Diversity patterns in terrestrial dipteran communities

Authors

  • Klaus. HO­Vemeyer

    1. Abteilung O­kologie, Institut für Zoologie und Anthropologie, Georg-August-Universität Göttingen, Berliner Str. 28, D-37073 Göttingen, Germany
    Search for more papers by this author

K. Hövemeyer, Abteilung O­kologie, Institut für Zoologie und Anthropologie, Georg August Universität Göttingen, Berliner Str. 28, D-37073 Göttingen, Germany. E-mail: khoevem@gwdg.de

Abstract

1. Dipteran communities were studied in five terrestrial habitats [beech forest (BE), oak and hornbeam forest (OH), hedgerow (HE), meadow (ME), alder and willow forest (AW)] using emergence traps and diversity patterns of three trophic groups with soil-dwelling larvae (zoophages, phytosaprophages and surface scrapers) were analysed in detail.

2. Across habitats, sampling effort was a poor predictor of species richness, and species richness increased more steeply with sample size in the zoophages than in the phytosaprophages and surface scrapers.

3. Point diversity (S/trap) of phytosaprophages and surface scrapers increased (as predicted) with resource heterogeneity in the litter layer, but that of zoophages did not. It is suggested that this may be due to differential resource requirements of the three trophic groups during adult life.

4. Zoophages attained the highest and surface scrapers the lowest values of Shannon diversity, while phytosaprophages were intermediate.

5. Predator:prey ratios differed between habitats; they were particularly high in the meadow and low in the oak and hornbeam forest communities.

6. Species identity (Sorensen index) of soil-dwelling Diptera arranged habitats in an order of decreasing similarity (BE–OH–HE–ME–AW), but percentage similarity indicated a closer relationship between the meadow and the hedgerow communities [BE–OH–AW–(HE–ME)]. Functional similarity, which was based on the proportional biomass of differently sized trophic groups, was highest between the BE and OH communities, i.e. in the habitats most similar in litter layer and vegetation structure.

Introduction

During the last few years biodiversity has become a major focus of animal ecology (e.g. Wilson 1988; Ricklefs & Schluter 1993; Huston 1994; Rosenzweig 1995; Gaston 1996). However, only few studies have considered soil communities (Giller 1996) and much of the ‘enigma of soil animal species diversity’ (Anderson 1975) is still unsolved.

The present study investigates diversity patterns of Diptera, a taxon which is among the least-studied groups of soil fauna. There are probably three reasons for this neglect: (i) efficient extraction of dipteran larvae from soil samples requires a special technique (see Healey & Russell-Smith 1970); (ii) soil-dwelling larvae do not conform well to the conventional distinction between mesofauna and macrofauna (e.g. Dunger 1983); and (iii) most dipteran larvae cannot be identified to species. However, it is possible to circumvent these problems using emergence traps (=‘ground photo-eclectors’, Funke 1971) to sample adults after emergence. The present paper is based on emergence trap studies conducted in five habitats, which differed in heterogenity of plant litter input: three woodland sites (beech forest, oak and hornbeam forest, alder and willow forest), a hedgerow and a mesoxerophytic meadow.  Animal communities are not only sets of taxonomic species, but can also be taken as composed of guilds (Root 1967) or functional groups (Cummins 1973), which comprise species that utilize similar food types in similar ways. The larval feeding habits of Diptera are highly diverse (e.g. Brauns 1954; Healey & Russell-Smith 1971; Stubbs & Chandler 1978; Ferrar 1987) and species of the communities studied were assigned to ‘trophic groups’ (of larvae). Within trophic groups, species may differ in body size implying further differences in the niche space occupied (Wilson 1975; Wiens 1982). Different roles of this kind were accounted for by assigning species to ‘trophic size classes’.

There is probably not a single habitat for which a comprehensive species list of the dipteran fauna has been completed (Disney 1986). This is often due to taxonomic effort being restricted to certain families (Hövemeyer 1991) and the present study too is selective in this respect, but care was taken to identify all members of certain trophic groups. This paper, however, will not deal with individual species (species lists have been given elsewhere: Hövemeyer 1992, 1995, 1996a,b, 1997), rather it is my objective to compare communities of selected trophic groups within and between habitats with respect to various aspects of diversity.Whittaker (1977) distinguished four levels of inventory diversity, two of which apply to the data: α diversity and point diversity. As to the latter, hypotheses are tested predicting that species diversity should increase with resource heterogeneity, i.e. in the present context, variety of litter types. Such relationships have been demonstrated for many types of organisms and habitats (see Bell, McCoy & Mushinsky 1991, and ecology textbooks for examples).

Predator:prey ratios (P:P ratios) have often been claimed to be roughly constant in the context of food web theory (Cohen 1978; see Wilson 1996, and Warren & Gaston 1992 for further references), but this view has been challenged for various reasons (Closs, Watterson & Donnelly 1993; Wilson 1996). While guild composition, of which P:P ratios are only a special case (Warren & Gaston 1992), may not always be constant (e.g. Cornell & Kahn 1989) and major habitat types appeared to differ in P:P ratios (Warren & Gaston 1992), P:P ratios have also been suggested to be constant within assemblages of higher taxa such as beetles by Gaston, Warren & Hammond (1992). The present study will demonstrate differences between dipteran communities in different habitats.

Point and α diversity are matched by two levels of differentiation diversity, pattern diversity and β diversity (Whittaker 1977), which will also be treated. However, β diversity will not only be analysed with respect to taxonomic composition of communities of soil-dwelling Diptera, but also in order to look for functional community convergence (Cody 1989), using the proportional biomass of trophic size classes as the data base.

Materials and methods

Study sites and sampling procedure

Five habitats located in the vicinity of Göttingen (Germany) were studied using emergence traps (Funke 1971): a beech forest (abbreviated: BE), an oak and hornbeam forest (OH), an alder and willow forest (AW), a hedgerow (HE) and a mesoxerophytic meadow (ME).

In the beech forest, an area the herb layer of which was dominated by the spring geophyte Allium ursinum L. (Dierschke & Song 1982) was studied for 7 years (1981–87) and additional catches were made in an area where Mercurialis perennis L. was the dominant species (1981–82). Mean annual canopy leaf litter input was 278 g dry wt m−2 year−1 and consisted almost exclusively of beech leaves (Jörgensen 1987). In the Allium area, litter input from non-Allium herbs was low and above-ground biomass of herbs measured in summer amounted to 19·2 g dry wt m−2 (Allium litter is virtually inaccessable to soil-dwelling dipteran larvae, as most of it is rapidly consumed by slugs).

In the oak and hornbeam forest (study years: 1988–89) canopy litter input (235 g dry wt m−2 year−1) was composed of approximately equal proportions of oak (46·6%) and hornbeam (43·0%) leaves with common maple as a third component (10·4%). Above-ground biomass of the herb layer in summer was 78·7 g dry wt m−2 (Hövemeyer 1997).

The alder and willow forest was situated on the bank of the River Oder (study years: 1987–88). Canopy litter input was 213·0 g dry wt m−2 year−1 (alder: 54·1%, willow: 45·9%); the field layer was characterized by large herbaceous plants such as nettles (Urtica dioica L.) and butterburs [Petasites hybridus (L.)] (Henke 1989), with a peak biomass of 166 g dry wt m−2 (Hövemeyer 1996a).

The meadow (study years: 1986–88) was rich in plant species (S = 41; Nauenburg 1980). It was mown once a year and autumnal litter input was estimated at ≈80 g dry wt m−2 year−1 (Hövemeyer 1995). In the hedgerows surrounding the meadow site (study years: 1987–88), traps were set up under hornbeam trees, common maple trees and hawthorn bushes. Leaf litter inputs from these woody plant species were 191, 279, and 255 g dry wt m−2 year−1, respectively, while the herb layer, if present at all, was usually sparse and almost exclusively formed by M. perennis (Hövemeyer 1996b).

Weather conditions were normal to moist in most study years comparing monthly precipitation rates with the long-term average (Göttingen Weather Station); only the summers of 1982 and 1983 were exceptionally hot and dry. Since emergence abundances of adults are affected by dry conditions during egg/larval development, i.e. in the year preceding the year in which adults emerge, results from BE83 and BE84 were omitted from the analyses where appropriate.

Traps were operated throughout the vegetation period (March to mid-December) in each habitat; results from some additional traps operated for a shorter period in the meadow (1988), were only used to augment the species list. The numbers of traps employed varied for study years and are given in Table 2. Further details on habitats and sampling procedures are reported elsewhere (Hövemeyer 1992, 1995, 1996a,b, 1997).

Table 2.  Annual catches of phytosaprophages (PS), surface scrapers (SS), and zoophages (Z) from five habitats arranged in decreasing order of Shannon diversity (base: log 10). Mean values and 95% CL of Shannon diversity were calculated from jack-knifed data; n = number of traps. See Appendix I for definitions of habitats; attached numbers indicate study years
 95% CL Trophic group
MeanUpperLowernPSSSZ
3·523·713·338  AW87
3·323·623·028  AW88
3·283·493·0712  ME87
3·163·362·9612  ME86
3·093·212·9710OH88  
3·033·242·829  HE88
2·873·172·5710  OH88
2·843·062·624  ME88
2·802·942·668AW87  
2·783·072·4910  OH89
2·743·012·478AW88  
2·682·862·5010OH89  
2·552·722·3812ME86  
2·552·862·245  BE87
2·492·972·016  BE82
2·482·902·066  BE81
2·373·461·283HE87  
2·322·721·929HE88  
2·242·422·0610 OH89 
2·112·192·0310 OH88 
2·052·281·828 AW88 
2·052·481·625BE85  
2·052·591·519 HE88 
2·022·291·754ME88  
1·942·041·848 AW87 
1·912·241·585BE87  
1·863·050·673  HE87
1·782·800·765  BE86
1·762·540·985  BE85
1·601·841·365BE86  
1·531·771·293 HE87 
1·471·651·2912 ME86 
1·441·631·256BE82  
1·431·711·1512ME87  
1·302·190·415 BE85 
1·202·190·215 BE87 
1·101·350·855 BE86 
1·061·170·956 BE81 
0·941·170·716 BE82 
0·731·090·3712 ME87 
0·640·950·334 ME88 

Fifty-seven out of 64 recorded dipteran families were selected for full identification, but Sciaridae (and a few other families) from the 1981 catches made in the beech forest were not identified to species.

Trophic groups, size classes and trophic size classes

Based on an intensive search of the literature, which left only some 1% of species and less than 0·1% of individuals unclassified, dipteran species (and sometimes higher taxa) were assigned to 15 types of larval feeding habits, i.e. trophic groups. Most of these trophic groups can be assigned to either (i) the herbivore food chain, or (ii) the saprovore food chain. The first group includes phytophages (abbreviated: PH), aphidivores (AV) and parasitoids of phytophages (PAPH). In the second group, species obligatorily associated with microhabitats, such as fungi or dead wood (macromycetophages: MAMY; xylophages: X) may be distinguished from others whose larvae live in the soil and litter layer: phytosaprophages (PS), microhumiphages (MI), surface scrapers (SS; see Healey & Russell-Smith 1971), hyphae piercers (HYPI; see Healey & Russell-Smith 1971), and zoophages (Z). Some parasitoids of saprovores (isopods) and zoophages (lithobiids; PAPS) belong here too. Phytosaprophages and zoophages may be subdivided further (see Appendix II), but most analyses will be made for the undivided groups. Zoosaprophagous larvae (ZS) utilize breeding media such as faeces and/or carcasses which may have originated from either food chain (Hövemeyer 1992).

Based on dry weight measurements (Thiede 1977; Hövemeyer, unpublished) species (and higher taxa) were assigned to one of 12 size classes, which formed a geometric series: 12, 24, . . . 24 576 × 10−6 g dry wt ind−1.

‘Trophic size classes’ are formed by species (and higher taxa) which were assigned to the same size class and the same trophic group. Assignations of individual species (and higher taxa) to trophic groups and size classes as used in the present study are given elsewhere (Hövemeyer 1995, 1996a,b, 1997).

Diversity measures

Species richness, the Chao 2 estimate, and the Shannon index were used to describe α-diversity, and β-diversity was assessed using the Sorensen index and percentage similarity (see Magurran 1988; Colwell & Coddington 1994; Wolda 1981).

Point diversity was described as species density (see Hurlbert 1971), i.e. the number of species per trap. Since in some study years (BE81/82, HE87 and ME88) large traps (1 m2) were used and small ones (0·25 m2) in the others, it was necessary to standardize results from the large traps in order to assess which species would have been caught had a small trap been used. It was assumed that individuals were distributed randomly and independently over the area covered by a large trap. This is always the case in species represented by just one individual, which make up the major part of species to be considered here, and corrections are thus conservative for highly aggregated species. Corrections were made for individual traps using formula (1) by McArdle (1990) and conducted for all species represented by 10 or fewer individuals per trap, as the probability of obtaining at least one individual of more abundant species is >95%. Species to be retained in the standardized data were randomly selected according to the probability at which they could be expected to be represented by at least one individual in a small trap ( m2); for example, in the species represented by singletons, corrected values were set to unity for one quarter and to zero for the remaining three-quarters of the species.

Pattern diversity was assessed calculating species turnover (TO) as:

TO (%) = [100 × (Ei + Ej)]/(Ei + Ej + Cij),

where Ei and Ej denote the number of species exclusive to traps i and j, respectively, and Cij is the number of species common to both traps.

Results

α Diversity

The study is based on a total of 157 554 individuals, 63 910 of which were identified and represented 650 dipteran species. Total species numbers were fairly similar for most habitats, but in the hedgerow (HE) fewer species were found (Fig. 1). Observed species numbers varied considerably among individual trophic groups: Phytosaprophages and soil-dwelling zoophages contributed most to species richness of both the catches from individual habitats (Fig. 2a) and the entire catch (Fig. 2b).

Figure 1.

Number of dipteran species recorded per habitat. Total species numbers are given above columns and proportions of species occurring in one to five habitats (=ocurrence classes 1–5) are indicated by different shadings. See Appendix for definitions of abbreviations; BE includes additional samples from Mercurialis plots.

Figure 2.

Species richness of selected trophic groups: (a) trophic groups in the five habitats, (b) species numbers of trophic groups pooled across habitats (total numbers are given above columnes). Occurrence classes are indicated as in Fig. 1. See Appendices I and II for definitions of abbreviations (BE includes additional samples from Mercurialis plots). Asterics indicate trophic groups fully identified to species.

Individual trophic groups differed in the ratios between observed species numbers and Chao 2 estimates (Table 1): ratios were rather low in the macromycetophages and phytophages, but higher for phytosaprophages, surface scrapers and zoophages. Since macromycetophages may thus have been inadequately sampled and phytophages were species-rich in only one habitat (ME), most of the following analyses will be restricted to the three last-named trophic groups.

Table 1.  Chao 2 estimates of species richness for selected trophic groups in five habitats (BE-values were calculated excluding catches from dry years and Mercurialis-plots). Percentages of species numbers actually recorded (%observed) are given in parentheses. Asterics indicate estimates for which no species occurred in exactly two traps, while singleton species were present. See Appendices I and II for definitions of abbreviations
Habitats
 BEOHHEMEAW
MAMY33·150·87·019·624·7
(%observed)(63)(63)(*)(82)(57)
PH6·09·55·045·119·0
(%observed)(*)(52)(*)(66)(68)
PS66·493·666·188·0122·1
(%observed)(78)(69)(73)(69)(79)
SS14·023·018·010·017·0
(%observed)(86)(83)(78)(80)(88)
Z55·451·048·389·074·6
(%observed)(83)(84)(75)(75)(92)

Sampling effort for annual catches in individual habitats varied and is here expressed as ‘square metre years’ (m2 × year). This equals the ground area covered by the set of traps operated in a habitat throughout the entire vegetation period. There were no significant positive correlations across habitats of species numbers recorded per year with sampling effort in any of the trophic groups considered in Fig. 3 (Spearman rank correlation; rs ≤ 0·370; n.s.), and also no consistent pattern in the scatter of data points: for example, maximal species numbers of phytosaprophages (Fig. 3a) and zoophages (Fig. 3c) were recorded from the alder and willow forest (AW), while species numbers of surface scrapers were highest in the oak and hornbeam forest (Fig. 3b: OH).

Figure 3.

Relationship between sampling effort and species numbers of (a) phytosaprophages, (b) surface scrapers and (c) zoophages recorded from annual catches in five habitats (see Appendix I for definitions of habitats). Numbers attached to symbols indicate the number of identical data points. For the beech forest (BE), results from ‘dry years’ (d) and Mercurialis-plots (m) are also indicated.

Across habitats, species numbers of zoophages increased more strongly with sample size (log–log linear regression, slope: 0·605; Fig. 4) than did those of surface scrapers (slope: 0·344). Phytosapropages were, on average, represented by only 1·55 times (SE = 0·15; n = 16) as many species as zoophages, but outnumbered the latter by a factor of at least 5·89 (mean = 15·19; SE = 2·17; n = 16) in the corresponding annual catches.

Figure 4.

Relationship between species numbers of phytosaprophages (PS), surface scrapers (SS) and zoophages (Z), and sample sizes of annual catches from five habitats (see Appendix I and Fig. 3 for definitions of abbreviations).

Species rank abundance curves of the three trophic groups in the five habitats (Fig. 5a–e) illustrate a second aspect of α diversity. Proportional abundances were calculated for the mean annual emergence abundance (ind. m−2) in a habitat (excluding ‘dry’ years in BE) in order not to give undue weight to study years with high numbers of individuals caught. While the curves for zoophages are fairly similar to those of phytosaprophages in some habitats (BE, OH, HE; Fig. 5a–c), proportional abundances of zoophages varied less than those of phytosaprophages in the meadow (ME) and the alder and willow forest (AW), and the differences are significant (Kolmogorov–Smirnov two-sample test: ME: P < 0·005; AW: P < 0·05). The curves for surface scrapers always decline steeply (Fig. 5a–e), but they only differ significantly (Kolmogorov–Smirnov two-sample test) from the corresponding curves for zoophages in ME (P < 0·005) and AW (P < 0·05).

Figure 5.

Species rank–abundance curves for communities of phytosaprophages (PS), surface scrapers (SS) and zoophages (Z) in five habitats. Relative abundances of species were calculated from mean annual abundances. Beech forest data are based on catches from Allium plots in moist to normal years.

Using the Shannon index as ‘an ordering device’ (Ghent 1991), annual communities of zoophages tended to attain high values (Table 2) and, overall, zoophages are significantly more diverse than both phytosaprophages and surface scrapers (two-sided Wilcoxon two-sample test for jack-knifed pseudo-values; P < 0·0001). Shannon diversity of phytosaprophages is intermediate and, overall, significantly higher than that of surface scrapers (P < 0·0001).

In a similar manner significance of differences between habitats was assessed for each of the trophic groups (Table 3). Compared with other habitats, Shannon diversity is low for the beech forest communities and mostly differs significantly from communities in other habitats. Apart from this, patterns of differences between communities are not consistent: for example, diversity of phytosaprophages in the oak and hornbeam forest and the alder and willow woodland do not differ significantly, while diversity of surface scrapers is higher in the former habitat and that of zoophages higher in the latter. These differences should somehow be related to specific properties of the respective habitats.

Table 3.  Between-habitat comparisons of Shannon diversity of three trophic groups (see Table 2; two-sided Wilcoxon two-sample tests). Significant results are indicated by ‘>’ (Shannon-diversity is larger for the habitat in the left-hand column than the one in the top row) and ‘<’ (the opposite is true), and p-values are indicated (*** P < 0·001; ** P < 0·01; * P < 0·05); non-significant results (P > 0·05) are indicated by ‘Ð’ and ‘≤’, respectively. Kruskal–Wallis-tests were run prior to pairwise comparisons and demonstrated high heterogeneity in the data (P < 0·0001 in each case)
 OHHEMEAW
(a) Phytosaprophages
BE<<<<
 *********
OH >
**
>
***
Ð
HE  Ð<
*
ME   <
***
(b) Surface scrapers
BE<<Ð<
 ****** ***
OH >>>
  *****
HE  >
***
ME   <
***
(c) Zoophages
BE<<<<
 **********
OH <
**
<
***
HE  <
*
<
**
ME   <
*

Point and pattern diversity

Collector curves for the three trophic groups were generated using species numbers standardized for trap size and only data from moist to normal years. The shapes of curves for both (i) individual trophic groups in different habitats and (ii) the trophic groups within one habitat differ considerably (Fig. 6). Resource heterogeneity was assumed to be the factor responsible for differences in shapes of collector curves of trophic groups. In order to evaluate this hypothesis, habitats were ranked according to the variety of plant litter input at the spatial scale of one trap ( m−2). The resource heterogeneity ranking was based on (i) the number of tree species which made up the litter layer (note that heterogeneity of tree canopy leaf litter for the hedgerow was given rank 1, because each trap was operated under just one of the woody plant species), (ii) the presence of dead wood, (iii) the ranked variety of non-woody plant species, and the presence of (iv) considerable amounts of grass litter and (v) the ‘thick stems’ of large herbaceous plants such as Petasites (Table 4). Based on these rankings the following predictions were made:

Figure 6.

Collector curves for (a) phytosaprophages, (b) surface scrapers and (c) zoophages in five habitats (see Appendix I for definitions of abbreviations). Data points are arranged chronologically and according to the original numbering of traps within study years.

Table 4.  Heterogeneity of litter input in the five habitats (see Appendix I for definitions). Resource heterogeneity was assessed at the spatial scale of one trap ( m2) and numbers estimate variety of resource types
Habitats
 BEOHHEMEAW
Canopy leaf litter1312
Dead wood1111
Variety of non-woody plants12132
Herbs with thick stems1
Grass litter1
Total36346

(i) Species density of phytosaprophages should increase with overall resource heterogeneity (Table 4: ‘Total’);

(ii) Species density of surface scrapers should increase with heterogeneity of canopy leaf litter input only, since surface scraper larvae were assumed to be typically associated with stratified litter layers formed by tree leaves;

(iii) Species density of zoophages should increase with species density of prey and, hence, with total resource heterogeneity [see (i)], because saprophagous dipteran larvae form an important part of their diets, and the assumed relationship between species density of phytosaprophagous Diptera and resource heterogeneity, if correct, should also hold for other phytosaprophagous taxa (e.g. earthworms, enchytraeids), which can serve as prey.

Predictions were tested using the specific non-parametric one-way anova (Meddis 1984; Barnard, Gilbert & McGregor 1993) and species densities averaged for habitats are shown in Fig. 7(a). Mean species densities of phytosaprophages and surface scrapers significantly conformed to the predicted order, but those of zoophages did not (Table 5).

Figure 7.

(a) Mean species density and (b) mean trap-to-trap species turnover of three trophic groups in five habitats. See Appendices I and II for definitions of abbreviations and text for further details.

Table 5.  Results of specific NPAR1WAY-anovas (Meddis 1984) testing predictions on the relationship between resource heterogeneity and species density (species/trap) of three trophic groups of soil-dwelling Diptera
(a) Phytosaprophages
Prediction:OH ≈ AW>ME>BE ≈ HE  
  1. z not calculated.

Result:
Species/trap22·77>15·71>13·30  
SE0·80 0·63 0·53  
n36 28 33  
z = 7·728, P < 0·00003.
(b) Surface scrapers
Prediction:OH>AW>BE ≈ HE>ME
Result:
Species/trap9·10>7·00>3·90>2·04
SE0·35 0·58 0·28 0·28
n20 16 39 28
z = 7·914, P < 0·00003.
(c) Zoophages
Prediction:OH ≈ AW>ME>BE ≈ HE  
Result:
Species/trap11·52leqslant R: less-than-or-eq, slant12·46>5·80  
SE0·93 0·49 0·42  
n36 28 39  

Pattern diversity (sensuWhittaker 1977) was assessed in terms of species turnover between traps within each habitat. Turnover rates as calculated here are mathematically equivalent to the concept of complementarity [Colwell & Coddington 1994; formula (17)], and were calculated for all combinations of pairs of traps within each study year; the resulting values were averaged for habitats. Mean species turnover of zoophages was always higher than both that of surface scrapers and phytosaprophages (Fig. 7b), and, considering corresponding annual catches, this was true for 12 out of 13 (binomial test: P = 0·004) and 11 out of 13 cases (binomial test: P = 0·022), respectively. In the surface scrapers, turnover rates were usually low; this is probably due to their feeding habit which involves high mobility of the larvae in the litter layer.

Predator:prey ratios

The results presented so far suggest that community structure may not be constant for the habitats studied and a closer inspection of the ratios between species numbers of zoophagous and non-zoophagous soil-dwelling Diptera seems appropriate. Predator:prey ratios (P:P ratios) were calculated for three levels of data aggregation.

Based on catches of individual traps, species numbers of predators are strongly correlated with the numbers of non-predatory species (Fig. 8; r = 0·4209, n = 106, P < 0·01). The slope of the reduced major axis (=RMA; see Sokal & Rohlf 1995) is 0·5014. Data points for BE, HE and AW are more or less evenly scattered to both sides of the RMA, but those for ME (binomial test: Nabove RMA:Nbelow RMA = 26:2, P < 0·002) and OH (binomial test: Nabove RMA: Nbelow RMA = 0:20, P < 0·002) are unequitably distributed. P:P ratios for the meadow traps (mean = 0·664) are significantly higher (Kruskal–Wallis test) than values found in other habitats: AW (mean = 0·447; H[28;16] = 16·01, P < 0·001), HE (mean = 0·440; H[28;12] = 10·54, P < 0·005), BE (mean = 0·280; H[28;30] = 35·52, P < 0·001), OH (mean = 0·214; H[28;20] = 34·29, P < 0·001). In addition to that, mean P:P ratios for the oak and hornbeam forest differ significantly from those of AW (means as above; H[20;16] = 25·95, P < 0·001) and HE (H[20;16] = 10·95, P < 0·001), but not from BE (H[20;30] = 1·66, P > 0·05).

Figure 8.

Relationship between species numbers of soil-dwelling dipteran predator and non-predator species in five habitats. The dashed line is the reduced major axis; stippled lines represent selected values of P:P ratios. See Appendix I for definitions of abbreviations.

Considering annual catches, the correlation between species numbers of predators and non-predators remains significant (r = 0·7612, n = 15, P < 0·01), and the slope of the resulting RMA is 0·6147. P:P ratios averaged for habitats are highest for the meadow (0·943), and lowest for the oak and hornbeam forest (0·388) with AW (0·557), HE (0529) and BE (0·424) taking the same ranks as in the previous analysis.

Finally, P:P ratios were calculated for the entire catch from each of the habitats and the following sequence was obtained: ME (0·910) > BE (0·597) > AW (0·580) > HE (0·554) > OH (0·422). The slope of the RMA is 0·7307, but the correlation is not significant for these highly aggregated data (r = 0·4622, n = 5, P > 0·05). Nevertheless, these results suggest marked variation in P:P ratios between habitats.

Between-habitat similarity (β-diversity)

Species were assigned to five occurrence classes, according to the number of habitats from which they had been recorded (Fig. 1 and 2a,b). Nearly two-thirds (n = 422) of the identified species (n = 650) were found in one habitat only (henceforth ‘exclusive species’), while 138, 50, 25 and 15 species were recorded from two, three, four and five habitats, respectively. All but two pairwise comparisons testing for differences in the representation of occurrence classes in the habitats yielded significant results (G-test: mostly P < 0·001); non-significant results were obtained only comparing (i) the beech forest (BE) with the oak and hornbeam forest (OH), and (ii) the meadow (ME) with the alder and willow woodland (AW).

Considering individual trophic groups, both observed species numbers and the proportions of occurrence classes varied widely (Fig. 2a). Exclusive species dominated in many cases (Fig. 2b), while species recorded from four or more habitats attained percentages larger than 10% only in the phytosaprophages (PS), surface scrapers (SS) and parasitoids associated with the saprovore food chain (PAPS).

The G-test was used to test for significant differences between the representation of occurrence classes in individual trophic groups, and (i) the overall pattern (see above) and (ii) the pattern found in the phytosaprophages (see Fig. 2b). Phytosaprophages (G = 12·730; p[d.f.=3] < 0·01) and surface scrapers (G = 7·483; p[2] < 0·025) differed significantly from the overall pattern. This was obviously due to an under-representation of exclusive species, while the significant differences for phytophages (G = 16·262; p[3] < 0·01) and parasitoids of phytophages (G = 4·504; p[1] < 0·05) were attributable to over-representation of exclusive species. Compared with the phytosaprophages, exclusive species contributed more to species richness in most other trophic groups (Z: G = 21·625, p[4] < 0·001; AV: G = 6·979, p[1] < 0·01; MAMY: G = 19·246, p[3] < 0·001; PH: G = 31·289, p[2] < 0·001; PAPS: G = 9·587; p[1] < 0·005); the number of exclusive species was lower than expected only in the surface scrapers, but the test was not significant (G = 2·656; p[2] > 0·05).

Owing to the preponderance of exclusive species in the catches species identity should be low between habitats and, indeed, the upgma-dendrogram generated for the communities of soil-dwelling Diptera (Fig. 9a) shows that (i) annual catches from a habitat were always more similar to catches from the same than to any other habitat, and (ii) communities from OH and HE were more similar to the beech forest communities (BE) than both to ME and AW. If relative abundance of species is taken into account and percentage similarity used as a similarity measure (Fig. 9b), (i) again annual catches from the same habitat were more similar to each other than to catches from any other habitat, but (ii) the hedgerow community appeared to be more closely related to that of the meadow.

Figure 9.

upgma-dendrograms showing similarity between annual catches of soil-dwelling Diptera in five habitats (see Appendix I and Fig. 3 for definitions of abbreviations): (a) species similarity (Sorensen-index); (b) percentage similarity; (c) functional similarity.

Combining the 12 size classes with the nine trophic groups of soil-dwelling Diptera considered (HYPI, MI, PAPS, PS, PSk, PSx, SS, Z, Zx; see Appendix II) gives 108 possible combinations, but this number reduces to 97 as the HYPI's (Lestremiinae: Cecidomyiidae) were not identified to species and always assigned to the same size class. Overall, 51 trophic size classes were recorded, 22 of which were common to all habitats. Total numbers of recorded trophic size classes were 41 for the alder and willow forest, and somewhat lower for the other habitats (ME: 38; OH: 35; BE: 33; HE: 31).

In order to describe the similarity in resource utilization between the soil-dwelling dipteran communities, trophic size classes were treated as ‘species’ and their respective proportional biomass was used to calculate a percentage similarity matrix. The resultant upgma-dendrogramm (Fig. 9c) differs from the two species-based graphs (Fig. 9a,b) in two respects: (i) catches from the oak and hornbeam forest mingled with the beech forest communities, and (ii) communities of the three other habitats were more or less isolated.

Discussion

Emergence traps are particularly suitable for studies of Diptera with soil-dwelling larvae (Funke 1971) and, therefore, most analyses were restricted to three fairly species-rich trophic groups of soil-dwellers, which were efficiently sampled and fully identified to species: phytosaprophages, surface scrapers and zoophages. Catches reflect breeding success of Diptera in the ground area covered by the traps (Hövemeyer 1992), and, hence, contrasting with some other sampling methods such as fogging or Malaise traps, the problem of ‘tourist species’ (Southwood 1996) occurring in the catches is avoided.

While each of the habitats studied harboured a more or less specific community of soil-dwelling Diptera (Fig. 9a), results presented here suggest that phytosaprophages and surface scrapers tend to be less habitat specific than members of the herbivore food chain and the soil-dwelling zoophages (Fig. 2b). Species identity (Sorensen index; Fig. 9a) arranged communities of soil-dwellers in an order of decreasing similarity: BE–OH–HE–ME–AW, suggesting that many ‘woodland species’ occurred in the hedgerow, but percentage similarity (Fig. 9b) indicates that ‘meadow species’ tended to dominate in the hedgerow community. While many trophic groups and individual families conformed more or less well with the overall patterns for soil-dwelling Diptera (Fig. 9a,b), communities of phytosaprophages and Sciaridae (the dominant family in this trophic group) did not and exhibited high species identity between the hedgerow and the meadow catches (Hövemeyer unpublished).

Different species may play more or less similar roles in different habitats, and community convergence may occur in spite of taxonomic dissimilarity (Cody 1989). Similarity in resource utilization between communities of soil-dwelling Diptera (Fig. 9c) as calculated here emphasizes similarity in the abundance of large trophic size classes, and was highest for the beech forest and the oak and hornbeam forest. These habitats resemble each other closely not only in ‘plantscape’ (Samways 1994), but also with respect to the litter layers, since both comprise litter types (beech and oak leaves, respectively) decomposition of which takes more than 1 year. Contrasting with this, litter layers are formed by rapidly decomposing leaf types in the hedgerow, and the alder and willow forest, while grassy litter produces a differently structured litter layer on the meadow. I would suggest that functional similarity of dipteran communities is influenced by such habitat properties.

It is difficult to assess species richness in a habitat from samples (Figs 1a and 2a) as observed species numbers will usually depend on sampling effort and sample size (see Magurran 1988; Gaston 1996). Across habitats, however, sampling effort was a poor predictor of species richness (Fig. 3), and different shapes of collector curves (Fig. 7), too, indicated that habitats differed in species numbers returned for a given sampling effort.

Species richness increasing with habitat heterogeneity has been demonstrated in many studies considering all sorts of communities (e.g. Anderson 1978; Moran 1980; Strong, Lawton & Southwood 1984). It was thus not surprising to find a positive relationship between species densities of phytosaprophages and surface scrapers, and heterogeneity of litter input (Table 5). Why, then, did species densities of zoophages not conform to the predicted order? This a priori hypothesis had been formulated before species densities were calculated, and it turned out that the oak and hornbeam forest and the alder and willow forest, which had been grouped together according to equal resource heterogeneity ranks (Table 4), differed considerably in species densities of zoophages (Fig. 7a). Obviously, heterogeneity ranks suitable to predict species densities of saprophages yielded inappropriate estimates of resource heterogeneity for zoophages. Therefore, soil-dwelling dipteran saprophages and zoophages should differ in some important aspect of life history. This difference is probably related to what has been termed ‘intraspecific diversity’ by Southwood (1978). Within the order of Diptera the importance of adult food gathering ranges from 0 to 100%. Examples are provided by the paedogenetic generations of Heteropezinae (Cecidomyiidae) and species of the so-called Pupipara, respectively, but even species from the same family may occupy different positions on this continuum (e.g. Sciomyzidae; Vala 1989). So, there will be some variation, but overall, soil-dwelling saprophagous and zoophagous dipteran species tend to differ in the importance of adult food gathering: While females of most saprophagous species (e.g. Sciaridae, Tipulidae, Bibionidae) emerge with fully developed eggs, apparently ready for oviposition and adult activity involves little (if any) feeding, adults of most dipteran species with zoophagous larvae do feed substantially, be it as predators (e.g. Hybotidae, Asilidae), as flower visitors (e.g. many Empididae) or both. According to prey capture strategies, predatory adults require perches (e.g. Hybos spp.) or specific substrates as hunting grounds (Tachypeza spp., Platypalpus spp.: Hybotidae) and phytophagous adults depend on the presence of flowers which can be exploited (Chvala 1976). Species richness of zoophagous Diptera may, thus, depend more heavily on the heterogeneity of above-ground structures utilized by adults than on habitat heterogeneity or prey diversity encountered by the soil-dwelling larvae. Indeed, not only species densities (Fig. 7a), but also species richness of zoophages (Fig. 2a) were highest on the meadow and in the alder and willow forest, i.e. in the habitats characterized by the richest and structurally most diverse field layers.

If individual trophic groups respond differently to habitat heterogeneity, this may affect community composition. Predator:prey ratios have often been assumed to be roughly constant within food webs. Data on species numbers of non-dipteran taxa are not yet available for all of the five habitats, nonetheless, I would not expect low species numbers of zoophagous Diptera to be balanced by high species numbers of taxa with different life histories, such as predatory mites or lithobiids. But predator:prey ratios have also been suggested to be constant in assemblages of higher taxa by Gaston et al. (1992), whose analysis of beetle communities suggests that much of the deviation from the general trend was accounted for by sampling method. Using just one sampling method, I too found significant overall correlations between the numbers of zoophagous and non-zoophagous species of soil-dwelling Diptera, but two habitats deviated consistently from the general trend (see Fig. 8: ME, OH), and communities differed significantly in P:P ratios. I would not argue that such variability should be ignored. In the present study, habitats that were species-rich for one trophic group were not always species-rich for others, a result which, if on a spatially and taxonomically more restricted scale, parallels the conclusions drawn by Prendergast et al. (1993), who demonstrated that ‘species-rich areas frequently do not coincide for different taxa’, and Scheu & Schulz (1996), who made a similar point in a study on the oribatid and macrofaunal communities of seral habitats.

This paper has considered diversity in dipteran communities with respect to sampling effort, sample size and habitat heterogeneity. Other factors such as size and age of habitat, productivity, disturbance, isolation, latitudinal gradients and the size of the regional species pool have been suggested to influence species richness (see Lawton & Strong 1981; Magurran 1988; Huston 1994; Gaston 1996). It is not possible to analyse the impact of these factors quantitatively as only one study site per habitat type was sampled. However, with its emphasis on diversity patterns at small spatial scales the present study may help to shape ideas for the design of future sampling programmes.

The German fauna comprises some 8000 dipteran species (Schumann 1992) with ≈4100 terrestrial species contained in the families selected for identification here. It was surprising to find that a small sampling programme covering only five, albeit quite different, habitats located in a small region should detect more than 15% of the German species list. Obviously, many more habitats will have to be studied until the relationship between local species richness and the regional species pool of Diptera can be understood.

Received 26 January 1998; revision received 3 July 1998

Appendices

Appendix I

Definitions of abbreviations for habitatsBE beech forest

BEm beech forest–Mercurialis plots

BEd beech forest–‘dry’ years

OH oak and hornbeam forest

HE hedgerow

ME meadow

AW alder and willow forest

Appendix II

Definitions of trophic groups of dipteran larvae

AV aphidivores: larvae feeding on aphids;

HYPI hyphae piercers: larvae sucking fluids from fungal hyphae and possibly from roots;

MAMY macromycetophages: larvae feeding on fruitingbodies of fungi or thick mycelia;

MI microhumiphages: larvae feeding on various types of small debris particles;

PAPH parasitoids of phytophagous hosts (e.g. Lepidoptera, Hemiptera);

PAPS parasitoids infesting hosts that belong to the saprovore food chain (isopods, lithobiids, snails);

PH phytophages (mostly leaf and stem miners, or gall formers);

PS phytosaprophages: soil-dwelling larvae feeding on dead plant material other than wood; this group may be subdivided into larvae feeding on dead tree leaves or grassy litter (PS sensu strictu), larvae usually associated with compost of dicots (PSk), and larvae for which an association with dead wood has been suggested in the literature, although this may not be obligatory (PSx);

SS surface scrapers: larvae scraping microorganisms from the surfaces of leaves in the litter layer;

X xylophages: larvae feeding on dead wood;

Z zoophages: predatory soil-dwelling larvae, some of which are believed to be associated with dead wood according to the literature (Zx);

ZS zoosaprophages: larvae living on dead animal material (faeces, carcasses).

Ancillary