Computational protocol: Rainfall can explain adaptive phenotypic variation with high gene flow in the New Holland Honeyeater (Phylidonyris novaehollandiae)

Similar protocols

Protocol publication

[…] At the time of banding, we measured seven morphological traits: (1) bill length from the tip of the bill to the back of the head (bill-head length); (2) bill length from the tip of the bill to the anterior extreme of the nostril (bill-nostril length); (3) bill depth measured at the base of the bill (bill depth), (4) bill width measured at the base of the bill (bill width); (5) length of the flattened wing (wing length); (6) tarsometatarsal length (tarsus length); and (7) body mass (mass). All traits, excluding mass, were measured to the nearest tenth of a millimeter using callipers. Mass was measured to the nearest tenth of a gram using scientific scales. All measurements were made by SK (N = 386) and SM (N = 117), who measured birds in all six study sites. Variation in measurement between researchers for all traits was not significant (t-test; P > 0.05), and was less than the variation across sites. A second test of measurement error tested morphological variation across sites for each researcher separately, which mirrored the findings for both researchers combined, indicating a low degree of measurement error.Multivariate analysis of variance (MANOVA) with sex as a fixed factor, as well as study site and year, showed that morphological variation differed significantly between the sexes (F = 38.51; P < 0.001; Wilk's Lambda = 0.66; Partial ETA2 = 0.34). No significant interaction effect was observed between sex and either study site or year, indicating that morphological variation between sexes did not change across study sites or years. Examination of trait means for each sex indicated that males had larger measurements for all traits (), consistent with previous observations of sexual dimorphism in P. novaehollandiae (Disney ; Rogers et al. ; Higgins and Peter ; Myers et al. ). Due to this variability between sexes, and anticipated variation in life histories (Greenwood ; Clarke et al. ), we chose to investigate the sexes separately, where possible, for all further analyses. We used MANOVA (SPSS 19.0; IBM® SPSS® Statistics, Armonk, NY) with study site as a fixed factor to test if P. novaehollandiae showed significant morphological variation across study sites. To further explore the potential influence of rainfall on morphological variation, we used multiple regression, with rainfall as the dependent variable, and the morphological traits as the independent variables. If rainfall is influencing morphological variation as we predicted, then a correlation that reflects the predicted interactions between rainfall and morphology should be observed. [...] We genotyped 330 individuals at 10 microsatellite loci: Pn2, Pn3, Pn4, Pn5, Pn8, Pn13, Pn15, Pn22, Pn23, and Pn25 (Myers et al. ). Polymerase chain reaction conditions were as outlined by Myers et al. (). Prior to performing analyses, we tested the suitability of our data for analysis with F- and R- statistics (Hardy et al. ). The results of these tests showed that all of our 10 loci were better analyzed by F-statistics. The number of alleles (NA), expected and observed heterozygosities (HE, HO), and the inbreeding co-efficient (FIS) were calculated for each locus by site and globally for each locus using genepop v4 (Raymond and Rousset ).We carried out tests of linkage disequilibrium for each locus by site using genepop v4. After Bonferroni correction (Rice ), significant departure from linkage disequilibrium was detected for 13 locus pairs across the sites (P < 0.01), although 12 of these pairs were found to depart from linkage disequilibrium at only one site. We followed recommendations by Kaeuffer et al. () and estimated the correlation co-efficient (rLD; Black and Krafsur ) for these locus pairs using Linkdos software (Garnier-Gere and Dillman ) (http://genepop.curtin.edu.au/linkdos.html). The rLD for each locus pair was <0.55 (P < 0.05), indicating a probable distance of greater than 3 cM between loci (Kaeuffer et al. ), which is sufficient distance that any linkage effect does not bias clustering analyses (Pritchard and Wen ); therefore, we retained all loci for further analyses. We tested for Hardy–Weinberg equilibrium within each site using genepop v4. After Bonferroni correction (Rice ), two loci, Pn5 and Pn15, differed significantly from Hardy–Weinberg equilibrium at one and two sites respectively, both showing heterozygote deficiency (P < 0.01). We investigated the effect of these loci on our analyses by comparing results obtained with and without them. Results were consistent in all cases and we concluded that the observed departures from Hardy–Weinberg equilibrium for these loci would most likely not be strong enough to significantly bias results. Therefore, keeping these violations in mind, we retained all loci for further analyses, assuming that the extra information in these loci outweighs any potential biases they may add. [...] Evidence from mark-recapture data of P. novaehollandiae indicates a limited capacity for dispersal (c.a. 110 km max.), shorter than the distances between some of our study sites (Paton et al. ). Museum vouchers and sight records (b), as well as a general survey of the land, suggest that, with the exception of the sea divides, there are no other obvious geographical or environmental discontinuities between our study sites that are likely to disrupt dispersal; therefore, we predict a pattern of isolation by distance. When dispersal is unrestricted in two dimensions, a positive regression slope of FST/(1–FST) on log of distance is expected (Rousset ), and we tested this. We used the program SPAGeDi v1.2 (Hardy and Vekemans ) to calculate pairwise FST/(1–FST) for each sex. We calculated pairwise distance measures as the shortest distance across land between two points (km), assuming dispersal between Kangaroo Island and mainland South Australia only across the smallest sea divide that separates the two (between the eastern tip of Kangaroo Island and the southern tip of Fleurieu Peninsula). Log of distance was subsequently computed from these distances. We evaluated the correlation between log of distance and pairwise FST/(1−FST) for each sex using 1 × 107 randomizations. We accounted for non-independence of distance correlations inherent with matrix data by using Mantel tests (Mantel ; Mantel and Valand ) in the software program zt (Bonnet and Peer ). […]

Pipeline specifications