Ecotone Properties and Influences on Fish Distributions Along Habitat Gradients of Complex Aquatic Systems

ECOTONE PROPERTIES AND INFLUENCES ON FISH DISTRIBUTIONS ALONG HABITAT GRADIENTS OF COMPLEX AQUATIC SYSTEMS

A Dissertation Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

by Nuanchan Singkran May 2007

ECOTONE PROPERTIES AND INFLUENCES ON FISH DISTRIBUTIONS ALONG HABITAT GRADIENTS OF COMPLEX AQUATIC SYSTEMS

Nuanchan Singkran, Ph. D. Cornell University 2007

Ecotone properties (formation and function) were studied in complex aquatic systems in New York State. Ecotone formations were detected on two embayment- stream gradients associated with Lake Ontario during June–August 2002, using abrupt changes in habitat variables and fish species compositions. The study was repeated at a finer scale along the second gradient during June–August 2004. Abrupt changes in the habitat variables (water depth, current velocity, substrates, and covers) and peak species turnover rate showed strong congruence at the same location on one gradient. The repeated study on the second gradient in the summer of 2004 confirmed the same ecotone orientation as that detected in the summer of 2002 and revealed the ecotone width covering the lentic-lotic transitions. The ecotone on the second gradient acted as a hard barrier for most of the fish species. Ecotone properties were determined along the Hudson River estuary gradient during 1974–2001 using the same methods employed in the freshwater system. The Hudson ecotones showed both changes in location and structural formation over time. Influences of tide, freshwater flow, salinity, dissolved oxygen, and water temperature tended to govern ecotone properties. One ecotone detected in the lower-middle gradient portion appeared to be the optimal zone for fish assemblages, but the other ecotones acted as barriers for most fish species. A spatially explicit abundance exchange model (AEM) was developed to predict

distribution patterns of five fish species in relation to their population characteristics and habitat preferences across the lentic-lotic ecotones on the two freshwater gradients associated with Lake Ontario. Preference indexes of each target fish species for water depth, water temperature, current velocity, cover types, and bottom substrates were estimated from field observations, and these were used to compute fish habitat preference (HP). Fish HP was a key variable in the AEM to quantify abundance exchange of an associated fish species among habitats on each study gradient. The AEM efficiently determined local distribution ranges of the fish species on one gradient. Results from the model validation showed that the AEM was able to quantify most of the fish species distributions on the second gradient.

BIOGRAPHICAL SKETCH

Nuanchan Singkran was born in Chantaburi province located in the eastern part of Thailand. After finishing high school, Nuanchan entered the faculty of Science at Chulalongkorn University in 1989 for her undergraduate study in Marine Science with a major in Marine Biology. Study at the Marine Science Department allowed her to obtain broad knowledge of both applied science and ecology, especially the relationship between aquatic creatures and environmental conditions along the coastal zone of Thailand. After graduation, Nuanchan began her career as an environmental reporter for Thai and English newspapers during 1993–1996. During her career as a journalist, Nuanchan received a 2nd place award for an environmental article supporting Eco-Tourism of Thailand in 1993, and the award for outstanding environmental news reporter from the Environmental Siam Association in 1994. Nuanchan’s work in the media provided her a unique opportunity to observe various environmental problems in Thailand. She realized that the deterioration of natural resources in the country comes primarily from public lack of understanding and awareness about the environment. Lack of suitable methodologies to solve problems and weak environmental policy are additionally significant factors that aggravate environmental problems. Although Nuanchan likes the challenge of media work, she needed to be engaged more actively in the problem-solving process. She believed that academic and research work is really the area that would allow her to contribute to public education and in-depth research on specific issues.

During 1996–1999, Nuanchan obtained scholarships from the Chin Sophonpanich Foundation for Environment and from the Biodiversity Research Grant to further her Master’s program in the multidisciplinary Department of Environmental Science, Chulalongkorn University. Her Master’s thesis was entitled “Species Composition of

iii

Fish in Mangrove Canals as Reflected from Coastal Land Use at Trat Bay, Thailand”. Her interest in this topic comes from the fact that a large number of fish species (over 2,000 across the country) have been seriously threatened by human activities, especially the conversion of coastal areas into shrimp farms, urban districts, or industrial zones, all done without proper management. Her Master’s research using fish as a biological indicator revealed that improper land use activities altered the coastal environment, which in turn affected fish and other living communities in coastal systems in terms of both abundance and species diversity. After completing her Master’s degree, Nuanchan received a government scholarship to pursue her PhD program in Natural Resources at Cornell University in Ithaca, New York in 2001. At Cornell, Nuanchan conducted her dissertation research on influences of ecotone properties (formation and function) on species distributions over spatial and temporal scales. Her research work dealt with both empirical and simulation modeling to determine ecotone formation in complex aquatic systems and ecotone function on fish species assemblages. Additionally, she used a simulation model to spatially and temporally forecast distribution of fish species assemblages across aquatic boundaries.

For all fish and other creatures across global boundaries

ACKNOWLEDGMENTS

My graduate committee played vital roles throughout the development of my dissertation research. I am grateful to all of them for giving me the chance to freely design and follow my research ideas. My committee Chair, Dr. Mark B. Bain, not only supported me in any situation during my PhD program at Cornell University, but also shared his philosophy of thinking and was patient in shaping my research work to continue on the right track. Without his strong academic support and other help, my graduate study at Cornell University could not have been completed. As one of my committee, Dr. Patrick J. Sullivan always gave me instructive comments and helpful suggestions that greatly improved the quality of my research work. His simple questions, e.g., “Why is an ecotone is important?” and “How many functions can an ecotone have?” caused me to think deeply, carefully, and quantitatively on my ecotone study. Prof. Daniel P. Loucks is not just one of my committee, but also one of the best teachers I have ever had during my lifetime as a student. Although Prof. Loucks said that he did not know anything about fish and ecology, his contributions to my research work deserve praise. Following his suggestion to take some core courses in his department of Civil and Environmental Engineering, I gained more modeling experience and recovered my mathematical background to achieve my dissertation research. The Royal Thai Government and Thai people whose tax money supported me to study in the USA deserve any credit for my accomplishments. Additionally, my special thanks are due to the Lake Ontario Biocomplexity Project (Natural Science Foundation OCE–0083625), the Hudson River Foundation (GF/01/05), and the College of Agriculture and Life Sciences at Cornell University, which partly supported my study and research. I thank Prof. Paul A. Delcourt and Prof. Hazel R.

Delcourt for suggesting ecotone analysis; John Young from ASA Analysis & Communication, Inc. for providing the beach seine fish and habitat data from the Utilities’ Hudson River estuary monitoring program; John Ladd from New York State Department of Environmental Conservation for providing the substrate and shoreline data; and the Cornell University’s Geospatial Information Repository for providing submerged aquatic vegetation data. The completion of my graduate program at Cornell University would be impossible without the warm support and encouragement from friends and collaborators. I am thankful to Kristi Arend, Takao Kumazawa, Kathy Mills, James Murphy, Marci Meixler, Emelia Deimezis, Geof Eckerlin, Gail Steinhart, Carol Steinhart, Geoff Steinhart, and all the other people I cannot name here for their support and help during my study period at Cornell. I thank my parents and my husband, Poonrit Kuakul, for their patience and understanding. I am thankful to my senior friends, Dolina Millar, Bruce Anderson, Wandee Anderson, and some sisters and brothers from the Taste of Thai restaurant who made my life in Ithaca not so lonely and never refused to help me in any manner they could.

vii

TABLE OF CONTENTS Biographical sketch iii Dedication v Acknowledgments vi List of figures ix List of tables xii Chapter One Ecotone properties in the freshwater system associated 1 with Lake Ontario, New York.

Chapter Two Ecotone properties in the Hudson River estuary system 26 New York.

Chapter Three An abundance exchange model of fish assemblage 60 response to changing habitat along embayment- stream gradient of Lake Ontario, New York.

Appendix 2.1. Cumulative abundance and species richness of fish 93 by region (RE) along the Hudson River estuary gradient (RE1–RE12) from the five power plant utilities’ Hudson River estuary monitoring program using beach seine survey during 1974 to 2001.

Appendix 2.2. Multiple functions (aggregator, mediator, soft barrier, 105 and hard barrier) of the ecotones (from all frequencies of detection) on the Hudson fish species in each time period during 1974–2001.

Appendix 2.3. Multiple functions (neutral, partly inhibit, and 109 completely inhibit) of the ecotone peripheries (from all frequencies of detection) on the Hudson fish species in each time period during 1974–2001.

References 112

viii

LIST OF FIGURES

Figure 1.1. Ecotone structures and functions. 5

Figure 1.2. Sampling stations along the Sterling gradient during the 9 summer of 2002 (June–August). Observed value of each habitat variable was averaged over the three summer months and compared among stations using different sizes (small, medium, large) of bar charts to display mean magnitude (low, intermediate, or high).

Figure 1.3. Sampling stations along the Floodwood gradient during the 11 summer of 2002 (June–August). Observed value of each habitat variable was averaged over the three summer months and compared among stations using different sizes (small, medium, large) of bar charts to display mean magnitude (low, intermediate, or high).

Figure 1.4. Twelve stations (rectangles) were sampled along the 13 Floodwood gradient in the summer of 2004 (June–August) compared to the eight stations (circles) sampled in the summer of 2002. The ecotone periphery was indicated monthly by a peak absolute difference in the beta diversity (δ beta A) of fish species (June = connected line, July = dashed line, and August = gray line with squared marker) in each study summer.

Figure 2.1. The Hudson River estuary gradient showing 13 regions 31 designated for the five power plant utilities’ Hudson River estuary monitoring program.

Figure 2.2. Frequency of ecotones detected monthly based on the 39 data collection along the Hudson River estuary gradient during 1974–2001.

Figure 2.3. Ecotones (thick bars) and ecotone peripheries (thick lines) 40 detected during 1974–2001on the Hudson River estuary gradient covering 228 river kilometers (RKM) from region (RE) 1 to RE12 are grouped into each frequency class. Numbers of times of the formation of an ecotone and an ecotone periphery in each location are shown in brackets in each frequency class on the right side of the graph.

Figure 2.4. Monthly mean values of dissolved oxygen observed along 47 the Hudson River estuary gradient in the detected ecotone (line with solid dot markers) and non-detected ecotone (line with open dot markers) periods.

Figure 2.5. Monthly mean values of salinity observed along the 49 Hudson River estuary gradient in the detected ecotone (line with solid dot markers) and non-detected ecotone (line with open dot markers) periods.

Figure 2.6. Monthly mean values of water temperature observed 50 along the Hudson River estuary gradient in the detected ecotone (line with solid dot markers) and non-detected ecotone (line with open dot markers) periods.

Figure 2.7. Values of dissolved oxygen (DO), salinity, and water 52 temperature were averaged by region over the 116 time periods that the ecotones were detected. Bed areas of two submerged aquatic vegetation (SAV) species (Trapa natans and Vallisneria american) vary along the gradient (data for region 1 were unavailable). Proportions of 10 substrate types were observed in the tidal river estuary.

Figure 3.1. The Floodwood (FL) and Sterling (ST) gradients with eight 65 sampling stations (stars, FL; circles, ST) for monthly collecting of target fish and habitat variables during June– August 2002. The fish and habitat collections were repeated at 12 stations (circles) along the FL gradient during June– August 2004.

Figure 3.2. A conceptual diagram of fish distribution within and among 70 habitats. Migration rates of fish at two life stages in the growing season (yk, 0+ yr old; ak, 1+ yr old) and for all fish in winter (k) among habitats n, n–1, and n+1 are dependent on fish habitat preference in the growing season (yHP, 0+ yr old; aHP, 1+ yr old) and in winter (HP), respectively.

Figure 3.3. Categorized mean values of all habitat variables (except 77 substrate composition) observed monthly during June– August 2002 along the Floodwood gradient.

Figure 3.4. Categorized mean values of all habitat variables (except 79 substrate composition) observed monthly during June– August 2002 along the Sterling gradient.

Figure 3.5. Seasonal distribution pattern of yellow perch over the 81 100-year simulation (A). The fish population size exponentially decreased or increased for the first 10-year period of the simulation when changes of 50% of carrying capacity Cy (B), birth fraction rb (C), or death fraction rd (D) were assigned to the abundance exchange model, but the population size returned to the seasonal pattern over the long- term prediction.

Figure 3.6. Observed abundances of yellow perch (YP), bluegill (BG), 83 bluntnose minnow (BM), logperch (LP), and fantail darter (FTD) along the Floodwood gradient in August 2004 were compared with the mean predicted abundance ± standard deviation of the prediction of the associated fish species in August over the 100-year simulation.

Figure 3.7. Mean predicted abundance of yellow perch (YP), bluegill 85 (BG), bluntnose minnow (BM), logperch (LP), and fantail darter (FTD) among months and stations on the Floodwood gradient over the 100-year simulation.

Figure 3.8. Observed abundance of yellow perch (YP), bluegill (BG), 86 bluntnose minnow (BM), and fantail darter (FTD) along the Sterling gradient in August 2002 were compared with the mean predicted abundance ± standard deviation of the prediction of the associated fish species in August over the 100-year simulation.

Figure 3.9. Mean predicted abundance of yellow perch (YP), bluegill (BG), 88 bluntnose minnow (BM), and fantail darter (FTD) among months and stations on the Sterling gradient over the 100- year simulation.

LIST OF TABLES

Table 1.1. Intensity of each habitat variable was average the three 17 summer months (June–August) of 2002 along the Sterling and Floodwood gradients.

Table 1.2. Relative abundance comparison of each fish species 20 among the downstream (DW, FL1–FL3), ecotonal (ECO, FL4–FL5), and upstream (UP, FL6–FL12) habitats on the Floodwood gradient over the summer months of 2004 (June– August).

Table 2.1. Functions of the ecotones detected ≥5 times at certain 41 regions (RE) and time periods on Hudson fish species assemblages.

Table 2.2. Functions of the ecotone peripheries detected ≥5 times 43 at certain regions (RE) and time periods on Hudson fish species assemblages.

Table 2.3. Results from multivariate analysis of variance for each 46 habitat variable and its delta (absolute difference between minimum and maximum values) between non-detected (ECO = 0) and detected (ECO = 1) ecotone periods and due to the interactions of ECO × RE (river region), ECO × Month, ECO × Year, ECO × RE × Month, ECO × RE × Year, and ECO × Month × Year.

Table 3.1. Initial values of the population variables (Cy, rb, and rd) 73 and individuals of each target fish species for simulating the abundance exchange model.

Table 3.2. Preference index (Pi) for each habitat variable of five 75 modeled fish species at two life stages (0+yr and 1+yr old) estimated from the fish and habitat data observed along the Floodwood gradient over the three summer months (June– August) of 2004.

xii CHAPTER ONE

Ecotone properties in the freshwater system associated with Lake Ontario, New York

ABSTRACT

Ecotone formations were detected on two embayment-stream gradients associated with Lake Ontario using abrupt changes in habitat variables and peak turnover rates in fish species compositions along the gradients. The study was conducted on both gradients over the three summer months (June–August) of 2002, and it was repeated at a finer scale along the second gradient over the same summer months of 2004. The results revealed that the ecotones were static in their orientations on both gradients during the study periods. In the summer of 2002, abrupt changes in the habitat variables (water depth, current velocity, substrates, and covers) and peak species turnover rate showed strong congruence at the same location on one of the two gradients. On the second gradient, only the peak species turnover rate could clearly define the ecotone periphery, whereas the abrupt changes in the habitat variables occurred at multiple locations. The repeated study at the finer scale on the second gradient in the summer of 2004 confirmed the same ecotone orientation as that detected in the summer of 2002 and revealed the ecotone width covering the lentic- lotic transition. Four ecotone functions (aggregator, mediator, soft barrier, and hard barrier) were inferred from the comparison of relative abundance of fish species among downstream, ecotonal, and upstream habitats. In general, the ecotone acted as a hard barrier for most of the fish species. The distinct physical discontinuities below and above the ecotonal habitat appear to be an important factor shaping the ecotone geometry and impeding fish distribution into the ecotone.

1 INTRODUCTION

The ecotone concept is almost as old as the field of ecology (Zalewski et al. 2001) although ecologists have used the concept loosely on a broad range of boundary conditions. This practice has impeded the development of theory and the recognition of patterns that could advance the understanding and use of the ecotone idea (Strayer et al. 2003). Nevertheless, ecotone studies in ecosystems have remained prominent and appealing (e.g., Clements 1897, Leopold 1933, Odum 1971, Risser 1990, Johnson et al. 1992, Winemiller and Leslie 1992, Lidicker 1999, Fortin et al. 2000, Willis and Magnuson 2000, Harding 2002, Cadenasso et al. 2003, Martino and Able 2003, Miller and Sadro 2003, Ries et al. 2004) because of the important roles of ecotones on biodiversity (di Castri and Hansen 1992, Neilson et al. 1992, Lachavanne 1997), biotic interactions (Wiens et al. 1993), flows of materials and nutrients in ecosystems (Hansen and di Castri 1992), and the ecological implications for natural resources management at the individual, population, and ecosystem levels (Sisk and Haddad 2002). Clements (1897) was the first who viewed an ecotone as the transition zone between different plant communities. He observed that the vegetation in this zone was often accentuated in size and density. Leopold (1933) later explained that shape and spatial configuration of an ecotone influence species richness and organism abundance. Successive ecotone studies (e.g., Lay 1938, Johnston and Odum 1956) worked with these early ideas. Odum (1971) defined an ecotone as a transition zone between two or more communities, and he labeled the increases in both species richness and abundance at the ecotone as a consequence of the “edge effect”. At present, an ecotone is widely viewed as a transition zone between adjacent ecological systems with influences on organism distributions going beyond the traditional “edge

2 effect” (e.g., Holland 1988, Naiman and Décamps 1990, Fortin et al. 2000, Fagan et al. 2003, Strayer et al. 2003). From over 900 empirical ecotone studies in terrestrial ecosystems (Ries et al. 2004), ecotones provide diverse influences on abundance distributions of organisms in and around the ecotones. For instance, results from bird communities (Sisk and Margules 1993, Baker et al. 2002, Watson et al. 2004) showed that the expected increase in density and species richness in ecotones was not easily detected because bird species responded differently, i.e., edge avoiders, edge exploiters, and habitat generalists. Although distinct changes in bird assemblages were observed at an ecotone, conditions in the ecotone appeared to be neutral to some species. Birds have been widely used as ecotone response variables in terrestrial ecosystems (Ries et al. 2004) because of their mobility, diversity, and diverse habitat specializations. Fish are also mobile organisms that often display high diversities in species richness and life histories and, thus, appear to be an excellent ecotone indicator (Kirchhofer 1995, Schiemer et al. 1995) in aquatic ecosystems. However, fewer ecotone investigations have been conducted in aquatic ecosystems (e.g., Przybylski et al. 1991, Willis and Magnuson 2000, Martino and Able 2003). Whereas locations of most terrestrial ecotones can be visually identified (e.g., an ecotone between grassland and forest) before hypothesis testing, it is more difficult to determine locations of aquatic ecotones. In aquatic ecosystems, environmental conditions are complex in space and time due to variations in hydrology (Wiens 2002) and climate-linked seasonal changes that can alter other aquatic habitat variables (e.g., water temperature, growth of aquatic plants and algae). Consequently, ecotone properties (formation and function) in water systems are highly related to system variability, and they reflect the dynamics of organism assemblages responding to the environmental changes over space and time.

3 The objectives of this study were thus to detect the ecotone formation on two embayment-stream gradients and determine the ecotone function on fish species assemblages. An ecotone is viewed as a dynamic structure in an ecosystem that constitutes a unique habitat and possibly a unique community (Fagan et al. 2003). In this study, it is defined as a zone of transition in the physicochemical properties and interactions among habitats and species. An ecotone may form, decline, disappear, shift location, or change in structure or function due to the influence of changing physicochemical conditions, species use of habitat, and interactions among species (Holland 1988, Delcourt and Delcourt 1992, Wiens 1992, Kolasa and Zalewski 1995).

Relevant hypotheses and aquatic ecotone studies The status of an ecotone may differ depending on the spatial scale considered and the prior delineation of the system of interest (Kolasa and Zalewski1995). Consequently, scientists across fields of study may view ecotone structure differently under different terms, such as an edge, a transition zone, or a boundary (e.g., Strayer et al. 2003). Hence, it is necessary to clearly specify the spatial structure of an ecotone (Kolasa and Zalewski 1995, Cadenasso et al. 2003) in a study. Ecotone structure may be defined according to dimensionality, grain size, sharpness, or contrast. Based on dimensionality (Strayer et al. 2003), an ecotone may be considered either as a thin line between adjacent habitats or as a zone with the same dimension as its adjacent habitats (Figure 1.1a). An ecotone may be defined according to different grain sizes between adjacent habitats (Figure 1.1b) or sharpness of formation (Figure 1.1c) depending on changes in the ecological process across the ecotone (Cadenasso et al. 1997, Bowersox and Brown 2001, Cadenasso et al. 2003, Fagan et al. 2003, Strayer et al. 2003). Lastly, an ecotone may form between adjacent habitats with high or low habitat contrast (Figure 1.1d; Cadenasso et al. 2003, Strayer et al. 2003).

Figure 1.1. Ecotone structures and functions. Ecotone structure may be defined based on dimensionality (a.1: a line, a.2: a broader zone), grain size (b.1: a fine grain size, b.2: a coarse grain size), sharpness (c.1: a sharp formation, c.2: a gradual formation), or contrast (d.1: high contrast, d.2: low contrast). As in a.2, an ecotone may function as an aggregator attracting some species into the ecotone, a hard barrier completely blocking some species into the ecotone, a soft barrier partially allowing some species to enter the ecotone, or a mediator having no effect on flux of certain species. (Figures 1a–1d are reproduced from Strayer et al. 2003, A classification of ecological boundaries, Copyright, American Institute of Biological Sciences.)

5 In this study, ecotone structure was considered as a zone having the same dimension as its adjacent habitats, and its formation was detected on each study gradient using both habitat changes and fish species turnover rate as indicators. These are based on the hypothesis that an aquatic ecotone can be determined using abrupt changes in either habitat conditions or fish species composition. An ecotone generally provides diverse permeability (Wiens 1992, Ross et al. 2005) on flux of organisms. This role is particularly significant for mobile animals, as they can migrate to their optimal habitats in and around an ecotone. According to various permeability of an ecotone on species distributions, an ecotone may simultaneously facilitate, inhibit, or have no effects on distributions of multiple species into the ecotone (Wiens et al. 1985, Wiens 1992, Kolasa and Zalewski 1995, Haddad 1999, Lidicker 1999, Ries and Debinski 2001, Matthysen 2002, Strayer et al. 2003). Four hypothetical functions of aquatic ecotones (Figure 1.1) on fish species distributions along the study gradients were proposed as follows. An ecotone may act as an aggregator for particular species by attracting them into the ecotone. Alternately, it may act as a hard barrier by completely blocking distributions of some other species into the ecotone, or as a soft barrier by partially allowing some species to enter the ecotone. In the latter case, most species populations entering the ecotone will return to the habitats from which they came, causing few of them to be observed in the ecotone. Lastly, an ecotone may act as a mediator, having no effect on the flux of particular species between the ecotone and the adjacent habitat(s). Some previous studies reported about the influences of aquatic ecotones in particular ecosystem settings and the way changing characteristics of habitat and species interact. For instance, the transitions between downstream stations and Lake Balaton (Hungary) acted as the zones aggregating the highest fish density and species richness (Przybylski et al. 1991). Similarly, Willis and Magnuson (2000) studied

6 patterns in fish species composition across the interface between lakes and streams in Wisconsin and concluded that both species richness and density of fish species composition were the highest at the lake-stream ecotones. Their study suggested that differences in the hydrologic and geomorphic properties of habitat strongly influence the patterns of change in fish communities. Features of aquatic habitats, such as water depth, current velocity, substrates, and covers, have been reported as important factors influencing fish distribution and species richness (Zalewski et al. 2001). Rakocinski (1992) revealed that water depth, emergent stem density, salinity, water temperature, and dissolved oxygen influenced both temporal and spatial changes in the community structure of marsh-edge fishes across the marsh-open water ecotone in a Louisiana estuary. Environmental variables related to habitat size and salinity showed the greatest correspondence with the changes in the structures of fish assemblages across a marine-freshwater ecotone in Tortuguero National Park on the Caribbean coast of Central America (Winemiller and Leslie 1992). In addition, the variability in species richness and abundance of fish assemblages in and around an ocean-estuarine ecotone on the river-bay-ocean gradient in southern New Jersey were influenced by habitat heterogeneity, water depth, and the fish species responses to the environmental gradient (Martino and Able, 2003). Miller and Sadro (2003) reported an alternative basis of an estuary-stream ecotone in Oregon as a habitat that aggregated juvenile anadromous fish during their physiological transition.

7 MATERIALS AND METHODS

Study gradients and field observations Field observations were conducted along two embayment-tributary stream gradients, Sterling (ST) and Floodwood (FL), located on the southeastern coast of Lake Ontario, New York. The ST gradient includes Sterling Pond and Sterling Creek. The total watershed area associated with the ST gradient was 210.76 km2, while the surface area of Sterling Pond alone was 0.38 km2. Sterling Creek ranges in width from 3 m (upstream) to more than 20 m (entering Sterling Pond) with a mean annual discharge of 1.9 m3/s (37-year record—USGS gauge 04232100 above the study area). The FL gradient comprises Floodwood Pond and Sandy Creek. Total watershed area for the FL gradient was 671.82 km2, while the surface area of Floodwood Pond alone was 0.078 km2. Sandy Creek ranges in width from 6 m (upstream) to more than 50 m (entering Floodwood Pond) with a mean annual discharge of 7.8 m3/s (37-year record—USGS gauge 04250750 above the study area). A Garmin Etrex Venture global positioning system was used to mark the locations of sampling stations on both gradients. In the summer months (June–August) of 2002, eight transect stations on the ST gradient were selected (between 43°20'34.5" N, 76°41'54.7" W and 43°16'51.6" N, 76°37'37.6" W) using a stratified random sampling method to cover all habitat types. The sampling stations (ST with number) were near the connection between Lake Ontario and Sterling Pond (ST1), the connection between Sterling Pond and the mouth of Sterling Creek (ST2), the downstream (ST3 and ST4), the midstream (ST5 and ST6), and the upstream (ST7 and ST8) of Sterling Creek (Figure 1.2). The length of the entire sampling gradient was about 16.3 km.

Figure 1.2. Sampling stations along the Sterling gradient during the summer of 2002 (June–August). Observed value of each habitat variable was averaged over the three summer months and compared among stations using different sizes (small, medium, large) of bar charts to display mean magnitude (low, intermediate, or high). Four substrate types were mud (Md), sand (Sd), gravel-cobble (Gv-Cb), and rock-bedrock (Rb). The three graphs below show the turnover rate in fish species composition (Beta A—connected line) and the peak absolute difference in the beta A, (δ beta A—dashed line) along the gradient in the associated months of observation.

9 The same stratified random method was used to select eight transect stations (between 43°43'16" N, 76°12'10" W and 43°48'47" N, 76°4'31" W) on the FL gradient. These sampling stations were at the connection between Lake Ontario and Floodwood Pond (FL1), the connection between Floodwood Pond and the mouth of Sandy Creek (FL2), the downstream (FL3, FL4, and FL5), the midstream (FL6), and the upstream (FL7, FL8) of Sandy Creek (Figure 1.3). The length of the entire sampling gradient was about 18.6 km. Both abundance and species richness of fish assemblages were observed once a month in the summer of 2002 (June–August) along the FL and ST gradients. Fish were sampled with electrofishing gear at all FL and ST stations. At the sampling stations where water depth was ≥1.0 m (ST1–ST6, FL1–FL4), the power supply gears (a 3500-W generator and a transformer—the DC voltage output controller of 250–350 V) and the electrode array were installed on a 4.9-m shallow draft, flat bottom boat. The boat hull served as the cathode array, and two 1-m-diameter rings with dropper anodes were suspended in the water off the front of the boat. A 20-m-diameter circular area was repeatedly electrofished for 15 minutes per station. While catching fish, the boat speed was controlled, as closely as possible, at the average walking speed of people, 4.85 km/hr (Fruin 1971), and this speed is assumed to be similar to the walking pace of the fieldwork crew in shallow water stations (ST7–ST8 and FL5– FL8, <1.0 m deep). Because the boat could not access the shallow water stations, the generator and transformer for generating electricity with the same DC voltage output as used on the boat were instead installed on shore. One end of a cathode arraying with 4-m steel cables was connected to the transformer and its other end was centrally positioned in a sampling station and occasionally repositioned. Using this configuration, the power supply gears generated a strong electric current across the sampling station.

Figure 1.3. Sampling stations along the Floodwood gradient during the summer of 2002 (June–August). Observed value of each habitat variable was averaged over the three summer months and compared among stations using different sizes (small, medium, large) of bar charts to display mean magnitude (low, intermediate, or high). Four substrate types were mud (Md), sand (Sd), gravel-cobble (Gv-Cb), and rock- bedrock (Rb). The three graphs below show the turnover rate in fish species composition (Beta A—connected line) and the peak absolute difference in the beta A, (δ beta A—dashed line) along the gradient in the associated months of observation.

11 For catching fish, one end of a 100-m electricity cable was connected to the transformer. The other end of the cable was connected to a handheld steel ring (0.5-m diameter) anode, and it was maneuvered through the sampling station for manually shocking fish. A 60-m-long channel segment was sampled from downstream to upstream by repeating passes for 15 minutes in each station. Juveniles and adults of each fish species caught at each station were enumerated and their total length was measured to the nearest millimeter. In the same summer months of 2002, the habitat variables were measured once a month in all stations where the fish species were collected, as follows. Water depth

(m) and water temperature (°C) were randomly measured five times in each sampling station. Current velocity (m/s) was randomly measured five times within each station using a Marsh-McBirney 2000-model (Bovee 1982). Bottom substrates (%) were visually observed once a month in all sampling fish stations on each study gradient; at the first observation, substrate grain sizes (diameter) were measured and classified into four types modified from Pusey et al. (1993): mud (<1 mm diameter), sand (1–16 mm diameter), gravel-cobble (16–128 mm diameter), and rock-bedrock (>128 mm diameter). The average percentage of each substrate type and average abundance of each cover type was calculated from their 10 random observations in each station. The abundance of aquatic plants, algae, and woody debris was coded on a scale of 0– 3: 0 = absent, 1 = low, 2 = common, and 3 = high. In the three summer months (June–August) of 2004, the fish collections were repeated at 12 sampling stations (FL1–FL12; Figure 1.4: squares) along the FL gradient. The 12 sampling stations were located between 43°43'20.6" N, 76°12'13.3" W and 43°51'11.7" N, 75°56'44" W with a total sampling gradient length of 28.2 km. Within each of the 12 sampling stations, three to six locations were randomly chosen for monthly collecting abundance and species richness of fish on a finer scale. In

12 other words, the fish assemblages were collected monthly at a total of 60 locations within the 12 stations along the FL gradient. With the finer scale of sampling, juveniles and adults of fish assemblages were collected for 10 minutes at each of the 60 locations using the same fishing gear and sampling protocols as used in the summer of 2002. Individuals of each fish species collected from all locations in each station were accumulated, enumerated, and measured for their total length to the nearest millimeter.

Figure 1.4. Twelve stations (rectangles) were sampled along the Floodwood gradient in the summer of 2004 (June–August) compared to the eight stations (circles) sampled in the summer of 2002. The ecotone periphery was indicated monthly by a peak absolute difference in the beta diversity (δ beta A) of fish species (June = connected line, July = dashed line, and August = gray line with squared marker) in each study summer. The ecotone comprises the lower ecotone periphery (e1), the ecotonal (ECO) habitat, and the upper ecotone periphery (e2). According to the ecotone orientation detected in the summer of 2004, the major downstream (DW) habitat was from stations 1 to 3, and the major upstream (UP) habitat was from stations 4 to 5.

13 Ecotone analysis Turnover rate or beta diversity (Whittaker 1975, Gauch 1982, Delcourt and Delcourt 1992, Winemiller and Leslie 1992, Wagner 1999, Hoeinghaus et al. 2004) in species composition of fish was used to detect ecotone formation on each study gradient in the summers of 2002 and 2004. Detrended correspondence analysis (DCA) in CANOCO 4.0 software (ter Braak and Smilauer 1998) was used to compute the beta diversity (hereafter called beta A) based on both occurrence and abundance of fish. In the computation, the sample scores on the DCA ordination axes are rescaled into the average standard deviation (SD) units of species turnover (Hill 1979). A gradient length of 1.0 SD represents approximately one-fourth of a complete turnover rate in species composition on the gradient (Gauch 1982). Species appear, rise to their modes, and disappear over a distance of about 4 SD. That is, if the beta A on the

DCA’s axis 1 (that accounts for the most variation) is ≥4.0 SD, it indicates that species compositions at both extreme ends of a gradient represent completely different communities (Gauch 1982). This indicates complete formation of at least one ecotone on the gradient. The peak absolute difference in the beta A (δ beta A) indicates an ecotone periphery (a margin where a sharp transition between adjacent habitats occurs). Abrupt changes in the study habitat variables, the other ecotone indicator, were compared among stations, months, and the station-month interactions along each gradient in the summer of 2002 using a repeated measures 2-way ANOVA in SPSS version 13 software. The habitat variables with significant differences among stations (p≤0.05) were further analyzed using Tukey’s honestly significant differences (HSD) test. The statistical results showing patterns of variations in the habitat variables along each gradient in the summer of 2002 were summarized graphically in low, medium,

14 and high intensity levels to allow visual synthesis of habitat and fish assemblage shifts along the gradients (Figures 1.2 and 1.3). With the finer scale of the field observations along the FL gradient in the summer of 2004, the abundance by species of the fish assemblages observed in and around the FL ecotone was used to interpret ecotone function. In the process, each fish species with ≥5 individuals observed on the FL gradient over the three summer months (June– August) was selected to compare its percentage abundance among three major habitats: ecotonal (ECO—area within the ecotone peripheries), downstream (DW), and upstream (UP) habitats. The four hypothetical ecotone functions (aggregator, mediator, soft barrier, and hard barrier) were then interpreted from relative abundances of fish among the three major habitats. This study proposed that an ecotone acts as an aggregator for a fish species with >60% abundance observed in the ECO habitat and as a mediator for fish species with 30–60% abundance observed in the ECO habitat. Concurrently, an ecotone acts as a soft barrier for a fish species with 5–30% abundance observed in the ECO habitat. Lastly, an ecotone acts as a hard barrier for a fish species with <5% abundance observed in the transitional habitat.

RESULTS

Ecotone formation In the summer of 2002, 2,071 individual fish from 39 species were collected along the ST gradient. Both fish abundance and species richness were used to estimate the fish species turnover rate (beta A) along the gradient. Results from the DCA analysis showed that the beta A was 2.1 SD, 12.6 SD, and 12.8 SD in June, July, and August, respectively (Figure 1.2). The beta A values >4.0 SD in July and August indicated the complete changeover in fish species composition on the gradient in these months. In

15 June, no change (the beta A = 0) in species composition of fishes was detected from ST1 to ST6 (Figure 1.2), but a clear increase of the beta A occurred between ST6 and

ST7. As a result, a peak of an absolute difference in the beta A (δ beta A) was detected between ST6 and ST7 in all months indicating this location as a persistent zone of transition in fish species composition on the ST gradient. Significant habitat variations along the ST gradient were detected among stations, months, and station-month interactions (the repeated measures 2-way ANOVA, p≤0.05). Significant differences among stations were found for all habitat variables except possibly current velocity (p=0.065). Habitat differences across the three sampling months were minor and significant only for current velocity, algae, and aquatic plants. These three habitat variables were the only ones showing significant station-month interactions, which were expected because of plant growth and declining flows during the study period. The intensity values of all habitat variables were averaged over the three summer months (Table 1.1), and they were graphically compared among the ST stations (Figure 1.2) based on the results from Tukey’s HSD post hoc. Consistent with the peak transition in fish species composition, abrupt changes of all habitat variables except current velocity were also detected between ST6 and ST7, reflecting this location as a zone of transition in habitat conditions as well (Figure 1.2). In the 2002 summer, 1,988 individual fish from 42 species were collected along the FL gradient, from which the turnover rate in fish species composition on this gradient was estimated. Results from the DCA analysis revealed the complete fish turnover rate on the FL gradient over the study period according to the beta A value (Figure 1.3) which exceeded 4.0 SD in June (6.3 SD), July (4.8 SD), and August (5.4

SD). The peak δ beta A occurred between FL5 and FL6 over the three summer months marking this location as a distinct zone of transition in fish species

16 composition on the FL gradient. Significant differences among stations were found for all habitat variables on this gradient (the repeated measures 2-way ANOVA, p≤0.05). Additionally, changing intensities of all habitat variables except substrate composition were significantly different among the sampling months.

Table 1.1. Intensity of each habitat variable was averaged over the three summer months (June–August) of 2002 along the Sterling and Floodwood gradients.

Station Habitat variable 12345678

Sterling gradient Water depth (m) 1.71 2.06 2.72 2.92 2.39 1.66 0.33 0.48 Current velocity (m/sec) 0.11 0.12 0.11 0.11 0.19 0.18 0.22 0.20 Substrate (%) mud 90 90 100 100 90 75 2 0 sand 106005104725 gravel-cobble 0400384550 rock-bedrock000027625 Cover1 algae 1.17 2.30 2.43 1.97 1.87 2.10 1.07 1.00 aquatic plants 1.97 2.27 1.90 2.43 2.03 2.10 0.30 0.37 woody debris 0.00 0.23 0.00 0.00 0.60 0.73 1.73 1.70

Floodwood gradient Water depth (m) 2.69 1.94 2.20 1.50 0.54 0.17 0.46 0.44 Current velocity (m/sec) 0.19 0.17 0.15 0.15 0.28 0.34 0.25 0.27 Substrate (%) mud 90100100125000 sand 10 0 0 48 14 5 5 5 gravel-cobble 0 0 0 40 40 35 45 46 rock-bedrock000041605049 Cover1 algae 0.80 1.47 0.10 0.10 1.27 1.67 1.37 0.97 aquatic plants 1.43 2.23 0.93 0.73 0.93 0.20 0.33 0.53 woody debris 0.33 0.73 1.27 0.67 0.20 0.13 0.80 0.60

1Occurrence and abundance of each cover type was coded: 0 = absent, 1 = low, 2 = common, and 3 = high.

17 Similar to the patterns observed on the ST gradient, current velocity, algae, and aquatic plants were significantly different by station-month interactions on the FL gradient. Again, these were expected due to plant growth and declining flows during the study period. The intensity values of all habitat variables were averaged over the three summer months of 2002 (Table 1.1), and they were graphically compared among the FL stations (Figure 1.3) based on the results from Tukey’s HSD post hoc. While the peak fish turnover rate was detected between FL5 and FL6 over the study period, abrupt changes in the habitat variables mostly occurred at the multiple stations (Figure 1.3). Variability and inconsistency of habitat changes along the FL gradient made it difficult to use the habitat variables to clearly define ecotone location on this gradient. Abrupt changes in both habitat variables and fish species composition between ST6 and ST7 on the ST gradient indicated the ECO habitat between the two stations. With respect to the peak turnover rate in fish species composition that was consistently detected between FL5 and FL6 over the three summer months, the ECO habitat on the FL gradient was most likely to be situated between these two stations. The field observations were repeated at a finer scale along the FL gradient in the summer of 2004 to: 1) test if the ecotone would be detected at the same location on the gradient with the complex mix of habitat changes during the summer period and 2) determine the ecotone width and ecotone functions. Over the three summer months (June–August) of 2004, 11,318 individual fish from 43 species were collected from 12 stations along the FL gradient (Figure 1.4: squares). The complete changeover in fish species composition occurred over the summer of 2004 according to the beta A values of 5.0, 4.4, and 4.2 SD in June, July, and August respectively. With the finer scale of the study, two peaks of the δ beta A were detected. One was between FL5 and FL6 over the three months, while the other was between FL3 and FL4 in July and August (Figure 1.4: Summer 2004). Results

18 from both summer studies revealed that the ECO habitat on the FL gradient was situated within the two ecotone peripheries (i.e., FL3–FL4 and FL5–FL6 peripheries) and had the width from FL4 to FL5 (Figure 1.4). In relation to the ecotone orientation (the ECO habitat plus its ecotone peripheries), a DW habitat was from FL1 to FL3 and an UP habitat was from FL6 to FL12 (Figure 1.4).

Ecotone function Relative abundances of fish species were compared among the DW, ECO, and UP habitats to interpret ecotone function on fish species distribution along the FL gradient in the summer of 2004. Of the total 43 fish species, 31 species with ≥5 individuals collected along the gradient over the three summer months were selected for comparison (Table 1.2). The FL ecotone acted as an aggregator for bluntnose minnow (Pimephales notatus), brook stickleback (Culaea inconstans), largemouth bass (Micropterus salmoides), logperch (Percina caprodes), smallmouth bass (Micropterus dolomieui), tessellated darter (Etheostoma olmstedi), and white sucker (Catostomus commersoni). Each of these species was observed >60% in the ECO habitat. The ecotone acted as a mediator for common shiner (Notropis cornutus), fathead minnow (Pimephales promelas), pumpkinseed (Lepomis gibbosus), and rock bass (Ambloplites rupestris). These species were observed 30–60% in the ECO habitat. The ecotone acted as a soft barrier for bluegill (Lepomis macrochirus), creek chub (Semotilus atromaculatus), cutlips minnow (Exoglossum maxillingua), golden shiner (Notemigonus crysoleucas), longnose dace (Rhinichthys cataractae), and yellow perch (Perca flavescens). These fish species showed 5–30% abundance in the ECO habitat. The ecotone acted as a hard barrier for the remaining 14 fish species (Table 1.2), each of these species showed <5% abundance in the ECO habitat.

19 Table 1.2. Relative abundance comparison of each fish species among the downstream (DW, FL1–FL3), ecotonal (ECO, FL4–FL5), and upstream (UP, FL6– FL12) habitats on the Floodwood gradient over the summer months of 2004 (June– August).

Abundance (%) by habitat Total Fish species 1 DW ECO %UP individuals

Alewife (Alosa pseudoharengus ) 100 0 0 7 Black crappie (Pomoxis nigromaculatus ) 100 0 0 13 Blacknose dace (Rhinichthys atratulus )0298885 Blacknose shiner (Notropis heterolepis ) 100 0 0 7 Bluegill (Lepomis macrochirus )78220521 Bluntnose minnow (Pimephales notatus ) 6 69 25 466 Bowfin (Amia calva )950510 Brook stickleback (Culaea inconstans )096415 Brown bullhead (Ictalurus nebulosus )8901174 Comely shiner (Notropis amoenus ) 100 0 0 13 Common carp (Cyprinus carpio ) 100 0 0 25 Common shiner (Notropis cornutus ) 6 53 41 315 Creek chub (Semotilus atromaculatus ) 0 11 89 473 Cutlips minnow (Exoglossum maxillingua ) 0 20 80 1,736 Fantail darter (Etheostoma flabellare )0298826 Fathead minnow (Pimephales promelas ) 0 39 61 45 Golden shiner (Notemigonus crysoleucas )83170 41 Largemouth bass (Micropterus salmoides )26740 334 Logperch (Percina caprodes )991074 Longnose dace (Rhinichthys cataractae )09912,000 Northern pike (Esox lucius ) 100 0 0 10 Pumpkinseed (Lepomis gibbosus )65350531 Redside dace (Clinostomus elongatus ) 0 0 100 19 Rock bass (Ambloplites rupestris ) 125830313 Sand shiner (Notropis stramuneus ) 0 0 100 28 Smallmouth bass (Micropterus dolomieui )18820 16 Spottail shiner (Notropis hudsonius ) 100 0 0 13 Stonecat (Noturus flavus )709332 Tessellated darter (Etheostoma olmstedi ) 4 90 6 307 White sucker (Catostomus commersoni ) 0 82 18 1,884 Yellow perch (Perca flavescens ) 77 23 0 255

1Cumulative individuals of each fish species collected along the gradient over the three summer months.

20 DISCUSSION

The peak turnover rates in fish species compositions indicated the ecotone peripheries between ST6 and ST7 on the ST gradient and between FL5 and FL6 on the FL gradient over the three summer months of 2002. However, using the abrupt changes in habitat variables as the ecotone indicator yielded different results between the two gradients. The congruence of the abrupt changes in habitat conditions and fish species composition at the ecotone periphery on the ST gradient revealed a strong fish- habitat relationship (e.g., Przybylski et al. 1991, Winemiller and Leslie 1992, Rakocinski et al. 1992, Wagner 1999, Willis and Magnuson 2000, Gido et al. 2002, Martino and Able 2003) across lentic-lotic transitions on the gradient. This consequence conformed to the hypothesis that the ecotone formation could be detected using abrupt changes in either habitat conditions or fish species composition. In contrast, the similar pattern of coincidental abrupt changes in both fish species composition and the habitat conditions was not detected on the FL gradient. While the peak turnover rate in fish species composition was detected at the same location, the changes in most of the habitat variables varied along the FL gradient. Using habitat condition to detect ecotone formation is difficult when the changes in the habitat variables are gradual (Hansen et al. 1988). In the summer of 2002, average intensities of many habitat variables showed gradual or no changes at the FL ecotone periphery. The complex mix of gradual and abrupt changes in habitat conditions along the FL gradient diminished the efficiency of this indicator to define ecotone location. From this result, the peak turnover rate in fish species composition appears to be the most powerful indicator for depicting the FL ecotone. This is based on the ideas that changes in species assemblages across a transition zone can

21 apparently reflect both gradual and abrupt changes in habitat conditions (Armand 1992, Strayer et al. 2003, Hoeinghaus et al. 2004). This study revealed that there were no shifts in ecotone location over the summer periods. The ecotone peripheries (that indicated the ecotone orientation between lentic and lotic systems) on both gradients were static in their locations throughout each study summer period. The fix in location of the ecotones may be due to environmental discontinuities (Pinay et al. 1990, Delcourt and Delcourt 1992). In this study, unidirectional water flow from upstream towards downstream on the gradients promotes mixing of materials and results in spatiotemporal heterogeneities and physical discontinuities (Ward and Stanford 1983, Pinay et al. 1990) at the particular gradient portions. The strong physical discontinuities (e.g., abrupt changes in water depth, current velocity, substrate composition, and stream geomorphology) can contribute to the stable formation of ecotones over space and time (Pinay et al. 1990). Previous studies on similar gradient patterns by Przybylski et al. (1991) and Willis and Magnuson (2002) concluded that river mouths connecting with lakes appeared to be the ecotones with the highest abundance and species richness of fish observed, and these conformed to the edge effect (Odum 1971). However, no quantitative methods were used to define the ecotone formation in those studies. For this study, the ecotone formation on both gradients was defined using statistical methods. The orientations of both ST and FL ecotones were detected between the downstream and midstream stations where the transitional conditions between lentic and lotic systems occurred instead of at the stations between Lake Ontario and the embayments. These findings indicated that differences in the sizes of the study gradients, habitat conditions, or geographical regions (Kolasa and Zalewski 1995) could also contribute to the different formation patterns and characteristics of the aquatic ecotones.

22 Methods of collecting fish species assemblages are another factor that can influence the determination of the ecotone locations on the lentic-lotic gradients. From this study, the ecotone periphery on the ST gradient was detected at the location where the sampling procedure was changed from boat to manual electrofishing. However, the ecotone periphery on the FL gradient was detected at the location where the same sampling procedure was used (i.e., manual electrofishing). Although the change from boat to manual electrofishing between the deep and shallow stations was unavoidable in this study, the same electricity output was generated for stunning fish with a similar pace of locomotion in all stations, and thorough samplings were conducted to cover all habitat types in each station. From these efforts, the method constraint for collecting fish between the deep and shallower stations on the gradients might only trivially affect the results. Additionally, Rakocinski et al. (1992) indicated that differences in fish species assemblages along environmental gradients represent species-specific differences in habitat use rather than gear bias. Ecotone distinctiveness can be scale dependent (Johnston et al. 1992, Wiens 1992, Fagan et al. 2003), thus a proper spatial sampling design is required to determine an ecotone (ecotone peripheries plus ecotone width—ECO habitat). For example, an ecotone that is very abrupt on a community scale may be too small to be detected on a continental scale because the resolution of data collection (on the continental scale) is too broad (Johnston et al. 1992, Fagan et al. 2003). Hence, the scale of data observation must be commensurate with the scale of an ecotone in a study (Johnston et al. 1992, Wiens 1992, Fortin et al. 2000, Csillag et al. 2001, Cadenasso et al. 2003). Using the fish species turnover rate as the ecotone indicator, the results from the field observations along the FL gradient in both summers of 2002 and 2004 confirmed that the FL ecotone was static at the same location during the summer periods. With the finer and proper scale of study in the second summer, the entire ecotone was

23 successfully determined, and it was situated within the ecotone periphery detected in the summer of 2002 (Figure 1.4). Conditions in habitats adjacent to an ecotone play a strong role in governing the ecotone geometry and permeability (Stamps et al. 1987, Wiens 1992, Fortin 1994, Fagan et al. 2003). In this study, the ECO habitat on the FL gradient represents a mixed zone of transition between the lentic and lotic systems. Based on the ecotone orientation, the lower portion of the ECO habitat is most likely to be influenced by the transition between the embayment and downstream conditions at the lower ecotone periphery (between FL3 and FL4, Figure 1.4: squares). Likewise, the upper portion of the ECO habitat tended to be affected by the transition between downstream and midstream conditions at the upper ecotone periphery (between FL5 and FL6, Figure 1.4: squares). Consequently, the ECO habitat appears to be a unique area providing diverse permeability (Wiens et al. 1985, Wiens 1992, Cadenasso et al. 1997, Ross 2005) for the distribution of fish species assemblages on the gradient. Of the four ecotone functions (i.e., aggregator, mediator, soft barrier, and hard barrier) on fish species assemblages, the “edge effect” of an increase in fish abundance and species richness (Odum 1971) was not observed at the FL ecotone. In contrast, the ecotone acted as a hard barrier allowing very low permeability for most of the fish species observed in the summer of 2004 (Table 1.2). The physical discontinuities in space or patch contrast appear to be another important factor affecting the species distribution (Pinay et al. 1990, Forman and Moore 1992, Wiens 1992, Wiens 2002). From the field observations, small cascades and long runs existing above the upper ecotone periphery possibly obstruct the distribution of some stream fish species into the ecotone. In addition, the decrease of water depth observed at the lower ecotone periphery acted as a hard obstacle for many fish species to migrate into the ecotone. While the FL ecotone provided low-to-zero permeability for most fish species, it was

24 concurrently a nursery ground for seven species (i.e., bluntnose minnow, brook stickleback, largemouth bass, logperch, smallmouth bass, tessellated darter, and white sucker). These particular species aggregated (>60%) in the ECO habitat during the study summer, and most of the individual fish by species were young (0+ year-old). The fundamental knowledge about ecotone formations on the two embayment-stream gradients and the ecotone functions on fish species assemblages are useful for better understanding ecotone properties in aquatic systems and future applications in resource management (e.g., Petts 1990, Sisk and Haddad 2002) and fish habitat restoration (Kirchhofer 1995, Kolasa and Zalewski 1995). Overall, the results from this study show that the turnover rate in fish species composition appears to be the best ecotone indicator on the embayment-stream gradients. While abrupt changes in habitat conditions and species composition may coincidentally occur at the same location reflecting the strong fish-habitat relationships, the habitat variables showed less efficiency in defining the ecotone formation on the gradient with the complex mix of habitat conditions. An appropriate scale of study is required to determine ecotone width. The distinct physical discontinuities below and above the ecotone detected on the FL gradient contributed to low permeability of the ecotone on flux of the fish species. This study is the first effort of its type using both biotic and non-biotic variables to detect ecotone formation in the freshwater system associated with Lake Ontario. To explore ecotone properties in a different water system, both functions and formations of ecotones were detected along the tidal Hudson River, one of the world’s largest marine-freshwater gradients. The methodology of study, results, and discussion are presented in Chapter 2.

25 CHAPTER TWO

Ecotone properties in the Hudson River estuary, New York

ABSTRACT

Ecotone properties were determined on the Hudson River estuary gradient during 1974–2001. Turnover rate in fish species composition was used to detect ecotone formations and abundances of fish species assemblages in and around ecotones were used to interpret ecotone functions. From the results, the Hudson ecotones showed both changes in location and structure spatially and temporally. In some time periods, the ecotones changed in their breadth (from broader to narrower or vice versa) or considerably declined to be just ecotone peripheries. The ecotone peripheries mostly detected at the lower gradient portion might result from the sharp changes in dissolved oxygen (DO) and water salinity within this gradient range. The ecotone peripheries mostly detected at the upper gradient portion might result from the abrupt changes in DO, water temperature, SAV abundance, and dam constructions within this gradient range. The ecotones were highly detected at the lower-middle portion of the gradient and their formations tended to relate with the changes in salinity, tide, and freshwater flow. Functional patterns of the Hudson ecotones were not distinctly different over time. One ecotone mostly detected in the lower-middle gradient portion appeared to be the optimal zone for distributions of diverse fish assemblages. However, the other detected ecotones acted as barriers for most fish species. Likewise, the ecotone peripheries inhibited distributions of most fish species. These results might relate with complex migratory behaviors and life histories of fish and their responses to variations in the estuarine conditions in space and time.

26 INTRODUCTION

Ecotone studies have been prominent and appealing (e.g., Clements 1897, Leopold 1933, Odum 1971, Risser 1990, Winemiller and Leslie 1992, Lidicker 1999, Willis and Magnuson 2000, Harding 2002, Martino and Able 2003, Ries et al. 2004) because of the important roles of ecotones on biodiversity (di Castri and Hansen 1992, Neilson et al. 1992, Lachavanne 1997), biotic interactions (Wiens et al. 1993), flows of materials and nutrients in ecosystems (Hansen and di Castri 1992), and the ecological implications for natural resources management at the individual, population, and ecosystem levels (Sisk and Haddad 2002). Clements (1897) was the first who viewed an ecotone as the transition zone between different plant communities. He observed that the vegetation in this zone was often accentuated in size and density. Leopold (1933) later explained that shape and spatial configuration of an ecotone influence species richness and organism abundance. Working with these early ideas, Odum (1971) defined an ecotone as a transition zone between two or more communities and he labeled the increases in both species richness and abundance at the ecotone as a result of the “edge effect”. At present, an ecotone is widely viewed as a transition zone between adjacent ecological systems and its influences on organism distributions have been considered beyond the traditional edge effect. Additionally, scientists may use diverse terms to call a transition zone depending on the spatial scale considered and the prior delineation of the system of interest. In this study, an ecotone periphery is used to represent a narrow transition zone or a boundary with one less dimension than a patch (e.g., Stamps et al. 1987, Fagan et al. 1999, Strayer et al. 2003), whereas an ecotone refers to a broader transition zone (Fagan et al. 1999) comprising ecotonal habitat and ecotone peripheries. An ecotone may form, decline, disappear, shift locations, or change in terms of structure (from an ecotone to be just an ecotone periphery) or

27 function due to the influence of changing physicochemical conditions, species use of habitat, and interactions among species (Holland 1988, Delcourt and Delcourt 1992, Wiens 1992, Kolasa and Zalewski 1995). As a zone of transition in biotic and physicochemical conditions, an ecotone provides different permeability on flux of organisms. In this study, ecotone permeability is considered in terms of ecotone function on organism distributions. Other than attracting particular organisms, the mechanism that produces the edge effect may simultaneously inhibit dispersion of other organisms into an ecotone. The same ecotone may be neutral and has no effect on distribution of some other species into the ecotone. Dynamic properties (formation and function) of ecotones can be obviously observed in an estuary, a complex ecosystem where the interaction between marine and freshwater systems occur, and tide and water flow play strong influence. As one of the world’s important estuaries, the tidal Hudson River drainage serves more than 200 fish species (Stanne et al. 1996, Daniels et al. 2005) from freshwater, estuarine, and marine assemblages. Other than providing variety of habitats for aquatic creatures, the Hudson River estuary has been heavily altered by human activities, such as urbanization, shoreline development, dam construction, power plant operation, and water diversion for commercial, industrial, and agricultural purposes (Limburg et al. 1986, Barnthouse et al. 1988, Schmidt and Cooper 1996, Wall and Phillips 1998, Baker et al. 2001, Phillips et al. 2002, Daniels et al. 2005). The tidal Hudson River portions where the intermediate conditions between fresh and saline waters meet are apparently vital ecotones for living of various fish species, especially in their first year of life (Weinstein et al. 1980, Stanne et al. 1996). At the lower-middle portion of the tidal Hudson River, the locations and spatial structures of the Hudson ecotones tend to vary in space and time due to the changes in salinity regime, tide, and freshwater runoff. At the upper portions of the tidal Hudson River,

28 effects of dam constructions and water pollutions, e.g., low dissolve oxygen, high coliform densities (Boyle 1979), and large loads of polychlorinated biphenyls discharged from several general electric factories above the Federal Dam (Baker et al. 2001) may contribute to sharp formations of ecotone peripheries affecting fish distributions below and above the Federal Dam. Although ecotone formations and their functions on organism distributions can be used as a baseline for resources managers to maximize healthy habitats for species assemblages of interest, these ecotone properties have never been revealed in the Hudson system. The first objective of this study was to detect ecotone formations on the Hudson River estuary gradient. Spatial structures of the Hudson ecotones were determined under the hypothesis that complex mix of abrupt and gradual changes in habitat conditions might contribute to the formation of an ecotone (comprising ecotonal habitat and ecotone peripheries), whereas abrupt changes in habitat conditions might contribute to a narrow formation of an ecotone periphery alone. Over space and time, some ecotones may be fixed in location, new ecotones may form, or former ecotones may decline or change in their breadths. The study habitat variables were dissolved oxygen, salinity, water temperature, two major types of submerged aquatic vegetation (Trapa natans and Vallisneria americana), and bottom substrates of water. Changing patterns of these habitat variables in and around the ecotone pay strong role on distribution patterns of fish assemblages along the Hudson River estuary gradient.

The second objective of the study was to detect ecotone functions on the Hudson fish species assemblages over time. Ecotone functions on fish species assemblages are interpreted from observed abundance of fish by species in and around an ecotone. An ecotone may act as an aggregator for particular fish species. Concurrently, it may act as a hard barrier completely blocking distribution of some fish species into the ecotone, or as a soft barrier partially allowing some fish species to enter the ecotone.

29 In the latter case, most species populations entering the ecotone will return to the habitats from which they came, causing few of them to be observed in the ecotone. An ecotone formed somewhere on the Hudson River estuary gradient may act as a mediator for marine and freshwater fish species with broad tolerance on changing salinity. Over space and time, an ecotone may decline in its structure to be just an ecotone periphery and may completely or partly inhibit distribution of particular fish species between adjacent habitats. Concomitantly, an ecotone periphery may be neutral and has no effect on fish distributions between adjacent habitats it separates.

MATERIALS AND METHODS

Study gradient and data sources The Hudson River estuary starts from Battery at river kilometer (RKM) 1 in Manhattan, New York City, to the Federal Dam at RKM 246 in Troy, Albany, New York State (Figure 2.1). In this study, the gradient was divided into four zones. The lowest zone (RKMs 1–19) was normally polyhaline with salinity concentration of 18– 30 ppt (Ristich et al. 1977, National Oceanic and Atmospheric Administration 1989). At this zone, the river channel was approximately 1 km wide and 9–15 m deep (Coch and Bokuniewicz 1986). The above three zones (the lower, the middle, and the upper zones) were classified following Schmidt et al. (1988).

Figure 2.1. The Hudson River estuary gradient showing 13 regions designated for the five power plant utilities’ Hudson River estuary monitoring program.

31 At the lower zone (RKMs 19–61), the river channel was 1.0–1.5 km wide and 13– 17 m deep (Coch and Bokuniewicz 1986). At the middle zone (RKMs 61–122), the river channel was 0.5–1.0 km wide and 18–48 m deep (Coch and Bokuniewicz 1986). At the upper zone (RKMs 122–246), the river channel was 0.6–1.2 km wide and 15– 31 km deep (Coch and Bokuniewicz 1986). Salinity in water varies temporally between mesohaline and oligohaline in the lower and the middle zones depending on tide and freshwater flow, while water was usually fresh in the upper zone (Strayer et al. 2004). Fish data used to determine ecotone properties on the Hudson River estuary gradient were from the beach seine survey (BSS) during 1974–2001 of the Utilities’ Hudson River estuary monitoring program. The Utilities collectively represents the five power plant utilities including Central Hudson Gas and Electric Corporation, Consolidated Edison Company of New York Inc, New York Power Authority, Niagara Mohawk Power Corporation, and Southern Energy New York. Based on the monitoring program, the Hudson River estuary was divided into 13 regions (RE) starting from RE0 (RKMs 1–19) to RE12 (RKMs 201–246; Figure 2.1). The monitoring program used stratified random samplings to conduct the BSS along the tidal river gradient. According to the stratification method, fish assemblages were randomly sampled along the shore zone (≤3-m water depth) using a 30.5-m total length beach seine with a 0.5-cm bag mesh size. At each sampling point, one end of the net (of the beach seine) was held on shore and the other end was towed perpendicularly away from shore by boat sweeping approximately 450 m2 of sampling area. Main purpose of the survey was to obtain abundance and distribution data of young of the year of target fish species including Atlantic tomcod (Microgadus tomcod), American shad (Alosa sapidissima), striped bass (Morone saxatilis), and

32 white perch (Morone americana) along inshore zone of the Hudson River estuary (Texas Instruments 1981). At least three observations were taken at each region in each sampling period of the BSS. Yearly numbers of the observations were 2,331–3,648 during 1974–1980, 500–800 during 1981–1984, and 1,000 during 1985–2001. The BSS was done weekly from: April to the 2nd week of December during 1974–1975 and 1978–1979, April through November during 1976–1977, April to the 3rd week of August and then biweekly through November in 1980. During 1981–1995, the BSS was done biweekly each year. The biweekly survey schedules were from: August to the 2nd week of October during 1981–1983, last week of July to the 1st week of October in 1984, the 2nd week of July to the 3rd week of November during 1985–1986, the 4th week of June to the 2nd week of November in 1987, June through October during 1988–1992 and 1994–1995, and last week of June to the 1st week of November 1993. During 1996– 2001, the BSS was done biweekly each year from July to the 3rd week of October. The BSS method in detail can be found from the Utilities’ early or recent monitoring programs (e.g., Texas Instruments 1977, ASA Analysis & Communication 2004). Habitat data including dissolved oxygen (DO), salinity, water temperature, submerged aquatic vegetation (SAV), and substrate composition were used to describe habitat conditions along the Hudson River estuary gradient (RE1–RE12) in this study. The first three habitat variables were from the Utilities’ BSS, and they were in-situ measured at approximately 0.3 m below the water surface and 15 m from the shoreline using a YSI Model 57 Dissolved Oxygen Meter (Texas Instruments 1976). Mean value for each of the three habitat variables at each sampling location and sampling week, weighted by stratum volume, was calculated using equation (1) as

33 nrw Wrw = 1/ nrw ∑ Wirw , (1) i =1

where Wrw is the mean of a habitat variable at river mile r during bi-week w of the

BSS, Wirw is a habitat measurement for location i at river mile r during bi-week w, and nrw is a number of habitat measurement taken at river mile r during bi-week w. DO was measured to the closet 0.1 mg/L, and water temperature was measured to the closest 0.1°C. Conductivity of water was measured in millisieman/cm at 25 °C (C25) and it was used to compute salinity concentration (ppt) using equation (2) as (Texas Instruments 1976)

Salinity = -100 ln (1-C25/178.5) (2)

SAV data were from the study by the collaboration among the Institute of Ecosystem Studies, the Hudson River National Estuarine Research, New York State Department of Environmental Conservation (NYSDEC), and the Cornell Institute for Resource Information Systems. The SAV study was conducted from RE2 to RE12 in August 1995 and July–August 1997. Photographic, digital, and grab sampling data were used to classify and map bed areas of Trapa natans (water chestnut—rooted with floating leaves) and Vallisneria americana (water celery). Substrate data from the Utilities’BSS (1974–2001) and NYSDEC (1998–2003) were combined to describe bottom condition along the study gradient. The Utilities’ substrate data were observed in shallow water areas (≤3 m deep) from RE1 to RE12 and its composition was classified into four types: mud, sand, gravel with <76 mm grain size, and gravel with ≥76 mm grain size. The NYSDEC’s substrate data were collected across the Hudson River estuary system, but they were not completely done

34 in shallow water <4 m deep (Bell et al. 2006). Both core and grab samples were categorized with some guidance from the acoustic backscatter data (Nitsche et al. 2004) and the sediment profiler imagery data (Rhoads and Cande 1971, Rhoads and Germano 1982). Based on grain size and percentage composition, the NYSDEC’s substrates were classified into: mud (<0.062 mm grain size), sand (0.062–2 mm), gravel (2–36 mm), and six mixed groups including sandy mud (mud with >10% sand), gravelly mud (mud with >10% gravel), muddy sand (sand with >10% mud), gravelly sand (sand with >10% gravel), muddy gravel (gravel with >10% mud), and sandy gravel (gravel with >10% sand).

Ecotone and habitat analyses Because the Utilities’ BSS was not conducted in RE0 prior to 1996, the ecotone and habitat analyses were done from RE1 to RE12 on the tidal Hudson River gradient based on the BSS fish and habitat data available from 1974 to 2001. Ecotone formation was detected monthly in all years the fish data were available for all 12 regions (RE1–RE12). Turnover rate or beta diversity (Whittaker 1975, Gauch 1982, Delcourt and Delcourt 1992, Winemiller and Leslie 1992, Wagner 1999, Hoeinghaus et al. 2004) in fish species composition was used to detect ecotone formation on the gradient. Detrended correspondence analysis (DCA) in CANOCO 4.0 software (ter Braak and Smilauer 1998) was used to compute the beta diversity (hereafter called beta A) based on both occurrence and percentage abundance by species of fish in each region. In the computation, the sample scores on the DCA ordination are rescaled into the average standard deviation (SD) units of species turnover (Hill 1979). A gradient length of 1.0 SD represents approximately one-fourth of a complete turnover rate in species composition on the gradient (Gauch 1982). Species appear, rise to their modes, and disappear over a distance of about 4 SD. That is, if the beta A on the

35 DCA’s axis 1 (that accounts for the most variation) is ≥4.0 SD, it indicates that species compositions at both extreme ends of a gradient represent completely different communities (Gauch 1982). This indicates complete formation of at least one ecotone on the gradient. The peak absolute difference (≥1 SD for this study) in the beta A (δ beta A) indicates an ecotone periphery (a margin where a sharp transition between adjacent habitats occurs). Multiple functions of an ecotone (an ecotonal habitat plus its ecotone peripheries) were interpreted from the relative observed abundance by species of fish among an ecotonal habitat, an upper adjacent habitat, and a lower adjacent habitat. This study proposed that an ecotone acts as an aggregator for a fish species with >60% abundance observed in the ecotonal habitat and as a mediator for fish species with 30–60% abundance observed in the ecotonal habitat. Concurrently, an ecotone acts as a soft barrier for a fish species with 5–30% abundance observed in the ecotonal habitat. Lastly, an ecotone acts as a hard barrier for a fish species with <5% abundance observed in the transitional habitat. Forming as an ecotone periphery (ep) between adjacent habitats (without ecotone width), the ep may be neutral allowing fish species with >30% abundance to distribute between the adjacent habitats. Concurrently, an ep may partly or completely inhibit distribution of a fish species from one habitat to the other adjacent habitat causing low (5–30%) or lower-absent (0–5%) abundance, respectively, of fish in the latter habitat. Each fish species with <5 individuals observed along the gradient in each analyzed time period (i.e., a certain month in a certain year) was not included in the comparison because of low catch. General linear model with multivariate analysis of variance (GLM MANOVA) procedure in SPSS version 14.0 software was used to compare mean values of DO, salinity, water temperature, and their associated delta values (i.e., absolute difference between minimum and maximum values) between two periods of ecotone analysis

36 (i.e., non-detected period, ECO = 0 and detected period, ECO = 1) and due to the interactions of ECO × RE, ECO × Month, ECO × Year, ECO × RE × Month, ECO × RE × Year, and ECO × Month × Year. Tukey’s honestly significant differences (HSD) test was used to monthly compare mean values of each habitat variable between paired regions when the significant difference of the variable was detected.

Bed areas (km2) of two SAV species (Trapa natans and Vallisneria americana) from the SAV shapefile were estimated by region along the Hudson River estuary gradient using ArcGIS 9.1. Area (km2) of each substrate type including mud (<0.062 mm grain size), sand (0.062–2 mm), gravel (2–76 mm), cobble (>76 mm), and the six mixed groups from the NYSDEC and Utilities’ datasets were combined and re-calculated. Because the NYSDEC’s substrate data were less completed along the Hudson inshore zone, only offshore area for each substrate type was retrieved from this dataset and summed up with the inshore area of the associated substrate type estimated from the Utilities’ BSS for each associated region. That is, after the area for each substrate type from the NYSDEC’s shapfile was estimated by region using ArcGIS 9.1, the offshore area for each substrate type in each associated region was re-computed using equation (3) as follows:

Area i(sub j ,DEC) Area = × Area − Area , (3) i(sub j ,offshore) ()i(total,DEC) i(shore,Utilities Areai(total,DEC)

where Areai(sub j, offshore) is the estimated area of substrate type j at region i (i = 1, 2,

3,…, 12) in the offshore zone, Areai(sub j, DEC) is the estimated area of substrate type j at region i based on the NYSDEC’s data, Areai(total, DEC) is the estimated area for all substrate types at region i based on the NYSDEC’s data, and Areai(shore, Utilities) is the estimated area in the inshore zone at region i based on the Utilities’ data.

RESULTS

A total of 127 fish species (excluded 14 unidentified species) from 47 families was recorded along the Hudson River estuary gradient (RE1–RE12) from the BSS during 1974–2001. Of the 47 families, four families: Cyprinidae, Centrarchidae, Clupeidae, and Sciaenidae had the highest species richness. Blueback herring (Alosa aestivalis) was the most abundant (34%) species and distributed throughout the Hudson River estuary. The latter four fish species with >8% observed abundance on the gradient were bay anchovy (Anchoa mitchilli, 9%), white perch (Morone americana, 9%), unidentified clupeid (9%), and American shad (Alosa sapidissima, 8%). Distribution patterns of fish along the Hudson River estuary gradient are apparently shaped by fish migratory behaviors (anadromous and catadromous) and salinity tolerance across marine-estuarine and estuarine-freshwater transitions. Based on these criteria, the Hudson fish species were classified into five assemblages, comprising anadromous, catadromous, estuarine, freshwater, and marine (Smith 1985). Of the total 127 identifiable fish species recorded along the gradient, nine species were anadromous fish, one species (American eel—Auguilla rostrata) was catadromus, 13 species were estuarine fish, 59 species were freshwater fish, and 45 species were marine fish (Appendix 2.1).

Ecotone formation and function The ecotones were detected in 116 time periods (76%) of the total 152 time periods the fish data were available for the analysis during 1974–2001 (Figure 2.2).

Figure 2.2. Frequency of ecotones detected monthly based on the data collection along the Hudson River estuary gradient during 1974–2001.

39 The ecotone was detected with the highest frequency (18 times) at RE2, while the ecotone peripheries alone were highly detected between RE1 and RE2 (18 times) and between RE11 and RE12 (30 times) on the Hudson River estuary gradient (Figure 2.3). Ecotones and ecotone peripheries were less detected at the remaining portions on the gradient (Figure 2.3).

Figure 2.3. Ecotones (thick bars) and ecotone peripheries (thick lines) detected during 1974–2001on the Hudson River estuary gradient covering 228 river kilometers (RKM) from region (RE) 1 to RE12 are grouped into each frequency class. Numbers of times of the formation of an ecotone and an ecotone periphery in each location are shown in brackets in each frequency class on the right side of the graph.

With the highest frequency (18 time periods) of detection at RE2, this ecotone acted as a hard barrier for most (>35%) of the total fish species observed along the

40 gradient in the associated time periods (Table 2.1). Similarly, the ecotones detected at other locations (except at RE2–RE4) acted as a hard barrier for most of the Hudson fish species (Table 2.1, see ecotone functions from all detected frequencies in Appendix 2.2). Detecting 10 times at RE2–RE4, this ecotone acted as an aggregator for most of the fish species observed from the six time periods (July 1986, August 1979, September 1982 and 1988, October 1975, and November 1986), but it acted as a mediator, soft barrier, or hard barrier for most of the fish species observed in the rest time periods (Table 2.1). Consistent with most of the ecotones functioning as hard barriers, the ecotone peripheries detected alone between particular adjacent regions on the gradient completely inhibited the distribution between the adjacent regions of most (>35%) fish species in the analyzed time periods (Table 2.2, see ecotone periphery functions from detected frequencies in Appendix 2.3).

Table 2.1. Functions of the ecotones detected ≥5 times at certain regions (RE) and time periods on Hudson fish species assemblages.

EcotoneFunction (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE2 May-75 3.1 21.9 28.1 46.9 32 May-76 6.7 26.7 26.7 39.9 30 May-78 15.6 12.5 21.9 50.0 32 Jun-74 8.8 11.8 20.6 58.8 34 Jun-75 10.9 16.2 24.3 48.6 37 Jun-76 12.2 9.8 34.1 43.9 41 Jun-77 16.7 13.9 25.0 44.4 37 Jul-78 7.0 18.6 14.0 60.4 43 Jul-79 26.2 28.6 11.9 33.3 42 Jul-84 31.0 21.4 9.5 38.1 42 Jul-90 32.3 19.4 12.9 35.4 31 Aug-78 10.2 18.4 20.4 51.0 49 Aug-84 36.6 14.6 22.0 26.8 41 Aug-96 21.4 10.7 17.9 50.0 28 Sep-84 31.3 12.5 15.6 40.6 32 Nov-76 18.2 13.6 22.7 45.5 22 Nov-79 37.5 17.5 12.5 32.5 40 Dec-78 0.0 11.1 27.8 61.1 18

41 Table 2.1 (Continued).

EcotoneFunction (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE2-RE3 Jun-91 19.0 23.9 21.4 35.7 42 Jul-00 20.0 17.1 8.6 54.3 35 Sep-79 42.9 34.3 11.4 11.4 35 Sep-81 27.9 9.3 23.3 39.5 43 Sep-87 46.0 16.2 18.9 18.9 37 Sep-99 12.5 6.3 18.8 62.4 32 Oct-83 36.8 5.3 21.1 36.8 19 Nov-78 16.7 33.3 13.3 36.7 30 RE2-RE4 Jul-86 42.1 15.8 18.4 23.7 38 Aug-74 22.0 30.0 22.0 26.0 50 Aug-79 36.4 25.0 18.1 20.5 44 Sep-82 38.5 25.6 18.0 17.9 39 Sep-88 51.5 9.1 15.2 24.2 33 Sep-96 30.0 20.0 30.0 20.0 20 Oct-75 38.9 22.2 22.2 16.7 36 Oct-98 22.6 19.4 6.5 51.5 31 Nov-80 23.5 29.5 23.5 23.5 17 Nov-86 36.0 28.0 12.0 24.0 25 RE3-RE4 Apr-74 50.0 11.1 16.7 22.2 18 May-78 37.5 37.5 15.6 9.4 32 Jun-75 29.8 40.5 10.8 18.9 37 Jul-75 25.6 30.2 23.3 20.9 43 Aug-81 18.0 35.9 28.2 17.9 39 Aug-97 3.6 25.0 42.8 28.6 28 Sep-01 3.5 13.8 24.1 58.6 29 Oct-84 20.1 23.3 33.3 23.3 30 Dec-78 38.9 27.8 11.1 22.2 18 RE3-RE5 Aug-98 15.2 18.2 33.3 33.3 33 Sep-98 7.0 10.3 37.9 44.8 29 Oct-85 20.0 20.0 50.0 10.0 30 Oct-92 21.4 25.0 14.3 39.3 28 Nov-87 14.3 7.1 35.7 42.9 14 RE4-RE5 Jul-90 3.2 9.7 12.9 74.2 31 Aug-99 8.2 10.8 32.4 48.6 37 Oct-83 15.8 26.3 31.6 26.3 19 Oct-93 13.3 20.1 33.3 33.3 30 Sep-81 4.7 9.3 44.2 41.8 43 RE5 Jun-91 0.0 4.8 26.2 69.0 42 Jun-92 0.0 8.3 25.0 66.7 24 Jul-93 0.0 13.9 22.2 63.9 36

42 Table 2.1 (Continued).

EcotoneFunction (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE5 (Continued) Sep-93 7.0 3.4 31.0 58.6 29 Sep-96 15.0 10.0 30.0 45.0 20 Dec-78 50.0 16.7 5.6 27.7 18 RE6 Jun-91 0.0 7.1 16.7 76.2 42 Jun-92 0.0 16.7 20.8 62.5 24 Jul-90 3.2 3.2 13.0 80.6 31 Aug-88 3.2 15.6 28.1 53.1 32 Dec-78 5.6 16.7 11.0 66.7 18 RE7-RE11 Aug-87 21.2 27.3 18.2 33.3 33 Aug-97 29.4 14.7 20.6 35.3 34 Sep-87 19.4 22.3 19.4 38.9 36 Sep-90 16.6 26.7 26.7 30.0 30 Oct-88 27.6 6.9 31.0 34.5 29

1A fish species with less than 5 individuals collected along the Hudson River estuary gradient in each of the 116 time periods was not included in the analysis.

Table 2.2. Functions of the ecotone peripheries detected ≥5 times at certain regions (RE) and time periods on Hudson fish species assemblages.

Ecotone periphery Function (%) for the total selected species No. of selected1 location between Neutral Partly inhibit Completely inhibit species/gradient

RE1-RE2 Apr-78 8.0 20.0 72.0 25 Apr-79 12.5 29.2 58.3 24 May-74 26.1 4.3 69.6 23 May-77 12.2 24.2 63.6 33 Jun-78 12.5 20.0 67.5 40 Jul-76 13.3 15.6 71.1 45 Aug-80 12.2 9.8 78.0 41 Aug-82 8.9 23.5 67.6 34 Aug-89 10.0 10.0 80.0 30 Aug-93 14.7 2.9 82.4 34 Sep-75 8.3 22.9 68.8 48 Sep-83 6.3 18.7 75.0 32 Sep-85 11.8 23.5 64.7 34 Sep-86 3.0 9.1 87.9 33 Oct-74 11.8 17.6 70.6 34

43 Table 2.2 (Continued).

Ecotone periphery Function (%) for the total selected species No. of selected1 location between Neutral Partly inhibit Completely inhibit species/gradient

RE1-RE2 (Continued) Oct-77 15.2 27.3 57.5 33 Oct-81 16.7 13.9 69.4 36 Nov-77 14.7 23.5 61.8 34 RE3-RE4 Jun-88 17.1 17.1 65.8 35 Jul-96 18.0 25.6 56.4 39 Sep-76 18.0 41.0 41.0 36 Sep-80 11.1 33.3 55.6 36 Sep-94 8.6 17.1 74.3 35 Oct-78 25.7 34.3 40.0 35 Oct-80 15.4 26.9 57.7 26 Oct-86 25.8 29.0 45.2 31 Oct-95 13.3 30.0 56.7 30 RE10-RE11 Jun-74 35.3 35.3 29.4 34 Jun-78 22.5 35.0 42.5 40 Jul-75 30.2 30.2 39.6 43 Aug-79 29.5 29.5 41.0 44 Sep-75 22.9 14.6 62.5 48 Sep-80 30.6 19.4 50.0 36 Oct-74 14.7 38.2 47.1 34 Nov-80 11.8 5.9 82.3 17 Nov-86 24.0 24.0 52.0 25 RE11-RE12 May-74 13.1 21.7 65.2 23 May-77 9.1 30.3 60.6 33 May-78 21.9 18.7 59.4 32 May-80 16.7 13.9 69.4 36 Jun-79 13.9 16.7 69.4 36 Jun-80 20.0 32.5 47.5 40 Jun-95 9.6 33.3 57.1 21 Jul-76 19.5 31.7 48.8 41 Jul-77 14.0 25.5 60.5 43 Jul-78 28.6 26.2 45.2 42 Jul-97 8.6 20.0 71.4 35 Jul-00 20.6 32.4 47.0 34 Aug-75 20.0 28.9 51.1 45 Aug-76 29.2 20.8 50.0 48 Aug-78 28.6 22.4 49.0 49 Aug-80 26.8 19.5 53.7 41 Sep-76 25.6 18.0 56.4 39 Sep-78 28.9 15.8 55.3 38 Sep-79 22.9 22.9 54.2 35 Sep-82 20.5 18.0 61.5 39 Sep-94 20.0 20.0 60.0 35 Sep-99 18.8 43.8 37.4 32

44 Table 2.2 (Continued).

Ecotone periphery Function (%) for the total selected species No. of selected1 location between Neutral Partly inhibit Completely inhibit species/gradient

RE11-RE12 (Continued) Sep-01 16.7 13.3 70.0 30 Oct-76 10.3 25.6 64.1 39 Oct-78 20.0 11.4 68.6 35 Oct-79 22.3 19.4 58.3 36 Oct-80 15.4 26.9 57.7 26 Oct-93 10.3 13.8 75.9 29 Nov-78 16.7 10.0 73.3 30 Nov-79 13.4 23.3 63.3 30

1A fish species with less than 5 individuals collected along the Hudson River estuary gradient in each of the 116 time periods was not included in the analysis

Changing habitat related to ecotone formation Only mean values of DO were significantly different (GLM MANOVA: p≤0.050) between the non-detected (ECO = 0) and detected (ECO = 1) ecotone periods, whereas mean values of DO, delta salinity, and delta water temperature were significantly differences due to the interactions of ECO × RE (Table 2.3). Mean values of all habitat variables and their associated deltas showed significant differences due to the interactions of ECO × Month, ECO × Year, ECO × RE × Month, ECO × RE × Year, and ECO × Month × Year (Table 2.3). Mean values of DO, salinity, and water temperature were then further examined over space (RE) and time (Month) based on the results from Turkey’s HSD post hoc. The DO along the Hudson River estuary gradient ranged from 11 to 13 mg/L in April, but decreased to 9–11 mg/L in May, and 7–10 mg/L during June–September. The DO increased again to 8–10 mg/L in October and 10–14 mg/L in December (Figure 2.4).

45 Table 2.3. Results from multivariate analysis of variance for each habitat variable and its delta (absolute difference between minimum and maximum values) between non- detected (ECO = 0) and detected (ECO = 1) ecotone periods and due to the interactions of ECO × RE (river region), ECO × Month, ECO × Year, ECO × RE × Month, ECO × RE × Year, and ECO × Month × Year.

Dependent Mean Sourcedf F p-value Variable Square

Dissolved oxygen 1 2.45 4.71 0.030 Delta dissolved oxygen 1 0.02 0.01 0.921 Salinity 1 0.03 0.06 0.804 ECO Delta salinity 1 1.27 1.03 0.311 Water temperature 1 0.07 0.12 0.730 Delta water temperature 1 2.01 0.60 0.440

Dissolved oxygen 11 1.08 2.07 0.020 Delta dissolved oxygen 11 2.70 1.13 0.336 Salinity 11 0.75 1.72 0.064 ECO × RE Delta salinity 11 3.60 2.90 0.001 Water temperature 11 0.97 1.78 0.053 Delta water temperature 11 9.41 2.78 0.001

Dissolved oxygen 6 3.06 5.89 0.000 Delta dissolved oxygen 6 8.59 3.59 0.002 Salinity 6 3.59 8.21 0.000 ECO × Month Delta salinity 6 8.95 7.23 0.000 Water temperature 6 11.90 21.83 0.000 Delta water temperature 6 23.04 6.81 0.000

Dissolved oxygen 18 3.53 6.79 0.000 Delta dissolved oxygen 18 12.97 5.42 0.000 Salinity 18 3.63 8.30 0.000 ECO × Year Delta salinity 18 5.81 4.69 0.000 Water temperature 18 9.73 17.86 0.000 Delta water temperature 18 30.19 8.93 0.000

Dissolved oxygen 154 1.51 2.90 0.000 Delta dissolved oxygen 154 2.94 1.23 0.043 ECO × RE × Salinity 154 2.11 4.84 0.000 Month Delta salinity 154 2.88 2.33 0.000 Water temperature 154 2.59 4.75 0.000 Delta water temperature 154 7.37 2.18 0.000

Dissolved oxygen 493 1.27 2.44 0.000 Delta dissolved oxygen 493 3.55 1.48 0.000 ECO × RE × Salinity 493 1.77 4.04 0.000 Year Delta salinity 493 2.86 2.31 0.000 Water temperature 493 0.63 1.16 0.026 Delta water temperature 493 6.99 2.07 0.000

46 Mean values of DO between the detected and non-detected ecotone periods were significantly different at particular region(s) in all eight months included in the analysis, however, inconsistent patterns were shown over time (Figure 2.4).

Figure 2.4. Monthly mean values of dissolved oxygen observed along the Hudson River estuary gradient in the detected ecotone (line with solid dot markers) and non- detected ecotone (line with open dot markers) periods.

47 Focusing on mean values of DO in the detected ecotone period, the abrupt changes of DO from RE1 to RE4 and between RE11 and RE12 in most of the analyzed months (Figure 2.4) were parallel with high frequencies of detecting ecotones and ecotone peripheries within this gradient range. Unlike DO, the salinity showed consistent patterns of change over time along the Hudson River estuary gradient. Mean values of salinity along the gradient were lower in April and May, but higher in June, October, and December, and reached the highest during July–September (Figure 2.5). In general, the salinity was the highest at RE 1 that was influenced by the marine conditions of New York Bay and it abruptly decreased over distance until became tidal fresh at the upper zone of the gradient (Figure 2.5). The mean values from the non-detected ecotone period were significantly higher than those from the detected ecotone period at RE1–RE2 in May, but they were significantly lower at RE6 in April, RE1–RE2 in October, and RE1–RE4 in December (Figure 2.5). The salinity along the gradient was similar between the two periods in the remaining months (Figure 2.5).

Figure 2.5. Monthly mean values of salinity observed along the Hudson River estuary gradient in the detected ecotone (line with solid dot markers) and non- detected ecotone (line with open dot markers) periods.

Water temperature along the Hudson River estuary gradient was 7–10°C in April and it increased to 14–17°C in May. The water temperature was higher during June– August (22–27°C), but lower in September (20–24°C) and October (14–18°C), and the lowest in December (2–7°C; Figure 2.6).

Figure 2.6. Monthly mean values of water temperature observed along the Hudson River estuary gradient in the detected ecotone (line with solid dot markers) and non- detected ecotone (line with open dot markers) periods.

Mean values of the water temperature in the detected ecotone period were significantly lower than those in the non-detected ecotone period at most regions in May and June, RE1–RE6 in July, and RE7–RE12 in December (Figure 2.6). In April,

50 August, and September, the mean values of water temperatures from the non-detected ecotone period was significantly higher at some particular regions, while there was no significant difference of the mean values along the gradient between the two periods in October (Figure 2.6). Two SAV species (Trapa natans and Vallisneria americana) and their bed areas varied along the tidal river gradient (Figure 2.7) above RE1, where the SAV data were unavailable. Trapa natans beds were not observed from RE2 to RE4, where the ecotones were highly detected within this range. The estimated bed areas of Trapa natans along the remaining portion on the gradient varied from 0.1 km2 at RE5 and RE12 to the highest (1.8 km2) at RE11 (Figure 2.7). Bed areas of Vallisneria americana were observed from RE2 to RE12 and ranged from the lowest (0.2 km2) at RE2 and RE8 to the highest (5.3 km2) at RE10 (Figure 2.7). Abrupt changes in bed areas of the two SAV species between RE11 and RE12 (Figure 2.7) were congruent with the highest frequency of the ecotone periphery detected between these two regions. Bottom of water along the study gradient was mainly a combination of mud and sandy mud from RE1 to RE9, and a combination of sand and muddy sand from RE10 to RE12 (Figure 2.7). Four substrate types including gravelly sand, gravel, sandy gravel, and cobble were observed at low percentage compositions along the gradient (Figure 2.7). Gravelly mud and muddy gavel were observed <2% and at RE1 only.

Figure 2.7. Values of dissolved oxygen (DO), salinity, and water temperature were averaged by region over the 116 time periods that the ecotones were detected. Bed areas of two submerged aquatic vegetation (SAV) species (Trapa natans and Vallisneria american) vary along the gradient (data for region 1 were unavailable). Proportions of 10 substrate types were observed in the tidal river estuary. Two of them (gravelly mud and muddy gravel) not shown in the graph were observed <2% and at RE1 only.

52 DISCUSSION

Formations of the Hudson ecotones were highly dynamic showing both shift in location and change in structure over the detection periods (1974–2001). The ecotone formations were related with the changes in habitat conditions over space and time. Changeovers in fish species composition influencing by the transitions in physicochemical conditions (Winemiller and Leslie 1992, Marshall and Elliott 1998, Whitfield 1999, Martino and Able 2003) indicated the ecotone formations at the particular portions on the Hudson River estuary gradient. Consistent with the study by Attrill and Rundle (2002), the tidal Hudson River represents two important transitional conditions between brackish and estuarine waters (from the lowest zone to the middle zone of the gradient) and between estuarine and freshwater (from the middle zone to the upper zone).Although the ecotone analysis was not determined at the lowest zone (RE0) because of the unavailable fish data at this region prior to 1996, it is speculated that the ecotone periphery between RE0 and RE1 is most likely to exist due to the sharp transition between polyhaline and mesohaline waters (Weinstein et al. 1980) between the two regions. At the lower zone, the sharp changes in DO and water salinity between RE1 and RE2 apparently contributed to the formation of the ecotone periphery (the narrow transition) that was highly detected between these two regions. Salinity was not a major factor influencing formations of the ecotone peripheries at the upper zone of the gradient where water condition becomes the tidal fresh (salinity <0.5ppt; Winemiller and Leslie 1992). At the upper gradient portion, the ecotone periphery frequently detected between RE11 and RE12 was most likely influenced by the abrupt declines in DO and water temperature from RE11 to RE12 in most months, the decreased abundance of Trapa natans and Vallisneria americana from RE11 to RE12, and the dam constructions

53 above RE12 and in close proximity. Along the gradient, SAV occupied approximately 8% of the river bottom area, and 18% of the shallow water areas (≤3 m deep) was available for SAV to colonize (Nieder et al. 2004). Below RE12, higher abundance of the two SAV species, especially from RE9 to RE11, created both food sources and shelters for inshore fish and other benthic animals in the shallow habitats (Abood and Metzger 1996, Noordhuis et al. 2002, Strayer et al. 2004). In contrast, the dams existing around and above RE12 alter flow regimes by reducing peak discharged downstream and meting out flow over time to maintain minimum depths for navigation (Lumia 1998). These consequences tend to affect distributions of fish assemblages to the downriver portion (Abood and Metzger 1996, Schmidt and Cooper 1996). While abrupt changes in both fish species composition and habitat conditions concomitantly reflected the formations of ecotone peripheries between the particular regions in the lower and upper zones, the ecotones were mainly detected within the lower-middle zone of the gradient. The shift in locations of these ecotones, but still within the gradient range from RE2 to RE5, tended to relate with the changes in salinity regime, tidal current, and freshwater runoff seasonally or yearly (Weinstein et al. 1980, Stanne et al. 1996, Attrill and Rundle 2002, Jaureguizar et al. 2003). Most of the ecotones detected in the lower-middle zone had the breadth covering one or two important nursery areas (i.e., Tappan Zee—RE2 and Haverstraw Bay—RE3) with low salinity (1.0–5.1 ppt) serving diverse fish species (Stanne et al. 1996, Daniels et al. 2005) from marine (e.g., bay anchovy—Anchoa mitchilli and Atlantic menhaden— Brevoortia tyrannus), estuarine (e.g., hogchoker—Trinectes maculates and banded killifish—Fundulus diaphanus diaphanus), and freshwater (e.g., green sunfish— Lepomis cyanellus and brown bullhead—Ameiurus nebulosus) assemblages.

54 However, in most time periods of the study, only one ecotone with the breadth from RE2 to RE4 acted as an aggregator for most fish species, while the other ecotones acted as soft or hard barriers for most of the Hudson fish (Table 2.1). Furthermore, functional patterns of these ecotones on fish species assemblages were not clearly different among seasons. From the comparison, while certain fish species from marine, estuarine, and freshwater assemblages were usually observed in most ecotones, total numbers of the fish species from the mixed assemblages in these ecotones were intermediate or lower than those from the marine-estuarine or freshwater assemblage observed in the adjacent habitat downstream or upstream, respectively. Additionally, the abundance (numbers of individuals) of most fish species was relatively low in the ecotones. These reflected complex distributions of the Hudson fish assemblages and their variety of life histories (Hurst et al. 2004, Daniels et al. 2005) responding to changing habitat conditions in and around the ecotones on the gradient (Kupschus and Tremain 2001, Able 2005). In an estuary, distributions of fish assemblages highly vary over short and long- term periods due to: 1) complex life histories of fish and their habitat preferences that may widely vary among life history stages (Able and Fahay 1998), 2) influences of the recruitment dynamics of individual species (Hurst et al. 2004) and their differential utilization of specific portions of the estuary (Weinstein et al. 1980), and 3) the behaviors of fish that are inadequately understood (Able 2005). Along the Hudson River estuary gradient, distributions of some fish species are restricted by their narrow salinity tolerances, whereas distributions of other fish species with broader salinity tolerances may be limited by other habitat variables that are correlated with salinity concentration (Winemiller and Leslie 1992), e.g., tide, freshwater flow, and water temperature (Vernberg and Vernberg 1976).

55 Influencing by river discharge and salinity concentration, young marine fish in the Hudson system generally move upriver on the incoming flow of dense saline water near the river bottom, whereas the young anadromous fish travel downriver on the outgoing freshwater nearer the surface (Stanne et al. 1996). These fish assemblages generally seek nursery areas with low salinity in the lower-middle zone of the tidal river to temporarily spend their first year of life here and eventually migrate to the Atlantic Ocean (Stanne et al. 1996, Able and Fahay 1998) as juveniles or adults (Weinstein et al. 1980). For freshwater fish, some species may periodically move farther downstream during high freshwater runoff with low salinity, while most freshwater species are residents in the tidal fresh regions. Conclusively, patterns of fish use of estuarine habitats in the lower-middle zone of Hudson is apparently a continuum (Attrill and Rundle 2002, Able 2005); some species have obligated life history stages (residents or diadromous) in the estuary, others are estuarine opportunists, and still others are simply strays that occasionally find their ways into the estuary (Able 2005). The variations in freshwater flow, DO, salinity, and water temperature from month to month also affected fish distributions (e.g., Gilmore 1995, Marshall and Elliott 1998, Whitfield 1999, Martino and Able 2003) in and around the Hudson ecotones. During low freshwater flow with the increase salinity along the gradient, distribution ranges of some marine and estuarine fish species show spatial change as they migrate farther upstream (Attrill and Rundle 2002) beyond the brackish-estuarine ecotones (at the lower-middle zone) leading to low species richness and abundance of fish in these ecotones. In addition, both species richness and abundance of fish are mostly likely to decrease at the freshwater-estuarine ecotones (at the middle-upper zone) of the gradient as freshwater species became intolerant of increased salinity, while only a few estuarine species can tolerate low salinity (Rundle at al. 1998).

56 Focusing at the lower-middle zone from RE2 to RE5 (where the ecotones were highly detected), DO and water temperature showed gradual changes within the range from RE2 to RE5 in a few months, but sharply fluctuated among these regions in most months. Wells and Young (1992) found that the seasonal cycle accounted for most (>99%) of the variation in water temperature in the Hudson system, whereas patterns of changes in DO reflected an inverse relationship with water temperature (Mancroni et al. 1992). Unlike DO and water temperature, salinity showed consistent pattern of abrupt decrease from RE2 to RE5 in all study months, except in April. The synergy of both gradual and sharp changes among these habitat variables plays important role in structuring fish species assemblages (Whitfield 1999, Marshall and Elliott 1998) in and around the Hudson ecotones. The bottom substrates are another factor influencing distributions of fish species assemblages in an estuary at some extent (Weinstein et al. 1980). However, this habitat variable may play less influence on the distributions of fish assemblages along the tidal Hudson River where the bottom of water is relatively homogeneous. More than half of the gradient length (from RE1 to RE9) comprises a major combination of mud and sandy mud, while sand and muddy sand were major combined substrates at the remaining gradient length (RE10–RE12). Since 1991, the zebra mussel invasion has been observed in the Hudson system (Strayer et al. 1996) and its effect might be an additional factor shaping ecotone functions on the Hudson fish species assemblages afterward. With their high filtration rate of 6 m3/m2.day (Strayer et al. 1996), the zebra mussel populations are suspected to bring about a selective decline in the density of phytoplankton and small zooplankton (rotifers, tintinnids, and copepod nauplii) and a loss of consumers that depended on phytoplankton and other edible particles in deep water of the Hudson (Pace et al. 1998, Strayer et al. 1999). In contrast, the filtration rates of the zebra mussel populations increased sunlight penetration through surface of shallow water of the

57 Hudson enhancing the growth of aquatic plants and the invertebrates associated with them (Strayer et al. 1999). These consequences interfered distribution patterns of the Hudson fish assemblages and caused 28% of the median decrease in abundance of open water fish species (e.g., Also spp.) due to the decline of food source in deep water, but caused 97% of the median increase in abundance of littoral fish (e.g., sunfishes) due to the increase of food source in shallow water (Strayer et al. 2004). As a result, both open-water and inshore fish assemblages move downstream and upstream, respectively, away from the middle portion of the Hudson River estuary gradient. In conclusion, the formations of Hudson ecotones showed both changes in location and structure spatially and temporally. In some time periods of detection, the ecotones changed in their breadth (from broader to narrower or vice versa) or considerably declined to be just ecotone peripheries. The ecotones were highly detected within the lower–middle zone (RE2–RE5) of the Hudson River estuary gradient, whereas the ecotone peripheries alone were highly detected at the lower zone (between RE1 and RE2) and the upper zone (between RE11 and RE12) of the gradient. Abrupt changes in one or more habitat variables and the peak turnover rates in fish species composition were coincidently observed at the ecotone peripheries. Functional patterns of the Hudson ecotones were not distinctly different across seasons. The ecotone with the breadth from RE2 to RE4 was most likely to be the optimal zone for distributions of diverse fish species assemblages, whereas the other ecotones acted as barriers for most of the Hudson fish species. Likewise, the ecotone peripheries on the gradient mainly inhibited distributions of most fish species between adjacent regions with high contrast of habitat changes. Relatively low abundances of most fish species observed in the Hudson ecotones were apparently related with complex migratory behaviors and life histories of fish responding to variations in the estuarine conditions

58 in space and time. The effect of zebra mussel invasion in the Hudson system might be an additional factor shaping the ecotone functions on distribution patterns of the Hudson fish assemblages.

To explore influences of changing habitat in and around aquatic ecotones on distributions of fish species assemblages in space and time, an explicitly abundance exchange model was developed to quantify distribution patterns of target fish species in association with their habitat preferences across lentic-lotic transitions on the two freshwater gradients located on the southeastern coast of Lake Ontario. The methodology of study, results, and discussion are presented in Chapter 3.

59 *CHAPTER THREE

An abundance exchange model of fish assemblage response to changing habitat along embayment-stream gradients of Lake Ontario, New York

ABSTRACT

A spatially explicit abundance exchange model (AEM) was developed to predict distribution patterns of five fish species in relation to their population characteristics and habitat preferences along two embayment-stream gradients associated with Lake Ontario, New York. The five fish species were yellow perch (Perca flavescens), bluegill (Lepomis macrochirus), logperch (Percina caprodes), bluntnose minnow (Pimephales notatus), and fantail darter (Etheostoma flabellare). Preference indexes of each target fish species for water depth, water temperature, current velocity, cover types (aquatic plants, algae, and woody debris), and bottom substrates (mud, sand, gravel-cobble, and rock-bedrock) were estimated from the field observations, and these were used to compute habitat preference (HP) of the associated fish species. Fish HP was a key variable in the AEM to quantify abundance exchange of an associated fish species among habitats on each study gradient. According to the results, the AEM efficiently determined local distribution ranges of the fish species on one study gradient. Results from the model validation showed that the AEM with its estimated parameters was able to quantify most of the fish species distributions on the second gradient. Overall, the AEM is rigorous for quantifying the distribution patterns of the target species along the changing habitat gradients.

*Singkran, N., An abundance exchange model of fish assemblage response to changing habitat along embayment–stream gradients of Lake Ontario, New York, Ecol. Model. (2006), doi:10.1016/j.ecolmodel.2006.10.012

60 With its flexible structure that is applicable for array functions and differential equations from both static and dynamic components, the AEM can be modified to determine patterns of organism distribution in complex systems with different environments and geography.

INTRODUCTION

The influence of changing landscapes and environments on organism distribution across ecotones (transition zones) of contrasting habitats (e.g., grassland-forest) has been increasingly studied in terrestrial ecosystems (e.g., Clements 1905, Leopold 1933, Odum 1971, Risser 1990, Johnson et al. 1992, Lidicker 1999, Fortin et al. 2000, Harding 2002, Cadenasso et al. 2003, Ries et al. 2004), whereas fewer studies of this similar research type have been conducted in aquatic ecosystems (Winemiller and Leslie 1992, Willis and Magnuson 2000, Martino and Able 2003, Miller and Sadro 2003). Additionally, although distribution of organisms is a key mechanism underlying the population dynamics (Nisbet and Gurney 1982, Tilman and Kareiva 1997), few empirical studies (Reyes et al. 1994, Gaff et al. 2000, Kupschus 2003) have quantitatively determined organism distribution in association with both changes in habitat and population characteristics (birth, death, and migration) over space and time. In this study, an abundance exchange model (AEM) was built to quantify the distribution of fish species assemblages in relation to both population characteristics and habitat preferences of fish across lentic-lotic transitions on two embayment-stream gradients associated with Lake Ontario, New York. For purposes of this study, the term “transition” referred to both abrupt and gradual changes in spatial habitat heterogeneity between lentic (embayment) and lotic (stream) systems on the study

61 gradients. As a result, this transition is often distinctly observed at the lentic-lotic periphery, and it may gradually or abruptly occur across the zone of transition between the two systems. The objectives of developing the spatially explicit AEM were to determine how changing habitat influences distribution patterns of fish species assemblages along the embayment-stream gradients and how population characteristics shape the size of fish population over time. To test if the AEM could efficiently capture distribution patterns of fish across different habitat conditions, five fish species, yellow perch (Perca flavescens), bluegill (Lepomis macrochirus), logperch (Percina caprodes), bluntnose minnow (Pimephales notatus), and fantail darter (Etheostoma flabellare) were selected for modeling. Yellow perch and bluegill commonly distributed in the embayment-downstream parts on both study gradients. Logperch was mainly observed at the downstream-middle part on one study gradient, but this species was not observed on the second gradient. Bluntnose minnow commonly distributed throughout both study gradients, while fantail darter mainly distributed at the upstream parts of both study gradients. In aquatic systems, high variability in hydrology (Wiens 2002) and climate-linked seasonal changes can alter other aquatic habitat variables (e.g., water temperature and growth of aquatic plants and algae). As a result, the migration rate of fish along a gradient is most likely to reflect fish habitat preference within a dynamic template. Additionally, it has been considered that each fish species has a specific range of distribution (e.g., Morrison et al. 1985, Guay 2000), the species stays within its range but will variably distribute among habitats therein. This conforms to nonlinear patterns of organism distribution along changing habitat gradients (Whittaker 1975, Lewis 1997). In this study, both dynamic habitat variables (water depth, water temperature, current velocity, aquatic plants, algae, and woody debris) and static

62 habitat variables (the bottom substrates, i.e., mud, sand, gravel-cobble, and rock- bedrock) were observed along the study gradients and used to estimate fish habitat preferences. To test the hypothesis that distributions of fish across lentic-lotic transitions on the study gradients are based on fish preferences for changing habitats within their local ranges, habitat preferences of the target fish species at two life stages (0+ years and 1+ years) were the key variable in the AEM to quantify distribution patterns of the associated fish species along each study gradient. STELLA® 7.0.3 Research (High Performance Systems, Inc. 2002) was used to run the AEM. Since it was developed by Barry Richmond in 1985, STELLA has been widely used by scientists across fields of study. This software is capable of simulating many components and interactions of complex systems, and is fast and efficient at calculating a diversity of array functions and differential equations (e.g., Reyes et al. 1994, Pan and Raynal 1995, Marín 1997, Gottlieb 1998, Ford 1999, Krivtsov et al. 2000, Dew 2001, Gertseva et al. 2003, Gertseva et al. 2004, Kilpatrick et al. 2004, McCausland et al. 2006). Besides model development and prediction, model validation, an important modeling process (Haefner 1996, Olden et al. 2002, Ottaviani et al. 2004), was the other objective of this study. Although numerous modeling approaches (e.g., statistics, system dynamics, and geographic information system (GIS)) have been used for modeling species distribution, few of the models were further validated after the predictions were completed (Olden et al. 2002). Model validation is crucial for evaluating the accuracy of the model’s predictions and future applicability (Olden et al. 2002, Ottaviani et al. 2004). In addition, proper ways of validating the model must be implemented. To verify that the model prediction is robust and unbiased and has minimal errors (i.e., over-fitting data, overrating model performance, or underestimating the error for future application), one must validate the model with an

63 independent data set that has not been previously used to estimate the model’s parameters (Guisan and Zimmermann 2000, Elith and Burgman 2002, Fielding 2002, Olden et al. 2002, Ottaviani et al. 2004). In this study, to evaluate if the AEM is robust and efficient for predicting the distribution of the target fish species in similar geographic and hydrologic systems, the model with its estimated parameters to quantify the distribution pattern of each target species on one study gradient was validated against the observed data of the same fish species on the second study gradient located in close proximity. Accuracy of the model prediction was determined using the chi-square (χ2) statistic to test the goodness of fit between predicted and observed percentage abundances of fish along each study gradient. The important modeling processes and the model performance presented in this study may contribute to applications for better predicting the distribution of fish species assemblages across different habitats. This knowledge will help resource managers evaluate and maximize healthy habitats among different aquatic ecosystems to support diverse fish assemblages at both local and regional scales.

STUDY GRADIENTS AND FIELD OBSERVATIONS

The empirical studies were conducted across lentic-lotic transitions on two embayment-stream gradients, Floodwood (FL) and Sterling (ST) gradients, located on the southeastern coast of Lake Ontario, New York (Figure 3.1). The FL gradient comprises Floodwood Pond and Sandy Creek. The total watershed area for the FL gradient was 671.82 km2, and the surface area of Floodwood Pond alone was 0.078 km2. Sandy Creek ranges in width from 6 m (upstream) to more than 50 m (entering Floodwood Pond) and has a mean annual discharge of 7.8 m3/s (37-year record—

64 USGS gage 04250750 above study area). The ST gradient includes Sterling Pond and Sterling Creek. The total watershed area associated with the ST gradient was 210.76 km2, and the surface area of Sterling Pond alone was 0.38 km2. Sterling Creek ranges in width from 3 m (upstream) to more than 20 m (entering Sterling Pond) with a mean annual discharge of 1.9 m3/s (37-year record—USGS gage 04232100 above study area).

Figure 3.1. The Floodwood (FL) and Sterling (ST) gradients with eight sampling stations (stars, FL; circles, ST) for monthly collecting of target fish and habitat variables during June–August 2002. The fish and habitat collections were repeated at 12 stations (circles) along the FL gradient during June–August 2004.

A Garmin Etrex Venture global positioning system was used to mark locations of sampling stations on both gradients. In the summer months (June–August) of 2002, eight transect stations on the FL gradient (Figure 3.1: stars) were selected (between 43°43'16" N, 76°12'10" W and 43°48'47" N, 76°4'31" W) using a stratified random sampling method to cover all habitat types. The length of the entire sampling gradient was about 18.6 km. The same stratified random method was used to select eight transect stations (between 43°20'34.5" N, 76°41'54.7" W and 43°16'51.6" N,

65 76°37'37.6" W) on the ST gradient (Figure 3.1). The length of the entire sampling gradient was about 16.3 km. Abundance of the five selected fish species (yellow perch, bluegill, logperch, bluntnose minnow, and fantail darter) was observed once a month along the FL and ST gradients during June–August 2002. Fish were randomly sampled for 15 minutes per station at all FL and ST stations using electrofishing gear (the power supply to generate electricity was a 3500-W generator and a transformer with DC voltage output controller of 250–350 V). Juveniles and adults of each target species caught at each station were enumerated and their total length was measured to the nearest millimeter. In the same summer months, the habitat variables were measured once a month in all stations where the fish species were collected as follows. Water depth (m) and water temperature (°C) were randomly measured five times in each sampling station. Current velocity (m/s) was randomly measured five times within each station using a Marsh-McBirney 2000-model (Bovee 1982). Bottom substrates (%) were visually observed once a month in all sampling fish stations on each study gradient; at the first observation, substrate grain sizes (diameter) were measured and classified into four types modified from Pusey et al. (1993): mud (<1 mm diameter), sand (1–16 mm diameter), gravel-cobble (16–128 mm diameter), and rock-bedrock (>128 mm diameter). The average percentage of each substrate type and average abundance of each cover type was calculated from their 10 random observations in each station. The abundance of aquatic plants, algae, and woody debris was coded on a scale of 0– 3: 0 = absent, 1 = low, 2 = common, and 3 = high. In the summer months (June–August) of 2004, the fish and habitat collections were repeated at 12 sampling stations (Figure 3.1: circles) along the FL gradient. The 12 sampling stations were located between 43°43'20.6" N, 76°12'13.3" W and 43°51'11.7" N, 75°56'44" W, the total sampling gradient length was 28.2 km. The

66 target fish species were sampled monthly at a total of 60 locations within the 12 stations along the FL gradient. With a fine scale of sampling, juveniles and adults of each target species were collected for 10 minutes in each of the 60 locations using the same fishing gear and sampling protocols as in the summer of 2002. Individuals of each target fish species collected from all locations within each station were accumulated, enumerated, and measured for total length to the nearest millimeter. In the same summer months of 2004, the same habitat variables were sampled monthly five times in each of the 60 locations within the 12 stations where the target fish species were collected using the same methods and instruments as in the summer of 2002. As a result, 900 samples for each habitat variable were collected along the FL gradient during the three summer months of 2004. Mean values of each habitat variable for the 180 samples along the FL gradient were then obtained by dividing the total 900 samples by the five times of collection in each location (i.e., 900/5 = 180) in summer 2004. The mean habitat variables and abundance data for the target fish species collected on the FL gradient in the 2004 summer were used to build the AEM and to estimate and calibrate the model’s parameters (i.e., a preference index of the target fish species for each of the habitat variables). The model with its estimated parameters was then validated against the fish and habitat data collected on the ST gradient over the three summer months of 2002 (see detail in next section).

MODELING FISH DISTRIBUTIONS ALONG EMBAYMENT-STREAM GRADIENTS

Applications of diffusion and random walk theories have been accepted as powerful methods for spatially modeling organism dispersion (Nisbet and Gurney 1982, Reyes et al. 1994, Turchin 1998, Fagan et al. 1999, Kot 2001, Okubo and Levin

67 2001, Sparrevohn et al. 2002). In homogeneous habitat conditions, population distribution of a given species in relation to certain habitat variables on a one- dimensional gradient can be described as a Gaussian curve (Whittaker 1975, Lewis 1997). This normal distribution is the main idea proposed in Fick’s law of particle/organism dispersion on a one-dimensional system expressed as (Okubo and Levin 2001)

n ⎛ x 2 ⎞ P ( x, t ) = 0 exp ⎜ − ⎟, (1) 2π σ ⎜ 2 ⎟ ⎝ 2σ ⎠

where P(x, t) is the total individuals of a given species at location x and time t, n0 is the initial number of individuals of a given species, and σ2 is the variance of the normal distribution of a given species. It has been statistically proven that σ equals 2Dt (e.g., Fischer et al. 1979), where D is a constant dispersion rate (distance2/time) of the population, and x is distance. Equation (1) after being rewritten becomes

n ⎛ x2 ⎞ P( x, t) = 0 exp ⎜ − ⎟ (2) 4 Dt ⎜ 4Dt ⎟ π ⎝ ⎠

However, using a constant D in equation (2) may be unrealistic for describing species distribution across heterogeneous habitat conditions (Johnson et al. 1992, Fortin et al. 2000, Okubo and Levin 2001). In this study, it was assumed that the migration rate of each target fish species varied depending on the fish preference for changing habitat conditions across the lentic-lotic transition on each study gradient. This assumption is based on the non-Fick’s law of dispersion (Johnson et al. 1992),

68 stating that although a given species acts as a random walker, it displays biased decision to distribute to its optimal habitat.

Abundance exchange model development A system dynamics approach, a rigorous modeling method that enables users to build formal computer simulations of complex dynamic systems (Ford 1999, Sterman 2000) was used for building an abundance exchange model (AEM) to determine distribution patterns of the target fish species along the two embayment-stream gradients (Figure 3.1). The AEM includes both within-habitat and across-habitat feedbacks to integrate population characteristics and habitat changes. The model parameters were estimated and calibrated based on the fish and habitat data collected along the FL gradient over the three summer months (June–August) of 2004. To construct the AEM for each target fish species with respect to the species preference for changing habitat, two major assumptions were made. First, distribution of the target fish species at two life stages (0+ years and 1+ years) within and across habitats on each study gradient was the result of functional migrations (Lucas and Baras 2001) of fish to seek preferred areas for refuge in winter (November–March) and growth (feeding and spawning) in growing season (April–October). Second, since small and shallow waters can be conceptualized as one-dimensional habitats (Skalski and Gilliam 2000), the distribution of each target fish species along the shallow gradients (<6 m deep) was considered on one dimension only (i.e., in a horizontal direction from downstream to upstream or vice versa). The conceptual diagram (Figure 3.2) shows how the distribution pattern of each target fish species within and across habitats was quantified along each study gradient. From the diagram, fish that successfully overwinter in habitat n will migrate from wintering areas to growing areas for breeding, feeding, or refuge in the growing

69 season. Mature fish will spawn at a certain period of time and have offspring in the same season. In the late growing season, young fish will grow up and aggregate with the adults to overwinter in the wintering areas. The same pattern of fish migration within habitats is assumed to occur in habitat n–1 and n+1. Concurrently, migration of a target fish species at both life stages also occurs among adjacent habitats.

Figure 3.2. A conceptual diagram of fish distribution within and among habitats. Migration rates of fish at two life stages in the growing season (yk, 0+ yr old; ak, 1+ yr old) and for all fish in winter (k) among habitats n, n–1, and n+1 are dependent on fish habitat preference in the growing season (yHP, 0+ yr old; aHP, 1+ yr old) and in winter (HP), respectively. The three graphs below show the estimated preference indexes (Pi) of the three example habitat variables that were used to compute the geometric mean HP for the target fish species at each life stage.

The migration rate (k) of each fish species varies in relation to the species habitat preference (HP) along each study gradient. For instance, if a fish species preferred

70 habitat n–1 to habitat n, more individuals of that species would migrate to habitat n–1, rather than stay in habitat n. In contrast, if suitable habitat conditions are not available in habitat n–1 and n+1, the fish species’ populations are most likely to distribute only within habitat n. In the AEM, the k of each target fish species among adjacent habitats was derived from the fish species HP, such that k = (1–HP) over time t of the simulation. The HP of each target species at each life stage is estimated using a geometric mean model, equation (3), as follows.

1 / N ⎛ N ⎞ ⎜ ⎟ , (3) HP = ⎜ ∏ Pi x ⎟ ⎝ x =1 ⎠

where Pix is a fish preference index (Degraaf and Bain, 1986) for each habitat variable x. The model in equation (3) was employed from the habitat suitability index method of Bovee (1986), which has been widely applied for determining health of fish habitats (e.g., Pajak and Neves 1987, Reyes et al. 1994, Guay et al. 2000) and characterizing species distributions (O’Connor 2002). In equation (3), Pix, is expressed as

(Fj / Ft) Pi = (4) x (Ej / Et) where Fj is the number of individuals of a target fish species observed at an intensity interval j of a habitat variable x, Ft is the total number of individuals of a target fish species observed from all intensity intervals j of a habitat variable x, Ej is the number of observations for a habitat variable x at an intensity interval j, and Et is the total number of observations for a habitat variable x from all intensity intervals j.

To obtain the index between 0 and 1, the Pix on each interval j is normalized (i.e.,

Pix,j/Pix,jmax). Examples functions of the Piwater depth, Picurrent velocity, and Piwater temperature of

71 bluntnose minnow at both life stages in the growing season in habitat n–1 were shown in Figure 3.2. The fish HP is then computed as a product of the fish species Pix for all habitat variables N. To derive the HP value between 0 and 1 over time t of the simulation, the HP, equation (3), is normalized with the power of 1/N. A summarized equation incorporating both population characteristics and HP to simulate the AEM for each target fish species along each study gradient is dP n = {}{}(ak × A ) + ( yk × Y ) + (ak × A ) + ( yk × Y ) + f (Y ) dt n −1,n n −1 n −1,n n −1 n +1,n n +1 n +1,n n +1 n (5) ⎧⎛ aHP ⎞ ⎛ yHP ⎞ ⎫ ⎜ n −1 ⎟ ⎜ n −1 ⎟ − ⎨⎜ ⎟ × (ak n × An ) + ⎜ ⎟ × ( yk n × Yn )⎬ ⎩⎝ aHP n −1 + aHP n +1 + z ⎠ ⎝ yHP n −1 + yHP n +1 + z ⎠ ⎭ ⎧⎛ aHP ⎞ ⎛ yHP ⎞ ⎫ ⎜ n +1 ⎟ ⎜ n +1 ⎟ − ⎨⎜ ⎟ × (ak n × An ) + ⎜ ⎟ × ( yk n × Yn )⎬ − (rd × Pn ), ⎩⎝ aHP n −1 + aHP n +1 + z ⎠ ⎝ yHP n −1 + yHP n +1 + z ⎠ ⎭

where Pn is the population of a target fish species in habitat n, and akn and ykn are migration rates of 1+ year-old fish (A) and 0+ year-old fish (Y), respectively, out of habitat n. Similarly, akn–1,n and akn+1,n are migration rates of A from habitat n–1 and n+1, respectively, to habitat n, and ykn–1,n and ykn+1,n are migration rates of Y from habitat n–1 and n+1, respectively, to habitat n. In this equation, akn equals (1–aHPn) over time t. The rest of the migration rates are computed in the same way. The aHPn, aHPn–1, and aHPn+1 are the HPs of A in habitats n, n–1, and n+1, respectively.

Likewise, the yHPn, yHPn–1, and yHPn+1 are the HPs of Y in habitats n, n–1, and n+1, respectively.

A small number z (e.g., z ≤ 10–6) is added to the model to avoid zero denominators in case the species HP in any habitat becomes zero at a certain time t of simulation. The birth of a target species in habitat n, for example, is represented by a logistic growth model, f(Yn), where f(Yn) equals rb × An × (1–Yn/Cy). That is, the birth of Yn depends on An abundance, birth fraction rb, and carrying capacity Cy of Yn. The

72 natural mortality of An in the growing season is ignored, but it is considered in winter, and it is computed by multiplying the death fraction rd of fish over the winter months by the fish population that survives winter in habitat n.

Model simulation and validation The model was initiated by assigning individuals of all target fish species to certain FL stations (Table 3.1) where the fish species were observed in high abundance (>70% of the total observed individuals along the FL gradient) in August

2004. Because of insufficient information to estimate carrying capacity Cy for each target species at each station, the same value of Cy for each target species was assigned to all stations in the growing season (Table 3.1). Ranges of birth fraction rb in the growing season and death fraction rd in winter for each target species were assigned to all stations (Table 3.1).

Table 3.1. Initial values of the population variables (Cy, rb, and rd) and individuals of each target fish species for simulating the abundance exchange model.

Initial no. of Fish species Station1 C 2 r 3 r 4 individuals y b d Yellow perch 25,000 FL1, ST1 6,000 0.40–0.60 0.05–0.10 (Perca flavescens ) Bluegill 55,000 FL1, ST1 8,000 0.40–0.60 0.05–0.10 (Lepomis macrochirus ) Logperch5 6,000 FL5 2,000 0.22–0.24 0.08–0.10 (Percina caprodes ) Bluntnose minnow 3,000 FL5, FL12, 2,500 0.30–0.35 0.12–0.15 (Pimephales notatus )ST7 Fantail darter 8,000 FL12, ST8 2,500 0.30–0.40 0.08–0.09 (Etheostoma flabellare )

1The initial number of individuals of each target species was assigned to certain stations on the Floodwood (FL) and Sterling (ST) gradients to start the simulation on December 1st. Unassigned stations had zero individuals of fish at the starting time of the simulation. 2 The same value of carrying capacity Cy (individuals) for each target species was assigned to all FL and ST stations in the growing season.

73 Table 3.1 (Continued).

3 -1 The same range of birth fraction rb (month ) for each target species was assigned to all FL and ST stations to generate series of uniformly distributed random values of rb in the growing season. 4 -1 The same range of death fraction rd (month ) for each target species was assigned to all FL and ST stations to generate series of uniformly distributed random values of rd in winter. 5Distribution pattern of logperch was not modeled on the ST gradient.

From these initial conditions, the model generates series of uniformly distributed random values of population variables for each fish species in the associated seasons. To explore how the model would behave in a system in which a sharp change in each population variable might occur for some reason (e.g., food increase/decline, predator), ±50% changes in value of each population variable were assigned one at a time at all FL stations to rerun the model. Mean values of each habitat variable from the 180 samples along the FL gradient were classified into intervals using natural breaks (Jenks), one of the classification types in ArcGIS 9.1. This method is useful for a data set that contains relatively big jumps in the data values. Thus, for each habitat variable, similar data values were grouped together in the same interval, and the differences between intervals were maximized. A fish species Pi for each habitat variable at each classified interval was then estimated using equation (4) (Table 3.2). Series of IF-THEN-ELSE statements (logical functions) were used to incorporate the species Pi at each interval for all habitat variables (Table 3.2) into the AEM for computing the species HP, equation (3). The species’ migration rate k in the model was then converted from the species HP and used to quantify abundance exchange of that species among FL stations in the growing season.

74 Table 3.2. Preference index (Pi) for each habitat variable of five modeled fish species at two life stages (0+yr and 1+yr old) estimated from the fish and habitat data observed along the Floodwood gradient over the three summer months (June–August) of 2004.

Pi Habitat variable Yellow perch Bluegill Logperch Bluntnose minnow Fantail darter 0+yr 1+yr 0+yr 1+yr 0+yr 1+yr 0+yr 1+yr 0+yr 1+yr

Water depth (m) 0.10-0.3 0.10 0.16 0.17 0.10 1.00 1.00 0.62 1.00 1.00 0.73 0.31-0.6 0.03 0.00 0.17 0.02 0.17 0.29 0.75 0.43 0.73 1.00 0.61-1.0 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.61 0.49 0.77 1.01-4.0 1.00 1.00 1.00 1.00 0.37 0.09 0.08 0.22 0.00 0.00 Current velocity (m/s) 0.03-0.06 1.00 1.00 1.00 1.00 0.19 0.00 0.05 0.12 0.00 0.00 0.07-0.12 0.18 0.46 0.59 0.90 1.00 1.00 1.00 1.00 0.92 0.68 0.13-0.24 0.04 0.00 0.07 0.06 0.66 0.91 0.48 0.41 0.37 0.70 0.25-0.36 0.04 0.00 0.00 0.06 0.76 0.86 0.64 0.60 1.00 1.00 0.37-0.48 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.13 0.42 Water temperature (°C) 16.1-18 0.71 0.68 0.90 0.57 0.51 0.58 1.00 0.45 0.82 0.74 18.1-20 1.00 0.45 1.00 0.27 1.00 1.00 0.59 0.73 1.00 1.00 20.1-22 0.27 0.00 0.63 0.32 0.00 0.00 0.49 0.27 0.18 0.66 22.1-24 0.30 0.27 0.00 0.28 0.51 0.00 0.19 0.10 0.05 0.17 24.1-27 0.78 1.00 0.90 1.00 0.57 0.78 0.56 1.00 0.02 0.16 Aquatic plants1 0.01-0.3 0.00 0.00 0.00 0.00 0.00 0.00 0.65 0.64 1.00 1.00 0.31-0.6 0.01 0.00 0.04 0.08 0.11 0.12 1.00 0.39 0.36 0.68 0.61-1.0 0.54 0.94 0.31 0.82 1.00 1.00 0.59 1.00 0.01 0.03 1.01-1.5 0.33 0.25 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 1.51-3.0 1.00 1.00 1.00 0.99 0.00 0.00 0.00 0.31 0.00 0.00 Algae1 0.00-0.2 1.00 1.00 0.92 1.00 0.33 0.09 0.15 0.30 0.00 0.00 0.21-0.6 0.40 0.26 0.00 0.47 0.23 0.00 0.04 0.22 0.08 0.05 0.61-1.2 0.09 0.00 0.00 0.07 1.00 1.00 1.00 0.37 0.01 0.10 1.21-2.0 0.28 0.14 1.00 0.11 0.00 0.00 0.13 0.15 0.06 0.26 2.01-3.0 0.06 0.08 0.24 0.05 0.42 0.33 0.87 1.00 1.00 1.00 Woody debris1 0.01-0.3 0.08 0.00 0.00 0.08 1.00 1.00 0.35 0.93 1.00 1.00 0.31-0.6 0.22 0.22 0.54 0.33 0.46 0.62 1.00 1.00 0.48 0.91 0.61-1.0 0.88 0.52 0.17 0.58 0.28 0.00 0.28 0.23 0.02 0.18 1.01-1.5 1.00 1.00 1.00 1.00 0.00 0.00 0.08 0.17 0.00 0.00 Mud (%) 0.0-5 0.04 0.08 0.11 0.04 0.51 0.52 0.62 1.00 1.00 1.00 5.1-15 0.08 0.00 0.24 0.06 0.56 1.00 1.00 0.34 0.01 0.07 15.1-30 0.25 0.47 0.29 1.00 1.00 0.24 0.13 0.41 0.00 0.00 70.1-100 1.00 1.00 1.00 0.54 0.00 0.00 0.05 0.21 0.00 0.00

75 Table 3.2 (Continued).

Pi Habitat variable Yellow perch Bluegill Logperch Bluntnose minnow Fantail darter 0+yr 1+yr 0+yr 1+yr 0+yr 1+yr 0+yr 1+yr 0+yr 1+yr

Sand (%) 0.0-5 0.00 0.00 0.00 0.00 0.00 0.00 0.40 0.29 0.06 0.29 5.1-15 0.19 0.28 0.33 0.17 0.80 1.00 1.00 1.00 1.00 1.00 15.1-30 1.00 1.00 1.00 0.30 0.00 0.00 0.36 0.30 0.12 0.20 30.1-50 0.27 0.58 0.32 1.00 1.00 0.26 0.16 0.35 0.00 0.00 Gravel-cobble (%) 0.0-5 1.00 0.99 1.00 0.38 0.00 0.00 0.22 0.16 0.03 0.19 5.1-15 0.56 0.77 0.86 0.68 0.00 0.00 0.26 0.14 0.07 0.82 15.1-30 0.28 0.42 0.47 1.00 0.22 0.19 0.98 0.44 1.00 1.00 30.1-50 0.52 1.00 0.86 0.54 1.00 1.00 1.00 1.00 0.07 0.17 Rock-bedrock (%) 0.0-5 1.00 1.00 1.00 0.54 0.00 0.00 0.05 0.12 0.00 0.00 5.1-15 0.25 0.47 0.29 1.00 0.77 0.16 0.11 0.24 0.00 0.00 30.1-50 0.00 0.00 0.00 0.00 0.00 0.00 0.55 0.25 0.21 0.54 50.1-75 0.12 0.16 0.24 0.10 1.00 1.00 1.00 1.00 1.00 1.00 75.1-100 0.00 0.00 0.06 0.00 0.00 0.00 0.24 0.21 0.07 0.67

1Abundance of each cover type was numerically coded, 0 = absent, 1 = low, 2 = common, and 3 = high.

To simulate habitat change along the FL gradient in the growing months, series of uniformly distributed random values of all dynamic habitat variables (except substrates) were generated based on the observed values from the summer of 2004 (Figure 3.3). The distributed random values of each habitat variable along the FL gradient in April and May were generated from the observed values in June. The distributed random values of the habitat variables in September and October were generated from the observed values in August. The substrate composition observed at each station along the FL gradient (Figure 3.3) was assumed to be constant across seasons (Hatzenbeler et al. 2000). Because the target fish species and habitat data were not collected during the winter, the HPwinter of each target species was assumed to

76 be the same as the species HP in the late growing season (October) estimated by the model.

Figure 3.3. Categorized mean values of all habitat variables (except substrate composition) observed monthly during June–August 2002 along the Floodwood gradient. Abundance of aquatic plants, algae, and woody debris was rated from absent (0) to high (3). Percentage composition of four substrate types was averaged over the three months along the gradient.

77 The model was validated against the fish and habitat data collected on the ST gradient in the summer of 2002. The AEM with all estimated Pi for each species from the FL gradient was used to predict the distribution pattern of the same species (except logperch) along the ST gradient. Logperch was not modeled on the ST gradient because this species was not observed at all ST stations over the three summer months (June–August 2002). To simulate the AEM on the ST gradient, the same individuals of the target species assigned to the FL stations were assigned to the particular ST stations (where the target fish species accounted for more than 70% of the total observed individuals in August 2002) (Table 3.1). The same ranges of Cy, rb, and rd of each modeled species assigned to the FL stations were also applied to all ST stations (Table 3.1). Series of uniformly distributed values of the habitat variables along the ST gradient in the growing months were generated from the values observed on this gradient during the summer of 2002 (Figure 3.4) based on the same assumption made for those on the FL gradient. STELLA® 7.0.3 Research (High Performance Systems, Inc. 2002) was used to run the AEM for a long-term prediction (100 hundred years) on each study gradient starting with month 0 on December 1st and ending at 1,200 months on November 30th. The interval of time between calculations (dt) was set to a small value of 0.0625 months to avoid artifactual delays and dynamics during the software calculation. The artifactual delays result from the fundamental conceptual distinction between stocks (state variables) and flows. Stocks exist at a point in time, whereas flows exist between points in time. Consequently, the two components cannot coexist at the same instant in time. For example, during migration of fish from one habitat to another habitat, it will take an “instant” for fish from the first habitat to actually arrive in the second habitat.

Figure 3.4. Categorized mean values of all habitat variables (except substrate composition) observed monthly during June–August 2002 along the Sterling gradient. Abundance of aquatic plants, algae, and woody debris was rated from absent (0) to high (3). Percentage composition of four substrate types was averaged over the three months along the gradient.

79 The “instant” constitutes the software’s artifactual delay. As long as dt is defined as a relatively small value, the artifactual delays are usually insignificant (High Performance Systems, Inc. 2002). Avoiding artifactual dynamics is the second reason for choosing a small dt value for running the model that generates a dynamic pattern of behavior caused by the way the software is performing its calculations. If the dt value is too large, the flow volumes will be too large, and pathological results are generated. However, the two problems are unlikely to occur when running the model with the dt at or below 0.25 (High Performance Systems, Inc. 2002).

The chi-square (χ2) statistic was used to test the goodness of fit between the mean predicted abundance (%) on August 31st from the 100-year simulation and the observed abundance (%) of the target fish species in the same month on each study gradient.

RESULTS

Model prediction Populations of the five modeled fish species showed seasonal patterns of distribution along the FL gradient over the 100-year simulation following results for yellow perch (Figure 3.5). The fish populations declined in winter as a result of winter mortality and increased in the growing season owning to birth and growth of offspring. Populations were relatively stable for a certain time period in the growing season after the spawning period when no offspring were born. After each growing season, the fish populations rapidly declined during the subsequent winter. As expected, the fish population size increased when higher value of either Cy (Figure

3.5B) or rb (Figure 3.5C), or lower value of rd (Figure 3.5D) was assigned one at a time for simulating the model.

Figure 3.5: Seasonal distribution pattern of yellow perch over the 100-year simulation (A). The fish population size exponentially decreased or increased for the first 10- year period of the simulation when changes of 50% of carrying capacity Cy (B), birth fraction rb (C), or death fraction rd (D) were assigned to the abundance exchange model, but the population size returned to the seasonal pattern over the long-term prediction.

81 In contrast, the fish population size declined when lower value of either Cy (Figure

3.5B) or rb (Figure 3.5C), or higher value of rd (Figure 3.5D) was used one at a time to run the model. Results from the model simulation showed that the fish population sizes exponentially increased or decreased for short-term predictions (e.g., around the first 10 years of the simulation) when any one of the three population variables was disturbed. The fish population sizes returned to their seasonal patterns of distribution (Figure 3.5B–D) when the population growth reached the carrying capacity of the system. Across both temporal and spatial scales, the mean predicted percentage abundance of each target species in August from the 100-year simulation agreed with the

2 observed percentage abundance in August 2004 (χ predicted < 19.68, p>0.05, d.f. = 11) on the FL gradient (Figure 3.6). Although the yellow perch abundance was somewhat underestimated at FL1 and overestimated at FL3 (Figure 3.6), the prediction was, in

2 general, acceptable (χ predicted = 12.15, i.e., < 19.68, p>0.05, d.f = 11) compared to the observation.

2 The AEM could efficiently quantify the distribution pattern of bluegill (χ predicted =

2 3.84) and logperch (χ predicted = 9.35) on the FL gradient with less underestimation or overestimation (Figure 3.6). The predicted abundance for bluntnose minnow and

2 2 fantail darter was also acceptable (χ predicted = 16.37 for bluntnose minnow, χ predicted = 15.33 for fantail darter), although the overestimation or underestimation occurred at some FL stations (Figure 3.6).

Figure 3.6. Observed abundances of yellow perch (YP), bluegill (BG), bluntnose minnow (BM), logperch (LP), and fantail darter (FTD) along the Floodwood gradient in August 2004 were compared with the mean predicted abundance ± standard deviation of the prediction of the associated fish species in August over the 100-year simulation.

83 The zero Pi of yellow perch (Table 3.2) for habitats with moderate-to-high proportions of rock-bedrock above FL5 (Figure 3.3) limited the distribution of this species to within FL1–FL5 as predicted by the model over the 100-year simulation (Figure 3.7). The yellow perch population mainly aggregated at FL1 during winter and early spring and migrated to FL2 in May and July, although a portion migrated to FL4 and FL5. In the late growing season, most yellow perch moved downstream to FL1 (Figure 3.7). Like yellow perch, the abundance exchange of bluegill occurred between FL1 and FL5. The bluegill population mainly aggregated at FL1 during winter and early spring. In June and July, bluegill were most abundant at FL2, but the population spread to FL1 and FL3 in August, and most moved downstream to FL1 in the late growing season (Figure 3.7). The zero Pi of logperch for habitats containing high proportions of mud, such as FL1–FL2, or moderate-to-high rock-bedrock content, such as FL6–FL12, defined the distribution range of this species within FL3–FL5 (Figure 3.7). Logperch mainly aggregated at FL4 and FL5 in the growing season, but at FL3 and FL4 in November (Figure 3.7). Most fantail darter aggregated at the upstream portion on the FL gradient across a year. Relatively few of them distributed to FL4 and FL5 in the growing season (Figure 3.7). The distribution of fantail darter below FL4 (Figure 3.7) was restricted by its zero Pi for habitats with little-to-no rock-bedrock content at FL1–FL3 (Figure 3.3). Unlike all other target species, bluntnose minnow distributed throughout the FL gradient. In general, most bluntnose minnow aggregated at FL4, FL5, and FL12 across a year (Figure 3.7).

Figure 3.7. Mean predicted abundance of yellow perch (YP), bluegill (BG), bluntnose minnow (BM), logperch (LP), and fantail darter (FTD) among months and stations on the Floodwood gradient over the 100-year simulation.

85 Model validation The AEM with the estimated Pi for each target species on the FL gradient was validated against the observed abundance of the same target species on the ST gradient in the summer of 2002. The AEM predictions for yellow perch and fantail darter

2 agreed with the observations (Figure 3.8) along the ST gradient (χ predicted < 14.07, p>0.05, d.f. = 7).

Figure 3.8. Observed abundance of yellow perch (YP), bluegill (BG), bluntnose minnow (BM), and fantail darter (FTD) along the Sterling gradient in August 2002 were compared with the mean predicted abundance ± standard deviation of the prediction of the associated fish species in August over the 100-year simulation.

86 However, the model prediction was inconsistent with the observations of bluegill and bluntnose minnow at some stations on the ST gradient. Significant differences between predicted and observed abundances of bluegill occurred at ST7 and ST8

2 (χ predicted > 3.84, p≤0.05, d.f. = 1). During the field sampling in August 2002, few (6.3%) bluegill were observed at ST7 and fewer (3.1%) at ST8, while the average abundance of bluegill over the 100-year simulation in the same summer month was overestimated (16.2%) at ST7 and underestimated (0.3%) at ST8 (Figure 3.8). Significant differences between predicted and observed abundances of bluntnose minnow occurred at ST1 and ST3 (Figure 3.8). Few (3.5%) bluntnose minnow were observed at ST1 and none at ST3 in August 2002, while the prediction of this species was underestimated (0.3%) at ST1 and overestimated (4.7%) at ST3 (Figure 3.8). According to the distribution patterns of the target species among months and ST stations over the 100-year simulation (Figure 3.9), yellow perch distributed within the range between ST1 and ST6, whereas bluegill distributed throughout the ST gradient (Figure 3.9). Most yellow perch and bluegill aggregated at ST1 in winter and early spring and migrated between ST1 and ST2 in the growing season. As on the FL gradient, bluntnose minnow distributed throughout the ST gradient in the growing season. Most of them aggregated at ST7 across a year (Figure 3.9). Predicted distribution of fantail darter was between ST7 and ST8 across a year (Figure 3.9). In general, the predicted abundance patterns of the modeled species in the summer months were consistent with those from the field observations in the same time period.

Figure 3.9. Mean predicted abundance of yellow perch (YP), bluegill (BG), bluntnose minnow (BM), and fantail darter (FTD) among months and stations on the Sterling gradient over the 100-year simulation.

88 DISCUSSION

The modeled fish species showed relatively stable seasonal patterns of distribution over the long-term prediction, and this was consistent with the findings of Reyes et al. (1994), which revealed a stable seasonal pattern of pigfish abundance in Laguna de Terminos, Mexico. The seasonal pattern of fish distribution can be explained in terms of the negative and positive feedback loops (Ford 1999, Sterman 2000). The negative feedback loop represents the contradictory conditions affecting fish distribution within habitats between winter and the growing season, i.e., if more fish rapidly decline over winter in these habitats, fewer fish will survive to have offspring or grow up in the successive growing season. The positive feedback loop represents the supportive condition increasing fish distribution within and across habitats between the two seasons, i.e., if more young fish are born and more fish migrate from the other habitats, a larger fish population in these habitats will be available to overwinter. These consequences produced the seasonal patterns of fish distribution along Lake Ontario’s coastal zone.

While the population variables (Cy, rb, and rd) shaped the population size of fish in time, the model’s key variable, HP, accounted for fish distribution in space. As the geometric mean of a species Pi for both static and dynamic habitat variables, HP could capture the local range of species distribution and reflect the influence of habitat changes (Morrison et al. 1985, Guay et al. 2000) on the abundance of a species within its local range. Specifically, the estimated Pi (Table 3.2) for the static variable (i.e., substrate composition) defined the local distribution ranges of all modeled species on the FL gradient and most of the species (except bluegill) on the ST gradient. Changes in major substrate type(s) from a combination of mud and sand at FL3 to a combination of gravel-cobble and rock-bedrock at FL4–FL5, and rock-bedrock at FL6

89 (Figure 3.3) marked these stations as the zone of transition in grain and extent (Morrison and Hall, 2002) of the bottom landscape between lentic (embayment) and lotic (stream) systems on the FL gradient. Similarly, changes in major substrate type(s) from mud at ST6 to a combination of gravel-cobble and rock-bedrock at ST7 (Figure 3.4) indicated the distinct transition of the bottom landscape between the lentic-lotic systems on the ST gradient. In accordance with its local ranges of distribution in the upstream portions of riffles and runs (Mundahl and Ingersoll 1983, Gray and Stauffer 1999) on both gradients, fantail darter showed a high preference for shallow water, higher current velocity, and rock-gravel streambeds (Table 3.2). In contrast, yellow perch and bluegill represent embayment species that prefer deeper water habitat with common to abundant aquatic plants and low current velocity (e.g., Harrington 1947, Wells 1968, Helfman 1979, Hatzenbeler et al. 2000, Paukert and Willis 2002). Logperch distributed within the lentic-lotic transition zone (FL3–FL5), habitats with moderate current velocities, shallow water, and a combination of gravel-cobble and rock- bedrock were preferred by this species (Smith 1985). Bluntnose minnow distributed in a wide variety of water habitats (Trautman 1957, Becker 1983), but this species preferred hard-bottomed, sandy, or gravelly shallows of pools for spawning activity (Houston 2001) in the growing season. Consistently, high numbers of 0+year-old bluntnose minnow were both observed and predicted at FL5 and FL12 that provide suitable nursery substrates for this species. Results from the model validation on the ST gradient revealed high levels of agreement between the model predictions and the field observations for yellow perch and fantail darter, but the predictions were not consistent with the observations for bluegill and bluntnose minnow. The disagreement between predicted and observed abundances of bluegill and bluntnose minnow at certain portions on the ST gradient

90 might be due to several causes. For bluntnose minnow, the small sample size (i.e., one sample/station/month) of fish along the ST gradient might not well represent the distribution pattern of this species along the gradient. Furthermore, increasing growth of aquatic plants at the downstream and midstream portions of the gradient in the late summer month could impede the ability to catch small fish. These factors might explain why bluntnose minnow were not collected at most downstream-midstream stations (ST3–ST6) on the ST gradient. Based on the broad tolerance of bluntnose minnow for a wide variety of habitats, except the deeper waters of lakes and rivers (Trautman 1957, Becker 1983), the model prediction revealed the distribution pattern of this species throughout the ST gradient as being the same as that on the FL gradient. The model parameters (i.e., Pi) for bluegill need some improvement to enhance the ability of prediction on the ST gradient. Whereas this species was observed throughout the ST gradient, results from the model simulation showed low efficiency in quantifying abundance exchange of this species in the upstream portion of the gradient. Pajak and Neves (1987) proposed three primary concerns that are generally implicit in model validation procedures based on fish species Pi, one has to ensure that (1) the fish species data to be validated are accurately collected, (2) the habitat condition is accurately measured, and it is limited primarily by the habitat variables included in the model, and (3) the sampling sites are representative sub-samples of the target fish species’ habitats. In this study, it is believed that methods of collecting both fish and habitat data were appropriate. Nevertheless, it was noticeable that the sampling sites on the FL gradient might not well represent all kinds of utilized substrates for bluegill on the ST gradient. While the Pi for the rest of the habitat variables fell within the observed ranges for bluegill on both study gradients, the Pi for substrate variables showed some gaps between the quantified intervals (Table 3.2). This observation indicates that

91 although the substrates along the FL gradient were collected at a fine scale (900 samples), they were still unable to represent diverse proportions of the bottom substrates occupied by bluegill on the ST gradient. Thus, the Pi for substrate variables for this species may need to be improved for better model prediction. Similar gaps in sampling coverage did not occur for the other modeled species. Results from this study suggest that the species Pi for a given habitat variable estimated from one study area may not completely represent the species Pi in another study area (Pajak and Neves 1987, O’Connor 2002). Thus, to apply the AEM for predicting fish species distribution in other water systems, one should crosscheck the fish species Pi from existing data at both local and regional scales before incorporating them into the AEM for computing the species HP. This is particularly necessary for fish species with broad tolerances on the habitat variables that can powerfully define local ranges of fish distribution (Hatzenbeler et al. 2000) or control distribution of fish species from cell to cell (e.g., substrate types; Reyes et al. 1994). For different water systems with different environmental conditions, influential habitat variables should be considered for developing fish species Pi. Additionally, in large water systems where fish movements in both vertical and horizontal directions are equally important, two or more dimensions of fish distribution may need to be added in the model (e.g., Morrison et al. 1985, Reyes et al. 1994). Overall, the AEM empirically developed from the fieldwork data performs rather well for quantifying distributions of the coastal fish species assemblages along Lake Ontario’s embayment-stream gradients. With its flexible model structure that can accommodate array functions and multiple differential equations for both static and dynamic components, the AEM can be modified to determine distribution patterns of other types of organisms in different environmental and geographic systems.

92 70104 along the Hudson River estuary y monitoring program using beach seine ichness of fish by region (RE) r plant utilities’ Hudson River estuar 0700030014717300 390900000102 000000050901 495 1919 1114 906 1038 1154 596 627 512 904 780 1824 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 2464 5132 5850 12268 13530 14429 7118 6885 4544 83021544 7330 3027 6382 1130 786 199 74 12 31 0 6 64 1 12700 25011 29086 33150 3169412037 58033 6314 4465763966 29278 1525 72611 24640 45225 662 47865 192501 63858 188958 166113 90 45563 243664 205780 149394 21 143780 16587527259 145809 10 60760 55534 23359 15016 11946 7610 4969 6054 9021 13376 8256 Fish species Fish oxyrinchus) oxyrinchus) (Alosa aestivalis) (Alosa (Alosa mediocris) (Alosa (Morone saxatilis) (Anguillar rostrata) (Alosa sapidissima) (Microgadus tomcod) (Petromyzon marinus) (Petromyzon (Acipenser oxyrinchus(Acipenser (Alosa pseudoharengus) nadromous lewife merican shad tlantic sturgeon tlantic tomcod (Osmerus mordax mordax) Catadromus Striped bass Sea lampreySea A A A A A American eel Hickory shad Hickory Blueback herringBlueback Rainbow smelt Rainbow Appendix 2.1. Cumulative abundance and species gradient (RE1–RE12) from the five power gradient (RE1–RE12) survey during 1974 to 2001.

93 00006 001000000000 010000000001 95 7955 1642115 31796 40825 1646 20664 9930 955 8609 1668 16774 2695 18514 230 33734 33322 381 4521136265282543812 639 41614 255 1068 114 573 202 26 39 38 44 35 46 23 86 331 1774282 1774 726 697996 2782 219 4152 2047 1607 509 31 283 70 11 30 30 3 5 12 8 0 15 36 150 1089 3050 4 1781 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 2416 1923 1432 918 505 556 126 25 95 129 365 25 20473 95912 67889 12073 4011 1290 204 1 143 2 1 2 ) ) ) ) ) ) aculeatus) Fish species Fish heteroclitus) Ameiurus catus Menidia menidia Menidia ( ( (Menidia beryllina) (Menidia choker Apeltes quadracus Apeltes ( Trinectes maculatus (Syngnathus fuscus) g Dormitator maculatus Fundulus diaphanous Fundulus ( (Fundulus heteroclitus (Fundulus ( ( (Gasterosteus aculeatus (Acipenser brevirostrum)(Acipenser Appendix 2.1 (Continued). Threespine stickleback Northern pipefish Estuarine silverside Atlantic killifish Banded sleeper Fat stickleback Fourspine Ho silverside Inland White catfish White Shortnose sturgeonShortnose Mummichog

94 001000001100 0 193190284734443059407131 0020200312687 100051000441 00000000004100 0131111103266 002000100000 1113112100300200 1269 12 160 8 526 6 1047 810 39 1094 20 835 493 4 385 0 456 0 566 450 0 2 54 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 9677 79929 128023 45597 37570 27099 22055 23737 17297 38740 41222 19336 ) ) ) ) ) ) ) ) ) s atratulus y (Continued) Fish species Fish (Mugil curema) (Mugil ill Ameiurus melas ( g Culaea inconstans Culaea Notropis bifrenatus Notropis Salvelinus fontinalis Salvelinus Pimephales notatus Pimephales ( ( (Morone Americana) (Morone Labidesthes sicculus Labidesthes Rhinichth ( ( Lepomis macrochirus Lepomis ( ( ( Pomoxis nigromaculatus Pomoxis ( Appendix 2.1 (Continued). White perch Estuarine Estuarine Black bullhead Black crappie Blacknose dace Blue Bluntnose minnow Bridle shiner Brook silverside stickleback Brook Brook trout White mullet Freshwater

95 0134154101025 47001200000001 759024129611001413 425000102252321 72800002230010017 150 383 106000421021418120 990180243571025 040230010012 97000002000000 2100908113413293142917510 173 12 376 456 481 420 303 271 205 245 161 124 187 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 ) ) ) ) ) ) ) ) ) ) ) ) ) gmaea minnow y y Continued (

prinus carpio prinus Umbra limi Umbra Esox niger Esox niger shiner Fish species Fish ( y ( y bognathus regius bognathus C Umbra p y ( Notropis cornutus Notropis Salmo trutta trutta ( Notropis amoenus Notropis Ictalurus punctatus Ictalurus ( Ictalurus nebulosus ( ( H ( ( ( Exoglossum maxillingua Exoglossum Semotilus atromaculatus ( ( Appendix 2.1 Continued). Freshwater Brown bullhead trout Brown Central mudminnow Chain pickerel catfishChannel Comel Common carp Common shiner Creek chub minnow Cutlips Eastern mudminnow Eastern silver

96 20500325102124157 3801006233819227 44001000001300 1026 154 119 870010000001024 123 105013103000200 3876 1180001000121428 7355 54 215 995 279 658 461 610 316 231 156 169 2233 9410 138 194 1029 354 85 1433 1006 272 5343 323 3417 432 419 5812 2212 4634 117 441 2633 80 706 3333 609 1515 72 2890 820 74 541 118 469 402 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 ) ) ) Continued (

Fish species Fish vermiculatus) Percina caprodes Percina (Esox americanus ( (Lepomis cyanellus) (Lepomis (Semotilus corporalis) (Notropis atherinoides) (Notropis (Pimephales promelas) (Pimephales (Aplodinotus grunniens) (Aplodinotus (Micropterus salmoides) (Dorosoma cepedianum) Appendix 2.1 (Continued). Carassius auratus auratusCarassius Fathead minnow Fathead Gizzard shad Freshwater drum Freshwater Fallfish Largemouth bass Freshwater Emerald shiner crysoleucas) (Notemigonus Goldfish Grass pickerel Golden shiner Golden Logperch Green sunfish (

97 002000520010 000411000024 000000010001 2000132134134 0001110003347 000000000020 000000010010 3003210120217112 44 20811157322113141715156 733 199 591 1543 1315 1033 1323 749 665 18 699 5724 6576 6508 5497 5025 6704 4811 3680 1829 1578 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 ) (Continued) Fish species Fish americanus) (Esox lucius) Esox americanus (Lepomis auritus) (Lepomis ( (Lepomis gibbosus) (Lepomis megalotis) (Notropis volucellus) (Notropis anogenus) (Notropis Ambloplites rupestris Ambloplites (Salvelinus gairdneri ) gairdneri (Salvelinus ( (Hypentelium nigricans) (Rhinichthys cataractae)(Rhinichthys Appendix 2.1 (Continued). Rock bass Redbreast sunfish Northern pike Longnose dace Rainbow trout Rainbow Freshwater Mimic shiner Mimic Northern hog sucker Pumpkinseed Longear sunfish Redfin pickerel Pugnose shiner

98 000000000001 000083037523225 000000010000 1 341133161323971 0 3 3 7 117 5000010030202 13000000000011 5 590000000000040 41000000010030 79 151 159 158 346 35 588 3203 8149 13026 8656 5282 4509 6099 8386 5965 15717 509 1784 12113 20052 50081 24333 35546 41154 24852 41041 42027 76124 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 ) ) ) ) ) ) ) ) ) Esox e g cross (Continued) Fish species Fish masquinongy) Percina peltata Percina Notropis procne Notropis prinella analostana prinella ( Notropis rubellus ( ( y Notropis spilopterus Notropis hudsonius Notropis er muskellun Etheostoma olmstedi ( ( C Micropterus dolomieui ( Esox lucius g ( ( ( tizostedion vitreum vitreumtizostedion Appendix 2.1 (Continued). Freshwater shiner Satinfin Shield darter Smallmouth bass Spotfin shiner Spottail shiner Swallowtail shiner darter Tessellated Ti (Percopsis omiscomaycus) Walleye Trout-perch Rosyface shiner

99 121212110024 0110421312178936 6000001000000 161 183 379 16 143 181 118 453 140 274130000000000 75 133 177 38 196 41 151 83 149 221 148 659 36592000000000 5000000000000 97 596 381 117 271 26 50 28 100 70 31 55 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 4086 27076 20420 6535 3316 9944 2241 2001 3034 1037 6751 91 193422 183599 23145 42592 41299 4581 4960 556 520 63 147 38 ) (Continued) Fish species Fish Anchoa mitchilli Anchoa ( (Ictalurus natalis ) (Ictalurus natalis (Perca flavescens) (Morone chrysops) (Morone (Selene setapinnis) (Selene (Pomoxis annularis) (Pomoxis (Strongylura marina) (Strongylura (Brevoortia tyrannus) (Brevoortia Appendix 2.1 (Continued). (Micropogonias undulatus) (Micropogonias White crappie Freshwater Atlantic croaker harengus) harengus (Clupea menhaden Atlantic Atlantic herring Atlantic Yellow perchYellow Atlantic moonfishAtlantic Yellow bullhead Yellow Marine Bay anchovy Atlantic needlefish Atlantic White sucker White bass (Catostomus commersoni ) (Catostomus commersoni

100 32400 140000000000 210000000000 000100000000 0114000000000 010000000000 010000000000 1091200000000 52136149218223400000 15432000000000 121913200000000 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 3751 6823 6591 4135 2078 386 59 ) ) ) ) ) ) ) ) ) ) ) ack j oxocephalus oxocephalus nodus foetensnodus y Fish species Fish (Continued) y Selene vomer Selene y Caranx hippos ( M ( Lutjanus griseus Lutjanus S horn sculpin ( snapper ( ( oxocephalus aenaeus oxocephalus y g Peprilus triacanthus Peprilus Lophius americanus Lophius Pomatomus saltatrix Pomatomus octodecemspinosus y ( ( ( Hippoglossina oblonga Hippoglossina ( M Tautogolabrus adspersus Tautogolabrus Appendix 2.1 (Continued). ( Bluefish Marine Butterfish Crevalle Cunner Fourspot flounder Goosefish Gra Grubb Inshore lizardfish Lon Lookdown (

101 8246000000000 4130000000000 100000000000 5120000000000 000000020000 001000000000 001000000000 3000690000000 1390171742100000 2152000000000 12313712000000000 286879677217341101021 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 ) ) ) ) ) ) ) ) ) ) ) ) sops ) y azer y fish g cis chuss g ob y y g ob Continued ( Fish species Fish g

espotted filefish h silverside Uroph g Gobiosoma bosc ( ( g Trachinotus blochii Trachinotus Prionotus carolinus Membras martinica Cantherhines pullus Cantherhines Merluccius bilinearis Merluccius ( ( ( Menticirrhus saxatilis Astroscopus guttatus Gobiosoma ginsburgi ( Stenotomus chr Stenotomus ( ( ( ( ( Sphoeroides maculatus Sphoeroides ( Appendix 2.1 (Continued). Marine Naked Northern kin Northern puffer Northern searobin Northern star Oran Permit hakeRed Rou Scup Seaboard hake Silver

102 417623300100100 2101400000000 110200000000 0000000000017 010000000000 100000000000 78028400000000 1260000000000 43197420000000 6565643514768866000000 102840000000000 1461679326106000000 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 ) ) ) s dentatus y Continued ( Fish species Fish

(Mugil cephalus) (Mugil (Urophycis regia) (Urophycis (Fundulus majalis) (Fundulus (Anchoa hepsctus) (Anchoa (Prionotus evolans) (Bairdiella chrysoura) (Bairdiella Paralichth (Etropus microstomus) ( (Leiostomus xanthurus) (Leiostomus Appendix 2.1 (Continued). (Eucinostomus argenteus) Smallmouth flounder Summer flounder Striped killifish Spotfin mojarra Marine perch Silver Spot Striped anchovy Striped Spotted goatfish Striped mullet Spotted hake Spanish mackerelSpanish Striped searobin Scomberomorus maculatus maculatus) (Pseudupeneus

(

103 910810000000000 710000000000 000000000001 001000000100 2000010000000 01118631014078 2720000000000 000000200000 000000300003 200040221101201 003000000000 108614331102202 07756251000000 739 171 361 196 134 108 61 1208 19 1612 545 2151 5754 28828 46055 70521 58371 90574 82058 78494 247 244 2792984250702000000 13 11 2102 186 0 947 0 2252 3782 0 2942 1458 0 2775 3290 1 1096 540 0 2032 RE1 RE2 RE3 RE4 RE5 RE6 RE7 RE8 RE9 RE10 RE11 RE12 Morone Fish species americanus) (Continued) (Tautoga onitis) (Tautoga (Cynoscion regalis) (Cynoscion (Pseudopleuronectes (Pseudopleuronectes (Scophthalmus aquosus) (Scophthalmus Appendix 2.1 (Continued). Weakfish Marine Esox spp. Fundulus spp. unidentified centrarchid unidentified clupeid unidentified cyprinid unidentified gobiid unidentified unidentified searobin unidentified sturgeon unidentified sucker unidentified spp. unidentified spp. unidentified spp. unidentified spp. Windowpane Unidentified species Winter flounder Tautog

104 Appendix 2.2. Multiple functions (aggregator, mediator, soft barrier, and hard barrier) of the ecotones (from all detected frequencies) on the Hudson fish species in each time period during 1974–2001.

Ecotone Function (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE2-RE4 Jul-86 42.1 15.8 18.4 23.7 38 Aug-74 22.0 30.0 22.0 26.0 50 Aug-79 36.4 25.0 18.1 20.5 44 Sep-82 38.5 25.6 18.0 17.9 39 Sep-88 51.5 9.1 15.2 24.2 33 Sep-96 30.0 20.0 30.0 20.0 20 Oct-75 38.9 22.2 22.2 16.7 36 Oct-98 22.6 19.4 6.5 51.5 31 Nov-80 23.5 29.5 23.5 23.5 17 Nov-86 36.0 28.0 12.0 24.0 25

105 Appendix 2.2 (Continued).

Ecotone Function (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE2-RE5 Jun-94 9.4 12.5 28.1 50.0 32 Aug-88 40.6 25.0 28.1 6.3 32 Aug-90 56.3 15.6 18.7 9.4 32

RE2-RE6 Jul-91 14.3 11.8 42.8 31.1 42 Aug-87 30.3 15.2 33.3 21.2 33 Sep-00 21.5 21.4 25.0 32.1 28

RE3 Aug-96 25.0 17.9 25.0 32.1 28 Sep-90 10.0 16.7 26.7 46.6 30

RE3-RE4 Apr-74 50.0 11.1 16.7 22.2 18 May-78 37.5 37.5 15.6 9.4 32 Jun-75 29.8 40.5 10.8 18.9 37 Jul-75 25.6 30.2 23.3 20.9 43 Aug-81 18.0 35.9 28.2 17.9 39 Aug-97 3.6 25.0 42.8 28.6 28 Sep-01 3.5 13.8 24.1 58.6 29 Oct-84 20.1 23.3 33.3 23.3 30 Dec-78 38.9 27.8 11.1 22.2 18

RE3-RE5 Aug-98 15.2 18.2 33.3 33.3 33 Sep-98 7.0 10.3 37.9 44.8 29 Oct-85 20.0 20.0 50.0 10.0 30 Oct-92 21.4 25.0 14.3 39.3 28 Nov-87 14.3 7.1 35.7 42.9 14

RE3-RE6 May-76 60.0 16.6 6.8 16.6 30 Aug-83 11.2 25.0 44.4 19.4 36 Aug-91 4.7 20.9 37.2 37.2 43

RE4 Jun-90 9.5 0.0 14.3 76.2 21

RE4-RE5 Jul-90 3.2 9.7 12.9 74.2 31 Aug-99 8.2 10.8 32.4 48.6 37 Oct-83 15.8 26.3 31.6 26.3 19 Oct-93 13.3 20.1 33.3 33.3 30 Sep-81 4.7 9.3 44.2 41.8 43

106 Appendix 2.2 (Continued).

Ecotone Function (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE4-RE6 Aug-86 2.5 20.0 35.0 42.5 40 Sep-87 11.1 25.0 36.1 27.8 36 Sep-90 10.0 33.3 20.0 36.7 30

RE4-RE7 Oct-00 4.5 45.5 22.7 27.3 22

RE4-RE8 Nov-93 57.2 28.6 7.1 7.1 14

RE5 Jun-91 0.0 4.8 26.2 69.0 42 Jun-92 0.0 8.3 25.0 66.7 24 Jul-93 0.0 13.9 22.2 63.9 36 Sep-93 7.0 3.4 31.0 58.6 29 Sep-96 15.0 10.0 30.0 45.0 20 Dec-78 50.0 16.7 5.6 27.7 18

RE5-RE6 Jun-90 4.6 13.6 27.3 54.5 22 Aug-97 8.8 11.8 32.4 47.0 34 Oct-98 6.5 16.1 41.9 35.5 31

RE5-RE7 Dec-75 50.0 17.0 16.5 16.5 12

RE6 Jun-91 0.0 7.1 16.7 76.2 42 Jun-92 0.0 16.7 20.8 62.5 24 Jul-90 3.2 3.2 13.0 80.6 31 Aug-88 3.2 15.6 28.1 53.1 32 Dec-78 5.6 16.7 11.0 66.7 18

RE6-7 Oct-83 5.3 15.8 31.5 47.4 19 Sep-93 6.9 17.2 34.5 41.4 29 Sep-95 3.8 18.5 44.4 33.3 27

RE6-RE9 Aug-90 7.7 15.4 50.0 26.9 26 Oct-94 18.6 29.6 11.1 40.7 27

RE6-10 Nov-87 28.6 14.3 7.1 50.0 14 Oct-92 11.1 11.1 29.6 48.2 27

107 Appendix 2.2 (Continued).

Ecotone Function (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE7 May-76 3.3 3.3 30.1 63.3 30

RE7-RE8 Jun-90 33.3 14.3 9.5 42.9 21

RE7-RE9 Dec-78 16.7 16.7 22.2 44.4 18

RE7-RE11 Aug-87 21.2 27.3 18.2 33.3 33 Aug-97 29.4 14.7 20.6 35.3 34 Sep-87 19.4 22.3 19.4 38.9 36 Sep-90 16.6 26.7 26.7 30.0 30 Oct-88 27.6 6.9 31.0 34.5 29

RE8 Aug-89 23.3 16.7 20.0 40.0 30 Oct-00 4.5 9.1 18.2 68.2 22

RE8-9 Apr-76 0.0 0.0 45.2 54.8 31

RE8-RE10 Nov-85 13.1 30.4 17.4 39.1 23 Dec-75 0.0 0.0 25.0 75.0 12

RE8-11 Apr-78 4.0 8.0 40.0 48.0 25 Sep-95 48.1 11.2 7.4 33.3 27

RE9 Nov-77 2.9 2.9 26.6 67.6 34

RE9-RE10 Jun-90 13.6 9.1 9.1 68.2 22 Oct-00 22.7 22.7 22.7 31.9 22

RE9-RE11 Nov-93 0.0 7.2 57.1 35.7 14

RE10-RE11 Apr-76 0.0 6.5 32.3 61.2 31 Nov-77 0.0 8.8 32.4 58.8 34

108 Appendix 2.2 (Continued).

Ecotone Function (%) for the total selected species No. of selected1 location Aggregator Mediator Soft barrier Hard barrier species/gradient

RE11 Jun-90 19.0 19.0 4.8 57.2 21 Oct-92 11.1 7.4 22.2 59.3 27 Oct-00 31.9 22.7 4.5 40.9 22 Nov-87 21.4 14.3 0.0 64.3 14

1A fish species with less than 5 individuals collected along the Hudson River estuary gradient in each of the 116 time periods was not included in the analysis.

Appendix 2.3. Multiple functions (neutral, partly inhibit, and completely inhibit) of the ecotone peripheries (from all detected frequencies) on the Hudson fish species in each time period during 1974–2001.

Ecotone periphery Function (%) for the total selected species No. of selected1 location between Neutral Partly inhibit Completely inhibit species/gradient

RE2-RE3 Jun-79 13.9 44.4 41.7 36 Jun-95 28.6 4.8 66.6 21 Aug-75 11.1 31.1 57.8 45 Nov-75 10.0 23.3 66.7 30

109 Appendix 2.3 (Continued).

Ecotone periphery Function (%) for the total selected species No. of selected1 location between Neutral Partly inhibit Completely inhibit species/gradient

RE3-RE4 Jun-88 17.1 17.1 65.8 35 Jul-96 18.0 25.6 56.4 39 Sep-76 18.0 41.0 41.0 36 Sep-80 11.1 33.3 55.6 36 Sep-94 8.6 17.1 74.3 35 Oct-78 25.7 34.3 40.0 35 Oct-80 15.4 26.9 57.7 26 Oct-86 25.8 29.0 45.2 31 Oct-95 13.3 30.0 56.7 30

RE4-RE5 Sep-92 17.3 31.0 51.7 29

RE5-RE6 Jul-97 8.6 40.0 51.4 35 Jul-99 21.2 12.1 66.7 33 Aug-94 16.2 29.0 54.8 31 Aug-95 12.9 38.7 48.4 31

RE7-RE8 Aug-00 14.3 40.0 45.7 35 Oct-06 12.5 54.2 33.3 24

RE8-RE9 Apr-79 16.7 25.0 58.3 24

RE10-RE11 Jun-74 35.3 35.3 29.4 34 Jun-78 22.5 35.0 42.5 40 Jul-75 30.2 30.2 39.6 43 Aug-79 29.5 29.5 41.0 44 Sep-75 22.9 14.6 62.5 48 Sep-80 30.6 19.4 50.0 36 Oct-74 14.7 38.2 47.1 34 Nov-80 11.8 5.9 82.3 17 Nov-86 24.0 24.0 52.0 25 RE11-RE12 May-74 13.1 21.7 65.2 23 May-77 9.1 30.3 60.6 33 May-78 21.9 18.7 59.4 32 May-80 16.7 13.9 69.4 36 Jun-79 13.9 16.7 69.4 36 Jun-80 20.0 32.5 47.5 40 Jun-95 9.6 33.3 57.1 21 Jul-76 19.5 31.7 48.8 41 Jul-77 14.0 25.5 60.5 43 Jul-78 28.6 26.2 45.2 42

110 Appendix 2.3 (Continued).

Ecotone periphery Function (%) for the total selected species No. of selected1 location between Neutral Partly inhibit Completely inhibit species/gradient

RE11-RE12 (Continued) Jul-97 8.6 20.0 71.4 35 Jul-00 20.6 32.4 47.0 34 Aug-75 20.0 28.9 51.1 45 Aug-76 29.2 20.8 50.0 48 Aug-78 28.6 22.4 49.0 49 Aug-80 26.8 19.5 53.7 41 Sep-76 25.6 18.0 56.4 39 Sep-78 28.9 15.8 55.3 38 Sep-79 22.9 22.9 54.2 35 Sep-82 20.5 18.0 61.5 39 Sep-94 20.0 20.0 60.0 35 Sep-99 18.8 43.8 37.4 32 Sep-01 16.7 13.3 70.0 30 Oct-76 10.3 25.6 64.1 39 Oct-78 20.0 11.4 68.6 35 Oct-79 22.3 19.4 58.3 36 Oct-80 15.4 26.9 57.7 26 Oct-93 10.3 13.8 75.9 29 Nov-78 16.7 10.0 73.3 30 Nov-79 13.4 23.3 63.3 30

1A fish species with less than 5 individuals collected along the Hudson River estuary gradient in each of the 116 time periods was not included in the analysis.

111

REFERENCES

Able, K. W. 2005. A re-examination of fish estuarine dependence: evidence for connectivity between estuarine and ocean habitats. Estuarine Coastal and Shelf Science 64:5–17.

Able, K. W., and M. P. Fahay. 1998. The first year in the life of estuarine fishes in the middle Atlantic Bight. Rutgers University Press, New Brunswic, New Jersey.

Abood, K. A., and S. G. Metzger. 1996. Comparing impacts to shallow-water habitats through time and space. Estuaries 19:220–228.

Armand, A. D. 1992. Sharp and gradual mountain timberlines as a results of species interaction. Pages 360–378 in Hansen and di Castri, editors. Landscape boundaries: Consequences for biotic diversity and ecological flows. Springer- Verlag, New York.

ASA Analysis & Communication, I. 2004. 2001 year class report for the Hudson River estuary monitoring program. Dynegy Roseton L.L.C., Entergy Nuclear Indian Point 2 L.L.C., Entergy Nuclear Indian Point 3 L.L.C., and Mirant Bowine L.L.C., New York.

Attrill, M. J., and S. D. Rundle. 2002. Ecotone or ecocline: ecological boundaries in estuaries. Estuarine, Coastal and Shelf Science 55:929–936.

Baker, J., K. French, and R. J. Whelan. 2002. The edge effect and ecotonal species: Bird communities across a natural edge in southeastern Australia. Ecology 83:3048–3059.

Barnthouse, L. W., R. J. Klauda, D. S. Vaughan, and R. L. Kendall, editors. 1988. Science, law, and Hudson River power plants: a case study in environmental impact assessment. American Fisheries Society, Monograph 4, Bethesda, Maryland.

Becker, G. C. 1983. Fishes of Wisconsin. University of Wisconsin Press, Madison.

112

Bell, R. E., R. D. Flood, S. Carbotte, W. B. F. Ryan, F. O. Nitsche, S. Chilrud, R. Arko, V. Ferrini, A. Slagle, C. Bertinato, and M. Turrin. 2006. Hudson River estuary program benthic mapping project. Hudson River Estuary Program, New York State Department of Environmental Conservation, New York.

Bovee, K. D. 1982. A guide to stream habitat analysis using the instream flow incremental methodology: Instream Flow Information Paper 5. FWS/OBS–82/26, U.S. Fish and Wildlife Service.

Bovee, K. D. 1986. Development and evaluation of habitat suitability criteria for use in the Instream Flow Incremental Methodology. Instream Flow Information Paper 21: 86(7), U.S. Fish and Wildlife Service, Washington DC.

Bowersox, M. A., and D. G. Brown. 2001. Measuring the abruptness of patchy ecotones. Plant Ecology 156:89–103.

Boyle, R. H. 1979. The Hudson River, Expanded edition. W.W. Norton and Company, New York.

Cadenasso, M. L., S. T. A. Pickett, K. C. Weathers, and C. G. Jones. 2003. A framework for a theory of ecological boundaries. BioScience 53:750–758.

Cadenasso, M. L., M. M. Traynor, and S. T. A. Pickett. 1997. Functional location of forest edges: gradients of multiple physical factors. Canadian Journal of Forest Research 27:774–782.

Clements, F. E. 1897. Peculiar zonal formations of the Great Plains. American Naturalist 31:968–970.

Clements, F. E. 1905. Research methods in ecology. University Pubishing Company, Nebraska, USA.

Coch, N. K., and H. J. Bokuniewicz. 1986. Oceanographic and geologic framework of the Hudson system. Northeastern Geology 8:96–108.

Csillag, F., B. Boots, M.-J. Fortin, K. Lowell, and F. Potvin. 2001. Multiscale characterization of boundaries and landscape ecological patterns. Geomatica 55:291–307.

113

Daniels, R. A., K. E. Limburg, R. E. Schmidt, D. L. Strayer, and R. C. Chambers. 2005. Changes in fish assemblages in the tidal Hudson River, New York. American Fisheries Society Symposium 45:471–503.

Degraaf, D. A., and L. H. Bain. 1986. Habitat use by and preferences of Juvenile Atlantic salmon in two Newfoundland Rivers. Transactions of the American Fisheries Society 115:671–681.

Delcourt, P. A., and H. R. Delcourt. 1992. Ecotone dynamics in space and time. Pages 19–54 in Hansen and di Castri, editors. Landscape Boundaries: Consequences for biotic diversity and ecological flows. Springer-Verlag, New York.

Dew, I. M. 2001. Theoretical model of a new fishery under a simple quota management system. Ecological Modelling 143:59–70. di Castri, F., and A. J. Hansen. 1992. The environment and development crises as determinants of landscape dynamics. Pages 3–18 in Hansen and di Castri, editors. Landscape Boundaries: Consequences for biotic diversity and ecological flows. Springer-Verlag, New York. di Castri, F., A. J. Hansen, and M. M. Holland, editors. 1988. A new look at ecotones: emerging international projects on landscape boundaries (special issue 17). Biology International. IUBS, Paris, France.

Fagan, W. F., R. S. Cantrell, and C. Cosner. 1999. How habitat edges change species interactions. American Naturalist 153:165–182.

Fagan, W. F., M. J. Fortin, and C. Soykan. 2003. Integrating edge detection and dynamic modeling in quantitative analyses of ecological boundaries. BioScience 53:730–738.

Fielding, A. H. 2002. What are appropriate characteristics of an accuracy measure? Pages 271–280 in Scott, Heglund, Morrison, and al., editors. Predicting species occurrences: issues of accuracy and scale. Island Press, Washington.

Fischer, H. B., E. J. List, R. C. Y. Koh, J. Imberger, and N. H. Brooks. 1979. Mixing in inland and coastal waters. Academic Press, San Diego, California.

114

Ford, A. 1999. Modeling the Environment: an introduction to system dynamics modeling of environmental systems. Island Press, Washington, D.C.

Forman, R. T. T., and P. N. Moore. 1992. Theoretical foundations for understanding boundaries in landscape mosaics. Pages 236–258 in Hansen and di Castri, editors. Landscape Boundaries: Consequences for biotic diversity and ecological flows. Springer-Verlag, New York.

Fortin, M.-J. 1994. Edge detection algorithms for two dimensional ecological data. Ecology 75:956–965.

Fortin, M.-J., R. J. Olson, S. Ferson, L. Iverson, C. Hunsaker, G. Edwards, D. Levine, K. Butera, and V. Klemas. 2000. Issues related to the detection of boundaries. Landscape Ecology 15:453–466.

Fruin, J. J. 1971. Pedestrian planning and design. Metropolitan Association of Urban Designers and Environmental Planners, New York.

Gaff, H., D. L. DeAngelis, L. J. Gross, and R. Salinas. 2000. A dynamic landscape model for fish in the Everglades and its application to restoration. Ecological Modelling 127:33–52.

Gauch, H. G., Jr. 1982. Multivariate analysis in community ecology, first edition. Cambridge University Press, New York.

Gertseva, V. V., V. I. Gertsev, and N. Y. Ponomarev. 2003. Tropho-ethological polymorphism of fish as a strategy of habitat development: a simulation model. Ecological Modelling 167:159–164.

Gertseva, V. V., J. E. Schindler, V. I. Gertsev, N. Y. Ponomarev, and W. R. English. 2004. A simulation model of the dynamics of aquatic macroinvertebrate communities. Ecological Modelling 176:173–186.

Gido, K. B., C. W. Hargrave, W. J. Matthews, and G. D. Schnell. 2002. Structure of littoral-zone fish communities in relation to habitat, physical, and chemical gradients in a southern reservoir. Environmental Biology of Fishes 63:253–263.

115

Gilmore, R. G., Jr. 1995. Environmental and bigeographic factors influencing ichthyofaunal diverity in the Indian River Lagoon. Bulletin of Marine Science 57:153–170.

Gottlieb, S. J. 1998. Nutrient removal by age-0 Atlantic menhaden (Brevoortia tyrranus) in Chesapeake Bay and implications for seasonal management of the fishery. Ecological Modelling 112:111–130.

Gray, E. V. S., and J. R. Stauffer, Jr. 1999. Comparative microhabitat use of ecologically similar benthic fishes. Environ. Biol. Fishes. 56:443–453.

Guay, J. C., D. Boisclair, D. Rioux, M. Leclerc, M. Lapointe, and P. Legendre. 2000. Development and validation of numerical habitat models for juveniles of Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Science 57:2065–2075.

Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147–186.

Haddad, N. M. 1999. Corridor use predicted from behaviors at habitat boundaries. American Naturalist 153:215–227.

Haefner, J. W. 1996. Modeling biological systems: principles and applications. Chapman & Hall, New York.

Hansen, A. J., and F. di Castri, editors. 1992. Landscape Boundaries: Consequences for biotic diversity and ecological flows. Springer-Verlag, New York.

Hansen, A. J., F. di Castri, and R. J. Naiman. 1988. Ecotones: what and why? Pages 9- 45 in di Castri, Hansen, and Holland, editors. A new look at ecotones: emerging international project on landscape boundaries (special issue 17). Biological International, IUBS, Paris.

Harding, E. K. 2002. Modelling the influence of seasonal inter-habitat movements by an ecotone rodent. Biological Conservation 104:227–237.

Harrington, R. W., Jr. 1947. Observations on the breeding habits of the yellow perch, Perca flavescens (Mitchill). Copeia 1947:199–200.

116

Hatzenbeler, G. R., M. A. Bozek, M. J. Jennings, and E. E. Emmons. 2000. Seasonal variation in fish assemblage structure and habitat structure in the nearshore littoral zone of Wisconsin Lakes. North American Journal of Fisheries Management 20:360–368.

Helfman, G. S. 1979. Twilight activities of yellow perch, Perca flavescens. J. Fish. Res. Board Can. 36:173–179.

High Performance Systems, Inc. 2002. STELLA(R) Research, version 7.0.3 for Windows. Hanover, NH.

Hill, M. O. 1979. DECORANA: a FORTRAN program for detrended correspondence analysis and reciprocal averaging. Cornell University, Ithaca, New York.

Hoeinghaus, D. J., K. O. Winemiller, and D. C. Taphorn. 2004. Compositional change in fish assemblages along the Andean piedmont-Llanos floodplain gradient of the Rio Portuguesa, Venezuela. Neotropical Ichthyology 2:85–92.

Holland, M. M. 1988. SCOPE-MAB technical consultations of landscape boundaries: report of a SCOPE/MAB workshop on ecotone, January 5–7, 1987. Pages 47–106 in di Castri, Hansen, and Holland, editors. A new look at ecotones: emerging international project on landscape boundaries (special issue 17). Biological International. IUBS, Paris.

Houston, J. 2001. Status of the bluntnose minnow, Pimephales notatus, in Canada. The Canadian Field-Naturalist 115:145–151.

Hurst, T. P., K. A. McKown, and D. O. Conover. 2004. Interannual and long-term variation in the Nearshore fish community of the mesohaline Hudson River estuary. Estuaries 27:659–669.

Jaureguizara, A. J., R. Mennic, C. Bremeca, H. Mianzan, and C. Lastaa. 2003. Fish assemblage and environmental patterns in the Rio de la Plata estuary. Estuarine, Coastal and Shelf Science 56:921–933.

Johnson, A. R., J. A. Wiens, B. T. Milne, and T. O. Crist. 1992. Animal movements and population dynamics in heterogeneous landscapes. Landscape Ecology 7:63– 75.

117

Johnston, C. A., J. Pastor, and G. Pinay. 1992. Quantitative methods for studying landscape boundaries. Pages 107–125 in Hansen and di Castri, editors. Landscape Boundaries: Consequences for biotic diversity and ecological flows. Springer- Verlag, New York.

Johnston, V. R., and E. P. Odum. 1956. Breeding bird populations in relation to plant succession on the piedmont of Georgia. Ecology 37:50–62.

Kilpatrick, H. J., A. M. LaBonte, J. S. Barclay, and G. Warner. 2004. Assessing strategies to improve bowhunting as an urban deer management tool. Wildlife Society Bulletin 32:1177–1184.

Kirchhofer, A. 1995. Morphologicla variability in the ecotone–an important factor for the conservation of fish species richness in Swiss rivers. Hydrobiologia 303:103– 110.

Kolasa, J., and M. Zalewski. 1995. Notes on ecotone attributes and functions. Hydrobiologia 303:1–7.

Kot, M. 2001. Elements of mathematical ecology. Cambridge University Press, Cambridge, UK.

Krivtsov, V., Corliss, J., Bellinger, E., Sigee, D., 2000. Indirect regulation rule for consecutive stages of ecological succession. Ecol. Model. 133, 73–81.

Kupschus, S. 2003. Development and evaluation of statistical habitat suitability models: and example based on juvenile spotted seatrout Cynoscion nebulosus. Marine Ecology Progress Series 265:197–212.

Kupschus, S., and D. Tremain. 2001. Association between fish assemblages and environmental factors in nearshore habitats of a subtropical estuary. Journal of Fish Biology 58:1383–1403.

Kupschus, S., and D. Tremain. 2001. Associations between fish assemblages and environmental fiactors in nearshore habitats of a subtropical estuary. Journal of Fish Biology 58:1383–1403.

118

Lachavanne, J.-B. 1997. Why study biodiversity in land-inland water ecotones? Pages 1–33 in Lachvanne and Juge, editors. Biodiversity in land-inland water ecotones. UNESCO, Paris.

Lay, D. 1938. How valuable are woodland clearings to birdlife? Wilson Bulletin 50:254–256.

Leopold, A. 1933. Game management. Charles Scribner's Sons, New York.

Lewis, M. A. 1997. Variability, patchiness, and jump dispersal in the spread of an invading population. Pages 46–69 in Tilman and Kareiva, editors. Spatial ecology: the role of space in population dynamics and interspecific interactions. Princeton University Press, Princeton, New Jersey.

Lidicker, W. Z. J. 1999. Responses of mammals to habitat edges: an overview. Landscape Ecology 14:333–343.

Limburg, K. E., M. A. Moran, and W. H. McDowell. 1986. The Hudson River ecosystem. Springer-Verlag, New York.

Lucas, M. C., and E. Baras. 2001. Migration of freshwater fishes. Blackwell Science, London.

Lumia, R. 1998. Flood of January 19–20, 1996 in New York State. Water-Resources Investigations Report, U.S. Geological Survey, Albany, New York.

Mancroni, W. J., M. W. Daley, and W. Dey. 1992. Recent Trends and Patterns in selected water quality parameters in the mid-Hudson River estuary. Pages 59–81 in Smith, editor. Estuarine Research in the 1980s. State University of New York Press, Albany.

Marín, V. H. 1997. A simple-biology, stage-structured population model of the spring dynamics of Calanus chilensis at Mejillones del Sur Bay, Chile. Ecological Modelling 105:65–82.

Marshall, S., and M. Elliott. 1998. Environmental Influences on the Fish Assemblage of the Humber Estuary. Estuarine, Coastal and Shelf Science 46:175–184.

119

Martino, E. J., and K. W. Able. 2003. Fish assemblages across the marine to low salinity transition zone of a temperate estuary. Estuarine Coastal and Shelf Science 56:969–987.

Matthysen, E. 2002. Boundary effects on dispersal between habitat patches by forest birds (Parus major, P. caeruleus). Landscape Ecology 17:509–515.

McCausland, W. D., E. Mente, G. J. Pierce, and I. Theodossiou. 2006. A simulation model of sustainability of coastal communities: aquaculture, fishing, environment and labour markets. Ecological Modelling 193:271–294.

Miller, B. A., and S. Sadro. 2003. Residence time and seasonal movements of juvenile coho salmon in the ecotone and lower estuary of Winchester Creek, South Slough, Oregon. Transactions of the American Fisheries Society 132:546–559.

Morrison, K. A., N. Therien, and B. Coupal. 1985. Simulating fish redistribution in the LG-2 reservoir after flooding. Ecological Modelling 28:97–111.

Morrison, M. L., and L. S. Hall. 2002. Standard terminology: toward a common languge to advance ecological understanding and application. Pages 43–52 in Scott, Heglund, Morrison, Haufler, Raphael, Wall, and Samson, editors. Predicting species occurrences. Island Press, Washington.

Mundahl, N. D., and C. G. Ingersoll. 1983. Early autumn movements and densities of johny (Etheostoma nigrum)and fantail (E. Flabellare) darters in a southwestern Ohio stream. Ohio J. Sci. 83:103–108.

Naiman, R. J., and H. Décamps, editors. 1990. The ecology and management of aquatic-terrestrial ecotones. The Parthenon Publishing Group, Paris.

National Oceanic and Atmospheric Administration. 1989. Estuaries of the United States: vital statistics of a natural resource base. Strategic Environmental Assessments Division, Rockville, MD.

Neilson, R. P., G. A. King, R. L. DeVelice, and J. M. Lenihan. 1992. Regional and local vegetation patterns: The responses of vegetation diversity to subcontinental air masses. Pages 129–149 in Hansen and Castri, editors. Landscape boundaries: Consequences for biotic diversity and ecological flows. Springer-Verlag, New York, New York.

120

Nieder, W. C., E. Barnaba, S. E. G. Findlay, S. Hoskins, N. Holochuck, and E. A. Blair. 2004. Distribution and abundance of submerged aquatic vegetation and Trapa natans in the Hudson River estuary. Journal of Coastal Research 45:150- 161.

Nisbet, R. M., and W. S. C. Gurney. 1982. Modelling fluctuating populations. Jonh Wiley & Son, New York.

Nitsche, F. O., R. Bell, S. M. Carbotte, W. B. F. Ryan, and R. Flood. 2004. Process- related classification of acoustic data from the Hudson River Estuary. Marine Geology 209:131-145.

Noordhuis, R., D. T. van der Molen, and M. S. van den Berg. 2002. Response of herbivorous watr-birds to the return of Chara in Lake Veluwemeer, the Netherlands. Aquatic Botany 72:349–367.

O'Connor, R. J. 2002. The conceptual basis of species distribution modeling: time for a paradigm shift? Pages 25–33 in Scott, Heglund, Morrison, Haufler, Raphael, Wall, and Samson, editors. Predicting species occurrences. Island Press, Washington.

Odum, E. P. 1971. Fundamentals of ecology, third edition. W.B. Saunders Company, Philadelphia.

Okubo, A., and S. A. Levin. 2001. Diffusion and ecological problems: Modern perspectives, second edition. Springer, New York.

Olden, J. D., D. A. Jackson, and P. R. Peres-Neto. 2002. Predictive models of fish species distributions: A note on proper validation and chance predictions. Transactions of the American Fisheries Society 131:329–336.

Ottaviani, D., G. J. Lasinio, and L. Boitani. 2004. Two statistical methods to validate habitat suitability models using presence-only data. Ecological Modelling 179:417–443.

Pace, M. L., S. E. G. Findlay, and D. Fishcher. 1998. Effects of an invasive bivalve on the zooplanktion community of the Hudson River. Freshwater Biology 39:103– 116.

121

Pajak, P., and R. J. Neves. 1987. Habitat suitability and fish production: a model evaluation for rock bass in two Virginia streams. Transactions of the American Fisheries Society 116:839–850.

Pan, Y., and D. J. Raynal. 1995. Decomposing tree annual volume increments and constructing a system dynamic model of tree growth. Ecological Modelling 82:299-312.

Paukert, C. P., and D. W. Willis. 2002. Seasonal and diel habitat selection by bluegills in a shallow natural lake. Transactions of the American Fisheries Society 131:1131–1139.

Petts, G. E. 1990. The role of ecotones in aquatic landscape managment. Pages 227– 261 in Naiman and Decamps, editors. The ecology and managment of aquatic- terrestrial ecotones. UNESCO and The Parthenon Publishing Group, Paris, France.

Phillips, P. J., D. A. eckhardt, D. A. Freehafer, G. R. Wall, and H. H. Ingleston. 2002. Regional patterns of pesticide concentration in surface waters of New York in 1997. Journal of the American Water Resources Association 38:731–745.

Pinay, G., H. Décamps, E. Chauvet, and E. Fustec. 1990. Functions of ecotone in fluvial systems. Pages 141–169 in Naiman and Decamps, editors. The ecology and managment of aquatic-terrestrial ecotones. UNESCO and The Parthenon Publishing Group, Paris, France.

Przybylski, M., P. Biro, M. Zalewski, I. Tatrai, and P. Frankiewicz. 1991. The structure of fish communities in streams of the northern part of the catchment area of Lake Balaton (Hungary). ACTA Hydrobiologica 33:135–148.

Pusey, B., A. H. Arthington, and M. G. Read. 1993. Spatial and temporal variation in fish assemblage structure in the Mary River, south-eastern Queensland: the influence of habitat structure. Environmental Biology of Fishes 37:355–380.

Rakocinski, C. F., D. M. Baltz, and J. W. Fleeger. 1992. Correspondence between environmental gradients and the community structure of marsh-edge fishes in a Louisiana estuary. Marine Ecology-Progress Series 80:135–148.

Reyes, E., F. H. Sklar, and J. W. Day. 1994. A regional organism exchange model for simulating fish migration. Ecological Modelling 74:255–276.

122

Rhoads, D. C., and S. Cande. 1971. Sediment profile camera for in situ study of organism-sediment relations. Limnology and Oceanography 16:110–114.

Rhoads, D. C., and J. D. Germano. 1982. Characterization of organism-sediment relations using sediment profile imaging: an efficient method of remote ecological monitoring of the seafloor (RemotsTM System). Marine EcologyProgress Series 8:115–128.

Ries, L., and D. M. Debinski. 2001. Butterfly responses to habitat edges in the highly fragmented prairies of central Iowa. Journal of Animal Ecology 70:840–852.

Ries, L., R. J. Fletcher, Jr., J. Battin, and T. D. Sisk. 2004. Ecological responses to habitat edges:mechanisms, models, and variability explained. Annual Reviews of Ecology, Evolution, and Systematics 35:491–522.

Risser, P. G. 1990. The ecological importance of land-water ecotones. Pages 7–21 in Naiman and Decamps, editors. The ecology and managment of aquatic-terrestrial ecotones. Parthenon Publishing Group, Paris, France.

Ristich, S. S., M. Crandall, and J. Fortier. 1977. Benthic and epibenthic macroinvertebrates of the Hudson River: Distribution, natural history and community structure. Estuarine, Coastal and Shelf Science 5:255–266.

Ross, J. A., S. F. Matter, and J. Roland. 2005. Edge avoidance and movement of the butterfly Parnassius smintheus in matrix and non-matrix habitat. Landscape Ecology 20:127–135.

Rundle, S. D., M. J. Attrill, and A. Arshad. 1998. Seasonality in macroinvertebrate community composition across a neglected ecological boundary, the freshwater- estuarine transition zone. Aquatic Ecology 32:211–216.

Schiemer, F., M. Zalewski, and J. E. Thorpe. 1995. The importance of aquatic- terrestrial ecotones for freshwater fish. Hydrobiologia 303.

Schmidt, R. E., and S. Cooper. 1996. A catalog of barriers to upstream movement of migratory fishes in Hudson River tributaries. Final report to Hudson River Foundation, New York.

123

Schmidt, R. E., R. J. Klauda, and J. M. Bartels. 1988. Distributions and movements of the early life stages of three species of alosa in the Hudson River, with comments on mechanisms to reduce interspecific completition. Pages 193–215 in Smith, editor. Fisheries research in the Hudson River. State University of New York Press, Albany, New York.

Sisk, T. D., and N. M. Haddad. 2002. Incorporating the effects of habitat edges into landscape models: Effective area models for cross-boundary management. Pages 208–240 in Liu and Taylor, editors. Integrating landscape ecology into natural resource management. Cambride University Press, Cambridge, UK.

Sisk, T. D., and C. R. Margules. 1993. Habitat edges and restoration: methods for quantifying edge effects and predicting the results of restoration efforts. Pages 45– 55 in Saunders, Hobbes, and Ehrlich, editors. Nature conservertion 3: Reconstruction of fragmented ecosystems. Surrey Beatty and Sons Pty Limited Chipping Norton, New South Wales, Australia, Chipping Norton, New South Wales.

Skalski, G. T., and J. E. Gilliam. 2000. Modeling diffusive spread in a heterogeneous population: a movement study with stream fish. Ecology 81:1685–1700.

Smith, C. L. 1985. The inland fishes of New York State. New York State Department of Environmental Conservation, Albany, New York.

Sparrevohn, C. R., A. Nielsen, and J. G. Stottrup. 2002. Diffusion of fish from a single release point. Canadian Journal of Fisheries and Aquatic Science 59:844–853.

Stamps, J. A., M. Buechner, and V. V. Krishnan. 1987. The effects of edge permeability and habitat geometry on emigration from patches of habitat. American Naturalist 129:533–552.

Stanne, S. P., R. G. Panetta, and B. E. Forist. 1996. The Hudson. Rutgers University Press, New Brunswick, New Jersey.

Sterman, J. D. 2000. Business dynamics: Systems thinking and modeling for a complex world. McGraw-Hill, Singapore.

Strayer, D. L., N. F. Caraco, J. J. Cole, S. Findlay, and M. L. Pace. 1999. Transformation of freshwater ecosystems by bivalves. BioScience 49:19–27.

124

Strayer, D. L., K. A. Hattala, and A. W. Kahnle. 2004. Effects of an invasive bivalve (Dreissena polymorpha) on fish in the Hudson River estuary. Canada Journal of Fisheries and Aquatic Science 61:924–941.

Strayer, D. L., J. Powell, P. Ambrose, L. C. Smith, M. L. Pace, and D. T. Fischer. 1996. Arrival, spread, and early dynamics of a zebra mussel (Dreissena polymorpha) population in the Hudson River estuary. Canada Journal of Fisheries and Aquatic Science 53:1143–1149.

Strayer, D. L., M. E. Power, W. F. Fagan, S. T. A. Pickett, and J. Belnap. 2003. A classification of ecological boundaries. BioScience 53:723–729. ter Braak, C. J. F., and P. Smilauer. 1998. Canoco reference manual and user's guide to Canoco for windows: Software for canonical community ordination (version 4). Microcomputer Power, Ithaca, New York.

Texas Instruments, I. 1976. 1975 Water qualtity standard operating procedures. Consolidated Edison Company of New York, New York.

Texas Instruments, Inc. 1977. 1974 year-class report for the multiplant impact study of the Hudson River estuary. Consolidated Edison Company of New York, New York.

Texas Instruments Inc. 1981. 1979 year class report for the muliplant inmpact study of the Hudson River estuary. Consolidated Edison Company of New York, New York.

Tilman, D., and P. Kareiva, editors. 1997. Spatial ecology: the role of space in population dynamics and interspecific interaction. Princeton University Press, Princeton, New Jersey.

Trautman, M. B. 1957. The fishes of Ohio. Ohio State University Press, Columbus, Ohio.

Turchin, P. 1998. Quantitative analysis of movement: measuring and modeling populaiton redistribution in animals and plants. Sinauer Associates, Inc., Sunderland, Massachusetts.

125

Vernberg, W. B., and F. J. Vernberg. 1976. Physicological adaptations of estuarine animals. Oceanus 19:48–54.

Wagner, C. M. 1999. Expression of the estuarine species minimum in littoral fish assemblages of the lower Chesapeake Bay tributaries. Estuaries 22:304–312.

Wall, G. R., and P. J. Phillips. 1998. Pesticides in the Hudson River basin, 1994–96. Northeastern Geology and Environmental Sciences 20: 299–307.

Ward, J. V., and J. A. Stanford. 1983. The serial discontinuity concept of lotic ecosystems. Pages 347–356 in Fontaine and Bartell, editors. Dynamics of lotic ecosystems. Ann Arbor Science Publishers, Ann Arbor, MI.

Watson, J. E. M., R. J. Whittaker, and T. P. Dawson. 2004. Habitat structure and proximity to forest edge affect the abundance and distribution of forest-dependent birds in tropical coastal forests of southeastern Madagascar. Biological Conservation 120:311–327.

Weinstein, M. P., S. L. Weiss, and M. F. Walters. 1980. Multiple determinants of community structure in shallow marsh habitats, Cape Fear River estuary, North Carolina. Marine Biology 58:227–243.

Wells, A. W., and J. R. Young. 1992. Long-term variability and predictability of Hudson River physical and chemical characteristics. Pages 29–58 in Smith, editor. Estuarine research in the 1980s. State University of New York Press, Albany, New York.

Wells, L. 1968. Seasonal depth distribution of fish in southeastern Lake Michigan. Fisheries Bulletin 67, U.S Fish and Wildlife Service.

Whitfield, A. K. 1999. Ichthyofaunal assemblages in estuaries: A South African case study. Reviews in Fish Biology and Fisheries 9:151–186.

Whittaker, R. H. 1975. Communities and Ecosystems, second edition. Macmillan Publishing, New York.

Wiens, J. A. 1992. Ecological flows across landscape boundaries: a conceptual overview. Pages 217–235 in Hansen and di Carstri, editors. Landscape boundaries:

126

Consequences for biotic diversity and ecological flows. Springer-Verlag, New York, New York, USA.

Wiens, J. A. 2002. Riverine landscapes: taking landscape ecology into the water. Freshwater Biology 47:501–515.

Wiens, J. A., C.S.Crawford, and J.R.Gosz. 1985. Boundary dynamics: a conceptual framework for studying landscape ecosystems. Oikos 45:421–427.

Wiens, J. A., N. C. Stenseth, B. Van Horne, and R. A. Ims. 1993. Ecological mechanisms and landscape ecology. Oikos 66:369–380.

Willis, T. V., and J. J. Magnuson. 2000. Patterns in fish species composition across the interface between streams and lakes. Canadian Journal of Fisheries and Aquatic Science 57:1042–1052.

Winemiller, K. O., and M. A. Leslie. 1992. Fish assemblages across a complex, tropical freshwater/marine ecotone. Environmental Biology of Fishes 34:29–50.

Zalewski, M. 1997. Fish diversity and ecotonal habitat. Pages 183-203 in Lachavanne and Juge, editors. Biodiversity in land-inland water ecotones. UNESCO and The Parthenon Publishing Group, Paris.

Zalewski, M., J. E. Thorpe, and R. J. Naiman. 2001. Fish and riparian ecotones-a hypothesis. Ecohydrology and Hydrobiology 1:11–14.

127