Improving our capacity to predict biodiversity changes requires a profound understanding of the mechanisms that drive the variation of life in space and time. Over the last decades, ecological research has greatly advanced in documenting spatial patterns of biodiversity at large scales (e.g.7,8) as well as their underlying environmental drivers9–11. Besides spatial patterns, many taxa show characteristic seasonal replacement of species caused by their particular timing of life events, i.e. species’ phenologies. The mechanisms driving phenological patterns of diversity are, compared to those driving spatial diversity, much more complex and poorly understood, which critically limits conclusions about the spatio-temporal changes that underpin species responses to climate change5.
Insects constitute the vast majority of terrestrial animal species, and their direct dependence on ambient temperature for development, reproduction and activity render them particularly vulnerable to temperature changes. In temperate regions, insects show distinct seasonal timing of their activity periods12. Insect phenology results from a complex interaction between developmental constraints and environmental life cycle regulation2–4, in which life cycles are synchronised to optimal seasonal moments where fitness is maximized – a process named phenological fundamental tracking1. Within this system, phenological events such as adult emergence are triggered by environmental cues that link to subjacent drivers of optimal timing1 (Fig. 1a). For instance, emergence of butterflies may be triggered by certain temperature and photoperiod levels2 that align to the seasonal appearance of their host plants13. Most phenological research has so far focused on describing species-specific phenological events in response to environmental factors14, but little is understood about the underlying mechanisms that may determine optimal timing5,15. Shifts in species´ phenologies constitute one of the most obvious consequences of climate change16–18, with many insect groups showing advances in their flight periods. However, we do not understand their consequences for species, which could range from positive to negative – depending on whether species track shifting optimal seasonal conditions19. A failure to recognise the mechanisms underlying phenological patterns of diversity is a key knowledge gap in understanding the impact of climate change on natural systems3,15,20.
Body colour is a key trait for thermoregulation, which is, in turn, a crucial mechanism regulating life cycles and occurrences of ectotherms21. Due to their lower reflectance, darker colours absorb more solar radiation than lighter ones22, making darker bodies heat up faster. As a result, darker individuals and species are able to occur in colder environments than their light-coloured counterparts – a pattern known as Thermal Melanism Hypothesis6 (TMH). TMH is well supported based on patterns of species’ distributions and community composition in a broad range of ectothermic taxa, including reptiles23 or insects such as ants24, Lepitoptera25,26 or dragonflies and damselflies27,28. The possible role of thermal melanism in determining insect phenology has never been assessed even though the environmental factors driving the TMH also characterise seasonality.
Here we test whether thermal melanism contributes determining insect phenological patterns of insects. We use Odonata (suborders dragonflies and damselflies) as study system because of their rich natural history record29. Odonata is one of the groups where the TMH has been most strongly supported, driving functional community assembly across Europe and North America27,28. The strength of thermal melanism in this warm-adapted30 group is related to their high thermoregulatory requirements. Odonates rely on – energetically costly – flight for all their essential activities (displacement, foraging, reproduction), for which particularly the larger-bodied dragonflies require thoracic temperatures above ambient levels (e.g. 27–36°C31). Odonate diversity in temperate latitudes shows characteristic replacement of species´ flight periods over the warm months of the year (Fig. 1d) as a result of environmental regulation of species´ life cycles based on photoperiod and temperature cues together with developmental constraints (Fig. 1c). Whether certain underlying environmental drivers may determine optimal timing of flight periods remains to be understood.
We expected the same mechanisms driving ectotherms spatial diversity patterns to contribute to determining insect flight periods. Specifically, we expected (1) colour lightness of odonate assemblages (CL) to follow predictions of the TMH and show phenological variation in response to seasonal changes in solar radiation intensity and temperature. Furthermore, we expected (2) thermal melanism responses to be stronger in the larger dragonflies than in damselflies, based on their higher thermoregulatory requirements31,32. Following reported phenological advances in odonate flight periods33,34, we expected (3) to see advances in the phenological pattern of CL over the last decades.
Recently available massive observational data allowed us to study spatio-phenological diversity patterns at unprecedented detail and extension. We downloaded a database of over one million odonate records for Great Britain35. After grouping observations within fine spatio-phenological units and controlling for sampling effort by using rarefaction curves, we obtained unique datasets of 8,159 and 4,134 ecologically meaningful assemblages of dragonflies and damselflies, respectively, between May and October from 1990 to 2020 (see Methods for details). Species´ body colour lightness was obtained from scientific illustrations36, see27,28. To quantify assemblage-level body colour lightness (CL), we used community-weighted means37, whose deviations from null expectations were then analysed in relation to the seasonal variation of the thermal environment. We also accounted for potential effects of phylogenetic relatedness as well as spatial autocorrelation (see Methods and Supplementary Fig. S3 and Fig. S5). We finally assessed changes in the phenological pattern of CL over the last 30 years to evaluate potential shifts in response to climate change.
We found colour lightness of dragonfly assemblages to vary as expected – both phenologically and with latitude (4th degree polynomial model: n = 8159, F5,8153 =1147, R2 = 0.41, P < 0.001; Fig. 2a, 2d). CL decreased linearly with latitude as predicted by the THM (Lat: t =–29.66, P < 0.001. Figure 2c), but the phenological effect was much stronger, with most explained variance (hierarchical partitioning: 84.8%) depending on the day of the year (Day: t = 15.55, P < 0.001; Day2: t =–13.8, P < 0.001; Day3: t = 11.92, P < 0.001; Day4: t =–10.11, P < 0.001; Fig. 2d, 2e). CL increased from May until mid-June to early July, and then gradually decreased until the end of August from where assemblages remained constantly dark until the end of the season in October (Fig. 2e). The phenological pattern of dragonfly CL was consistent when applying spatially restricted null models (Supplementary Fig. S1) which allows isolating the phenological component (see methods for details). In contrast to dragonflies, CL of damselfly assemblages did not show latitudinal nor phenological patterns (Linear regression: n = 4134, F2,4131 =1.5, P = 0.22; Fig. 2b, Supplementary Fig. S2). The observed spatio-phenological patterns of CL of both dragonflies and damselflies were robust to potentially confounding effects of phylogenetic non-independence of traits (see methods and Supplementary Fig. S3). Variation in CL of dragonfly assemblages followed expectations from the TMH as it increased non-linearly with solar radiation received (Fig. 3) (2nd degree polynomial model: n = 7901, F2,7898 = 1789, R2 = 0.31, P < 0.001; radiation: t =–14.92, P < 0.001; radiation2: t = 23.14, P < 0.001). Spatio-phenological components and drivers of dragonfly CL were consistent to alternative definitions of assemblages (Supplementary Table S1).
Independent polynomial models of the phenological pattern of CL for each of the 24 years were consistent in their shape and explained between 34 and 48 percent of CL variation (all P < 0.001; Fig. 4a, Supplementary Fig. S4, Table S2). Phenological CL patterns advanced over years by 3.6 days per decade for the day when CL turned lighter than expected by chance (CL > 0) (Fig. 4b; F1,22 =6.12, R2 = 0.18, P = 0.021), and by 3.8 days per decade for the day when CL peaked (Fig. 4b; F1,22 =11.25, R2 = 0.31, P = 0.003). The day when CL became darker than expected by chance (CL < 0) and the length of the period where CL was lighter than expected by chance did not change over the years (Fig. 4b; F1,22 =1.87, P = 0.185 and F1,22 =1.34, P = 0.259, respectively). The magnitude of the peak with maximum CL showed a non-significant positive tendency (Fig. 4a, F1,22=2.58, R2 = 0.06, P = 0.122).
Patterns of darker species occurring at colder locations have so far supported thermal melanism as a crucial mechanism driving spatial patterns of ectotherm diversity6. Our study based on a uniquely high resolution and comprehensive dataset showed that colour lightness of dragonfly assemblages varied phenologically closely linked to seasonal changes in solar radiation, which provides for the first time support for the fundamental role of thermal melanism as a driver of phenological diversity patterns. Seasonal timing of dragonfly flight periods would therefore be optimized based on the relation between prevailing radiation and species´ thermoregulatory performance driven by body colour lightness. Medium and dark colours seemingly allow early and late flight season species to deal with corresponding intermediate and low radiation, respectively, while light colours enable mid-season species to thermoregulate well under highest seasonal radiation conditions.
Our analysis reveals a novel mechanistic explanation driving optimal timing of insect flight periods. One of the few similar studies showed that phenological changes of body sizes of wild bees in Catalonia (Spain) followed Bergmann´s rule, allowing larger species to deal with cold temperatures in early and late in the year39. In our study, radiation was the primary driver of the variation of CL, in line with its direct mechanistic effect on thermoregulation22 and with previous studies on thermal melanism23, although others also support the contribution of temperature27 in combination with radiation24,28.
The phenological operation of the TMH in dragonflies underlines the well-known dependency of adult odonates on thermoregulation31. However, we did not find any support for either spatial or phenological dimensions of the TMH in the suborder of the much smaller damselflies, which could be explained by the generally lower thermoregulatory requirements of small flying insects32. Smaller insect species have lower thoracic temperature requirements, possibly because flight is energetically less demanding in those31. A similar differential trait response driving phenology depending on group physiology was found in39 where only the large – more endotherm – bees obey Bergmann’s rule and not the small, more thermally conformist species. Previous assemblage-level support for TMH in odonates27,28 may therefore be mostly driven by the dragonflies. Our contrasting results between taxonomically closely related taxa stresses the importance of accounting for the fundamental physiological difference of taxa in mechanistic ecological studies.
According to our analyses, the phenological pattern of dragonfly CL advanced over the last decades at rates of almost four days per decade, which aligns with previously reported advances of odonate flight periods of 1.5 days per decade in Great Britain33, or 8.7 in the Netherlands34. In the latter, flight period length of species did not change, similarly to our result for the length of the period above average CL. Phenological advances of dragonfly flight periods imply that temperature, whose seasonal patterns are advancing over years40, plays a primary role for determining dragonfly emergence, either by accelerating development41 or as environmental cue. This relegates radiation, whose seasonality is static over years42, to play only a weak role as an environmental cue triggering dragonfly emergence, even though our analysis suggests that it is the main factor determining optimal flight periods. Poor alignment between the cues triggering seasonal regulation and the drivers of optimal timing may likely prevent species to track shifting seasonal conditions1. Altogether, our results suggest that phenological advances may desynchronize dragonfly flight periods from ideal seasonal conditions. Earlier emergence of spring species may expose them to lower radiation than optimal based on body colour lightness, while summer species may be confronted with higher radiation than optimal. Similar results have been suggested for flowering plants (e.g.43). Species´ can adjust to seasonal shifts by modifying their phenological responses to environmental cues1 through either evolutionary adaptation or phenotypic plasticity mechanisms44,45. Similarly, species´ body colour lightness may respond to changes in climate or photoperiod, either evolutionarily46 or via phenotypic plasticity47,48. The degree to which these mechanisms would be able to compensate observed and potential future phenological mismatches is, however, unknown (see49,50). More research on intraspecific colour variation, plasticity and adaptability in relation to seasonal environmental variation may offer clues about insect´s potential to respond to a changing climate1.
Our study contributes to filling a gap in the comprehension of the essential but vastly understudied dimension of phenology, by providing support for a phenological extension of the TMH can drive ideal timing of ectotherm´s phenologies. Our results, which rely on key mechanisms regulating species occurrences and life histories, may be representative for a broad spectrum of ectotherm taxa and stress the fundamental ecological importance of colour-based thermoregulation in ectotherms. The complexity of the mechanisms driving phenology make predictive consequences of the phenological component of the TMH which we report in this study far more complex than the direct implications of its spatial operation28. However, our results point to body colour as a key trait mediating the mechanisms and repercussions of recent phenological advances, which opens new integrative research avenues that help elucidating general responses of species under global warming both in space and time51. We call for more research that does not only report phenological responses to triggering cues and phenological shifts, but addresses mechanisms determining phenology, including the relative contributions of unregulated and regulated phenological tracking as well as mechanisms behind cue systems across taxa. Integrating the growing availability of massive high-resolution species’ occurrence and environmental data together with increasingly comprehensive trait datasets opens opportunities for unprecedentedly detailed mechanistic insights into the spatio-temporal variation of diversity – now and in the future.