Global changes and animal phenotypic responses: melanin-based plumage redness of scops owls increased with temperature and rainfall during the last century
Supplementary Material
Paolo Galeotti, Diego Rubolini, Roberto Sacchi and Mauro Fasola
(a) Origin of specimens and dataset
We collected data from 281 scops owl specimens dated from 1870 to 2007 from Italian natural history museums (81.1%, see Acknowledgements for details), private collections (16.1%) and bird ringing activities (2.8%). Most of specimens provided by museums or collections were shot or captured birds, and we may consider them as a random sample, representative of the natural occurrence of different colour morphs in the living population. This is because the species has crepuscular habits, and it is unlikely that birds could be collected differentially according to morph type, because morphs are hardly distinguishable under the dim light conditions of their activity period. When available, we recorded sex, age (young of the year or adult), finding locality, and date of collection (month, year). The geographical coordinates of the collection site were recorded for each specimen.
The relevant data for all the 85 specimens included in the analyses (adult birds collected during May-September, which were likely to be breeding; see Methods) are shown in Table S1, and their geographical and temporal distribution in Figures S1 and S2. The temporal interval of the specimens used in the analyses was 1875-2006 (see Table S1 and Fig. S2), with a mean of 0.64 ± 1.03 s.d. specimens per year (range 0-5). We restricted the analyses to the adult birds collected during the breeding period because we speculated that these birds could be affected by local climatic variation more strongly than juveniles, which were highly likely to have been collected during dispersal. Inclusion of juveniles in the analyses could therefore confound the effects of local climate on colour variation.
(b) Measurement of colour variation
We scored plumage colour by measuring the relative dominance of red and grey coloration in four body regions: head, back, underparts (breast, belly and flanks pooled) and wings. If all relevant body regions were prevalently red, a specimen was assigned a score of 4 (i.e. a “dark-reddish” individual), while 1-3 values represented plumage scores for intermediate individuals, and a score of 0 characterized “pale-reddish” birds. Between-observer repeatability of our scoring method was very high (F12,13 = 100.75, p < 0.001, R2 = 0.98), as judged by repeated measurements of 13 individuals by two independent observers. Descriptively, pure red birds had upperparts and underparts largely rufous-cinnamon with limited silvery-grey on outer scapulars; facial disc rufous-chestnut or rufous; grey birds were mainly silvery-grey with some rufous visible near feather-bases on central crown, mantle or inner scapulars; intermediate birds had mainly red-grey mosaic plumages. Specimens kept in museum collections for decades do not fade (Roulin 2003). Indeed, even in skins collected during the nineteenth century the different colour morphs were clearly distinguishable (dark-reddish vs. pale-reddish vs. intermediates).
In the entire sample, pale-reddish birds prevailed (38.4%, n = 281), whereas 35.6% of individuals were intermediates and only 26.0% were dark-reddish. These frequencies slightly varied, though non-significantly, when considering the subsample of breeding adults we used in the analyses (n = 85, pale-reddish: 43.5%, intermediates pooled: 24.7%, dark-reddish: 31.8; χ22 = 3.56, p > 0.10, Fig. S3). The distribution frequency of colour variation was bimodal, as expected for a colour polymorphic species showing two main colour morphs (Fig. S3).
Though we could not record any morphological data from museum specimens, we obtained additional information on colour score and morphological variables (body mass and wing length) from 51 adult individuals (data kindly provided by F. Silvano and S. Fasano) captured during ringing activities at several localities in Central and Northern Italy in recent years (2003-2008). In this sample, there was no significant association between colour score and wing length or body mass, respectively (rs = 0.22, p = 0.13 and rs = 0.05, p = 0.76).
(c) Climatic data
Climatic data for all the years in the period 1865-2003 were kindly provided by Dr. M. Brunetti (ISAC-CNR, Bologna). Data referred to the whole year (January to December), as well as to winter (December to February), spring (March to May), summer (June to August), and autumn (September to November) seasons. Owl specimens of known geographic origin were assigned to one of the three homogeneous temperature regions and to one of the six homogeneous rainfall regions of Italy (see Brunetti et al. 2006 for more details), and the relative yearly and seasonal minimum temperatures and rainfall recorded over a five-year period (starting from the year before specimen collection) were used in the analyses. Temperatures were expressed as anomalies from the long-term 1961-1990 seasonal or annual means for each temperature region and for the whole Italy (Brunetti et al. 2006). Similarly, rainfall data were expressed as the ratios with the 1961-1990 seasonal or annual means for each temperature region and for the whole Italy (Brunetti et al. 2006).
Rainfall data for Sahel were found at http://jisao.washington.edu/data_sets/sahel/022208/. These data were available for only a subset of our study years (i.e. 1898-2004). We considered only the data referring to the five years prior to owl collection. The amount of Sahel wet season (June to October) rainfall (expressed as the Sahel rainfall index) shows strong interannual variability and has profound influences on the ecology of the Sahelian savannahs, by determining marked variation in primary productivity and thus habitat quality for long-distance migratory birds. Sahel rainfall has indeed been shown to influence survival and phenotype of long-distance Afro-Palearctic migrants (e.g. Møller 1989; Peach et al. 1991; Szép 1995; Saino et al. 2004).
(d) Statistical analyses
To investigate the effects of climate on colour variation, we performed an analysis of covariance, where colour index was the dependent variable, sex was a fixed factor, while geographical coordinates and a set of climatic variables were the covariates. To avoid model overfitting, and to avoid multicollinearity among climatic variables and among climatic variables and year (most of the climatic variables were obviously intercorrelated, and there were significant temporal trends of climatic variables in Italy, see Brunetti et al. 2006), we adopted the following procedure. The set of climatic variables that best explained colour morph variation among all the possible combinations of year of collection, the 25 temperature and the 25 rainfall variables (5 years lag x 4 seasons, plus 5 yearly values) was selected by running regression analyses where colour index was the dependent variable and all the possible combinations of one temperature variable, one rainfall variable and the year of collection were the predictors (n = 1351 models, including all models with individual variables or with two variables). These models, built on a subset of the original 85 specimens (n = 83 speciments) that had values for each of the 50 climatic variables, were sorted based on their AIC value, and those variables included in the regression model with the lowest AIC were entered as covariates in the analysis of covariance model described above.
Climatic data for the Sahel were not included in this first selection of climatic variables because the number of specimens with data for each of the 5 yearly Sahel rainfall index values was small (n = 59 cases). Therefore, to identify the temporal lag at which the Sahel rainfall index explained most variance in the colour index, we added each of the 5 yearly Sahel index values in turn to a regression model where the best subset of climatic variables (identified according to the procedure described above) were the predictors. These models were run on an even smaller subset of specimens that had values for all the Italian and Sahel climatic variables (n = 58 specimens). The Sahel rainfall index yearly value that resulted in the lowest AIC was identified as a potential additional predictor to the minimal adequate analysis of covariance model (see below).
The initial analysis of covariance model was subjected to a step-down simplification procedure, where non-significant (p > 0.05) terms were dropped at each step starting from the least significant terms until a minimal adequate model, containing only significant (p < 0.05) terms, was obtained (Crawley 1993). The effect of the Sahel rainfall index was tested by adding this predictor to the minimal adequate model, and was thus not included in the model simplification procedure. This was done because this variable was available for only a small subset of cases (see above).
Although the dependent variable had a clear bimodal distribution (see Fig. S1), the residuals of the analysis of covariance models were normally distributed (Kolmogorov-Smirnov test, all p-values ³ 0.10), indicating that a normal error distribution and an identity link function (as assumed by the analysis of covariance) were appropriate to model the colour index data. This was confirmed by a simple randomization test (Manly 1997). Randomization tests are distribution-free statistical tests that can be very useful in cases of data with complex distributions (Manly 1997). Briefly, the coefficients obtained from the model run on the observed data were compared with the frequency distributions of coefficients that were obtained by randomly shuffling colour score and rerunning the regression calculations 9999 times. Based on these frequency distributions, we then calculated the probability to obtain a value more extreme than the observed ones by chance alone. For the final model, the p-values obtained by the randomization test were as follows: year of collection, p = 0.030; temperature of the year(n-3), p = 0.006; summer rainfall of the year(n-1), p = 0.015; these p-values were highly consistent with those based on standard F-tests (see Results).
e) Test of morph-specific range shifts
We argued that, if the temporal increase in plumage redness was at least partly due to differential shifts in the geographical distribution of the colour morphs, this should have resulted in differential temporal variation of geographic coordinates of specimen collection according to colour morph. We tested this possibility by running an analysis of covariance where latitude was the dependent variable while colour index, year and the interaction between colour index and year were the covariates. There was no differential temporal trend of the latitude of specimen collection according to colour index (colour index x year, F1,81 = 0.01, p = 0.94). The same result was obtained if longitude was tested as a dependent variable (effect of colour index x year, F1,81 = 1.19, p = 0.28). Qualitatively similar results were obtained if the colour index was recoded as a three-level factor (pale-reddish, intermediate colour, dark-reddish; latitude, effect colour morph x year, F2,79 = 0.07, p = 0.93; longitude, effect of colour morph x year, F2,79 = 0.29, p = 0.75).
Acknowledgements
We are very grateful to Dr. M. Brunetti for providing climatic data, and to Dr. G. Tavecchia for collecting part of the data. Dr. R. Ambrosini helped us with the randomization test. F. Silvano and Dr. S. Fasano kindly provided additional data on scops owl morphology. Dr. A. Roulin and two anonymous referees provided constructive comments on an earlier version of the manuscript. We are also indebted to a number of public and private collections and their curators, without whose assistance this study would not have been possible: Museo di Storia Naturale, Università di Pavia (Dr. E. Razzetti); Museo Civico di Storia Naturale, Milano (Dr. G. Chiozzi); Museo Friulano di Storia Naturale, Udine (Dr. C. Morandini, R. Parodi); Museo Regionale di Scienze Naturali, Torino (Dr. R. Toffoli); Museo Provinciale di Storia Naturale, Livorno (Dr. E. Arcamone); Museo di Zoologia, Università degli Studi di Palermo (Dr. M. Sarà); Museo Civico, Rovereto (TN) (Dr. F. Finotti); Museo Ornitologico F. Foschi, Forlì (Dr. S. Gellini); Museo Civico di Storia Naturale, Venezia (Dr. M. Bon); Museo Civico di Scienze Naturali, Brescia (Mr. P. Brichetti); Museo di Zoologia “La Specola”, Università degli Studi di Firenze (Mr. Fausto Barbagli); Museo Civico di Storia Naturale, Genova (Dr. G. Doria); Museo di Zoologia, Università degli Studi di Napoli (Dr. M. Fraissinet); Museo Civico di Zoologia, Roma (Dr. C. Marangoni); Istituto Nazionale per la Fauna Selvatica, Ozzano Emilia (BO) (Dr. L. Serra); Museo Civico di Storia Naturale, Pordenone (Dr. U. Chalvien); Museo Tridentino di Scienze Naturali, Trento (Dr. P. Pedrini); Museo Regionale, Terrasini (PA) (Dr. V. Orlando, Dr. G. Viviano); Collezione Pazucconi, Pavia (Mr. A. Pazucconi); Museo Ornitologico e di Scienze Naturali, Ravenna (Ms. G. Grillanda, A. Di Girolamo).
Brunetti, M., Maugeri, M., Monti, F. & Nanni, T. 2006 Temperature and precipitation variability in Italy in the last two centuries from homogenised instrumental time series. Int. J. Climatol. 26, 345-381.
Crawley, M. J. 1993 GLIM for Ecologists. Oxford: Blackwell Science.
Manly, B. J. F. 1997 Randomization, bootstrap and Monte Carlo methods in Biology, 2nd Edition. London: Champan & Hall.
Møller, A. P. 1989 Population dynamics of a declining swallow Hirundo rustica L. population. J. Animal Ecol. 58, 1051–1063.
Peach, W., Baillie, S. & Underhill, L. 1991 Survival of British Sedge Warblers Acrocephalus schoenobaenus in relation to West African rainfall. Ibis 133, 300–305.
Roulin, A. 2003 Geographic variation in sexual dimorphism in the barn owl Tyto alba: a role for direct selection or genetic correlation? J. Avian. Biol. 34, 251–258.
Szép, T. 1995 Relationship between West African rainfall and the survival of Central European Sand Martins Riparia riparia. Ibis 137, 162–168.
Saino N., Szép, T., Romano, M., Rubolini, D., Spina, F. & Møller, A. P. 2004 Ecological conditions during winter predict arrival date at the breeding quarters in a trans-Saharan migratory bird. Ecol. Lett. 7, 21–25.
Figure S1. Geographic distribution of scops owl specimens (n = 85) used in the analyses. Small dots = 1-2 specimens; medium dots = 3-4 specimens; large dots = 5-6 specimens. Distance between the northernmost and southernmost specimen was 1,072 km.
Figure S2. Temporal distribution of scops owl specimens (n = 85) used in the analyses.