1. Termites play a key role in decomposition processes in tropical ecosystems. Rapid assessment of local termite assemblages requires a standardized sampling protocol capable of producing an accurate picture of species composition. This paper evaluates the efficacy of a sampling protocol designed to assess termite species richness and functional diversity in tropical forests.
2. The protocol entails a 100 m long transect consisting of 20 sections (each 5 × 2 m). One hour of human sampling effort per section is used to search for and collect termites from dead wood, soil, termite nests and other microhabitats up to a height of 2 m above ground. The protocol was tested in three forest sites where the local termite fauna was already comprehensively documented. Two transects were run at Danum Valley (Sabah, Borneo), one at Pasoh Forest Reserve (Peninsular Malaysia) and one at Mbalmayo Forest Reserve (Cameroon).
3. At the three sites the transect samples contained 31–36% of the known local termite species pool. The taxonomic and functional group composition of the transect samples did not differ significantly from that of the known local fauna. The two transects run at Danum Valley gave very similar patterns, suggesting that the protocol produces consistent within-site results. After sampling 20 sections, pseudoturnover between the two Danum transects had declined to a relatively low level.
4. The transect method is effective because it utilizes collecting expertise within a protocol that standardizes sampling effort and area. The protocol provides a much more rapid and cost-effective method for studying termite assemblage structure than sampling regimes designed to estimate population abundances. It was demonstrated that one supervised training transect was sufficient to ensure the protocol was conducted with the required level of sampling efficiency.
5. The protocol offers a rapid tool for investigating spatial and temporal patterns of termite assemblage structure in tropical forest sites. Existing data also suggest that termites warrant further investigation as ecological indicators. Termite assemblage composition shows a strong response to habitat disturbance and may be indicative of quantitative changes in the decomposition process. The termite transect has potential as a useful addition to any suite of organisms recommended for monitoring functional processes in tropical forests.
Given the continued destruction, disturbance and fragmentation of tropical forests, and the associated extinction of species (Sayer & Whitmore 1991; May, Lawton & Stork 1995; Turner 1996), there is an urgent need to develop and test standardized methods for sampling biodiversity (Sutton & Collins 1991; Stork & Samways 1995). This has prompted a growing literature on the search for ‘bioindicators’, particularly for insects, because they form the bulk of species and animal biomass in terrestrial habitats and have a strong influence over many ecosystem processes. McGeoch (1998) reviewed the selection and utility of insects as indicators, and emphasized the need for explicit aims and hypothesis testing when nominating a species or group of taxa as a bioindicator. The basic purpose of bioindicators, whatever the ultimate aim, is to indicate a relationship with another biotic or abiotic variable. These relationships with other variables are, however, only as good as the sampling method employed to gather the data on the ‘target’ group (sensuHammond 1994; i.e. the taxa being collected). If a sampling method cannot reproduce a reliable picture of the target group, then any attempt to use the results to infer relationships with other variables is likely to lead to spurious conclusions. Therefore, an important preliminary question that must be addressed before a target group can be evaluated as a bioindicator is: How well does a sample of the target taxon characterize the variation within its own group? In this paper we focus on termites (Insecta; Isoptera), and ask: Can a rapid assessment technique produce a representative sample of the local termite assemblage that provides reliable information on the composition and function of the group as a whole?
Amongst insects, Brown (1991) identified termites as potentially one of the most important indicator taxa, not least because they are at the ecological centre of many tropical ecosystems (Wilson 1992). Termites can achieve very high population densities; for example, in the forests of southern Cameroon termites are one of the most numerous of all arthropod groups (Watt et al. 1997), with abundances of up to 10 000 m–2 and live biomass densities of up to 100 g m–2 (Eggleton et al. 1996). As the dominant arthropod detritivores, termites are important in decomposition processes (Wood & Sands 1978; Matsumoto & Abe 1979; Collins 1983) and thereby play a central role as mediators of nutrient and carbon fluxes (Jones 1990; Lawton et al. 1996; Bignell et al. 1997). They exhibit a wide range of dietary, foraging and nesting habits, with many species showing a high degree of resource specialization (Wood 1978; Collins 1989; Sleaford, Bignell & Eggleton 1996). Their influence on decomposition processes at any site is likely to be governed to a large extent by the species composition of the local termite assemblage (Lawton et al. 1996).
Termites are taxonomically tractable, relatively sedentary, and the sterile castes (workers and soldiers) are usually present throughout the year and can therefore be sampled directly, unlike solitary and more mobile groups. However, the varied sampling regimes adopted in previous studies do not allow reliable direct comparisons of termite assemblages among sites (Eggleton & Bignell 1995). Therefore, a sampling method is required that can characterize accurately the structure of the local termite assemblage at different sites. This would allow reliable comparative analyses and help to elucidate spatial patterns of termite-mediated decomposition (Eggleton & Bignell 1995; Bignell et al. 1997).
An important requirement of standardized sampling is the ability to estimate turnover in species composition between sites (i.e. β-diversity: species that occur at one site but not the next). Site comparisons using inventories (the list of all species in the target group that occur at a site) will give true levels of turnover, whereas comparisons using samples will artificially increase species turnover, a source of error called pseudoturnover (Nilsson & Nilsson 1985). As the proportion of the local species assemblage collected in a sample increases, so the potential level of pseudoturnover decreases. By extension, the more efficient a sampling method is at accumulating species, the lower the pseudoturnover for the same amount of collecting effort. A rapid assessment protocol must therefore involve a compromise between trying to minimize the size of the samples in order to reduce the time spent collecting and sorting, and the need for samples that are large enough to be representative and have low levels of pseudoturnover.
Many groups of insects can be sampled easily using techniques such as Malaise and flight-interception trapping, baiting, and devices that extract organisms from leaf litter and soil. Several tropical forest studies have shown, however, that the Winkler method and Tullgren funnels fail to extract a high proportion of the termites occupying litter and soil (Stork & Brendell 1993; Olson 1994; Didham 1996). Light-trapping of termite alates (the winged reproductive caste) is not a suitable method for use in the rapid assessment of species richness because interspecific synchronicity of temporal flight patterns is very low (Martius, Bandeira & Medeiros 1996). Because there are no effective automatic sampling methods for termites (Eggleton & Bignell 1995), visual searching had to be adopted in our protocol. Furthermore, as with any sampling method that relies on humans having to search for small, cryptic arthropods, the amount of collecting experience a person has will affect his or her ability to sample termites.
With all these factors in mind, we have developed a standardized belt transect sampling protocol designed to measure termite species richness and functional diversity in tropical forests. The aim of this study was to answer the following questions arising from the use of a rapid assessment method of this kind:
1. What proportion of the local termite assemblage is sampled by the transect method?
2. How representative of the local termite assemblage, in terms of taxonomic and functional composition, is a transect sample?
3. What level of pseudoturnover does the sampling protocol produce?
4. How efficient, in terms of effort, is the transect protocol at sampling an assemblage?
5. How much training is required to run the transects at the required level of collecting efficiency?
To address these questions we ran transects in three forest sites where the local termite fauna is well recorded due to previous extensive sampling programmes. A detailed analysis comparing the termite assemblages of the three sites will be published elsewhere.
Materials and methods
Danum Valley, Sabah, Malaysia
The Danum Valley Conservation Area in south-east Sabah (4°57′ N, 117°36′ E; altitude c. 100 m a.s.l.), on the island of Borneo, consists of primary lowland mixed dipterocarp forest (Marsh & Greer 1992), with a mean annual rainfall of about 2700 mm. In the Holdridge Life Zone system (Holdridge et al. 1971) this habitat is classified as tropical moist forest. Two transects were run in June 1995.
Pasoh Forest Reserve, Peninsular Malaysia
Pasoh Forest Reserve (2°59′ N, 102°18′ E; altitude c. 100 m a.s.l.) is located in the state of Negeri Sembilan, and has a mean annual rainfall of about 2000 mm (Soepadmo 1978). The reserve (described in Lee 1995) consists of primary lowland mixed dipterocarp forest (Kochummen, LaFrankie & Manokaran 1990), and is classified as tropical moist forest in the Holdridge Life Zone system. One transect was run in November 1994.
Mbalmayo Forest Reserve, Cameroon
The Mbalmayo Forest Reserve is an area of disturbed semi-deciduous forest in southern Cameroon with an annual rainfall of about 1520 mm (Lawson et al. 1990). The site selected for the transect (approximately 3°27′ N, 11°28′ E; altitude c. 650 m a.s.l.) was logged about 70 years ago but retains a few large mature trees and a closed canopy and was considered to be very similar to primary forest (Eggleton et al. 1996). In the Holdridge Life Zone system this habitat is classified as tropical premontane moist forest. One transect was run in March 1994.
The termite fauna at each site
The known termite fauna recorded from each site was based on all available species records, in addition to large collections examined by the authors. Had these local species checklists been compiled only from repeated transects, it could be expected that the test transects simply sampled the same subset of the local assemblage, and therefore produced a similar pattern to that seen in the checklists. However, any such sampling bias was greatly reduced because the checklists were compiled from surveys conducted using various sampling methods, from casual collecting and small-scale studies, to large-scale labour-intensive sampling programmes, plus the species found in the transects. Each checklist represented the assemblage of a single forest type, and excluded other habitats such as adjacent riverine forest or agricultural areas. We estimated that the checklist at each site was collected from an area of between 10 and 100 ha.
The transect protocol
The transect (protocol described in Davies 1997; Eggleton et al. 1997) was 100 m long and 2 m wide, and divided into 20 contiguous sections (each 5 × 2 m) and numbered sequentially. Each section was sampled by two trained people for 30 min (a total of 1 h of collecting per section). In order to standardize sampling effort, the collectors worked steadily and continuously during each 30 min collecting period. In each section the collectors searched the following microhabitats, which are common sites for termites: 12 samples of surface soil (each about 12×12 cm, to 10 cm depth); accumulations of litter and humus at the base of trees and between buttress roots; the inside of dead logs, tree stumps, branches and twigs; the soil within and beneath very rotten logs; all subterranean nests, mounds, carton sheeting and runways on vegetation, and arboreal nests up to a height of 2 m above ground level.
The protocol was designed to offer a flexible approach to the sampling, whereby collectors used their experience and judgement to search for, locate and sample as many species of termite as time allowed in each section. Because species can only be separated accurately with the aid of a microscope, collectors were trained to sample specimens from every termite population encountered. Care was taken to search termite mounds because they are known to harbour inquiline species (Eggleton & Bignell 1997). Priority was given to finding soldiers and workers, as these are the easiest to identify, but alates were also collected if present because these could be used in future taxonomic studies. Termites were placed in vials filled with 80% ethanol and labelled with the section number.
Identification of material
Quality control in biodiversity studies necessitates taxonomic accuracy and precision (Hammond 1994; New 1996), and therefore parataxonomists were not used to identify any material. We determined binomial identifications for much of the material examined by reference to the taxonomic literature and the extensive termite collections at the Natural History Museum, London, UK. Remaining taxa were assigned to morphospecies applied consistently across regional voucher collections.
1. Soil-feeding: termites that feed on humus and mineral soil.
2. Wood-feeding: termites that feed on dead wood.
3. Soil/wood interface-feeding: termites that feed on extremely decayed wood that has lost its structure and become soil-like. This is synonymous with the ‘intermediate feeders’ of De Souza & Brown (1994).
4. Litter-feeding: termites that feed predominantly on leaf litter and small items of woody trash.
5. Epiphyte-feeding: Hospitalitermes is known to feed on lichens and other free-living non-vascular plants, which they graze from the surface of tree trunks (Collins 1979; Jones & Gathorne-Hardy 1995).
Analyses of species composition
Chi-squared tests of independence were used to detect possible significant within-site differences between the composition of the transect samples and the local species assemblages. The proportion of species in each family (or subfamily in the case of the Termitidae) was used to compare taxonomic composition, while the proportion of species in each feeding group was used to compare functional composition.
The rate of species accumulation per section, and the level of pseudoturnover inherent in this sampling protocol, were estimated by examining the two Danum Valley transects. To construct the species accumulation curve, the 20 sections from both transects were pooled and 500 sequences of the 40 sections were generated at random without replacement. For each sequence the cumulative number of species was calculated. The mean cumulative number per section across the 500 sequences was then computed and used to draw the accumulation curve.
To investigate how the level of pseudoturnover between transects varied with the number of sections sampled, each Danum transect was split into contiguous blocks of five sections. The block size was then increased by one section at a time up to 20 sections. For each size of block the mean observed turnover of all possible pairwise comparisons between transects was calculated (i.e. for blocks of five sections the number of comparisons between transects = 16; with 6–10 sections, n = 4; for 11–20 sections, n = 1). Observed turnover was calculated as follows:
where S1 and S2 are the number of species in the block from transect 1 and transect 2, respectively, and U1 and U2 are the number of species unique to each set of blocks from the respective transects (adapted from Nilsson & Nilsson 1985). Observed turnover will be made up of ‘real’ turnover (i.e. differences between the total species checklists of each transect sample) and pseudoturnover. As the number of sections increases the level of pseudoturnover should decrease and the observed turnover will approach real turnover.
Estimating sampling efficiency
To investigate the ‘cost-effectiveness’ of the transect protocol we compared the sampling efficiency of the transects with that of standardized population sampling regimes designed to measure the population and biomass densities of termite species in forested systems (Eggleton & Bignell 1995). Sampling efficiency is defined as the number of species collected per unit effort (where effort is measured as the number of days required for one trained person to collect, sort and identify the samples). Population sampling has been conducted in Danum Valley (Homathevi Rahman, unpublished data) and Mbalmayo (Eggleton et al. 1996). These population sampling regimes involved quantifying the density and distribution of termite populations within defined forest plots. For example, in Danum Valley the sampling area was 0·25 ha, within which all termite populations were estimated in epigeal mounds and arboreal nests up to 2 m above ground, 20 soil pits (25 × 25 × 30 cm depth) and all items of dead wood from 20 quadrats (2 × 2 m) (Homathevi Rahman, unpublished data).
Training and the analysis of collecting experience
Training in the transect protocol was provided on-site by experienced termite field biologists, and involved two elements. First, a detailed demonstration was given to illustrate how to search the microhabitats in which termites were likely to be found. Secondly, trainees conducted one practice transect under the conditions specified by the protocol while supervised by the trainers.
The four transects discussed in this paper were all run by experienced collectors. However, to test the effect of sampling experience on collecting efficiency, we used all transects from tropical lowland moist forest in South-east Asia run by experienced collectors as a baseline (Eggleton et al. 1997; Jones & Brendell 1998; D.T. Jones & F. Gathorne-Hardy, unpublished data). The Mann–Whitney U-test was used to examine whether the first (practice) and second transects run by inexperienced collectors in Danum and Pasoh were significantly different from the range of baseline values. The Mann–Whitney test is a non-parametric test and was chosen because the number of baseline transects (n = 11), first transects (n = 3) and second transects (n = 2) were too small to allow accurate tests of normality.
What proportion of local species richness is sampled by the transects?
The number of termite species collected from transects in the three forest sites and the total known assemblages are given in Table 1. The proportion of the total known assemblage sampled in the transects varied from 31·2% (Danum 1) to 36·3% (Pasoh). The species collected in the Pasoh and Mbalmayo transects are listed in the Appendix, while the species collected in the Danum transects are given by Eggleton et al. (1997).
Table 1. The number of termite species collected from transects in three forest sites. The total number of known species recorded from each site is based on all available records, from labour-intensive sampling programmes to casual collecting. These totals represent the best available estimates of the species richness of each assemblage
Species sampled in transect
Total known species
Proportion of total fauna in transect (%)
Eggleton et al. (1997); Homathevi Rahman, unpublished data.
Is the transect sample representative of the local assemblage?
None of the transects differed significantly in taxonomic composition from the total known termite assemblage at each site [Danum Valley (Fig. 1a) transect 1: χ2 = 2·59, P = 0·76; transect 2: χ2 = 5·43, P = 0·37; Pasoh (Fig. 1b): χ2 = 3·93, P = 0·56; Mbalmayo (Fig. 1c): χ2 = 2·19, P = 0·82]. Similarly, when considering the functional composition, none of the transects had proportions of species in each feeding group that were significantly different from the feeding group composition of the total known termite assemblage at each site [Danum Valley (Fig. 2a) transect 1: χ2 = 3·11, P = 0·54; transect 2: χ2 = 2·92, P = 0·57; Pasoh (Fig. 2b): χ2 = 1·79, P = 0·77; Mbalmayo (Fig. 2c): χ2 = 0·37, P = 0·95]. The two transects run at Danum Valley gave very similar results, suggesting the protocol can achieve good sampling replication at the same site.
What level of pseudoturnover does the sampling protocol produce?
When the 40 sections of the two Danum transects were combined, the rate of species accumulation slowed rapidly (Fig. 3). The rate of increase per unit sampling effort (i.e. per section) was 2·16 species at section 5 but had slowed to only 0·29 species by section 39. Accordingly, the number of additional sections that must be sampled in order to increase the species richness by 10% increased from only one at section 5 to 14 at section 39 (Fig. 3). As each section requires 1 h of sampling, after 20 sections another 7 h of sampling effort would be needed to increase the species richness of the transect sample by 10%.
As expected, the observed turnover between the two Danum transects declined (Fig. 4) as the species accumulation curve increased. Observed turnover was 60% when blocks of only five sections were compared, falling to 29% when 20 sections were compared. As the two transects were run within 500 m of each other we assume that real turnover between the species pools sampled by each transect was low. Therefore, a majority of the observed turnover between the transects was pseudoturnover. The gradient of the line in Fig. 4 indicates that after 20 sections further decline in pseudoturnover will be more gradual.
How efficient is the transect protocol at sampling species richness?
The sampling effort involved in the transect protocol was considerably less than for the population sampling regimes (Table 2). The transect accumulated species much more rapidly than population sampling (see Fig. 5 for Danum Valley data; the Mbalmayo data, not shown, produced a similar pattern). In addition, population sampling generated more specimens than the transect method and thus required far greater taxonomic processing time for the same species richness. The estimates of sampling efficiency (Table 2) suggested that population sampling requires approximately four to five times more effort to obtain and identify roughly the same number of species as a single transect.
Table 2. A comparison of the approximate effort required to conduct transects and plot-based population sampling regimes, the cumulative number of termite species collected, and the sampling efficiency of both methods in forest at Danum Valley (Sabah) and Mbalmayo (Cameroon). Sampling efficiency is defined as the number of species collected per unit effort, where effort is measured as the number of person days required to collect and process the samples. Taxonomic processing is the time taken for one expert to sort and identify specimens and, in the case of the population sampling, to count specimens. See text for description of the population sampling methods
Collecting time (days)
Taxonomic processing (days)
Total effort (days)
Cumulative number of species
Sampling efficiency (number of species collected per day)
Many termite studies have used soil pits for investigating the diversity of soil-dwelling termites (Abe & Matsumoto 1979). When the transect protocol was first devised it also included one soil pit (20 × 20 × 50 cm depth) in every section (Eggleton et al. 1995). As one of those prototype transects was run at the same site in Mbalmayo, Cameroon, as the current study it was possible to compare the protocol with and without 20 soil pits. The transect without soil pits collected a total of 47 species (Table 1), whereas the transect with soil pits sampled a similar richness (46 species; Eggleton et al. 1995) but required considerable additional labour to hand-sort the large volume of soil. There is clearly no benefit gained from including soil pits when using this rapid assessment protocol because the dozen small surface soil samples in each section capture a similar number of soil-dwelling species for considerably less effort. It should be noted that although the second transect was run 2 years later, both were run in a wet season and it is highly unlikely that the local species pool would have changed significantly during the intervening period.
How much training is required to run the transects at the required level of collecting efficiency?
After the initial training by experienced termite field biologists, six new collectors were able to locate and sample between 23 and 26 species of termite in their training transect (three pairs conducting one transect each: two transects at Danum and one transect at Pasoh). These transects were significantly lower in species richness compared with the 11 South-east Asian baseline transects conducted by experienced collectors (Mann–Whitney U-test; Z = 2·49, P = 0·01). However, two pairs of these new collectors conducted second transects at Danum that produced 28 species each, which was not significantly different from the richness of the baseline transects (Z = 1·48, P = 0·13). Therefore, the data strongly suggest that field assistants who are trained by experienced collectors can sample termites to the required standard after one practice transect. The critical factor is that collectors must be sufficiently motivated to work continuously at a steady pace throughout the 30-min sampling period in each section of the transect.
The efficacy of the transect protocol
The transect method produces samples that are representative of the taxonomic and functional composition of the overall local termite assemblage. The four transects each captured approximately 31–36% of the known local termite fauna (Table 1), giving a reasonably high degree of sampling consistency among the three sites. Any sampling method must be reliable and informative, i.e. it must reflect real differences between sites (Hammond 1994). This was demonstrated recently with the transect protocol, when studies at Danum Valley (Eggleton et al. 1997) and Maliau Basin, Sabah (Jones, Tan & Bakhtiar 1998), using paired transects within different forest types (the forest types were within close proximity to each other), showed that termite taxonomic and functional composition was more similar within forest types than between them.
When designing sampling protocols, Coddington et al. (1991) emphasized the value of utilizing the traditional ‘museum collecting’ approach (i.e. inventories), which is efficient at maximizing the number of species recorded. The strength of the transect method is that it incorporates this qualitative collecting expertise within a protocol that standardizes sampling effort and area. This is because the flexible nature of the transect protocol allows a collector to search all suitable microhabitats within each section, thus giving a higher likelihood of finding termites. In comparison, the strictly defined and prescriptive method used in population sampling is restricted to dead wood, termite nests and a limited number of soil pits. These labour-intensive population sampling regimes can, however, produce reliable estimates of termite population density (Eggleton et al. 1996) that can be used to quantify the impact of termites on ecosystem processes such as carbon fluxes (Bignell et al. 1997).
Pseudoturnover cannot be eliminated without the use of complete species inventories. However, this is not practical if a sampling method is to be used for rapid assessment purposes, and so a suitable trade-off must be found between minimizing collecting effort while maximizing the number of species accumulated and thereby reducing pseudoturnover. The Danum data indicate that by 20 sections the rate of species accumulation (Fig. 3) and the decline in pseudoturnover (Fig. 4) had both slowed considerably. Additional sampling effort would not be justified by the minimal improvements in species accumulation and pseudoturnover. Thus, a transect of 20 sections appears to provide the optimal amount of sampling effort required to maximize the efficiency of the protocol. Simberloff's (1976) detailed study of insect communities on mangrove islands clearly highlighted how mobile organisms, particularly transients or ‘tourists’, can greatly increase pseudoturnover. The eusocial lifestyle of termites means that collecting soldiers or workers is proof of colony establishment and its functional input to that ecosystem.
Applications of the termite transect protocol
McGeoch (1998) defined ecological indicators as taxa that are sensitive to, and demonstrate the effects of, anthropogenic environmental stress or disturbance on biotic systems. Evidence suggests that termites may prove useful and versatile as ecological indicators. For example, several studies in West Africa have described how termite assemblages show a marked response to modification of the forest structure (Eggleton et al. 1995, 1996; Dibog et al. 1999) and are thus sensitive indicators of habitat disturbance. Recent work conducted along a land-use intensification gradient in central Sumatra (Jambi, Indonesia) using the transect protocol reveals that termite species richness and relative abundance both show highly significant correlations with various plant parameters, particularly the reduction in canopy cover (D.T. Jones, unpublished data).
The transect protocol may also be useful in assessing the impact of termites on ecosystem processes. A detailed knowledge of spatial patterns of variation in termite assemblages will help to quantify termite-mediated decomposition (Eggleton & Bignell 1995; Bignell et al. 1997). For example, variation in the dominance of taxonomic groups such as the fungus-growing Macrotermitinae and the soil-feeding Apicotermitinae can be expected to produce significant functional differences among sites (Wood & Sands 1978; Bignell et al. 1997). In South Africa, Muller et al. (1997) proposed the use of termites as an indicator taxon for ecosystem processes across a network of sites designed for monitoring the impact of land-use changes. However, it is still uncertain to what extent differences in the species richness and functional diversity of termite assemblages may result in different levels of decomposition (Lawton et al. 1996). Work currently in progress in French Guiana (R.G. Davies, unpublished data) should help to quantify the relationship between termite diversity and ecosystem function.
With the current pace of terrestrial habitat modification, it would be prudent to invest greater resources into monitoring the healthy functioning of ecosystems. Kremen et al. (1993) advocated sampling arthropod assemblages in order to provide information for the management of optimal ecosystem functioning. This would require standardized sampling methods for a suite of target taxa from unrelated groups that represent all major ecosystem processes (di Castri, Robertson-Vernhes & Younes 1992; Pearson 1994). In such a suite of bioindicators, termites may serve as the best indicator of decomposition processes in lowland tropical ecosystems. If termites are adopted as an indicator taxon, then the transect protocol can be regarded as a relatively quick, simple, cost-effective and reliable method for assessing termite species assemblages in tropical forests.
These results represent considerable field work, and we are grateful to our colleagues and field assistants for all their help. In particular, we thank Drs Maryati Mohammed, Homathevi Rahman, Jeeva Dharmarajah, David Bignell and Luc Dibog. We are grateful to Yayasan Sabah and the Danum Valley Management Committee for their support. The work was partly funded by Darwin Initiative (grant 162/4/059) and NERC via the Terrestrial Initiative in Global Environmental Research programme (award GST/02/625). Bill Sands provided invaluable assistance with identification. Helpful comments on the manuscript were provided by Allan Watt, Bill Sands, Peter Hammond, Richard Davies, Dick Vane-Wright, Charlotte Jeffery and Malcolm Scoble. We are very grateful to British Airways for providing free international flights as part of the BA Assisting Conservation scheme. This is paper A/222 in the Royal Society's South-East Asian Forest Research Programme.
Received 3 October 1998; revision received 21 September 1999
Pasoh forest reserve
Checklist of the 29 species from the transect run in November 1994. Morphospecies lettering as in Jones & Brendell (1998).
RHINOTERMITIDAE: Parrhinotermes minor Thapa, P. pygmaeus John, P. sp. A, Schedorhinotermes javanicus Kemner, S. tarakanensis (Oshima). TERMITIDAE: Macrotermitinae: Macrotermes malaccensis (Haviland), Odontotermes denticulatus Holmgren, O. grandiceps Holmgren, O. longignathus Holmgren, O. oblongatus Holmgren, O. sarawakensis Holmgren, O. sp. D, Ancistrotermes pakistanicus (Ahmad). Apicotermitinae: Euhamitermes sp. Termitinae: Prohamitermes mirabilis (Haviland), Labritermes emersoni Krishna & Adams, Microcerotermes dubius (Haviland), Procapritermes ?neosetiger Thapa, Kemneritermes sp. A, Coxocapritermes sp. B, Pericapritermes nitobei (Shiraki), Pericapritermes sp. C, Dicuspiditermes nemorosus (Haviland). Nasutitermitinae: Nasutitermes longinasus (Holmgren), Nasutitermes sp. C, Bulbitermes sp. A, Oriensubulitermes inanis (Haviland), Proaciculitermes sp. D, Subulioiditermes sp. B.