Introduction
Climate change is considered one of the biggest threats to biodiversity today, and many species risk extinction due to a changed climate (Thomas et al.
2004; Parmesan
2006; Cahill et al.
2013). Species interactions make up an important part of biodiversity. Yet, knowledge of how such interactions are influenced by climate and habitat change is comparatively sparse (Lavergne et al.
2010). A change in climate or other environmental conditions may influence the strength of species interactions by relatively rapid plastic responses and by evolutionary changes over generations (Visser and Both
2005; Visser
2008; Altermatt
2010; Singer and Parmesan
2010). For example, if the phenology of an herbivore and its host plants in a seasonal environment is differentially influenced by temperature, a change in climate may lead to changes in the temporal overlap between the herbivore and its hosts (e.g., Singer and Parmesan
2010). As a result, the intensity of the interaction might change, or it may even disappear (Dewar and Watt
1992; Harrington et al.
1999). In herbivores using multiple hosts, climate change may lead to changes in the relative overlap with different hosts and thus to changes in host use. Such changes in interaction patterns are important to study as they influence both population dynamics and selection regimes, and are fundamental to understand how climate change might influence natural communities (Visser
2008).
A clear trend among many temperate species, including birds, plants and insects, during the past decades is that they have started to reproduce earlier during the spring and summer (Walther et al.
2002; Menzel et al.
2006; Parmesan
2007). Butterflies are temperature sensitive and all their life history stages are influenced by temperature (e.g., Dennis
1993; Karlsson and Wiklund
2005). Several studies have observed positive correlations between ambient temperatures during growth and development and date of the adult flight period, with an average advancement around 4 days/°C (Sparks and Yates
1997; Karlsson
2013). Some authors have also documented recent advancements in butterfly phenology in response to a warmer climate (Sparks and Yates
1997; Stefanescu et al.
2003; Menzel et al.
2006; Altermatt
2010; Diamond et al.
2011; Karlsson
2013).
However, recent comparative studies of butterflies in the UK (Diamond et al.
2011) and in Sweden (Karlsson
2013) reveal that shifts in phenology show a profound variation among species, making a more thorough inspection of the phenological responses justified. Previous studies have shown that variation in phenology shifts among butterfly species is associated with several life history traits, including overwintering stage, seasonal appearance, food plant species as well as several other factors, like food availability, habitat, altitude, and latitude (Altermatt
2010,
2012; Diamond et al.
2011; Illán et al.
2012; Karlsson
2013). For example, species overwintering as adults or as pupae tend to advance their phenology more than species overwintering as larvae or in their egg stage (Altermatt
2010; Diamond et al.
2011; Karlsson
2013).
Butterflies critically depend on plants as larval hosts and for nectar, and it is likely that optimal butterfly phenology in many cases strongly depends on the phenology of their host plants. Butterflies and host plants may respond differently to a warming climate, either because they use partly different cues or because their sensitivity to given cues differ (e.g., Menzel and Fabian
1999; Menzel et al.
2001; Parmesan
2007). Moreover, the direct effects of increased availability of CO
2 may affect plant phenology more than the insects that use them as a resource. The relative importance of cues also varies among plants species (e.g., Rathcke and Lacey
1985), which may result in climate-dependent variation in relative abundances of different host species during the period of reproduction and growth of the butterflies (Schweiger et al.
2008). Such differences in reaction norms should lead to changes in species interactions with changes in climate.
Given that butterflies are strongly selected to maximize synchrony with their host plants and that host plants to some extent differ from each other and from butterflies in their response to increased temperatures, we expect butterfly responses to be related to the specific set of host plants that they depend on. For example, Diamond et al. (
2011) showed that butterfly species with a small diet breadth, i.e., with only a few species of larval host plants, have higher advancement rates compared to species with a large repertoire of host plants. It can also be expected that butterfly species that feed exclusively on specific developmental stages of their hosts, e.g., flowers, young fruits, or young leaves, shift their phenology more strongly in response to warming than species that are not restricted to specific developmental stages.
An additional factor affecting plant and animal phenology is geographic location. The effects of latitude have been extensively scrutinized, and due to climate gradients stretching from south to north, growth and reproduction are generally occurring later in the northern parts (e.g., Myneni et al.
1997; Karlsen et al.
2007; Rötzer and Chmielewski
2001; Doi and Takahashi
2008). Butterflies show a relatively straightforward pattern with northern populations flying at later dates (Roy and Asher
2003; Karlsson
2013). However, not only phenology may vary along latitudinal gradients, but also the relative importance of different cues. Such differences would imply that plant populations of the same species along a latitudinal gradient respond differently to climate warming. This may lead to different responses among butterfly populations in order to maximize synchronization. Moreover, many butterfly species depend on multiple host plants, which use partly different environmental cues for start of development and that vary in relative abundance along latitudinal gradients. In combination, these relationships suggest that the realized pattern of host use will be affected by variation in climate, whether it is due to latitudinal differences or to long-term climate change. Such climate effects on host use are likely to be particularly important in butterfly species that are specializing on feeding on a specific phenological stage of their hosts. However, the effects of climate variation on patterns of host utilization in phenological specialists have rarely been studied. Indeed, detailed data on climate-induced changes of insect–host plant interactions over long periods of time are overall very rare (Visser and Both
2005; Singer and Parmesan
2010). One way forward is therefore to explore spatial variation in butterfly–host interactions along the climatic gradients of latitude or altitude.
Here, we review phenological changes in temperate butterflies over the last decades in Sweden and present results from an ongoing project exploring variation in the interaction between one phenological specialist, Anthocharis cardamines, and its multiple host plants along a latitudinal cline representing large variation in climate. More specifically we ask (1) How much has mean flight date changed in butterfly species in general, and in A. cardamines in particular, during the last 20 years in the same geographical area? (2) How well do temporal changes in mean flight dates for these species agree with the spatial trend along a latitudinal gradient? (3) To what degree do life history traits such as voltinism and overwintering stage correlate with changes in mean flight dates of butterflies in general? and (4) How does among- and within-species host plant use in Anthocharis cardamines differs along a latitudinal gradient?
Discussion
There has been a general trend toward earlier flight periods in Swedish butterflies the last 20 years, and Anthocharis cardamines is among the species that has advanced its adult emergence most. Moreover, most Swedish butterfly species follow the typical pattern of later flight dates in more northern populations but this cline is steeper in A. cardamines. This type of correspondence appears to be a general trend as the rate of phenological change over time shows a significant correlation with the degree of change in flight date with latitude. This was true for both the full dataset with all Swedish butterflies as well as for the subgroup of univoltine, pupal diapausers, to which A. cardamines belongs. The results also show the quite intuitive pattern that butterfly species that are bivoltine start reproduction earlier in the year compared to univoltine species. This is most likely because selection in bivoltine species favors individuals that can use a longer period of the favorable season to produce two rather than one generation. In this respect, the early spring flight period of A. cardamines is clearly atypical for an univoltine butterfly in Sweden, occurring on average more than a month earlier than the other species (May 31 as compared to July 5). The early emergence of A. cardamines is very probably a direct consequence of that newly hatch larvae feeds on flowers and developing fruits of early flowering Brassicaceae plants.
During the last decades, there have been substantial phenological changes in a large number of animal and plant species (Walther et al.
2002; Menzel et al.
2006; Parmesan
2007). As the typical direction of change has been an advancement of phenological events, it has been causally linked to recent climate change and in particular the global increase in temperatures (Sparks and Yates
1997; Stefanescu et al.
2003). The results presented here add to this literature. More interestingly, this study and that of Karlsson (
2013) found that the rate of phenological change over time was correlated with the phenological changes across latitudes. This suggests that species of butterflies that show strong latitudinal variation in phenology, presumably due to spatial variation in climate, also tend to show strong effects of changes in climate over time. This correspondence is expected if the adaptations that control butterfly life cycles and phenology include response to aspects of climate that changes in a similar way over time and space, and that populations along the latitudinal gradient respond in similar ways to climatic cues. While temperature is one obvious and important aspect of climate, other cues, such as the photoperiod, will not show this type of parallel change in time and space, i.e., the photoperiod at a given time of year varies with latitude while it is not influenced by temporal changes in climate at any given location. For our particular study system, this pattern suggests that it is reasonable to use the “space for time” paradigm to get a rough idea of how climate is likely to affect the phenology of
A. cardamines and how this might influence its host utilization (Hodgson et al.
2011). Indeed, it seems likely that both temporal and spatial changes in the phenology of
A. cardamines are reflecting strong effects of thermal conditions on the hatching of adults in comparison with other butterfly species. In support of this idea, the flight date of
A. cardamines shows a strong response to ambient spring temperature during pupal development where an increase of 1°C advances flight date with 6.4 days. Mean value for other univoltine butterflies overwintering in the pupal stage is an advancement of 3.3 days/°C (cf. Karlsson
2013).
The regulation of life cycles of temperate insects is typically due to plasticity in relation to seasonal cues such as photoperiod and temperature (Tauber et al.
1986; Nylin and Gotthard
1998). Given the patterns shown here, it seems likely that the part of the life cycle determining adult emergence of
A. cardamines in the spring is highly dependent on temperature. As this species spends the overwinter period in the pupal stage, it is the post-diapause pupal development in spring that will determine when the adults hatch. Hence, variation in adult emergence is likely to be strongly affected by the thermal reaction norms of pupal development. The advancement of spring phenology during the last decades as well as the latitudinal variation is likely to be largely a consequence of plasticity in response to variation in temperature (Gienapp et al.
2008; Merilä and Hendry
2014). However, thermal reaction norms have a genetic basis and may evolve in response to environmental changes. Indeed, recent experimental studies demonstrate that thermal reaction norms of post-diapause development in
A. cardamines varies among populations from different latitudes suggesting that a part of the spatial variation in phenology seen here is due to local adaptation in these thermal reaction norms (Posledovich et al.
2014; Ståhandske et al.
2014). This also indicates that natural selection due to consistent directional change in climatic conditions over time will alter adaptations that are central for the evolution of phenology. From a climate change perspective, such evidence of local adaptation in thermal reaction norms suggests that responses to similar changes in temperatures will differ between regions along latitudinal gradients.
In the field survey examining host plant use of
A. cardamines in three regions along a latitudinal gradient, we documented significant differences among regions in which of the host species that were used for oviposition. Given that the butterfly has strong preferences for plants in a given phenological stage (Arvanitis et al.
2008; this study), it is likely that effects of climate on the temporal overlap between the butterfly and each of the host plant species were important for these among-region differences. Such differences in temporal overlap between the butterfly and the different host plant species in response to latitudinal variation in temperature are to be expected if the thermal reaction norms differ between host plants and between the butterfly and its preferred host plants. It might seem reasonable to assume that phenological specialists, such as
A. cardamines, are particularly sensitive to changes in climate. However, while the butterfly is expected to be under strong selection to match its phenology with the temporal distribution of Brassicaceae flowers in the spring, it is simultaneously strongly selected to be able to use multiple hosts given that the temporal overlap with one given species varies among years (Wiklund and Friberg
2009). As a result, the specialized feeding on the young fruits and seeds of its hosts is combined with the ability to utilize a quite wide host range of Brassicaceae species. Such a notion, that the species can be characterized as a phenological specialist but a host species generalist, is strongly supported not only by our data on latitudinal variation in host use but also by data on between-year variation in host use at a given site. During a 5-year study of the species at one locality in Sweden (the central location in this study), the species oviposited on 16 of the 18 available Brassicaceae species (Wiklund and Friberg
2009). A tentative conclusion is therefore that an assumed sensitivity of herbivores specializing in particular phenological stages of their host plants to climatic variation might sometimes be buffered by an ability to switch host plant species. If such host plant switching does not occur, we should expect very strong selection on consumer reaction norms to match the reaction norms of their resources.
Our results also show that within species, the phenological state and size of the hosts at the time of butterfly reproduction are important for oviposition. For most of the plant species, we found that later-flowering individuals attracted more eggs, although in one of the main hosts,
C. paludosa, early flowering plants were significantly more used for oviposition. These results are important in two respects. First, they provide further evidence that phenological stage is important for butterfly host plant selection and that not only among-species choice but also choices within species are influenced by the phenological stage of the host plant. Moreover, several within-species patterns varied among regions suggesting that the exact temporal overlap between butterfly oviposition and host plant flowering had a strong effect on the realized host use across the climatic gradient described by the latitudinal range and that this overlap differed among regions. This suggests that the effect of climatic variation on host plant phenology, both in space and over time, will be of major importance for the realized host use of
A. cardamines. Second, given that butterfly attack has strong negative effects on plant fitness (König
2014), the documented patterns of butterfly preferences translate to butterfly-mediated selection on plant flowering phenology. Given that butterfly attacks are relatively frequent in host plant populations, our documented patterns suggest that butterfly-mediated selection on plant flowering phenology may differ not only among different host plant species but also among regions within species.