typical successional stages or zones of development: (I) the open marl zone, (2) the sedge meadow zone, and (3) the shrub meadow zone. The open marl zone, characterized by shallow pools on mucky, well-exposed marl is sparsely populated by Chara sp., Jimctis brachycephalus and Rhynchospora capillacea. Immediately surrounding the open marl is an association of Scirpus acutus and R. capillacea. Other fen indicator species found around the open marl include Gentiana procera. Lobelia kalmii, Lysimachia quadriflora, Pycncmthemum virginicmum and Rudbeckia hirta.

The herbaceous open meadow zone, characterized by sedges, is primarily dominated in cover by Carex spp. As flowering period for sedges are limited, it was often difScult to identify different Carex species. However, in many areas it could be determined that Carex stricta was dominant, as this hummock sedge clearly forms tussocks. Other Carex species were often found in smaller patches. C. buxbaiimii, C. hystricina and C. suberecta occurred throughout the meadow, but were particularly abundant around the open marl zone. In contrast, the potentially threatened C. trichocarpa was only found in an open area at the extreme west of the fen.

Although Carex spp. generally dominated coverage, sedge-herb associations characterized certain areas. Dominants varied among areas, and included Carex-

51 Sanguisorba, Carex-Solidago, and Carex-Eupatorium. In addition, marsh fern

(Dryopteris thelypteris) was consistently found across the sedge meadow zone.

Populations of the potentially threatened Filipendula rubra, Gentiana procera,

Sanguisorba canadensis and Solidago ohioensis are found throughout the open

meadow. In particular, Sanguisorba canadensis was abundant throughout much of the

area, while the other three species seemed concentrated in smaller patches. Filipendida

rubra was observed at the immediate west of the marl center, close to the riparian

woods boundary zone. Patches of Gentiana procera were found at the extreme west of

the open meadow, and around the open marl zone. Finally, Solidago ohioensis seems

concentrated between the open marl and the shrub meadow zone, often associated with

other Solidago species, i.e., S. canadensis, S. gigantea, S. patula and S. uliginosa.

The shrub meadow successional stage is dominated by Salix interior, along with

Carex spp. in the herbaceous layer. Other fen indicator species can also be found in this

layer. However, as this zone is located farther away from the center groundwater source

and presumably receives a higher oxygen supply from surface water (Stuckey and Denny

1981), fen species are found interspersed with common wetland species which may not be indicative of fens. These include Eupatorium perfoliatum, Galium sp., and Impatiens capensis.

In addition to the three vegetation zones described above, there are distinct stands of Acorns calamus, a monocotyledonous Araceae species. These populations of

A. calamus appear as almost pure stands with little else growing with them, except for the occasional Carex stricta and Scirpus atrovirens.

52 Woodland areas surround the fen meadow, both along the creek and towards the

upland area. Tree, shrub and herbaceous species composition represented typical species

of wet and mesic environments. Tree species diversity is relatively low, and dominated by Acer negundo, Gleditsia triacanthos, Jtiglans nigra, Plaiams occidentalis, Pmmis serotina and Ulmus rubra.

In general, plant species found in the south fen were also found in the north fen.

A list of species observed and identified during this study is presented in Appendix A.

Vegetation mapping

Five major plant communities were distinguished for mapping purposes based on the above observations. These are: (1) Scirpus central marl zone, (2) Carex sedge meadow zone, (3) Accrus stand, (4) Salix shrub meadow, and (5) woodland (Figure

4.1).

The Scirpus central marl zone includes the actual open marl area and its immediate surrounding. Although Scirpus acutus can be found scattered in other parts of the fen, the abundance of this species around the open marl zone was clearly evident.

As mentioned above, Carex sp. was generally dominant in determining cover in the sedge meadow zone, despite various sedge-herb associations. Therefore, these associations were collectively designated as the Carex sedge meadow zone. Although the Accrus calamus stands were also found in the herbaceous open meadow of the fen, these patches were distinguished fi'om the rest of the sedge meadow because of the clear dominance of A. calamus.

53 Figure 4.1 ; Vegetation map showing the five major plant communities in Betsch Fen.

54 M m

AAAAAAAAAAAA AAAAAAAAAA AAAAAAAAA AAAAAAAAAA AAAAAAAAAAAAA AAAAAAAAAAA^ AAAAAAAA A A A A A A AAAAAAAAAAAAAA AAAAAAAAAAAAAA AAAAAAAAAAAAAAA 0 AAAAAAAA AAAAAAAAAAAA AAAAAAAAA AAAAAAAAAAAAAAA Ui AAAAAAAAAAAAAAAAA A w'wW f '^ 'V LA AAAAAAAA A ^ y t y ^ w r

^IwlwTwiwTwf^^ ( W « w * w * w lw l A I A yTwTyTwTyTwT^ A a A A I ^TwTwTYTYTyryi.» a a a a X* A •wTwTw>wTw>w««vT^ A a a a / a X w Vw Tw Tw Tw I w ^ a a a a 88 ^TwTw *Y'Y'Y'V'V'^ AAA ih !! :’ : ! A'Tw Tw I w Tw Iw Tw Vw a a a a

Central marl [^>>?] Sedge meadow Acorus stand ml Eftwa Shrub meadow ü vybodland 1 HilitUiitlti: I Figure 4.1 The shrub meadow and woodland areas were differentiated based on

physiognomy of dominant species. Salix sp.clearly dominated the shrub meadow zone,

while woodlands were not characterized by the dominance of a particular species.

Woodland areas consisted of both upland woods bordering the fen and riparian woods

along Blackwater Creek. These were not differentiated because woodland areas

intergrade with each other and share similar species composition.

Historical changes

Image measurements on aerial photographs suggest vegetational dynamics which may have affected the extent of open fen area, although the time of year in which photographs were taken must be taken into consideration (Table 4.1). Measurements

Year Time of year South fen (ha) North fen (ha) Total (ha)

1938 October 2.882 + 0.030 2.282 + 0.030 5.164 1957 July 2.944 + 0.038 2.806 + 0.026 5.750 1964 May 2.844 + 0.022 2.559 + 0.026 5.403 1973 September 2.828 ± 0.023 2.526 + 0.051 5.354 1981 November 4.030 + 0.038 2.694 + 0.031 6.724 1994 October 3.746 + 0.026 2.869 ± 0.026 6.615

Table 4.1: Estimate of fen area (i.e. central marl, sedge meadow and shrub meadow) by image analysis of aerial photographs. Areas reported as means and standard errors from ten repeated measurements.

56 presented are totals of open fen area consisting of areas occupied by plant associations

in the open fen meadow, i.e. consisting of the central marl, sedge meadow

(including Acorus stand) and shrub meadow. Results suggest that there has been no

significant loss in fen area. On the contrary, area measurements from recent years were

the highest.

4.4 Discussion

Strongly minerotrophic ‘rich’ fens of western Ohio generally have a higher

species composition than weakly minerotrophic ‘poor’ fens of eastern Ohio (Stuckey and

Denny 1981, Bryan and Andreas 1988). Betsch Fen, located in the south-central part of

the state, is unique in that it shares typical wet prairie species of western Ohio fens

(Stuckey and Denny 1981), yet is within the Glaciated Allegheny Plateau which primarily

contains weakly minerotrophic peatlands (Andreas and Bryan 1990).

Stuckey and Denny (1981) listed 74 plant species considered indicators of Ohio fens. From this list, they marked 29 species found in Betsch Fen, 27 of which were also recorded during the present study (Carex buxbaumii, marked as ‘not found’ in Betsch

Fen by Stuckey and Denny, was in addition recorded during this study). In general, species found in Betsch Fen during this study well represented common taxa found in

Ohio fens, as described by Gordon (1969), Denny (1979), Stuckey and Denny (1981), and Andreas (1985). However, Potentilla fniticosa (shrubby cinquefoil), considered an indicator species which is almost always found in fens (Anderson 1982, Andreas 1985), was not found in the south fen during this study. With the exception of one shrub in the

57 western area of the north fen, this species seems nearly absent. Similar observations were reported by Anderson in 1980 (unpublished TNC files).

Knoop and Andreas (1987) found the plant community in Sinking Creek Fen, in western Ohio, as being floristically rich with approximately 150 species of vascular plants. They noted that a single square-meter quadrat may contain as many as 30 species.

Similarly, Slack et al. (1980) characterized strongly minerotrophic fens as containing numerous species, each species having low dominance, but high constancy throughout the fen. In this study, however, it was found that most of the area was characterized by dominance of a particular taxon, as will be shown by quantitative data discussed later in chapter 7. For example, in the sedge meadow zone, Carex spp. generally constituted more than 75% of plot coverage. Similarly, Acorus calamus accounted for nearly 90% of coverage in the Acorus stands. This therefore facilitated the designation of community types based on dominant species.

In the open meadow, the number of species recorded within a 1.0 x 0.5 m^ quadrat (data discussed in chapter 7) was generally fewer than ten. However, this number increased in areas away fi"om the fen center, where other wetland species that are not necessarily indicative of fens become established.

The establishment of typical fen species has been attributed to their tolerance to high alkalinity ground water which is deficient in both oxygen and nitrogen (Denny 1979,

1994). Establishment of these species consequently reduces the ability of non- indigenous plants to invade and colonize the fen (Hillmer 1991). However, typical fen species must still compete with other “non-fen” wetland species. For example, a few

58 individuals of Typha latifolia were found at the center marl zone during this study.

Typha spp. have been considered undesirable by many because they are rapid colonizers

but are of limited ecological value (Mitsch and Gosselink 1993, based on Odum 1987).

Such potential threats to typical fen species have been of particular concern to TNC.

Another current threat is that of woody plants which may spread into the open area and

shade out herbaceous fen plants (Hillmer 1991). The extent to which interspecific

can affect overall vegetation patterns in this system is still open to question.

The major vegetation zones observed fi’om this study were similar to those

reported in a qualitative survey by Anderson in 1980 (unpublished TNC files), although

he did not mention the establishment of Acorns stands which potentially can also spread

into the high quality fen and compete with rare fen species. Continued observations over

three field seasons of this study (1994 to 1996) recorded slight shifts in vegetation limits

which suggest the importance of competition between Acorus calamus and other

species. For example, in the boundary zone between Acorus calamus and Carex spp. in

the southwest area of the south fen open meadow, Carex spp. replaced Acorus calamus

where the latter had dominated in the previous year.

Other than interspecific competition, another factor which may affect vegetation

patterns in the future is grazing and trampling. An important large mammal population in the fen area is white-tailed deer (Odocoileus virginiamis). In 1980, Anderson

(unpublished TNC files) reported limited grazing damage to the site, but cautions that grazing could become a future threat. The effects of this disturbance on fen species establishment is still unknown. The abundance of thorny plants, such as Crataegus sp.,

59 Gleditsia triacanthos and Madura pomifera, in the woodlands suggest the historical significance of plant defense against grazing. Such plants act as inhibitors to crossings of large mammals, and may have functional importance in determining movements of species across landscape boundaries (Forman 1995).

Along with the actual grazing, trampling of herbaceous vegetation can be severe in certain areas. Other than in established deer trails, however, vegetation seems to recover very well fi-om trampling. For example, paths made for sampling purposes during one growing season in this study were not at all detectable in the following spring.

Direct comparisons between aerial photographs taken during different decades can only be made with caution as these photographs were taken at different times of the year, and also varied in quality and resolution. As would be expected, larger open areas were detected during autumn months, possibly due to some reduction in canopy coverage. However, visual inspection of photographs did not suggest significant changes in this coverage. Another factor which may affect satisfactory comparisons is the potential weakness of using scanned photographs, instead of direct digitizing (Clarke

1990). Furthermore, the quality of measurements made using SigmaScan™ can be affected by errors caused by measurement precision, resolution and accuracy. Despite these problems, however, comparisons between photographs can still provide useful information about general patterns and establishments of plant communities. For example, visual inspections revealed the early establishment of Acorus calamus stands and the development of dense woodland canopy during the 1950s.

60 In summary, Betsch fen is a high quality fen exhibiting typical fen successional stages and containing many typical plant species representative of Ohio fens. Preliminary observations during this part of the study provided some information on the three dimensions of edge (Forman and Moore 1992, Forman 1995), i.e., width, verticality and length, including curvilinearity. The clear dominance of certain taxa produced relatively sharp boundaries, i.e. short edge widths, in vegetation. However, changes in vegetation did not seem to correspond entirely with boundaries of soils and hydrology, as saturated substrate was still found under woodland canopy.

Although earlier surveys have provided important floristic information on Betsch

Fen, little is known about vegetation dynamics at this site, and how patterns may change in the future. The suggestion that there have been temporal shifts in vegetation patterns calls for further ecological investigations.

61 CHAPTERS

WATER CHEMISTRY

5.1 Background and objectives

Hydrology and geochemistry have been considered to be the two main factors determining the development and vegetation composition of peatlands. As a particular type of peatland, fens have been identified by the presence of distinct vegetation associated with highly alkaline substrate and groundwater (Gordon 1969, Denny 1979,

1994). Ground water alkalinity is relatively high in these systems because it dissolves limy materials and becomes charged with calcium and magnesium bicarbonates as it percolates through the underlying limestone-rich gravel deposits (Denny 1979, 1994).

Carbonates, bicarbonates and carbonate-containing solids are components of the carbonate system, the most important acid-base system in natural waters (Snoeyink and

Jenkins 1980, Stumm and Morgan 1981).

Several studies have found that the presence of typical fen species are closely related to alkalinity or substrate calcium content (see Beltman and Verhoeven 1988,

Boeye et al. 1994). The effect of calcium (and magnesium) is believed to be primarily that of a conditioning factor which secondarily controls important conditions for plant growth, such as pH, resultant solubilities of other elements, cation-saturation of

62 exchange sites and microbiological processes that affect nutrient availability (Wassen et

al. 1990). These observations suggest that studies attempting to relate water chemistry

to vegetation in fens should consider water alkalinity as an important environmental

factor.

The concept of alkalinity is often confused with the concept of pH. The alkalinity

of a water, defined as its acid-neutralizing capacity, is a function of all its titratable bases

(APHA 1985). Any dissolved substance with the capacity to consume protons can

contribute to alkalinity. However, because the alkalinity of many waters is primarily

determined by dissolved carbonates, bicarbonates and hydroxides, it is used as an

indicator of the concentration of these constituents (APHA 1985, Ludwig 1985). The presence of carbonates and bicarbonates imparts a buffering capacity for water; as alkalinity is reduced, the buffering capacity of that water is lowered, and pH is more easily modified (Ludwig 1985). Therefore, the stability of pH is directly proportional to the alkalinity of that water. Substrate pH strongly affects the availability of nutrients for plant use; circumneutral to alkaline (=basic) conditions have been found to decrease the availability of nutrients such as P, K, Mg, Fe, Mn, B, Cu and Zn (Lucas and Davis 1961).

Water conductivity, a measure of total ionic concentration, can be used as an indicator of a water’s overall nutrient content (Holland 1996). In a study relating fen vegetation to hydrology and nutrient dynamics, Verhoeven et al. (1988) found that the ionic composition of the fen water strongly affected the species composition of vegetation. They also found that during periods of flooding from an outside source, conductivity of both surface and ground water decreased most rapidly in the fen ecotone

63 areas, and suggested that these changes were due to active plant nutrient uptake in the ecotone (Verhoeven et al. 1988, Holland 1996).

Studies on the correlation between vegetation and element content in European peatland (mire) waters are well documented (eg. Sjors 1950, Gorham 1950, 1956,

Gorham and Pearsall 1956, Beltman and Verhoeven 1988, Verhoeven et al. 1988,

Wassen et al. 1989 and 1990, Wassen and Barendregt 1992, Boeye et al. 1994). There have been fewer attempts, however, to correlate vegetation patterns and floristic diversity with physical gradients in North American peatlands (Vitt and Bailey 1984).

Because of differences in vegetation (Slack et al. 1980) and environmental conditions, direct comparisons between European and North American fens may not always be appropriate. Recent studies linking vegetation to water chemistry in US fens include those by Glaser et al. (1990), Cooper and Andrus (1994) and Walbridge (1994).

Long-term studies on water chemistry in Ohio fens are generally lacking. Bryan and Andreas (1988) present baseline data on various ground water chemistry parameters from a number of fens in western Ohio, and also examined northeastern Ohio fens in an earlier study (Bryan and Andreas 1986). Although they conducted some temporal monitoring, they did not examine spatial variability within sites.

The objective of this part of the study was to collect baseline data to examine temporal and spatial variability in shallow groundwater alkalinity, pH and conductivity in

Betsch Fen. This information should provide insight into temporal shifts in water chemistry patterns, particularly alkalinity, as a functional determinant which may affect

64 structural boundaries of vegetation. As a preliminary study, however, it is not intended to be a comprehensive investigation of chemical and physical characteristics of water at this site. Analyses relating results from this part o f the study to other objectives o f the project will be discussed in subsequent chapters.

5.2 Methods

Field sampling

Sixty permanent water sampling points, consisting of 11 points in the north fen and

49 points in the south fen, were established along transects designed to capture the variation of vegetation, soils and topography present in Betsch Fen. In the north fen, transects ran SW to ME, and covered the NW and NE comers of the preserve. To capture the distinct zonation of vegetation in the south fen, four parallel transects ran from W to E, supplemented by additional SW to NE transects (map in Appendix B).

Location of points within transects were selected visually to represent site variability as judged by vegetation.

Based on field observations and vegetation studies (Chapter 4), communities in the south fen were distinguished into riparian woods, Scirpus central marl, Carex sedge meadow, Acorus stand, Salix shrub meadow and upland woods. An additional category of upland boundary zone was necessary to designate sampling point locations at fen- upland boundary zones in which wells were under dense canopy outside of fen vegetation, but were still within saturated soil.

65 The number of sampling points varied among plant communities. As requested by

TNC, no wells were placed within the open marl proper. To represent this area, six sampling points were established in the Scirpus community immediately surrounding the open marl. The Carex open sedge meadow was the largest area traversed by transects, and contained 21 sampling points. The number of sampling points in other communities varied between one and seven. The upland community was only represented by one sampling well because presampling observations indicated that wells placed in the substrate, an excessively drained gravelly loam soil, do not collect water easily. This suggested that no shallow ground water is available for plant use at the depth at which water is sampled in other wells. Therefore, it was assumed that even if more sampling points were established, little water would be collected, and this would have had little effect on overall data collection.

Designation of plant communities in the north fen was more difficult due to the absence of distinct vegetation zones typical of fens. For the purposes of data analysis, however, sampling point locations were broadly categorized into three areas: open meadow zone, riparian woods-open meadow boundary zone, and upland woods-open meadow boundary zone.

Sampling points were established by driving sampling wells of polyvinyl chloride

(PVC) pipes with a diameter of 2.5 cm vertically into the ground. In order to sample shallow ground water around the plant root zone, pipes were designed approximately 50 cm in length, with lower segments perforated and covered with mesh screen to allow water to enter and be collected for analysis (Figure 5.1).

66 Removeable Cap

2.5 cm dlam. PVC pipe

Substrate Surface

Root zone Groundwater Table

j Sampling Zone

Fixed End Cap

Figure 5.1: Principle of ground water sampling well (adapted from Kadlec 1989).

67 Sampling was conducted biweekly between 28 June and IS November 1994, and between 29 April and 26 October 1995. At each sampling, water was drawn out of the wells by using a clear PVC tube pipette and then collected in acid washed polyethylene bottles.

Short-term monitoring of water level fluctuations in a number of representative wells was conducted between August and October 1995 to provide supplementary data.

Water level was measured as the distance from the soil surface to water level within a sampling well.

Sample analysis

Water pH and conductivity were measured directly in the field using a Coming

Checkmate™ modular testing system. Samples were then transported under refrigeration to the laboratory where they were immediately analysed for alkalinity by potentiometric titration with 0.1 N HCl to an end point of pH 4.5 (as determined by

APHA 1985). All samples were analysed within eight hours of sample collection.

Data analysis

Data on water alkalinity, conductivity and pH were analysed for temporal and spatial patterns. To examine seasonal and plant community effects, water alkalinity data were analysed using repeated measures analysis of variance (RMANOVA) (SAS 1985).

Geostatistical analyses on water alkalinity, conductivity and pH data were conducted by semivariance analysis using (Gamma Design Software 1992). This

68 analysis quantifies the degree of spatial autocorrelation among samples and calculates

the maximum distance at which the samples of the measured parameter are significantly

correlated (Boemer et al. 1996).

Spatial autocorrelation analysis is based on the assumption that variation for a given attribute is a function of distance; small separations denote similarity, while large

separations denote dissimilarity (Clarke 1990). GS* calculates an autocorrelation index

(the semivariance) among groups of sample pairs separated by a given distance, and then produces a composite graph (the semivariogram) relating the semivariance among samples and the distance between samples. The two main applications of the semivariogram are for structure recognition and optimum interpolation (Burrough 1995).

GS^ also performs unweighted least squares analysis to fit a range of models to the semivariogram.

The total model variance of a semivariogram can be divided into two components:

(1) the structural variance, C, which is related to spatial pattern or structure in the data set, and (2) the nugget variance, Co, which can be interpreted as the measurement error associated with the method of sampling. Relative structural variance, i.e. the ratio of structural variance to total model variance [ C/(C + Co) ] can be used as a measure of spatial dependence. Where this ratio approaches 1, spatial dependence is high, and the system is strongly spatially structured. Where this ratio approaches 0, spatial dependence is low (see Burrough 1995, Robertson and Freckman 1995, Boemer et al. 1996).

69 Best-fit semivariogram models were used to map spatial patterns in water

chemistry by kriging, an optimum method of interpolation (Fortin et al. 1989, Burrough

1995), using SURFER® (Golden Software Inc. 1989 ).

To acquire a better understanding of how spatial patterns in water chemistry

differed temporally, analyses were also conducted on data from 1994 and 1995

separately, as well as from spring (April-May), summer (June-August) and autumn

(September-November) separately. Data used were mean measurements from individual

water sampling wells, taking into account their spatial coordinates within a 190 x 105 m^

area in the south fen which encompasses all wells. Due to small sample size,

geostatistical analysis was not performed on the north fen data.

5.3 Results

Temporal patterns

During the sampling periods, water alkalinity levels exhibited marked temporal

fluctuations, with mean values and ranges which varied among plant communities

(Figures 5.2 and 5.3). Within a plant community, however, relative readings from individual wells were consistent, i.e. wells which exhibited relatively low alkalinity levels were consistently low, while those which exhibited high alkalinity levels were consistently high.

In the south fen, temporal fluctuations in alkalinity were more pronounced in vegetation zones away from the fen center. Readings in the Scirpus central marl zone

(Figure 5.2B) fluctuated the least, and were also most consistent among wells, as shown

70 Figure 5.2 A-G: Temporal fluctuations in water alkalinity (closed circles) and conductivity (open circles) in the south fen according to plant community. Vertical bars represent standard errors of the mean. In G, data points are missing for observation periods in which well was empty, or insufhcient sample volume was collected (see text for explanation).

71 800 -/X------y /_ 1400 A. riparian woods (n=7) B. Scirpus central marl (n=6) 700 1200 600 1000 500 600 400 600 300 400 200

100 200

-i—i —I—j—I 1. iL I— y 1 '

800 1400 C. Carex sedge meadow (n-21) 700 1200 600 1000 500 800 400 600 300 E 400 200

200 D. Acorus stand (n=5) 5 > 800 ------03 700 E. Salix shrub meadow (n=3) "O c 600 8 500 400 1 300

200

100

JJASON MJJASO 800 -T" /- 1994 1995 . G. upland woods (n=1) 700 1200 sampling month 600 1000 500 800 400 600 300 w ater alkalinity 400 200 w ater conductivity 100 200

sampling month

Figure 5.2

72 Figure 5.3 A-C: Temporal fluctuations in water alkalinity (closed circles) and conductivity (open circles) in three areas of the north fen. Vertical bars represent standard errors of the mean.

73 A. npanan woods - open meadow boundary zone (n=2)

800 1400 CO B. open sedge meadow (n=5) 700 1200 I CO 600 3 1000 J ? 500 > 800 u 400 3 "D 600 C 300 400 8 200 a 2 00 CO 100 $

800 / / ------1400 C. upland woods - open meadow boundary zone (n=4) 700 1200

600 1000 500 800 400 600 300 400 200 water alkalinity water conductivity 200 100

JAS OMNJ JJA S O 1994 1995 sampling month Figure 5.3

74 by the low standard errors. In contrast, the widest range in means was recorded in the

Acorus stand (Figure 5.2D). Overall, there was a highly significant plant community effect (p < 0.0001, Table 5.1). Further analyses by a Ryan-Einot-Gabriel-Welsch

Multiple F test (SAS 1985) revealed that for all repeated measures, alkalinity levels in the Acorus stand were always significantly different (at a = 0.05) from all other communities.

Source DF MS F Value Pr>F

South fen data Community 5 679254.675 26.01 0.0001 Season 2 38024.909 1.46 0.2379 Community X Season 10 13757.548 0.53 0.8678 Error 104 26117.568

North fen data Community 2 20319.879 1.60 0.2245 Season 2 9839.802 0.77 0.4735 Community X Season 4 3214.012 0.25 0.9053 Error 23 12737.430

Table 5.1: Repeated measures analysis of variance results for water alkalinity levels in both the south and north fens.

Readings in the north fen also appeared to be consistent among wells and among sampling periods in the central open sedge meadow area (Figure 5.3B), and fluctuated

75 more in the boundary zones. However, differences between these areas were not statistically significant.

Although seasonal effects were not statistically significant in either the north or south fen (Table 5.1), alkalinity levels in most communities seemed to decrease towards autumn, and were initially lower in the spring (Figures 5.2 and 5.3). Temporal patterns in water conductivity paralleled those of alkalinity measures (Figures 5.2 and 5.3), suggesting that in this case, conductivity may have been a good predictor of alkalinity.

Measurements in water pH showed less variation among sampling periods and among samples within the entire study site. Ranges in mean values recorded throughout the study indicate circumneutral conditions in all plant communities (Table 5.2). Lowest individual readings were recorded in the upland community. In general, individual readings in the north fen were higher than in the south fen, i.e., often above 7.0, and approaching 8.0 in some cases.

Short-term monitoring of water levels in wells within different plant communities showed an obvious pattern of lower and more variable water levels in areas away from the fen center (Figure 5.4). Field observations confirmed that in the central communities, ground water produced a stable water table near the surface, and substrate was consistently saturated. Precipitation seemed to be transferred to direct run-off

Spatial patterns

As expected, alkalinity levels tended to decrease along the creek and towards the upland area where the soil substrate differed. However, gradients within the fen itself

76 Fen Area pH Range

South Fen Riparian woods 6.29 - 7.48 Scirpus central marl 6.43 - 7.40 Carex sedge meadow 6.36 - 7.46 Acorus stand 6.42-7.21 Salix shrub meadow 6.19-7.30 Upland boundary zone 6.40 - 7.59 Upland woods 5.73 - 7.22

North Fen Riparian woods/meadow boundary zone 6.47 - 7.69 Open meadow 6.41-7.53 Upland woods/meadow boundary zone 6.43 - 7.49

Table 5.2: Range of mean water pH recorded during the study. Values are means from sampling points within a community type for a particular sampling period.

were not as previously assumed. Along a transect extending through the fen center and covering the major plant communities (Figure 5.5), it was found that, contrary to previous assumptions, highest alkalinity levels were not found in the central marl area, but instead within the Acorus stand.

Throughout the study, mean readings in alkalinity from the central marl and sedge meadow areas of the fen varied between 261 to 342 mg CaCOs/L and 273 to 392 mg

CaCOs/L, respectively. In woodlands surrounding the fen, relatively high readings were recorded in the riparian woods (between 190 and 350 mg CaCOs/L), as compared to the upland woods (below 62.5 mg CaCOs/L). As mentioned above, the Acorus stand exhibited the highest alkalininy levels (between 403 and 649 mg CaCOs/L). These

77 10 5 - E 0 - -5 _ o -10 - A.. ' g 3 -15 . A ' (n #.. 6 -20 _ ♦. U) - ▼. -25 ...... T o -30 - 0) ...... ^ . ^ ■ ...... ** -36 _ œ V.. .. V -40 ** ■ ■ • • • ScApuscentral malt • • ■ • • Carex sedge meadow L_ -45 A Acorus stand -so - upland border * upland woods -55 - -60 1 1 1 . _ L _ 1 1 1 1 8/29 9/5 9/13 9/23 10/3 10/10 10/19 10/26 Date

Figure 5.4; Short-term monitoring of water level fluctuations in a few representative wells in the south fen. Asterisks (**) indicate maximum depth of monitoring well: actual water level is below this depth.

78 700 1200

w ater alkalinity 1100 600 CO water conductivity 1000

900 Ü 500 5> 800

400 700

600

300 500

400 200 300

200 100

Carex Sdrpus Carex Salix 100 meadow central marl meadow shrub meadow

0 20 40 60 8 0 100 120 140 160 Distance (meters)

Figure 5.5; Mean water alkalinity (closed circles) and conductivity (open circles) of water samples collected along a transect in the south fen which traverses the major plant communities.

79 observations are further clarified by the spatially explicit results fi-om semivariance

analyses.

Semivariance analyses produced significant (r^ > 0.200) semivariograms for all of

the data sets on water alkalinity and conductivity, except for the autumn alkalinity data

set. Overall, there was a high proportion of total model variance attributable to spatial

structure in all data sets. Analyses of entire data sets revealed that spatial structure

accounted for 91.2% of total variance among samples of alkalinity, and 92.6% of total

variance among samples of conductivity (Table 5.3).

Interpolated maps fi-om best-fit models of alkalinity indicated spatial gradients

from distinct areas of high alkalinity levels within the fen to low alkalinity levels towards

upland areas (Figures 5.6 and 5.7). High alkalinity readings were concentrated in two

locations: at the immediate west of the central marl zone (at approximate coordinates 90

m east, 50 m north) and about 120 meters northwest of the central marl (at approximate

coordinates 30 m east, 90 m north). Both areas of high alkalinity coincided with distinct

patches of Acorus calamus.

Although annual patterns were generally similar (Figure 5.6), higher alkalinity

readings and stronger spatial structure (=94.6%) were detected in 1994, compared to

1995 (=86.5%). Comparisons among maps constructed by season (Figure 5.7) support previous observations (Figure 5.2) that readings were highest in summer, and lowest in spring. Semivariogram models also indicate strongest spatial structure (=91.3%) in summer and weakest (=69.7%) in the autumn.

80 Nugget Total Relative Data set Model form Model fit variance variance Range structure (r^) (Co) (C + Co) (Ao) (m) C/(C + Co)

Alkalinity ail data linear/sill 0.395 1790 20350 106.8 0.912 1995 linear/sill 0.401 1710 12660 106.7 0.865 1994 linear/sill 0.380 1650 30830 106.9 0.946 spring linear/sill 0.286 1580 10440 110.9 0.849 summer linear/sill 0.434 2200 25190 106.8 0.913 autumn gaussian 0.194 3870 12780 124.6 0.697

Conductivity all data linear/sill 0.368 4400 59100 106.8 0.926 1995 linear/sill 0.393 4000 44740 106.8 0.910 1994 linear/sill 0.345 3400 87800 106.8 0.961 spring linear/sill 0.308 3300 41730 110.9 0.921 summer linear/sill 0.400 5100 80600 106.8 0.937 autumn linear/sill 0.266 8200 42500 106.7 0.807 pH all data exponential 0.154 0.0017 0.0050 24.8 0.660 1995 linear/sill 0.202 0.0018 0.0058 110.9 0.690 1994 linear/sill 0.349 0.0060 0.0280 96.9 0.786 spring linear/sill 0.227 0.0070 0.0270 110.9 0.741 summer linear/sill 0.109 0.00001 0.0109 11.9 0.999 autumn linear/sill 0.425 0.0001 0.0410 108.5 0.998

Table 5.3 Semivariogram model parameters for water chemistry based on complete data sets and selected subsets. Relative structural variance, as a measure of spatial dependence, is expressed as the proportion of total model variance (C+Co) represented by structural variance (C).

81 B

to L— 0) 0) £

!1

9

65 05 10512525 105 *165 meters east

Figure 5.6; annual spatial patterns in water alkalinity levels (in mg CaCOs/L) in a 190 X 105 m" area in the south fen based on (A) 1994 data only; (B) 1995 data only; and (C) 1994 and 1995 data combined. Maps generated by kriging using the best-fit semivariogram model. 82 300 94

74 o

B

o c 74

c/) 54 0) o -4—» 0 E

9 4

74

25 65 35 105 125 145 1655 185 meters east

Figure 5.7: seasonal spatial patterns in water alkalinity levels (in mg CaCOs/L ) in a 190 X 105 m' area in the south fen based on (A) spring data only; (B) summer data only; and (C) autumn data only. Maps generated by kriging using the best-fit semivariogram model. 83 Consistent with alkalinity patterns, conductivity readings decreased towards the

fen boundary, and were highest around Acorus stands (Figure 5.8). For all data sets of

conductivity, semivariance analysis indicated strong spatial structure, in each case

accounting for more than 80% of total variance among samples.

Model fit for the best fit semivariogram based on the entire data set for water pH

was not significant (r^ < 0.200). Stronger model fits and spatial structures were found

when annual data were analysed separately. The semivariogram based on the autumn

data set showed particularly strong model fit (r^ = 0.425) and spatial structure (=99.8%).

A kriged map based on the entire data set of water pH (Figure 5.9) show spatial

patterns which do not seem to correlate well with either alkalinity or conductivity

patterns. The map indicates possible gradients centered around the locations of high

alkalinity and conductivity. However, contrary to expectations, readings increase towards the upland area in the southwest comer. This does not support previous observations (Table 5.2) that pH readings are lower in the upland area.

5.4 Discussion

Bryan and Andreas (1988) found that in addition to sharing many species in common, Ohio’s strongly minerotrophic peatlands also exhibit similar water chemistry.

Mean values for water alkalinity and conductivity found in the present study of Betsch

Fen are comparable to data presented by Bryan and Andreas (1988). From several fens in western Ohio, they reported a mean alkalinity range of between 363 mg CaCOg/L and

627 mg CaCOg/L, and a mean conductivity range of between 519.0 and 762.7 nS/cm.

84 94

o C

ÔOo

E 900 34

5 2545 6565 105 125 145 165 185 meters east

Figure 5.8: spatial patterns in water conductivity (in pS/cm) in a 190 x 105 m" area in the south fen based on complete data set collected during the study. Map generated by kriging using the best-fit semivariogram model.

5 25 45 65 85 105 125 145 165 185 meters east

Figure 5.9: spatial patterns in water pH in a 190 x 105 m" area in the south fen based on complete data set collected during the study. Map generated by kriging using the best-fit semivariogram model.

85 They note, however, that overall mean alkalinity in these western Ohio fens was

considerably higher than in eastern Ohio fens (438 mg CaCOs/L and 122 mg CaCOs/L,

respectively).

From a study of Michigan fens, Schwintzer and Tomberlin (1982) presented a

mean alkalinity reading of 153 + 74 mg CaCOs/L. This lower mean value may be

attributed to the fact that the fens studied did not have areas of open marl deposition

which are characteristic of many western Ohio fens (Bryan and Andreas 1988). In

contrast, Wilde and Randall (1951, cited in Schwintzer and Tomberlin 1982) reported an

alkalinity range of between 144 and 332 mg CaCOs/L from a Wisconsin fen. Similar

values for water conductivity have been reported from other fens, both in Europe and in

North America (eg.. Slack et al. 1980, Schwintzer and Tomberlin 1982, Verhoeven et al.

1988, Wassen et al. 1989, Boeye et al. 1994).

Results from this study suggest that water alkalinity is more stable at the fen center

which presumably receives a continuous supply of alkaline groundwater. In contrast,

readings were most variable in boundary zones. Closer examination of hydrology,

including water level fluctuations in the boundary zone is required. As shown by short­

term monitoring, water table level within the fen itself is relatively constant. This is as

expected because water levels in a fen fluctuate the least among wetlands since it is

maintained primarily by continuous ground water discharge (NRG 1995).

Although analysis of variance did not reveal significant seasonal variations, both temporal and spatial analyses suggested that water alkalinity levels are higher during the summer months than during either spring or autumn. This temporal pattern may be

86 explained by (1) seasonal differences in precipitation and run-off and/or (2) increased biological activity during the growing season.

Lower alkalinity readings may be a result of the dilution of ground- and surface water by increased precipitation and run-off particularly during the spring. Biological activity may be a factor since alkalinity is affected by all chemical processes that yield or consume îT or OH', including photosynthesis and respiration. As alkalinity is associated with charge balance, the assimilation of ions such as NO 3', NH»"^ and HPO4" accompanying photosynthesis must be balanced by the uptake or release of H" or GHT, resulting in alkalinity change (Stumm and Morgan 1981). For example, photosynthetic

NO3 assimilation is accompanied by an increase in alkalinity; conversely, bacterial decomposition of organic matter to NO 3 decreases alkalinity. Other processes which may increase alkalinity include denitrification, sulfate reduction and CaC 0 3 dissolution, while alkalinity will decrease as a result of forward reactions of nitrification and sulfide oxidation (Stumm and Morgan 1981).

Bryan and Andreas (1988) presented a water pH range of between 7.45 and 7.88, while Stuckey and Denny (1981) have suggested that pH range in fens could be as high as 8.0 to 8.5. Values recorded in Betsch Fen were not that high; although there were individual readings which exceed pH 7.0, values were generally below 7.0. Values recorded in the present study are similar to those reported by Schwintzer and Tomberlin

(1982) in Michigan, and Slack et al. (1980) in Western Alberta.

Obviously, continued monitoring of Betsch Fen is required to understand long­ term temporal changes in water chemistry, as results presented here must be interpreted

87 within the context of this study. Events such as recent rainfall or drought, run-off during

rainy season, recent decay of organic material, or trampling of vegetation prior to

sampling are expected to affect variations in water chemistry (Bryan and Andreas 1988).

Overall, geostatistical analyses showed that there is strong spatial structure in the

water chemistry parameters measured. Spatial analyses of water alkalinity patterns

confirmed that there are gradients fi"om high to low alkalinity and conductivity as we

move from the center of the fen towards upland. However, maps also demonstrated

spatial patchiness and complexities in gradients. While it is generally assumed that water

in the central marl area of a fen is the most alkaline (Denny 1979, 1994), this was not the

case for this site.

The strong plant community effect suggests that water alkalinity does affect overall

plant distribution in the south fen. Similarly, the absence of distinct plant zonation in the

north fen corresponded with no significant differences in water alkalinity within the area

sampled. This brings about the question of how temporal shifts in water alkalinity,

suggested from this study, may affect plant establishment. If in fact boundaries of water

alkalinity are functional determinants of structural boundaries, then this may have

important implications to the process of fen delineation. Bridgham et al. (1996),

however, caution against inferring that alkalinity gradients (and correlated factors such

as pH and Ca^"^ concentration) are coincident with nutrient availability gradients. They point out that few studies have examined the distinction between pH/alkalinity and nutrient availability in controlling growth of peatland plants. Furthermore, the nutrients

88 that actually limit plant growth, as well as the role of hydrology in supplying these

nutrients, remain open to question (Bridgham et al. 1996).

Spatial patterns observed for water pH are somewhat difiBcult to explain, because

areas of high alkalinity do not exhibit correspondingly high pH values, as would be

expected. Because these areas are located away from the central marl, considered as the

source of high carbonate containing water, other explanations for alkalinity are possible.

Alkalinity can be interpreted in terms of specific substances only when the chemical composition of the water sample is known (APHA 1985). Other dissolved buffer components such as ammonia, organic bases, sulfides and phosphates may be titrated in an acidimétrie titration of alkalinity, although their concentrations are usually small

(Stumm and Morgan 1981). In particular, conditions of high alkalinity accompanied by relatively low pH may be encountered in waters with high iron content. It is known that for some organic soils high in iron content, submergence does not always increase pH

(Fennessy and Mitsch 1989, Mitsch and Gosselink 1993). The possibility of perhaps microtopographical pockets of iron concentration in this site is supported by qualitative observations that water samples were often rust colored, indicative of possible metal content. In contrast, water sampled from the central marl and sedge meadow zones were clear.

Factors affecting the congruency between high alkalinity patches with Acorns stands require further investigation, although several hypotheses may be suggested. One hypothesis can be related to the above speculation that perhaps sites of Acorus establishment coincide with substrate that contains relatively high iron concentrations. If

89 this is tme, toxicity of high iron content has been found to affect the growth and distribution of various wetland plant taxa, and also determine the species composition of herbaceous fen vegetation (see Snowden and Wheeler 1993).

In a study examining the toxicity of ferrous iron to seedlings of several fen plant species, Snowden and Wheeler (1993) found that iron tolerance was negatively correlated with titratable water alkalinity. They also found that monocotyledonous wetland species were more tolerant of iron than were dicotyledonous species, and discussed the various adaptations which might explain this. Acorus calamus, a monocotyledon, is known as a vigorous and competitive invader that is very efficient in allocating nitrogen and maintaining high carbohydrate content throughout the year

(Weber and Brandie 1994).

Other hypotheses may be based on hydrological and topographical reasons: sites may coincide with hummocks or hollows which collect high alkalinity water, or there may be multiple springs, as suggested by TNC (unpublished files). These possibilities, however, do not explain the relatively low pH suggested by individual measurements and resultant maps. They also do not explain the high fluctuation in readings, which contrasts with the stability of readings in the central springs of the fen.

In summary, results of this part of the study indicate that there is a relationship between water alkalinity levels and plant community distribution, although spatial patterns may be more complex than previously assumed. This supports the assumption that gradients in water alkalinity is important in determining the limits of typical fen species establishment. However, it is important to realize that the distributions of

90 species is governed by their individual tolerances to different environmental gradients, and that there may be multiple limiting gradients (i.e., in alkalinity, plant nutrient availability, nutrient mineralization, hydrology, decomposition) which do not necessarily coincide (Bridgham et al. 1996).

91 CHAPTER 6

SOILS

6.1 Background and objectives

Wetland soils are the primary storage compartment of available nutrients for most

wetland plants, and are the media in which many wetland chemical transformations take

place (Mitsch and Gosselink 1993). These soils, referred to as hydric soils, are defined as

soils that are saturated, flooded, or ponded long enough during the growing season to

develop anaerobic conditions in the upper layers (US Soil Conservation Service 1987,

Mitsch and Gosselink 1993). As a result of reducing conditions associated with prolonged

inundation and/or soil saturation, characteristic soil properties typically develop (NRC

1995).

Definitions and descriptions of hydric soils have been very important in the

identification of wetlands, as the presence or absence of these soils determines whether a substrate meets the criteria for a site to be considered wetland (NRC 1995). In general, wetland soils may be either (1) mineral soils, or (2) organic soils (called Histosols). Both soil types contain organic material; however, a soil is considered a mineral soil when it contains less than 20 to 35 percent organic matter on a dry weight basis (Mitsch and

Gosselink 1993).

92 Organic matter tends to accumulate in wetlands as a result of the imbalance

between primary production and decomposition (Mausbach and Richardson 1994), which

is slowed as a result of anaerobic conditions. In organic soils, including soils commonly

termed peat and muck, two important characteristics are the botanical origin of the

organic material, and the degree to which it is decomposed (Clymo 1983).

Soil organic carbon content is frequently used to estimate soil organic matter

because carbon is the principal element of soil organic matter and can be readily measured

quantitatively (Allison 1965, Nelson and Sommers 1982). However, since there is

considerable variation in the ratio of carbon to organic matter among different soils, and

also among horizons of the same soil, it is preferable to simply report organic carbon as

such (Allison 1965, Nelson and Sommers 1982).

Inorganic soil carbon can be of either primary (parent material) or secondary

(pedogenic) origin, and commonly occurs as the carbonate minerals calcite (CaCOs), dolomite [CaMg(C 0 3)2] and magnesian calcites (Cai-xMg^COs) (Goh et al. 1993). Soil carbonates have been shown to impact root and water movement, soil pH (Nelson 1982), and exchange complexes (St. Arnaud and Herbillon 1973). Quantitative determination of soil carbonates is also useful, for example, in understanding micronutrient and phosphorus sorption (Goh et al. 1993). 1 hypothesized that carbonate content might be important in the present study because of the base-rich substrate underlying the site, and the resulting high alkaline conditions. The topography and surficial geology of fens allows the accumulation of base-rich water, and fen peats become base-rich as a result of continual flushing by ground water (Fitter and Hay 1987).

93 Soil pH significantly afifects the solubility of mineral nutrients and their uptake by

plants. In general, plants in high pH soils must adapt to certain nutrient deficiencies

(notably P), while plants in low pH soils must contend with toxicities (Fitter and Hay

1987). Soils formed over parent materials containing high proportions of CaCOs (e.g.,

limestones, chalks, and other glacial deposits) tend to have pH values near or above 7.

These soils are well-buffered by a reservoir of CaCOs and only in extreme conditions will

the pH fall below 5 (Fitter and Hay 1987). Circumneutral pH conditions are also a general

consequence of flooding previously drained soils: alkaline soils tend to decrease in pH

because of the buildup of CO 2 and carbonic acid (Mitsch and Gosselink 1993).

Soils that are fi'equently or chronically flooded display a number of characteristics

which indicate waterlogging or anaerobic soil conditions. These indicators include the

presence of certain soil colors, as well as chemical deposits associated with reactions that

occur in the absence of oxygen (Lyon 1993). For example, flooded soils often develop a gray, greenish, or blue-gray color as a result of a process known as gleying or gleization,

associated with the chemical reduction of iron (Mitsch and Gosselink 1993). Another

characteristic is the formation of mottles, spots of different color shades interspersed within the dominant matrix color in a soil layer. Orange/reddish-brown mottles (due to iron) or dark reddish-brown/black mottles (due to manganese) found in an otherwise gray

(gleyed) soil matrix suggest soils that are intermittently exposed and contain iron and manganese oxides in a reduced environment (Mitsch and Gosselink 1993).

94 The objective of this part of the study was to obtain baseline information on spatial variability in soil physicochemical characteristics, particularly soil organic carbon, carbonate content, and pH, within different plant communities in Betsch Fen. In accordance with the main objectives of the project, it is hoped that these data will add insight into the location of ecosystem or community boundaries. As was the case with water chemical characterization, this preliminary study was not intended to be a comprehensive investigation of soils at the site.

6.2 Methods

Field sampling

To acquire an understanding of how soil organic carbon, carbonate content and pH varied among the major plant communities in the south fen, samples were collected along a transect of water sampling wells which traversed the major communities. To allow comparisons with water chemistry, sampling was conducted along the same transect presented ni Chapter 5 (Figure 5.5). Three soil cores, 2.0 cm in diameter and 20.0 cm in depth, were removed from around each water sampling well. In addition to the eleven established sampling points, a point was added within the Salix shrub meadow, a zone which did not contain a water sampling well.

In order to observe soil variability at smaller spatial intervals, soil samples were also collected at every two meters along a transect which had been quantitatively sampled for vegetation (transect code SB, discussed later in chapter 7). This transect was selected because it extended through the major plant communities close to the open marl zone, but

95 did not run through the open marl proper. Therefore disturbance to the open marl by

additional soil sampling was avoided.

In the north fen, three soil samples were collected from around each water

sampling well along a transect which extended from the riparian woods-open meadow

boundary zone to the open meadow-upland woods boundary zone.

In addition to quantitative sampling, qualitative observations on soils were also

recorded throughout this study. These include general verification of soil survey maps, and

the determination of soil color using the Munsell Soil Color Charts (Kollmorgen

Corporation 1975), a useful, qualitative tool to estimate color characteristics of soil (Lyon

1993). By matching the color of field sampled soils to colors on these charts, one can determine a soil’s hue (spectral color), value (color depth or intensity), and chroma (color strength or darkness). These color characteristics provide information about a soil’s wetness, i.e. whether it is a hydric or nonhydric soil (Environmental Laboratory 1987,

Mitsch and Gosselink 1993). Soil color chromas of 1 and 2 are generally indicative of wet soils and hydric conditions (Lyon 1993), and include soils that are black, various shades of gray, and dark shades of brown and red.

Sample analysis

Soil samples were air-dried, then sieved through a 5 mm mesh screen. Organic carbon content of soils was determined by Walkley-Black oxidation (Allison 1965), and carbonate carbon content was estimated through the approximate gravimetric method

(Allison and Moodie 1965; Raad 1978, cited in Goh et al. 1993). Soil pH was determined

96 by measuring the pH of 10 g of mineral soil, or 2 g o f organic soil, in a 20 mL solution of

0.01 M CaCl2 (Hendershot et al. 1993).

Data analysis

Data on soil organic carbon, carbonate content and pH were analysed separately to examine the effects of plant community by one-way analysis of variance (ANOVA) using

SigmaStat™ (Jandel Corporation 1994a). Where significant plant community effects were detected, pairwise multiple comparisons were conducted using the Student-Newman-

Keuls method to determine which pairs of communities were different from each other.

Mean readings along the water-sampling transect, as well as individual data points along the supplemental transect, were plotted against distance to observe any spatial patterns or gradients.

6.3 Results

Along a water-sampling transect which extended through the central open marl zone in the south fen, mean organic carbon content varied between 1.6 % in the upland boundary zone and 14.3 % in the central marl zone (Figure 6.1 A). In general, readings along the transect exhibited the expected trend of higher organic carbon within the open fen than in the upland area. This was also observed in the additional vegetation transect, although the more extensive sampling along that transect revealed greater spatial variability within zones (Figure 6.2A). In the latter transect, the range of carbon content

97 Figure 6.1 : Mean readings for (A) soil organic carbon, (B) soil carbonate content, and (C) soil pH of samples collected along a water-sampling transect in the south fen which traverses the major plant communities. Vertical bars represent standard errors of the mean.

98 20 A. 19 18 17 16 15 c 14 o 13 ■s 12 g 11 10 9 8, 8 7 o 6 03 5 4 3 2 1

40 B 36

IS

JQ

14 c. 13 12

11 10 9 Wo 8 7 6 S 4 3 2 hçMltn A conjs tftnjb mmdow 1 woods

140 150 160 170 Distance (meters) Figure 6. l 99 Figure 6.2; Readings for (A) soil organic carbon, (B) soil carbonate content, and (C) soil pH from samples collected at every two meters along a vegetation transect in the south fen.

100 O DO Soil pH Soil carbonate content (%) Soil organic carbon (%) -A -A ^ hJ N) •n r\> ^ o> 0» o N> ^ roAO)oooro^o)oooK> I o\ k)

o

9, 0) 3 8 3

Ô) was slightly higher; between 3.3% in the upland boundary zone and 19.1% in the sedge meadow zone.

Overall, there was a highly significant difference in soil organic carbon among communities (p<0.0001. Table 6.1). Pairwise multiple comparisons revealed that there was a significant difference (p<0.05) between most pairs of communities, although there were also cases where no difference was detected. These were generally found when comparing communities that were located close to each other on the topographic gradient.

For example, there were no differences between organic carbon in the central marl zone and the Carex sedge meadow zone, or between Salix shrub meadow and upland.

Soil carbonate content along both transects in the south fen generally varied between 1.9% and 6.2%, except for one sampling point in the central marl zone which exhibited very high readings (Figures 6.IB and 6.2B). Measurements from three samples taken at this particular point ranged fi-om 17.6% to 34.9%, producing a mean of 27.4%.

Other than this obvious high reading, measurements around other water-sampling wells suggested fairly constant readings with no observable spatial pattern or gradient (Figure

6. IB). Greater variability was again demonstrated through more extensive sampling along a transect (Figure 6.2B). Soil pH was generally circumneutral along both transects

(Figures 6.1C and 6.2C): readings varied only between 5.76 and 7.01, except for one low pH mean around a single, but different, water sampling well in the central marl zone

(Figure 6.1C). There were no significant differences among communities in either soil carbonate content or pH (Table 6.1 ).

102 Source DF SS MS F Value P

Soil organic carbon Community 5 629.3 125.85 17.2 <0.0001 Residual 28 204.6 7.31 Total 33 833.9

Soil carbonate content Community 5 308.9 61.80 1.14 0.3632 Residual 28 1518.5 54.20 Total 33 1827.3

Soil pH Community 4 6.61 1.653 1.77 0.1669 Residual 25 23.38 0.935 Total 29 29.99

Table 6.1: Results of one-way analysis of variance for soil organic carbon. carbonate content and pH in the south fen.

103 Soil chemical properties at the three sampling points in the north fen paralleled those in the south fen; both soil organic carbon and carbonate content decreased towards the upland boundary zone, while soil pH was circumneutral along the transect (Figure

6.3). Soil organic carbon varied between 5.6% and 12.7%, while carbonate content varied between 2.7% and 6.7%. Higher minimum values for these two measures as compared to the south fen may be attributed to the fact that sampling towards the upland area was conducted within the open meadow-upland woods boundary zone, and thus soils were not representative of actual upland areas.

Among the three areas sampled in the north fen, there were significant differences in soil organic carbon (p=0.0005) and soil pH (p=0.0020) (Table 6.2). Soil organic carbon in the open meadow-upland woods boundary zone was significantly different (p<0.05) from both the open meadow and the open meadow-riparian woods boundary zone.

Despite the generally circunmeutral soil pH readings, pairwise multiple comparisons revealed significant differences (p<0.05) among all three areas sampled. There were no significant differences, however, in soil carbonate content among sampling areas.

Field reconnaissance was conducted to verify draft soil survey maps from the unpublished Ross County soil survey (Gillmore et al. in press). Soils in the south fen fit the descriptions of Carlisle muck, while Adrian muck was found in the north fen (Figure

6.4). Both of these soils have a slope range of 0 to 2 percent and are very poorly drained, with seasonable high water table depths at near or above the surface. However, Adrian muck has more rapid permeability in the substratum (ODNR DSWC 1993). During this

104 Figure 6.3: Mean readings for (A) soil organic carbon, (B) soil carbonate content, and (C) soil pH of samples collected along a water-sampling transect in the north fen. Vertical bars represent standard errors of the mean.

105 o CD Soil pH Soil carbonate content (%) Soil organic carbon (%)

T) -» ro cj cn O) 00 CO o Î ON U) o

o

o D I O O n § "3 ê (nI

8

8

O Source DF SS MS F Value P

Soil organic carbon Community 2 87.48 43.74 35.5 0.0005 Residual 6 7.40 1.23 Total 8 94.88

Soil carbonate content Community 2 15.96 7.98 4.89 0.1136 Residual 3 4.89 1.63 Total 5 20.85

Soil pH Community 2 1.146 0.5731 20.8 0.0020 Residual 6 0.165 0.0275 Total 8 1.311

Table 6.2; Results of one-way analysis of variance for soil organic carbon, carbonate content and pH in the north fen.

107 Figure 6.4: General boundaries of major soil types in Betsch Fen based on draft maps of the unpublished soil survey of Ross County, and field observations during the present study. Aa=Adrian muck, Cd=Carlisle muck, KaB=Kendallville silt loam 2-6% slopes, KeC2=Kendallville-Eldean complex 6-12% slopes, RdE2=Rodman gravelly loam, 20-35% slopes.

108 study, soil conditions in both open fen areas were generally saturated, often with some standing water present.

Muck soil is characterized by highly decomposed organic matter in which original plant material is unrecognizable, and thus differ from peat soils in which the identification of constituents is still possible (Allaby 1994). Indeed, samples collected from central areas of the fen did not contain much peaty material and thus may be considered saprists, i.e. soils in which two-thirds or more of the material is decomposed (Mitsch and Gosselink

1993).

Distinctly different types of soils were found outside of the wetland boundary zone. Soils underlying woodland vegetation bordering the south of the southern fen were classified as Rodman gravelly loam (Figure 6.4), an excessively drained soil with rapid permeability (ODNR DSWC 1993). This soil is typically found in upland areas, with a slope range of 12 to 35 percent. Kendallville and Eldean soils, which border the north fen, are also well drained soils, but have moderate slow permeability. Transitional areas were found within the wetland-upland boundary zone: these were characterized by mostly non­ fen vegetation or dense canopy on relatively wet soil.

Low soil color chromas were found along transects in both the south and north fen

(Table 6.3). In general, black or dark gray soils were found in the central areas of the fen, particularly in the open sedge meadow and central marl zones. Soil color was less consistent in the Salix shrub meadow and Acorus stands, while higher chromas were recorded as soils approach upland areas. Other hydric soil characteristics, such as gleying.

109 Sampling Transect Vegetation Soil color point location (m) zone

South Fen I 5 Riparian woods 5YR 3/1 (very dark gray) 2 23 Carex sedge meadow 2.5YR 2.5/0 (black) 3 34 Carex sedge meadow 5YR 2.5/1 (black) 4 55 Scirpus central marl lOR 2.5/1 (reddish black) 5 58 Scirpus central marl 5YR 2.5/1 (black) 6 61 Scirpus central marl 5YR 4/1 (dark gray) 7 68 Scirpus central marl 5YR 2.5/1 (black) 8 98 Acorus stand 5YR 4/1 (dark gray) 9 107 Acorus stand lOYR 6/3 (pale brown) 10 117 Salix shmb meadow lOYR 5/1 (gray) 11 159 Upland woods lOYR 5/3 (brown) 12 169 Upland woods 5YR 6/3 (light reddish brown)

North Fen 1 0 Riparian woods-open 5 YR 2.5/1 (black) meadow boundary zone 2 20 Open sedge meadow 5 YR 2.5/1 (black)

3 60 Open meadow-upland 5YR 4/1 (dark gray) woods boundary zone

Table 6.3: Qualitative soil color characteristics along water-sampling transects from both south and north Betsch Fen, as determined through use of the Munsell Soil Color Charts (KoUmorgen Corporation 1975).

110 mottling, and oxidized rhizospheres, were found in various soil samples. For example,

orange mottles were observed in several samples collected from the Acorus stand.

6.4 Discussion

Wetland mineral soils are different from organic soils in several physiochemical

features (Mitsch and Gosselink 1993). Based on the percentage of organic carbon, soils

containing less than 12 to 2 0% organic carbon are considered mineral soils, while those

containing greater than 12 to 20% organic carbon can be considered organic soils (Mitsch

and Gosselink 1993). If only the minimum percentage of 12% is considered in

differentiating these two soil groups, data from both transects in south Betsch Fen suggest

that organic soils are found in the open sedge meadow zones, the central marl zone, and the Acorus stands. By the same standard, soils in the riparian woods, Salix shrub meadow, and uplands woods would be considered mineral soils.

Quantitative sampling of organic soil carbon thus confirms to some extent the organic nature of soils in Betsch Fen. However, other physicochemical features must also be considered. The distinction of hydric soils into organic and mineral soils becomes particularly diflBcult in boundary zones and in vegetation zones which may not be typical of particular wetland types. For example, the Salix shrub meadow and Acorus stands in the south fen are mapped as organic, Carlisle muck soil in the draft copy of the Ross

County soil survey (Gillmore et al. in press), yet individual samples taken from these areas show mineral soil characteristics. There are indeed limitations to the level of detail and

111 scale in soil survey maps (Lyon 1993, NRC 1995). Surveys are usually published at scales

of 1:15,840 or 1:20,000, with a minimum map unit size of 2-3 acres (0.08-0.12 ha) (NRC

1995). Each soil map unit may have up to 25% inclusions of other soils. While the scales

of soil survey maps may be adequate for most agricultural uses, they do not provide

enough detail for the delineation of boundaries or landscape features (NRC 1995).

While organic carbon content generally increased towards the center of Betsch

Fen, a comparison of two transects in the south fen showed higher organic carbon along

the transect which extended through larger areas of open meadow and which did not

extend through the open marl zone. This would be expected, as marl areas support less

vegetation biomass and consequently accumulate less organic matter (c f Bernard et al.

1983). In the north fen, my sampling also suggests that organic soil is found in the central

sedge meadow zone, while more mineral soils are found towards the riparian and upland

woods.

In a study of Byron-Bergen swamp, a rich fen in western New York, Bernard et al.

(1983) reported mean soil carbonate readings of 4.7% under shrubs and trees, 6.1% in the sedge meadow, and 8.2% in marl pools. These means all fall within the range of values recorded in the present study, although in general, readings in Betsch Fen were slightly lower. Estimation of carbonate content can be affected by many factors, including the methods used for analysis (Goh et al. 1993). The method used in this study, for example, is a gravimetric method which only provides approximate measures: the accuracy of results depends upon the accuracy of weighing and the degree to which CO 2 retained in the analysing solution is compensated for by loss of water vapor (Goh et al. 1993).

112 Soils and overlying waters of wetlands occur over a wide range of pH. Organic soils, particularly those which do not receive significant groundwater inflow, are often acidic, while mineral soils often have more neutral or alkaline conditions (Mitsch and

Gosselink 1993).

As Betsch Fen is clearly an alkaline fen which receives groundwater inflow, neutral or basic pH was to be expected in the central areas. However, measurements along transects showed circumneutral conditions throughout the study area. The absence of significant pH gradients in both soil and interstitial water (Chapter 5) suggested that this environmental factor is not a good indicator of boundary locations at this site. Similar pH values have also been found in other fens and peatlands (e.g., Gorham 1956, Bares and

Wali 1979, Bernard et al. 1983).

Qualitative observations of soil color characteristics were helpful in describing differences among soils underlying different vegetation zones. Again, I found the distinction between organic and mineral soils was quite obvious when considering the extremes (i.e. central marl zone vs. upland woods), but less clear in intermediate or transitional areas. In both the south and north fen, soil color and appearance in the riparian woods were more similar to central fen soils than to upland woods soil, despite the two forest types having similar vegetation. This may be attributed to the topography of the riparian zones which receive some run-off from the fen as well as overflowing from the creek. These inputs, however, do not generally produce standing water, and thus conditions may be more favorable to the establishment of certain woodland species than is the case in the fen proper.

113 Hydric soil characteristics, such as mottling and gleying, were noted in various samples but did not provide significant information. In some hydric soils, mottles may not be visible due to masking by organic matter (Hurt and Brown 1995). Furthermore, both gleization and mottle formation are strongly influenced by the action of certain microorganisms. Under certain conditions, these two processes will not proceed, even though the soil has been saturated for prolonged periods (Diers and Anderson 1984,

FICWD 1989).

In summary, physicochemical analysis of soils in Betsch Fen reveal spatial variability among soils underlying different vegetation zones. However, while differences are obvious when comparing the extremes of the topographic and vegetation gradient (i.e. between the central marl zone and upland woods), it is more difficult to characterize intermediate areas. Analysis of soil organic carbon, carbonate and pH suggest some patchiness, but no clear gradients. The absence of any significant gradients, or obvious boundaries, makes it difficult to relate soil characteristics to vegetation patterns, and therefore yield little insight into system boundaries.

114 CHAPTER?

BOUNDARY DETERMINATION

7.1 Background and objectives

The renewed interest in landscape boundaries has reinforced the need for reliable

boundary detection methods (Chapter 2). Boundary locations have become important not

only because of their perceived ecological signiJBcance, but also because of their potential

regulatory implications, as in the case of determining legal boundaries to wetlands. While

the fields o f both vegetation analysis and landscape ecology have developed quantitative

methods related to boundary detection, the application of these methods to the detection

of wetland boundaries has only recently been recognized (NRC 1995).

Quantitative techniques have been developed to determine boundaries by locating

and characterizing ecological discontinuities along one-dimensional transects (Webster and

Wong 1969, Ludwig and Cornelius 1987, Wierenga et al. 1987, Johnston et al. 1992), as

well as fi"om two-dimensional data (Johnston et al. 1992, Fortin 1994). The use of

gradient-oriented transects in collecting one-dimensional data has been shown to be an

eflBcient and statistically sound sampling method (Ludwig and Cornelius 1987), and has also been recommended by the principal US federal wetland delineation manuals (NRC

1995).

115 When considering a two-dimensional landscape, transect data can show edge locations at points where the transect crosses the perimeter of a patch. As a landscape patch can be viewed as a set of edges, edge locations can theoretically be extended to two-dimensions (Brunt and Conley 1990). Therefore, advancements in the identification and interpretation of edges will have subsequent use in addressing questions about patches and landscapes (Brunt and Conley 1990).

Several studies have investigated the reliability of boundary detection algorithms based on simulated data (e.g.. Brunt and Conley 1990, Fortin 1994), while others have used field data to test the effectiveness of a specific detection method (e.g., Ludwig and

Cornelius 1987, Wierenga et al. 1987). However, few studies, if any, have actually applied and compared different approaches to actual field data.

The following chapter examines different approaches to landscape boundary determination by applying them to quantitative field data collected from Betsch Fen. Three methods will be examined: ( 1 ) gradient analysis by detrended correspondence analysis

(Hill and Gauch 1980), (2) the moving split-window technique (Johnston et al. 1992), and

(3) the federal method of wetland delineation (Environmental Laboratory 1987).

Methods 1 and 2 represent approaches by vegetation analysis and landscape ecological analysis, respectively. Although the emphasis will be on the application of these methods in detecting discontinuities in vegetation, the relationships among vegetation, water and substrate will also be examined. Both methods were applied to locate boundaries between vegetation zones in general, i.e., both within the fen, and between the fen and upland area. Locations of discontinuities were then used to test whether these two

116 methods confirm boundary locations previously hypothesized through field observations

(chapter 4).

As it is hoped that this study can have future implications to outer wetland

boundary determinations, comparisons will later be made to results fi'om method 3, which

is the regulatory approach specifically aimed at delimiting wetlands. Method 3 followed

procedures outlined by the currently used federal method for wetland delineation (Mitsch

and Gosselink 1993), i.e. the 1987 US Army Corps of Engineers technical manual

(Environmental Laboratory 1987).

This chapter will present results from these three methods separately, while comparisons among them will be discussed in the next chapter.

7.2 Methods

Method 1: Gradient analysis by Detrended Correspondence Analysis (DCA)

Data collection

Vegetation:

Vegetation was sampled in the field along belt transects consisting of contiguous

1.0 X 0.5 m^ quadrat plots. Six parallel transects (coded SA to SF) running northeast to southwest were established across six different areas of the south fen. Transect SA cut through a steel post close to the open marl zone, and subsequent transects were placed at

20 meter intervals parallel to SA (with the exception of a 10 meter interval between SD and SE). Transects SB, SC, SD and SE ran parallel to the west of SA, while SF ran parallel to the east of SA. Placement of transects is mapped in Appendix C.

117 The number of quadrats in transects SA, SB, SC, SD, SE and SF were 8 6, 104,

128, 92, 134 and 49 respectively, for a total of 593 samples. All transects covered the fen meadow zone, and extended from riparian woods boundary zone towards upland woods boundary zone. In each transect, quadrats were numbered beginning from sample plot 1 at the riparian woods boundary zone, and increasing as the transect moves towards upland woods. As this study aimed to locate the external boundary lines of the fen, the change of vegetation from fen meadow to upland woods was considered more important than the change from riparian woods to fen meadow. Thus transects did not include riparian woods vegetation.

In the north fen, two parallel transects (coded NA and NB) were placed 32 meters apart, and ran southwest to northeast, also extending from riparian woods to upland boundary zone. The number of quadrats in transects NA and NB were 85 and 72 respectively, for a total of 157 samples.

In each quadrat, vegetation frequency and cover were recorded using the Braun-

Blanquet scale, a semiquantitative method which gives a combined estimate of abundance and percent cover (Braun-Blanquet 1964, cited in Bonham 1989). For the purposes of data analysis, ranges of cover percentage were later converted to absolute values (Table

7.1).

Water and soil chemistry:

Transect SB was selected for closer examination to understand the relationships between vegetation and water alkalinity, as well as soil carbonate content and pH.

118 Rating Number of plants Area occupied Conversion value by a species for data analysis

+ sparse or very sparse very small 1% I plentiful small 5% 2 very numerous 10-25 % 18% 3 any number 25-50 % 38% 4 any ntimber 50-75 % 63% 5 any niunber > 75 % 88%

Table 7 .1 : Braun-Blanquet scale used for field data collection, and conversion values used for data analysis in this study. Conversion values were chosen arbitrarily for ratings + and I, and by taking the midpoint of the percentage range for ratings 2 to 5.

Because water sampling wells had been placed in a stratified random design along

transects and across communities, it was not possible to obtain water chemistry data at

regular increments along a vegetation transect. Placing wells this way would have made it unfeasible to sample and analyse routinely, and would also have increased disturbance to the site. To overcome this problem, values for water chemistry parameters along transect

SB were extrapolated fi’om contour maps produced through semivariance analysis

(chapter 5). Collection and analysis of soil samples along transect SB were as described in

Chapter 6.

Data analysis

Vegetation data fi’om each transect were analysed by detrended correspondence analysis (DCA) using PC-ORD™ software (McCune and Meflford 1995). In each analysis, the option of downweighting rare species was selected, i.e. the abundances of species rarer

119 than one-fifth of the frequency of the commonest species were downweighted in

proportion to their frequency. As boundaries will be perceived as discontinuities in species

cover across sample plots within the transect, plot ordinations were conducted instead of

species ordinations.

To relate vegetation to both soil and water chemistry, data from transect SB was

further subjected to direct gradient analysis by canonical correspondence analysis (CCA)

(ter Braak 1986), an option within PC-ORD™. In this case, the environmental matrix was

constructed by arranging data on water alkalinity, soil carbonate content, and soil pH

along the transect in a second matrix which corresponded to the same sample units as

vegetation (McCune and Mefford 1995).

Method 2: Moving split-window analysis (MSW)

Data collection

Data used for this method of analysis were the same data used for Method 1

above. Procedures for data collection are as described above.

Data analysis

Discontinuities along transects were located and characterized by placing a double window, consisting of eight plots, over multivariate vegetation data points, and then statistically comparing the dissimilarity between attribute values in each window half

(Figure 7.1). The window was then moved sequentially by one plot along the transect until statistical comparisons have been computed for the entire transect length.

120 • Sample points

Figure 7.1: The moving split-window technique for analysis of one­ dimensional data, using a window width of eight sample points. The dissimilarity between window halves is calculated based on mean sample values in each half. The window is then moved sequentially by one plot along the transect, and calculation of dissimilarity is repeated until the entire transect is covered (reproduced from Johnston et al. 1992).

121 Dissimilarities between adjacent window halves were calculated using the squared

euclidean distance (SED) metric which is computed as the square of the difference

between the means of each variable in adjacent windows, summed across all variables

measured (see Johnston et al. 1992):

S E D m a — to a C^iAw ~ ^iB w Ÿ

where: A and B denote window halves, n is a station or midpoint between window halves,

w is window width, and a is the number of variables sampled at each station (Brunt and

Conley 1990, Turner et al. 1991). As the window is moved along the transect, a series of

values that represent successive differences between window halves is produced.

Discontinuities or boundary locations occur at maximum SED values, indicating that the

rate of attribute change is at a maximum. When SED values are plotted against station

location or transect length, tall and narrow peaks occur at abrupt boundaries, while lower

and wider peaks occur at more gradual changes.

Dissimilarity metrics were computed using Lotus 123™ (Lotus Development

Corporation 1993) spreadsheet software, and later exported to SigmaPlot™ (Jandel

Corporation 1994b) for graphic plotting.

Method 3: Federal method of wetland delineation

Data collection

Wetland delineation was conducted according to the following routine onsite determination method outlined in the US Corps of Engineers Wetlands Delineation

Manual (Environmental Laboratory 1987) for areas greater than five acres (two hectares):

122 A baseline was first established parallel to the major watercourse, in this case

Blackwater Creek, and perpendicular to the perceived hydrologie gradient. This baseline extended across the whole open fen area and measured 242 m. The length of the baseline was divided by the number of required transects (i.e., three), and transects were then extended perpendicular to the baseline, using the midpoint of each baseline increment as a starting point (Figure 7.2).

Representative locations which suggest distinct plant communities were selected as observation points along each transect. At each observation point, vegetation, soils and hydrology were characterized by completing a data form 1 fi'om the Manual (Figure 7.3).

The procedures can be summarized as follows:

I. Vegetation:

Dominant plant species were recorded from each vegetation layer in the immediate vicinity of the observation point, i.e., within a 5-ft (1.5 m) radius for herbs and shrubs, and a 30-fi (9.0 m) radius for trees and woody vines. Dominant species were subjectively determined by estimating those having the largest relative basal area (woody understory), greatest height (woody understory), greatest percentage of areal cover (herbaceous understory), and/or greatest number of stems (woody vines). Dominant species that have known morphological or physiological adaptations for occurrence in wetlands, as listed in

Appendix C, Section 3 of the Manual, were recorded.

By referring to the species list on Appendix C, Section 1 of the Manual (list for wetlands in region 1, i.e.. Northeast US) the wetland indicator status for each dominant species was recorded according to the criteria in Table 7.2. A particular observation point

123 NW

80 m

40 m

80 m

120 m

80 m

200 m

Baseline = 242 m

SE

C

Figure 7.2: Placement of transects (A, B and C) to conduct wetland delineation.

124 DATA FORM I WETLAND DETERMINATION

Applicant Application Project Name:______Number:______Name:___ State: County: Legal Description: Townshtp:^*^^ Range: Date: P lo t Wo.: ^ { T n jn s e c f 3 ) S e c tio n : ______

Vegetation [list the three dominant species in each vegetation layer (S if only 1 or 2 layers)]. Indicate species with observed morphological or known physiological adaptations with an asterisk Indicator Indicator Species Status Species Status Trees Herbs 1. £ ///n w rubra. FAC y^Jmpatieas Capcytîis 2. i.Z.eerS/a orYXoidei CBL. 3. 9 .Sym/iApco/fws fottidoi O&L. Saplings/shrubs Woody vines U. Acer n tq u n d o * FA C i- 10.Lonicera ‘/oponiccu FAC. 5. Cornus ttotoniftra. FAC^-r- l l . 6, 12. Z of species th a t a re OBL, FACW. and/or FAC:^f£_^ O ther in d ic a to rs :_ Hydrophytic vegetation: Yes No . Basis : cA an S c % o A i. . Fa CHf. F AC.

Soil Series and phase: ^ u c k On hydric s o ils l i s t ? Yes ^ ; No . M ottled: Yea _____ ; No . M ottle color: ______; M atrix c o lo r: S y A X . S " / / Cleyed: __Yes_____ No Ocher Indicators: ______. Hydric s o ils : Yes No _____ ; Basts: Indicatory peaent______.

Hydrology Inundated: Yes ; No . Depth of standing water: Saturated s o ils : Yes ; No _____ . Depth to saturated soil: Ocher in d ic a to rs: Wetland hydrology: Yes ^ ; No_____ . Basis: ^ndfcafprx prttM ^ ______Atypical s itu a tio n : Yes ; No ^ . Normal Circumstances? Yes ^ No Wetland Determination: __Wetland_____ ^ ; Nonuetland ______Comments:

Determined hv: B2

Figure 7.3: Example of Data Form 1 from the 1987 US Army Corps of Engineers Manual, completed during field observations for wetland determination.

125 Indicator category Code Definition

Obligate wetland OBL Plants that occur almost always (estimated probability >99%) in wetlands under natural conditions, but which may also occur rarely (estimated probability < 1 %) in nonwetlands.

Facultative wetland FACW Plants that occur usually (estimated probability of 67 to 99%) in wetlands, but also occur (estimated probability of 1 to 33%) in nonwetlands.

Facultative FAC Plants with a similar likelihood (estimated probability of 33 to 67%) of occurring in either wetlands or nonwetlands.

Facultative upland FACU Plants that occur sometimes (estimated probability of 1 to 33%) in wetlands, but occur more often (estimated probability of 67% to 99%) in nonwetlands.

Obligate upland UPL Plants that occur rarely (estimated probability <1%) in wetlands, but occur almost always (estimated probability of >99%) in nonwetlands under natural conditions.

Table 7.2: Plant indicator status categories, from the 1987 US Army Corps of Engineers Wetlands Delineation Manual (Environmental Laboratory 1987).

126 was considered to have hydrophytic vegetation if more than 50 percent of the dominant species have an indicator status of OBL, FACW, and or FAC, or if two or more dominant species have observed morphological or physiological adaptations to wetlands.

2. Soils;

Hydric soils could be assumed to be present in an observation point if either (1) all dominant plant species had an indicator status of OBL, or (2) all dominant plant species had an indicator of OBL and/or FACW, with at least one dominant species being OBL. If neither ( 1 ) nor ( 2) applied, but the vegetation qualified as hydrophytic, a soil core sample was collected and examined for hydric soil characteristics. Soil at the observation point was considered a hydric soil if a positive hydric soil indicator as listed in the Manual (e.g., definite organic soils, or characteristic soil colors) was present. Conversely, if no positive soil indicator was found, the observation point did not have hydric soil, and thus the area was not a wetland.

3. Hydrology:

Each observation point was examined for indicators of wetland hydrology as listed by the Manual, e.g. through visual observation of inundation, soil saturation or watermarks. The observation point was considered to have wetland hydrology if a positive wetland hydrology indicator was present. If no positive wetland hydrologie indicator was present, the area at the observation point was not a wetland.

127 Based on the above, in order for an observation point to be considered wetland, it

was necessary to find; (1) 50% or more of dominant plants having a high probability of

occurring in a wetland, (2) hydric soils, and (3) evidence of wetland hydrology. Failure to

meet one of these three criteria under "normal" or typical conditions implies that the

particular area was not a wetland according to jurisdictional definition. However, strong

evidence of hydrophytic vegetation and hydric soils may be used to indicate presence of

wetland hydrology.

Along a transect, the wetland-nonwetland boundary was determined within the

area where an observation point considered wetland is followed by an observation point

considered nonwetland (or vice versa). When such an area was detected, closer

examination was conducted to look for subtle changes in the plant community (e.g., the

first appearance of upland species, disappearance of apparent hydrology indicators, or

slight changes in topography). Where such features were noted, an additional observation

point was established, and all three parameters were again characterized by field

observation. If the area at this last observation point satisfied the wetland criteria, then the

transect was followed towards the nonwetland observation point until upland indicators

were more apparent.

Characterization procedures described above were repeated until a nonwetland

observation point was found, in which case it was required to move half-way back along

the transect toward the last documented wetland observation, and again repeat

observations. This procedure was continued until the wetland-nonwetland boundary was found. It was not necessary to complete a data form 1 for all intermediate points, although

128 a form was required for the wetland-nonwetland boundary location. The position of the wetland boundary was then marked on the base map, and characterization of observation points was continued along the remaining transects.

Data synthesis

Plant community types were marked on the base map based on all completed copies of data form I. Each community was identified as either a wetland or nonwetland, and observation points that represented their boundaries were marked. The outlined procedures called for these points to be connected on the map by generally following contour lines in order to separate wetland from nonwetland. Field observations needed to be conducted along the contour line between transects to confirm the boundary. In cases of anomalies, it would have been necessary to establish and characterize short transects in these areas to make necessary adjustments on the base map.

7.3 Results

Method 1: Gradient analysis by Detrended Correspondence Analysis (DCA)

DCA results varied among transects in the south fen, although there were similarities among transects (Figure 7.4). In most cases, it was difScult to separate sample plots into groups by subjective examination of the ordination space. Although spatial discontinuities between consecutive plot numbers could be detected through critical study of the ordination space and by considering actual score coordinates, clusters were generally not readily visible. Plot separations were relatively clear in certain transects (e.g..

129 Figure 7.4 A-F: DCA plot ordinations of vegetation data (percent cover by species) from six belt transects of contiguous 1.0 x 0.5 plots in the south fen. Numbers correspond to sequential plot numbers along the transect, beginning from plot 1 at the riparian woods boundary zone, and increasing as transect moves towards upland woods boundary zone.

130 A. Transect SA •85

200 •83 N •82 o 30 1 mc -S> 100 •66 •74 CM CO ■75 *71 Acorus stand

0 •64

0 100200 300 400 500 600 DCA AXIS 1 (Eigenvalue 0.873)

B. Transect SB •29 200

T f Scrpus sedge CM 0 #2g meadwv e 3 •28 CO •482 •97 1 •gg Sa/K shrub •88 "87 •98

•104 •79 •95 upland woods <

0 100 200 300 400 500 600 700 DCA AXIS 1 (Eigenvalue 0.901)

Figure 7.4 (to be continued) 131 Figure 7.4 (continued)

400 C. Transect SC

in % 300 o 3m CO Ê o 200 ÜL CM 1 3 Ss/»r shrub meadow CO ■66 *69 !3 T il6 100 Q

100 200 300 400 500 600 DCA AXIS 1 (Eigenvalue 0.627)

D. Transect SD

200

CM O •71 m Impatiens 3 plots •gQ Sa/pr shrub meadowdupland >CD oC g 100 (M CO

Q

400 DCA AXIS 1 (Eigenvalue 0.886) (to be continued) 132 Figure 7.4 (continued)

500 E. Transect SE

400 ■125 128 § in o 3(D 300 CD > ■130 OC

CM ■109 OT 100 Sa/K shrub ^ Camx sedge meadow meadow $ Q

0 100 200 300 400 500 600 DCA AXIS 1 (Eigenvalue 0.696)

F. Transect SF *19 *23 g d 100 Solaoga/lmpatKns 0 plots 3 *46 CD 1 *4519 *41 i *48 CM *28 ^6*31 ' co •29 *30 *27 47 -34 *13 *39 *38 meadowSa/ct shrub •11

100 200 300 DCA AXIS 1 (Eigenvalue 0.387)

133 in transects SA and SD) but nearly undetected in others (e.g., in transects SC and SE).

DCA was thus not equally successful in separating major vegetation zones along different

transects, and therefore in locating boundary lines between adjacent zones. Boundary lines

subjectively determined through previous field measurements and qualitative observations

(Chapter 4) were not entirely confirmed by DCA results (Table 7.3). In three transects,

DCA detected boundaries which were not detected during field observations (Table 7.3).

The eigenvalue, equal to the maximized dispersion of scores on an ordination axis,

is a measure of axis importance; values greater than 0.5 generally denote a significant

separation of points along the axis (ter Braak 1995). In transects SA to SE which

extended through major vegetation zones, high eigenvalues for axis 1 suggest the

importance of gradients represented by this axis. In contrast, much lower eigenvalues for

axis 2 suggest that these axes represent relatively insignificant gradients. Separation of

plot clusters along DCA axis I can generally be attributed to a hydrologie or moisture

gradient where sample plots at presumably lower elevations near riparian woods occurred

towards one end of the axis, while plots on drier areas approaching upland occurred at the

other end.

Transects SA SC, SD and SE share a general trend of exhibiting little or no

separation among plots along axis 2 at lower scores of axis 1, but clear separation at higher axis 1 scores. This suggests that gradients represented by axis 2 only affect plots at one end of axis 1. These plots generally corresponded to locations away from the fen center, and approaching upland. Description of results for individual transects are as follows:

134 Hypothesized boundary Support of hypothesis by Detection of boundary line locations quantitative analysis: not observed in field: DCA MSW DCA MSW

Transect SA 70 m {Carex/Acorus) no no no yes 77 tn (Acorus/Salix) no no

Transect SB 60 m (CarexlAcorus) yes yes no no 79 m (Acorus/Salix) yes yes 98 m (&z/f%/upland) yes yes

Transect SC 74 m (sedgdSalix) yes yes yes yes 102 m (S'a/fx/upland) no yes

Transect SD 74 m (Carex/Acorus) no yes yes yes 89 m (Acorus/Salix) no no

Transect SE 94 m (Carex/Salix) yes no no yes 101 m (Sa//j:/upland) no no

Transect SF 26 m (Carex/Salix) yes yes no no

Transect NA 63 m (Carex/upland) yes no no yes

Transect NB none yes yes

Table 7.3: Results of boundary detection by detrended correspondence analysis (DCA) and moving split-window technique (MSW) as compared to major vegetation zone boundary lines subjectively determined through field measurements and qualitative observations.

135 Transect SA:

In transect SA, plot number 55 was problematical in that it was found within the

Carex sedge meadow zone, but unlike all other plots in this zone, no Carex coverage was

recorded. This plot was dominated instead by Solidago canadensis. An initial DCA run

showed this plot as an outlier, while the rest of the plots were grouped together with no

separation among them. Because of this problem, plot 55 was removed from analysis, and

DCA was repeated. The second analysis (Figure 7.4A) separated the plots into three

distinct clusters: plots 1-63 {Carex/Scirpus sedge meadow), plots 64-80 {Acorns stand),

and plots 81-86 {Salix shrub meadow).

Transect SB:

Three discontinuities in sample plots along transect SB produced four plot

groupings: plots 1-60 {Carex/Scirpus sedge meadow), plots 61-80 {Acorus stand), plots

81-97 {Salix shrub meadow) and plots 98-104 (upland woods). Considering the sequence of plant zones along axis 1, this axis could be interpreted as a gradient of decreasing water alkalinity from left to right.

Unlike in transect SA, axis 2 produced some separation within the sedge meadow zone: plots dominated by Carex were found at the lower end of axis 2, and those dominated by Scirpus were found at the higher end of the axis. Axis 2 scores of between

80 and 140 contained an intergrading of plots dominated by either of these two sedge species. In the field, Scirpus was generally found in proximity to the open marl zone

136 characterized by shallow pools of exposed marl or by saturated substrate, while Carex was

found in more peaty and less saturated substrate.

Ordination for transect SB is thus suggestive of two gradients: a water alkalinity gradient along axis 1, and a substrate gradient along axis 2.

Transect SC:

Discontinuities in consecutive plot numbers along transect SC separated plots according to the major plant zones traversed, i.e. Carex sedge meadow (plots 1-74), Salix shrub meadow (plots 75-116) and upland woods (plots 117-128). Plots in the upland woods were separated into two clusters along DCA axis 2: plots 117-121 at the lower end of the axis were dominated in cover by Gleditsia triacanthos, while plots 122-128 were dominated by Acer negundo and Ulmus rubra.

Transect SD:

Plots in transect SD fell into four clusters. The largest part of the transect (plots I-

69) constituted Carex sedge meadow, while the remaining plots ran through a narrow strip 0 Î Acorus stand (plots 73-81) and subsequently into a Salix shrub meadow/upland woods intergradation (plots 82-92). Plots 70-72, which contained high coverage of

Impatiens capensis, were found within the boundary zone between Carex sedge meadow and Acorus stand.

137 Transect SE:

DCA results separated the major vegetation zones traversed by transect SE.

However, gradients were not easily interpretable. Plots within the Carex sedge meadow

(plots 1-94) were clearly clustered at the lower end of axis I, while an indistinct discontinuity separates these plots from Salix shrub meadow plots in which Impatiens capensis are often dominant (plots 95-117). An axis 2 score of about 200 then separates upland woods (plots 119-134) from all other plots. Plot number 118 was inconsistent within the sequence as it was found between the shrub meadow and upland woods, but was more similar in species composition to sedge meadow plots.

Transect SF:

This transect, consisting of only 49 quadrats, provided supplementary information on open meadow vegetation, but did not extend to upland boundary zone because of difficulty in field sampling. DCA was successful, however,in distinguishing between the

Carex sedge meadow zone (plots 1-17) and Salix shrub meadow (plots 26-49). Plots 18-

25 constituted a boundary zone of plots containing Carex in association with Solidago ohioensis and Impatiens capensis. The low eigenvalue for axis 1 (0.387) suggests the relative unimportance of any gradients underlying the distribution of species along this transect.

138 Interpretation of results from the north fen are more straightforward (Figure 7.5),

partly due to the absence of distinct vegetation zonations in this area:

Transect NA:

DCA results for transect NA is quite clear: as the open fen area is almost entirely dominated in cover by Carex, ordination clearly separated this open sedge meadow (plots

1-63) from surrounding upland (plots 64-85). A high eigenvalue (0.969) indicates the importance of axis 1, which can be interpreted as an elevational gradient corresponding to change from saturated substrate to drier, upland soil.

Transect NB:

DCA results separated transect NB into Carex sedge meadow (plots 1-67) and upland woods (plots 68-72). Within the sedge meadow zone, plots dominated by Salix cover (plots 15-28) were separated from the rest of the plots. It should be noted that despite the presence of plots containing Salix, a distinct shrub meadow zone comparable to that in the south fen was not found here. However, the transect did run through a cluster o f Salix shrubs found close to the riparian woods boundary zone. Therefore, axis 1 could still be interpreted as an elevational or hydrologie gradient from upland woods to riparian woods.

139 Figure 7.5 A and B: DCA plot ordinations of vegetation data (percent cover by species) from two belt transects of contiguous 1.0 x 0.5 plots in the north fen. Numbers correspond to sequential plot numbers along the transect, beginning from plot 1 at the riparian woods boundary zone, and increasing as transect moves towards upland woods boundary zone.

140 200 A. Transect NA

•25

:"23 •85 :% •2P

100 ^ • S é

CM

: Cara* sedge upland woods • * a c 7 meadow

•S3

0 100 200 300 500400 600 DCA AXIS 1 (Eigenvalue 0.969)

B. Transect NB

200 •24 R O O •70 •23 3 •72 •27 >CD ■19 Q)C ■1» .S> 100 •32 CM CO •3¥»o •35 *37

Cara* sedge *31 •28 meadow •30 *59 *29

0 •Ifr

0 100 200 300 DCA AXIS 1 (Eigenvalue 0.437)

Figure 7.5 141 Additional analysis by Canonical Correspondence Analysis (CCA)

CCA results from transect SB indicate that from the three environmental factors

tested, water alkalinity and soil pH were both important in determining plant species

distribution (Table 7.4). Based on the number of variables and samples tested, the critical

value for correlation coefBcients to be significant (at a = 0.01) is 0.327 (Snedecor 1956,

cited in Rohlf and Sokal 1969). Therefore, it could be inferred that in the CCA ordination

diagram (Figure 7.6), the first axis is a water alkalinity gradient (r^ = 0.998), while the

second axis is a soil pH gradient (r^ = -0.911), separating the plots into groups found in;

(1) areas with relatively low water alkalinity and moderate soil pH (plots 1-50), (2) areas

with moderate to high water alkalinity and low soil pH (plots 51-62 and 73-84), (3) areas

with moderate water alkalinity and moderate to high soil pH (plots 85-104), and (4) areas

with high water alkalinity and moderate soil pH (plots 63-72).

CoefBcients Correlations Variable Axis 1 Axis 2 Axis 3 Axis I Axis 2 Axis 3

Water alkalinity 0.822 0.245 0.037 0.998 0.013 0.063 Soil carbonate -0.053 0.054 0.338 -0.144 0.082 0.986 SoU pH 0.007 -0.596 0.031 0.397 -0.911 0.114

Table 7.4: Canonical coefficients and intraset correlations of water alkalinity, soil carbonate content and soil pH with the first three axes of CCA for transect SB.

142 2

•73 ■74

CO R 1 •76 •77 •62 •51 •78 •56 1» '40 O) •72 WATER ALKAUNITY 0 •32 •86 •80 -66 Salix shrub meadow/ •68 upland woods -H) Acorus stand •38

1 SO IL PH

1 0 1 2 CCA AXIS 1 (Eigenvalue 0.690)

Figure 7.6: CCA plot ordinations of vegetation data from transect SB, with environmental variables represented by arrows. Soil carbonate effects were not significant in the first two ordination axes.

143 Although the importance of gradients in water alkalinity and soil substrate support

general results obtained from DCA, the grouping of plots to suggest boundaiy line

locations revealed different results. CCA produced four plot groupings which did not

correspond entirely with plot groupings detected by DCA. Plots in the Carex/Scirpus

sedge meadow zone were scattered into two groups. A discontinuity suggested in plot no.

62 approximates the sedge meadowMcon/j stand boundary location in plot no. 60 as

suggested by DCA. Other discontinuities were detected in plots no. 72 (within the Acorus

stand) and no. 84 (within the Salix shrub meadow zone). Thus CCA did not separate field

vegetation zones observed in the field as clearly as DCA. Furthermore, none of the

hypothesized boundary line locations (Table 7.3) were confirmed.

Method 2: Moving split-window analysis (MSW)

Plots of squared euclidean distances (SED) as a function of transect length

revealed peaks at locations where the rate of change in vegetation cover was high, thus

denoting discontinuities in vegetation structure (Figure 7.7 and 7.8). As maximum SED

values are a function of the number of variables (species) sampled and the maximum

abundances per species for the sample unit employed (Turner et al. 1991), these values

varied among transects. SED values generated from transect data in both the south and

north fens fell within a range of 5 to 12,573. Based on comparisons of peaks among

split-window graphs, as well as recorded field observations, a minimum SED value of

3,500 was used as the criterion to indicate a significant peak or boundary location.

144 Figure 7.7 A-F: Peaks of squared euclidean distance (SED) determined by moving split-window analysis (window width=8) of percent vegetation cover data from six belt transects of contiguous 1.0 x 0.5 m^ plots in the south fen. Transects begin from plot 1 at the riparian woods boundary zone and moves towards upland woods boundary zone. Numbers above peaks indicate points which satisfy the minimum SED value used to determine a significant peak or boundary line location.

145 14000 13000 A. Transect SA 12000 (D ü 11000 jS 10000 Î2 Q 9000 (0c 8000 0) -g 7000 y 6000 3 LU 5000 ■o 4000 2 (03 3000 g 2000 CO 1000

10 20 3 0 4 0 50 6 0 70 80

14000 13000 B. Transect SB 12000 Ud) 11000 i5 10000 (A Q 9000 (0C 8000 ^ 7000 ^ 6000 3 LU 5000 4000 1 (D 3000 3 O ’ 2 0 0 0 CO 1000

llllllllllllllllllllllllirillllliirilllllllltlllllllllllllll 10 20 30 40 50 60 70 80 90 100 Transect (meters)

Figure 7.7 (to be continued)

146 Figure 7.7 (continued)

14000 13000 - C.TransectSC 12000 0) o 11000 f 10000 b 9 0 0 0 coc 8 0 0 0 7000 I 60 0 0 3o LU 50 0 0 40 0 0 30 0 0 g 2000 1000 1/

INI II III lllillll Illllllillllliriillilllliilillllllllillllllllllllllllllliliilirli IlliiliiliIllllilirilMlilillilriliilr 10 20 30 40 50 60 70 80 90 100 110 120 130

14000 - 13000 - D. Transect SD

12000 -

8 11000 - I 10000 - 9 0 0 0 - coc 80 0 0 - 0) ;g 70 0 0 - 75 6 0 0 0 - 5 0 0 0 - 4 0 0 0 - 1 8 30 0 0 - 2000 - S 1000 - 0 - 11 n 1 M t t m M {I I M I I 11 I I m M M 1111 I n 1111 t 111 M 11 I I I j t I M M 11

Transect (meters)

(to be continued)

147 Figure 7.7 (continued)

14000 13000 - E. Transect SE 12000 d) O 11000 jS 10000 (g Q 9000 (0C 8000 (D ;g 7000 "Ô 6000 3 U J 5000 4000 3000 2000 1000

imiiiiliiillllliliiiniilllmiiiiiilrliilmilliiiiiiiilllllllillluilliililmiiiiiiliiiiilllilirliiiinlin il iririririirii 10 20 30 40 50 60 70 80 90 100 110 120 130

14000 13000 - F. Transect S F 12000 8 11000 I 10000 b 9000 c 8000 CD 7000 I 6000 3 UJ 5000 T 3 4000 E CD 3000 O - 2000 1000

5 10 15 20 25 30 35 40 45 Transect (meters)

148 Figure 7.8 A and B: Peaks of squared euclidean distance (SED) determined by moving split-window analysis (window width=8) of percent vegetation cover data from two belt transects of contiguous 1.0 x 0.5 m^ plots in the north fen. Transects begin from plot 1 at the riparian woods boundary zone and moves towards upland woods boundary zone. Numbers above peaks indicate points which satisfy the minimum SED value used to determine a significant peak or boundary line location.

149 14000 13000 - A. Transect NA 12000 11000 10000 9000 8000 7000 6000 5000 4000 3000

1000

10 20 30 40 50 60 70 80

14000 13000 B. Transect NB 12000 11000 10000 9000 8000 7000 6000 5000 4000 3000

D " 2000 1000

10 20 30 40 50 60 70 Transect (meters)

Figure 7.8

150 In transects SA to SE which extended through the major vegetation zones, highest peaks generally corresponded to locations of boundary lines between these major zones.

However, there were also cases where peaks that appear significant in the split-window graphs correspond to only minor shifts in species composition which do not appear significant in the field. MSW confirmed hypothesized locations of certain boundary lines, but discrepancies were also noted (Table 7.3). In sbc of the eight transects analysed, MSW detected boundaries locations which were not observed during field surveys. Description of results for individual transects are as follows:

Transect SA:

Seven significant peaks defined seven discontinuities in vegetation structure.

However, not all peaks denote significant discontinuities between major vegetation zones.

Within the sedge meadow zone (0-63 m), peaks 1 to 5 denoted transect locations where dominance shifted between Carex and Scirpus. The wide bases of peaks 2 to 5 suggest relatively gradual changes in structure. The first shift between zones considered as major vegetation zones, i.e. peak 6 between Carex sedge meadow and Acorus stand, appears as a narrower yet lower peak. Finally, peak 7 denotes change from Acorus stand to Salix shrub meadow.

151 Transect SB:

Major vegetation zones were clearly detected by split-window analysis on transect

SB where peaks I, 2, 3 and 4 corresponded with transect locations where vegetation changes towards dominance by Scirpus, Acorus, Salix and woodland species, respectively.

Transect SC:

In transect SC, MSW was more sensitive at locating discontinuities within zones than between major zones. Although the transition from Carex sedge meadow to Salix shrub meadow is quite clear and abrupt in the field, it is only detected as a relatively low peak (peak 1 at 74 m). Peaks 2 and 3 within the Salix shrub meadow zone (74-II7 m) denote slight shifts in cover dominance within the shrub understory (by Eupatorium maculaium and Galium sp.) and canopy (by Camus amomum). Similarly, peak 5 within the upland vegetation ( II7-128 m), which is both tall and narrow, indicate shifts in species composition from Gleditsia triacanthos to Acer negundo.

Transect SD:

Changes in structure between major vegetation zones were detected by relatively low peaks. Peaks 2 and 3 correspond to transect locations where vegetation shifts towards dominance by Acorus and Salix, respectively. In contrast, peaks I and 4 do not represent significant changes observable in the field.

152 Transect SE:

Moving split window analysis on transect SE failed to produce significant peaks

between major vegetation zones. Transition betwen Carex sedge meadow and Salix shrub

meadow is detected by a relatively insignificant peak (peak 2), while change to upland

vegetation (peak 3) is also unclear. Within zones, peak 1 indicates a decrease in both

Carex stricta and Dryopteris thelypteris in the sedge meadow zone. Within upland

vegetation, peaks 4 and 5 denote species composition shifts.

Transect SF:

The major structural change along this transect is the transition from Carex sedge

meadow to Salix shrub meadow denoted by peak 1.

Transect NA:

The discontinuity between Carex open meadow and upland vegetation is detected

by peak I at 65 m. However, despite a relatively clear transition in the field, this peak does

not appear as a distinct, tall peak. Peaks 2 and 3 correspond to shifts in species

composition within the upland vegetation, marked particularly by coverage of Lonicera japonica.

Transect NB:

Transect NB was not characterized by any distinct changes in vegetation structure,

and this is reflected in the moving split-window results . Peaks I and 2 indicate changes in

153 cover due to the presence of several Salix shrubs. Peak 3 characterizes an increase in

Filipendula rubra coupled with a decrease in Solidago ohioensis cover.

Method 3: Federal method of wetland delineation

Characterization of wetland indicator parameters along three transects detected both wetland and nonwetland areas (Table 7.5). Thus parts of the area studied were identified as jurisdictional or potential jurisdictional wetlands. The number of necessary observation points varied along the three transects; there were II, 10 and 8 points in transects I, 2 and 3, respectively. Notes on each of these observation points are presented in Appendix D.

Observations along transects began fi"om the riparian woods and proceeded towards upland area. In most of the open fen area, vegetation was mainly herbaceous species, and thus required only a 1.5 m radius for observation. The hydrophytic vegetation criteria was met in fen areas which corresponded with the open sedge meadow zone, the

Acorus stand, and Salix shrub meadow. The presence of facultative wetland (F ACW) and facultative (FAC) species, as well as the total number of species, increased in the shrub meadow-upland woods boundary zone.

Boundary zones were generally characterized by significantly different species composition within substrate that was still saturated. Plant species that are not obligate

154 Wetland Indicator Criteria Distance firom Determination baseline (m) Hydrophytic Hydric Wetland vegetation: soils: hydrology:

Transect A 0 + + + wetland 22.0 + + + wetland 54.4 + + + wetland 67.4 + + + wetland 78.3 + + + wetland 91.0 + + + wetland 101.0 + + + wetland 107.0 + + + wetland 117.0 + - - nonwetland 122.0 - -- nonwetland 127.0 - - - nonwetland

TransectB 4.0 + + + wetland 17.0 + + + wetland 35.0 + + + wetland 47.0 + + + wetland 62.5 + + + wetland 68.0 + + + wetland 72.0 + + + wetland 75.0 + - - nonwetland 78.0 -- + nonwetland 81.0 - -- nonwetland

Transect C 0 + + + wetland 9.8 + + + wetland 22.8 + + + wetland 31.5 + + + wetland 46.4 + + + wetland 54.7 - - - nonwetland 57.0 - - - nonwetland 59.0 nonwetland

Table 7.5: Results of wetland determination along three transects characterized according to routine onsite determination method o f the 1987 US Army Corps of Engineers Delineation Manual (Environmental Laboratory 1987).

155 wetland species could generally be found in saturated substrate. In contrast, few obligate

species were found in drier soils.

All three criteria for wetland determination (hydrophytic vegetation, hydric soil

and wetland hydrology) were generally found in areas previously designated as upland

boundary zones. Thus these areas were still considered wetland. Following procedures

outlined in the Manual, boundaries between wetland and nonwetland in each transect

could be determined (Table 7.5)

7.4 Summary

Method I: Gradient analysis by Detrended Correspondence Analysis (DCA)

DCA was not equally successful in separating major vegetation zones along

different transects, and therefore in locating boundary lines between adjacent zones.

However, in all cases, sequential plots within a vegetation zone appeared close to one

another within the ordination space, and could therefore be grouped together.

The clustering of plots by DCA can generally be attributed to a topographic or

hydrologic/moisture gradient where sample plots at lower elevations occurred towards

one end of DCA axis 1, while plots on drier areas at higher elevations were found at the

other end. However, it should be noted that the establishment of plant communities is

affected by a complex combination of gradients which in this case may include water level,

water alkalinity, substrate conditions and pH. Although the relative strengths of these gradients may vary in different transects, there were suggestions of which gradient was

more important in a particular transect.

156 Method 2: Moving spllt-window analysis (MSW)

Use of MSW to detect boundary lines between communities provided readily

interpretable graphs. In general, significant peaks corresponded to locations of boundary

lines between major vegetation zones. However, MSW also detected boundary lines at

positions which only corresponded to minor shifts in species composition which were not

observed during field observations.

The relative strengths or importance of each boundary line could be determined by

considering the relative height and width of the significant peak. However, such

determinations could only be made when an adequate number of transects and significant

peaks are available for comparison.

Method 3: Federal method of wetland delineation

Wetland delineation following the US Army Corps of Engineers technical manual

(Environmental Laboratory 1987) detected both wetland and nonwetland areas, and could be used to determine boundary lines along transects. However, observations were mostly qualitative and subjective. In addition, it was difficult to extrapolate transect data to a two-dimensional map as suggested in the Manual, due to difficult field conditions and scale limitation in topographic maps.

157 CHAPTERS

GENERAL DISCUSSION AND COMPARISON OF METHODS

The preceding chapter has demonstrated how detrended correspondence analysis

(DCA) and moving split-window analysis (MSW) may be applied to the detection of

structural discontinuities along one-dimensional transects. Results from both methods

indicate, however, that different methods may not be equally effective when applied to data sets from different transects. The following is a discussion on the strengths and weaknesses o f each of these methods, as well as a comparison between them. The significance o f these strengths and weaknesses in relation to the procedure of wetland delineation will also be discussed.

Detrended correspondence analysis (DCA)

DCA was not equally successful in separating major vegetation zones along different transects, and therefore in locating boundary lines between adjacent zones.

Groupings of sequential plots within a vegetation zone were relatively clear in the ordination space of certain transects, but were not very obvious in other transects.

Many problems associated with gradient analysis techniques (Kent and Coker

1992, ter Braak 1995) must be considered when interpreting DCA results. As one

158 example, outlier plots will have an effect on ordination, regardless of which technique is applied (Causton 1988, Kent and Coker 1992). The removal o f an outlier plot from transect SA significantly affected the dispersion of points within the ordination space. Of course the removal of a plot does not imply that it can be disregarded from interpretation. It should rather be explained why the plot or plots were outliers, and possible ecological reasons for the problem should be investigated (Kent and Coker

1992).

Sampling design for vegetation analysis description and analysis must consider factors such as the time and resources available for study, the type of habitat, and methods of data analysis and presentation (Kent and Coker 1992). Sampling of data in the present study was conducted systematically along continuous belt transects.

However, vegetation analysis for the purpose of community description is often based on data from either stratified or random sampling design.

In order to examine how the number and location of sample plots analysed would affect DCA results, analyses were conducted on subsets of plots from transect SB (which consisted of 104 contiguous sample plots). Subsets of 40 and 20 plots were selected randomly from the transect using random numbers generated by SigmaStat™ (Jandel

Corporation 1994a), and each subset was then subjected to DCA. Results indicate that the subset of 40 plots still adequately represented the transect in describing plant zone separations, while much information was lost when using only 20 plots (Figure 8.1; compare to Figure 7.4B or inset of Figure 8. IB). Therefore, although collecting data randomly from a limited number of plots along the transect may provide information on

159 Figure 8.1: DCA plot ordinations of vegetation data from (A) a subset of 40 plots from transect SB, and (B) a subset of 20 plots from transect SB. Inset = DCA plot ordination based on all plots in transect SB.

160 400

350

O 0 250 0 ® 2“

W ^ 100 ••

o a so

0 100 200 300 400 500 600 700 800 DCA AXIS 1 (Eigenvalue 0.961)

CM 100

200 400 600 800 1000 DCA AXIS 1 (Eigenvalue 0.981)

Figure 8.1

161 plant community structure, it is at the risk of losing information on patch size, or

boundary locations. The number of sample plots for boundary studies could conceivably

be reduced by maintaining the systematic sampling design by increasing distances

between points and/or increasing plot size. However, this would first involve

determination of optimum distance, and must also take into consideration the area and

scale of the site being studied. For example, in using DCA to detect boundaries in the

Great Dismal Swamp, Carter et al. (1994) divided their transects into 25 meter

increments.

Moving split-window analysis (MSW)

Analysis of different transects suggested that an adequate number of transects need to be sampled in order to understand variations in SED values, and to distinguish between significant peaks and random noise. Furthermore, MSW requires some degree of trial and error in determining the optimum window size. Brunt and Conley (1990) examined the behavior of the SED metric on data with known properties and found that edges (i.e., boundary lines) become more distinct fi"om background noise as window- width was increased.

In addition, edge recognition becomes difficult as edges become smaller relative to background heterogeneity. This is also demonstrated by data fi’om the present study.

As an example, transect SB was analysed using window widths of 2, 4, 8 and 16 (Figure

8.2). Although increasing window width resulted in more distinguishable peaks (c/'Brunt and Conley 1990), a window size that was too large reduced important resolution. In this

162 Figure 8.2; Comparison of results from moving split-window analysis on transect SB using four different window widths.

163 30000 A. Window width = 2 25000

20000

15000

10000

5000

.iLllli.liJlimiinljliiiriiiliriJiimliiiLiiiiiliiiiJimliitiiiiiiliiiiiiiiiltirrririiliiiiiiiiilririiiir

10000

14000 B. Window width = 4

12000

® "WM

0000

^ 2000 c m < D III!It II till 11II II 11 II III!I ill III III III ILILIllllllllllllllllllllllllllllllllllllllllllll T3 10000 C. Window width - 8 «000

0000

4000

2000

Iiniiiiliiiiiiinliiiiiiiiiliiiiiiii'l ■1 LI LI I I 11 11 11 I I I I I I I I I I I 11 11 I I I I I I I I I I I I I I I I I I I I I I I II 11 I I I I

D. Window width = 16

11IIIIII111111III il III I nil I lull II11 li 11II11II li III11 null IIII III I liiiii II111...... 11111111111111 ii 111 10 20 30 40 60 60 70 80 90 100

Transect (meters) Figure 8.2 164 case, window widths of 2 and 4 both produced high sample-to-sample noise which made

results difficult to interpret. In contrast, important information was lost when using a

window width of 16. A window size of 8 produced optimum results from the widths

tested, and was thus selected for this study. In general, the choice of window width will

vary with the purpose of the study, the resolution of the data used, and the

characteristics of the attributes involved (Turner et al. 1991). In addition to window

width, distance between sampling points is another factor which may afreet the

resolution of edges (Turner et al. 1991).

Comparison o f results

Although results from DCA and MSW have been interpreted independently, they

were in fact based on the same data sets. Therefore, a comparison can be made between

their effectiveness in locating discontinuities, or boundary lines, within a given transect

(Figures 8.3 and 8.4). In general, the MSW technique was more sensitive at detecting

sample plot discontinuities; in almost all transects, it detected boundary lines at positions

where DCA did not. In other locations within a particular transect, there were cases where the two methods agreed precisely, although there were also cases of slight discrepancies. Thus agreement between these methods varied among transects.

The conclusion is further supported through direct comparisons in which results from one method is substituted into the other. In transect SB, MSW detected peaks at positions which generally supported the hypotheses of where vegetation boundary lines should be according to field observations. Based on these results, individual plots were

165 SA V V

1 111 II II 1 1 [ i-H fi+ H 11 i I m I ‘ n 111 i + H m 1111 i-»m 11 I I ! 11 » n 11 1 I II I H r i [ H 111 i+ i t| ! i i i I i i !■; I 0 10 A 20 30 A A 50 A A 6o A 70 aA 90

T T ▼

0 10 20A 30 40 50 A 70 Â 90 A 100 110

T ▼ ▼ |in iiin i[niiiiw[+iHin ii[iiiH m t|iiiiiiiii|iH n n ii|iiiin iii|iin iin i[n iniiii[iiiiiiiii[iiiiiiiii|iH Hi‘ii|ini!iifi| 0 10 20 30 40 50 60 70 A 80 90 l À A 10 A A 130

SD vv T I l-H H H H 1 1 n H H H 1 1 1 1 i I II II 1 1 1 1 1 1 1 1 I I 11 i 1 11111 l-jiH -H I II 11 1 H-r-H t H 1 1 11 H H H | i 11 H H H [ 11 i I i i H ! | 0 10 20 30 40 50 60 Ao A sA A 90 100

▼ T |lfW titf|H IH IIItfllHIHIl[llliHHl|lllllllll|llH m il|lllllllll|nW HIl|lllllH»|tim i lll|llH H H l[lllllllll[nillilll[lllil!i!!| 0 10 20 30 40 A o 60 70 80 90 A 100 lA lA A 130 140

S F V V

I I i l i I I I i I I I I H - l - l I I l - | - | ' l I I I I I I I I I - + H -I t I 1 - 1 - 1 I I I l - i I I I I i I 0 10 20 A 30 40 50 Transect length (meters) A moving split-window analysis ^ detrended correspondence analysis

Figure 8.3: Comparisons of boundary line positions detected by detrended correspondence analysis (DCA) and moving split-window analysis (MSW) in six transects in the south fen.

166 NA ▼ [-1 m il i-it-j 1111111111 H-m i I n III w h h | i-h-i m i i [ i i i h it [-i n m 11 i | n n - f i i n | 0 10 20 30 40 50 60 A A?0 A 80 90

T T T I I I I I I I I H - f H H -t m I I I I I I i I I I | -| t I I I I I H I I I I I I I I I I I I H I I'l II I I I I I I I I I' H I I I H + H H i { 0 10 A A 20 A 30 40 so 60 70 80 Transect length (meters)

A moving split-window analysis ^ detrended correspondence analysis

Figure 8.4: Comparisons of boundary line positions detected by detrended correspondence analysis (DCA) and moving split-window analysis (MSW) in two transects in the north fen.

167 coded according to which vegetation zone they were found in. Using different symbols

to designate plots according to plant zones detected by MSW, a DCA graph was then

generated (Figure 8.5A). Results indicate relatively clear separation of plot groups

according to vegetation zones. In this case, therefore, results from the two methods were

in close agreement and could be used to support each other.

However, MSW assumes linear change by comparing adjacent plots while not

recognizing similarités between plots that are far apart. Thus it may detect discontinuities

which may not be ecologically significant. This is the case with transect MB where

significant peaks were detected by MSW, but DCA, which takes into account species composition, shows that the plots intergrade with no clear pattern of separation (Figure

8.5B). In this case, results from MSW need to be interpreted with caution. Although peaks of maximum values suggest that there may be structural discontinuities, these values are a result of shifts in dominance between Carex and Scirpus within the sedge meadow zone, and do not appear as distinct zones in the field.

The detection of edges usually involves analytical exploration of graphs (Orloci and Orloci 1990). Indeed the interpretation of both DCA and MSW results is facilitated through the use of visual representation. However, as demonstrated in Chapter 7, it is generally more difficult to visually interpret results from DCA. While significant peaks detected by MSW are readily identifiable, the determination of plot clusters in DCA required a certain degree of subjective judgment.

In terms of data analysis, the development of computer software programs have greatly facilitated calculations for both DCA and MSW. Gradient analyses can now be

168 Figure 8.5: Direct substitutions of MSW results into DCA for (A) transect SB, and (B) transect NB. Different symbols indicate different plant zones as separated by MSW.

169 0L \

DCA AXIS 2 Squared Euclidean Distance 0 5 a a 1 I 00 W 1/1 I 3

CD DCA AXiS 2 Squared Euciidean Distance o 5 3 > Xw aI

p *>■, easily conducted through the use of programs such as PC-ORD™, although it is

important for the user to be familiar with the background theory behind the analyses in

order to select the most appropriate technique. Similarly, calculations in MSW can be readily conducted through the use of spreadsheet programs such as Lotus 123™ or

Excel™.

In order to understand the structure and function of boundaries, it is important to observe them at both community and landscape levels. While DCA can detect vegetation changes at the community level, it is often diflBcult to extrapolate this information to a landscape level. In contrast, MSW may detect changes at a landscape level by overestimating minor shifts in species composition at the community level. While ideally these two methods should be used in conjunction to provide maximum information on boundary location and ecological significance, this would be unrealistic in practical applications. Although comparison of results from the two methods show that neither method is clearly more effective than the other, a method must be selected which satisfies the specific objectives most, taking into consideration the compromise between practicality and reliability.

Many problems have been associated with the procedure of wetland delineation

(e.g., NRC 1995, Tiner 1996). Furthermore, the challenging of accepted paradigms in wetland ecology have presented new factors to consider. For example, Brinson (1993) questioned whether there has been too much emphasis on the wet-to-dry continuum in wetlands, as gradients of moisture do little to explain the transport of water and water­ borne materials through the landscape. Similarly, Bridgham et al. (1996) suggest a

171 reexamination of the relative importance of resource gradients in peatlands. As the response of plant communities to environmental gradients is fundamentally important to wetland characterization (NRC 1995), it is perhaps also important to clarify the confusion that has resulted from failure to distinguish among at least three types of environmental gradients, i.e. indirect environmental gradients (e.g., altitude, aspect), direct environmental gradients (e.g., pH, temperature), and resource gradients (e.g. available nutrients), as species responses to these three types of gradients may differ

(Werger et al. 1983).

Results from wetland delineation in the present study cannot be directly compared to the previous two methods because it was not used to examine the same transects. However, indirect comparisons can be made based on the procedures of data collection and analysis, as well as results obtained. In particular, data collection and analysis for all three methods examined were based on transect data.

In general, the disadvantage of using transects is that they are nonrandom; the placement and intervals of sample-spacing are usually determined subjectively (Johnston et al. 1992). As boundaries are studied through transects which are placed perpendicular to perceived boundaries, data become biased toward boundaries that are readily observable, and may therefore be unrepresentative of prevalent conditions (Johnston et al. 1992). The extrapolation of results from transects to two-dimensions (i.e., to determine patch boundaries) would require numerous transects, or additional data derived from two-dimensional sources.

172 Table 8.1 presents a comparison among the three approaches examined in this study, i.e., DCA, MSW and wetland delineation (WD). While, as mentioned above, transects provide information that are biased toward a visible gradient, DCA can detect and quantitatively rank the relative importance of underlying gradients which were not previously assumed during data collection. However, a weakness of this technique is that it indirectly identifies boundaries by focusing on the contents of patches, rather than on their boundaries.

Feature DCA MSW WD

Unbiased toward visible gradient +

Focus on boundaries, rather than contents - + + Precise species identification not required + + -

Can indicate boundary width and strength - + -

Provides information on seasonal effects -- -

More time and effort required - - +

Table 8.1: Summary of strengths and weaknesses of detrended correspondence analysis (DCA), moving split-window analysis (MSW) and wetland delineation (WD). A "+" sign indicates a strength or desirable feature of the method; a sign indicates a weakness or undesirable feature of the method.

Both DCA and MSW can be used to detect structural changes in vegetation without identifying plants to the species level. This information, combined with information on other environmental factors can establish the location of significant

173 boundary lines within a transect. Indeed, edge detection methods which are not species- based have been proposed in vegetation analysis literature (e.g., Orloci and Orloci 1990).

In contrast, the principle of WD requires precise species identification in order to determine a plant's wetland indicator status, and to consequently determine the presence or absence of hydrophytic vegetation.

A comparison of features presented in Table 8.1 suggests that MSW may have the most desirable properties as it is specifically designed to focus on boundaries, and has the advantage of providing information on boundary strength and abruptness through the widths and heights of peaks produced in its graphical output.

The determination of boundaries in general, and wetland boundaries in particular, must consider seasonal variations in both structural and functional attributes. The present study has demonstrated how seasonal shifts in water chemistry boundaries may have some effect on the resultant structural boundaries of vegetation. Similarly, Pinay and

Naiman (unpublished manuscript, cited in Johnston et al. 1992) have also shown that an ecotonal area may actually contain numerous ecotones related to biogeochemical processes as well as vegetation cover, and these may be temporally asynchronous.

Therefore, it is important to consider the temporal component of spatial change by determining boundaries during different times of the year. However, this is normally not considered when conducting wetland delineations according to the specified procedures, as delineations tend to be conducted at only one time. Wetland functions are a product of all components of the system (not just vascular plants), and it must be realized that the

174 wetland functions all year round (not just during active growth of vascular plants)

(Holland 1996).

Finally, the issue of time and effort relates to the question of how ecological methods can be applied to regulatory wetland delineation. Cost-efiBcient ecological and natural resource surveys require flexible, logistically simple and statistically sound sampling methods, as well as sensitive and computationally simple data analysis (Ludwig and Cornelius 1987). Studies such as the present study should continue to provide input to weigh the trade-offs between practicality and reliability in both data collection and analysis. Although wetland deineation should be based on the professional judgment of a scientist, it is not a "scientific" procedure in that it should be relatively clear for non­ scientists. Thus it is more important to consider the issues of time, manpower, expenses and practicality, with the aim of minimizing time and effort while still considering the importance of all three factors, i.e., vegetation, water and soil.

Observations on water chemistry and soil in the present study have demonstrated that structural boundaries may not coincide with functional boundaries. Although there may be clear structural gradients and discontinuities, the functional determinants may not be as clear. Therefore, while structural boundaries may be easier to observe and can be used as surrogate measures of function, they should be interpreted with caution. If it can be established that structure and function do not coincide, then of course, the next step is to conduct more studies on how to relate these two. Comprehensive boundary determination would require a combination of different approaches (e.g., spatial statistics and GIS) using both coarse and fine scales, depending on our particular objective.

175 In summary, different approaches to boundary determination are not equally effective or efBcient. Therefore, we must evaluate the strengths and weaknesses of each approach as it is applied to our particular objectives. In addition, it is important to realize that the structural and functional boundaries of a system may not coincide, and should be related to one another.

176 CHAPTER 9

RECOMMENDATIONS AND SUMMARY

9.1 Recommendations

Selection o f method

Various quantitative approaches are available for the detection of ecological

boundaries. However, this study has demonstrated that different methods do not necessarily

produce the same results, and thus cannot be used interchangeably. In particular, landscape

ecological approaches may be more sensitive to boundary changes, and are more readily

interpretable, but they may not provide adequate information about community structure, as

would be acquired from approaches based on vegetation analysis. Therefore, a landscape-

based approach should be used when the objective is to understand landscape structure (i.e.,

the specific location of boundaries), and when insight is needed into landscape processes of

species, material and/or energy movement. However, results from such an approach should be interpreted with caution, as they are acquired at the expense of losing information on intrapatch variation in species and community structure. In contrast, if the interest is in community structure and species zonation, then obviously an approach based on vegetation

177 analysis should be selected. The appropriate method to be selected will thus clearly depend

on the particular objectives, and the questions being asked.

The procedure of wetland delineation is concerned with the determination of where a

wetland ends, and takes into consideration boundaries of vegetation, soils and hydrology. In

standardizing field methods to ensure accurate identification of wetland boundaries, Tiner

(1996) suggests, among others, that the method should be (1) technically sound by making

use of current scientific knowledge, (2) precise enough to produce repeatable results, (3)

practical and easy to use, and (4) efficient and requiring minimal effort.

Analysing vegetation data using formal wetland delineation procedures as conducted

in this study (Chapter 7) suggested several weaknesses:

(1) observation points within communities along a transect were determined randomly and

subjectively, while the present study has demonstrated that random samples may

misrepresent vegetation zonations and boundaries.

(2) collection of vegetation data was only based on the presence of dominant species, while

discontinuities may occur in subtle ways, i.e. distinct boundaries may not be obvious

through a discontinuity in dominant vegetation, but are demonstrated mathematically in

groups of subdominant species (Brunt and Conley 1990).

(3) the actual determination of a boundary line between wetland and nonwetland (i.e.,

through the establishment of additional observation points in boundary zones) was

time-consuming, and also involved a degree of subjective judgment.

178 The evaluation of ecological methods such as gradient analysis and moving split-

window analysis conducted in this study can provide input into the development of a more

scientifically sound method of vegetation analysis in wetland delineation. For this particular

study site, I would suggest analysing vegetation through the use of the moving split-window

technique because:

(1) results are easily analysed and clearly interpretable,

(2) data collection is not based solely on dominant plant species, and thus considers the

multivariate nature of the data,

(3) data are not randomly collected, and

(4) collection of data through belt transects ensures the detection of changes in boundary

zones, which contrasts with both WD and DCA which tend to focus on contents rather

than on boundaries (i.e., sampling points are established within vegetation zones).

Of course the application of MSW should take into consideration the type of wetland and area covered. In large areas where using a continuous belt transect would be too laborious, more effort should be concentrated in presumed or potential boundary locations. In addition, the size of quadrats and the window width used in analysis would need to be adjusted according to the specific conditions of the area studied.

In this study, the structural boundaries of wetland vegetation did not coincide entirely with functional boundaries of hydrology or hydrochemistry. Previous studies have shown that information on vegetation, soils, and hydrology can give mixed indications: data on vegetation can indicate conditions either wetter or drier than shown by data on soils or hydrology (NRC 1995). However, the method selected for wetland boundary determination

179 should not be solely based on the analysis of vegetation, as hydrology is widely

acknowledged as the driving force creating and maintaining wetlands (e.g. Mitsch and

Gosselink 1993). On the other hand, Tiner (1993) has proposed a Primary Indicators

Method which does not use hydrology as an indicator. He pointed out that visual signs of

hydrology, including direct observations of water, are only indicators of hydrologie events

and not of their duration and frequency.

In improving the procedures for wetland delineation, the National Research Council

(1995) recommends that boundary determinations involving vegetation analysis should be

confirmed by analysis of substrate. The present study has found that little information was

gained by conducting extensive sampling of soil along transects (Chapter 6). In contrast to

vegetation sampling, it was found that representative random sampling of soil around

boundary zones, supplemented by available information fi'om soil surveys, provided

adequate information to characterize boundaries. Therefore, the collection of samples

regularly along transects and the analysis of data through a method such as MSW is not

recommended. In this case, data collected through wetland delineation procedures provided

adequate and reliable information.

In summary, the evaluation of boundary detection methods in this study suggests the

moving split-window technique to be a practical and reliable method of analysis for vegetation data in this particular study site, and which may have potential application to the procedure of wetland delineation. Although Betsch Fen may not be representative of wetlands in general, it is a wetland type which shares similar hydrogeomorphic characteristics with other groundwater-dominated wetlands, such as seeps and certain

180 marshes. Brinson (1993) has suggested the classification of wetlands based on landscape- level hydrologie movement, i.e., wetlands as water donors, receptors, or conveyors. In this case, fens can be considered receptor wetlands characterized by relatively high primary productivity (compared to precipitation-dominated wetlands), fairly stable geomorphic location, slow flow (low kinetic energy), and insufficient turbulence for sediment transport

(Brinson 1993). As many wetlands share these characteristics, observations and results from

Betsch Fen should be applicable to some similar wetland types.

Further research

The scientific understanding of wetland soils and of correlations between plant distribution and wetland soils should be improved through research and monitoring (NRC

1995). In addition, the physiological effects of flooding and waterlogging on plants under different wetland conditions should continue to be investigated.

As vegetation distribution is a criterion used in locating boundaries, the response of plant communities to environmental gradients is fundamentally important to the characterization of wetlands (NRC 1995). In particular, more research needs to be conducted on the changes in plant community structure at boundary locations as determined by (I) the differing adaptations of plant species to abiotic conditions, and (2) competition among species (NRC 1995). Research may include both field experiments, as well as controlled greenhouse experiments, to further understand plant adaptive mechanisms to specific wetland conditions. An important first step in understanding the relationships

181 between environmental factors and zonation is to gain knowlege at the community level about species composition along gradients (Blom et al. 1996).

Development o f a knowledge based system to facilitate wetland delineation

Various groups have complained that wetland delineation manuals are too technical, complicated, and difficult to understand, and that they have been written for the federal regulator rather than the ultimate users, i.e., landowners and their consultants (Kusler 1992).

A technological tool which may help in this respect is the development of a computer program that would incorporate available scientific knowledge and databases in addition to the actual quantitative techniques required for wetland delineation.

Knowledge-based systems, or knowledge systems, are computer programs that solve problems ordinarily requiring specialized knowledge, and are becoming widely used in natural resource management (Schmoldt and Rauscher 1996). Of course efforts should be made to develop a system that can take the place of a technical manual by being interactive and user-fiiendly. The developed program should then be made available on loan, or provided free, to delineators.

The process of wetland delineation essentially involves three steps; (1) site reconnaissance and the planning of field work, (2) field data collection, and (3) data analysis and boundary determination. A knowledge-based system would have the potential to assist delineators in the first and third steps, e.g. the preliminary determination of field transects, and the analysis of collected data on vegetation, water and soil. The system would also (I) incorporate quantitative ecological methods into data analysis and (2) minimize work for the

182 landowner or delineator by facilitating analytical procedures. Data analysis will be strengthened by a knowledge base consisting of ecological methods and mathematical models, and a database including information on plant species indicator status, hydric soils list, hydrologie thresholds and findings fi'om previous research. Problems such as plant species identification could also be addressed through the interactiveness of this program.

9.2 Summary

9.2.1 The general objective of this study was to compare and evaluate different approaches

to landscape boundary determination by applying them to field data collected from a

wetland system. Three boundary detection methods were examined; (I) gradient

analysis by detrended correspondence analysis (DCA), (2) the moving split-window

(MSW) technique, and (3) the federal method of wetland delineation as outlined by

the US Army Corps of Engineers technical manual. DCA and MSW were used to test

hypothetical boundary locations determined through field measurements and

qualitative observations. Although results fi'om MSW were more easily interpretable,

neither method completely confirmed hypothesized boundary locations based on field

observations. DCA was more successful in detecting vegetation changes at the

community level, but it was often dfficult to extrapolate this information to a

landscape level. In contrast, MSW detected changes at a landscape level which

overestimated minor shifts in species composition at the community level. While DCA

and MSW should ideally be used in conjunction to provide maximum information on

boundary location and ecological significance, this would be unrealistic in practical

183 applications. Although comparisons between the two methods showed that neither is

clearly more efifective than the other, a method must be selected which satisfies the

specific objectives most, taking into consideration the compromise between

practicality and reliability.

9.2.2 Indirect comparisons between DCA and MSW results to wetland delineation results

emphasized the consideration of time, manpower, expenses and practicality issues. In

the case of wetland delineation, the objective should be to minimize time and effort

while employing an ecologically sound quantitative method which considers the

importance of vegetation, water and soil. For this purpose, the moving split-window

technique is suggested as a method which can be applied to the analysis of vegetation

data in wetland delineation. However, for this particular study site, more extensive

sampling of soil along transects did not provide important additional information on

soils, and is not recommended.

9.2.3 Through the application of difierent boundary detection methods, this study also

aimed to describe and characterize the internal and external structural boundaries of

the wetland. Internal vegetation boundaries between communities within the fen and

external boundaries between the fen and upland vegetation were detected and

presented for eight representative transects. These were then related to vegetation

patterns mapped through field reconnaissance and mapping surveys. Functional

boundaries, as defined by groundwater chemistry, were mapped through geostatistica!

analyses. Spatial patterns suggested that structural and functional boundaries do not

coincide. This observation is important when considering the formal procedures of

184 wetland delineation whose aim it is to locate the overall functional boundary of a

system.

9.2.4 Specific objectives of this study were to provide baseline ecological data on

vegetation, water and soil for the particular study site, i.e., Betsch Fen Preserve.

Vegetation surveys and description during the study have found that Betsch Fen is a

high quality fen exhibiting typical fen vegetation zonations and containing many plant

species representative of Ohio fens. Five distinct vegetation zones were mapped in the

south fen: (1) a central marl zone dominated by Scirpiis acutus, (2) an open sedge

meadow zone dominated by Carex spp., (3) a shrub meadow zone dominated by Salix

spp., (4) stands of Acorns calamus^ and (5) woodland vegetation, including riparian

woods and upland woods.

9.2.5 Baseline data were collected to provide information on temporal fluctuations and

spatial distributions in water chemistry. Two years’ monitoring of water chemistry

suggested that there is a highly significant relationship between water alkalinity levels

and plant community distribution. However, spatial patterns were more complex than

previously assumed, i.e. highest water alkalinity levels were not found in the center of

the fen as often suggested in the literature. Furthermore, functional boundaries of

water alkalinity did not coincide entirely with structural boundaries of vegetation.

Temporal patterns in water chemistry were not statistically significant, although

geostatistical analyses suggested seasonal shifts in spatial patterns.

9.2.6 Physicochemical analysis of soils in Betsch Fen revealed spatial variability among soils

underlying different vegetation zones. However, while differences were obvious when

185 comparing the extremes of the topographic and vegetation gradients, it was more difBcuIt to characterize intermediate areas. Thus soil chemistry was generally not as important in determining vegetation establishment as water chemistry. Direct gradient analysis relating vegetation patterns to environmental conditions confirmed water alkalinity as the most important determining factor.

186 LITERATURE CITED

Ailaby, M., editor. 1994. The concise Oxford dictionary of ecology. Oxford University Press, Oxford, UK. 415 pp.

Allison, L.E. 1965. Organic carbon. Pp. 1367-1378 in C.ABlack, editor. Methods of soil analysis. Part 2. American Society of Agronomy, Inc., Madison, Wisconsin.

Allison, L.E. and C D. Moodie. 1965. Carbonate. Pp. 1379-1400 in C.A Black, editor. Methods of soil analysis. Part 2. American Society o f Agronomy, Inc., Madison, Wisconsin.

Altsys Corporation. 1994. Macromedia Freehand™ version 5.0. Altsys Corporation, Richardson, TX.

American Public Health Association (APHA). 1985. Standard methods for the examination of water and wastewater. 16th. edition. American Public Health Association, Washington, DC. 1268 pp.

Anderson, D M 1982. Plant communities of Ohio: a preliminary classification and description. Division of Natural Areas and Preserves, Ohio Department of Natural Resources, Columbus, Ohio. 183 pp.

Andreas, B.K. 1985. The relationship between Ohio peatland distribution and buried river valleys. Ohio Journal of Science. 85 (3): 116-125.

Andreas, B.K. 1989. The vascular flora of the Glaciated Allegheny Plateau region of Ohio. Ohio Biol. Surv. Bull. New Series Vol. 8 No. 1 viii + 191

Andreas, B.K. and G. Bryan. 1990. The vegetation of three 6^Aag7n//M-dominated basin- type bogs in Ohio. Ohio Journal of Science. 90(3): 54-66.

Bares, R.H. and M.K. Wali. 1979. Chemical relations and litter production of Picia maricma and Larix laricina stands on an alkaline peatland in northern Minnesota. Vegetatio. 40: 79-94.

187 Beals, E.W. 1969. Vegetational change along altitudinal gradients. Science 165: 981-985.

Bedford, B.L., M B. Brinson, R. Sharitz, A. van der Valk and J. Zedler. 1992. Evaluation of proposed revisions to the 1989 "Federal manual for identifying and delineating jurisdictional wetlands". Report of the Ecological Society of America's Ad Hoc Committee on Wetlands Delineation. Bulletin of the Ecological Society of America 73(1): 14-23.

Beltman, B. and J.T.A. Verhoeven. 1988. Distribution of fen plant communities in relation to hydrochemical characteristics in the Vechtplassen area, the Netherlands. Pp. 121-135 in J.T.A. Verhoeven, G.W. Heil and M.J.A. Werger, editors. Vegetation structure in relation to carbon and nutrient economy. SPB Academic Publishing, The Hague, The Netherlands.

Bernard, J.M, F.K. Seischab and H.G. Gauch, Jr. 1983. Gradient analysis of the vegetation of the Byron-Bergen swamp, a rich fen in western New York. Vegetatio. 53: 85- 91.

Blom, C.W.P.M, H.M van de Steeg and L.A.C.J. Voesenek. 1996. Chapter 7: Adaptive mechanisms of plants occurring in wetland gradients. Pp. 91-112 in G. Mulamootil, B.G. Warner and E.A. McBean, editors. Wetlands: environmental gradients, boundaries, and buffers. Lewis Publishers, Boca Raton. 298 pp.

Boemer, R.E.J., B.G. DeMars and P.N. Leicht. 1996. Spatial patterns of mycorrhizal infectiveness of soils along a successional chronosequence. Mycorrhiza. 6: 79-90.

Boeye, D., L. Clement and R.F. Verheyen. 1994. Hydrochemical variation in a ground­ water discharge fen. Wetlands. 14(2): 122-133.

Bonham, C D 1989. Measurements for terrestrial vegetation. John Wiley and Sons, New York. 338 pp.

Braun-Blanquet, J. 1964. Pflanzensoziologie. Springer-Verlag Inc., New York.

Bray, J R. and J.T. Curtis. 1957. An ordination of the upland forest communities of southern Wisconsin. Ecological Monographs 27: 325-349.

Bridgham, S.D., J. Pastor, J.A. Janssens and C. Chapin. 1996. Multiple limiting gradients in peatlands: a call for a new paradigm. Wetlands 16(1): 45-65.

Brinson, M.M. 1993. Changes in the functioning of wetlands along environmental gradients. Wetlands 13(2): 65-74.

188 Brooks, R.P. 1989. Chapter two: an overview of ecological functions and economic values of wetlands. Pp. 11-20 in S.K. Majumdar, R.P. Brooks, F.J. Breener and R.W. Tiner, Jr., editors. Wetlands ecology and conservation: emphasis in Pennsylvania. The Pennsylvania Academy of Science.

Brunt, J.W. and W. Conley. 1990. Behavior of a multivariate algorithm for ecological edge detection. Ecological Modelling 49: 179-203.

Bryan, G.R. and B.K. Andreas. 1986. Chemical and physical characteristics of ground waters in eight norteastem Ohio peatlands. Report to the Division of Natural Areas and Preserves, Ohio Department of Natural Resources, Columbus, Ohio.

Bryan, G.R. and B.K. Andreas. 1988. Chemical and physical characteristics of ground waters in Western Ohio fens. Report to the Division of Natural Areas and Preserves, Ohio Department of Natural Resources, Columbus, Ohio.

Burrough, P. A. 1986. Principles of geographical information systems for land resources assessment. Oxford University Press, Oxford, UK. 194 pp.

Burrough, P.A. 1995. Chapter 7: Spatial aspects of ecological data. Pp. 213-255 in R.H.G. Jongman, C.J.F. ter Braak and O.F.R. van Tongeren, editors. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge. 299 PP

Carter, V. 1994. Ecotone dynamics and boundary determination in the Great Dismal Swamp. Ecological Applications 4(1): 189-203.

Causton, D R. 1988. An introduction to vegetation analysis: principles, practice and interpretation. Unwin Hyman, London, UK. 342 pp.

Clarke, K.C. 1990. Analytical and computer cartography. Prentice Hall, Englewood Cliffs, New Jersey. 290 pp.

Clymo, R.S. 1983. Peat. Pp. 159-224 in A.J.E. Smith, editor. Mires: swamp, bog, fen, and moor. Ecosystems of the world 4 A. Elsevier, Amsterdam.

Cooper, D.J. and R E. Andrus. 1994. Pattern of vegetation and water chemistry in peatland of the west-central wind river range, Wyoming, USA. Canadian Journal of Botany. 72: 1586-1597.

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

189 Cusick, A.W. and K.R. Troutman. 1978. The Prairie Survey Project; a summary of data to date. Ohio Biological Survey Informative Circular No. 10. The Ohio State University, Columbus, Ohio. July 1978.

Daubenmire, R. 1968. Plant communities: a textbook of plant synecology. Harper and Row Publishers, New York. 300 pp.

Denny, G.L. 1979. Relicts of the past—Bogs. Pages 141-150 in M B. Lafferty, editor-in- chief. Ohio’s natural heritage. The Ohio Academy of Science, Columbus. 324 pp.

Denny, G.L. 1994. Ohio’s fen communities. Newsletter of the Ohio Department of Natural Resources, Division of Natural Areas and Preserves 16(3): 3. Columbus, Ohio. di Castri, F. and A.J. Hansen. 1992. The environment and development crises as determinants of landscape dynamics. Pages 3-18 in A.J. Hansen and F. di Castri, editors. Landscape boundaries: consequences for biotic diversity and ecological flows. Springer-Verlag Inc., New York. 452 pp.

Diers, R. and J.L. Anderson. 1984. Part 1: development of soil mottling. Soil survey horizons (Winter): 9-12.

Environmental Defense Fund and World Wildlife Fund. 1992. How wet is a wetland? The impact of the proposed revisions to the federal wetlands delineation manual. Environmental Defense Fund and World Wildlife Fund, Washington DC. 175 pp.

Environmental Laboratory. 1987. Corps of Engineers Wetlands Delineation Manual. Technical Report Y-87-1, US Army Engineer Waterways Experiment Station, Vicksburg, Miss.

Fasham, M.J.R. 1977. A comparison of non-metric multidimensional scaling, principal components analysis and reciprocal averaging for the ordination of simulated coenoclines and coenoplanes. Ecology 58: 551-561.

Fennessy, M.S. and W.J. Mitsch. 1989. Design and use of wetlands for renovation of drainage from coal mines. Pp. 231-253 in W.J. Mitsch and S.E. Jorgensen, editors. Ecological engineering: an introduction to ecotechnology.

Fitter, A H. and R.K.M Hay. 1987. Environmental physiology of plants, second edition. Academic Press Inc., San Diego. 423 pp.

Forman, R.T.T. and M. Godron. 1986. Landscape ecology. John Wiley and Sons, Inc., New York. 619 pp.

190 Forman, R.T.T. and P.N. Moore. 1992. Theoretical foundations for understanding boundaries. Pp. 236-258 in A.J. Hansen and F. di Castri, editors. Landscape boundaries: consequences for biotic diversity and ecological flows. Springer- Verlag Inc., New York. 452 pp.

Forman, R.T.T. 1995. Land mosaics: the ecology of landscapes and regions. Cambridge University Press, Cambridge. 632 pp.

Fortin, M-J., P. Drapeau and P. Legendre. 1989. Spatial autocorrelation and sampling design in plant ecology. Vegetatio 83: 209-222.

Fortin, M-J. 1994. Edge detection algorithms for two-dimensional ecological data. Ecology. 75(4): 956-965.

Gamma Design Software. 1992. GS"^ professional geostatistics for the PC, version 2. Ganuna Design Software, Plainwell, Michigan.

Gill, D. 1970. Application of a statistical zonation method to reservoir evaluation and digitized log analysis. Bulletin of the American Association of Petroleum Geologists 54: 719-729.

Gilmore, G., T.E. Lucht and S.J. Hamilton. Draft of Soil Survey of Ross County, Ohio. US Soil Conservation Service and Division of Soil and Water Conservation, Ohio Department of Natural Resources. In press.

Gittins, R. 1969. The application of ordination techniques. Pp. 37-66 in l.H. Rorison, editor. Ecological aspects of mineral nutrition in plants. Blackwell Scientific, Oxford.

Glaser, P H., J.A.Janssens, and D.I. Siegel. 1990. The response of vegetation to chemical and hydrological gradients in the lost river peatland, northern minnesota. Journal of ecology. 78: 1021-1048.

Glavac, V., C. Grillenberger, W. Hakes, and H. Ziezold. 1992. On the nature of vegetation boundaries, undisturbed flood plain forest communities as an example - a contribution to the continuum/discontinuum controversy. Vegetatio 101: 123-144.

Goh, T.B., R.J. St. Arnaud and A.R. Mermut. 1993. Chapter 20: Carbonates. Pages 177- 185 in M R. Carter, editor. Soil sampling and methods of analysis. Lewis Publishers, Boca Raton. 823 pp.

Golden Software, Inc. 1989. SURFER* version 4.07. Golden Software, Inc., Golden, Colorado.

191 Goodall, D.W. 1954. Vegetational classification and vegetational continua. Angew. Pflanzensoziologie, I; 168-182.

Gordon, R.B. 1969. The natural vegetation of Ohio in pioneer days. Bulletin of the Ohio Biological Survey EH (2). The Ohio State University, Columbus, Ohio.

Gorham, E. 1950. Variation in some chemical conditions along the borders of a carex lasiocarpa fen community. Oikos 2(2) : 217-239.

Gorham, E. 1956. The ionic composition of some bogs and fen waters in the English lake district. Journal of ecology. 44: 142-152.

Gorham, E. and W.H. Pearsall. 1956. Acidity, specific conductivity and calcium content of some bog and fen waters in northern Britain. Journal of Ecology. 44:129-141.

Gosz, J.R., C.N. Dahm and P.G. Risser. 1988. Long-path FTIR measurement of atmospheric trace gas concentrations. Ecology 69: 1326-1330.

Gosz, J R. 1991. Fundamental ecological characteristics of landscape boundaries. Pages 8- 30 in M.M. Holland, P.G. Risser and R.J. Naiman, editors. Ecotones: the role of landscape boundaries in the management and restoration of changing environments. Chapman and Hall, New York. 142 pp.

Gosz, J R. 1992. Gradient analysis of ecological changes in time and space: implications for forest management. Ecological Applications 2(3): 248-261.

Greig-Smith, P. 1964. Quantitative plant ecology, second edition. Butterworth and Co. Ltd., Washington, D C.

Greig-Smith, P. 1983. Quantitative plant ecology, third edition. Blackwell Scientific, Oxford, UK.

Hansen, A.J. and F. di Castri, editors. 1992. Landscape boundaries: consequences for biotic diversity and ecological flows. Springer-Verlag, New York. 452 pp.

Heathwaite, A.L., Kh. Gottlieb, E.-G. Burmeister, G. Kaule and Th. Grospietsch. 1993. Chapter 1: Mires: definitions and form. Pp. 1-75 in A.L. Heathwaite and Kh. Gottlich, editors. Mires: process, exploitation and conservation. John Wiley and Sons, Chichester. 506 pp.

Hendershot, W.H., H. Lalande and M. Duquette. 1993. Chapter 16: Soil reaction and exchangeable acidity. Pages 141-145 in M R. Carter, editor. Soil sampling and methods of analysis. Lewis Publishers, Boca Raton. 823 pp.

192 Herrick, J.A. 1974. The natural areas project (of Ohio), a summary of data to date. Ohio Biological Survey Informative Circular No. 1. OSU Columbus, Ohio. 60 pp.

Hill, M.O. 1973. Reciprocal averaging: an eigenvector method of ordination. Journal of Ecology 61: 237-250.

Hill, M.O. 1979. DECORANA - a FORTRAN program for detrended correspondence analysis and reciprocal averaging. Cornell University, Department o f Ecology and Systematics, Ithaca, New York.

Hill, M.O. and H.G. Gauch. 1980. Detrended correspondence analysis, an improved ordination technique. Vegetatio 42: 47-58.

Hillmer, J. 1991. Preserve highlight: Betsch Fen. Newsletter of the Nature Conservancy: Ohio Chapter. Spring 1991.

Holland, M.M., P.G. Risser and R.J. Naiman, editors. 1991. Ecotones: the role of landscape boundaries in the management and restoration of changing environments. Chapman and Hall, New York. 142 pp.

Holland, M.M. and P.G. Risser. 1991. The role of landscape boundaries in the management and restoration of changing environments: introduction. Pages 1 -7 in M.M. Holland, P.G. Risser and R.J. Naiman, editors. Ecotones: the role of landscape boundaries in the management and restoration of changing environments. Chapman and Hall, New York. 142 pp.

Holland, M.M. 1996. Wetlands and environmental gradients. Pages 19-43 in G Mulamootil, B.G. Warner and E.A. McBean, editors. Wetlands: environmental gradients, boundaries, and buffers. Lewis Publishers, Boca Raton. 298 pp.

Hook, D.D., B. Davis, J. Scott, J. Struble, C. Bunton and E.A. Nelson. 1995. Locating delineated wetland boundaries in coastal South Carolina using global positioning systems. Wetlands 15(1): 31-36.

Hurt, G.W. and R.B. Brown. 1995. Development and application of hydric soil indicators in Florida. Wetlands 15(1): 74-81.

Jandel Corporation. 1994a. SigmaStat™ image measurement software. Jandel Scientific, San Rafael, CA.

Jandel Corporation. 1994b. SigmaPlot™ image measurement software. Jandel Scientific, San Rafael, CA.

193 Jandel Corporation. 1995. SigmaScan™ image measurement software. Jandel Scientific, San Rafael, CA.

Johnston, C.A. and J.P. Bonde. 1989. Quantitative analysis of ecotones using a gegraphic information system. Photogrammetric Engineering and Remote Sensing 55; 1643- 1647.

Johnston, C.A., J. Pastor and G. Pinay. 1992. Quantitative methods for studying landscape boundaries. Pages 107-125 in A.J. Hansen and F. di Castri, editors. Landscape boundaries: consequences for biotic diversity and ecological flows. Springer- Verlag Inc., New York. 452 pp.

Jongman, R.H.G. 1995. Chapter 1: Introduction. Pages 1-9 in R.H.G. Jongman, C.J.F. Ter Braak and O.F.R. van Tongeren, editors. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge. 299 pp.

Jongman, R.H.G., C.J.F. Ter Braak and O.F.R. van Tongeren, editors. 1995. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge. 299 PP

Kadlec, J.A. 1989. Hydrology. Pp. 8-11 in E.J. Murkin and H R. Murkin, editors. Marsh ecology research program: long-term monitoring procedures manual. Paper no. 54 of the Marsh Ecology Research Program, Delta Waterfowl and Wetlands Research Station and Ducks Unlimited Canada. Manitoba, Canada.

Kent, M. and J. Ballard. 1988. Trends and problems in the application of classification and ordination methods in plant ecology. Vegetatio 78:109-124.

Kent, M. and P. Coker. 1992. Vegetation description and analysis: a practical approach. John Wiley and Sons, Chichester.

Kershaw, K.A. and J.H.H. Looney. 1985. Quantitative and dynamic plant ecology, third edition. Edward Arnold, London.

Knoop, J.D. and B.K. Andreas. 1987. Vegetational survey of Sinking Creek Fen, Clark County, Ohio. Ohio Journal of Science. 87(2):4.

KoUmorgen Corporation. 1975. Munsell soil color charts. Macbeth division ofKollmorgen Corporation. Baltimore, Maryland.

Kiichler, A.W. 1988a. Chapter 28: Küchler’s comprehensive method. Pp. 393-399 in A.W. Kiichler and I S. Zonneveld, editors. Vegetation mapping. Kluwer Academic Publishers, Dordrecht. 635 pp.

194 Kiichler, A.W. 1988b. Chapter 4A: A physiognomic and structural analysis of vegetation. Pp. 37-50 in A.W. Kiichler and I.S. Zonneveld, editors. Vegetation mapping. Kluwer Academic Publishers, Dordrecht. 635 pp.

Kusler, J. 1992. Wetlands delineation: an issue of science or politics? Environment 34(2): 7-11,29-37.

Leopold, A. 1933. Game management. Scribners, New York. 481 pp.

Lotus Development Corporation. 1993. Lotus 123™. Lotus Development Corporation, Cambridge.

Lucas, R E. and J.F. Davis. 1961. Relationships between pH values of organic soils and availabilities of 12 plant nutrients. Soil Science. 92(3): 177-182.

Ludwig, L.S. 1985. Influences of water pH, alkalinity and acid additions upon floral crop growth and nutrient relationships. Master of Science thesis. The Ohio State University, Columbus. 116 pp.

Ludwig, J.A. and J.M. Cornelius. 1987. Locating discontinuities along ecological gradients. Ecology 68(2): 448-450.

Lyon, J.G. 1993. Practical handbook for wetland identification and delineation. Lewis Publishers, Boca Raton, Florida. 157 pp.

Lyon, J.G. and J. McCarthy, editors. 1995. Wetland and environmental applications of GIS. Lewis Publishers, Boca Raton. 373 pp.

Matheron, G. 1971. The theory of regionalized variables and its applications. Les Cahiers du centre de morphologie mathématique de Fontainebleu. Ecole Nationale Supérieure des Mines, Paris, France.

Mausbach, M.J. and J.L. Richardson. 1994. Biogeochemical processes in hydric soil formation. Pp. 68-127 in Current topics in wetland biogeochemistry, volume 1. Wetlands Biogeochemistry Institute, Louisiana State University.

McCance, R.M., Jr. and J.F. Bums, editors. 1984. Ohio endangered and threatened vascular plants: abstracts of state-listed taxa. Division of Natural Areas and Preserves, Department of Natural Resources, Colmbus, Ohio. 635 pp.

McCoy, E D., S.S. Bell and K. Walters. 1986. Identifying biotic boundaries along environmental gradients. Ecology 67(3): 749-759.

195 McCune, G. and M.J. Mefiford. 1995. PC-ORD™. Multivariate analysis of ecological data, version 2.0. MJM Software Design, Gleneden Beach, Oregon, USA.

Metzger, J.P. and E. Muller. 1996. Characterizing the complexity of landscape boundaries by remote sensing. Lanscape Ecology 11(2): 65-77.

Milne, B.T. 1991. Chapter 9: Lessons fi’om applying fractal models to landscape patterns. Pp. 199-235 in M.G. Turner and R.H. Gardner, editors. Quantitative methods in landscape ecology. Springer-Verlag Inc., New York. 536 pp.

Mitsch, W.J. and J.G. Gosselink. 1993. Wetlands. 2nd. edition. Van Nostrand Reinhold, New York. 722 pp.

Moore, P.D. and D.J. Bellamy. 1974. Peatlands. Elek Science, London. 221 pp.

Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and methods of vegetation ecology. John Wiley and Sons Inc., New York. 547 pp.

Musick, H.B. and H.D. Grover. 1991. Chapter 4: Image textural measures as indices of landscape pattern. Pp. 77-103 in M.G. Turner and R.H. Gardner, editors. Quantitative methods in landscape ecology. Springer-Verlag Inc., New York. 536 PP

Naiman, R.J. and H. Decamps, editors. 1990. The ecology and management of aquatic- terrestrial ecotones. The Parthenon Publishing Group Inc. New Jersey, NJ. 316 pp.

National Research Council (NRC). 1995. Wetlands: characteristics and boundaries. National Academy Press, Washington, D C. 306 pp.

National Wetland Policy Forum. 1988. Protecting America's wetlands: an action agenda. Conservation Foundation, Washington DC. 69 pp.

Nelson, R E. 1982. Chapter 11: carbonate and gypsum. Pp. 181-197 in A.L. Page, R.H. Miller and D R. Keeney, editors. Methods of soil analysis. Part 2: chemical and microbiological properties. Second edition. American Society of Agronomy, Madison, Wisconsin.

Nelson, R E. and L.E. Sommers. 1982. Chapter 29: total carbon, organic carbon, and organic matter. Pp. 539-579 in A.L. Page, R.H. Miller and D R. Keeney, editors. Methods of soil analysis. Part 2: chemical and microbiological properties. Second edition. American Society of Agronomy, Madison, Wisconsin.

Nwadialo, B E. and F.D. Hole. 1988. A statistical procedure for partitioning soil transects. Soil Science 145: 58-62.

196 Odum, E.P 1971. Fundamentals of ecology, third edition. W.B. Saunders Company, Philadelphia. 574 pp.

Odum, W.E. 1987. Predicting ecosystem development following creation and restoration of wetlands. Pp. 67-70 in J. Zelazny and J.S. Feierabend, editors. Wetlands; increasing our wetland resources. Proceedings of the conference wetlands: increasing our wetland resources, Washington, DC, Corporate Conservation Council, National Wildlife Federation, Washington, DC.

Ohio Department of Natural Resources, Division of Soil and Water Conservation (ODNR DSWC). 1993. Information about the soils of Ross County, Ohio: their characteristics and capabilities. ODNR Division of Soil and Water Conservation, Columbus, Ohio.

Orlôci, L. 1966. Geometric models in ecology. I. The theory and application of some ordination methods. Journal of Ecology 54: 193-215.

Orlôci, L. and M. Orlôci. 1990. Edge detection in vegetation: Jornada revisited. Journal of vegetation science 1:311-324.

Petro, J.H., W.H. Shumate and M R. Tabb. 1967. Soil survey of Ross County, Ohio. USDA Soil Conservation Service, Washington DC.

Prentice, I.C. 1977. Non-metric ordination models in ecology. Journal of Ecology. 65: 85- 94.

Pringle, J.G. 1980. An introduction to wetland classification in the Great Lakes region. Technical Bulletin no. 10. Royal Botanical Gardens, Hamilton, Canada.

Raad, A.A. 1978. Carbonates. Pp. 86-98 in J.A. McKeague, editor. Manual on soil sampling and methods of analysis. Second edition. Canadian Society of Soil Science, Ottawa, Ontario.

Reed, P B., Jr. 1988. National list of plant species that occur in wetlands: national summary. US Fish and Wildlife Service Biological Report 88(24). 244 pp.

Richardson, C.J. 1994. Ecological functions and human values in wetlands: a framework for assessing forestry impacts. Wetlands 14(1): 1-9.

Ricklefs, R E. 1990. Ecology, third edition. W.H. Freeman and Company, New York. 896 PP

Ripley, B.D. 1981. Spatial statistics, series in probability and mathematical statistics. John Wiley, New York.

197 Risser, P.G. 1990. The ecological importance of land-water ecotones. Pages 7-21 in R.J. Naiman and H. Decamps, editors. The ecology and management of aquatic- terrestrial ecotones. The Parthenon Publishing Group, New Jersey.

Risser, P.G. 1993. Ecotones at local to regional scales from around the world. Ecological Applications 3(3): 367-368.

Risser, P.G. 1995a. The status of the science examining ecotones. Bioscience 45(5): 318- 325.

Risser, P.G., editor. 1995b. Understanding and managing ecotones. Proceedings of the 3rd. Intemation SCOPE/UNEP workshop on ecotones. Ecology International 1995: 22. Published by the International Association for Ecology.

Roberts, M.L. and T.S. Cooperrider. 1982. Dicotyledons. Pages 48 to 84 in T.S. Cooperrider, editor. Endangered and threatened plants of Ohio. Ohio Biological Survey Biological Notes no. 16. The Ohio State University, Columbus, Ohio. 92 PP

Robertson, G.P. and D.W. Freckman. 1995. The spatial distribution of nematode trophic groups across a cultivated ecosystem. Ecology. 76(5): 1425-1432.

Rohlf F.J. and R.R. Sokal. 1969. Statistical tables. WH Freeman and Company, San Francisco. 253 pp.

SAS. 1985. Statistical analysis system user’s guide: statistics. SAS Institute, Cary, NC.

Schmoldt, D.L. and H.M Rauscher. 1996. Building knowledge-based systems for natural resource management. Chapman and Hall, New York.386 pp.

Schneider, G.J. 1992. Cyperaceae of Ohio fens: floristic analysis and phytogeographical relationships. Master of Science thesis. The Ohio State University, Columbus.

Schwintzer, Christa.R. and T.J. Tomberlin. 1982. Chemical and physical characteristics of shallow ground waters in northern Michigan bogs, swamps and fens. American Journal of Botany. 69(8) : 1231- 1239.

Sjors, H. 1950. On the relation between vegetation and electrolytes in north Swedish mire waters. Oikos. 2(2): 241-258.

Slack, N.G., D.H. Vitt and D.G. Horton. 1980. Vegetation gradients of minerotrophically rich fens in western Alberta. Canadian Journal of Botany. 58: 330-350.

198 Smith, R.L. 1966. Ecology and field biology. Harper and Row Publishers, New York, 686 PP

Snedecor, G.W. 1956. Statistical methods. Fifth edition. The Iowa State University Press.

Snoeyink, V.L. and D. Jenkins. 1980. Water chemistry. Wiley, New York.

Snowden, R.D. and B.D. Wheeler. 1993. Iron toxicity to fen plant species. Journal of Ecology. 81: 35-46.

St. Arnaud, R.J. and A.J. Herbillon. 1973. Occurrence and genesis of secondary magnesium-bearing calcites in soils. Geoderma. 9: 279-298.

Stohlgren, T.J. and R.R. Bachand. 1997. Lodgepole pine {Pirtus contorta) ecotones in Rocky Mountain National Park, Colorado, USA. Ecology 78(2): 632-641.

Stuckey, R.L. and G.L. Denny. 1981. Prairie fens and bog fens in Ohio: floristic similarities, differences, and geographical affinities. Pages 1-33 in R.C. Romans, editor. Geobotany H. Plenum Publishing Corporation, New York.

Stuckey, R.L. and M.L. Roberts. 1982. Monocotyledons. Pages 27-47 in T.S. Cooperrider, editor. Endangered and threatened plants of Ohio. Ohio Biological Survey, Biological Notes no. 16. Ohio State University, Columbus, Ohio.

Stumm, W. and J.J. Morgan. 1981. Aquatic chemistry: an introduction emphasizing chemical equilibria in natural waters. Second edition. John Wiley and sons. New York. ter Braak, C.J.F. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67: 1167-1179. ter Braak, C.J.F. 1988. CANOCO - an extension of DECORANA to analyse species- environment relationships. Vegetatio 75: 159-160. ter Braak, C.J.F. 1995. Chapter 5: Ordination. Pages 91-173 in R.H.G. Jongman, C.J.F. ter Braak and O.F.R. van Tongeren, editors. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge. 299 pp.

Tiner, R.W. 1993. The primary indicators method - a practical approach to wetland recognition and delineation in the United States. Wetlands 13(1): 50-64.

Tiner, R.W. 1996. Chapter 8: Practical considerations for wetland identification and boundary delineation. Pp. 113-138 in G. Mulamootil, B.G. Warner and E.A.

199 McBean, editors. Wetlands; environmental gradients, boundaries, and buffers. Lewis Publishers, Boca Raton. 298 pp.

Turner, S.J., R.V. O’Neill, W. Conley, M R. Conley and H.C. Humphries. 1991. Chapter 2: Pattern and scale: statistics for landscape ecology. Pages 17-49 in M.G. Turner and R.H. Gardner, editors. Quantitative methods in landscape ecology. Springer- Verlag Inc., New York. 536 pp.

US Soil Conservation Service. 1987. Soil taxonomy: a basic system of soil classification for making and interpreting soil surveys. US Soil Conservation Service agricultural handbook 436. Washington, DC. 754 pp. van der Maarel, E. 1990. Ecotones and ecoclines are different. Journal of vegetation science 1: 135-138. van Leeuwen, C.G. 1966. A relation-theoretical approach to pattern and process in vegetation. Wentia 15: 25-41. van Tongeren, O.F.R. 1995. Chapter 6: Cluster analysis. Pp. 174-212 in R.H.G. Jongman, C.J.F. ter Braak and O.F.R. van Tongeren, editors. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge. 299 pp.

Verhoeven, J.T.A., W. Koerselman, and B. Beltman. 1988. The vegetation of fens in relation to their hydrology and nutrient dynamics: a case study. Pp.249- 282. in J.J. Symoens, editor. Vegetation of inland waters. Kluwer Academic Publishers, Dordrecht. Netherland. 385 pp.

Vitt, D.H. and S. Bayley. 1984. The vegetation and water chemistry of four oligotrophic basin mires in northwestern Ontario. Canadian Journal of Botany. 62:1485-1500.

Wallbridge, M R. 1994. Plant community composition and surface water chemistry of fen peatlands in West Virginia’s Appalachian Plateau. Water, air and soil pollution 77:247-269.

Wassen, M.J., A. Barendregt, M.C. Bootsma, and P.P. Schot. 1989. Ground water chemistry and vegetation of gradients from rich fen to poor fen in the Naardermeer (the Netherlands). Vegetatio 79:117-132.

Wassen, M.J., A. Barendregt, A. Palczynski, J.T. De Smidt and H. De Mars. 1990. The relationship between fen vegetation gradients, groundwater flow and flooding in an undrained valley mire at Biebrza, Poland. Journal of Ecology. 78: 1106-1122.

Wassen, M.J., and A. Barendregt. 1992. Topographic position and water chemistry of fens in a Dutch river plain. Journal of Vegetation Science. 3:447-456.

200 Weber, M. and R. Brandie. 1994. Dynamics of nitrogen-rich compounds in roots, rhizomes, and leaves of sweet flag (Aconis calamus L.) at its natural site. Flora. 189:63-68.

Webster, R. and I.F.T. Wong. 1969. A numerical procedure for testing soil boundaries interpreted fi'om air photographs. Photogrammetria 24: 59-72.

Webster, R. 1973. Automatic soil-boundary location fi'om transect data. Mathematical Geology 5: 27-37.

Weishaupt, C.G. 1971. Vascular plants of Ohio - a manual for use in field and laboratory. Third edition. Kendall/Hunt Publishing Company, Dubuque, Iowa. 292 pp.

Werger, M.J.A., J.M.W. Louppen and J.H.M. Eppink. 1983. Species performances and vegetation boundaries along an environmental gradient. Vegetatio 52: 141-150.

WesthoflF, V. 1971. The dynamic structure of plant communities in relation to the objectives of conservation. Pp. 3-14 in E. Duffey and A.S. Watt, editors. The scientific management of animal and plant communities for conservation. Blackwell Scientific Publications, Oxford, UK. 652 pp.

Whittaker, R.FI. 1967. Gradient analysis of vegetation. Biological Review 49: 207-264.

Whittaker, R.H. 1960. Vegetation of the Siskiyou Mountains, Oregon and California. Ecological Monographs 30: 279-338.

Wiens, J.A., C.S. Crawford and JR. Gosz. 1985. Boundary dynamics: a conceptual fi'amework for studying landscape ecosystems. Oikos 45: 421-427.

Wierenga, P.J., J.M.H. Hendrickx, M.H. Nash, J.A. Ludwig and L A. Daugherty. 1987. Variation of soil and vegetation with distance along a transect in the Chihuahuan desert. Journal of Arid Environments 13: 53-63.

Wilde, S.A., and G.W. Randall. 1951. Chemical characteristics of ground water in forest and marsh soils of Wisconsin. Trans. Wisconsin Academy of Science, Arts and Letters 40: 251-259.

Wilson, M.V. and C.L. Mohler. 1983. Measuring compositional change along gradients. Vegetatio 54: 129-141.

Zonneveld, I.S. 1988. Chapter 4: Composition and structure of vegetation. Pp. 25-35 in A.W. Kiichler and I.S. Zonneveld, editors. Vegetation mapping. Kluwer Academic Publishers, Dordrecht. 635 pp.

201 APPENDICES

202 APPENDIX A

Table A. 1 : List of vascular plant species recorded in Betsch Fen (de­ noted by "x" sign) from vegetation analyses and surveys conducted during this study (1994-1996). Observations mainly conducted in fen area south of Blackwater Creek, unless otherwise noted by N (north). Nomenclature follows Weishaupt (1971).

203 Species Area of fen;

Central Sedge Shrub Wood- mari meadow meadow lands

Acer negundo L. - - - x Acer saccharinum L. - - - x Acorus calamus L. x x - Agrimonia sp. x Andropogon gerardi Vitman x x - - Angelica sp. - - - x Asclepias incamata L. - x - - Aster puniceus L. - x x - Calamagrostis inexpansa Gray - x - - Campanula aparinoides^nrsh x x - - Carex buxbaumii Wahl. - x x - Carex frankii Kunth. - x - - Carex hystricina Muhi. - x x - Carex lamtginosa Michx. x x - Carex rosea Schkuhr - x - - Carex stricta Lam. x x x - Carex suberecta (OIney) Britt. - x - - Carex trichocarpa Muhi. - x - - Circaea quadrisulcata (Maxim.) Franch. & Sav. - - x x Cirsium muticum Michx. x x x Convolvulus arvensis L. - x - x Comus amomum Mill. - x - x Comus florida L. - - - x Comus stoloniferayildax. - x - x Crataegus sp. x Cryptotaenia canadensis (L.) DC. x Cuscuta gronovii Wû\é. - - - x Dryopteris thelypteris (L.) Gray var. pubescens (Lawson) Nakai (* = Thelypterispalustris Schott.) x x x - Elymus virginicus L. x Eupatorium maculatum L. x x - Eupatorium perfoliatum L. - x x - Eupatorium rugosum Houtt. Filipendula rubra (Hill) Robins

Table A. 1 (to be continued) 204 Table A. I (continued)

Species Area of fen:

Central Sedge Shrub Wood- marl meadow meadow lands

Fraxinus omehcanaL. - - - x Galium asprellum Michx. - x x - Galium sp.2 x Gentiana procera Holm. - x - - Geum sp. x Gerardia (=Agalinis) purpurea L. x x - - Gleditsia triacanthos L. - - - x Glyceria striata (Lam.) Hitchc. - x - x Helenium autumnale L. x x - - Hydrophyllum virginiamtm L. x Impatiens capensis Meerb. {biflora Wiild) x x x Juglans cinereaL. x Jugions nigra L. - - - x Juncus acuminatus Michx. - x - - Juncus brachycephalus (Engelm.) Buch. x x - - Juncus torreyi Cov. - x - - Juniperus virginiana L. x Leersia oryzoides (1.) Swartz - x x - Ligustrum vulgare L. - - - x Lindera benzoin (L.) Blume x Lobelia kalmii L. x x - - Lobelia siphiliticaL. - x x - Lonicera japonica Thunb. - - x x Lysimachia nummulariaL. - - - x Lysimachia (=Steironema) quadriflora Sims. x x - - Lythnim alatum Pursh - x - - Madura pomifera (Raf.) Schneid - - - x Morus rubra L. x Muhlenbergia glomerataQfl\\\â.)Tnn. - x - - Muhlenbergia sylvatica Ton. - x - x Parthenocissus quinquefolia(L.) x Pedicularis lanceoiata Michx. - x - x

(to be continued)

205 Table A. 1 (continued)

Species Area of fen; CenUal Sedge Shrub Wood- marl meadow meadow lands

Pileapumila (L.) Gray - - - x Platams occidentalis L. x Poa palustris L. x Podophyllum peltatum L. x Polygonum virginianum L. x Potentilla fruticosa L - x (N) Pycnanthemum virginicmum (L.) x x x - Pyrus coronaria L. - - - x Quercus macrocarpa^fJlichx.. - - - x Rhus radicans L. - - - x Rhynchospora capillacea Torr. x x - - Rosa palustris Marsh. x x - Rosa setigera Michx. - x x - Rubus allegheniensis?on. - - - 4 Rudbeckia hirta L. x x - - Rumex orbiculatus Gray - - - x (N) Salix discolor - x x - Salix interior Rowles (* = Salix exigua Nutt.) x Salix nigra Maxûi. x Sanguisorba canadensis L. x x - - Sanicula canadensis L. x Scirpus acutusyiuhl. x x - - Scirpus atrovirens Willd. - x x - Sparganium americanum Nutt. - - - x (N) Smilax tamnoides L. van hispida (Muhl.) Fern x x Solidago canadensis L. x x x - Solidago gigantea Ait. - x - - Solidago ohioensis Riddell. x x - - Solidago patula Muhl. - x x - Solidago uliginosa Nutt. x x - Sorghastnim rtutans(L.) Nash x x - - Symplocarpusfoetidus{L.)'i^\x\X. - x x x Typha latifolia L. - x - -

(to be continued) 206 Table A. I (continued)

Species Area of fen; Central Sedge Shrub Wood­ marl meadow meadow lands

Ulmus americana L. X

Ulmtis ntbra Muhl. — - - X Urtica dioica L. — — - X

Verbena hastata L. X X -

Verbesina altemifolia (L.) Britt. - - X

Vemonia altissima Nutt. X X - Viburnum prunifoliiim L. - - X Viola sp. - - X vais sp. -- X

207 APPENDIX B

Figure B. 1 : Map of water-sampling well locations (denoted by "w") established in the field. Map shows mainly the south fen area.

208 liiilllli 0

miHi É AAAAAAAAA ImMi n A A A A AAAAAAAAAAAAAA Al*!%M! A Am ..... AAAAAAAA A. A. A A A A *:*»!* A A A A:)

^ ^ ^ * * %* A A n l|{«A È r A A A AmiiiiiiiiiiM A A A A A A A A

ÎY^t VO mBlimilUllH * * îIyîck I lllËWtÂÂÀA 'Y'WVY V 'Y Y A A A A AJH

A A**W»W»V*VN^*V*'#';'^

FigineB.l APPENDIX C

Figure C. I ; Map of the placement of transects (SA to SF) established for vegetation analysis in the south fen.

210 I!!!!! mi!!:

###11 IIH#0

. J! . gnp".'.".*.".".'.#™'AAAAAAAAA

1hi! ilii Ilii ii ii ■ « ■ k J i i p

N> I I I ! rJCI$K y Iy Iy T

Central marl Sedge meadow II Æ# Acorns stand Shrub meadow ^ • ^ • < { jP a a a a a Wbodland AAA mmiiimmii

Figure C.l APPENDIX D

Field notes on observations of dominant plant species and field conditions during wetland delineation.

TRANSECT A:

Sampling point A-I (0 m): Plant species; Carex stricta Dryoptehs thelypteris Scirpus acutus Sanguisorba canadensis Solidago canadensis Soil wet and saturated.

Sampling point A-2 (22.0 m from baseline) Plant species: Carex stricta Dryopteris thelypteris Sanguisorba canadensis Soil wet and saturated

Sampling point A-3 (54.4 m from baseline) Plant species: Acorus calamus Carex stricta Comus amomum Salix interior Some standing water present

Sampling point A-4 (67.4 m from baseline) Plant species: Acorus calamus Carex stricta Impatiens capensis Salix interior Symplocarpus foetidus Soil wet and saturated

Sampling point A-5 (78.3 m from baseline) Plant species: Carex stricta Impatiens capensis Rosa palustris 212 Salix interior Symplocarpus foetidus Soil very wet with some standing water

Sampling point A-6 (91.0 m from baseline) Plant species; Carex hystricina Carex stricta Comus amomum Impatiens capensis Salix interior Symplocarpus foetidus Ulmus rubra Soil very wet

Sampling point A-7 (101.0 m from baseline) Plant species: Acer negundo Cirsium muticum Comus amomum Gleditsia triacanthos Impatiens capensis Symplocarpus foetidus Ulmus rubra Soil drier than previous sampling point

Sampling point A-8 (107.0 m from baseline) Plant species: Carex spp. Comus stolonifera Impatiens capensis Leersia oryzoides Platanus occidentalis Rosa setigera Salix discolor Symplocarpus foetidus Soil relatively dry

Sampling point A-9 (117.0 m from baseline) Plant species: Acer negundo Acer saccharinum Comus sp. Elymus virginicus Lonicera japonica Rhus radicans Soil dry

Sampling point A-10 (122.0 m from baseline) Plant species: Acer saccharinum Elymus virginicus Galium sp. 213 Gleditsia triacanthos Poa sp. Verbesina altemifolia Soil dry

Sampling point A-II (127.0 m from baseline) Plant species; Acer saccharinum Elymus virginicus Galium sp. Gleditsia triacanthos Poa sp. Rosa setigera Rhus radicans Rubus allegheniensis Verbesina altemifolia Soil dry

TRANSECT B:

Sampling point B-I (4.0 m from baseline): Plant species: Carex spp. Comus amomum Pycnanthemum virginianum Salix interior Sanguisorba canadensis Solidago canadensis Solidago uliginosa Soil wet and saturated

Sampling point B-2 (17.0 m from baseline) Plant species: Carex spp. Comus stolonifera Rosa sp. Salix interior Symplocarpus foetidus Standing water present

Sampling point B-3 (35.0 m from baseline) Plant species: Carex stricta Symplocarpus foetidus More standing water than previous sampling point

Sampling point B-4 (47.0 m from baseline) P lant species : Carex stricta Impatiens capensis Salix interior Soil very wet, some standing water present 214 Sampling point B-5 (62.5 m from baseline) Plant species; Impatiens capensis Leersia oryzoides Rosa sp. Salix interior Symplocarpus foetidus No standing water, but saturated soil

Sampling point B-6 (68.0 m from baseline) Plant species: Acer negundo Comus stolonifera Impatiens capensis Leersia oryzoides Lonicera Japonica Symplocarpus foetidus Ulmus rubra Soil somewhat wet

Sampling point B-7 (72.0 m from baseline) Plant species: Acer negundo Convolvulus arvensis Galium palustre Leersia oryzoides Lonicera japonica Ulmus rubra Vitis sp. Much drier than previous sampling point

Sampling point B-8 (75.0 m from baseline) Plant species: Impatiens capensis Polygonum virginianum Rosa setigera Symplocarpus foetidus Relatively dry soil

Sampling point B-9 (78.0 m from baseline) Plant species: Muhlenbergia sylvatica Elymus virginicus Ulmus rubra Viburnum prunifolium soil moist Sampling point B-10 (81.0 m from baseline) Plant species: Madura pomifera Poa palustris Ulmus rubra Viburnum prunifolium Soil dry

215 TRANSECT C:

Sampling point C-1 (0 m from baseline); Plant species: Impatiens capensis Platanus occidentalis Pycnanthemum virginianum Salix interior Scirpus atrovirens Symplocarpus foetidus Typha latifolia Soil not saturated

Sampling point C-1 (9.8 m from baseline) Plant species: Carex hystricina Impatiens capensis Leersia oryzoides Pycnanthemum virginianum Salix interior Scirpus atrovirens Symplocarpus foetidus Typha latifolia Soil not saturated, relatively dry

Sampling point C-3 (22.8 m from baseline) Plant species: Comus amomum Galium sp. Impatiens capensis Leersia oryzoides Rosa sp. Salix interior Scirpus atrovirens Soil wetter than previous sampling point

Sampling point C-4 (31.5 m from baseline) Plant species: Carex spp. Comus sp. Impatiens capensis Leersia oryzoides Salix interior Scirpus acutus Symplocarpus foetidus Inundated soil

Sampling point C-5 (46.4 m from baseline) Plant species: Comus sp. Impatiens capensis Leersia oryzoides Salix interior 216 Symplocarpus foetidus Relatively drier soil

Sampling point C-6 (54.7 m from baseline) Plant species; Impatiens capensis Platanus occidentalis Ulmus rubra Viburnum prunifolium Soil dry

Sampling point C-7 (57.0 m from baseline) Plant species: Convolvulus arvensis Platanus occidentalis Ulmus rubra Viburnum prunifolium Soil dry

Sampling point C-8 (59.0 m from baseline) Plant species: Convolvulus arvensis Poa sp. Platanus occidentalis Ulmus rubra Viburnum prunifolium Soil dry

217