Wallacea, bound by Wallace’s Line in the West and Lydekker’s Line in the East, is widely recognized as one of the world’s leading biodiversity hotspots (Mittermeier et al. 1999; Myers et al. 2000). The area has very high levels of vertebrate endemism, with about 40 % of known bird species, more than 60 % of the known mammal, amphibian, and reptile species, and 75 % of known fish species found nowhere else (Sodhi et al. 2004; Lohman et al. 2011). One of the primary mechanisms leading to the rich biodiversity of the region is its complex earth history. On the one hand, Wallacea is characterized by unusually rapid tectonic movements, making it one of the only areas on earth where the distribution of land masses has changed dramatically within the last ~5 mya (Hall 1998, 2002, 2012). On the other hand, Wallacea has endured the repeated appearance and disappearance of land bridges between some of its landmasses during the sequential ~120 m sea level changes in the last 2.5 mya (Voris 2000; Siddall et al. 2003; Caputo 2007), subjecting populations to alternating cycles of isolation and contact, and turning Wallacea into a natural laboratory for evolution and speciation. Despite this, there has been a paucity of biodiversity research conducted in the region and Wallacean biodiversity is in general poorly understood (Lohman et al. 2011).

The White-faced Cuckoo-dove (Turacoena manadensis) (Quoy and Gaimard 1830) is a Wallacean endemic inhabiting Sulawesi and its satellites, including the Togian, Banggai and Sula archipelagoes. These birds are arboreal frugivores (Baptista et al. 2013) and can be found in a wide range of habitats from primary forests to degraded orchards, where they help as seed dispersers of a range of tree and plant species. They have been documented ranging up to 500 m above sea level on Taliabu and to over 900 m on Peleng (Rheindt 2010; Rheindt et al. 2010).

The species was first described as Columba manadensis (Quoy and Gaimard 1830) based on a specimen from Manado, now stored at the Muséum National d’Histoire Naturelle in Paris (Voisin et al. 2005). Bonaparte then divided it off into the genus Turacoena in 1854 (Bonaparte 1854). Forbes and Robinson (1900) allocated populations from the Sula archipelago subspecies status T. manadensis sulaensis based on smaller size and minor differences in plumage (Forbes and Robinson 1900). However, this distinction has not been corroborated subsequently, and T. manadensis is considered by most modern sources to be monotypic (White and Bruce 1986; Coates and Bishop 1997; de By 2002; Gibbs 2010; Clements et al. 2012; Baptista et al. 2013).

We analyzed a dataset of sound recordings of the White-faced Cuckoo-dove collected by us (e.g. Rheindt 2010; Rheindt et al. 2010, 2014) and other field ornithologists, and spanning several island groups including the ranges of both the often-synonymized subspecies sulaensis as well as the nominate. Vocal analyses can be more diagnostic compared to morphology as species delimitation tools in pigeons and doves: many columbid species differ only in minute plumage details from one another but have completely different, innately-coded vocalizations, e.g. the Streptopelia radiation in Africa (de Kort and ten Cate 2001; Johnson et al. 2001; Sinclair et al. 2003), several Patagioenas species in South America (Mata et al. 2006), and several Ptilinopus members in South-east Asia (Rheindt et al. 2011). This makes vocal traits even more important in columbid taxonomy than in oscines. We report on distinct variation in the vocalizations of the White-faced Cuckoo-dove from across its geographic range in Wallacea, which provides new insights into species boundaries within this complex. The objectives of our publication are to (1) document the vocal character differences, (2) combine the differences in vocal trait pattern with Pleistocene earth-historic information to revise species limits in this taxon, and (3) explore the implications this would have on the conservation status of the taxa nestled within the complex.

Due to SM1 having a shorter duration and lesser amplitude than SM2, the signal from SM1 was lost in approximately 30 % of the lesser-quality recordings. Overall, full call data (i.e. SM1 and SM2) were obtained for 24 out of the 34 individuals, comprising 70 % of the full dataset. For the remaining 10 individuals, only SM2 call data were obtained.

To the human ear, birds from mainland Sulawesi, Buton, and Togian Island (henceforth called the nominate group) sound similar to each other, while birds from Peleng and Taliabu (henceforth called the satellite group) sound similar to each other but different from the nominate group (Fig. 3). The SM1 of nominate cuckoo-doves comprises a single, relatively short element (0.06 ± 0.021 s). SM2 comprises a single element that in poorer recordings can sometimes appear bisyllabic. SM2 has a longer duration than SM1 (0.41 ± 0.07 s). Overall, call duration was short at 0.8 ± 0.11 s (Additional file 1: Appendix).

In contrast, the calls of satellite birds comprise more elements in both SM1 and SM2. SM1 of satellite birds comprises 3.78 ± 1.35 elements, while SM2 comprises 6.02 ± 1.83 elements. These differences in the number of elements translate directly into much longer call durations for SM1, SM2 and the full call (SM1 duration of 0.52 ± 0.16 s, SM2 duration of 1.97 ± 0.36 s, full call duration of 2.57 ± 0.39 s; see Additional file 1: Appendix).

Boxplots of means and standard deviations suggest that five parameters differ substantially between the nominate and satellite groups while the remaining eleven parameters do not (Fig. 4; uninformative data not shown). Within the groups (i.e. Sulawesi-Togian and Peleng-Taliabu), no discernible differences were identified.

Principal component analysis supports this distinction, with the PCA plot showing White-faced Cuckoo-dove populations clustering in two separate spatial agglomerations consisting of the satellite birds in a single cluster and nominate birds in a separate cluster (Fig. 5). Principal components 1 and 2 (PC1 and PC2) accounted for 64.57 % of observed variability in the full dataset, with the remaining principal components being much less informative and each accounting for less than 12 % of variation.

We provide documentation of the call of the White-faced Cuckoo-dove from different localities across its natural range in Wallacea: Togian island, Buton island, Sulawesi, Peleng in the Banggai archipelago, and Taliabu in the Sula archipelago (Fig. 1). Two vocal types exist, with deep bioacoustic differentiation separating the nominate populations of Sulawesi/Buton and Togian from satellite populations inhabiting the Banggai and Sula archipelagoes to the east.

The five vocal characters which showed diagnosability are all temporal in nature: the number of elements in SM1, total duration of SM1, number of elements in SM2, total duration of SM2, and total call duration (Fig. 4; Table 1). The remaining three temporal parameters (inter-motif duration, SM1 and SM2 average element lengths) were not informative. None of the eight frequency measures were sufficiently differentiated to be informative (data not shown). This is in agreement with earlier observations of frequency parameters being less informative than temporal parameters in several columbid genera including Turacoena, Ptilinopus, and Streptopelia (Beichle 1989; Slabbekoorn and Ten Cate 1999; Rheindt et al. 2011).

Vocal patterns agree with Pleistocene sea level changes

This observed pattern of vocal differentiation is in agreement with the presence and absence of land bridges between Sulawesi and its satellites during Pleistocene glaciation events (Fig. 6). Despite Peleng being only 14 km distant from Sulawesi, the presence of a 400–700 m deep sea trench between the two has resulted in them never being connected by a land bridge even during Pleistocene glacial maxima (Becker et al. 2009). The deep vocal differentiation between cuckoo-doves from Sulawesi and Peleng suggests that this has been sufficient to prevent gene flow between the Sulawesi and Peleng populations. Conversely, the lack of divergence in vocalization within the nominate and satellite groups suggests that gene flow has occurred when land connections formed between Sulawesi, Buton and Togian (~8 and ~27 km distant) and between Peleng and Taliabu (~80 km distant). These land bridges, which have appeared in 10,000–50,000 year intervals, have resulted in sufficient gene flow to eliminate any vocal differentiation that might have occurred during interglacial periods when sea levels are high. Additionally, these findings support the idea that Wallacean forest birds such as cuckoo-doves and their relatives are weak over-water dispersers, a notion which arose when Alfred Russell Wallace noted the rarity of bird colonizations over narrow sea channels characterized by deep sea trenches (Wallace 1869).

Using highly diagnostic vocalization data, we have demonstrated that two distinct species of White-faced Cuckoo-dove exist across its natural range in Wallacea: nominate taxon T. manadensis from Sulawesi and the Togian archipelago, and satellite taxon T. sulaensis from the Banggai and Sula archipelagoes to the east. These patterns of divergence agree with the presence and absence of land bridges between islands in the region resulting from sea level changes during Pleistocene glaciations. These changes to the taxonomic status of the White-faced Cuckoo-dove have important implications on the conservation statuses of the taxa nestled within.