Using ants as bioindicators in land management: simplifying assessment of ant community responses potx

This intensive ant survey yielded 174 species from 24 genera, and revealed seven key patterns of ant community structure and composition in relation to habitat and SO2 levels.. Ant speci

Journal of Applied

Ecology 2002

39, 8–17

Using ants as bioindicators in land management:

simplifying assessment of ant community responses

ALAN N ANDERSEN*, BENJAMIN D HOFFMANN*, WARREN J MÜLLER† AND ANTHONY D GRIFFITHS*‡

*Tropical Savannas Cooperative Research Centre, Division of Sustainable Ecosystems, CSIRO Tropical Ecosystems Research Centre, PMB 44 Winnellie, NT 0822, Australia; and †CSIRO Mathematical and Information Sciences, GPO Box 664, Canberra, ACT 2601, Australia

Summary

1 The indicator qualities of terrestrial invertebrates are widely recognized in the

context of detecting ecological change associated with human land-use However, the use of terrestrial invertebrates as bioindicators remains more a topic of scientific discourse than a part of land-management practice, largely because their inordinate numbers, taxonomic challenges and general unfamiliarity make invertebrates too intimidating for most land-management agencies Terrestrial invertebrates will not be widely adopted as bioindicators in land management until simple and efficient protocols have been developed that meet the needs of land managers.

2 In Australia, ants are one group of terrestrial insects that has been commonly

adopted as bioindicators in land management, and this study examined the reliability

of a simplified ant assessment protocol designed to be within the capacity of a wide range of land managers.

3 Ants had previously been surveyed intensively as part of a comprehensive assessment

of biodiversity responses to SO2 emissions from a large copper and lead smelter at Mt Isa in the Australian semi-arid tropics This intensive ant survey yielded 174 species from

24 genera, and revealed seven key patterns of ant community structure and composition in relation to habitat and SO2 levels.

4 We tested the extent to which a greatly simplified ant assessment was able to

reproduce these results Our simplified assessment was based on ant ‘bycatch’ from bucket-sized (20-litre) pitfall traps used to sample vertebrates as part of the broader biodiversity survey We also greatly simplified the sorting of ant morphospecies by considering only large (using a threshold of 4 mm) species, and we reduced sorting time

by considering only the presence or absence of species at each site In this manner, the inclusion of ants in the assessment process required less than 10% of the effort demanded by the intensive ant survey.

5 Our simplified protocol reproduced virtually all the key findings of the intensive

survey This puts effective ant monitoring within the capacity of a wide range of land managers.

Key-words: environmental assessment, land-use impacts, monitoring, sampling

protocols, SO2

Journal of Applied Ecology (2002) 39, 8– 17

© 2002 British

Ecological Society

Correspondence: Alan Andersen, Tropical Savannas Cooper-ative Research Centre, Division of Sustainable Ecosystems, CSIRO Tropical Ecosystems Research Centre, PMB 44 Win- nellie, NT 0822, Australia (e-mail

Alan.Andersen@csi r o.au)

‡Present address: Key Centre for Tropical Wildlife Manage-ment, Northern Territory University, Darwin, NT 0909, Australia

Introduction

The indicator qualities of terrestrial invertebrates are widely recognized in the context of detecting ecological change associated with human land use (Rosenberg, Danks & Lehmkuhl 1986) This contrasts with the use

of particular invertebrate groups as indicators of general

Simplified ant

assessment

© 2002 British

Ecological Society,

Journal of Applied

Ecology, 39,

8–17

diversity patterns (Pearson & Cassola 1992; Kremen

1994), which has been widely disputed (Lawton et al.

1998; Kotze & Samways 1999) Invertebrates make good indic- ators of ecological condition because they are highly diverse and functionally important, can integrate a variety of ecological processes, are sensitive

to environ- mental change, and are easily sampled (Greenslade & Greenslade 1984; Brown 1997;

McGeoch 1998)

Despite these qualities, however, the use of terrestrial invertebrates as bioindicators remains more

a topic of scientific discourse than a part of land-management practice With few exceptions, invertebrates are rou- tinely ignored in land monitoring and assessment pro- grammes, largely because their inordinate numbers, taxonomic challenges and general unfamiliarity are too intimidating for most land-management agencies (New 1996) This contrasts with the situation in aquatic systems, where relatively simple protocols for assessing macroinvertebrates have been widely applied in studies of river health (Hellawell 1978;

Norris & Norris 1995) Terrestrial insects and other inverte- brates will not be widely adopted as bioindicators in land management until simple and efficient protocols have been developed that meet the needs of land man- agers (Andersen 1999)

In Australia, ants are one group of terrestrial insects that has been commonly adopted as bioindicators in land management (Majer 1983; Andersen 1997a) In particular, ants have frequently been used by the min- ing industry as indicators of restoration success (Majer

1984; Andersen 1997b) Ant species richness and com-position show predictable colonization patterns at mine sites undergoing rehabilitation (Andersen 1993;

Majer & Nichols 1998; Bisevac & Majer 1999), with these patterns reflecting those of other invertebrate groups (Majer 1983; Andersen 1997b) and key eco-system processes (Andersen & Sparling 1997) More recently, ants have been used as indicators of off-site mining impacts (Read 1996; Hoffmann, Griffiths &

Andersen 2000) and for other land uses such as forestry (York 1994; Vanderwoude, Andersen &

House 1997) and pastoralism (Landsberg, Morton &

James 1999; Hoffmann 2000; Read & Andersen 2000) However, in virtually all these cases ant surveys have involved spe- cialist entomologists, and comprehensive ant surveys remain largely beyond the capacity of most environ- mental practitioners

In any sampling programme there will inevitably be

a trade-off between simplicity on one hand, and reli-ability on the other When endeavouring to make insect surveys more accessible to land managers, there

is no point in developing simplified sampling protocols if reliability is seriously compromised This study examined the reliability of a simplified ant assessment protocol designed to be within the capacity

of a wide range of land managers

The study was conducted as part of a comprehensive assessment of biodiversity responses to SO2 emissions from a large copper and lead smelter at Mt Isa in the

Table 1 Key findings of a comprehensive ant sampling

programme conducted as part of an assessment of the

(Hoffmann, Griffiths & Andersen 2000)

1 The two regionally dominant landforms (rocky ridges and alluvial plains) supported distinct ant faunas

2 Ant abundance declined with increasing levels of SO2

3 Species richness declined with increasing levels of SO2

4 Species composition varied systematically with increasing levels of SO2

5 Several common species showed clear abundance

others increasing

6 Ant functional group composition (sensu Andersen 1995)

showed relatively little change in relation to SO2

7 Ant responses varied according to biogeographical affinity, with the overall abundance of Eyrean (arid-adapted) taxa increasing in relation to SO2, Torresian (tropical) taxa decreasing, and widespread taxa showing no change

Australian semi-arid tropics Vegetation had been dra-matically affected immediately downwind from the smelter, and the influence of the smelter could be detected for at least 15 km (Griffiths 1998) A key ques-tion was whether or not faunal biodiversity was sim-ilarly affected Routine vertebrate sampling indicated that bird and reptile assemblages were significantly influ- enced by high SO2 levels (up to 5 km from the smelter), with mammals providing too few records for statistical treatment (Griffiths 1998) A survey of ants,

on the other hand, revealed an effect of the smelter for

up to 35 km (Hoffmann, Griffiths & Andersen 2000) Ants were therefore a far more sensitive indicator of the effects of the smelter on faunal integrity than vertebrates

The ant survey was based on catches in small (4·5 cm) pitfall traps that were partly filled with pre-servative, which is the most widely used technique for obtaining quantitative assessments of ant communities

in open habitats (Andersen 1991; Bestelmeyer & Wiens 1996; Fisher 1999) The survey yielded 174 species from

24 genera, and revealed seven key patterns of ant com-munity structure and composition in relation to habi-tat and SO2 (Table 1) Our simplified protocol used the ant ‘bycatch’ from the bucket-sized pitfall traps used to capture vertebrates, which are routinely used in wildlife surveys We tested the extent to which our simplified ant assessment was sufficient to reproduce the seven key results of the intensive ant sampling programme

Methods

 

Mt Isa is located in north-western Queensland (2943S

13927E), Australia, with its 400-mm average annual rainfall being heavily concentrated within a summer wet season Temperatures are high year-round, with mean monthly maxima ranging from about 25 C

(July) to 38 C (November), and minima 10

C to

A.N Andersen

et al

© 2002 British

Ecological Society,

Journal of Applied

Ecology, 39,

8–17

24 C (Bureau of Meteorology, Mt Isa) The major landforms in the region are erosional Tertiary surfaces with skeletal soils, rock (sandstone, shale and quart-zite) outcrops and alluvial plains The predominant vegetation is low open woodland dominated by

species of Eucalyptus, Acacia and Atalaya, with the ground- layer dominated by hummock grass (Triodia

spp.) on skeletal soils and tussock grasses (species of

Aristida and Cenchrus) on alluvial loams.

  

Biodiversity sampling at Mt Isa was stratified accord-ing to four SO2 zones arranged along the direction of prevailing winds: background (soil sulphate levels 1– 5 p.p.m., 15 – 30 km upwind from the smelter), low

(10 – 30 p.p.m., 30 – 35 km downwind), medium (30 –

70 p.p.m., 7 –15 km downwind) and high (70 –120 p.p.m.,

3 – 5 km downwind) (Griffiths 1998) The intensive ant survey was conducted at 40 sites, comprising four rocky ridge and four alluvial plain sites at each of the low, medium and high SO2 zones, and eight sites from each habitat in the background zone (Hoffmann, Griffiths & Andersen 2000) Ants were sampled using 4·5-cm diameter pitfall traps partly filled with ethylene glycol as a preservative A 5  3 grid of traps with 10-m spac- ing was established at each site, and operated for 48 h All ants falling into traps were sorted to species, iden- tified and enumerated

Vertebrate pitfall traps (20-litre buckets, 28-cm diameter) were installed at 14, 14, 21 and 16 sites, respectively, within the background, low, medium and high SO2 zones, distributed amongst the three major habitat types as follows: rocky ridges, 23 sites; rocky plains, 20 sites; alluvial plains, 22 sites (Table 2)

These sites included all those used in the intensive ant survey, except for eight from the background zone

At each site, four pitfall traps (28-cm diameter plastic buckets), each with 10 m of drift fencing (height 30 cm), were randomly located within a 100  100-m plot Our sim- plified protocol therefore considered ants from 220 (large) traps distributed across 65 sites, compared with 600 (small) traps from 40 sites for the intensive ant survey Ants were collected from vertebrate traps early in the morning and late in the afternoon during 5 consecutive days between late October and early December 1997 This was done within 2 weeks of the intensive ant survey In addition to simplified sampling (i.e using the ant bycatch from routine vertebrate traps), we also greatly simplified the sorting of specimens in the laboratory by considering only large species A 4-mm total length threshold was used to

designate genera and species groups to be

considered The taxa we considered were Anochetus,

Bothroponera, Leptogenys, Odontomachus, Rhytidoponera, Calomyrmex, Camponotus, Opisthopsis and Polyrhachis (i.e all species within these genera),

as well as the diversus species group of Meranoplus, the mayri and purpureus groups of Iridomyrmex, and

the bagoti and aeneovirens groups of Melophorus

follows

Andersen 2000)

The use of

higher-level taxa

rather than a

strict size

criterion avoided

potential

confusion caused

by polymorphic

or otherwise

variably sized

species in which

some speci- mens

fall below but

others are above

the threshold

Sorting time was

further reduced

by considering

only the presence

or absence of

species at each

site, rather than

their abundance

quantify it, but

esti- mated that

our simplified

protocol required

less than

10% of the

laboratory time

This was despite

the sim- plified

protocol

covering more

sites

















 was used to

analyse

abundance data

in the inten- sive

(Hoffmann,

Griffiths &

Andersen 2000),

alternative

analytical

strategy was

needed for the

species by site presence/absence matrix from our simplified protocol Each species had a binary response at each site, producing a number of occupied sites out of a total possible number of sites for each zone  habitat combination Our analytical strategy was to fit generalized linear models (GLM) with bino- mial error and logit link (Dobson 1990) These ana- lyses give rise to analysis

of deviance tables When the residual deviance is greater than one, extra-binomial variation may be

present and the deviance ratios are approximate

F-ratios and can be tested for significance in a manner

similar to F-ratios in  When the residual deviance is less than or equal to one, binomial variation is assumed and testing for significance

is based on the deviance for each term being approxim- ately distributed as a chi-squared variate The outputs of such analyses are lists of factors with their deviances or deviance ratios, and their level of significance For significant terms we provided adjusted means and standard errors where appropriate The means were the average proportions of sites occupied by individual species for all species in the model

Three sets of analyses were performed, fitting indi- vidual species, functional groups (Andersen 1995) and groups based on biogeographical affinities (Andersen

2000) as factors, respectively Five functional groups were common enough for analysis: Dominant Dolichoderinae (species of Iridomyrmex),

Subordinate Camponotini (primarily species of

Camponotus and Polyrhachis), Hot Climate

Specialists (species of Melophorus and

Meranoplus), Opportunists (primarily species of Rhyti- doponera) and Specialist Predators (primarily species of Bothroponera and Leptogenys) Similarly,

three bio- geographical groups, Eyrean, Torresian and widespread, were common enough for analysis Analyses were also conducted on the eight individual species that were recorded from at least

15 sites

Patterns of species composition in relation to habitat and SO2 were explored by multivariate analysis, as was the case for results from the intensive survey (Hoffmann, Griffiths & Andersen 2000) Site species data were combined within habitat types, and a similarity

Table 2 Records of ant species across SO2 zones and habitat type (RR, rocky ridge; RP, rocky plain; AP, alluvial plain) Data are numbers of sites at which species were recorded Species codes follow Hoffmann, Griffiths & Andersen (2000) (those not recorded by Hoffmann, Griffiths & Andersen 2000 are indicated by an asterisk) The functional group (FG; DD, Dominant Dolichoderinae; HCS, Hot Climate Specialist; OPP, Opportunist; SC, Subordinate Camponotini; SP, Specialist Predator) and biogeographical affinity (BIOG; E, Eyrean; T, Torresian; W, widespread) of each species is also given

SO2 zone

Habitat

RR 7

RP 3

AP 6

RR 6

RP 8

AP 7

RR 5

RP 5

AP 4

RR 5

RP 4

AP 5

RR 23

RP 20

AP 22

12

A.N Andersen

et al

© 2002 British

Ecological Society,

Journal of Applied

Ecology, 39,

8–17

matrix of habitat type/SO2 zone combinations was con- structed using the Jaccard index, based on presence/ absence data for all species Habitat type/

SO2 zone combinations were then ordinated using semi-strong hybrid multidimensional scaling (SSH option of the PATN software package; Belbin 1994)

Results

   

A total of 41 species from 12 genera was recorded by

our simplified protocol, with Camponotus (14 species) and Polyrhachis (eight) collectively contributing half of

all species (Table 2) Twelve (29%) species were not among the 174 species recorded during the intensive ant survey of Hoffmann, Griffiths & Andersen (2000) This high proportion can be explained by the greater efficiency of drift fences in capturing uncommon

species We did not consider abundance per se,

because we considered only presence/absence data However, the occurrence of species across sites can

be used as an abundance surrogate Considering all species, mean occurrence decreased significantly

(deviance = 9·63, d.f = 3, P = 0·02) with increasing

levels of SO2, and mean site species richness showed

a similar pattern, although this was not quite

statistically significant (P = 0·07, one-way ; Fig 1) Individual species exhibited a wide range of responses in relation to SO2 (Fig 2) Three of the eight most common species showed statistically significant responses (Table 3), with the most marked

being shown by Camponotus sp A (denticulatus

group), which occurred primarily at higher SO2 zones (Fig 2d) Four of these eight species showed significant habitat effects (Table 3) The most marked

was for Camponotus sp D (novaehollandiae group),

which occurred primarily at rocky ridge sites and was not recorded at all in alluvial plain habitat (Table 2)

The functional group–zone interaction was not

sig-nificant (deviance ratio = 1·51, d.f = 12,396, P > 0·05),

indicating that functional group composition was rel-atively uniform across SO2 zones There was a tendency for specialist predators to favour low levels of SO2 (the nine records occurred exclusively in the low or back- ground zones; Table 2), but their numbers were too low to influence the overall analysis In contrast, the bio- geographical group–zone interaction was highly sig- nificant (deviance ratio = 4·57, d.f = 6,418,

P < 0·001) The mean occurrence of widespread

species was relat- ively constant, but the occurrence of Torresian species declined systematically in relation to

SO2, and Eyrean taxa showed a curvilinear response (Fig 3)

 

Multivariate ordination showed strong separation of habitats according to species composition, with rocky ridge and alluvial plain habitats at the two extremes,

13

Simplified ant

assessment

0·25 0·2 0·15 0·1 0·05

Occurrence Richness

8 6 4 2

Fig 1 Mean (+ SE) site occurrence and species richness of species at high (H), medium (M), low (L) and background (B) SO2

zones A mean occurrence of 0·1 means that on average each species occupied 10% of sites Only those species occurring at >

2 sites (n = 26) were considered for occurrence.

0·4 0·3 0·2 0·1 0 (a) Rhytidoponera sp nr cornuta

0·8 0·6 0·4 0·2 0 (b) Rhytidoponera sp nr rufithorax

0·4 0·3 0·2 0·1 0 (c) Iridomyrmex sp nr mayri

0·5 0·4 0·3 0·2 0·1 0 (d) Camponotus sp A (denticulatus gp)

0·8 0·6 0·4 0·2 0 (e) Camponotus fieldae

0·4 0·3 0·2 0·1 0 (f) Camponotus sp D (novaehollandiae gp)

0·5 0·4 0·2 0·1 0 (f) Camponotus sp F (subnitidus gp)

0·8 0·6 0·4 0·2 0

(g) Camponotus sp Q (discors gp)

Fig 2 Occurrence (proportion of sites occupied) of common species at high (H), medium (M), low (L) and background (B) SO2

zones

Table 3 Results of GLM tests of the effects of habitat type and SO2 zone on the occurrence of the eight most common species

(those recorded from at least 15 sites) Data are P-values, with significant values indicated in bold

© 2002 British

Ecological Society,

Journal of Applied

Ecology, 39,

8–17

and rocky plains intermediate (Fig 4) This was sup-ported by a highly significant species–habitat

inter-action in GLM (deviance = 174·84, d.f = 80, P <

0·001) Similarly, SO2 zones were also clearly evident

in ordination space (Fig 4), and this was supported

by a highly significant species–zone interaction in

GLM (deviance = 222·18, d.f = 120, P < 0·001).

Discussion

   

The intensive ant survey at Mt Isa, yielding 174 species, documented seven clear responses of ant communities

to variation in habitat and SO2 (Table 1) Our simplified

14

A.N Andersen

et al

0·25 0·2 0·15

Eyrean Torresian