INTRODUCTION

The woodpecker genus Dinopium has historically consisted of 5 species from South and Southeast Asia, namely Dinopium rattlesii, D. shorii, D. javanense, D. everetti, and D. benghalense (Winkler et al. 1995, Winkler and Christie 2002, Clements et al. 2012). They are commonly called flamebacks due to their golden-colored mantle, scapulars, and folded wings. Among Dinopium, D. benghalense consists of 6 subspecies or races, distributed across South Asia (Gill and Donsker 2014), with 2 in Sri Lanka: D. b. jaffnense (Golden-backed Woodpecker) in the north and D. b. psarodes (Red-backed Woodpecker) in the south (Rasmussen and Anderton 2012, Gorman 2014). Many color variations have been observed within the Sri Lankan forms (e.g., Legge 1880, Baker 1927, Wait 1931, Whister 1944, Philips 1953, Henry 1971, Rasmussen and Anderton 2012, Lamsfuss 2013, Gorman 2014, Fernando and Seneviratne 2015, Freed et al. 2015).

The red-backed form of the Dinopium complex had long been considered as endemic to Sri Lanka. Layard (1853) classified the red-backed form as Brachypternus ceylonus and the golden-backed form (D. b. jaffnense) as B. aurantius. Legge (1880) reported the Red-backed Woodpecker B. ceylonus as an endemic species to Sri Lanka. Both Legge (1880) and Wait (1931) commented on the possible hybridization of the 2 island forms. Intermediate phenotypes of Dinopium were reported as far back as 1868–1877 (Legge 1880). As a result, after Wait (1931), all authors except del Hoyo et al. (2014) continued to adopt the same taxonomic treatment of the 2 Sri Lankan forms as members of D. benghalense (Baker 1927, Whistler 1944, Ripley 1946, Ripley 1949, Philips 1953, Ali and Ripley 1983, Ripley and Beehler 1990, Rasmussen and Anderton 2012).

The recent checklist of Birdlife International (del Hoyo et al. 2014) elevated D. b. psarodes to a full species based on a system proposed by Tobias et al. (2010). In this system, a taxon is considered a separate species if an overall differentiation score reaches 7. The overall score is a sum of 2 elements: the quantification of differentiation in 3 main traits (morphology, voice, and biometrics) and 1 subsidiary trait (ecology and behavior), and the quantification of geographical relationship (allopatry or parapatry and the nature of the hybrid zone). In elevating the subspecies D. b. psarodes into a full species, del Hoyo et al. (2014) used the differences in its coloration (shining scarlet on back shading to dull red on wings, lack of black in carpal area), measurements (longer bill, wings, and tail) and vocalization (louder and higher-pitched call) compared to D. b. jaffnense.

Based on observations of plumage of Dinopium in the wild, a recent paper (Freed et al. 2015) has described 8 intermediate plumage types, and assumed that these types are of hybrid origin. Furthermore, Freed et al. (2015) showed that these intermediate types are found in the north-central part of the island. However, a quantitative analysis of these different plumage patterns in the Dinopium complex has not been attempted. Here we provide the first comprehensive examination of phenotypic (body shape/size and plumage) and genotypic variation across the 2 taxa of the Dinopium benghalense complex in the island to determine how well the criteria proposed by Tobias et al. (2010) for avian taxonomy apply to this species complex and to shed light on the complex biogeographic patterns displayed in Dinopium in the South Asian region (de Silva et al. 2014). We hypothesized that (1) D. psarodes is genetically differentiated from the D. benghalense cluster in a way corresponding to their phenotypic difference; (2) hybridization occurs in the contact zone, resulting in hybrids that have intermediate phenotypes and genotypes; and (3) partial reproductive isolation is achieved through a narrow hybrid zone maintained by selection against hybrids; if so, D. psarodes is best treated as a separate species.

METHODS

Cline Construction

We collapsed latitude and longitude data of sampled birds to a single dimension by measuring the shortest geographic distance from the capture location of each bird to the southeastern coast of mainland India (as shown in Figure 1A; we call this distance the “Geographic Distance”). We used the program CFit-7 (Gay et al. 2008, Lenormand and Gay 2008) to estimate parameters for the best-fitting cline for the Cytb marker (as in Seneviratne et al. 2012). The cline was estimated as a simple sigmoidal curve across the contact zone. We treated the genetic trait as a simple two-allele system with each individual carrying a single Cytb allele.

FIGURE 1.

Allopatric distribution of D. b. jaffnense and D. psarodes and the existence of a hybrid zone between the 2 forms as seen from the capture locations of Dinopium flamebacks, frequency distribution of mitochondrial haplotype (Cytb), and the frequency distribution of plumage color in Sri Lanka. (A) Capture locations of field and museum specimens. The lines indicate the distance from the southeastern coast (Dhanushkodi) of mainland India. Squares denote field capture locations and triangles denote the locations of museum specimens: green = D. b. jaffnense, orange = Dinopium intermediate, red = D. psarodes. (B) Frequency distribution of mitochondrial haplotype (Cytb): green = northern haplotype, red = southern haplotype. (C) Frequency distribution of plumage color: green = D. b. jaffnense, yellow/ red and yellow = Dinopium intermediate in the hybrid zone, red = D. psarodes. The 2 lines indicate the northern and southern borders of the hybrid zone. The province boundaries are also shown.

We also used CFit-7 (with the unimodal model) to fit clines for PC1 scores of morphological and plumage characters. However, to better visualize the clinal variation of those phenotypic traits, we also fit cubic splines in program R (R Development Core Team 2010). We used CFit-7 to test whether the plumage, morphology, and genetic clines are coincident (i.e. have the same location) and concordant (i.e. have the same width); differences in location and width can be due to different patterns of selection, dispersal, inheritance, and associations between genes on these traits (Barton and Hewitt 1985). We compared each of the 4 combinations (unconstrained, centers constrained, slopes constrained, and both center and slope constrained) of the 3 phenotypic and genetic clines. The likelihood-ratio test and Akaike information criterion (Akaike 1974) were used to determine the best model with varying sets of constraints on cline centers and slopes.

Evaluation of del Hoyo et al. (2014) Criteria of Splitting D. benghalense

We compared our dataset with the data used by del Hoyo et al. (2014) to split D. psarodes as a full species following the criteria proposed by Tobias et al. (2010; Table 1). Tobias et al. (2010) used the strength of differentiation in various characters according to effect sizes computed from the means and standard deviations. Effect sizes are most commonly presented as the Cohen's d statistic [d = X₁ – X₂/S_{pooled ,} where X = mean of species 1 and 2, S = standard deviation, and S_pooled = where n = number of individuals sampled in species 1 and 2], which expresses the difference in means in terms of the amount of within-group variation (Tobias et al. 2010). On the basis of the distribution of effect sizes produced by empirical tests of divergence in undisputed species, they have scored character differences with an effect size of 0.2–2 as minor, 2–5 as medium, 5–10 as major, and >10 as exceptional.

TABLE 1.

Comparison of del Hoyo et al. (2014) analysis for splitting Dinopium psarodes as a full species with present data shows a similar conclusion. “Scores” = introduced by del Hoyo et al. (2014).

RESULTS

We sampled 70 individuals across the island along a 430 km transect; based on plumage these were identified as 41 D. psarodes, 17 D. b. jaffnense, and 12 intermediate forms of varied levels of orange plumage. These birds included 32 males and 38 females (sex was determined using plumage; see Methods). We did not observe any sexual differences in morphological traits in either species (one-way ANOVA, F = 0.40, df = 1 and 70, P = 0.53), other than the slight color difference in forehead plumage. From the national collection we sampled 25 D. psarodes, 8 D. b. jaffnense, and 22 intermediate types (Figure 2G). Since museum and field data originated from different sampling locations (Figure 1A), we analyzed them separately.

FIGURE 2.

Color variation of Dinopium flamebacks in Sri Lanka: (A) crimson-red plumage (D. psarodes) (image courtesy of V. Weeratunge), (B) reddish-orange form, (C) orange form, (D) orange-yellow form, (E) orange/reddish mantle, (F) golden-yellow plumage (D. b. jaffnense) (image courtesy of V. Weeratunge), and (G) variation of back color in Dinopium flamebacks in the national collection (image courtesy of the Department of National Museums, Sri Lanka).

Plumage color of the back ranges from red to yellow, with many intermediate colors (Figure 2) as suggested by previous authors (e.g., Fernando and Seneviratne 2015, Freed et al. 2015). A clinal change of plumage characters was observed along the line transect from Jaffna (northern tip of the island: 9.46°N, 80.07°E) to Matara (southern tip of the island: 6.04°N, 80.38°E; Figures 1C, 3C). All individuals captured from Jaffna peninsula had yellow backs (Figure 1A). All individuals captured from the core of the wet zone (southern parts of the island) had crimson-red backs (Figure 1A). The intermediate plumage colors (orange) were found in varying proportions in the north-central dry zone, from the southern border of Wilpathu National Park (08.37°N, 80.07°E) up to Kilinochchi (9.22°N, 80.25°E).

FIGURE 3.

Clinal variation of allele frequency, morphology, and plumage of Dinopium along an axis representing distance from the southeastern coast of mainland India showing the allopatric distribution of D. b. jaffnense and D. psarodes and the presence of the hybrid zone. Zero kilometers represents the nearest southeastern coastal line of India (Figure 1A). Points represent mitochondrial haplotype frequency (red indicates southern Cytb haplotype and green indicates northern haplotype) in groups of local individuals (A) or PC1 scores (morphological and plumage characters) of individuals (B–C: red dots indicate red forms, orange triangles indicate intermediate forms, and green dots indicate yellow forms). The lines represent the best-fitting clines generated by CFit-7. Dark gray area indicates the width of the genetic cline. (A) Cytb gene, (B) Morphology, (C) Plumage.

A clinal change of genetic characters (in Cytb) was observed in the line transect across north-central Sri Lanka (Table 2; Figures 1B, 3A). Allopatric D. b. jaffnense has a G nucleotide and allopatric D. psarodes has an A at the diagnostic polymorphic site in Cytb gene. In the sympatric area a mixture of these Cytb types was found among the intermediate phenotypes.

TABLE 2.

A composition of genotypes of flameback woodpeckers at Cytb gene along the transect line showing a mixture of genotypes among the intermediate phenotypes in the sympatric area.

Our Cfit analysis positioned cline centers 83–149 km (mean center = 115 km) from the southeastern coast of mainland India (Table 3). The Cytb-based genetic cline and the center- and slope-constrained cline based on all traits positioned the cline center at 115–117 km from the coast of India (Table 3; Figure 3A). The widths of these clines range from 18 to 142 km among different traits. The constrained clines, which Akaike Information Criterion (AIC) values indicate should be taken as the best model, gave the center as 117 km and width as 118 km (Table 3). Cubic splines of both morphological and plumage traits showed clinal variation in the same area that of Cytb: ~117 km from the southeast coast of India (line north of Putlum to Vavunia; Figures 1C & 3). These 3 measured traits are generally concordant in their centers and widths (Figure 3).

TABLE 3.

Comparison of genetic, plumage, and morphometric clines based on maximum log-likelihood estimates obtained from Cfit-7. Cline centers are in kilometers from the nearest coast of mainland India. Likelihood ratio tests and Akaike Information Criterion (AIC) indicate that a single cline (Model 4) best represents variation in all of these traits. *Asterisks indicate similar centers or widths due to constrained models.

When we compare the empirical clines with hypothetical neutral clines unaffected by selection, the empirical clines were always narrower than the neutral clines. Assuming that the secondary contact of D. psarodes and D. b. jaffnense took place during the last complete exposure of the Adams Bridge due to the drop of sea level (~10,000 years ago: Ripley 1949, Hopkins 1967), and given the dispersal rate and generation time of C. auratus, a neutral cline would have now reached a width of ~1,700 km. A highly conservative (i.e. minimal) estimate of time since recent contact between taxa (160 years, based on Layard's [1853] original description of the red form and dispersal rate of songbirds) would give a width of 213 km. Just 15 years after this first description came the documentation of intermediate phenotypes between D. psarodes and D. b. jaffnense (Legge 1880). Assuming 1868 as the time of secondary contact would give a cline width of 192 km—still a much wider cline compared to the observed clines (112 km). The narrowness of the empirical clines suggests some form of selection constrains the width of the cline; such selection can be considered a form of partial reproductive isolation that hinders the blending of the 2 forms and stabilizes them as distinct entities (Barton and Hewitt 1985, 1989).

Dinopium flamebacks in Sri Lanka all have red crowns, and plain black napes and upper mantles (Figure 2A–F). However, their backs have 2 predominant color types (Figure 2A–F), crimson-red and olive-yellow, corresponding to the 2 forms. The intermediate color forms range from golden yellow (Figure 2F) to orange (Figure 2C). The orange color is sometimes confined to the mantle, and sometimes extends more prominently into the wings. Secondaries and coverts show olive-yellow, golden yellow, orange, and crimson red (Figure 2A–F). D. psarodes in the extreme south has crimson-red shading in the primaries and in the otherwise black rump as well (Figure 2A). Similarly, the amount of black and white plumes on the facial region and upper wing coverts shows slight variation along the north–south transect. The northern (dry-zone) birds of all color types have broader white supercilium and malar stripes (mustache). Birds in the south (wet-zone) have narrower white facial stripes (hence darker faces; Figure 2A). Upper wing coverts are dotted with 2 rows of white spots in northern birds (Figure 2E, 2F); these spots are less prominent in the southern birds (Figure 2A). In all forms along the transect the throat is black, bill blackish grey, iris chestnut, breast and belly have dark chevrons and streaks, and legs are gray.

Bill length (i.e. total culmen), wing length, and tail length are smaller in D. b. jaffnense than in D. psarodes and intermediates (Table 4). In the morphology-based PCA, PC1 explained 64.6% of the variation and described size differences between the species (Table 5, Figure 4A). In the morphological PCA, many traits contributed heavily to PC1 (Table 5) and there was a clear separation of D. psarodes and D. b. jaffnense in PC1 (ANOVA; P < 0.001). In the plumage PCA, PC1 captured 66.2% of the variation, correlating well with all characters that tend to differ between D. psarodes and D. b. jaffnense (ANOVA; P < 0.001; Table 5 and Figure 4B).

TABLE 4.

Morphological and plumage traits measured in Dinopium flamebacks showing that beak, wing, and tail length of D. psarodes are longer and supercilium and mustache width are narrower in D. psarodes than in D. b. jaffnense (mean ± SE, with data range in parentheses).

TABLE 4.

Extended.

TABLE 5.

Variance explained and factor loading of the first 2 principal components (PC) produced in principal components analyses (PCAs) of phenotypic traits: A) morphological traits and B) plumage traits. In the morphological PCA, PC1 explained 64.6% of the variation and described size differences between the species, with each standardized morphological variable being roughly equally correlated with PC1 (Figure 4). PC1 scores of morphological traits separate the 2 species. The PC1 of plumage captured 66.2% of the variation. PC2 of both morphological and plumage traits captured minor amounts of variation.

FIGURE 4.

Variation of morphological (A) and plumage (B) characters in adult D. b. jaffnense and D. psarodes showing clear separation of D. psarodes only in plumage. Red = D. psarodes, orange = intermediate, green = D. b. jaffnense.

Both Sri Lankan and Indian yellow forms of D. benghalense are clustered together, and form a sister group to Sri Lankan red form (D. psarodes) at Cytb (Figure 5).

FIGURE 5.

Molecular phylogenetic analysis by maximum-likelihood method for Cytb gene.

Our analysis of biometrics following Tobias et al. (2010) and del Hoyo et al. (2014) is consistent with the latter analysis (which resulted in splitting the D. psarodes into a full species) but with a greater effect size (4.21 rather than 2.92). Likewise, the quantitative analysis of plumage and bare parts (effect size 6.73), ecology, behavior, and geographical relationship too are in accordance with del Hoyo et al. (2014) (Table 1).

DISCUSSION

We examined phenotypic and genotypic relationships within 2 distinct forms of Dinopium flamebacks in Sri Lanka to test the validity of the newly elevated species D. psarodes, the Red-backed Woodpecker or the Lesser Sri Lanka Flameback (del Hoyo et al. 2014), a distinct form endemic to the island of Sri Lanka. We documented clear phenotypic and genetic evidence for a hybrid zone, and find that this hybrid zone is narrower (112 km in width) than that expected for a neutrally expanding zone. The narrow hybrid zone suggests some form of selection acts against hybrids; such selection is considered a form of partial reproductive isolation (Barton and Hewitt 1985, 1989, AOU 1998, Brelsford and Irwin 2009).

Is D. psarodes a good species under the biological species concept? We argue that despite extensive hybridization in sympatry, there are several key factors suggesting that D. psarodes is best considered a distinct species. It is phenotypically (Figures 2 and 4) and genetically (Figures 3A and 5) a distinct entity. Except for hybrids in this narrow contact zone (Table 3), it does not show intermediate phenotypes in the rest of its range, and breeding areas of D. psarodes and D. b. jaffnense remain largely separated from each other. When our data is evaluated under the species-delimiting criteria used in Tobias et al. (2010), we find strong support for the species status of D. psarodes.

Phylogenetic affinities based on Cytb support D. psarodes's separation from D. benghalense cluster (Figure 5). A much broader phylogenetic analysis on this group using very large amount of genomic sequence (152,311bp) generated by Restriction Site Associated DNA (RAD) sequencing also revealed a similar pattern (de Silva et al. 2014), where D. b. jaffnense cluster with the Indian forms and the D. psarodes separates out as the sister group with posterior probability of close to 1 (de Silva et al. 2014).

Our multivariate analysis further showed that D. psarodes is a distinct phenotypic entity diagnosed by crimson back, black rump and upper tail coverts, and slightly longer tail and tarsus. Intermediate phenotypes were clustered between the 2 parental phenotypes (Figures 3 and 4). The intermediates were found in the contact area between the parental groups, in a narrow band north of Puttalum to Killinochi in the north-central and northwestern parts of Sri Lanka (Figure 1). The center of the hybrid zone lies roughly in the middle of this band (~112 km from the mainland; Figure 3). We found crimson-red, red, orange-red, orange, golden yellow, and olive-yellow phenotypes of Dinopium in this area (Figure 2). Legge (1880) and several others over the past century have made similar observations (see Introduction). We encountered heterospecific pairs at the center of the hybrid zone as well. For example, in Wilpattu National Park in northwestern Sri Lanka (08.37°N, 80.08°E), we observed a nest with nestlings where the parent male was a D. psarodes and the female was a D. b. jaffnense. A previous observer also has noted similar heterospecific pairs in this same region (Freed et al. 2015).

Despite the plumage differences, these woodpeckers are vocally very similar (Rasmussen and Anderton 2012). We observed only subtle differences in the frequency in their calls (S. Seneviratne personal observation). In fact both species and their hybrids respond to each other's playback calls.

Sri Lanka has 2 main climatic zones. The southern one-third of the island is the wet zone with over 2,500 mm annual rainfall. The rest of the island, including the northern parts near mainland India, is the dry-zone with seasonal rainfall averaging ~1,500 mm (Ashton et al. 1997). The Sri Lankan endemic D. psarodes is bigger, and has longer bill breadth and length, than D. b. jaffnense. A possible reason is that the trees in the wet zone forests are much bigger, such that the bigger bill of D. psarodes can better excavate the thick bark. The supercilium and mustache width of D. psarodes are narrower than that of D. b. jaffnense (Table 4). The darker plumage coloration and the longer beak are general patterns shown by the birds in the wet zone of Sri Lanka compared to their dry-zone and Indian counterparts (Ripley 1946, 1949, Ripley and Bheeler 1990). This fits the general pattern of higher humidity being associated with darker and bigger individuals (Prum et al. 2012, Fernando and Seneviratne 2015). Mainland India and northern Sri Lanka are similar in humidity, and a similar plumage can likewise be found in both the mainland and the island (Fernando and Seneviratne 2015). D. psarodes is found in the dry zone of Sri Lanka as well, but there they are found closer to water bodies such as lakes and rivers, which are abundant in the dry zone. This may indicate that D. psarodes prefers humid habitats compared to its dry-zone counterpart.

Here we have provided the first comprehensive analysis incorporating both phenotypic and genetic variation across a hybrid zone for an Old World woodpecker group. We conclude that D. psarodes of Sri Lanka is both genetically and phenotypically a distinct entity that warrants full species status. Our study provides an example in which the phenotype-based criteria proposed by Tobias et al. (2010) for avian taxonomy are consistent with conclusions arrived at through genetic and cline-based analyses. Hence the conclusion of del Hoyo et al. (2014) to elevate D. psarodes to a distinct species. Detailed studies on the amount of introgression in the context of hybridization, the stability of the hybrid zone and the fate of its hybrids, and the ecology of the new species (D. psarodes) would shed more light into this charismatic group of poorly known Asian woodpeckers.