Urban Ecosystems

, Volume 13, Issue 4, pp 517–533

Emerging dragonfly diversity at small Rhode Island (U.S.A.) wetlands along an urbanization gradient

Authors

    • Department of Plant Sciences and Entomology, Woodward HallUniversity of Rhode Island
  • Howard S. Ginsberg
    • Department of Plant Sciences and Entomology, Woodward HallUniversity of Rhode Island
    • USGS Patuxent Wildlife Research Center, Coastal Field Station, Woodward-PLSUniversity of Rhode Island
Article

DOI: 10.1007/s11252-010-0133-8

Cite this article as:
Aliberti Lubertazzi, M.A. & Ginsberg, H.S. Urban Ecosyst (2010) 13: 517. doi:10.1007/s11252-010-0133-8

Abstract

Natal habitat use by dragonflies was assessed on an urban to rural land-use gradient at a set of 21 wetlands, during two emergence seasons (2004, 2005). The wetlands were characterized for urbanization level by using the first factor from a principal components analysis combining chloride concentration in the wetland and percent forest in the surrounding buffer zone. Measurements of species diversity and its components (species richness and evenness) were analyzed and compared along the urbanization gradient, as were distributions of individual species. Dragonfly diversity, species richness, and evenness did not change along the urbanization gradient, so urban wetlands served as natal habitat for numerous dragonfly species. However, several individual species displayed strong relationships to the degree of urbanization, and most were more commonly found at urban sites and at sites with fish. In contrast, relatively rare species were generally found at the rural end of the gradient. These results suggest that urban wetlands can play important roles as dragonfly habitat and in dragonfly conservation efforts, but that conservation of rural wetlands is also important for some dragonfly species.

Keywords

OdonataExuviaeLenticWetland

Introduction

Small wetlands are common in the urban landscape, and they often differ in abiotic characteristics (such as substrate and basin shape) from wetlands in less urban areas (Ehrenfeld 2000). Urban habitats are typically highly disturbed relative to natural areas, and four hypotheses about invertebrate diversity in disturbed habitats have been proposed. One holds that disturbed habitats (such as urban sites) are merely ecological sinks (Pulliam 1988), with low diversity and only widespread, generalist species present. On the other hand, Bock et al. (2008) suggest that an “oasis effect” can occur when disturbed areas provide exceptional resources, and hence foster higher diversity. A third hypothesis suggests that diversity is greatest in areas with moderate levels of disturbance (the “Intermediate Disturbance Hypothesis” of Connell 1978), which suggests that diversity should be greatest at habitats in the suburban or urban-edge zones of the urbanization gradient. A fourth possibility is that diversity does not change along the urbanization gradient.

Recent studies of terrestrial invertebrate distributions along urban gradients have had varying results with regard to diversity patterns. Blair and Launer (1997) found butterfly diversity to be highest at intermediate sites along the gradient, whereas Magura et al. (2004) found carabid beetle diversity to be lowest at intermediate sites; however, there is no clear definition of “intermediate” urbanization levels across studies. Winfree et al. (2007) found highest bee diversity in areas with fragmented forest, whereas Gibbs and Stanton (2001) found silphid beetle diversity to be higher in more intact forest areas. Overall, most studies have found terrestrial invertebrate diversity to be lower in the more human-modified areas of the gradient (e.g., spiders [Shochat et al. 2004], bees [McIntyre and Hostetler 2001], ants [Thompson and McLachlan 2007]). Aside from study of urban mosquito habitats (e.g., Fischer et al. 2000), most urban studies of aquatic invertebrates have concentrated on lotic systems (e.g., Paul and Meyer 2001; Kaushal et al. 2005). Furthermore, studies of adult odonates (dragon- and damselflies, Insecta: Odonata) in urban environments in the northeastern U.S. (e.g., Creveling 2003) likely include species that have flown there from other wetlands, and did not utilize the study wetlands as natal habitat (Buskirk and Sherman 1984; Pulliam 1988). Hence, the conservation value of lentic urban wetlands for odonate faunas remains unclear.

In this study we sampled exuviae—the last nymphal exoskeleton, which is shed upon emergence from the aquatic habitat—to assess dragonfly natal habitat. Standardized collection of dragonfly exuviae can be a direct and low-impact method for monitoring the emerging dragonfly communities at small wetlands, because exuviae indicate that the individuals sampled developed in and emerged from the wetland of interest (Morin 1984a; Corbet 1993; Foote and Rice Hornung 2005; Aliberti Lubertazzi and Ginsberg 2009). Furthermore, exuvial surveys have low impact on the local population because live individuals are not removed or disturbed.

It is difficult for some odonate species to successfully complete their nymphal stages in waters with fish (e.g., Morin 1984b; Pierce et al. 1985; McPeek 1990a), resulting in various traits to avoid fish predation (e.g., Johnson and Crowley 1980; Pierce et al. 1985; Blois-Heulin et al. 1990; McPeek 1990b; Johansson et al. 2006). Rubbo and Kiesecker (2005) found both fish presence and wetland permanence to be associated with urbanization. Hence, we would expect some definitive patterns in species composition along the urban gradient when fish vs. fishless sites are compared.

In this paper we describe the emerging dragonfly faunas at small wetlands along the urban to rural landscape gradient in Rhode Island, U.S.A. We surveyed dragonfly exuviae at twenty-one wetlands over two field seasons to determine whether small wetlands surrounded by urbanized landscapes provide natal habitat to fewer species of dragonflies than wetlands in more natural areas, and to assess the effects of fish presence on this pattern.

Materials and methods

Twenty-one small, lentic wetlands in Rhode Island were surveyed for dragonfly exuviae over two field seasons, 2004 and 2005. They were selected for size (approximately 1 ha), accessibility, long hydroperiod (mostly permanent), and to represent a variety of positions along the urbanization gradient. Figure 1 shows wetland locations on a map of human population density in Rhode Island.
https://static-content.springer.com/image/art%3A10.1007%2Fs11252-010-0133-8/MediaObjects/11252_2010_133_Fig1_HTML.gif
Fig. 1

Map of Rhode Island with human population density (by quantile) in Rhode Island (source: RIGIS 2002) and location of wetland sites. Inset map shows location of the state of Rhode Island on the east coast of the United States. Note: The large, dark area (denoting high human population density) in the northeastern region of the state indicates the location of metropolitan Providence, the capitol city of Rhode Island

We conducted timed searches for dragonfly exuviae on 6 visits per site in 2004, and 5 visits per site in 2005. Each site was sampled at roughly 3-week intervals from mid-May through September each year. Search routes around the wetlands were selected to include all potential habitat types at each wetland, and the same route was searched on each visit (sampling dates and aerial photos with sampling routes are reported in Appendix I of Aliberti Lubertazzi 2009). Exuviae were identified to species in the laboratory, using taxonomic keys (Walker 1958; Walker and Corbet 1975; Soltesz 1996; Bright and O’Brien 1999; Needham et al. 2000) and validated (by rearing and/or consensus identification) specimens, and counted. Exuviae of some species could not be reliably distinguished, so we used the following combination categories: Tramea spp. (Tramea lacerata and Tramea carolina), Sympetrum vicinum/semicinctum and Libellula vibrans/axilena. For each species, the number collected per hour was summed for all sampling dates of each season, by site, for a season-wide score (see Conrad et al. 1999). Fish presence-absence was determined by extensive dip-net sampling of all microhabitats throughout the sampling seasons at each wetland.

Chloride concentration can be a useful measure of watershed urbanization in the northeastern U.S. (A. Gold, pers. comm.; Kaushal et al. 2005; URI Watershed Watch 2006) because most towns in Rhode Island treat roads with salt during winter ice conditions and background chloride levels are very low (L. Green, pers. comm.). The amount of forest surrounding a wetland is an inverse measure of urbanization (Miller et al. 1997), as it indicates the lack of anthropogenic land-clearing and development (Booth et al. 2002). Although pH was also a strong environmental variable, and usually correlated with chloride concentration, it can be related to forest features, and is not as direct a measure of urbanization as is chloride. We measured pH (spring, summer and fall samples in 2004, spring and fall samples in 2005) and chloride concentration (spring and fall both years) at each wetland. pH was measured with an Accumet model AR20 research pH and conductivity meter (Fisher Scientific, Pittsburgh, PA) within 24 h of water collection. In fall 2004, water samples were measured for chloride concentration by titration with the Argentometric Method (Clesceri et al. 1998). Water samples from the 2005 field season were analyzed for chloride in the URI Watershed Watch Analytical Laboratory using an Astoria-Pacific International model 303a segmented continuous flow autoanalyzer (Astoria-Pacific International, Clackamus, OR), using method SM 4500-Cl-E in Standard Methods for the Examination of Water and Wastewater (Clesceri et al. 1998). Spring chloride values were used in the analyses (except for one site in 2005, NBGROUND, where the spring value was compromised so the fall value was used; see Aliberti Lubertazzi 2009, Appendix II, for details).

All study wetlands were located and digitized from the RIGIS03/04 Digital Orthophotos of Rhode Island 2003–2004, using ArcMap software (Environmental Systems Research Institute, Redlands, CA). The resulting shapefiles were used for wetland buffer analysis by constructing 100 m buffers around each of the study wetlands. The buffer polygons were used to clip the 1995 RI state land-use datalayer (RIGIS 2005), which had been recoded to 6 land-use categories by merging Anderson land-use classes (Anderson et al. 1976; see Marchand and Litvaitis 2004; Price et al. 2004): high-medium density residential, low-medium density residential, commercial/industrial, open/other, forest, and wetland. The study wetland area was erased, leaving a donut-shaped buffer polygon. We then calculated the percent area in each buffer polygon that consisted of each land-use category (Marchand and Litvaitis 2004).

Environmental patterns were assessed using canonical correspondence analysis (CCA) of species (seasonal score = number of exuviae collected per hour per season for each species, at each site) and environmental data (pH, chloride concentration, and the 6 land-use values) for each year using CANOCO software (terBraak and Smilauer 1997–1999). A strong environmental pattern emerged that was related to “urbanization”, characterized by the percent forest and chloride concentration vectors (Fig. 2). Therefore, an urbanization variable was formulated using the PCA Factor 1 from a principal components analysis (STATISTICA 6.0, StatSoft, Inc., Tulsa, OK 1984–2002) of percent forest cover and chloride concentration at each site, for each year. Level of urbanization was used as the independent variable for ANOVA and regression analyses, with species richness as the dependent variable. To assess the independent effects of the variables used to characterize “urbanization”, standard least squares analyses were calculated for species richness vs. percent forest and chloride concentration for both years. JMP (JMP® 7.0, SAS Institute, Inc., Cary, NC) and STATISTICA software were used for these analyses. Additionally, a Student’s t-test was used to evaluate the urbanization variable at sites with vs. without fish each year. Species richness was normally distributed in our samples in both years (Kolmogorov-Smirnov tests: 2004, Dmax = 0.102, p > 0.1; 2005, Dmax = 0.128, p > 0.1; BIOMSTAT 3.3, Exeter Software, Setauket, NY).
https://static-content.springer.com/image/art%3A10.1007%2Fs11252-010-0133-8/MediaObjects/11252_2010_133_Fig2_HTML.gif
Fig. 2

Canonical correspondence analysis plots of sites (“X”-marks; based on species composition) with environmental variable vectors in 2004 (a) and 2005 (b). FOREST = percent forest in 100 m buffer; PH = water pH; CL = water chloride concentration; OTHER WETLAND = percent cover of (other) wetland in 100 m buffer; COMMER/INDUS = percent cover of commercial/industrial land-use in 100 m buffer; HI MED RES = percent cover of hi-medium density residential land-use in 100 m buffer; for detailed information on land-use categories, see Aliberti Lubertazzi (2009)

Individual species found at >2 sites/category (all sites, fishless and fish sites) were also evaluated for patterns along the urbanization variable each year. Their seasonal scores (log (x + 1) transformed to eliminate trends in residuals) served as dependent variables in univariate regression analyses, with urbanization as the independent variable.

We analyzed spatial autocorrelation in species richness at these sites by performing the Moran’s I analysis (Fortin and Legendre 1989) (GS+: GeoStatistics for the Environmental Sciences, Version 7; Gamma Design Software, Plainwell, MI).

The Shannon-Wiener Diversity Index H’ (Shannon and Weaver 1949), and species richness, were calculated for each site, each year. We also calculated Simpson’s Diversity Index (Simpson 1949) and three measurements of evenness for 2005, and compared their patterns along the gradient for that year. Evenness was assessed using three different measurements: the Berger-Parker dominance index (proportion of dominant species in total catch; Southwood and Henderson 2000), H′/log(species richness) (Southwood and Henderson 2000), and the slope of a log-log plot of the dominance-diversity curve at each site (the yearly abundance value for each species was log (x + 1) transformed and plotted against its log-transformed rank to linearize the slope relationship). With this last measurement, evenness is inversely related to the magnitude of the negative slope. Each of these 6 measurements was then plotted along the urbanization variable for 2005 to compare and assess the patterns of species diversity along the urbanization gradient. Evenness was analyzed separately from species richness so that these two factors, which both contribute to the value of standard diversity indices, could be clearly interpreted.

Results

Environmental variables related to urbanization and dragonfly diversity measures for both years are given in Tables 1 and 2. pH and chloride values for the 21 wetlands surveyed were highly correlated between years (2004: r = 0.637, p < 0.002; 2005: r = 0.948, p < 0.0001). Chloride concentration differed among sites (F = 15.144, df = 20, 20, p < 0.0001) and between years (F = 4.960, df = 1, 20, p = 0.038).
Table 1

Fish presence, chloride concentration, %forest in 100 m wetland buffer, and urbanization variables (Factor 1 of PCA: chloride and % forest) for both 2004 and 2005

SITE

FISH?

Chloride

(mg/l)

%Forest

Urbanization

Variable

Y/N

2004

2005

(100 m buffer)

2004

2005

AMTRAK

N

1

2

0.60

0.76

0.73

BLACKSTONE

Y

121

159

0.74

−0.22

−0.09

BRISTOLSK

Y

148

154

0.57

−0.91

−0.44

CAMPUS

Y

115

159

0.03

−1.71

−1.65

CAROLBIG

N

2

6

0.85

1.30

1.26

DEXTER

Y

29

42

0.28

−0.25

−0.26

EIGHTROD

Y

−2

4

0.42

0.42

0.33

GLOBE

Y

17

11

0.35

0.03

0.12

INDUSTRIAL

Y

23

33

0.50

0.31

0.30

KITTBIG

N

2

3

0.89

1.40

1.37

NBGROUND

N

30

4

0

−0.87

−0.59

PAINTBALL

Y

25

49

0.46

0.20

0.09

PHELPS

Y

125

288

0.24

−1.37

−2.13

RUMFORD

Y

231

334

0

−3.04

−2.99

SAILADUMP

N

0

8

0.80

1.23

1.14

SANDY

Y

41

53

0.77

0.70

0.74

SKLT

N

0

3

1.00

1.66

1.60

SLATERFRIEND

Y

1

18

0

−0.54

−0.69

SPECTACLE

Y

106

112

0.22

−1.19

−0.88

STRATHMORE

N

3

2

0.65

0.86

0.85

WAJONES

Y

4

6

0.83

1.23

1.21

Table 2

Dragonfly diversity measurements at all study sites. See text for definitions

SITE

(H′)

Species richness

Simpson’s index

1-BP index

H′/ logSR

log-log slope

2004

2005

2004

2005

2005

2005

2005

2005

AMTRAK

1.11

1.54

10

12

3.41

0.54

−1.43

−2.01

BLACKSTONE

1.58

1.51

11

11

4.05

0.68

−1.45

−2.31

BRISTOLSK

1.61

2.23

15

17

7.64

0.79

−1.81

−1.65

CAMPUS

1.15

1.39

8

9

3.22

0.57

−1.45

−2.34

CAROLBIG

1.04

2.23

16

15

7.48

0.77

−1.89

−1.42

DEXTER

1.88

2.10

18

20

6.11

0.74

−1.62

−1.78

EIGHTROD

0.66

0.47

8

5

1.27

0.11

−0.67

−2.82

GLOBE

1.50

1.02

8

7

2.06

0.34

−1.21

−2.08

INDUSTRIAL

1.24

0.69

6

2

1.98

0.44

NAa

NAa

KITTBIG

1.23

1.37

10

17

2.36

0.38

−1.12

−2.03

NBGROUND

0.43

0.90

4

8

1.63

0.23

−1.00

−2.03

PAINTBALL

1.28

0.82

14

13

1.57

0.21

−0.74

−2.75

PHELPS

1.83

1.48

16

15

3.30

0.57

−1.26

−2.13

RUMFORD

1.19

0.73

12

15

1.40

0.16

−0.62

−2.43

SAILADUMP

0.51

1.38

6

9

2.75

0.44

−1.45

−2.07

SANDY

1.62

1.31

13

13

2.20

0.34

−1.18

−2.08

SKLT

0.80

0.92

4

9

1.64

0.23

−0.96

−1.96

SLATERFRIEND

1.38

1.70

13

12

3.26

0.48

−1.58

−1.37

SPECTACLE

1.63

1.71

10

10

4.49

0.67

−1.71

−1.83

STRATHMORE

1.53

1.50

8

7

3.53

0.57

−1.78

−1.42

WAJONES

1.56

1.15

18

15

1.94

0.30

−0.98

−1.81

aValues for this site were not used in analyses because only 2 species were recorded there in 2005

In 2004, Factor 1 in the PCA of percent forest and chloride concentration (which was used as the urbanization variable) explained 72.5% of total variability (eigenvalue = 1.45). In 2005 the urbanization variable explained 72.2% of the total variation (eigenvalue = 1.44). Although chloride differed between years, the urbanization values for the sites were highly correlated between years (r = 0.982, p < 0.0001). Both percent forest and chloride concentration had equally strong Factor I coordinates both years. It is important to note that sites with negative values of this index are more urban.

Greater than ten thousand exuviae were collected at all sites in 2004, and almost nine thousand were collected in 2005. Overall, species richness at individual sites ranged from 4 to 18 in 2004, and 2 to 20 in 2005. Species richness differed among sites (F = 8.961, df = 20, 20, p < 0.0001), but not between years (F = 1.052, df = 1, 20, p = 0.317).

Measures of species diversity and species richness are given in Table 2. The 2005 data were used to compare various indices of species evenness, as well as to assess relationships of urbanization with evenness (Table 2). The Shannon-Wiener Diversity Index (H′) was not related to urbanization either year, with or without fish (2004: overall R2 = 0.052, p = 0.322; fish R2 = 0.0002, p = 0.962; no fish R2 = 0.154, p = 0.385; 2005: overall R2 = 0.002, p = 0.853; fish R2 = 0.010, p = 0.740; no fish R2 = 0.114, p = 0.458). Species richness was not significantly related to the degree of urbanization in either year, even when fish and fishless sites were analyzed separately (Table 3, Fig. 3). Species richness was correlated with all of the diversity and evenness measurements in 2005 (p < 0.10) except for the log-log dominance diversity slopes (p = 0.277) and H′/log(species richness) (p = 0.468). None of the 2005 diversity, richness, or evenness measures showed a relationship with the urbanization variable or its component variables (Table 4). One site (INDUSTRIAL) was excluded for analyses with the log-log slopes and the H′/log(species richness), as only two species were recorded there in 2005. Moran’s I statistic indicated no regular pattern of autocorrelation for species richness among the sites. All of the Moran’s I correlations were low and nonsignificant.
Table 3

Univariate regression analyses of species richness with urbanization variable, chloride concentration, and forest cover in 2004 and 2005

 

Urbanization variable

Chloride concentration

% Forest in buffer

Coefficient

P-value

Coefficient

P-value

Coefficient

P-value

All sites

2004

−0.422

0.610

0.014

0.361

0.114

0.971

2005

−0.478

0.576

0.013

0.194

1.034

0.746

Fishless

2004

1.723

0.453

−0.140

0.419

4.251

0.464

2005

2.174

0.352

0.124

0.871

4.756

0.351

Fish

2004

0.304

0.764

0.002

0.916

2.767

0.492

2005

−0.946

0.442

0.015

0.243

−0.001

1.000

https://static-content.springer.com/image/art%3A10.1007%2Fs11252-010-0133-8/MediaObjects/11252_2010_133_Fig3_HTML.gif
Fig. 3

Species richness along the urbanization variable for 2004 (a) and 2005 (b). Black squares are wetlands with fish populations, open squares are fishless wetlands. The negative side of the urbanization scale denotes the “more urban” side of the variable

Table 4

Relationships between dragonfly diversity and evenness measures, and urbanization (2005 data). See text for details

Measure

# Sites

Urbanization variable

Chloride concentration

Percent forest in buffer

Coefficient

P-value

Coefficient

P-value

Coefficient

P-value

Diversity

Shannon-Wiener (H′)

21

−0.018

0.853

−0.001

0.689

−0.239

0.555

Simpson’s

21

0.006

0.987

0.003

0.544

−0.953

0.533

Species richness

21

−0.478

0.576

0.019

0.110

3.526

0.306

Evenness

1 - BP Index

21

−0.008

0.847

0.000

0.442

0.097

0.568

H′/logSR

20

−0.036

0.629

0.000

0.774

−0.066

0.835

log-log slope

20

0.082

0.279

−0.001

0.323

0.042

0.894

Dragonfly species richness tended to differ at sites with fish compared to sites without fish in 2004 (t = 2.03, p = 0.067), with apparently greater species richness in wetlands with fish, but not in 2005 (t = 0.37, p = 0.716). Additionally, the urbanization variable was significantly different at sites with vs. without fish populations both years (2004: t = -3.12, p = 0.007; 2005: t = -3.31, p = 0.004); in effect, the sites with fish were more urban than the sites without fish. We calculated all analyses of diversity and distribution patterns along the gradient for all, fish and fishless sites separately to account for this.

In contrast to overall diversity, abundances of some common species (i.e., those found at >2 sites/category/year) were correlated with the degree of urbanization (Tables 5 and 6). For example, at all sites Libellula luctuosa and Tramea spp. were positively correlated with urbanization (negative correlation with the urbanization variable; Table 5). At sites with fish, Pachydiplax longipennis and Epitheca cynosura tended to be more abundant at urbanized sites (p < 0.10; Table 6). Gomphus exilis tended to be more common at less-urbanized sites with fish in 2004 only (p < 0.10), and Sympetrum janae was generally more abundant at less-urban sites overall (Table 6). Rare species (those found only at 1 or 2 sites) were found most often at sites on the less-urban side of the gradient (Fig. 4).
Table 5

Univariate analyses of selected common species (seasonal scores, log [x + 1] transformed) vs. urbanization variable at all sites. Only species that were present at >2 sites per year with significant relationship at least one year were included. Note: P-values < 0.05 in bold

Species

Year

# of Sites

Coefficient

P-value

Family Aeshnidae

Anax junius

2004

16

0.071

0.603

2005

15

0.220

0.065

Family Corduliidae

Epitheca cynosura

2004

13

−0.249

0.061

2005

13

−0.340

0.014

Family Libellulidae

Erythemis simplicicollis

2004

14

−0.240

0.048

2005

14

−0.159

0.242

Libellula luctuosa

2004

6

−0.169

0.011

2005

3

−0.209

0.013

Sympetrum janae

2004

16

0.377

<0.001

2005

18

0.166

0.104

Sympetrum vicinum/semicinctum

2004

20

−0.018

0.903

2005

17

−0.253

0.099

Tramea spp.

2004

8

−0.323

0.009

2005

10

−0.206

0.018

Table 6

Univariate analyses of selected common species (seasonal scores, log [x + 1] transformed) vs. urbanization variable at sites with (n = 14) or without (n = 7) fish. Only species that were present at >2 sites per category per year, with significant relationship at least one year, were included

Species

Year

No fish

Fish

# of Sites

Coefficient

P-value

# of Sites

Coefficient

P-value

Family Gomphidae

Gomphus exilis

2004

 

ND

NA

5

0.256

0.067

2005

 

ND

NA

4

0.260

0.120

Family Corduliidae

Epitheca cynosura

2004

3

0.131

0.737

10

−0.303

0.084

2005

2

 

NA

11

−0.391

0.032

Family Libellulidae

Libellula luctuosa

2004

1

 

NA

5

−0.234

0.017

2005

0

 

NA

3

−0.266

0.048

Pachydiplax longipennis

2004

4

0.698

0.368

12

−0.437

0.069

2005

7

0.181

0.681

12

−0.420

0.093

Sympetrum janae

2004

7

0.393

0.157

9

0.296

0.045

2005

6

0.325

0.402

12

0.130

0.313

Tramea spp.

2004

2

 

NA

6

−0.464

0.009

2005

4

−0.574

0.023

6

−0.265

0.018

Bold values indicate <0.05 significance level

ND not detected

https://static-content.springer.com/image/art%3A10.1007%2Fs11252-010-0133-8/MediaObjects/11252_2010_133_Fig4_HTML.gif
Fig. 4

Abundance of rare species (those found at ≤2 sites) along urbanization gradient in 2004 (a) and 2005 (b). Abundance is the sum of the seasonal scores of all rare species at sites with rare species present

Discussion

Dragonfly diversity did not change along the urbanization gradient in this study. The components of diversity, species richness and evenness, also did not change along the urbanization gradient. In contrast, some individual species displayed strong trends with urbanization. Most of the abundant species with distinct patterns of distribution along the urbanization gradient favored the urban, and not the rural end of the gradient (Tables 5 and 6). This may indicate a tendency to utilize or select specific habitat types, which could be considered habitat specialization, as opposed to broad, generalized habitat use. This result contrasts with those of several studies that found the proportion of specialized invertebrate species to be lower in urban areas than in natural, intact areas (Gibbs and Stanton 2001; McIntyre and Hostetler 2001; Koh and Sodhi 2004; Clark et al. 2007; Thompson and McLachlan 2007). Our finding that less-common species were generally found at more rural sites is compatible with this pattern. However, the lack of a relationship between overall species diversity and urbanization suggests that urban sites have high value as dragonfly habitat.

Like the present study, McIntyre et al. (2001) found continuous richness along an urbanization gradient, and similar to this study, the communities consisted of very different species assemblages along the gradient. However, that study looked at a very broad group of invertebrates—all ground arthropods—so it is difficult to compare to narrower, taxon-based patterns. Other studies of terrestrial invertebrate diversity support the intermediate disturbance hypothesis (e.g., Blair and Launer 1997). In contrast, we found no region along the urbanization axis with distinctly higher species richness (Table 3, Fig. 3), even when sites with or without fish were considered separately.

Several studies have found that one native species can account for roughly half of the individuals collected along urbanization gradients. Examples include bees (McIntyre and Hostetler 2001), stream dragonflies (Hawking and New 2003), and carabid beetles (Magura et al. 2004). Similarly, Samways and Steytler (1996) found low (adult) dragonfly diversity but highest abundance at a city site comprised of just a few, super-abundant species. Presence of a super-abundant species might indicate poor environmental conditions (e.g., D’Amico et al. 2004), and steep dominance-diversity curves are common features in urban invertebrate surveys (e.g., McIntyre and Hostetler 2001; McFrederick and LeBuhn 2006; Clark et al. 2007). At our sites, however, the steep, negative slopes on the log-log plots of dominance-diversity curves, which denote low diversity and low evenness, did not predominate in any zone of the urbanization gradient.

An “oasis effect” (Bock et al. 2008) does not appear to apply for dragonflies in this study, since the urban sites do not show higher diversity than rural sites. It is hard to distinguish whether urban ponds supply unique structural conditions that some species might favor, compared to rural sites. The surrounding land-use, however, could be a visual cue for some species. Additionally, the urban wetlands are not located within continuous land-use regimes in the very heterogeneous New England landscape, compared to exurban developments in the southwestern U.S. studied by Bock et al. (2008).

Measurement of urbanization

Researchers have taken many positions on how to measure or characterize urbanization (Theobald 2004). While some have assumed it to be an “understood,” qualitative state of the landscape (e.g., Blair 1999; Thompson and McLachlan 2007), others have based urbanization gradients on human population density (e.g., Rubbo and Kiesecker 2005), or impervious surface (e.g., Winfree et al. 2007). Others have categorized urban land-use by measuring conditions chosen for their direct relevance to the organisms under study (McIntyre 2000).

Booth et al. (2004) found that impervious surface alone cannot effectively predict the biological condition of lotic systems in urbanizing regions. Kaushal et al. (2005) suggest that chloride concentration should be actively monitored in lotic systems that drain urbanizing areas, because of its potential toxicity to the biota. Critical threshold concentrations (e.g., 250 mg/l; Environment Canada 2001), which can damage the aquatic fauna, are becoming commonplace in the northeastern U.S., and some wetlands in this study exceeded this threshold (Table 1). Therefore, we selected chloride concentration as a measure of urbanization because of the strong effects of road salt runoff on water quality in lentic wetlands that have limited “flushing” outflow (Environment Canada 2001). We selected forest cover because of the known effects of forest patterns on aquatic invertebrate recruitment to wetlands (e.g., Ormerod et al. 1990; Nilsson and Svensson 1995). Nevertheless, dragonfly species diversity did not change along the urbanization gradient, or along the gradients of chloride concentration or forest cover at our sites (Tables 3 and 4; Fig. 3).

Factors that might influence dragonfly distribution patterns

Many studies of invertebrates and urbanization are focused on plant-feeding or otherwise plant-dependent invertebrates, which respond directly to changes in increased impervious surface or herbaceous plant distributions (e.g., ornamentals, agriculture crops; e.g., Blair and Launer 1997; Denys and Holger 1998; McIntyre and Hostetler 2001; Shapiro 2002; Koh and Sodhi 2004). In fact, many insect species may be generalists with regard to food resources, but specialists in nesting habitat features, as are some birds. Perhaps species that are limited, either directly or indirectly, by natal habitats or nesting sites (such as dragonflies in our study, and taxa such as bees [McIntyre and Hostetler 2001; McFrederick and LeBuhn 2006; Matteson et al. 2008] and birds [Blair 2004]), might show different trends than species that respond more directly to the abundance of food, including larval food resources.

Indeed, some definitive patterns in species composition were found when fish vs. fishless sites were compared along the urbanization gradient. This is particularly evident for the individual common species with strong relationships to the gradient (Table 6). However, even with few “urban” sites in this study that do not have fish populations (see Fig. 3), species diversity did not change along the gradient when fish vs. fishless sites were analyzed separately (Table 3). Regardless, a larger sample size of fishless, permanent ponds in urban areas would help to further clarify this issue. Dragonfly development and diversity are not considered to be strongly affected by water quality, relative to other aquatic insect taxa (e.g., pH [Bell 1971, Hudson and Berrill 1986]), but their vertebrate predators might be (Eriksson et al. 1980; Bendell and McNicol 1987; Johansson and Brodin 2003). However, it appears that some fish populations tolerate high chloride levels in lentic wetlands, to some degree, as Rubbo and Kiesecker (2005) and this study (M.A.L., pers. obs.) found an association between fish presence and urbanization. Therefore, further study is needed to distinguish the effects of urbanization from the effects of fish presence.

Like birds, dragonflies utilize distinct habitats in a given landscape for reproduction (Moore 1991). Indeed, some native bird species are known to prefer, and flourish, in urban areas of the northeastern U.S., which may indicate an “oasis effect”. As avian landscape ecology has played a pivotal role in conservation decisions, natal habitat use by dragonflies might have an analogous role for small wetlands in the urbanizing northeast. Our results suggest that urban, suburban, rural, and “pristine” wetlands can all play important roles in conservation of biodiversity on our landscape (Moore 1991; Clark and Samways 1997). Urban habitats in general (Simberloff 1997; Faeth et al. 2005), and specifically wetlands (Ehrenfeld 2000), can be viewed as novel types of habitat, and distinct from the more “natural” habitats in rural landscapes. The more “urban” sites might offer a distinct habitat type (i.e., open, no surrounding woods), regardless of fish populations, that is very attractive to some dragonfly species.

The lack of a relationship between diversity and urbanization indicates that urban wetlands provide natal habitat to many dragonfly species. Therefore, urban wetlands support a diverse dragonfly fauna, and not just a few, tolerant species. In fact, for some species, ponds in urban parks and restoration sites serve as dragonfly natal habitat more commonly than do ponds in natural areas (Tables 5 and 6). Currently, rare lentic odonate species in southern New England are often linked to rare habitats that can be heavily influenced by surrounding land-use (e.g., Williamsonia lintneri ; Biber 2002). Nevertheless, some rare dragonfly species utilize urban wetlands elsewhere (Johnson et al. 2001), further suggesting the potential conservation value of small urban ponds (McKinney 2006). Our results suggest that there is value in conserving wetlands all along the urbanization gradient, because some species do well specifically in urban wetlands, and others (including some relatively rare species) appear to only use wetlands in natural areas. With changes in urbanization patterns—and increasing rates of change—responses of dragonfly species to the management of existing wetlands (and their upland surroundings) requires more attention.

Acknowledgments

We thank Ryan Abney and Megan Priede for assistance with field work, and the URI Coastal Fellows program for funding their work. Art Gold, Kelly Addy, and Linda Green provided guidance on water quality analyses. Peter August performed the spatial autocorrelation analysis. Mary-Jane James-Pirri assisted with formatting Fig. 1. We thank Peter August, Eric Biber, Sam Droege, Patrick Logan, Nancy McIntyre and two anonymous reviewers for constructive comments on early drafts of the manuscript.

Copyright information

© US Government 2010