Hydrobiologia

, Volume 621, Issue 1, pp 49–62

Aquatic Hemiptera community structure in stormwater retention ponds: a watershed land cover approach

Authors

  • Sarah J. Foltz
    • Zoology DepartmentUniversity of Wisconsin
    • Zoology DepartmentUniversity of Wisconsin
Primary research paper

DOI: 10.1007/s10750-008-9631-6

Cite this article as:
Foltz, S.J. & Dodson, S.I. Hydrobiologia (2009) 621: 49. doi:10.1007/s10750-008-9631-6

Abstract

Stormwater ponds are increasingly common aquatic habitats whose biotic communities are largely unexplored. As anthropogenic development continues to alter the landscape, watershed land use is gaining recognition for its potential to predict species compositions in aquatic systems. This study reports species composition of five aquatic hemipteran families (Notonectidae, Corixidae, Belostomatidae, Nepidae, Pleidae) in 28 permanent, artificial stormwater ponds in watersheds with different land covers and associated contaminant input. We hypothesized that land cover variables would be significant drivers of aquatic hemipteran community structure in ponds, and that ponds with a high percentage of agricultural and lawn cover in the watershed would be characterized by the absence of species intolerant of the chemical, physical, and ultimately biotic changes associated with these watersheds. Non-metric multi-dimensional scaling (NMS) was used to identify dominant gradients of species composition and environmental variables. Pond morphology variables, watershed lawn, watershed agriculture, and predatory fish abundance were each found to have statistically significant correlations with hemipteran community structure. The abundance of Notonecta undulata, the species responsible for creating the largest (ranked) distance in species structure among ponds, was positively correlated with shallow, fishless ponds and independent of land use variables. The abundances of four species of corixids were negatively correlated with watershed agriculture, and hemipteran richness was positively correlated with watershed lawn and negatively correlated with pond surface area. Heirarchical cluster analysis revealed non-random hemipteran species assemblages in which congeneric corixid species tended to co-occur, contradicting traditional niche theory. Since artificial stormwater ponds are chemically different from natural-pond habitat and rapidly increasing in number, knowledge of which insect species are capable of thriving in this environment and their relationship to land use in the watershed is of both environmental and evolutionary interest.

Keywords

NotonectidaeCorixidaeLawnAgricultureLand useAnthropogenic change

Introduction

Suburban development in the United States is growing at a conservative rate of 906 thousand hectares a year (USDA, 2000), approximately 23% of which is devoted to monoculture lawn (Robbins & Birkenholtz, 2003), with the remaining land largely covered by impervious surfaces. Since impervious surfaces release storm water more quickly than undeveloped land, retention ponds are designed to collect this run-off and prevent its direct discharge into streams and lakes (England, 2001). While virtually every water body in the United States contains anthropogenic chemicals (Gilliom et al., 2006), stormwater ponds situated immediately in the drainage system of high impact land cover (such as commercial agriculture and chemically treated lawn) are incredibly vulnerable to ground and surface water pollutants, including high nutrient loads and chemicals intended specifically to prevent, repel, or kill organisms (EPA, 2008). Widely used organophosphate insecticides, for example, bind to neurotransmitter enzymes and disrupt nervous impulses, leading to death or altered behavior in insects. Likewise, the herbicide 2,4-d selectively interferes with broad leaf plant growth, and is known to kill several aquatic insects (PAN Pesticides Database, 2007).

Since chemical input into water catchment basins varies with land usage in the watershed (Tong & Chen, 2002), watershed land use may serve as a potential predictor of aquatic communities (Hoffman & Dodson, 2005; Dodson, 2008). Studies have shown that biotic response variables are strongly impacted by both agricultural (Allan et al., 1997; Karaouzas & Gritzalis, 2006) and residential watershed land cover (Dodson, 2008; Lussier et al., 2008), both of which are often associated with relatively homogeneous assemblages of species capable of tolerating high levels of chemical and physical disturbance (Delong & Brusven, 1998; Leisnham et al., 2007). While watershed agriculture has a history of examination in landscape-scale studies of stream communities (Allan, 2004), the impact of agriculture on pond systems is less understood. Additionally, the increasing sales and application of insecticides, herbicides, fungicides, and fertilizers in lawn care maintenance (Robbins & Sharp, 2003) necessitates broadening our environmental focus on agricultural to include the high-input suburban lawn (Robbins et al., 2001; Robbins & Birkenholtz, 2003).

While studies have focused on the performance of retention ponds and their ability to protect streams and lakes (Hancock & Popkin, 2005), few studies have examined the community structure patterns occurring in the stormwater ponds themselves (Gingrich et al., 2006). We chose to focus on fully aquatic hemipterans because they are widespread insects capable of colonizing nearly all types of aquatic habitats, and are often one of the first successional stages in newly created water bodies, such as artificial ponds (Papacek, 2001). Both herbivorous (most Corixidae) and carnivorous (Notonectidae, Belostomatidae, Nepidae, Pleidae), these bugs have considerable influence on aquatic food webs, feeding on zooplankton, insects, and small vertebrates, and serving as prey for large insects and fish (e.g., Murdoch et al., 1984; Hickley et al., 1994; Blaustein, 1998; Gilbert & Burns, 1999; Hampton et al., 2000). Additionally, the economic importance of at least three of these families as biological control agents of disease-carrying mosquitos is well established (Nam et al., 2000; Saha et al., 2007, reviewed in Papacek, 2001). Notonectids, in particular, have been shown to prey preferentially on mosquito larvae, drive mosquitoes to local extinction (Murdoch et al., 1984), and even deter oviposition behavior in gravid mosquitoes, including Anopheles gambiae, the primary vector of human malaria (Blaustein et al., 2005; Munga et al., 2006).

The goal of this project was to study landscape-scale aquatic hemipteran community composition in a multispecies system whose organisms are confronted with numerous anthropogenic inputs in some of the habitats. Chemical contaminants have profound lethal and sublethal effects in aquatic systems, altering community structure and function (Rohr & Crumrine, 2005) via the physiological and behavioral processes of aquatic organisms within (Fleeger et al., 2003). In natural systems, however, limited pesticide application data and increasingly unmanageable numbers of manufactured chemical products, break-down products, and undisclosed inert ingredients often hinder the study of direct chemical-to-community relationships. Given the elusive identity of the contaminants whose combined effects are under scrutiny, we proposed an investigation in which community structure was studied in relation to high-impact land cover associated with elevated levels of pesticide use, such as commercial agriculture and chemically treated lawns. In addition to watershed land cover, six other environmental variables recognized in the literature as drivers of aquatic hemipteran assemblages were measured: fish abundance, macrophyte abundance, conductivity, pond age, water depth, and surface area (e.g., Briers & Warren, 2000; Svensson et al., 2000). Multivariate techniques were used to test the relationships between environmental variables and hemipteran community structure, and to examine the co-occurrence tendencies of the species in this system. We hypothesized that land cover variables would be significant drivers of aquatic hemipteran community structure in ponds, and that ponds with a high percentage of agricultural and lawn cover in the watershed would be characterized by the absence of certain species intolerant of the chemical, physical, and ultimately biotic changes associated with these watersheds. Since artificial stormwater ponds are chemically different from natural-pond habitat and rapidly increasing in number, knowledge of which insect species are capable of thriving in this environment and their relationship to land use in the watershed is of both environmental and evolutionary interest.

Methods

Twenty-eight permanent artificial stormwater retention ponds in Dane County, Wisconsin, were selected for this study (Dodson, 2008). Ponds differ in watershed land cover, age, surface area, depth, and vegetation structure, but are relatively similar in landscape position, conductivity, and substrate composition (Dodson, 2008). Each pond is the sole body of water in its watershed basin. The extent of each watershed was determined by walking each site and using USGS topographic maps and orthographs (30-cm resolution) from http://accessdane.co.dane.wi.us/. Watershed land cover was categorized as lawn, agriculture (corn/soy), impervious surfaces, and low-impact (meadow/woods), and the area of each category was quantified from high-resolution orthographic photos, available through Wisconsin Department of Natural Resources Webview (http://dnrmaps.wisconsin.gov/imf/imf.jsp?site=webview). Webview was also used to determine the surface area of the ponds. Water depth was measured or estimated at the deepest point in each pond in June. The sites varied in age from two to 22 years old, as determined from city records, land owner interviews, or by personal observation (Dodson, 2008). Water temperature, specific conductance, total dissolved solids (TDS), salinity, and dissolved oxygen percent were measured at the time of each insect sampling. Specific conductance was the only water chemistry variable included in the multivariate analyses.

Hemiptera sampling

Each pond was sampled for aquatic insects between June 10th and 20th, and again between September 8th and 16th, 2007. Relative abundance sampling was conducted by pulling a 4.35 m-by-1.15 m beach seine (1 mm-by-1.5 mm mesh) through a 13 m stretch of the pond, spanning both littoral and limnetic zones. “Target” sampling was also conducted by running a D-frame aquatic net through several sites in the pond chosen to maximize the differences in vegetation or substrate type. This method also helped ensure the collection of very small or fast bugs. Seine and “target” samples were preserved separately in 70% ethanol for identification. Adult hemipterans were identified to species, and the sex and age class of notonectids and corixids were determined. Species-level identification was possible for notonectid instars III–V, but all other hemipteran instars were identified to genus (notonectids, nepids) or family (corixids), unless species designation could be inferred by repeated single-species collections (e.g., only one species of Belostoma, Neoplea, and Buenoa were collected in this study regardless of season, therefore the nymphs were also assigned to the pertinent species).

Biotic community variables

Each pond was seined for predatory fish in June and September by pulling a beach seine (described above) through three 13-m stretches of each pond. Fish abundance was combined across seasons and categorized between one and five, based on natural breaking points in the distribution of abundance values. Actual abundance values ranged from 0 to 349. Vegetation surveys were also conducted at this time, and the percentage cover occupied by submerged macrophytes was estimated by walking, wading, and raking the pond perimeter. Dominant plant species were collected and identified. For comparability with Dodson (2008), emergent macrophytes, floating macrophytes, and algae were not included in percentage cover values.

Statistical analysis

Individuals collected in “target” samples were not included in the analyses unless they represented a species not collected by the seine sample of that pond in either season. This was most commonly the case for the very smallest bugs (Corisella tarsalis Fieber, Trichocorixa species, and Neoplea striola Fieber), although all of these species were capable of collection by the seine as well. Rare species (found in less than three ponds) were omitted from multivariate analyses (McCune & Grace, 2002), but included in the instability/seasonality analysis. Species instability (I) was calculated for each species as the number of times it changed its presence–absence status in the ponds between June and September:
$$ I = \sum\limits_{i = 0}^{28} {|p_{i}^{\text{Sept}} - p_{i}^{\text{June}} |{\text{ where }}p_{i}^{\text{season}} \in \{ 0,1\} = {\text{absence}},\,{\text{presence}}} $$
The seasonality (S) of these changes was calculated to show the net change in presence–absence between the sample dates:
$$ S = \sum\limits_{i = 0}^{28} {p_{i}^{\text{Sept}} - p_{i}^{\text{June}} {\text{ where }}p_{i}^{\text{season}} \in \{ 0,1\} = {\text{absence,}}\,{\text{presence}}} $$

Apart from these analyses, the June and September data were summed (for abundance data) or unionized (for presence–absence data).

Resemblance matrices were created for the ponds using species abundance data (Bray–Curtis distance measure) and normalized environmental data (Euclidean distance measure) using PC-ORD (McCune and Mefford 2006). The relationships between species composition and environmental variables were examined using the Mantel test, which tests the hypothesis that species and environmental similarity matrices are not related in multivariate structure using the Pearson correlation method and Mantel’s asymptotic approximation for test-statistic evaluation (McCune & Grace, 2002). The Best analysis (BIOENV algorithm, in PRIMER) was used to select the environmental variables ”best explaining” species community pattern by maximizing a rank correlation between their respective resemblance matrices. Community structure (species by ponds) was analyzed with Non-metric multi-dimensional scaling (NMS) from PCORD5 (McCune & Mefford 2006). NMS avoids the assumption of linear relationships among variables (species abundances) and is recommended for data with lots of zeros (McCune & Grace 2002). In our case, the data matrix was 61% zeros, which represented species that were absent from a particular pond. Correlation of species data with the NMS axes indicated the species most important in determining community structure. Correlation of environmental variables with the NMS axes indicated which variables were candidates for drivers of community structure. The environmental variables included watershed landcovers, pond age, depth, surface area, conductivity, fish abundance, and macrophyte abundance. For visualization, the joint biplot generated by the ordination was rotated to align the first axis with the species causing the largest (Bray–Curtis) differences between ponds (Notonecta undulata Say). Three ponds were extreme outliers in the ordination biplot, probably because they were the only ponds that contained only one species. Therefore, they were subsequently omitted from further multivariate analyses in order to more clearly see the relationship between species and environmental variables in the other ponds.

Heirarchical cluster analysis of species assemblages based on presence–absence data was conducted using the Bray–Curtis distance measure and the group average linkage method (McCune & Grace, 2002) in PRIMER. The SIMPROF permutation procedure was used to test for significance (5% level) of the resulting clades (Clarke & Gorley, 2006). The use of presence–absence data in this analysis served to uncover relationships involving species with low abundance but high frequencies of occurrence.

Results

Twenty-six of the 28 artificial retention ponds were found to support aquatic hemipterans. In total, 9989 specimens were identified to species, representing 26 species from five families (Notonectidae, Corixidae, Belostomatidae, Nepidae, Pleidae) (Table 1). Corixidae was the most species-rich family; 17 corixid species were identified, one of which, Corisella inscripta Uhler, had not been previously documented in Wisconsin (Hilsenhoff, 1995; Chordas, et al., 2002). Although some species were very common and abundant, others were relatively rare, occurring in low numbers and at few sites (Table 1). Notonecta undulata (Notonectidae) was by far the most common and abundant species, occurring in 23 ponds, and accounting for 39% of the total species abundance. Sex ratios of notonectid species and Hesperocorixa corixid species were generally not different from 1:1, while the sex ratios of Sigara and Trichocorixa corixid species were generally female-biased (Table 1). Percentage of adults generally increased from June to September (Table 1).
Table 1

The 25 aquatic hemipteran (Nepomorpha) species encountered in the 28 retention ponds over two sample periods

Family

Species

Occurrence

Abundance

% Fem

% Ad June

% Ad Sep

Belostomatidae

Belostoma flumineum

16

77

 

8.33

39.29

Corixidae

Corisella inscripta

5

792

55.7

  

Corixidae

Corisella tarsalis

5

28

74.2

  

Corixidae

Hesperocorixa atopodonta

4

6

50.0

  

Corixidae

Hesperocorixa laevigata

5

49

52.0

  

Corixidae

Hesperocorixa obliqua

17

1130

55.4

  

Corixidae

Hesperocorixa scabricula

8

96

51.0

  

Corixidae

Hesperocorixa vulgaris

10

62

63.9

  

Corixidae

Sigara alternata

21

668

83.2

  

Corixidae

Sigara decorata

5

179

68.7

  

Corixidae

Sigara defecta

7

15

86.7

  

Corixidae

Sigara grossolineata

2

3

0.0

  

Corixidae

Sigara solensis

1

1

0.0

  

Corixidae

Trichocorixa borealis

3

12

77.8

  

Corixidae

Trichocorixa calva

6

27

64.1

  

Corixidae

Trichocorixa kanza

6

55

59.7

  

Corixidae

Trichocorixa sexcincta

6

30

38.7

  

Nepidae

Ranatra sp. (fusca,kirkaldyi,nigra)

8

13

 

20.0

50.0

Notonectidae

Buenoa margaritaceae

13

2143

58.4

33.3

74.7

Notonectidae

Notonecta insulata

13

481

49.3

32.8

98.8

Notonectidae

Notonecta lunata

3

3

0.0

 

100.0

Notonectidae

Notonecta undulata

23

4078

50.8

47.3

76.5

Pleidae

Neoplea striola

7

40

 

84.8

0.0

Occurrence indicates the number of ponds that the species was found in, while abundance indicates the total number of individuals that were sampled from those ponds. The sex ratio (% Fem) indicates the percentage of adults that were female; the age ratio (% Ad) indicates the percentage of the species that were adults, separated by sample date. Blank cells are shown for sex or age ratios not acquired. Age ratios were not acquired for corixid species (blank cells) due to the inability to identify the instars of this taxon (Hilsenhoff 1995). The age ratio of Corixidae (16 species combined) was 52% adults in June and 79% in September. Due to their low numbers and similarity in habitat and resource use, Ranatra species were combined in our analyses; R. fusca was the most common of the three species. Hemipteran species with occurrence values less than three were not included in the NMS ordination or cluster analysis

Aquatic hemipteran richness ranged from one to 16 species in the occupied ponds (Table 2); seven was both the mean and the mode species richness value. Corixid richness ranged from one to 12 species (Table 2), with two species the mode and 4.6 species the mean. The 26 ponds also showed variation in morphology, biotic variables, and percent composition of each watershed land cover category (Table 2). Sixteen of the 28 ponds supported fish, with a total of five species (all predatory) collected. The fathead minnow (Pimephales promelas, a pioneer species highly tolerant of a wide range of environmental conditions (Scott & Crossman, 1973)) was most common, occurring in 11 ponds. Green sunfish (Lepomis cyanellus), bluegills (Lepomis macrochirus), and goldfish (Carassius auratus) occurred in six, four, and three ponds, respectively, while largemouth bass (Micropterus salmoides) were found in just one pond. Potamogeton species were the most common dominant submerged macrophytes in this system, although a few of the well-vegetated ponds had relatively homogeneous populations of Eurasian milfoil (Myriophyllum spicatum).
Table 2

Environmental variables and richness values (Hemiptera and Corixidae) for each pond

Pond

Latitude

Longitude

Area (m²)

Depth (m)

Cond (μhos/cm)

Age (years)

% Imperv

% Lawn

% Ag

% Mead/Wd

% Macro

Fish (Ab.C.)

Het SR

Cor SR

A1

43.0400

89.2454

3441

3.0

0.333

9

6.3

12.5

26.4

51.4

100

4

0

0

A2

43.0166

89.3010

809

2.0

0.264

3

37.2

43.1

0.0

16.8

30

4

16

11

A3

43.0709

89.2597

405

1.0

0.129

2

13.3

79.3

0.0

4.4

80

2

12

6

A4

43.0916

89.2663

4452

2.0

0.163

7

48.4

41.1

0.0

8.9

5

5

2

1

A5

43.0749

89.2798

3642

2.2

0.196

14

24.3

24.8

28.8

21.6

0

2

12

8

A7

43.0603

89.4992

8903

2.0

0.318

13

66.9

26.5

0.0

5.4

0

5

7

4

A8

43.0396

89.5241

4047

2.0

0.152

6

34.8

9.4

28.7

27.1

1

5

1

0

A9

43.1151

89.4921

809

1.0

0.434

3

72.2

20.6

0.0

6.2

40

1

7

5

A10

43.1087

89.4956

1619

1.5

0.299

12

54.2

15.6

0.0

26.6

30

1

7

2

A11

43.0764

89.2058

405

1.5

0.142

2

84.0

11.8

0.0

0.0

50

3

7

5

A12

43.0766

89.2070

4047

1.5

0.192

4

34.0

20.7

8.7

36.7

10

4

0

0

A13

43.1506

89.2884

5261

2.0

0.180

6

27.2

20.6

0.0

50.7

20

5

12

8

A14

43.0197

89.2996

1214

1.0

0.211

2

6.3

45.2

0.0

47.6

0

5

15

12

A15

43.0722

89.7648

93

1.0

0.146

2

4.0

4.7

3.4

87.2

2

1

7

2

A16

43.1437

89.2755

405

1.5

0.207

2

74.4

23.2

0.0

0.0

50

1

9

5

A17

43.2281

89.7222

186

0.7

0.339

2

0.0

0.0

0.0

100.0

80

1

4

1

A18

43.2301

89.7206

287

1.5

0.251

2

0.0

0.0

0.0

100.0

5

1

9

4

A19

43.0723

89.1320

174

1.0

0.215

12

0.7

0.0

0.0

97.8

60

1

8

2

A20

43.0706

89.8131

182

2.0

0.597

19

0.0

0.0

1.6

97.5

80

1

5

2

A21

43.0261

89.2842

10522

2.0

0.281

8

34.0

64.2

0.0

0.9

0

3

1

1

A22

43.0266

89.2843

1619

2.0

0.162

4

40.1

48.2

0.0

10.9

10

3

9

6

A23

43.0843

89.2676

279

0.7

0.352

2

13.2

23.3

0.0

63.6

0

1

6

4

A25

43.1602

89.2710

40000

3.0

0.374

6

8.1

1.6

71.8

16.9

15

5

2

0

A27

43.1656

89.4273

405

1.5

0.084

6

37.8

22.8

33.8

5.3

30

1

5

2

A28

43.0374

89.4289

445

1.0

0.297

22

52.0

40.0

0.0

8.0

45

1

9

5

A29

43.0371

89.4331

3035

2.0

0.254

22

32.5

22.3

0.0

44.2

10

1

10

7

A30

42.9987

89.5273

2023

2.0

0.089

20

3.6

0.0

0.0

93.5

0

5

8

4

A32

43.0877

89.2568

5261

2.0

0.143

8

47.2

24.0

18.1

9.7

0

3

1

1

The conductivity values and percent submerged macrophyte values reported here were measured during the June sample period. Richness values reflect species occurrences over both the June and September sample periods. Ponds with zero (A1, A12) or one (A8, A21, A32) aquatic hemipteran species were excluded from multivariate analyses. Cond = conductivity, Imperv = impervious land cover, Lawn = lawn land cover, Ag = agricultural land cover, Mead/Wd = least impact meadow and wooded land cover, Macro = submerged macrophytes, Fish (Ab.C.) = Fish abundance category (1 = no fish, see methods), Het SR = Aquatic hemipteran species richness, Cor SR = Corixidae species richness. The land cover percentages do not appear to sum to 100 because the percent water in the watershed is not shown

Strong seasonality was expressed by the notonectids Buenoa margaritacea Torre-Bueno and Notonecta insulata Kirby, both of which expressed consistently positive (B. margaritacea) or consistently negative changes (N. insulata) in presence/absence status in the ponds between June and September (Fig. 1). Buenoa margaritacea was found in six ponds in June (total abundance = 29) and in six additional ponds in September (total abundance = 474). In contrast, N. insulata was found in 13 ponds in June (total abundance = 407), and no longer found in six of those ponds in September, when its total abundance dropped to 79. The highest instability values were found for the corixid species Sigara alternata Say (10 changes) and Hesperocorixa vulgaris Hungerford (nine changes), the changes of which were only weakly seasonal (Fig. 1). Belostoma flumineum Say and Hesperocorixa laevigata Uhler were also relatively unstable, non-seasonal species (Fig. 1). Species that were fairly consistent in their occurrence across seasons (detected by the combination of low instability and low seasonality values) included both very rare (e.g., Sigara solensis Hungerford) and very common species (e.g., N. undulata) (Table 1, Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-008-9631-6/MediaObjects/10750_2008_9631_Fig1_HTML.gif
Fig. 1

Occurrence instability and seasonality of 25 aquatic hemipteran species sampled from 28 ponds in June and September. Species whose seasonality values were highly positive tended to be absent in June but present in September, particularly when the seasonality value was equal to the instability value (e.g., B. margaritacea). Species with highly negative seasonality values tended to decline in September, again, particularly when the seasonality value (absolute value) was equal to the instability value (e.g., N. insulata). Species with seasonality values close to zero were highly consistent in their presence/absence status in ponds, if their instability value was also low. Low seasonality values coupled with high instability values indicates that the species had several countering (non-directional) occurrence changes (e.g., B. flumineum)

Non-metric multi-dimensional scaling (NMS) revealed patterns of hemipteran assemblages (Fig. 2), and agreement was found between rank dissimilarities in the Bray–Curtis matrix and distances among sites in ordination space (stress = 6.20%). The correlation between environment and species variables in multidimensional space was validated by the Mantel test (R = 0.363, P = 0.003). Water depth and watershed agriculture were identified as the environmental variables contributing the most to this relationship (Best test, see methods). The ordination joint biplot enabled differences in community structure between ponds to be evaluated in terms of environment and species variables. All species and environmental variables portrayed as vectors on the joint biplot are significant at the 0.05 level (Table 3). Since N. undulata was responsible for creating the largest (ranked) distance in species structure between ponds, it was chosen as the species variable to align with axis 1 (Fig. 2). The abundance of this species was negatively correlated water depth and fish abundance, and independent of watershed land cover. The abundance of Hesperocorixa obliqua Hungerford, a large corixid species, also followed this same pattern. Axis 2 was largely driven by four smaller corixid species (Corisella tarsalis Fieber, Corisella inscripta Uhler, Trichocorixa calva Say, and Sigara decorata Abbott) each expressing a negative association with agricultural land cover in the watershed (Fig. 2). Aquatic hemipteran richness, although not strongly correlated with either axis individually, was positively correlated with watershed lawn, and negatively correlated with pond surface area. Agriculture and lawn were the only land cover categories significantly correlated with the relationship between ponds based on species composition. Low-impact land cover (woods, meadow), impervious surface, pond age, macrophyte abundance, and conductivity had no relationship to species variables in multidimensional space.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-008-9631-6/MediaObjects/10750_2008_9631_Fig2_HTML.gif
Fig. 2

The NMS biplot of environmental and species correlates overlayed on the pond-by-species ordination. Ponds were separated based on Bray–Curtis distance in species abundances. Notonecta undulata was the species variable most correlated with one of the axes (i.e., did the best job separating the ponds based on ranked distance), therefore the graph was rotated to align this species with Axis 1. Each species and environmental vector displayed in the graph is significantly correlated with one of the axes at the 0.05 level (Pearson and Kendall test). Vector lengths indicate the strength of the correlations with the axes. Of the environmental variables tested, pond morphology variables, watershed land cover percentages, and fish abundance were all correlated with species structure. Of the land cover variables, only the high-impact categories (percent agriculture and percent lawn) were significantly correlated with species structure. In addition to N. undulata, five corixid species (C. tarsalis, C. inscripta, S. decorata, T. calva, and H. obliqua), and hemipteran richness have significant relationships with the projected community structure. Ag = Agriculture, WaterDep = Water Depth, H Rich = Hemipteran Richness, H.obl = Hesperocorixa obliqua, N.und = Notonecta undulata, S.dec = Sigara decorata, T.cal = Trichocorixa calva, C.ins = Corisella inscripta, C.tar = Corisella tarsalis

Table 3

Significant correlations (Pearson and Kendall test) of environmental and species variables with the NMS ordination axes (Critical r value for significance at the 0.05 level = 0.404)

 

Axis 1

Axis 2

r

r²

r

r²

Pond area

−0.585

0.343

0.569

0.324

Water depth

−0.429

0.184

0.068

0.005

Lawn

0.355

0.126

−0.368

0.135

Agriculture

−0.286

0.082

0.519

0.269

Fish abund

−0.46

0.211

−0.177

0.031

C. inscripta

0.197

0.039

−0.523

0.274

C. tarsalis

0.38

0.144

−0.632

0.4

H. obliqua

0.432

0.187

0.073

0.005

S. decorata

0.107

0.011

−0.433

0.187

T. calva

0.23

0.053

−0.436

0.19

N. undulata

0.759

0.576

0.139

0.019

Hem richness

0.609

0.371

−0.586

0.343

Cluster analysis using Bray–Curtis similarity revealed the non-random co-occurrence of certain hemipteran species in these ponds (Fig. 3). Notonecta undulata, S. alternata, and H. obliqua have 89% similarity in the ponds in which they occur. This clade is significant at the 0.05 level (SIMPROF test). Interestingly, the three other significant clades with high similarity values consisted of species-pairs belonging to the same corixid genus (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-008-9631-6/MediaObjects/10750_2008_9631_Fig3_HTML.gif
Fig. 3

Habitat co-occurrence tendencies of 21 aquatic hemipteran taxa sampled from 28 ponds. Cluster analysis of species assemblages was based on presence/absence data, and performed using the Bray–Curtis distance measure. The hierarchy of the dendrogram was determined by group average fusion. Circles indicate significant clades (0.05 level, SIMPROF test) with high similarity values (>75%). Clades 1, 2, and 3 each consist of corixid species belonging to the same genus. Clade 4 consists of three extremely common species with relatively high abundance values (Table 1). Taxa are abbreviated as: F(amily): G(enus) s(pecies); species names can be found in Table 1

Discussion

Despite their vulnerability to chemical and physical disturbance, permanent stormwater retention ponds are an important habitat for many species of aquatic bugs (Table 1). The tendency of B. margaritacea to increase in abundance and occurrence between early summer and early fall (Fig. 1) is consistent with reports in the literature. This species differs from the other notonectids in this system in both its overwintering strategy (as eggs, rather than adults) (Hilsenhoff, 1995), and in the lengthy gestation period of the eggs. Buenoa species generally have a gestation period of approximately 42–77 days (Bare, 1926), while Notonecta eggs develop in as little as 3–14 days (Clark & Hersh, 1939). The tendency of N. insulata to decrease in abundance and occurrence between early summer and early fall may be due to this species’ intolerance of high water temperatures (Streams, 1987b). Warm, shallow pond-conditions (more common at the end of summer) are probably inhospitable to this species (Streams, 1987b). Several species were found to be highly consistent in their occurrence across seasons (low instability and seasonality), usually denoting their very rare (e.g., S. solensis) or ubiquitous (e.g., N. undulata) pattern of distribution (Fig. 1, Table 1). The high instability of S. alternata and H. vulgaris and, to a lesser extent, B. flumineum and Hesperocorixa laevigata, indicate that these species frequently migrate in and out of ponds, or else maintain their presence in a pond but are difficult to capture and detect. Based on these findings, one sample event is not enough to assess the presence of all hemipterans in a small aquatic habitat, and repeated sampling regimes are especially critical for detection of the seasonal and instable species reported here (Fig. 1).

Non-metric multi-dimensional scaling (NMS) revealed the importance of both local and landscape scale variables in governing hemipteran species structure in ponds. This result is consistent with a recent study of Heteroptera species assemblages in streams, in which local and regional variables accounted for the most variation in community structure, and assemblages at the regional scale were mainly divided into those found in agricultural and forested landscapes (Karaouzas & Gritzalis, 2006). The abundant and ubiquitous distribution of N. undulata and H. obliqua in stormwater retention ponds indicate their tolerance to a wide array of species compositions and anthropogenic land use in the watershed. Although N. undulata is a strong flier with high dispersal potential and the ability to colonize almost any freshwater environment (Hilsenhoff, 1984; Papacek, 2001; Chordas et al., 2005), the tendency of this species to colonize shallow, fishless ponds (Fig. 2, Bennett & Streams, 1986) may be due to direct avoidance of fish predators (e.g., sunfish, Crowder & Cooper, 1982) or to indirect selection of smaller, shallower ponds (e.g., Svensson et al., 2000) in which fish are less likely to prosper. Streams (1987a) noted a major division in the Notonecta species colonizing habitat with and without insectivorous fish, and proposed that the foraging strategy of N. undulata (active search) when compared to N. lunata Hungerford and other congenerics (sit and wait), may explain this species’ poor adaption to fish (Streams, 1987a). In controlled fishless environments, however, the foraging efficiency (prey capture/predator, three types of prey) of N. undulata far exceeds that of N. lunata and N. insulata (Streams, 1987a), the two less common and abundant Notonecta species in our study (Table 1). Additionally, unlike N. insulata, N. undulata displays tolerance to high water temperatures which often occur in shallow ponds (Streams, 1987b). Artificial stormwater ponds tend to be shallow, sparsely vegetated, and prone to high levels of contamination (Dodson, 2008). Tolerance to such factors may give N. undulata and other opportunistic species a competitive advantage in anthropogenically influenced habitat, and their potential to displace less tolerant species in the current landscape should be examined.

Axis 2 of the NMS ordination biplot was largely driven by four smaller corixid species (C. tarsalis, C. inscripta, T. calva, and S. decorata) each expressing a negative association with agricultural land cover in the watershed (Fig. 2). This relationship suggests an anthropogenic source of pollution affecting agricultural ponds, and also indicates that these species are more sensitive to the aquatic environment imposed by watershed agriculture than that by lawn or impervious surface. Specific impacts of agricultural pollution on corixids have been documented; for example, the dietary exposure of Trichocorixa reticulata Say to subsurface agricultural drainage can result in the accumulation of selenium in the insect’s body (Thomas et al., 1999). Tissue residues of the aquatic herbicide dichlobenil have also been found to accumulate rapidly in exposed Sigara dorsalis Leach and Corixa punctata Illiger, inhibiting elytral pigmentation in both nymphs and newly developed adults (Tooby & Macey, 1977).

Given the negative impact of watershed lawn on the richness and abundance of other aquatic taxa (e.g., zooplankton and amphibians) (Dodson, 2008), the positive relationship between aquatic hemipteran richness and watershed lawn was unexpected (Fig. 2). If lawn is indeed the main driver of richness increases in this system, then the potential shift in hemipteran composition from sensitive to insensitive taxa (e.g., Lussier et al., 2008) deserves attention.

Neither impervious surface nor conductivity was related to community structure, suggesting inconsistent or complex effects of these variables on pond habitat and community. Pond age and submerged macrophyte abundance, both of which have been shown to increase richness in other insect taxa (e.g., dragonflies, Kadoya et al., 2004) also had no relationship to species variables in multidimensional space (Fig. 2). Although it is clear that aquatic vegetation is a major food and resource for many of these bug species (Hilsenhoff, 1984), submerged macrophytes also impose structural complexity which can have negative effects on predatory species by reducing predation rates and providing a refuge for prey (Cook & Streams, 1984). The lack of relationship between macrophytes and hemipteran community structure is consistent with the finding that species richness is not negatively affected by herbicide-intensive lawns in the watershed (Fig. 2). Herbicides kill both terrestrial and aquatic plants, but aquatic hemipterans (particularly N. undulata, the main driver of community structure in this system) do not appear to be affected by the loss of plants. While the limited influence of vegetation on Notonecta species has been reported (Bennett & Streams, 1986; Briers & Warren, 2000), the apparent lack of relationship between corixid abundance and submerged macrophytes is somewhat surprising, since these herbivores use aquatic vegetation to meet dietary needs (Hilsenhoff, 1984). A more sensitive (species level) measure of plant and algal resources would likely enable clearer assessment of their influence on corixids and other herbivorous taxa. In this system, Potamogeton species were the most common dominant macrophytes, although a few of the well-vegetated ponds had relatively homogeneous populations of Eurasian milfoil (Myriophyllum spicatum), which was not found to support aquatic hemipterans (e.g., Pond A1, Table 1, 100% M. spicatum cover).

Heirarchical cluster analysis identified non-random distribution patterns of certain species of hemipterans. Notonecta undulata, S. alternata, and H. obliqua were the three species with the highest similarities (89%) in pond occurrence (Fig. 3). Although two of these species share the positive association with smaller, shallower ponds (Fig. 2), all three were among the most common and abundant species in this system (Table 1), indicating their tolerance to a wide array of species compositions and anthropogenic land use in the watershed. Interestingly, H. obliqua has recently been identified as a species of greatest conservation need (Wisconsin Department of Natural Resources, 2005), probably due to reports by Hilsenhoff (e.g., 1970) noting the rarity of this species compared with H. vulgaris. Natural ponds and wetlands, such as those sampled by Hilsenhoff, were once the major habitat for aquatic bugs. However, in the past 150 years, more than half of Wisconsin’s original 10 million acres of wetlands and ponds have been filled in or drained (Dodson & Lillie, 2001), most of this loss occurring in Southern Wisconsin. In the current landscape, natural lentic habitat has largely been replaced by artificial ponds, many of which are the result of 2001 state legislation (NR151) which dictates the creation of retention ponds for storm water management practices (Dodson, 2008). The opposite-of-expected relative abundances of H. obliqua and H. vulgaris encountered in this study (Table 1) may reveal a species shift due to anthropogenic habitat change.

Cluster analysis also highlighted the co-occurrence tendencies of several corixid congenerics (Fig. 3). Traditionally, the tendency of closely-related species to co-occur is assessed in terms of similar habitat requirements (which would drive species together) and competition for limited resources (which would drive them apart) (e.g., Briers & Warren, 1999). Body size is an important factor in the competitive potential between aquatic hemipterans, and species of similar body sizes should have increased competition relative to species of different body sizes (Svensson et al., 2000). Indeed, the distribution patterns of corixids in this system appear to be best explained by genus-level differences in body size; however, body size similarity serves to unite, rather than separate, closely related species (Fig. 3). Species of Trichocorixa are the smallest corixids in Wisconsin (~3 mm), those of Hesperocorixa are the largest (~11 mm) and those of Corisella are intermediate (~7.5 mm) (Hungerford, 1948). Given those size disparities, our results contradict predictions of traditional niche theory in which similar-sized species do not co-occur (Hutchinson, 1962). Rather, it appears that the distribution pattern of these corixids is strongly influenced by non-competitive size-related interactions with the pond habitat or community, such as size-selective predation by waterfowl, fish, and/or aquatic insects (Dodson et al., 1994).

In summary, results of this study suggest that both agriculture and lawn in the watershed alter habitat quality in such a way as to have consistent impacts on hemipteran community structure. Chemical inputs, which vary with land usage in the watershed (Tong & Chen, 2002), are the most obvious mechanism for the relationships between land cover and species reported here, since pesticides, like predators, elicit density and behaviorally mediated changes in select species (Fleeger et al., 2003; Relyea & Hoverman, 2006), which can, in turn, have profound impacts at the population and community levels (Dill et al., 2003, Rohr & Crumrine, 2005, Rothley & Dutton, 2006, reviewed in Schmitz et al., 2004 and Werner & Peacor, 2003).

Community structure is traditionally evaluated in terms of random dispersal followed by non-random competition, predation, and resource-related mortality. The chemical sensitivity of insects, however, suggests the importance of non-random dispersal in systems of heterogeneous chemical habitat. Oviposition habitat selectivity (OHS) provides an interesting, largely overlooked, mechanism for community assembly in which species composition is largely limited by the frequency and success of surrounding populations’ selective dispersal events via the use of highly specific visual, tactile, and/or chemical cues (e.g., Åbjornsson et al., 2002; Blaustein et al., 2004, 2005; Binckley & Resetarits, 2005). As increasing research demonstrates the importance of OHS in structuring aquatic insect populations, communities, and metacommunity assemblages (Resetarits, 2005; Encalada & Peckarsky, 2006), questions arise as to the maintenance of this highly evolved, often chemically mediated behavior in anthropogenically altered chemical environments. Anthropogenic contaminants are a ubiquitous (Gilliom et al., 2006), but relatively recent environmental phenomena and it is unclear whether they are general enough, or have persisted long enough, to be perceived by the odor and taste receptors of ovipositing/migrating aquatic insects (receptors reviewed in Hallam et al., 2006; olfaction in Rutzlerm & Zwiebel, 2005). Comparison between the field data, presented here, and experimental monitoring of hemipteran migration into water from different watersheds would help reveal if the species present/absent in high impact ponds are able to perceive and select/avoid water chemistry associated with watershed land use, or if observed community differences in ponds with different watershed land cover are better explained by non-contaminant-related selection criteria and/or variation in post-colonization survival and success. Despite their undesirable presence in the environment, aquatic contaminants associated with watershed land cover may provide useful insight into the sensory perceptions, migration behaviors, and community structure patterns in aquatic insect assemblages. Artificial stormwater ponds are readily accessible, highly replicated, and understudied aquatic environments in which to observe these unique toxicological effects. The need for this line of investigation is growing as the abundance of these ponds within the current landscape begins to alter what limnologists think of as aquatic habitat, and arguably, what aquatic invertebrates call home.

Copyright information

© Springer Science+Business Media B.V. 2008