Abstract The arboreal ant fauna was investigated in Budongo Forest, a seasonal rain forest in Uganda, usingthe insecticidal fogging technique. Ants were collectedfrom 61 trees, between 7 and 33 m in height, belongingto four tree species. Trees were growing in adjacent plotsof forests characterized by different use and structure: anold primary forest, a primary swamp forest along a smallriver, and a secondary forest where selective logging wascarried out for 30 years. A total number of 37,065 ants,belonging to 161 species in 30 genera were collected.Considering the high number of species found only once,the completeness of the canopy ant fauna was relativelyhigh and of relatively similar magnitude as samples fromthe Neotropics or the Oriental region. Up to 37 ant species on a single tree, with an average of 18.2 speciesper tree, were found. Forty-four ant species (28.1%)were found only once, less than ten individuals werefound for each of 88 species (54.7%), but 64.0% of allindividuals belonged to one of five species. Consideringthe high numerical dominance of a few ant species like a Pheidole sp., Tetramorium aculeatum (Mayr) and aCrematogaster sp., there is some evidence for an ant mo-saic in the lower canopy of the Budongo Forest. Individ-ual numbers of ants were strongly correlated with nestsin the fogged tree, though the ants were not homoge-neously distributed in the tree crowns. Diversity mea-sures that strongly depend on individual numbers such asthe Morisita-Horn index or rarefaction methods werecalculated, but results were not concordant with those of incidence-based estimates such as jack-knife calcula-tions. Differences in ant species richness and faunalcomposition between tree species were low, but moresignificant between forest types. The ant fauna in the

secondary forest was less diverse with 12.6% fewer spe-cies compared to the primary forest sites. The averagenumber of ant species per tree was significantly lower inthe secondary forest (<20% of the species; F=8.03,df=59, P<0.01) than in the undisturbed forest types. Cataulacus, Leptothorax, Tetraponera, and Polyrhachis,which are typical canopy-dwelling ant genera, had a significantly higher diversity and frequency in the twoprimary forest types (F=4.17, df=53, P<0.05). Second-ary forest trees are often younger, lacking dead branchesand epiphytes which are important requisites for ant colonization on trees.

Keywords Arboreal ants · Distribution pattern · Speciesrichness · Tropical forest · Biodiversity

Introduction

Tropical rain forests harbour the most species rich animal communities on earth (Erwin and Scott 1980;Stork 1988). Their structure and dynamics are poorly understood and the ongoing destruction of tropical forests highlights the urgent need for more research (cf. Terborgh 1993). The canopies of these forests arerecognized as habitats of enormous species richness. Theeffective examination of tree crown arthropods was firstpossible when the insecticidal fogging technique was introduced into tropical ecology (Gagné 1979; Erwin andScott 1980).

In many studies, ants are found to be by far the mostabundant insect group in the canopy of tropical lowlandforests (Erwin 1983; Stork 1991; Wagner 1997). Oftenonly the number of individuals or the biomass of antshas been reported, but a few authors have also identifiedindividuals to species (e.g. Adis et al. 1985; Wilson1987; Majer 1990; Floren and Linsenmair 1997, 2000;Harada and Adis 1997). These studies were carried outin Central and South America, the Oriental region, orNorth Australia, while there are only few studies on canopy ants from Africa (cf. Lévieux 1978; Lawton et al.

T. Wagner (✉ )Universität Koblenz-Landau, Institut für Biologie, Universitätsstrasse 1, 56070 Koblenz, Germanye-mail: thwagner@uni-koblenz.deTel.: +49-261-2872231, Fax: +49-261-2872222

A. SchulzStaatliches Museum für Naturkunde, Erbprinzenstrasse 13,76133 Karlsruhe, Germany

Oecologia (2002) 133:224–232DOI 10.1007/s00442-002-1010-9

C O M M U N I T Y E C O L O G Y

Andreas Schulz · Thomas Wagner

Influence of forest type and tree species on canopy ants (Hymenoptera: Formicidae) in Budongo Forest, Uganda

Received: 29 November 2000 / Accepted: 16 February 2002 / Published online: 30 August 2002© Springer-Verlag 2002

225

1998; Dejean et al. 2000). Lawton et al. (1998) and Wattet al. (2002) described the ant diversity of secondary andstrongly degraded forests in Cameroon, but no primaryforest was studied.

The data presented here focus on the influence of treespecies and forest type on the canopy ant species rich-ness in adjacent plots of a primary forest, a primaryswamp forest, and a secondary forest. We also analysedthe data with respect to an ant mosaic, and differentmethods used to calculate species richness and diversityof arboreal ants.

Materials and methods

Study site

Research was carried out during the rainy season in June and July1995 in the Budongo Forest Reserve in Western Uganda aroundthe field station of the Budongo Forest Project (1 45′N, 31°35′E,1,200 m above sea level), a semideciduous part of the Guineo-Congolian Forest vegetation. This forest consists of selectivelylogged compartments and old primary forest areas within a contin-uous forest vegetation (Fig. 1) and has a well-known history(Eggeling 1947; Plumptre and Reynolds 1994). The mean annualrainfall is about 1,600 mm, with relatively constant precipitationfrom March to November, and a dry season from late December toFebruary, when about a quarter of the trees lose leaves. Dense for-

est covers an area of 428 km2, which is divided into compartmentsfor logging purposes. Ants were collected from four tree species,Cynometra alexandri C. H. Wright, Rinorea beniensis (Welwitschex Olivier) Kuntze, Teclea nobilis Delile, and Trichilia rubescensOlivier. The most common tree species is C. alexandri, whichforms up to 70% of the upper canopy trees.

Around the field station (compartments N3, N4) a largeamount of timber (80 m3 ha–1) was selectively logged between1947 and 1952 (Plumptre and Reynolds 1994). This area is nowcovered with a secondary forest (Fig. 1), with greater gaps in theupper canopy, no distinct canopy stories, and a dense cover ofyoung shrubs and trees in the understorey. Pioneer trees frequentlygrow here. The primary forest is located about 4 km away fromthe research station, which has been protected as a nature reservesince the 1930s. The ironwood, C. alexandri, is one of the mostcommon tree species of the upper canopy between 30 and 40 m inheight. The vegetation of the lowest storey, between 10 and 12 m,is dense, with R. beniensis as one of the most frequent tree spe-cies. Because of the dense canopy strata, the ground is intensivelyshaded and lacks a shrub layer. The vegetation on the poorlydrained soils next to the river is distinct primary swamp forest.The canopy is much lower, more open, and less homogeneousthan in the primary forest on slightly higher elevations (Eggeling1947). Plant communities of primary and secondary forest com-partments are more similar to each other than each type to theswamp forest. Subsequently the terms secondary, primary andswamp forest were used, but note that the swamp forest is also pri-mary forest.

Selection of trees

Tree species selected for fogging (Fig. 2) were abundant in allcompartments. Tree species were not closely related taxonomicallyand differed in shape of leaves, growth, and size. Rinorea beniensis

Fig. 1 Study site with locations of primary, secondary and swampforest plots investigated

226

(Violaceae) has a dense canopy and many small leaves, Trichiliarubescens (Meliaceae) has an extensive canopy with relatively fewlarge leaves, Cynometra alexandri (Caesalpiniaceae) and Tecleanobilis (Rutaceae) are between these extremes. Because foggingwas performed from the ground, the selected trees were of smallstature in the understorey or middle canopy stratum of the sameheight. Fogging trees of this size is practicable from the ground,but with increasing height one needs extremely calm weather con-ditions, which are rare. Six of the 15 fogged trees of Cynometra,four in the primary and two in the secondary forest, were higher,and the tree crowns of these trees were fogged up to a height of23–35 m. Maximum tree height was about 50 m. The fogged treeswere neither flowering nor fruiting. The canopies of trees fromstrata above the selected tree were at least 5 m higher.

Arthropod collecting, sample analysis

Fogging was carried out between 9.30 a.m. and 3.30 p.m. Beforefogging, 16 funnel-shaped nylon sheets, 1 m2 each, were fixed un-der the tree. The square sheets were packed around the trunk cor-responding to the shape of the canopy. This sampling regime wasused for a maximum of standardization. Arthropods on branchesin larger tree canopies outside of the area of the sixteen 1-m2

sheets were not collected. Each tree was fogged for approximately4 min with a Motan Swingfog SN-50 and a solution of natural pyrethrum with 1% active ingredient in diesel oil. The arthropodswhich fell were collected for 90 min after fogging. In the laborato-ry the material was sorted into groups, counted and determined atspecies/morphospecies level. The whole collection of voucherspecimens was deposited in the Staatliches Museum für Natur-kunde Karlsruhe, Germany and additional voucher specimenswere deposited in the British Museum, London.

Diversity measures

For the comparison of ant species richness, EstimateS 5.0 (Colwell1997) was used to compute cumulative species curves of thepooled samples and the extrapolated incidence was based on first-order jack-knife calculations (Burnham and Overton 1978). All cal-culations were randomized 100 times. Additionally, we computedthe diversity index alpha of the logarithmic series (Magurran1988), which is less dependent on the number of individuals, in or-der to compare the species richness of ant faunas between singletrees. For the conformity of ant species of the samples we used the

widely known Soerensen index. Furthermore, Pielou evenness,Berger-Parker dominance, rarefaction methods, and the Morisita-Horn index as a beta diversity measure were calculated as individu-al dependent diversity measures. To test the correlation between thedependent Morisita-Horn index and the independent Soerensen index we use the non-parametric Mantel test with 1,000 permuta-tions.

Results

Species richness, abundance, frequency of the ant fauna

On the 61 trees fogged, 37,065 ants, which could be assigned to 161 species, were collected. Three specieswere found as winged females unaccompanied by workers. Cumulative curves and jack-knife data for the species observed indicate slightly asymptotic graphs(Fig. 3). The jack-knife estimator calculates an extrapo-lated maximum of 205 species, and the pooled-samplesalgorithm computed an additional three ant species forthe last five plotted trees of the species accumulationcurve (Fig. 3). The jack-knife algorithm hypotheticallyestimates 43 (27%) more ant species than recorded in thefogging samples. The 161 observed ant species repre-senting 81% of this hypothetical species pool was esti-mated with the jack-knife method. Considering the largenumber of species found only once, this estimate is rather high.

With an average of 29.1% of all individuals, and onaverage 608 ants per tree, ants were the most abundantarboreal arthropod group in Budongo Forest (Wagner2001). Between different collection units, the percent-age of ants in the whole arthropod canopy fauna rangedbetween 23.5% on Rinorea, secondary forest, to 38.6%on Trichilia, primary forest. Exceptionally high num-bers of individuals were observed on a secondary foresttree of Cynometra, where 2,376 ants (90% Pheidolesp. 51) could be found, and on a swamp forest tree of Rinorea, with 2,268 ants (73% of them Tetramoriumaculeatum). Also on these two trees, ants reached the highest proportion of up to 55.0% of all arthropods

Fig. 2 Numbers, size, and forest type of fogged trees

Fig. 3 Cumulative curve of ant species (S) observed and jack-knife curve of all trees

227

collected. The lowest ant abundance found was 56 indi-viduals on a small primary forest Cynometra tree, thelowest percentage of the canopy arthropod fauna wasfound to be 5.1% on a Rinorea tree in the primary forest.

Forty-four ant species (28.0%) were found only once(Fig. 4). Pheidole sp. 51 was the most frequent species,being found on 54 trees, followed by Technomyrmexsp. 01 fogged from 41 trees. Only one individual wasfound for each of 37 species (23.0%); less than ten indi-viduals were found for each of 88 species (54.7%). De-spite the large number of trees investigated of all foresttypes, the number of rare species was remarkably high.However, some species were extremely abundant onsome trees, namely Pheidole sp. 51 (27.8% of all ant in-dividuals), Crematogaster sp. 32 (13.3%), and Tetra-morium aculeatum (8.1%). Sixty-four percent of all antindividuals belonged to the five most abundant species.Only a small number of species and individuals of Tetra-ponera, a common arboreal ant genus of the Old WordTropics, was found.

Diversity structure, influence of tree species and forest type

Between seven and 37 ant species, with an average of18.2 species, were found per tree. Species richness anddiversity on single trees was highly variable, rangingfrom 2,376 ants belonging to 15 species to 336 ants and37 species. Observed species accumulation curves hadon average steeper slopes on Teclea in swamp forest, andon both Cynometra and Trichilia, in primary forest. Thehighest species richness in terms of collected species andalpha diversity (Table 1) was found on Teclea in theswamp forest (mean species number 26.9 species pertree; mean alpha 6.06), while on Cynometra in the sec-ondary forest, the lowest average species number wasfound (13.9 species per tree; alpha 2.96; Table 1). Diver-sity calculated by the jack-knife method was similarlyhigh in both undisturbed forest types, but lower in thesecondary forest (Fig. 5).

Species richness on the same tree species betweenforest types was higher on average in the primary forest,

Table 1 Average species no.,total no. of species and speciesrichness (alpha) for collectionunits and forest types

Tree Forest type No. ant species Ant species No. of Alpha per treespecies per tree (cumulative) trees mean (SD)

Cynometra Primary 20.6 78 8 4.33 (1.10)Cynometra Secondary 14.6 54 7 2.96 (0.93)Cynometra Primary and secondary 17.8 97 15Rinorea Primary 17.8 65 8 3.86 (0.75)Rinorea Secondary 16.4 56 7 3.40 (1.68)Rinorea Primary and secondary 17.2 77 15Trichilia Primary 24.9 82 8 5.17 (1.25)Trichilia Secondary 18.1 66 8 4.05 (1.15)Trichilia Primary and secondary 21.5 106 16Rinorea Swamp 18.3 66 8 3.77 (0.90)Teclea Swamp 26.9 83 7 6.06 (2.77)

Primary forest all trees 21.1 119 24Secondary forest all trees 16.5 104 22Swamp forest all trees 22.3 101 15

Fig. 4 Frequency of ant species. No. of ant species collected=161,no. of trees fogged=61

Fig. 5 Cumulative curve of ant species (observed and jack-knifeestimate) for each tree species in the primary, secondary, andswamp forests

228

but significantly so only for Cynometra (F=6.69, df=14,P<0.05) (Fig. 5). Comparisons of species numbers between trees of one forest type showed a significant dif-ference only between Teclea and Rinorea in the swampforest (F=5.37, df=14, P<0.05). Secondary forest treesharboured on average 21.8% fewer ant species than treesfrom the primary and swamp forest (F=8.03, df=59,P<0.01; Table 1, Fig. 6). On the other hand the absoluteant species richness is reduced by 12.6% in the second-ary forest. In comparison to the other tree species, Cyno-metra showed a larger variation in height. Higher(>19 m) trees of Cynometra had more ant species (n=6,mean 19.8 species), than smaller (<19 m) trees (n=9,mean 16.4 species), but the difference was not signifi-cant.

Sixty-four ant species (39.7%) could be found in allthree forest types, 28 species exclusively in the primaryforest (18.1%), 16 species in the secondary forest(10.7%), and 18 in the swamp forest (12.8%). The Soerensen index estimated between 27 and 57% confor-mity between collection units. The conformity of specieswas high, but some abundant species, which represent>10% of all individuals collected, were restricted to onlyone forest type (Table 2). Four Leptothorax species,which are typical canopy dwellers in Afrotropical moist

forests, were exclusively found in the two primary foresttypes. In Budongo Forest, four species of Tetraponerawere found in primary forest on 26 trees and three species of this genus in secondary forest on 12 trees.

Table 2 Distribution patterns of selected abundant ant species correlated with forest types and tree species

Species Frequency Primary forest Secondary forest Swamp forest Abundance

Cynometra Rinorea Trichilia Cynometra Rinorea Trichilia Rinorea Teclea

Crematogaster sp. 32 31 1,350 12 491 – 2 – 1,601 1,492 4,948Crematogaster sp. 35 14 1 509 117 – 2 – – 3 632Lepisiota sp. 09 27 220 7 25 2 7 13 119 86 479Axinidris sp. 03 29 6 1 4 28 6 116 10 13 184Crematogaster sp. 45 16 1 2 2 – 60 210 1 8 284Tetramorium brevispinosus 16 1 1 1 62 725 3 1 – 794Crematogaster sp. 48 10 – 1 – 325 621 41 – – 988

Table 3 Average Morisita-Horn indices (MHI) between collection units and Soerensen indices (SI) in percentage conformity of pooledant species of each combination. PF Primary forest, SF secondary forest, SwF swamp forest

Cynometra alexandri Trichilia rubescens Rinorea beniensis

PF SF PF SF PF SF SwF

C. alexandri SF MHI 0.06SI 36

T. rubescens PF MHI 0.07 0.31SI 47 35

T. rubescens SF MHI 0.27 0.17 0.24SI 36 35 40

R. beniensis PF MHI 0.06 0.25 0.38 0.21SI 47 38 50 46

R. beniensis SF MHI 0.03 0.30 0.30 0.18 0.23SI 41 38 39 49 57

R. beniensis SwF MHI 0.35 0.13 0.19 0.39 0.20 0.12SI 33 28 36 27 32 31

T. nobilis SwF MHI 0.18 0.22 0.39 0.26 0.28 0.22 0.27SI 49 33 45 45 47 43 48

Fig. 6 Cumulative curves of ant species observed for each collec-tion unit. SwF Swamp forest, PF primary forest, SF secondary forest

229

Furthermore, the four typical aboreal ant genera Catau-lacus, Leptothorax, Tetraponera and Polyrhachis weresignificantly more abundant in the two primary forest areas (F=4.17, df=53, P<0.05).

When data on species abundance were used for calcu-lating beta-diversity indices like in the Morisita-Horn index, where all ant species are included (Table 3), theant fauna on Cynometra, primary forest, turned out to bevery different from those of other collection units.

The Morisita-Horn index and Soerensen index (Table 3) were not significantly correlated (Mantel test:r=–0.02, g=–0.109, P>0.05). The differences in theMorisita-Horn matrix were mainly due to a single Pheidole species which was very abundant on Cynome-tra trees in the secondary forest (seven trees, 2,666 spec-imens), and nearly absent on the same tree species in theprimary forest (six trees, 11 specimens).

Discussion

Ant species richness of the Budongo Forest

The ant canopy fauna of Budongo Forest was effectivelyrecorded by canopy fogging. The jack-knife estimate isabout 30% higher than the number of observed species,mainly due to the high number of singletons, while moreabundant species could be collected in sufficient num-bers. Many of these singletons might not be typical arbo-real ants, but workers of predominantly ground-nestingants, which forage temporarily on the understorey treesfogged. With 61 tree crowns fogged, and 161 ant speciescollected, the diversity of the canopy ant fauna in theBudongo Forest is of relatively similar magnitude orslightly lower than that of samples from the Neotropicsor Oriental region (Wilson 1987; Harada and Adis 1997;Floren and Linsenmair 1997, 2000; Bolton 1998) (cf.

Table 4). In Cameroon, Lawton et al. (1998) also foggedsmall trees from the ground and described 96 ant speciesfrom 45 canopies.

Differences in ant species richness between the fourexamined tree species were comparatively low, but theinfluence of forest type was higher (Figs. 5, 6). Cumula-tive curves for observed species showed the highest spe-cies numbers for swamp and primary forest trees. Ant diversity was not strongly reduced by selective logging,because undisturbed forest was in close proximity. Wattet al. (in press) fogged approximately 45 trees of Termi-nalia ivorensis A. Chev, an introduced tree species in atimber plantation in the Mbalmayo Forest Reserve inCameroon and found 97 ant species. In Budongo Forestthe species richness was much higher, where 15–16 con-specific trees were colonized by the same number of antspecies (Cynometra: 97 species; Trichilia: 107 species).As in our study, one species of Technomyrmex wasfound as a very common ant species in Mbalmayo For-est. The availability and quality of nesting sites, food,and age of trees seem to be more crucial factors for antdistribution than forest structure in general (e.g. David-son 1988; Blüthgen et al. 2000a, 2000b). Dead wood, ahighly structured bark, and a dense cover with epiphyticplants have favourable abiotic and microclimatic condi-tions for species- and colony-rich ant communities.These structures are less available in the secondary for-est, where the trees are often younger and have fewerepiphytes. Consequently, smaller (younger) trees ofCynometra either in the primary or the secondary forestharboured on average fewer ant species than higher andolder trees. However, there are different habitat struc-tures between the forest types, since old swamp foresttrees were generally smaller and had a much densercover of mosses, lichens and other epiphytes which mayexplain higher species numbers in the swamp forest es-pecially on Teclea trees.

Table 4 Comparison of selected studies on tropical canopy ant diversity. Approx. Approximately, max. maximum

Author Site No. of No. of ant Individuals Commentstrees species

This study Budongo Forest, 61 161 37,065 30 genera, max. 37 ant species Uganda (16 genera) per tree, 29% ants of

all arthropodsAdis et al. (1985) Manaus region, Brazil 9 69 4,922Stork (1991) Brunei, Borneo 10 98 23,275Wilson (1987) Tambopata Reserve 14 135 >100,000 40 genera, max. 43 ant species

(Peru) (26 genera) per treeFloren and Linsenmair Kinabalu Park, 19 192 88,502 39 genera, max. 54 ant species (1997) Borneo, Malaysia per tree, 54% ants of all arthropodsHarada and Adis (1997) Adolpho Ducke Forest, 2 100 2,613 21 genera, 82 ant species on one

Brazil tree after four foggingsBolton (1998) Pasoh Forest, 6 146 Only large trees were fogged

West MalaysiaLawton et al. (1998) Mbalmayo Forest 30 96 Only trees of secondary forest

Cameroon, Africa were foggedWatt et al. (in press) Mbalmayo Forest Approx. 45 97 48,500

Cameroon, Africa

Another characteristic of disturbed tropical land-scapes is the existence of ant mosaics. In tropical treecrop plantations, only a few ant species occur in the can-opy (e.g. Room 1971; Leston 1973; Majer 1993; Dejeanet al. 1994; Dejean and Gibernau 2000), where a smallnumber of aggressive species defend large territoriesforming a three-dimensional mosaic and thereby reduc-ing ant diversity partially. Ant mosaics are not found inthe lower canopy of Asian pristine tropical forest (e.g.Floren and Linsenmair 2000) but can be expected inmore degraded forests or in the higher canopy of pristineforests (Dejean and Corbara, in press). In the BudongoForest, one Crematogaster species was recorded with ahigher abundance on five trees in the secondary forest.On two very high trees this species occurs together withonly seven and eight other ant species, but on severallower trees this species was found together with numer-ous other species. Dejean and Corbara (in press) re-corded Tetramorium aculeatum as a typical dominant antspecies. We found this species with high abundance onRinorea trees in the swamp forest where the canopy ismuch lower, more open, and less homogeneous than inthe other forest types. This distribution pattern givessome evidence for an ant mosaic, but a clear character-ization cannot be given on the basis of canopy foggingdata (see below).

Tetraponera which have frequently been found intropical forests (Ward, personal communication) are notparticularly common in the Budongo Forest. It seems thatspecies of Tetraponera and the related neotropical genusPseudomyrmex occur at a higher frequency in secondaryforest (Dejean et al. 1994; Delabie et al. 1998; Dejeanand Gibernau 2000) than in primary forests. However, thearboreal Oecophylla longinoda Latreille, a common antin plantations and shrublands and often part of ant mosa-ics (Majer 1978; Dejean and Corbara in press) was notfound in Budongo Forest. Watt et al. (2002) recorded O. longinoda as the fifth most common ant species of theMbalmayo Forest Reserve in Cameroon. The absence ofthis species in Budongo Forest is remarkable, but at themoment not explicable.

Methods comparing canopy ant diversity

With the canopy fogging method only a snapshot of theindividual activity of ants foraging on the tree surface ispossible, and it does not allow complete quantitativesampling. Therefore, diversity indices, like evenness in-dices or rarefaction methods, which are strongly depen-dent on the number of individuals, were found to be lessuseful for comparing the diversity structure (Hughes1986; Romero and Jaffe 1989; Blüthgen et al. 2000b).However, the results of diversity indices which dependon numbers of individuals, like the Morisita-Horn index,were not concordant with those of methods based onspecies numbers only.

Canopy fogging allows for the assessment of a part ofthe canopy ant fauna, but only arthropods on the tree sur-

face can be collected quantitatively (Adis et al. 1985).Litter species can live in the humus of the forest floor, aswell as in humus-filled depressions in the canopy and inepiphytes. Longino and Nadkarni (1990) found 21 antspecies only in canopy leaf litter in a montane cloud for-est in Costa Rica. For the Budongo Forest, species of thegenera Anochetus, Pachycondyla, Centromyrmex, Cre-matogaster (Part), Oligomyrmex, Melissotarsus, Paedal-gus, Leptothorax (Part), Pyramica and Tetramorium(Partim) presumably nest in the canopy leaf litter or sus-pended soils, or in part in the bark of the trees. Especial-ly individuals of such species are likely to be proportion-ally under-represented in canopy fogging samples, asmany litter ant species are cryptic or leave their nest onlyat night. Pachycondyla, Odontomachus, Dorylus, andPheidologethon are taxa whose nests are mainly in theground, but frequently enter trees to forage, sometimesin large numbers. This might lead to strongly biased esti-mates of species abundance, but these species do certain-ly influence the arboreal community and must thereforebe taken into account.

The behaviour of ants in response to the foggingtreatment may also have a significant effect on the num-ber of individuals observed. For example, Orr et al.(1996) and Pfeiffer (1997) described a low number(16%) of foraging workers of the well-investigated antCamponotus gigas (Latreille) during their nocturnal ac-tivity phase. After disturbance, foragers of some Cam-ponotus species retreated into the nest and did not startattacking a possible enemy (Schulz, unpublished data).On the other hand, aggressive ants like some Cremato-gaster and Pheidole species could leave their nests inthousands during and after fogging. These species werefound in extremely high abundance on single trees, aswas shown for Pheidole in our sampling.

Generally, there is no doubt that the abundance ofspecies is an important part of community structures, butfor ants it is risky to compare individual numbers with-out examining nest densities (cf. Hughes 1986; Romeroand Jaffe 1989). Diversity indices which are strongly in-fluenced by the numbers of individuals, like evennessand dominance indices (Pielou-evenness, Berger-Parker-dominance, rarefaction-methods), are not useful for thedescription of canopy-ant faunas. Beta-diversity mea-surements which include abundance, such as the Mori-sita-Horn index, can also be misleading. For example theMorisita-Horn indices revealed a very low faunal over-lap in the primary forest between Cynometra trees andTrichilia and Rinorea trees, and between Cynometra andRinorea trees in the secondary forest with a faunal over-lap of between 3% and 7% only, whilst faunal overlapwith other collection units was much higher (18–35%;Table 3). This was caused by a single species of Phei-dole, the most abundant ant species recorded in the Budongo Forest (see Results), which nests favourably inthe rough bark of trees, particularly in larger canopybranches. It was surprisingly rarely found on Cynometratrees in the primary forest. The Morisita-Horn indicesalone suggest strong differences between canopy ant di-

230

Delabie JHC, DoNascimento IC, DosMariano SF (1998) Impor-tance de l’agriculture cacaoyère pour le maintien de la biodiv-ersité: étude comparée de la myrmécofaune de différentes mi-lieux du sud-est de Bahia, Brésil (Hymenoptera: Formicidae).In: Proceedings of the XIIth International Cocoa ResearchConference, Salvador, Bahia, Brazil. pp 75–83

Eggeling WJ (1947) Observations on the ecology of the Budongorain forest, Uganda. J Ecol 34:20–87

Erwin TL (1983) Beetles and other insects of tropical forest canopies at Manaus, Brazil, sampled by insecticidal fogging.In: Sutton SL, Whitmore TC, Chadwick AC (eds) Tropicalrain forest: ecology and management. Blackwell Scientific,Oxford, pp 59–75

Erwin TL, Scott JC (1980) Seasonal and size patterns, trophicstructure, and richness of Coleoptera in the tropical arborealecosystem: the fauna of the tree Luehea seemannii Triana andPlanch in the Canal Zone of Panama. Coleopt Bull 34:305–322

Floren A, Linsenmair KE (1997) Diversity and recolonization dynamics of selected arthropod groups on different tree spe-cies in a lowland rainforest in Sabah, Malaysia with specialreference to Formicidae. In: Stork NE, Adis J, Didham RK(eds) Canopy arthropods. Chapman and Hall, New York,pp 344–381

Floren A, Linsenmair KE (2000) Do ant mosaics exist in pristinelowland rain forests? Oecologia 123:129–137

Gagné WC (1979) Canopy-associated arthropods in Acacia koaand Metrosideros tree communities along an altitudinal tran-sect on Hawaii Island. Pac Ins 21:56–82

Harada AY, Adis J (1997) The ant fauna of tree canopies in Central Amazonia: a first assessment. In: Stork NE, Adis J,Didham RK (eds) Canopy Arthropods. Chapman and Hall,New York, pp 382–400

Hughes RG (1986) Theories and models of species abundance.Am Nat 128:879–899

Lawton JH, Bignell DE, Bolton B, Bloemers GF, Eggelton P, Hammond PM, Hodda M, Holt RD, Larsen TB, Mawdsley NA,Stork NE, Srivastava DS, Watt AD (1998) Biodiversity inven-tories, indicator taxa and effects of habitat modification ontropical forest. Nature 391(1):72–76

Leston D (1973) The ant mosaic – tropical tree crops and the limiting of pests and diseases. Pest Artic News Summ19:311–341

Lévieux J (1978) La structure du nid de quelques fourmis arbori-coles d’Afrique tropicale (Hymenoptera, Formicidae). AnnUniv Abidjan Ser C 7:5–22

Longino JT, Nadkarni NM (1990) A comparison of ground andcanopy leaf litter ants (Hymenoptera: Formicidae) in a Neo-tropical montane forest. Psyche 97:81–94

Magurran AE (1988) Ecological diversity and its measurements.Croom Helm, London

Majer JD (1978) The influence of blanket and selective sprayingon ant distribution in a West African cocoa farm Oecophyllalonginoda, Macromischoides aculeatus, Crematogaster spp.beneficial ants. Rev Theobroma 8:87–93

Majer JD (1990) The abundance and diversity of arboreal ants inNorthern Australia. Biotropica 22:191–199

Majer JD (1993) Comparison of the arboreal ant mosaic in Ghana,Brazil, Papua New Guinea and Australia – its structure and influence on arthropod diversity. In: LaSalle J, Gauld ID (eds)Hymenoptera and biodiversity. CAB International, Walling-ford, pp 115–141

Orr AG, Charles JK, Yahya HR, Sharebini HN (1996) Nesting andcolony structure in the gaint forest ant Camponotus gigas(Latreille) (Hymenoptera: Formicidae). Raffles Bull Zool44:247–251

Pfeiffer M (1997) Sozialstruktur und Verhaltensökologie von Riesenameisen Camponotus gigas Latreille 1802 im Regen-wald Malaysias auf Borneo. Wissenschaft und Technik, Berlin

Plumptre AJ, Reynolds V (1994) The effect of selective loggingon the primate populations in the Budongo Forest Reserve,Uganda. J Appl Ecol 31:631–641

231

versity in the primary forest, but a comparison on the ba-sis of species composition, observed species number,Soerensen and diversity alpha index showed a muchhigher similarity between tree species in one forest type(Table 1, Fig. 6). It is not clear why this Pheidole specieswas found to be extremely abundant on Cynometra treesin the secondary forest, and its absence on primary foresttrees may have been by chance. Therefore, we recom-mend excluding strongly individual-dependent indicesfor the description of ant diversity, when the studied antnests and individuals are not subject to intensive obser-vation.

Acknowledgements Thomas Wagner is grateful to Andy Plumptre,Chris Bakuneeta and Vernon Reynolds for the possibility to work inthe Sonso Station of the Budongo Forest Project, to the Uganda National Council for Science and Technology, and the Uganda Forest Department for acceptance and permission to carry out thefield study. Furthermore, we thank the National Geographic Society,Conservation International (Margot Marsh Fund), the Departmentfor International Development, and the Norwegian Agency for Development for supporting the Budongo Forest Project. We aregrateful to Alan Dejean for his valuable comments of an earlierdraft of this paper, and to his wife Andrea who kindly improved theEnglish. This research was supported by the Stifterverband für dieDeutsche Wissenschaft and the Deutsche Forschungsgemeinschaft.

References

Adis J, Lubin YD, Montgomery GG (1985) Arthropods from thecanopy of inundated and terra firme forest near Manaus, Brazil, with critical considerations on the pyrethrum-foggingtechnique. Stud Neotrop Fauna Environ 19:223–236

Blüthgen N, Verhaagh M, Goitía W, Blüthgen N (2000a) Ant nestin tank bromeliads – an example of non-specific interaction.Insect Soc 47:313–316

Blüthgen N, Verhaagh M, Goitía W, Jaffé K, Morawetz W, Barthlott W (2000b) How plants shape the ant community inthe Amazonian rainforest canopy: the key role of extrafloralnectaries and homopteran honeydew. Oecologia 125:229–240

Bolton B (1998) A preliminary analysis of the ants (Formicidae)of Pasoh Forest Reserve. In: Lee SS, Dan YM, Gauld ID,Bishop I (eds) Conservation, management and development offorest resources. pp 84–95

Burnham KP, Overton WS (1978) Estimation of the size of aclosed population when capture probabilities vary among animals. Biometrika 65:623–633

Colwell RK (1997) EstimateS: statistical estimation of speciesrichness and shared species from samples, version 5. User’sguide and application published temporarily at: http://viceroy.eeb.uconn.edu/estimates

Davidson DW (1988) Ecological studies of neotropical ant gardens. Ecology 69:1138–1152

Dejean A, Corbara B (in press) Review on mosaics of dominantants in rain forests and plantations. In: Basset Y, Novotny V,Miller SE, Kitching RL (eds) Arthropods of tropical forests,spatio-temporal dynamics and resource use in the canopy.Cambrigde University Press, Cambridge

Dejean A, Gibernau M (2000) A rainforest ant mosaic: the edgeeffect (Hymenoptera: Formicidae). Sociobiology 35:385–401

Dejean A, Amougou A, Djieto-Lordon C, Lenoir A (1994) Mosaicant territories in an African secondary rain forest (Hymenoptera:Formicidae). Sociobiology 23:275–292

Dejean A, McKey D, Gibernau M, Belin M (2000) The arborealant mosaic in a Cameroonian rainforest (Hymenoptera: Formicidae). Sociobiology 35:403–423

Wagner T (1997) The beetle fauna of different tree species in forests of Rwanda and East Zaire. In: Stork NE, Adis J, Didham RK (eds) canopy arthropods. Chapman and Hall, New York, pp 169–183

Wagner T (2001) Seasonal changes in the canopy arthropod faunain Rinorea beniensis in Budongo Forest, Uganda. Plant Ecol153:169–178

Watt AD, Stork NE, Bolton B (2002) The diversity and abundanceof ants in relation to forest distribution and plantation establishmeat in Southern Cameroon. J Appl Ecol 39:18–30

Wilson EO (1987) The arboreal ant fauna of Peruvian Amazonforests: a first assessment. Biotropica 19:245–251

232

Romero H, Jaffe K (1989) A comparison of methods for samplingants (Hymenoptera, Formicidae) in savannas. Biotropica21:348–352

Room PM (1971) The relative distributions of ant species in Ghana’s cocoa farms. J Anim Ecol 40:735–751

Stork NE (1988) Insect diversity: facts fiction and speculation.Biol J Linn Soc 35:321–337

Stork NE (1991) The composition of the arthropod fauna of Bornean lowland rain forest trees. J Trop Ecol 7:161–180

Terborgh J (1993) Lebensraum Regenwald Zentrum der biologi-schen Vielfalt. Spektrum, Heidelberg