Introduction

Geographical isolation and ecological divergence contribute to species diversification. Ecological divergence has been suggested as an important factor promoting speciation (Orr and Smith 1998; Schluter 1998, 2001). Populations in different environments are under different selection regimes, which could result in genetic divergence leading to reproductive isolation (Mayr 1963). Alternatively, reproductive isolation could result from processes not linked directly to ecological divergence, such as genetic drift (Coyne and Orr 2004).

Integrating phylogenetic information with environmental factors and geographical distributions of closely related species can be used to indicate the relative evolutionary importance of ecological divergence and geographical isolation. For example, Graham et al. (2004) combined geographical distribution data, environmental geographic information system layers, environmental niche models, and phylogentic information to investigate speciation processes in dendrobatid frogs. They concluded that ecological niche played a significant role in speciation because sister taxa typically differ ecologically regardless of their geographical distributions. Peterson et al. (1999) used ecological niche modeling to compare geographical distributions of sister taxa of birds, mammals, and butterflies in southern Mexico. Their results indicated that geographic isolation was the primary factor driving speciation, whereas ecological divergence evolved later.

Distributions of closely related species of black flies often are related to different ecologies (Adler 1988; Adler and McCreadie 1997; McCreadie and Adler 1998). For example, sibling species in the S. vittatum complex occupy streams with different thermal profiles (Adler and Kim 1984), and those in the African S. damnosum complex are associated with different forest types (Boakye et al. 1998), suggesting that ecological conditions play a significant role in the evolution of black flies. Ecological divergence, for example, has been suggested as a driving force in the speciation of black flies (Joy and Conn 2001; Joy et al. 2007; Rothfels 1989).

The species-rich subgenus Gomphostilbia contains nearly 10 % of all nominal black flies worldwide (Adler and Crosskey 2012), including more than 30 species in Thailand, one of which is a complex of seven cytoforms (Kuvangkadilok et al. 2008; Pramual and Wongpakam 2011). Gomphostilbia has been investigated morphologically (Takaoka and Choochote 2004), cytologically (Jitklang et al. 2008; Kuvangkadilok et al. 2008; Phasuk et al. 2005; Pramual and Wongpakam 2011), molecularly (Pramual et al. 2011a; b), and ecologically (Pramual and Kuvangkadilok 2009; Pramual and Wongpakam 2010). Previous phylogenetic studies indicate that this subgenus is monophyletic (Phayuhasena et al. 2010; Pramual et al. 2011a; Thanwisai et al. 2006), making it an ideal candidate for investigating the importance of ecology in black fly evolution.

Closely related species of the subgenus Gomphostilbia in Thailand are difficult to distinguish morphologically and often require chromosomal or molecular identification (Jitklang et al. 2008; Pramual et al. 2011a), raising the question of how these similar species occupy available habitat. Our objective, therefore, was to examine ecological divergence and geographical isolation by integrating phylogenetic information, based on multiple gene sequences, with geographical distributions and ecological conditions of the larval habitats of a set of closely related, nearly isomorphic species in the subgenus Gomphostilbia in Thailand.

Materials and methods

Samples, identifications, and ecological measurements

Larval black flies were collected throughout Thailand (Table 1) and fixed in Carnoy’s solution (3:1, 95 % ethanol: glacial acetic acid) for cytogenetic and molecular analysis (Pramual et al. 2011a). Samples were identified morphologically (Takaoka and Choochote 2004), and cytologically (Jitklang et al. 2008; Kuvangkadilok et al. 2008). Ecological data for larval habitats, including altitude, conductivity, depth, pH, velocity, and width, were obtained from previous publications (Pramual and Kuvangkadilok 2009; Pramual and Wongpakam 2010) and unpublished data (C. Kuvangkadilok, P. Pramual, S. Jitklang, U. Tangkawanit).

Table 1 Black fly species, collection sites, and GenBank accession numbers for COI, COII, and 18 S/ITS1 sequences
Full size table

DNA extraction, polymerase chain reaction, and sequencing

Treatment of larvae for molecular and cytogenetic studies followed the protocol of Pramual et al. (2005, 2011a). DNA was extracted from larval heads using a Genomic DNA extraction kit (RBC BioScience, Taiwan) and kept at −20°C. Polymerase chain reaction (PCR) was conducted for cytochrome oxidase II (COII), using primers TL2-J-3034 (5′- ATTATGGCAGAT TAGTGCA-3′) and TK-N-3785 (5′- GTTTAAGAGACCAGTACTTG-3′), and 18 S/ITS1, using primers 18 s/sd5′ (5′-TGGTGCATGGCCGTTCTTAG-3′) and 5.8 s/sd3′ (5′-GTCGATGTTCATGTGTCCTGC-3′) (Simon et al. 1994). We used the PCR conditions of Conflitti et al. (2010). PCR products were checked with 1 % agarose gel electrophoresis and cleaned using a HiYieldTM Gel/PCR DNA Extraction Kit (RBC BioScience, Taiwan). Sequencing was performed at Macrogen sequencing service (Seoul, Korea), using the same primers as in the PCR. A total of 38 samples representing 12 species of the subgenus Gomphostilbia were sequenced for mitochondrial (COII) and nuclear genes (18 S/ITS1). Representative haplotypes were deposited with GenBank (Table 1). Previously published COI barcoding sequences (Pramual et al. 2011a) were also included in phylogenetic analyses. The final alignment and trees obtained have been deposited with TreeBASE (http://purl.org/phylo/treebase/phylows/study/TB2:S12315).

Data analysis

DNA sequences were aligned using CLUSTAL X (Thompson et al. 1997), with a final visual inspection. Phylogenetic analyses were conducted for the combined data set (COI, COII, and 18 S/ITS1), with sequences from S. fenestratum (accession number JF916884) and S. takense (accession number JF916877) as outgroups. Congruence between separate genes was tested using the partition homogeneity test (Farris et al. 1995) with 1,000 replicates implemented in PAUP* v. 4.10b (Swofford 2002). The results indicated no significant differences between the separate gene regions (P = 0.970). To select the best-fit DNA substitution model for phylogenetic analysis, we used the program jModeltest v.0.1.1 (Posada 2008), based on the Akaike information criterion (AIC) algorithm. The best-fit model for combined sequences was the TIM2+I+G (Tamura and Nei 1993) with gamma shape parameter (G) of 0.7650 and proportion of invariable sites (I) of 0.6540.

The neighbor-joining (NJ) tree was calculated in PAUP*, based on the best-fit selected model. Bootstrap support was estimated using 1,000 replicates. Maximum parsimony analyses were performed in PAUP* (Swofford 2002), using a heuristic search with 1,000 random addition sequence replicates, TBR branch swapping, and MulTrees effect. Bootstrap support was estimated for 1,000 replicates. Phylogenetic relationships were analyzed by Bayesian methods using MRBAYES 3.04b (Huelsenbeck and Ronquist 2001). The best-fit model for Bayesian analysis was selected by hierarchical likelihood ratio tests implemented in MrModeltest (Nylander 2004). The general time-reversible (GTR) model (Rodriguez et al. 1990) with gamma distribution shape parameter of 0.8406 and proportion of invariable sites of 0.6625 was selected. Bayesian analysis was run for 2,000,000 generations, with a sampling frequency of 100 generations. Tracer version 1.5 (Rambaut and Drummond 2004) was used for visual inspection of the point where the log likelihood is stationary. Trees sampled before this point were discarded as burn-in. The remaining trees of two simultaneous runs were included in posterior probability calculations.

To determine ecological divergence among closely related taxa, we used principal components analysis (PCA), reducing the number of stream variables into groups of independent components (PCs). PCs with eigenvalues greater than 1.0 were retained as variables. To interpret PCs, Spearman rank correlations were used to detect relationships between stream variables and PC scores (McCreadie et al. 2006). To determine if species differed significantly along environmental space, multivariate analysis of variance (MANOVA) was used in which species were assigned as a fixed factor and the PC score for each axis was the dependent variable (Graham et al. 2004).

Results

DNA sequence variation and phylogenetic relationships

The combined dataset was 1,875 bp (586 bp for COI, 697 bp for COII, and 592 bp for 18 S/ITS1), with 460 variable positions, of which 408 were parsimony informative. All three phylogenetic analyses revealed similar tree topologies; thus, only one of the ten maximum parsimony (MP) trees is shown (Fig. 1), with tree length of 1,246 steps and consistency index (CI) of 0.491. The MP tree revealed two main clades (I and II) among the 12 species. Clade I comprised nine species in four subclades: S. asakoae, S. doisaketense, S. inthanonense, S. curtatum (subclade I-1), S. sheilae, S. trangense (subclade I-2), S. gombakense, S. decuplum (subclade I-3), and S. siamense (subclade I-4). Clade II consisted of S. angulistylum, S. chumpornense, and S. kuvangkadilokae.

Fig. 1
figure 1

Maximum parsimony tree for Gomphostilbia in Thailand, based on combined dataset of the cytochrome oxidase I, cytochrome oxidase II, and 18 S rRNA/ITS1 sequences. Bootstrap values for maximum parsimony/neighbor joining and posterior probability for the Bayesian analysis are shown above or near the branch. Bar Ten changes

Full size image

The monophyletic ceylonicum group was derived within the subgenus. The batoense group was polyphyletic; one species was clustered with the S. varicorne group in Clade II, and three species belonged to Clade I, which included members of the ceylonicum group. Two species of the varicorne group (S. chumpornense and S. kuvangkadilokae) formed a monophyletic clade and were clustered with S. angulistylum, which currently is recognized as a member of the batoense group.

Three of the six species of the ceylonicum group were monophyletic. Four samples of S. asakoae formed a clade with strong bootstrap support (100 %) but two samples clustered with S. doisaketense. The clade of S. asakoae was sister to the remaining members of Clade I-1. Simulium inthanonense was monophyletic and sister to the clade comprised of individuals of S. asakoae and S. doisaketense. Simulium curtatum was monophyletic and sister to the clade that comprised S. asakoae, S. doisaketense, and S. inthanonense. Simulium trangense was monophyletic but in the clade of S. sheilae, making the latter paraphyletic. Simulium doisaketense was not monophyletic; one individual clustered with an individual of S. asakoae sister to the clade of S. inthanonense. Another individual clustered with an individual of S. asakoae and was sister to the S. inthanonense and S. asakokae+S. doisaketense clades.

All members of the batoense group were monophyletic. Simulium angulistylum was monophyletic, with strong bootstrap support, but clustered with S. chumpornense and S. kuvangkadilokae of the varicorne group. Simulium siamense was monophyletic and the sister group of all other members of Clade I, which consisted of members of the batoense group (in part) and ceylonicum group. Simulium gombakense and S. decuplum were monophyletic sister species in MP analysis, albeit with low bootstrap support (59 %). This cluster was not resolved in NJ and Bayesian analyses.

Geographical and ecological patterns

Clade I-1: S. asakoae, S. curtatum, S. doisaketense, and S. inthanonense

This clade was represented by members of the ceylonicum group. Among the 37 sampling sites with at least one member of this clade, S. asakoae was found at 64.9 % (24 sites), S. curtatum at 21.6 % (8 sites), S. doisaketense at 16.2 % (6 sites), and S. inthanonense at 13.5 % (5 sites). Although their geographical ranges overlapped, these species rarely co-occurred. Simulium asakoae was found with S. inthanonense at two sites and with S. curtatum at one site. Simulium doisaketense was found with S. asakoae at one site. Simulium asakoae occurred over a wide range of altitudes, from 395 m to 1,615 m above sea level (a.s.l.) (Table 2). Simulium doisaketense also was found over a range of altitudes, but at lower elevations (304–870 m a.s.l.). Simulium curtatum and S. inthanonense were restricted to high-altitude streams (>1,100 m a.s.l.) (Table 2).

Table 2 Stream variables for larval habitats of 12 species in the black fly subgenus Gomphostilbia in Thailand
Full size table

PCA revealed two principal components (PCs) with eigenvalues >1.0. These components explained 60.51 % of the total environmental variation among species (Table 3). PC-1 explained 33.75 % of the variation and was related significantly positively to stream width, depth, velocity, pH, and conductivity, and negatively to altitude. PC-2 explained 26.76 % of the variation and was related significantly positively to stream width and depth and negatively to pH and conductivity (Fig. 2). MANOVA analysis revealed significant separation of the species along both PCs (PC-1, F 3, 37 = 6.389, P = 0.001; PC-2, F 3,37 = 4.325, P = 0.010).

Table 3 Results of principle complment ananlysis (PCA) and Spearman’s rank correlations between stream variables and principal components (PCs) for phylogenetically derived clades
Full size table
Fig. 2
figure 2

Principal components analysis (PCA) of stream variables for four species in Clade I-1 of subgenus Gomphostilbia in Thailand. Factor score 1 (PC-1) explained 33.75 % of the total variation and was related positively to stream width, depth, velocity, pH, and conductivity and negatively to altitude. Factor score 2 (PC-2) explained 26.76 % of the total variation and was positively related to stream width, depth, and velocity and negatively to pH and conductivity. The four taxa of Clade I-1 were separated in environmental space along the two axes

Full size image

Clade I-2: S. sheilae and S. trangense

Simulium sheilae was restricted to southern Thailand, whereas S. trangense was distributed widely in southern, central, and northeastern Thailand. Both species occupied a wide range of stream conditions (Table 2) but were not found above 600 m. Among the 13 sampling sites where at least one member of this clade was found, S. trangense occurred at 84.6 % (11 sites), S. sheilae at 38.5 % (5 sites), and both species at 23.1 % (3 sites).

PCA revealed three components that explained 91 % of the total variation in stream variables. PC-1 explained 49.49 % of the variation (Table 3), and was significantly positively related to pH and conductivity. PC-2 accounted for 24.80 % of the variation, and was related significantly positively to stream depth and negatively to altitude (Fig. 3). PC-3 accounted for 16.71 % of the variation but was not significantly related to stream variables. MANOVA analysis indicated that the species were not significantly different along PC-1 and PC-2 (PC-1, F 1, 14 = 0.430, P = 0.525; PC-2, F 1, 14 = 0.050, P = 0.827; PC-3, F 1, 14 = 1.142, P = 0.258).

Fig. 3
figure 3

PCA of stream variables for two species in Clade I-2 of subgenus Gomphostilbia in Thailand. Factor score 1 (PC-1) explained 49.49 % of the total variation and was positively related to water conductivity and pH. Factor score 2 (PC-2) explained 24.80 % of the total variation and was positively related to stream depth and negatively to altitude. The two taxa of this clade were not separated in environmental space

Full size image

Clade I-3: S. decuplum and S. gombakense

Simulium decuplum and S. gombakense were geographically widespread in Thailand over a broad range of stream conditions, and were among the few species inhabiting a wide range of altitudes (50–1,302 m a.s.l.) (Table 2). Of 27 sites where they were recorded, S. decuplum was found at 70.4 % (19 sites) and S. gombakense at 29.6 % (8 sites). Although these species overlapped geographically, they never occurred together.

PCA extracted two PCs with eigenvalues >1.00, which explained 67.77 % of the total variation in stream conditions (Table 3). PC-1 accounted for 38.36 % of the variation and was related significantly positively to stream width, depth, velocity, and conductivity. PC-2 accounted for 29.41 % of the variation, which largely explained the chemical conditions of the streams, as this component was related significantly positively to pH and conductivity (Fig. 4). MANOVA analysis revealed no significant differences along PC axes (PC-1, F 1, 27 = 0.267, P = 0.610; PC-2, F 1, 27 = 1.568, P = 0.222).

Fig. 4
figure 4

PCA of stream variables for two species in Clade I-3 of subgenus Gomphostilbia in Thailand. Factor score 1 (PC-1) explained 38.36 % of the total variation and was related positively to stream width, depth, velocity, and conductivity. Factor score 2 (PC-2) explained 29.41 % of the total variation and was related positively to water conductivity and pH. The two taxa of this clade were not separated in environmental space

Full size image

Clade II: S. angulistylum, S. chumpornense, and S. kuvangkadilokae

Simulium chumpornense was restricted to southern and western Thailand, whereas S. kuvangkadilokae was found only in northeastern Thailand. Simulium angulistylum was widespread throughout the country. Among 25 sites with at least one member of this clade, S. angulistylum was found at 48 % (12 sites), S. kuvangkadilokae at 40 % (10 sites), and S. chumpornense at 16 % (4 sites). Simulium angulistylum and S. kuvangkadilokae co-occurred at one site, but no stream had all three species.

Simulium chumpornense and S. kuvangkadilokae inhabited streams of similar size, depth, velocity, and pH, but different conductivity and altitude (Table 2). Simulium chumpornense occurred at higher conductivity and lower altitude, whereas S. kuvangkadilokae was found in streams with lower conductivity but higher altitude (Table 2). Simulium angulistylum occupied a wide range of stream conditions from 110 m to 1,153 m a.s.l.

PCA revealed three PCs with eigenvalues >1.00 (Table 3). These PCs accounted for 76.91 % of the total variation in stream conditions among species. PC-1 accounted for 39.01 % and was related positively to stream chemistry (pH, conductivity) and negatively to physical conditions (velocity, altitude). PC-2 accounted for 19.64 % of the total variation and was related significantly negatively to stream depth (Fig. 5). PC-3 accounted for 18.26 % of the variation but showed no significant relationship with stream conditions. MANOVA indicated significant differences along PC-1 (F 1, 25 = 13.381, P < 0.001) but not PC-2 (F 1,25 = 0.992, P = 0.387) or PC-3 (F 1,25 = 3.226, P = 0.059).

Fig. 5
figure 5

PCA of stream variables for three species in Clade II of subgenus Gomphostilbia in Thailand. Factor score 1 (PC-1) explained 39.01 % of the total variation and was related positively to stream width and related negatively to velocity and altitude. Factor score 2 (PC-2) explained 19.64 % of the total variation and was related positively to stream depth. Simulium chumpornense was distributed separately from other taxa of the clade along the PC-1 axis

Full size image

When S. angulistylum was omitted from analysis, PCA revealed three PCs that explained 81.68 % of the total variation in stream conditions. PC-1 accounted for 41.93 % of the variation and was positively related to stream width and significantly negatively related to stream velocity and altitude. PC-2 accounted for 21.44 % of the variation and was related significantly positively to stream depth. PC-3 accounted for 18.31 % of the variation but was not significantly related to the stream variables. MANOVA revealed significant differences along PC-1 (F 1,14 = 48.128, P < 0.001) but not PC-2 (F 1,14 = 0.128, P = 0.727) or PC-3 (F 1,14 = 0.393, P = 0.543).

Discussion

Phylogenetic relationships of Gomphostilbia in Thailand

Previous phylogenetic analyses have shown that Gomphostilbia in Thailand is a monophyletic group but placement of the species into species groups needs revision (Phayuhasena et al. 2010; Pramual et al. 2011a;; Thanwisai et al. 2006). Our results confirm the monophyly of the ceylonicum group, in accord with previous morphological (Takaoka and Choochote 2004), chromosomal (Jitklang et al. 2008), and molecular studies (Phayuhasena et al. 2010; Pramual et al. 2011a; Thanwisai et al. 2006). As with previous studies (Phayuhasena et al. 2010; Pramual et al. 2011a; Thanwisai et al. 2006), we also found that the batoense group is not monophyletic. Four species of the batoense group in our study formed three clades. Simulium decuplum and S. gombakense were the sister clade to the ceylonicum clade, in agreement with the phylogeny based on the COI barcode gene (Pramual et al. 2011a). Morphologically, females of S. gombakense and S. decuplum have an enlarged sensory vesicle (Takaoka and Davies 1995; Takaoka et al. 2010), which also is found in four species of the ceylonicum group (S. curtatum, S. inthanonense, S. sheilae, and S. trangense) and one species of the batoense group (S. parahiyangum). Although our analysis did not include S. parahiyangum, Phayuhasena et al. (2010) found that this species is sister to S. gombakense and that these two species are the sister clade of the ceylonicum group. We therefore suggest a transfer of S. decuplum and S. gombakense to the ceylonicum species group, in which case the principal morphological criteria currently used to define the species groups of Gomphostilbia (Takaoka et al. 2011) would no longer apply universally.

Simulium angulistylum was clustered with the varicorne group, in agreement with analyses based on ITS2 (Thanwisai et al. 2006) and COI (Pramual et al. 2011a). However, a close relation between S. angulistylum and the varicorne group was not observed in the multiple gene analysis of Phayuhasena et al. (2010). Given the different phylogenetic analyses and lack of shared morphological characters between S. angulistylum and the varicorne group, placement of this species in a group needs further study.

Patterns of geographical isolation and ecological divergence among closely related species

Our results revealed three patterns of geographical isolation and ecological divergence for closely related, nearly isomorphic taxa: (1) species overlap geographically but are isolated by habitat and are ecologically divergent (Clade I-1), (2) species overlap geographically but are isolated by habitat while having similar larval ecologies (Clades I-2 and I-3), and (3) species are geographically isolated and ecologically divergent (Clade II).

PCA indicated that five stream variables (altitude, stream conductivity, depth, width, and velocity) differ among the members of clade I-1, and combinations of altitude and stream velocity reveal ecological distinction of the species. Simulium curtatum and S. inthanonense are found at high altitudes (>1,000 m a.s.l.) but occupy streams of different velocity, averaging about twofold greater for S. curtatum (0.49 m/s) than for S. inthanonense (0.23 m/s). Species distributions of black flies often are related to stream velocity (Palmer and Craig 2000; Zhang and Malmqvist 1996). Morphological adaptation to different stream velocities could drive adaptive radiation (Zhang 2006). Thus, habitat isolation between S. inthanonense and S. curtatum might be a result of adaptation to exploit different stream velocities. Further morphological analyses could be used to test this hypothesis. Simulium doisaketense inhabits lower altitudes, relative to other members of Clade I-1, and is not found above 900 m. Simulium inthanonense and S. curtatum are found at high elevations, whereas S. asakoae occurs over a wide range of altitudes (304–1,615 m a.s.l.). The results are consistent with previous studies showing that elevation is associated with species distributions (Jitklang et al. 2008; Kuvangkadilok et al. 1999).

The second pattern that emerged from our analyses is that closely related species overlap geographically but are isolated by habitat and have similar ecologies. Similar ecological niches among closely related species have been attributed to historical constraints, with species tending to retain their ancestral niche (Weins 2004). Four factors prevent ecological niche divergence of closely related species, viz., natural selection, pleiotropy, gene flow, and lack of variability (Weins 2004), but identifying the responsible factor(s) can be difficult. Simulium gombakense and S. decuplum (Clade I-3) overlap geographically, but do not co-occur in streams. Simulium sheilae and S. trangense (Clade I-2) also geographically overlap; although they are not isolated at the habitat scale, they rarely coexist. These two species are nearly identical morphologically, and S. trangense was confirmed and described as a distinct species only after chromosomal study revealed fixed-banding differences (Jitklang et al. 2008). What prevents similar species from coexisting if they overlap geographically and prefer the same ecological niche? One possibility is interspecific competition. Species that use the same ecological niche potentially compete for resources (Morin 1999). Given that competition decreases fitness, natural selection should favor traits that reduce interspecific competitive interactions. We hypothesize that habitat separation of these species is related to oviposition preferences. If a female lays eggs in the same stream as that of another species with a similar ecological niche, its fitness might be compromised by larval competition. Natural selection, therefore, should favor females that lay eggs in streams without competitors.

The third pattern involves closely related species that are geographically isolated and ecologically divergent. Simulium kuvangkadilokae and S. chumpornense, which have nearly identical larvae, are possible sister species (Pramual and Tangkawanit 2008), but S. kuvangkadilokae is found in northeastern Thailand, whereas S. chumpornense is found in southern and western Thailand. Both physical and chemical factors of the streams differ significantly between these species. Geographical isolation of the species could be due to vicariance or dispersal. However, no vicariance event is apparent that could result in separation of S. chumpornense from S. kuvangkadilokae. Although a seaway putatively existed during the Neogene period (Woodruff 2003), it was located at a latitude lower than the current distribution of these species. Geographic isolation of S. chumpornense and S. kuvangkadilokae is therefore more likely the result of dispersal. Phylogeographic study of another group of black flies, the S. tani complex, indicated colonization of the eastern and southern areas from northern populations through the western forest corridor (Pramual et al. 2005). Simulium chumpornense is found in the western and southern regions, and we hypothesize that the ancestral population colonized these regions by dispersal. A phylogeographic approach could be used to test this hypothesis. Among the stream variables that differ significantly between S. kuvangkadilokae and S. chumpornense, altitude, conductivity, and stream velocity are markedly divergent. This situation resembles that of Clade I-1, suggesting that these factors could have played an important role in population isolation leading to species divergence.

A model of black fly speciation driven by chromosomal rearrangements, coupled with biological divergence, was proposed by Rothfels (1989). According to this model, closely related species differ in chromosomal inversions, and speciation results from different co-adapted gene systems leading to divergence in life cycle, host choice, or niche preference. Our results support this model; closely related species occupy different ecological niches. Previous studies have shown extensive chromosomal polymorphisms in many species of Gomphostilbia, which could be partitioned into distinct cytoforms (Jitklang et al. 2008; Kuvangkadilok et al. 2008). Ecological differentiation among cytoforms also has been demonstrated (Jitklang et al. 2008; Kuvangkadilok et al. 2008; Pramual and Wongpakam 2011), suggesting that chromosome inversions play a role in ecological preference. The Rothfels (1989) model also suggests that sympatric speciation is probable, based on empirical evidence that sibling species tend to be sympatric. In our study, 8 of the 12 species overlap geographically. These species are ecologically divergent, suggesting that adaptation to different ecological niches could have promoted speciation.

Pleistocene climate and environmental changes also could have facilitated population divergence leading to speciation or perhaps aiding chromosome-mediated speciation of some members of Gomphostilbia. These Pleistocene changes promoted population fragmentation and speciation in a number of different organisms (e.g., Carstens and Knowles 2007; Johnson and Cicero 2004; Knowles 2001). Although the effects of Pleistocene glaciations might have been less severe in the tropics than in the temperate regions, the environmental changes promoted speciation (Haffer 1969; Mayr and O’Hara 1986) and in several Southeast Asian mainland species, influenced genetic structure and diversity (e.g., Cannon and Manos 2003; Morgan et al. 2010; O’Loughlin et al. 2008). The dry conditions produced during glaciations reduced running-water habitats, although streams persisted in high mountainous areas (Gathorne-Hardy et al. 2002). Pleistocene glaciations, thus, could have driven population fragmentation of black flies, as demonstrated by the phylogeographic study of the S. tani complex (Pramual et al. 2005). Climate changes during the Pleistocene also played a significant role in the demographic history of simuliids. Recent population expansions, for example, have been found in the S. tani (Pramual et al. 2005) and S. siamense complexes (Pramual et al. 2011b). Geographical isolation of S. chumpornense and S. kuvangkadilokae also might have been a result of Pleistocene effects, the geographic distributions of these species reflecting the historical dispersal route of the S. tani complex (Pramual et al. 2005).

Integrating information on geographical distributions, larval ecology, and phylogeny enabled us to infer factors contributing to habitat exploitation of black flies. Both geographical isolation and ecological divergence are found among similar Gomphostilbia species. Most closely related species overlap geographically but are ecologically divergent, suggesting that ecological shifts could have been involved in species divergence in Gomphostilbia, although allopatric speciation driven by Pleistocene climate changes also could have contributed to speciation. If further evidence, such as broader phylogenetic study, supports the finding that some of the species pairs in our study (e.g., S. kuvangkadilokae and S. chumpornense) are true sister species, speciation driven by adaptation to different ecological niches might have been a prevalent mode in the subgenus Gomphostilbia.