The evidence for range shifts as a way of coping with ongoing climate change is overwhelming (Parmesan 2006), and examples of phenotypic changes in relation to climatic variation abound. The latter include changes in phenology (reviewed in Parmesan 2006) as well as in morphology (Millien et al. 2006). Among phenotypic traits taken into account when studying animal responses to climate changes, intraspecific colour variation (i.e. colour polymorphism (CP); Gray & McKinnon 2007) has received little attention, except for two recent studies (Cameron & Pokryszko 2008; Lepetz et al. 2009). This is surprising, since CP is a widespread phenomenon in many animal taxa, and colour morphs can be considered as phenotypic genetic markers whose fitness effects are related to various ecologically important factors (Roulin 2004). Colour morphs may have directly evolved under both natural and sexual selection and also as an indirect response to selection exerted on genetically correlated attributes, e.g. on genes that regulate both melanogenesis and other physiological processes (Roulin 2004). Thus, CP may be appropriate for studying short- and long-term changes in gene frequencies under various sets of environmental conditions (e.g. Sinervo & Lively 1996).

A comparative study suggested that polymorphic species of owls showed a wider niche than monomorphic ones, since they frequented many different habitats, both open and closed, lived in seasonally alternating dry/wet climates and were active during both day and night (Galeotti & Rubolini 2004). Such results suggest that different colour patterns may be adaptive in different environmental conditions by providing behavioural or physiological advantages to their bearers. For example, in the Italian tawny owl (Strix aluco) populations, dark-reddish birds may suffer greater mortality in cool-dry years while being favoured in warm-wet conditions (Galeotti & Cesaris 1996). This may occur because of differences in thermoregulatory physiology among morphs (Mosher & Henny 1976). Therefore, the prevalence of dark- or pale-reddish morphs in a given population may reflect adjustments to local environment in this species, i.e. local adaptation.

If the fitness of different morphs differs between habitats, then polymorphism can be established with different equilibrium gene frequencies in different habitats or in the same habitat under different conditions. This equilibrium between morphs (and in gene frequency) may be disrupted or shifted by environmental changes, among which climate changes may play a major role through their direct and indirect effects on fitness (Lepetz et al. 2009). Differential survival of morphs because of climate variation could thus lead to an evolutionary change in the morph ratio.

Here, we examined long-term variation in the plumage colour of the scops owl (Otus scops), a small (60–135 g) nocturnal raptor of the Mediterranean region (Cramp 1998), in relation to climatic factors. Similar to many related owl species (Galeotti & Cesaris 1996), scops owls show two main colour morphs that are independent of sex and age, ‘dark-reddish’ and ‘pale-reddish’ (the latter often termed ‘grey’; e.g. Cramp 1998), with the former morph probably having more pheomelanin and eumelanin in feathers than the latter (Gasparini et al. 2009). However, intermediates are frequent (see the electronic supplementary material), and CP in this species should be considered as ‘continuous’ (see Huxley (1955) for a definition of continuous morphism and Roulin (2004) for a discussion). The trait may thus vary continuously between two extreme colour values as an outcome of codominance, multigenic control or variation in gene expression (Buckley 1987).

The Akaike information criterion (AIC)-based procedure we adopted to select potential climatic predictors of CP (electronic supplementary material) identified a combination of the year of collection, minimum annual temperature of year_(n−3) and summer rainfall of year_(n−1) as the subset of variables best explaining colour index variation. The Spearman correlation coefficients between climatic variables and the colour index were, respectively, 0.30 and 0.26 (both p < 0.019, n = 85 and 83). The Sahel rainfall index having the lowest AIC value when added to the subset of these three variables (electronic supplementary material) was at year_(n−4) and had an r_s-value of −0.14 (p = 0.28, n = 60) with the colour index.

The temporal change in plumage coloration we detected is compatible with a shift in balance of selective pressures favouring dark-reddish morphs. The spread of dark-reddish birds in the population from the early twentieth century was related to a concomitant increase in the annual minimum temperature of year_(n−3) before specimen collection, a lag that is consistent with selection taking some time before displaying its effects (Ridley 2003). Such lag may reflect a decreased fecundity and survival of pale-reddish birds born in warmer years. Actually, between 1865 and 2003, the minimum annual temperature increased significantly in Italy (1.1°C per century), while the mean annual rainfall remained stable (Brunetti et al. 2006). Thus, a warmer climate might have favoured dark-reddish individuals, which may have a colour-linked susceptibility to cold (Mosher & Henny 1976; Gehlbach 1994), while concomitantly disadvantaging pale-reddish individuals. Climatic effects on morph frequencies may also be mediated by a differential parasite susceptibility of colour morphs under changing environmental conditions (Galeotti & Sacchi 2003). In both these cases, selection would be indirect, acting on physiological attributes (e.g. metabolic rate, immune response, resistance to stressors) that differ among colour morphs rather than on coloration per se. In fact, melanin-based coloration is often associated with variation in physiological and behavioural traits, stemming from the pleiotropic effects of genes regulating the synthesis of brown to black eumelanin (Ducrest et al. 2008).

However, the significant temporal trend, independent of climatic factors, suggested that other environmental variables varying over time could have simultaneously been at work to produce the morph shift towards a dark-reddish coloration. The increase in woodland extent in Italy since the early twentieth century (Piussi 2005) is a likely candidate, and could have favoured survival of dark-reddish birds because in this habitat they are more cryptic than pale-reddish ones. Several studies have documented that dark- and pale-reddish colour morphs are differently cryptic depending on habitat background and light conditions, and dark-reddish birds may be particularly cryptic in closed forests (e.g. Majerus 1998). We might thus speculate that dark-reddish scops owls in Italy have been favoured both by the increase in temperatures and by the concomitant increase in vegetation cover over the last century.

Three alternative hypotheses may account for the observed temporal change in plumage colour. First, it could be owing to differential range shifts of the different morphs. However, this should have resulted in morph-specific temporal trends in the geographical coordinates of specimen collection, which were not observed in the present sample (electronic supplementary material). Second, any change in morph frequencies might be explained by selective emigration (Lepetz et al. 2009) of pale-reddish individuals from the population or selective immigration of dark-reddish individuals, following range shifts occurring in nearby European populations. An extension of these analyses to the entire Palearctic population of scops owls could provide additional insights on this possibility. Finally, directional temporal changes in colour variation in this species might be regarded as a micro-evolutionary response to climate and habitat changes over time, with the dark-reddish morph being in the process of replacing other morphs. If the current trend persists, we may predict the extinction of pale-reddish morphs among Italian scops owls (i.e. the evolution of a monomorphic population, as occurs in females Tetrao tetrix, which are monomorphic at either end of their range; Stegmann 1932).