Variation of thorax flight temperature among twenty Australian butterflies (Lepidoptera: Papilionidae, Nymphalidae, Pieridae, Hesperiidae, Lycaenidae)

Thermal requirements for fl ight in butterfl ies is determined by a combination of external factors, behaviour and physical constraints. Thorax temperature of 152 butterfl ies was monitored with an infra-red thermometer in controlled laboratory conditions. The temperature at take-off varied from 13.4°C, for a female Heteronympha merope to 46.3°C, for a female Junonia villida. Heteronympha merope, an understorey species, had the lowest recorded take-off temperatures, with females fl ying at a much lower thorax temperatures than males. Among the tested butterfl y species, warming-up rate was positively correlated with take-off temperature and negatively with body mass. Wing loading is a major variable in determining the thorax fl ight temperature. Butterfl ies with the highest wing-loadings experienced the highest thorax temperatures at take-off. A notable exception to this rule is Trapezites symmomus, the only Hesperiidae of our data set, which had thorax fl ight temperatures of 31.5°C and 34.5°C, well within the range of the observed butterfl ies, despite a wing load ca. fi ve times higher. The high thorax temperature recorded in J. villida is probably linked to its high fl ight speed. The results highlight the importance of physical constraints such as body size on the thermal requirements for fl ight across a range of butterfl y species.


INTRODUCTION
Butterfl ies, being facultative endotherms (Bartholomew, 1981), rely on both external factors and internally produced heat for the maintenance of their body temperature.The thorax temperature depends on ambient temperature, irradiant heat gained by basking behaviour and heat generated by muscle movements, while heat loss occurs through body irradiance and air convection around the body (Wickman, 2009).Thermoregulation in insects is therefore dependant ultimately on behaviour, either directly through shivering or basking, or indirectly by moving to or away from sunlit patches (Kingsolver, 1985a).Butterfl ies of the genera Papilio, Colias and Pieris, need a thorax temperature between 28 and 42°C for fl ight, while rigorous fl ight is restricted to thorax temperatures between 33 and 38°C (Kingsolver, 1985b;Srygley & Chai, 1990).The lowest recorded fl ight thorax temperatures are found in large species, such as Parnassius phoebus, which may fl y with a thorax temperature of 17 to 20°C (Guppy, 1986).
The wing loading, defi ned as p w = m/S, where m is the total individual mass and S the total wing surface is a major variable determining the energy required for insect Upon capture, specimens were put into glassine envelopes and cooled to ca. 11°C in a portable fridge.The same day, in an air conditioned room at 22 to 25°C lit with fl uorescent lamps, each specimen was taken out of the fridge and placed on a piece of cotton under a 150 W neodynium daylight lamp, giving an irradiance of ca.280 W/m² at the butterfl y level.The thorax temperature of each specimen was measured with a testo®895 IR thermometer with the emissivity set at = 0.95, and the recorded temperature was automatically input into a computer database, one datum per second.The behaviour of individuals was reported as they were warming up, especially the basking behaviour, and the position of the wings.The take-off temperature was taken as the last temperature recorded before the butterfl y took off.A 4-th degree curve was fi tted to the raw data, and its slope was used to estimate the for fl ight can have a signifi cant impact on species fi tness.For example, the ability to fl y at lower temperatures may lead to more time available for dispersal and searching for mates and food (Kingsolver, 1983).Conversely, the more thorax temperature increases above ambient air temperature, the greater the risk of overheating, potentially leading to reduced survival and fi tness.For example, in Colias butterfl ies, a thorax temperature of above 40°C causes individuals to cease fl ying to avoid overheating (Kingsolver & Watt, 1983).With the predicted increase in global temperatures and frequency of extreme temperatures, it is increasingly important to understand physical constraints on thermal tolerances which ultimately impact species fi tness and dispersal ability.
Within a locality, the temperatures experienced by individual butterfl ies will largely depend on both their habitat preferences and timing of fl ight activity.Generally, species inhabiting forests will experience more stable conditions than species found in open habitats.A previous study, carried out in Mediterranean France (Nève, 2010), suggested that forest habitat species show lower thorax fl ight temperatures compared to open habitat species because of the lower temperature of the forest fl oor relative to open habitats.A secondary aim of the present study was to test this hypothesis with the available Australian species.

MATERIAL AND METHODS
A total of 152 butterfl ies were captured with hand nets (Upton & Mantle, 2010) in the fi eld, 85 near Adelaide (mainly Bedford Park 35°01´S, 138°34´E and Wotton Scrub, 34°59´S, 138°47´E, South Australia) and 67 at Cedar Creek, 27°49´S, 153°11´E (Queensland) (Table 1); only Pieris rapae was collected at both locations.The tested specimens belonged to twenty species, mostly from open habitat, or linked with the canopy (e.g.Graphium sarpedon).Only four species were specialist understorey species (pers.observ., Braby, 2000): Trapezites symmomus, Melanitis leda, Heteronympha merope and Nacaduba kurava.rate of warming, in °C/s (Fig. 1).The warming up rate at 26°C was used as an index of warming-up speed.This estimate was not possible on 21 specimens for which too few temperature points could be recorded during the warming-up process, hence the lower numbers of available warming-up rate estimates compared with available take-off temperatures.As a control, three Heteronympha merope specimens were tested with a testo® thermocouple sensor set inside their thorax, at the same time as the outside thorax temperature was monitored.The temperature inside the thorax had a mean of 1.3°C warmer than the one measured on the surface, with a low variance of this difference (variance = 0.31).
After the warming-up experiment, each individual was weighed.Twenty-eight of the specimens were released after the experiment, while the rest were retained for references and measurements.These specimens were set with the trailing edge of the hind wing perpendicular to the thorax (Upton & Mantle, 2010).All 125 set specimens were photographed in a standard way with a 60 mm Macro Canon Lens, together with a scale, and the total wing surface was estimated for each specimen using Im-ageJ (Rasband, 2012).The area of overlap between the forewing and the hind wing is taken into account only once, as this area is effective as a single area in the wing loading of fl ying butterfl ies (Dudley, 1990).The wing loading for each specimen was estimated as the ratio of the weight divided by the total wing surface.
All statistical analyses were performed using R (R Development Core Team, 2011).Differences in thorax temperatures at take-off between males and were investigated using t-tests, however only two species (Heteronympha merope and Melantis leda) had adequate numbers of both males and females for testing.Linear regression was used to assess the relationship between warming up rate and take-off temperature both among species and also within four species.Linear regression was also used to assess the relationship between wing loading and take-off temperature among species.Allometry between wing loading p w (in mm²) and body mass w (in mg) was studied directly using the R function nls, a nonlinear regression modelling function, by fi tting the power function p w = b w a directly rather than through its logarithmic transformation (Zar, 1968).
A total of 126 Pieris rapae specimens had been subjected to the same experimental design in France (Nève, 2010), and direct comparisons with that data set were performed where appropriate, using Wilcoxon Signed Rank tests in R.
Voucher specimens have been deposited either in the private collection of Roger Kitching (Griffi th University) for Queensland species, or at the South Australian Museum, Adelaide, for South Australian species and Eurema species (South Australian Museum Accession numbers 31-015517 to 31-015538).Details of capture locations, take-off temperatures, body weights, and wing surfaces are given in the supplementary Table S1.

RESULTS
The observed take-off temperatures varied widely among species, ranging from 13.4 to 46.3°C, with an amongspecies median value of 32.4°C (Fig. 2).The minimum temperature (13.4°C) was recorded for a Heteronympha merope specimen, and the highest (46.3°C) for a Junonia villida specimen.Warming-up rate at 26°C also varied widely among the twenty tested species, with the lowest values found in Melanitis leda (0.02°C/s) and Hypolimnas bolina (0.02°C/s) specimens.The highest value were found for Juniona villida (0.54°C/s) (Fig. 3).
The species with the highest recorded thorax temperature was Juniona villida, with temperatures ranging from 36.8°C to 46.3°C, implying an excess temperature (thorax temperature minus ambient temperature) of up to ca. 20°C.The three species with the highest recorded take-off thorax temperatures were Juniona villida (median = 39.7°C.),Belenois java (median = 37.5°C) and Pieris rapae (median = 34.7°C).The latter species is the only non-native species in our data set, having been introduced from Europe (Lokkers & Jones, 1999), allowing comparison with data from its natal range.Australian specimens of Pieris rapae fl ew with a temperature on average one degree warmer (median = 34.7°C,n = 20) than the ones from France (median = 33.7°C,n = 126; W = 821, P < 0.05, Fig. 5).Australian specimens also warm up more quickly (median = 0.16°C/s, n = 19) than French specimens (median = 0.12°C/s, n = 126, W = 618, P < 0.001).
The among species variation of warming-up rate did correlate signifi cantly with take-off temperature (t 18 = 4.89, P < 0.001, Fig. 6).When the species were looked at individually for the same relationship, among the ones for which at least ten individuals had been tested, only Melanitis leda showed such a positive relationship (t 12 = 2.66, P = 0.02, Fig. 7).
With a value of 0.524 mg/mm 2 , Trapezites symmomus had a much larger wing loading than all other species, which had values ranging from 0.035 mg/mm 2 (for Hypocysta metirius males) to 0.127 mg/mm 2 (for Catopsilia pomona females).Trapezites symmomus is also the only species in the data set to have a linear relationship between thorax temperature and time during the warming-up process, suggesting a different warming-up process than in other Lepidoptera lineages.Trapezites symmomus was then discarded from the data set for the analysis of the effect of wing load on take-off temperature.Among species variation showed that temperature at take-off was positively correlated with wing-loading (t 17 = 2.28, P = 0.036, Fig. 8).
Wing loading p w and body mass w were correlated by the allometric function p w = 0.030 w 0.225 (t = 3.86 and t = 4.31, P = 0.001 and P < 0.001; Fig. 9).

DISCUSSION
Very few temperature data on the studied species has been published before.Notable exceptions are Papilio aegeus, Melanitis leda and Pieris rapae.Heinrich (1986) gave thorax temperatures of 35°C for Papilio aegeus fl ying in the shade and 42.8°C when fl ying in the sunshine.Our data on this species shows lower values (30.4 and 35.4°C).Studies by Kemp (2002) gave inside thoracic temperatures of Melanitis leda males ranging from 28.2 to 38.9°C.Our data on the same species are lower, ranging from 23.7 to 34.8°C.Even allowing for the difference between inside and outside (about 1.3°C), our estimates in both cases remain lower, probably as a result of our laboratory conditions testing the thorax temperature from which fl ight is possible, rather the thorax temperature at which individuals usually fl y or, in the latter case, perform their courtship.
Pieris rapae was the only species tested from European and Australian populations, and as such provides an interesting comparison between its native and introduced range.The 1°C higher thorax take-off temperature, and the quicker warming-up in Australia may be due its adaptation to the higher temperatures experienced in Australia compared with the native range.Such physiological plasticity  in thermal requirements for fl ight may impact not only on the invasive ability of a species but also on its ability to adapt to climate change (Matilla, 2015).
As in European butterfl ies (Nève, 2010), understory specialists require a lower thorax temperature for fl ight activities when compared to open habitat species.Melanitis leda and Heteronympha merope are two butterfl ies active in the forest understorey; as such they experience cooler temperatures than species found in more open habitats (Wickman, 2009), and they have fewer opportunities to bask to gain higher temperatures.In Heteronympha merope, females may fl y at a much lower thorax temperature than males.This sex difference is unusual, as females need a higher thorax temperature for fl ight than males in Colias philodice (Kingsolver, 1983) and in Thymelicus lineola (Pivnick & McNeil, 1986), whereas no difference were found in Heo des virgaureae (Douwes, 1976).Unlike these species, H. merope shows mainly disjunct population peaks for males and females, with males most abundant in spring and females in autumn (pers.observ., Braby, 2000).The difference in fl ight thorax temperatures between males and females of H. merope may refl ect the different climatic conditions experienced by each, and hence an adaptation to minimize excess temperature, thus reducing metabolic costs.Alternatively, differences between sexes may also refl ect difference in wing shape and fl ight behaviour, as in the case of Pararge aegeria (Berwaerts et al., 2006).
Melanitis leda, unlike H. merope, fl ies mainly at dusk and dawn (Braby, 1995;Kemp, 2002Kemp, , 2003)), and may be present throughout the year in many localities (Braby, 2000).Their habit of fl ying during the coolest parts of the day may be linked to their low temperature requirements for fl ight activities.
The case of Juniona villida, with the highest recorded thorax temperatures, half of which close to or above 40°C, raises interesting questions.Previous studies (Kingsolver & Watt, 1983;Matilla, 2015) suggest that J. villida may be Fig. 6.Warming-up rate at 26°C as a function of take-off temperature, for median values observed in twenty Australian species.Fig. 7. Warming-up rate at 26°C as a function of take-off temperature, for four Australian species in which at least ten specimens had been tested.NS -not signifi cant.The relationship is signifi cant only in Melanitis leda.at risk of heat shock at such high temperatures.This species, however, has a fl ight pattern with quick reactions and fast fl ights (Braby, 2000;pers. observ.),suggesting a high amount of muscular activity and thus may have adapted to cope with higher than normal fl ight temperatures.Juniona villida may then be an extreme case of high wingbeat frequency related to a high temperature (Pivnick & McNeil, 1986).This species seems thus to have evolved a fast escape strategy against insectivorous bird predators, which also requires high thorax fl ight temperature, as palatable species often do (Chai & Srygley, 1990).
It should be noted that Trapezites symmomus, which has a wing loading ca.fi ve times higher than all other tested butterfl ies only showed take off temperatures of 31.5°C and 34.5°C, well within the range of the observed butterfl ies (Fig. 3).It has been shown that the genotypes of the individuals was a major determinant of the ambient temperature at which individual fl y (Colias butterfl ies: Watt et al., 1983, Monarch Danaus plexippus: Hughes & Zalucki, 1993).Whether Juniona villida holds special variants of the loci adapted to high temperature remains an open question.
In our data set, wing loading p w showed an allometric relationship with body mass w (p w = 0.030 w 0.225 ).As the exponent is <1, this is a negative allometric relationship (Huxley & Teissier, 1936;Gould, 1966) between wing loading and body mass.Dudley (1990) also obtained a negative allometry (which he erroneously called "positive allometry"), but with a much higher exponent (0.474); only a small part of the difference between our result and Dudley's are due to the use of the direct function for the estimate: an estimate using the classical log-log relationship gives an exponent factor of 0.238 with our data.It should be noted here that for comparative purposes, the exponent does not depend on the units being used, only the multiply- ing factor does; this was the cause of some debate between Huxley and Teissier (Gayon, 2000).Despite the fact that wing loading and body mass are correlated variables, our analysis clearly shows a correlation between thorax temperature at take-off and wing loading, but not with body mass.Heinrich (1986) could not differentiate these two effects, due to insuffi cient data.Our data shows that wing loading is more relevant than body mass in respect to thorax fl ight temperature.
In many insect species, the rate of warming-up is related to muscular activity and body mass, whereas heat loss depends of their external body surface, smaller individuals and species taking longer to warm up than large ones (May, 1976), although in within species comparison, the reverse has occasionally been found (Thymelicus lineola: Pivnick & McNeil, 1986;Hypolimnas bolina: Kemp & Krockenberger, 2004).In our experimental setting, when looking at multiple species comparison, large butterfl ies warm up more quickly than small ones.The lack of signifi cance for tests in individual species may be due to the small sample sizes or lower variance of intra-specifi c variation compared with inter-specifi c variation.
In the present data set, there is also a trend for the butterfl ies to warm up faster if they have to reach a higher temperature for fl ight activity.In this way, the time spent to reach the needed temperature does not increase in proportion with the required fl ight temperature.However, the energy required to reach a high temperature must be accounted for.While solar energy pays a role in increasing body temperature in butterfl ies (Wasserthal, 1975), muscular activity is also needed to reach this temperature, especially, as in the case of Juniona villida, it is well above the ambient air temperature.
Species showing a low wing loading generally fl y with lower wing beat frequencies than the ones with higher wing loadings (Casey & Joos, 1983), and low wing beat frequencies require a lower thorax temperature for fl ight activity.It is thus no surprise that the species with the lowest take-off temperature, Eurema brigitta, Nacaduba kurava and Hypocysta adiante were also among the ones with the lowest wing loading and total weight.On the other hand, wing loading per se is not suffi cient to infer the thorax temperature necessary for fl ight, as microhabitat and behaviour also have an infl uence.
The increase of thorax temperature of Trapezites symmomus in a linear fashion is similar to the sphinx moth Manduca sexta (Heinrich & Bartholomew, 1971), while other butterfl ies generally show a curve with decreasing rates until the take-off temperature is reached in a plateau fashion (Nève, 2010).This may be related to the shape of these two species, which have a lower surface to volume ratio than butterfl ies of the families others than Hesperiidae; they have thus a lower heat loss rate (Heinrich, 2003).The increase in temperature is therefore mainly the result of the muscular activity, heat loss being negligible in comparison.

CONCLUSIONS
Temperature needs of butterfl ies are an important component of their ecological requirements (Wickman, 2009), both at the adult and larval stages (Serratore et al., 2013).The thorax temperature needed for fl ight activity is the result of multiple effects.Individual weight and total wing surface are physical factors infl uencing wing beat frequency, but individual differences in behaviour also plays a role.The optimal fl ight thorax temperature is adaptively determined through selection of enzymes variants active at specifi c temperatures.Species living in cooler environments, such as the forest understorey, may be adapted to fl y at lower temperature as they would mainly rely on their muscular activity for thorax heating.The temperature needed for fl ight activity is thus the result of trade-off between environmental variables and internal physical and biochemical processes.The results highlight the importance of understanding this dynamic in the face of environmental change.While butterfl ies can behaviourally adapt to changes in temperature, factors such as body size and wing loading may impose physiological limits to future climate adaptation.

Fig. 1 .
Fig. 1.Example of individual warming up process, here a male Pieris rapae.The solid line indicates the 4 th degree function fi tted to the raw data.Its derivative at y = 26°C gives the warming up rate at this temperature, here 0.13°C/s, indicated by the slope at this point.

Fig. 2 .
Fig. 2. Thorax take-off temperatures of twenty Australian butterfl y species.The vertical lines indicate the median values, and the boxes the range from fi rst to third quartile; dotted lines indicate extreme values if they fall within 1.5 times the interquartile difference, otherwise open circles indicates outliers.All data points are shown in fi lled circles for species with less than fi ve data points.

Fig. 3 .
Fig. 3. Warming up speeds at 26°C for all tested Australian butterfl y species.Symbols as in Fig. 2.

Fig. 4 .
Fig.4.Thorax temperatures at take-off for males and females of Heteronympha merope and Melanitis leda.The difference in temperatures between the sexes is highly signifi cant in H. merope, but not in M. leda.

Fig. 5 .
Fig. 5. Thorax take-off temperature for Cabbage White butterfl ies Pieris rapae, from Australia and France.Australian specimens have a take-off temperature on average 1°C warmer than the French specimens.

Fig. 8 .
Fig. 8. Temperature at take-off according to wing loading for the 19 species (Trapezites symmomus is omitted; its wing loading is 0.524 mg/mm² and median thorax temperature at take-off was 33°C); dots indicate median values per species.The line indicates the linear correlation for all depicted points.

Fig. 9 .
Fig. 9. Wing loading according to mass in 19 Australian butterfl ies.Trapezites symmomus is omitted; its wing loading is 0.524 mg/mm² and its mass is 340 mg.Both scales are logarithmic.The fi tted allometric function is y = 0.030x 0.225 .