Zosterops, O’Connell & Kelly & Lawless & O’Brien & Marcaigh & Karya & Analuddin & Marples, 2019

Zosterops View in CoL sp. nov.

The ‘Wangi-wangi white-eye’ is strongly distinct from all Z. consobrinorum populations (ND2: 6.23% and COI: 8.35% at a minimum) and all Z. chloris populations (ND2: 5.24% and COI: 7.17% at a minimum). The most closely related populations are Z. griseotinctus (ND2: 5.08%) and the lowland white-eye, Zosterops meyeni (Bonaparte, 1850) (COI: 6.78%). The ‘Wangi-wangi white-eye’ shows minor within-group variability (ND2: 0.29%, COI: 0.16%).

MOLECULAR SPECIES DELIMITATION

Automatic Barcode Gap Discovery (ABGD) analysis finds the barcoding gap between Zosterops species to be 3.5% (COI) and 1.3% (ND2) K2P genetic distance. For both genes, ABGD groups individuals from our focal Zosterops populations in Sulawesi into four putative species;

1. Zosterops chloris from mainland south and south-east Sulawesi, the continental islands of Buton, Muna, Kabaena and Wawonii and Runduma Island (ND2 hapCH01-08, COI hapCH01-03; Supporting Information, Table S1) .

2. Zosterops c. flavissimus from the Wakatobi Islands (ND2 hapCH09-12, COI hapCH04-06).

3. Zosterops consobrinorum – all sampled individuals (ND2 CO01-13, COI CO01-04)

4. Zosterops sp. nov. ‘Wangi-wangi white-eye’ – all sampled individuals (ND2 hapCX01-03, COI hapCX01-02).

MORPHOMETRIC ANALYSES

A total of 752 Zosterops individuals from 11 islands were measured for these analyses; 575 Z. chloris , 139 Z. consobrinorum and 38 ‘Wangi-wangi white-eyes’ (Supporting Information, morphometric trait summaries, Tables S8–S11). The full morphometric database is available at https://figshare.com/articles/ SE_ Sulawesi _ Zosterops _morphology/7998299/1. For analysis, thesampledindividualsweregroupedalongthe splits provided by the molecular phylogenies. Zosterops chloris individuals were classified into the groupings: mainland (Sulawesi mainland and the continental

2.5 2.0 1.5 1.0 0.5 0 Ma islands N = 168), Wakatobi (Z. c. flavissimus from the six Wakatobi Islands, N = 362) and Runduma (N = 45). Zosterops consobrinorum individuals were split into mainland Sulawesi (N = 48), Buton and Muna (N = 68) and Kabaena (N = 23) groups. ‘Wangi-wangi white-eyes’ (N = 38) were analysed with Z. consobrinorum to establish the level of separation between them.

For Z. chloris morphometrics, PC1 (78% of the variance) and PC2 (8.3% of the variance) had eigenvalues> 1 and were carried forward for analyses. PC1 was loaded equally between the seven morphometric traits, giving a general indicator of body size (Supporting Information, Table S12). PC2 was largely loaded by bill length and skull length, giving a general indicator of bill to skull ratio. The Z. chloris populations are significantly different from each other in body size (PC1, ANOVA: F 2, 572 = 554.5, P <0.001), with the mainland, Wakatobi and Runduma populations all significantly different from each other (Tukey HSD, P adj. <0.001 for all comparisons). Runduma individuals are the largest, followed by mainland individuals, with Wakatobi Z. c. flavissimus individuals being the smallest ( Fig. 6 View Figure 6 ; Supporting Information, Tables S8, S9). Zosterops chloris populations also significantly differ in bill to skull ratio (PC2, ANOVA: F 2, 572 = 17.56, P <0.001), with the Runduma population differing from mainland (Tukey HSD, P adj. <0.001) and Wakatobi (Tukey HSD, P adj. <0.001) populations ( Fig. 6 View Figure 6 ). Mainland and Wakatobi populations do not differ for PC2. This strong difference in the Runduma population in bill to skull ratio (PC2) is likely due to Runduma birds having the longest bill of any of the Z. chloris populations measured (Supporting Information, Tables S8, S9).

For Z. consobrinorum and ‘Wangi-wangi white-eye’ morphometrics, only PC1 (88.7% of the variance) has an eigenvalue> 1 and is carried forward for analysis (Supporting Information, Table S12). PC1 is equally weighted between all seven morphometric traits and provided a general indicator of body size. The Z. consobrinorum populations and ‘Wangi-wangi white-eye’ differ significantly in body size (PC1, ANOVA: F 3,173 = 918.1, P <0.001) ( Fig. 7 View Figure 7 ; Supporting Information, Tables S10, S11). The ‘Wangi-wangi white-eye’ is larger than all Z. consobrinorum populations (Tukey HSD, P adj. <0.001 for all comparisons). The Z. consobrinorum Kabaena population is significantly larger than both the mainland Sulawesi and Buton/Muna population (Tukey HSD, P adj. <0.001 for both comparisons). The mainland Sulawesi and Buton/Muna population do not differ in morphometric traits.

SONG ANALYSES

A total of 120 Zosterops individuals from seven islands had their song recorded for these analyses: 52 Z. chloris and 68 Z. consobrinorum (Supporting Information, song trait summaries, Tables S13–S16). No ‘Wangi-wangi white-eye’ songs were recorded. An additional three recordings were sourced from xeno-canto: two Z. chloris maxi recordings taken on Lombok (Lesser Sunda Islands; XC166854 and XC166855) and one Z. consobrinorum recording from Buton Island (XC333521). The full song database is available at https://figshare.com/articles/SE_ Sulawesi _ Zosterops _ song/7998353. As with the morphometric analyses, for the song analyses Z. chloris individuals were split into mainland (N = 24) and Wakatobi (Z. c. flavissimus , N = 28) groups, with the addition of a Lombok group (N = 2). Zosterops consobrinorum individuals were split into mainland Sulawesi (N = 11), Buton and Muna (N = 31) and Kabaena (N = 27) groups.

For Z. chloris song, PC1 (39.8% of the variance), PC2 (24.0%) and PC3 (16.4%) had eigenvalues> 1 and were carried forward for analyses (Supporting Information, Table S15). PC1 is most heavily loaded by the number of notes, duration, maximum frequency and bandwidth. PC2 is most heavily loaded by the temporal traits duration and pace. PC3 is most heavily loaded by minimum frequency and pace. The Z. chloris populations differ significantly in all comparisons (PC1, ANOVA: F 2, 51 = 52.89, P <0.001; PC2, ANOVA: F 2, 51 = 6.073, P <0.005; PC3, ANOVA: F 2, 51 = 3.196, P <0.05). All three populations are distinct ( Fig. 8 View Figure 8 ; Supporting Information, Tables S13–S15). The −6 −4 −2 0 2 4

PC1 (body size)

−6 −4 −2 0 2 4 PC1 (body size)

Z. chloris mainland population differs significantly from the Z. c. flavissimus Wakatobi population in PC1 (Tukey HSD, P adj. <0.001) and from the Lombok population in PC1 and PC2 (Tukey HSD, P adj. <0.001 and P adj. <0.05, respectively). The Wakatobi and Lombok populations differ significantly in PC2 and PC3 (Tukey HSD, P adj. <0.01 and <0.05, respectively).

For Z. consobrinorum song, PC1 (41.7% of the variance), PC2 (21.8%) and PC3 (16.2%) had eigenvalues> 1 and were carried forward for analyses (Supporting Information, Table S17). PC1 is most heavily loaded by duration, maximum frequency and bandwidth. PC2 is most heavily loaded by pace and peak frequency. PC3 is most heavily loaded by the number of notes, maximum frequency and bandwidth. The Z. consobrinorum Kabaena and Buton /Muna populations differ significantly in song PC1 (PC1, ANOVA: F 2, 66 = 4.133, P <0.05; Tukey HSD, P adj. <0.05) ( Fig. 9 View Figure 9 ). There are no other significant differences in Z. consobrinorum song.

CLASSIFICATION BASED ON MORPHOMETRIC AND SONG TRAITS

Discriminant Function Analysis (DFA) classification of Z. chloris individuals suggests a close match of morphometric and song traits for the taxonomic groupings identified in our molecular phylogeny ( Table 2; Figs 3 View Figure 3 , 4 View Figure 4 ). The sampling location of the majority of individuals can be accurately predicted from these traits. The ‘Wangi-wangi white-eye is 100% distinguishable in morphometrics from all Z. consobrinorum populations in the DFA analysis ( Table 3). There is only a weak distinction between Z. consobrinorum populations. The Kabaena Z. consobrinorum is the most accurately classified in morphometrics and the Buton/Muna population shows the greatest classification accuracy in song, but both show a large degree of overlap with other Z. consobrinorum populations. The mainland Sulawesi population cannot be accurately classified, particularly with song traits. More Sulawesi individuals are classified as belonging to other islands than to Sulawesi.

TOBIAS SCORING

For the Tobias scoring of phenotypic traits, the Wakatobi Z. c. flavissimus population is compared to the Z. chloris population from mainland south-east Sulawesi and its continental islands and the ‘ Wangi-wangi white-eye’ was compared to Z. consobrinorum . Both Z. c. flavissimus (Tobias score: nine) and the ‘Wangi-wangi white-eye’ (Tobias score: seven) are identified as distinct species. Detailed scoring is provided in the Supplementary Information (Tables S18, S19).