Study system
The study was performed from 2012 to 2018 in the Hoya of Guadix-Baza, Granada, southeast of Spain (37º18’N, 3º11’W). The area is an extensive agricultural landscape with scattered holm oaks (Quercus ilex) where cork-made nest-boxes have been set up to palliate the low density of natural holes and favour the reproduction of medium sized hole-nesting birds (see details in (50)).
The scops owl is a medium-sized nocturnal and trans-Saharan migrant owl arriving into the study area in April (51,52) and starting its reproduction throughout May (52). Scops owls make one clutch per year of about 2-6 eggs that are laid every 1-3 days. Females start incubating after laying the second egg, and incubation takes 24-25 days (53). Nestling rearing takes 21-29 days on average (51).
Sampling procedure
Every year, starting at the last week of April, nest-boxes are visited once a week until egg-laying is detected. After detection of a breeding attempt, nests are visited once more after the end of laying, and only once again just before the estimated hatching date to avoid nest desertion. After owlet hatching, nests are visited weekly to record reproductive parameters.
Capture of adults for this study was done by hand, while sleeping at nests in the case of incubating females, and, in the case of males, with nest-traps while they were delivering food to offspring (44). This capture methodology has a negligible effect on nest desertion in this species (44). All individuals were metal ringed and sexed based on inspection of the brood patch (only present in females). Moreover, upon capture, all adults were photographed for colour assignment, blood samples extracted and feathers collected for assessment of corticosterone profile and some of the behavioural traits measured (see below).
Colour characterization
We systematically took two standardized photos for each captured individual: one head-on, in which we could observe head and breast plumage; and other to the back part in which we observed the back and wings. Photographs were taken using a digital camera (Canon EOS 1300D, Lens: EF-S 18-55 IS II) mounted on a tripod at a constant distance of 50 cm and with a flash (aperture: 4.5, shutter speed: 1/200, ISO: 800). Owls were gently fixed with a harness inside a neutral-coloured box that ensured stable light conditions and with the head placed next to a colour chart (X-Rite ColorChecker® Passport). Photos were standardized using the Adobe® Photoshop Lightroom 6 plugin and used to determine coloration by focusing on redness extension at the head, breast and wings–back. Each body part was scored among 1 to 3 points depending if they were predominantly greyish or reddish (44). Previous results have shown that scores of the three body parts are highly correlated within individuals and that scores assigned by different observers on the same individual are highly repeatable (44), hence scores of the three body parts were summed to get an individual score for every bird (ranging from 3 to 9). Based on this scores owls are classified as grey, intermediate and brown given that the frequency of scores is trimodal in our population (44). Pigment analyses have revealed that although eumelanin is the most abundant pigment in scops owl feathers, redness is related to phaeomelanin: the higher the score the larger the amount of phaeomelanin pigment in head and breast feathers (43). However, most of variation in phaeo-melanin content in head and breast feathers occurs between brown and grey morphs, and not between these two morphs and the intermediate one (43). Hence, in a first step we use colour scores in a continuous way to characterize variation in plumage coloration of scops owls, but also compare between grey and brown morphs (i.e. excluding intermediate individuals) to address the specific role of phaeomelanin-based colour variation.
Behavioural traits
Male territoriality
Territoriality was measured in 30 males from 2014 to 2018 by recording behavioural responses to a simulated territorial intrusion made by broadcasting calls of foreign males near the nest. All trials were conducted when clutches were completed and between nightfall and 01:00 a.m., when owls were expected to be more active. Territorial intrusions were simulated by broadcasting calls of a male scops owl with a MP3 player (takeMS MP3 Player “Deseo”) connected to a speaker (MOLGAR 3” 20W 4 ohm) placed under the closest tree to the target nest. Broadcasted records consisted of an initial 2 minutes silent track as an acclimation period, followed with a 2 minutes track with male territorial calls followed by, another 10 minutes of silence track and a final territorial call track of 2 minutes of the same male. To avoid recognition by familiarity, territorial tracks came from 3 unknown males to our population that were randomly chosen for each territorial intrusion simulation. Territorial tracks were extracted from xeno-canto (https://www.xeno-canto.org/). Male territorial behaviour was measured using two different variables: 1) Latency of response to the playback was measured as the time in seconds from broadcasting to the first male hooting response; 2) Duration of response to the playback, measured as time lasted in seconds from the first to the last male hooting response.
To confirm that individuals responding to the playback were the owners of the territory and not neighbours or floaters, 9 breeding males were captured and deployed with radio transmitters tags (PIP Ag392 de Biotrack Ltd., Wareham, UK) the night before the intrusion experiment in 2016. Tags with a weight of 1.10-1.90 g, were attached with cyanoacrylate glue onto the feathers of the back which is below the 2.5% of the adults’ weight threshold suggested by Rodríguez_Ruiz et al. (54). Individuals were located by means of receivers Yaesu FT-290R II antennas (frequency range of 150 MHz). All the individuals hooting back to the simulated intrusion carried the transmitter suggesting that they were the owners of the territory. The 9 males were re-captured the night after the experiment to remove the tag without any apparent harmful effect. None of the nests owned by these males were abandoned after tag deployment.
Female response to researchers’ visits
Response to researchers was measured in 47 females based on standardized video recordings (video camera Sony DCR-SR32) made at the nest boxes during the day, when females usually sleep. Behaviour was recorded during 20 seconds inside the nest box after the careful opening of the roof, while slowly approaching the camera to the female, and 10 seconds more while holding it in the observers´ hand after its capture. From the recordings, females were classed in two different categories: When females clicked the beak, hissed, swelled their body, laid on their back with claws raised, grabbed with bill or claws and/or tried to get away through that 30 seconds, were classified as proactive. Females feigning death in the nest and in the hand and not exhibiting any aggressive behaviour were classified as reactive.
Breath rate
Breath rate, estimated as the number of breast movements during 30 seconds, was measured from 2015 to 2018 in 63 females and 44 males as a measure of individual response to handling stress (55).
Parental care
We measured parental provisioning in most scops owl nests from 2012 to 2018 (98 nests) at the beginning of the chick-rearing period (3 days after the hatching of the last egg). Parental activity inside nest-boxes was recorded at night for at least 60 minutes using infrared cameras (KPC- S500, black and white CCD camera, Esentia Systems Inc.). Upon capture females were marked with a white Tippex spot on the head that allowed their identification in recordings. In subsequent visits to the nests and in video recordings we did not find any apparent effect of these marks on females.
From recordings, we determined: 1) latency of entering the nest-box in minutes after setting the microcamera, and, 2) adults’ feeding rates as the number of prey per hour.
Repeatability of behaviour
Our study design, where individuals were not assayed several times for the same behaviour in the same context (either because many birds do not return into our population from one year to another; or because the number of individuals with repeated observations varied according to the behaviour analysed due to logistic problems), does not allow providing a sound test for individual repeatability in behaviour (56). To account for this potential limitation, we measured repeatability in those individuals in which we had measured the same behavioural trait in more than one year (male territoriality n = 15; breath rate, males = 15 and females = 11). Repeatability of female response to researchers’ visits was not analysed because in the few repeated females the trait was measured in every individual in different hours of the day potentially conditioning the test. Moreover, parental repeatability was not estimated either because number of nestlings varied between years, which could have affected this behaviour. Analyses of covariation were subsequently conducted only on repeatable behaviours. We are aware that this might have rendered unrealistic low repeatability for these behaviours, and, therefore that our analyses should be considered conservative in this sense. However, we think that this approach is still highly valuable to study phenotypic integration in relation to phaeomelanin coloration because we targeted on evolutionary relevant behaviours for the studied species, rather than focusing on stereotyped behaviours against a stimulus that can hardly be interpreted in the context of reproduction.
Corticosterone levels
Adrenocortical response to stress was determined through stress-induced blood CORT levels and CORT deposited in feathers. In 28 breeding females from 2013 to 2015, 3 heparinized capillary tubes (225µl) were collected by puncturing the brachial vein 30 minutes after capturing in order to measure stress-induced CORT levels. We decided not to measure stress-induced CORT levels in males because we were unsure about the time it took males to enter traps and hence cannot estimate the time in the CORT increase curve in which the blood is extracted. Upon capture, we also collected the third covert feather of the left wing of adult individuals (27 males and 43 females) from 2012 to 2015 to determine CORT in feathers. Feathers were kept in hermetic plastic bags until analysis. The two CORT measures reveal different information. Stress-induced CORT reflects individual’s physiological state at the instant of sampling, and meanwhile CORT in feathers provides information about longer time periods, because the deposition of CORT in feather parallels its growth (57), showing thus the individual stress during the growing of feathers. Both measures are not correlated for our sample (r = -0.093, P = 0.673, N = 23). Blood samples were kept in coolers until plasma separation, which were done the same day by centrifugation at 368 g during 5 minutes. Plasma was stored at – 20 ºC until analyses.
Hormonal analyses were performed by ME at the Centre d’Etudes Biologiques of Chizé, France. Concentration of CORT in plasma samples was determined using an ethyl ether extraction technique following Lormée et al. (58). CORT levels in feathers were estimated using the method described by Bortolotti et al. (57), where a methanol-based extraction technique was used to extract CORT from feathers. Radioimmunoassay was used to measure the CORT extracts of plasma and feathers (58), with a highly cross-reactive rabbit anti-mouse antibody from Sigma (C8784). Different plasma steroids cross-react with the corticosterone antiserum: 11-deoxycorticosterone 20%, progesterone 15.7%, 20α-hydroxyprogesterone 8.8%, testosterone 7.9%, 20β-hydroxyprogesterone 5.2%, cortisol 4.5%, aldosterone 4.4%, cortisone 3.2%, androstenedione 2.6%, 17-hydroxyprogesterone 1.8%, 5α-dehydrotestosterone 1.4%, androsterone <0.1%, estrone <0.1%, estriol <0.1%. Overnight incubation of the extracts was done with 3H-corticosterone and antiserum at 4ºC. After centrifugation, a liquid scintillation counter was used to count the bound fraction in the aliquots of the extracts. Samples were frozen at -20ºC unless they were assayed the same day. The detection limit of the method was 0.28 ng/mL (lowest measure was 1.23 ng/mL). Although CORT in feathers was quantified in ng/mL, values were transformed to ng/mm for which feathers length (without calamus) were previously measured with a calliper to the nearest 0.1 mm. For plasma samples measured from 2012 to 2014, the intra- and inter-assay coefficients were 9.71% and 11.01%, respectively, and 6.18% and 10.29%, respectively, for those analysed in 2015. For feather samples analysed between 2012 and 2014 intra- and inter-assay coefficients of variation were 7.03% and 8.83%, respectively, whereas in 2015 samples, these values were respectively 8.82% and 12.45%. These intra- and inter-assay coefficients of variation were determined by distributing a minimum of two duplicate samples of commercial rabbit plasma. Intra-assay coefficients of variation were calculating as the average of coefficients of variation of each assay. These coefficients were getting as the division of the standard deviation by the average value and multiplied by one hundred. The Inter-assay coefficients were calculating at the same way but getting the standard deviation and the average of all assays at the same time.
Statistical analysis
Analyses were fit using SAS 9.3 software (SAS Institute Inc., Cary, NC).
We estimated repeatability for the subset of individuals with repeated behavioural samples in different years by performing a linear mixed model following methods described in Nakagawa and Schielzeth (59), analysing variance component with the trait measure as the dependent variable and the individual ID as the random intercept. This allowed us to obtain among-individual variance and within-individual variance that are used to estimate repeatability following Lessells and Boag (60). A behaviour was considered as repeatable whenever among individual variance was significantly higher than within individual variance (i.e. P<0.05), which is a reasonably assumption given low repeatability of behavioural traits (see (61)). Non-repeatable behavioural traits were not considered in subsequent analyses based in one observation randomly selected per individual (55 males and 137 females).
General linear models were run to investigate the relationships between the individual colour morph and continuous behavioural traits (i.e. latency of response of males to territorial intruders, breath rate, latency to enter the next-box and feeding rates). Behavioural traits were entered in models as dependent variables and the colour score, the year (as a categorical variable with seven levels) and its interaction with colour morph were included as explanatory fixed terms. The date when the behavioural trait was measured was introduced in models as a covariate to account for possible variation in individual quality through the season. Also, we entered filming duration (in minutes) and filming time (minutes until sunset) as two further covariates in models of parental care (i.e. latency to enter the next-box and feeding rates) to control for their possible influence. In addition, we also included the colour morph of the mate as a fixed factor in the models to take into account the fact that the behaviour of a member of a couple could be modulated by the behaviour of its mate and hence by its plumage coloration. Finally, brood size was included as a further covariate in models as having large broods may change parental investment and several studies have shown a covariation between aggressiveness and increased parental investment (62–64).
In addition, a generalized linear model was used for analysing females’ response to researchers as a binomial dependent variable (as proactive vs reactive). We replicated the same model structure than that for continuous behavioural traits, but included in the model the hour of the day (as time in minutes until sunset) when the response was measured and the interaction between colour morph and hour of the day as additional fixed factors to account for the fact that we captured females during the day at different hours and that females differing in colour may differ in activity rhythms along the day (see introduction).
We tested for differences in the levels of CORT in blood and feathers in relation to colour morph using General Linear Models. In a first analysis, stress-induced blood CORT of females was included as the dependent variable, the colour morph, the year and its interaction as fixed factors, and date and hour of the day as covariates. Hour was included to control for variations in CORT levels due to circadian rhythms (65). We also included the interaction between the colour morph and hour of the day to account for temporal variation in CORT levels associated to melanism (66). To analyse whether CORT in feathers differ with colour morph, we performed a model including sex, individual colour, year and the interactions as fixed factors.
Those models where colour scores explained variation in the behaviour and/or hormonal profiles of scops owls, were re-run including the colour morph (greyish vs reddish, excluding individuals of intermediate scores) as explanatory variable to address the specific roles of colour variation due to phaeomelanin (see above).
Standard model validation graphs (67) revealed that model assumptions of homogeneity of variance and normality of residuals were fulfilled after corticosterone in feathers and feeding rates and latency to return to nests were inverse-transformed, and log-transformed, respectively. P values smaller than 0.05 were considered significant. Given that the number of cases differed for the different explanatory variables in each model (i.e. we did not always capture the two adults in a nest or some of the individual measurements were not taken due to logistic problems (see table 1 appendix), besides saturated models we report reduced models based on backward stepwise procedure. Non-significant variables are left in the reduced models when they participated in a significant interaction. Pairwise differences in significant models were checked by comparisons of least-squared means of each treatment.