Journal of Insect Conservation

, Volume 14, Issue 6, pp 689–700

Habitat and conservation of the enigmatic damselfly Ischnura pumilio

Authors

    • School of Biological SciencesUniversity of Liverpool
  • Mike G. Le Duc
    • School of Biological SciencesUniversity of Liverpool
  • David J. Thompson
    • School of Biological SciencesUniversity of Liverpool
Original Paper

DOI: 10.1007/s10841-010-9297-z

Cite this article as:
Allen, K.A., Le Duc, M.G. & Thompson, D.J. J Insect Conserv (2010) 14: 689. doi:10.1007/s10841-010-9297-z

Abstract

Ischnura pumilio is threatened in the UK and its habitat requirements are not well understood. This study tests previously held notions of the habitat requirements of I. pumilio, investigates the features of a habitat influencing odonate species composition and provides recommendations for habitat creation and management for I. pumilio persistence. Thirty-one sites across south west England with past I. pumilio records were surveyed in 2006. Environmental variables and odonate abundance were recorded. Odonate species composition and I. pumilio abundance were related to environmental variables using multivariate techniques and GLM. Ischnura pumilio was found at a wide variety of habitat types; key habitat features were a muddy substrate with some open ground, turbid water, and low levels of shade. It was associated with increased structural diversity of vegetation away from water but low maximum height; characteristic of early-successional sites. The variables predicting odonate composition were location, shade, level of disturbance, water depth, and cover of terrestrial dwarf shrubs and Sphagnum species. Vegetation height and structure were also highly influential to at least 20 m from water. This study indicates that odonate habitat management should include adjacent hinterland. Management for I. pumilio may be complicated by the species’ use of two habitat types, each with associated problems. Furthermore, odonate species diversity was negatively associated with I. pumilio abundance, which may cause conflict of interest when managing habitats.

Keywords

Ischnura pumilioHabitat extentVegetation structureConservationOdonata

Introduction

The abundance and status of Ischnura pumilio (Charpentier) (Odonata: Coenagrionidae) in the UK is not well known, but a decline in numbers was reported at the end of the last century (Cham 1991, 1996) and the species is classed as near threatened in the Odonata Red Data List for Great Britain (Daguet et al. 2008). Recording of the species has been notoriously difficult due to a sparse, localised distribution (Askew 2003) and frequent misidentification (Cotton 1981; Fox 1987). The species was thought to be highly transient; colonising newly formed habitat and persisting only a few years (Fox 1989; Cham 1996; Askew 2003) earning it a reputation as a “wandering opportunist” (Fox 1989; Cham 1996). However, recent work has shown that in large, maintained areas of habitat, colonies may be extremely sedentary (Allen and Thompson 2010).

Ischnura pumilio is often reported to have very specific habitat requirements (slow flowing, shallow water with limited emergent vegetation) (Fox 1989; Cham 1991), but it is found at a variety of habitat types in the UK (Cotton 1981; Fox and Cham 1994). It is therefore important to distinguish which habitat features are important, from those which are incidental. The species is largely restricted to south-western UK regions where it reaches its northern global range margin (Fig. 1; Chelmick 1980; Fox 1990). However, species’ ranges are predicted to extend towards the poles and higher altitudes as global temperatures rise, both across taxa (Hickling et al. 2006) and in the Odonata specifically (Hickling et al. 2005; Hassall and Thompson 2008). This may be accompanied by changes in availability of habitat and an understanding of habitat requirements is therefore crucial, if successful conservation management is to be achieved.
https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9297-z/MediaObjects/10841_2010_9297_Fig1_HTML.gif
Fig. 1

Main: All sites surveyed with codes given in Table 1. Filled circles were occupied by I. pumilio and open circles were not. Inset: Map showing UK distribution (10 km grid squares with records of I. pumilio held by NBN gateway). Dotted box indicates extent of main map

This study aimed to test the hypothesis that I. pumilio has highly specific habitat requirements, and determine key habitat features in order to identify existing potential sites and inform conservation management via habitat creation and maintenance. The study also considers the hypothesis that the species does not occur where odonate species diversity is high (Dapling and Rocker 1969; Fox 1987) and determines which species I. pumilio is positively and negatively associated with. The species has been under-recorded due to its small size and ephemeral habitat use (Dapling and Rocker 1969; Cotton 1981; Fox 1987) and a set of environmental and species indicators could help determine where surveying efforts should be focussed. The study combines field measures of abiotic conditions, plant species composition and the presence of other odonates to determine the habitat most suitable for I. pumilio.

Methods

Study sites

During June and July 2006, 31 sites were surveyed across Cornwall, Devon and Hampshire, UK (Fig. 1; Table 1). Sites were chosen at random from those with British Dragonfly Society records of I. pumilio during the last 10 years (Smallshire 1996; Jones 2006). Sites varied from natural, undisturbed areas to previously industrialised and other disturbed or manmade sites. Some areas were surveyed as two or more sites because they formed distinct areas of potential habitat (indicated by numeric suffixes in Table 1). One site was surveyed per day. No surveys were conducted during rain or strong winds. One site (Menadue) was omitted from analyses as no water was present at the time of the survey.
Table 1

Table of site names and codes, British National Grid references (centre of site), area (m2), NVC community codes, NVC goodness of fit (GOF) values, I. pumilio abundance and level of disturbance as defined in Table 2

Site

Code

Grid reference

Area (m2)

NVC

GOF

Number I. pumilio

Disturbance

Great Wheal Seton

Aa

SW655417

2,500

S11c

18

26

2

Roscroggan

Ab

SW651420

900

W1

21

0

0

Rosewarne 1

Ac

SW644417

4,800

S10

16

0

1

Rosewarne 2

Ad

SW644418

4,500

S25

12

3

1

Bell Lake Marsh

Ae

SW621418

3,500

S10

12

0

0

Peter’s Point

Af

SW577410

300

MC9c

26

0

1

Chapel Porth 1

Ag

SW699493

1,400

S10

14

0

1

Chapel Porth 2

Ah

SW698494

500

MC8

19

11

1

Wimal Ford

Ai

SX212734

1,600

M15

49

0

1

Sand Cottages

Aj

SW789397

13,600

A24

13

1

2

Penwithick

Ak

SX020562

3,800

M23

16

7

1

Stepper Point

Al

SW914783

1,300

SD17c

10

53

1

Menadue

Am

SX027593

1,300

M29

24

0

0

Tolgus tin

An

SW688447

3,800

W25b

15

0

0

Carbis Moor

Ao

SX027558

3,200

W4

7

2

1

North Tresamble

Ap

SW747404

200

A24

18

0

1

Newlyn East Downs

Aq

SW836543

1,300

H9e

17

0

1

Prewley Moor

Br

SX542909

48,200

M25b

15

4

1

Sheepstor

Bs

SX578682

1,400

M25a

42

0

0

Cadover Bridge

Bt

SX556648

12,500

M23

29

51

0

Blackabrook

Bu

SX569639

2,700

M6a

27

13

0

Smallhanger Waste

Bv

SX576594

25,900

M25a

31

29

1

Whitchurch Down

Bw

SX508741

18,800

M25

33

0

1

Wigford Down

Bx

SX537649

61,300

M25a

45

0

0

Lydford Railway

By

SX500825

15,600

M25

40

7

0

Walla Brook

Bz

SX670771

144,000

M6a

33

6

0

Latchmoor 1

Ca

SU194127

28,800

M25

20

58

0

Latchmoor 2

Cb

SU191127

7,200

M29

23

23

0

Latchmoor 3

Cc

SU190127

21,600

M25a

29

19

0

Shipton Bottom

Cd

SU362996

9,900

M25a

26

3

0

Millersford Bottom

Ce

SU186167

109,000

M25a

50

0

0

Site codes commencing “A” are in Cornwall, “B” in Devon and “C” in Hampshire. NVC community definitions are: A24, Juncus bulbosus community; H9, Calluna vulgaris-Deschampsia flexuosa heath; M6, Carex echinata-Sphagnum recurvum/auriculatum mire; M15, Scirpus cespitosus-Erica tetralix wet heath; M23, Juncus effusus/acutiflorus-Galium palustre rush-pasture; M25, Molinia caerulea-Potentilla erecta mire; M29, Hypericum elodes-Potamogeton polygonifolius soakway; MC8, Festuca rubra-Armeria maritima maritime grassland; MC9, Festuca rubra-Holcus lanatus maritime grassland; S10, Equisetum fluviatile swamp; S11, Carex vesicaria swamp; S25, Phragmites australis-Eupatorium cannabinum tall-herb fen; SD17, Potentilla anserina-Carex nigra dune-slack community; W1, Salix cinerea-Galium palustre woodland; W4, Betula pubescens-Molinia caerulea woodland; W25, Pteridium aquilinum-Rubus fruticosus underscrub. Lower case letters refer to sub-communities which are not defined here due to low GOF values

A suite of environmental variables was recorded both at site level and within quadrats (Table 2). Quadrats were positioned along transects emanating from points within water. Start points and direction of transects were determined prior to arrival at a site using a numbered grid superimposed onto a site map. Randomly generated numbers determined each start point and a second point through which to draw the transect direction vector. Where water at the start point was more than 1 m deep, the quadrat sequence was shifted along the transect direction vector to start at 1 m depth, thus limiting the survey to the marginal areas of water which I. pumilio inhabits. Six transects were defined at each site, each consisting of 5 quadrats placed at 0, 2, 5, 10 and 20 m from the start point within water. Percentage cover of plant species was estimated and UK National Vegetation Classification (NVC) classes (Rodwell 1991) assigned for each site using Tablefit 1.0 (Hill 1996).
Table 2

Potential predictors of odonate species composition and I. pumilio abundance, and transformations applied (in parentheses)

Variable

Description

Easting

British national grid location (Log10)

Northing

British national grid location (Log10)

Altitude

As recorded by handheld GPS (m) (Log10)

Area

Area of potential habitat (m2)estimated from map (Log10)

Bankangle (bank)

Clinometer reading between transect points 0 & 10 (m) (None)

Disturbance

Human disturbance: 0, relatively untouched; 1, partially disturbed, some management, pathways, nearby roads; 2, industrial, heavily used sites (None)

Grazing

0, none; 1, light/occasional; 2, heavy. (None)

Shade

% shade over quadrat (Arcsine)

Type

Stream/flush/bog (Arcsine)

Mudcover

% cover of bare mud (Arcsine)

Muddepth (mud)

Depth of mud (cm) in centre of quadrat (Log10)

Watercover

% cover of water (Arcsine)

Waterdepth (water)

Depth of water (cm) in centre of quadrat (Log10)

Substrate

Silt/silt and gravel/gravel

Turbidity

Proportion of standard grayscale visible through standard tube of water from quadrat (Arcsine)

Conductivity (cond)

μSiemens measured in quadrat

Flowrate

Measured in m s−1 using 1 cm2 of polystyrene (Arcsine)

Maxheight

Maximum height of vegetation within quadrat (cm) (None)

Structure at 0 m (S00)

Structural complexity index at transect point 0 m (None)

Structure at 2 m (S02)

Structural complexity index at transect point 2 m (None)

Structure at 5 m (S05)

Structural complexity index at transect point 5 m (None)

Structure at 10 m (S10)

Structural complexity index at transect point 10 m (None)

Structure at 20 m (S20)

Structural complexity index at transect point 20 m (None)

Spcount (sp)

Number of odonate species observed at site (Log10)

ACG

Aquatic sedge/grass (% cover) (Arcsine)

AFO

Aquatic forb (% cover) (Arcsine)

AH

Aquatic horsetail (% cover) (Arcsine) (a)

MC

Marginal sedge (% cover) (Arcsine) (b)

MFO

Marginal forb (% cover) (Arcsine)

MG

Marginal grass (% cover) (Arcsine) (b)

MM

Marginal moss (% cover) (Arcsine)

MR

Marginal Rush (% cover) (Arcsine)

TC

Terrestrial sedge (% cover) (Arcsine) (c)

TD

Terrestrial dwarf shrub (% cover) (Arcsine)

TFE

Terrestrial fern (% cover) (Arcsine) (a)

TFO

Terrestrial forb (% cover) (Arcsine)

TG

Terrestrial grass (% cover) (Arcsine) (c)

TR

Terrestrial Rush (% cover) (Arcsine)

TS

Terrestrial shrub (% cover) (Arcsine) (b)

TT

Terrestrial tree (% cover) (Arcsine)

Variable names in parentheses are used in Fig. 2

Plant species were grouped according to functional group and habitat (aquatic, marginal or terrestrial) to reduce the number of variables to be considered (see Table 2). Site means (geometric means for percentages) were calculated for explanatory variables measured at transect or quadrat level. Coverage of each plant group at a site was calculated as the sum of the geometric mean coverage of all species in the group.

Plant structural complexity was measured by placing a marked pole in the centre of each quadrat, and recording the number of contacts by plants in each of the vertical intervals 0–10, 10–30, 30–70, 70–150 and over 150 cm (Lawton and Woodroffe 1991). The first principal component of these measurements was used as a compound measure of structural complexity in analyses to avoid multicollinearity (Graham 2003). Structural complexity was considered at each transect point, as Ischnura pumilio was expected to prefer limited emergent vegetation for oviposition, but increased structural complexity away from water as a source of shelter and food.

All I. pumilio sighted during surveying between 11 am and 4 pm were captured, counted and marked to prevent recounting. This included half an hour of searching exclusively for I. pumilio at each site. Abundance of other odonates was estimated in categories according to the British Dragonfly Society’s abundance classification system (1, 2–5, 6–20, 21–100, 100–500 and 500+) and the median of each category used in analyses.

Odonate habitat associations

Associations between odonate species composition and the abiotic factors and plant species at a site (hereafter referred to collectively as environmental variables) were investigated using multivariate techniques. A correspondence analysis (CA) (Oksanen 2008) of odonate species produced a scatter of 6 ordination score units along axis 1, and 5 ordination score units along axis 2. Enallagma cyathigerum (Charpentier) and Orthetrum cancellatum (L.) were outliers which strongly influenced this CA solution, and down-weighting species present in less than 20% of sites lowered the variation in scatter of species points. Constrained correspondence analysis (CCA) was then performed on the same data to examine the variation in species data due to measured environmental variables.

Model selection followed several stages due to the large number of potential explanatory variables. Initially, the set of environmental variables was split into three random subsets and each subset was used to define a maximal model. Variables with variance inflation factors (VIFs) greater than 10 (indicating correlation with other variables) were removed from each maximal model before model selection (Oksanen 2008). Maximal models were then subjected to backwards and forwards selection based on Akaike’s Information Criterion (AIC). This procedure was repeated 10 times. Variables in the resultant minimal models were then combined for further model selection following the same procedure. The significance of terms in the resulting final model was calculated using a permutation test (Oksanen 2008).

Non-normally distributed variables were subjected to an appropriate transformation (see Table 2). All explanatory variables were then standardised to zero mean and unit variance. Multivariate analyses were performed in R 2.4.0 (R Development Core Team 2005) using the package Vegan 1.13-0 (Oksanen et al. 2008).

Ischnura pumilio habitat associations

The relationship between I. pumilio abundance and other measured variables was investigated using Generalized Linear Models (GLM). A logarithmic link function was used with quasi-Poisson errors to account for overdispersion resulting from an aggregated I. pumilio distribution. This applies a scale parameter to the model (Pearson’s χ2/df), such that the variance increases with the mean. The maximal model was simplified using backwards selection based on F tests on deletion, until all variables were significant at P ≤ 0.05 (Crawley 2007). All non-normally distributed variables were subjected to an appropriate transformation (Table 2). Due to the large number of potential explanatory variables, model building was performed in several stages. Continuous variables were fitted first (along with their quadratic terms to test for non-linear effects), in sets which ensured the maximal model did not have more than n/3 parameters (Crawley 2007). The minimum adequate model at the end of each stage was used as the base for the following stage of model building. Categorical variables were fitted in the later stages and when they remained in the minimum adequate model, aggregated categories were also tested (Crawley 2007). D2 and adjusted D2 were calculated using standard formulae (Guisan and Zimmermann 2000).

To reduce collinearity, the covariance matrix was examined and pairs of variables with significant Pearson correlation coefficients (df = 29, P < 0.01) were removed from the analysis. Variables correlated with the largest number of other variables were removed one by one until no significant correlations remained. Variables removed from the analysis were easting, northing, altitude, percentage cover of water, conductivity, maximum height of vegetation and structural indices at transect points 2, 5 and 10 m.

Results

Description of sites

Ischnura pumilio was recorded at 17 of the 31 sites surveyed (Table 3). NVC classes varied greatly, from aquatic (A24) to woodland communities (W1, W4 and W25). Mires were the most common NVC class and accounted for 11 of the 17 sites with I. pumilio, mostly those in Devon and Hampshire. Most sites were shallow flushes or boggy areas with slow-flowing or still water, and a silt or silt and gravel substrate. Juncus bulbosus L. was present at every site where I. pumilio was recorded and Juncus effusus L. at all except one (Table 4).
Table 3

Odonate species, codes used in ordination plots, frequency of occurrence across the 31 sites surveyed and frequency of co-occurrence with I. pumilio

Species

Code

All sites

I. pumilio sites

Calopteryx virgo

C.v

8

3

Ceriagrion tenellum

C.t

13

10

Coenagrion mercuriale

C.m

6

5

Coenagrion puella

C.p

21

13

Enallagma cyathigerum

E.c

5

5

Ischnura elegans

I.e.

20

13

Ischnura pumilio

I.p

17

n/a

Lestes sponsa

L.s

5

5

Pyrrhosoma nymphula

P.n

22

12

Aeshna juncea

A.j

1

7

Anax imperator

A.i

12

0

Cordulegaster boltonii

C.b

9

5

Libellula depressa

L.d

9

4

Libellula quadrimaculata

L.q

7

3

Orthetrum cancellatum

O.ca

2

1

Orthetrum coerulescens

O.co

23

14

Sympetrum danae

S.d

1

1

Sympetrum striolatum

S.s

2

1

Table 4

Plant species occurring in 10 or more I. pumilio sites, number of occurrences in all 31 sites and in the 17 I. pumilio sites

Species

All sites

I. pumilio sites

Juncus bulbosus

30

17

Juncus effusus

29

16

Juncus articulatus

22

14

Erica tetralix

21

13

Festuca rubra

23

13

Lotus pedunculatus

21

13

Ulex europaea

23

13

Calluna vulgaris

25

12

Potamogeton polygonifolius

18

12

Holcus lanatus

17

11

Potentilla erecta

17

11

Agrostis canina

14

10

Molinia caerulea

18

10

Viola sp.

15

10

Odonate habitat associations

Inevitably, several models described the data well (Table 5). However, the most significant terms in the final model (Table 6; Fig. 2) were highly significant in all models and are likely to be the most important. The most important variables predicting odonate composition at a site were GPS location (only easting was included in the model due to correlation with northing), shade and terrestrial dwarf shrub cover. Also highly significant were level of disturbance, maximum height of vegetation, Sphagnum species cover, structural complexity away from water (S20) and water depth (Fig. 2).
Table 5

Model specification and AIC values for the best five CCA models of odonate species composition (log10 count data). Variable names are explained in Table 2

Model

AIC

Bankangle + conductivity + disturbance + easting + maxheight + MC + MM + muddepth + shade + spcount + TD + TFO + TG + waterdepth + S00 + S10 + S20

29.81

Bankangle + conductivity + disturbance + easting + maxheight + MC + MM + shade + spcount + TD + TFE + TFO + TG + waterdepth + S00 + S10 + S20

30.1

Bankangle + conductivity + disturbance + easting + maxheight + MC + MM + shade + spcount + TD + TFO + TG + waterdepth + S00 + S10 + S20

30.56

Bankangle + conductivity + disturbance + easting + maxheight + MC + MM + muddepth + shade + spcount + TD + TFE + TG + waterdepth + S00 + S10 + S20

32.25

Bankangle + conductivity + disturbance + easting + maxheight + MC + MM + muddepth + shade + spcount + TD + TG + waterdepth + S00 + S10 + S20

32.39

Table 6

Significance of predictors of odonate composition derived from a minimum adequate CCA model based on AIC

Variable

Constrained inertia

Easting

0.1101***

Shade

0.1251***

TD

0.0617***

Disturbance

0.0382**

Maxheight

0.0389**

MM

0.0411**

Structure zone 20

0.0442**

Waterdepth

0.0394**

Bankangle

0.0344*

Conductivity

0.0298*

MC

0.0341*

Muddepth

0.0344*

Spcount

0.0366*

Structure zone 0

0.0346*

TFO

0.0336*

TG

0.0393*

Structure zone 10

0.0275

Residual

0.1878

For each significant variable, constrained inertia and the P-value derived from permutation tests (number of permutation = 9,999) are shown (* P < 0.05, ** P < 0.01, *** P < 0.001). Addition of terms in alternative sequences generated different significance of some variables (here the test arbitrarily added terms in the order they appear in Table 3.2 resulting in S10: 0.01 < P < 0.05, symbolised “.”). However, all terms in the best model were significant in at least one of five sequences tested and the order of significance of variables was generally maintained. Each significant variable has 1df with 12 residual df. The final model accounted for 81% of the variation in species composition

https://static-content.springer.com/image/art%3A10.1007%2Fs10841-010-9297-z/MediaObjects/10841_2010_9297_Fig2_HTML.gif
Fig. 2

Ordination diagram of the final CCA model of odonate community composition. Species codes contain a full stop (Table 3) and site codes do not (Table 1). Some variable names have been abbreviated to reduce label overlap (Table 2). The “+” adjacent to C.t represents C.m due to label overlap. Arrows represent the direction of the gradient of significant predictors. Sites and species are scaled symmetrically by the square root of eigenvalues (Oksanen et al. 2008). Only community structure related to variables in the final model is shown. The final CCA model had inertias of 0.803 (constrained) and 0.188 (unconstrained), and constrained axis eigenvalues for the first four axes (λ1–4) = (0.251, 0.124, 0.095, 0.083)

Orthetrum cancellatum, E. cyathigerum and to a lesser degree Anax imperator Leach are associated with large water bodies and can tolerate brackish, eutrophic or mineral rich water (Smallshire and Swash 2004). Their grouping and location on Fig. 2 indicate association with deeper water and higher levels of disturbance, which were common features of the clay mining sites surveyed. Aeshna juncea (L.), Calopteryx virgo L., Sympetrum striolatum (Charpentier), S. danae (Sulzer) and Cordulegaster boltonii (Donovan) formed another group, all preferring acid water (Smallshire and Swash 2004). This was supported by the association with Sphagnum species which are a common feature of acid bogs and mires. Along with Pyrrhosoma nymphula (Sulzer), which has broader habitat requirements, this group was associated with greater odonate species richness. Libellula quadrimaculata L. also prefers acidic habitat and was associated with this group, but was not observed outside Cornwall and is therefore plotted at a lower easting. In the centre of the plot are the species with broad habitat requirements but preferring smaller sites with standing water such as Libellula depressa L., Coenagrion puella (L.) and to a lesser degree Ischnura elegans (van der Linden) which is less common in acidic waters and is plotted further from those species preferring acid conditions.

Lestes sponsa (Hansemann), Ceriagrion tenellum (de Villers) and Coenagrion mercuriale (Charpentier) were very closely associated, and O. coerulescens was plotted nearby. These species are all associated with heathland sites, characterised by Sphagnum species, except C. mercuriale which is more specific in its requirements; preferring base rich, open, slow flowing waters (Smallshire and Swash 2004). Ischnura pumilio also prefers small, open and shallow water bodies, but tolerates a wide variety of water quality conditions. It was plotted relatively far from other species, suggesting that its habitat requirements differ from all other species considered. Ischnura pumilio’s central placing on the first CCA axis, which represents the major environmental gradient influencing species composition, indicates that the factors correlated with this axis do not strongly influence its distribution (easting, shade, water depth, disturbance, maximum vegetation height, bank angle, conductivity and the percentage cover of several plant groups). The positioning of I. pumilio at a negative value along the second CCA axis suggests the species prefers shallow mud, a low species count and increased vegetation structural diversity.

Ischnura pumilio habitat associations

GLM analysis supported the results of the CCA analysis in terms of I. pumilio habitat preferences. Sites with low but present levels of shade and mud coverage were favoured by I. pumilio, indicated by quadratic terms (Table 7). Numbers were greater in water with moderate to high turbidity ratings, but at the highest levels abundance decreased. Silt substrates were preferred to gravel, and bogs preferred to flushes. Increased structural complexity away from water (S20) was associated with increased numbers, although vegetation structure at transect points near the water did not feature in the final model, despite being closely associated with I. pumilio in the CCA plot. A positive association with area was marginally significant and an F test on deletion indicated a significant decrease in explained variation, therefore the term was retained. No significant effects of plant groups were found. However, I. pumilio abundance is unlikely to depend on broadly defined plant groups. Field observations suggested that important vegetation features were emergent vegetation for oviposition and male perching; and tussocks of graminoids and heathers for shelter during windy conditions. This is supported by the proportion of plant species frequently co-occurring with I. pumilio which were graminoids or heathers (62%; Table 4).
Table 7

Significant predictors of numbers of I. pumilio at a site derived from a generalized linear model with quasi-Poisson errors

Model summary

Variable

Parameter estimates

SE

t

Deviance = 63.903

Mud cover

4881.00

1150.00

4.248***

df = 19

Mud cover2

−1.22 × 106

2.79 × 105

−4.361***

Dev/df = 3.36

Shade

27710.00

5510.00

5.026***

D2 = 0.901

Shade2

−4.04 × 107

8.42 × 106

−4.801***

Adj. D2 = 0.850

Substrate–silt

3.623

0.91

3.971***

 

Type–flush

−2.223

0.51

−4.365***

 

Structure at 20 m

0.9387

0.31

3.046**

 

Turbidity

21.04

8.46

2.488*

 

Turbidity2

−8.369

3.25

−2.572*

 

Area

0.5685

0.32

1.777

 

Intercept

−1.87E + 01

5.39

−3.473**

The F-value and associated P-value, df, D2 and adjusted D2 are shown. For each significant variable the P-value derived from t-tests, parameter estimates on the logarithmic scale and standard errors (SE) are shown (* P < 0.05, ** P < 0.01, *** P < 0.001)

When odonate species abundances (log10 transformed) were used as predictors in a separate GLM analysis of I. pumilio abundance Orthetrum coerulescens and I. elegans were positively associated with I. pumilio, and Cordulegaster boltonii negatively associated (Table 8). A negative association with Libellula quadrimaculata was marginally significant and an F test on deletion indicated a significant decrease in explained variation, therefore the term was retained. However, when these predictors were added to the previous final model they were non-significant. Therefore, habitat features better predicted I. pumilio abundance than the abundance of other recorded odonates.
Table 8

Significant predictors of numbers of I. pumilio at a site derived from a generalized linear model with quasi-Poisson errors

Model summary

Variable

Parameter estimates

SE

t

Deviance = 321.02

C. boltonii

−5.219

2.036

−2.563*

df = 25

O. coerulescens

1.335

0.490

2.728*

Dev/df = 12.84

I. elegans

1.177

0.512

2.298*

D2 = 0.505

L. quadrimaculata

−2.531

1.271

−1.990

Adj. D2 = 0.425

Intercept

0.513

0.864

0.594

The F-value and associated P-value, df, D2 and adjusted D2 are shown. For each significant variable the P-value derived from t-tests, parameter estimates on the logarithmic scale and standard errors (SE) are shown (* P < 0.05, ** P < 0.01, *** P < 0.001)

Discussion

Odonate habitat associations

Vegetation height and structure were highly influential on odonate species composition. Previous studies have shown vegetation structure to be more important than species in predicting odonate composition (Samways and Steytler 1996) and composition of other insect taxa (e.g. aphids; Strauss and Biedermann 2005). This was attributed to its effect on microclimate, as dense vegetation provides shelter from extreme temperatures, wind, rain and predators. Field observations support this as I. pumilio individuals regularly descended into tussocks of vegetation when approached on cooler, windier days. Surveys of odonates and their habitat are commonly conducted near to water, ignoring structural features of the hinterland (but see Stoks 2001), despite their importance for roosting and feeding (Foster and Soluk 2006; Rouquette and Thompson 2007). This study emphasises the importance of these structural components to at least 20 m from water.

The most commonly observed damselfly species were the nationally common and eurytopic P. nymphula, Coenagrion puella and I. elegans. However, the most commonly observed dragonfly was O. coerulescens, which is less common with more specialised habitat requirements. The species has a similar south–west distribution in the UK to I. pumilio (Smallshire and Swash 2004) and frequently co-occurs in Cornwall (S. Jones, personnel communication). The GLM indicates an association between the species; however its more frequent occurrence in these sites suggests that O. coerulescens can persist at a site after I. pumilio has been excluded.

Ischnura pumilio habitat associations

Despite previous notions that I. pumilio has highly specific habitat requirements, this study found the species at a wide variety of natural and manmade sites with very different profiles. The species was found at sites with a range of water depths, management regimes and levels of pH, pollution, grazing and disturbance. However, the unifying features of I. pumilio sites were mostly as previously suggested (Fox 1987, 1989; Cham 1990, 1991, 1992; Fox and Cham 1994). As the sites had previously supported I. pumilio populations, significant predictors may be considered those required for persistence at a site. However, I. pumilio populations are thought to exist at most sites for only a short time and then to move on (Fox 1989; Cham 1996; Askew 2003) making it difficult to determine whether the species’ absence is due to changes in habitat or to stochastic population dynamics. To our knowledge, no data exist on the extinction rate of colonies, but this information would assist in determining whether the loss of I. pumilio from the sites surveyed is part of normal dynamics or represents an overall decline.

Ischnura pumilio was most abundant at sites which were somewhat muddy, with silt rather than gravel substrates and some turbidity. This is characteristic of slow flowing or standing water where silt can accumulate and small particles become suspended in water, particularly where poaching by livestock occurs. Ischnura pumilio was associated with increased structural diversity of vegetation away from water but low maximum height, which is characteristic of the early-successional sites known to be preferred by the species (Fox 1989; Daguet 2005), where low-level vegetation may be dense before larger, over-shading plants have colonised.

Grazing and disturbance did not feature in this model despite being frequently reported as important factors (Fox 1987, 1989; Cham 1991). Additionally the CCA analysis did not find grazing to be important to odonate species composition; and while disturbance was an important factor, I. pumilio was placed centrally on the corresponding axis, indicating that disturbance does not strongly influence its distribution. However, the openness of most I. pumilio sites was maintained by one of these factors. The sparse and unusual vegetation at sites such as Great Wheal Seton may be the result of residual toxicity from tin mining, whereas at other key sites such as Latchmore it is intensive grazing which suppresses vegetation growth and succession. A degree of bare ground is known to be favoured by I. pumilio (Table 7; Fox et al. 1992) and it is openness, rather than the method of maintaining it, which is likely to be important. Many of the sites from which I. pumilio was absent appeared to be overgrown and undergoing succession to later stages, accompanied by drying in some cases. This suggests maintaining an early successional stage is key to management of sites for I. pumilio persistence.

The plant species most frequently co-occurring with I. pumilio (Table 4) were different to those previously listed (Fox 1987, 1989; Cham 1990), featuring only one common species: Juncus effusus. However, Ranunculus flammula L. listed by Fox (1987) was recorded at 8 of the 17 I. pumilio sites and one female was observed ovipositing into this species. It is unlikely that particular plant species are preferred by odonates, but they may either indicate favourable habitat or have soft stems suitable for oviposition (Rouquette and Thompson 2005 and refs therein).

Water depth was omitted from GLM analysis due to collinearity. However, Ischnura pumilio was recorded at the deepest water bodies in this study, albeit in small numbers, some of which may have been several metres deep at the centre. This contradicts previous notions that I. pumilio is restricted to shallow, spring-fed water courses due to thermal requirements (Cham 1991; Fox and Cham 1994; Strange et al. 2007).

Conservation implications

Odonate species diversity was negatively associated with I. pumilio abundance in the CCA analysis. This may be coincidental due to different habitat requirements, or may be due to competition with or predation of I. pumilio adults or larvae by larger species at more diverse sites. Similarly, the negative association with Cordulegaster boltonii and Libellula quadrimaculata indicated by the GLM may be due to predation by these species (observed once by L. quadrimaculata) or to their preference for acidic conditions and in the case of C. boltonii, fast flowing water (Smallshire and Swash 2004). This raises the question of whether I. pumilio has specific habitat requirements or is simply excluded from all but the most inhospitable sites, which it can tolerate due to greater ecological flexibility; previously attributed to a high chromosome number compared to other Coenagrionidae (Kiauta 1979). Management will depend on whether the priority is conservation of I. pumilio or the maximum number of species. This may jeopardise the persistence of I. pumilio given the financial and land limitations on conservation management. Maintenance of low level but structurally diverse vegetation at sites with shallow water will favour less common species such as I. pumilio, O. coerulescens and the nationally rare Coenagrion mercuriale. However, less intensive habitat management may favour greater species diversity, and may ultimately be preferred in conservation economic terms.

Development of effective conservation strategy requires detailed knowledge of the habitat requirements of all life stages (Thompson et al. 2003). Therefore, this study cannot represent the full range of I. pumilio habitat requirements. Odonates can be strongly affected by predator distribution (e.g. Stoks and McPeek 2003) and larval predation is likely to be a factor in determining I. pumilio presence at a site. No data on vertebrate predators was available to this study, although the shallow, boggy habitat often inhabited by I. pumilio is unlikely to support large fish populations. However, a negative effect of increased odonate diversity was detected and this may be due to larval predation by other odonates. Larval habitat is also important in determining a site’s suitability as larvae are confined to the water body (Thomas 1995; Hardersen 2008). Larval vegetation preferences may be stronger as plants provide shelter from predators (Thompson 1987; Elkin and Baker 2000) and are also important for foraging (Convey 1988). Water temperature is also likely to be important for larvae (Hickling et al. 2005) along with the physical conditions influencing it such as depth of water and over-shading (Strange et al. 2007). A study of larval populations may be useful, as adult abundances can be misleading in predicting larval habitats because adults may move away from the natal habitat due to different requirements (Raebel et al. 2010). However, I. pumilio can be particularly sedentary and no teneral dispersal was observed from high quality habitat in a previous study (Allen and Thompson 2010) suggesting dispersal to unfavourable larval habitats is unlikely.

Any designated conservation area should include the hinterland adjacent to water (Conrad et al. 2002; Bried and Ervin 2006), which may not be covered by current laws restricting wetland development (Foster and Soluk 2006), but is recognised by the IUCN Odonata Specialist Group as key to conserving odonates globally (Moore 1997). Creation of habitat for I. pumilio may be difficult as the species inhabits two perceivably distinct habitat types, each with associated problems. The first are high quality maintained habitats with existing conservation status such as most of the New Forest and Devon sites. Overgrazing to maintain low level vegetation at these sites is difficult to implement successfully, particularly where other species of conservation concern must also be considered. The second are polluted, manmade sites where there is a high level of past or present human intervention. While these are ideal, self-sustaining sites for I. pumilio, they are generally undesirable and it is unlikely their conservation or creation will be advocated.

Before applying these findings to conservation projects, particularly at lower latitudes, the models should be validated using data sets from other regions (Guisan et al. 2002). Transferability of habitat models has been demonstrated in other insects (e.g. (Bonn and Schröder 2001; Binzenhöfer et al. 2005) but has not been attempted with odonate species. At the northern limit of its range, the niches I. pumilio is able to occupy in the UK may become narrower towards its range margin, as conditions become increasingly prohibitive. Recorded movement distances and lack of evidence for metapopulation structure suggest that I. pumilio populations are significantly isolated from each other (Allen and Thompson 2010). If populations are genetically distinct they may have evolved different habitat selection (Whittingham et al. 2007) making these results less useful in other regions.

Threats to I. pumilio’s persistence have been identified as continued mineral extraction at quarry sites; lack of habitat management; succession and scrub encroachment; water pollution; and disturbance of aquifer fed spring lines leading to loss of habitat (Daguet 2005). However, it seems that I. pumilio can inhabit a broad range of habitat types including the potentially toxic waters of previous tin mining sites (Jones 1985). Continued mineral extraction may not present a problem provided new habitat is created by the process for colonisation. However, absence of I. pumilio from sites with past records is likely to be due to succession, scrub encroachment and over-shading, and lack of management to prevent these processes will be detrimental.

Acknowledgments

This work was funded by NERC and the Environment Agency. We thank field assistants Sara Ball, Phil Corney, Chris Gamble, Justine Saelens and Helen Turner. We are also grateful to Steve Jones, Dave Smallshire and Lesley Kerry for advice on sites. Robby Stoks, Mike Begon and two anonymous referees made valuable comments on this work.

Copyright information

© Springer Science+Business Media B.V. 2010