Evidence of recent climate change impacts on butterfly ranges and hybridization is extensive [
1,
45,
46,
47,
48,
49,
50,
51,
52,
53,
54]. Mean annual growing degree-days (above base 50° F, = 10° C) across eastern North America reflect strong elevational as well as latitudinal clines in selection pressures. They are used here to illustrate the thermal landscape that shapes the voltinism patterns of
Papilio species of the tiger swallowtail butterflies across eastern North America (
Figure 1). At our major collecting sites in Florida, Georgia, Ohio, Michigan and Alaska (from 1984 to 2013;
Table 1) female forewing sizes of these butterflies reflect a basically stable multi-year latitudinal size cline, despite significant increases across the landscape in thermal units (as much as 400–700 D-days F; as seen in 2005;
Figure 3; and even larger increases in 2010 and 2012) due to significant regional climate warming during the past 15 years. This clinal size pattern of smaller size at higher latitudes reflects the Converse of Bergmann’s Rule, which mostly occurs in larger insects with one (or few) generations per year compared to smaller insects with many generations per year or many years per generation [
19,
55,
56,
57]. Seasonal time (thermal) constraints result in smaller sizes and shorter developmental times for many ectotherms at higher latitudes/altitudes [
58,
59,
60,
61] and also in local cold pockets [
5,
17].
The major collection sites we have sampled for 25–30 years have not shown much variation in mean adult female sizes, however, some locations at the 1–2 generation “voltinism transition zone” (
Figure 1), with hybrid interaction and genetic introgression [
4,
27] have been shown to harbor late-flight (July) homoploid recombinant hybrids that reflect diagnostic traits of both parental species (the univoltine northern
P. c. and the bivoltine
P. g.). The forewing lengths of both these early and delayed flight hybrid zone populations of
Papilio have remained different but steady from 1999–2012. While at the same latitude (43° N) the Battenkill populations of EF(Pc) and LF hybrids are significantly different in size (48 mm
vs. 51 mm average;
Table 1), which illustrates the potential introgression-influenced variation for the “voltinism/size/D-day” model (see
Figure 1). A similar increase in forewing sizes of late flight (July) females compared to the early flight
P. canadensis across the State of Massachusetts (from 1986–1990) showed mean forwewings to be larger for the July (55.0 ± 3.2 mm, n = 26 females)
vs. May–June (47.3 ± 2.6 mm, n = 15 females). The areas where late flights were reported in Massachusetts are shown (
Figure 13, where dotted lines delineate the thermal landscape areas just north of the historical hybrid zone). Clearly, the 2010 landscape (inset in
Figure 13) recently suggests a capability for the bivoltine potential to move northward extensively across most of Massachusetts, with likely introgression of various traits from
P. glaucus [
1].
However, even without any obvious genetic introgression in northern Michigan and Wisconsin, since 1997, recent and significant increases in forewing sizes of females have occurred in localized “climatic cold pockets” and some adjacent populations of northern Michigan and Wisconsin, presumably due to unique local landscape patterns of warming during the past 15 years. These local “cold pockets” were seen to reflect a significantly greater rate of winter warming that anywhere else east of the Mississippi River (
Figure 5). We also found significant recent increases accumulation of summer Degreeday heat units during the past two decades in and nearby these cold pockets (compared to the major latitudinal sites described in
Table 1 and
Figure 6; e.g., Isabella, Mackinac/Chippewa, Ontonogon and Alaska). We also found rapid wing size increases in local populations of
P. canadensis at the “cold pocket”of Michigan’s lower peninsula (Otsego Co.;
Figure 8) and nearby (Charlevoix Co.,;
Figure 9; and Cheboygan Co.,;
Figure 10). Similarly, in the upper peninsula “cold pocket” of northern Wisconsin (near Iron Mt. at the eastern edge of the “cold pocket”) experienced significant summer D-day warming and the forewing lengths showed a corresponding increase (
Figure 11a,b).
Figure 13.
Female hindwing black band widths for the Otsego “cold pocket” as a function of year, showing a slight (non-significant) increase in the trendline. A decrease in band width would be expected if introgression from P. glaucus had occurred during this period.
The “thermal depth” of the cold pockets in northern Michigan and northern Wisconsin can be seen (
Figure 2 and
Figure 3), and the recent warming temperatures at the edge of these cold pockets can also be seen to reflect rapid mean forewing size increases in local females of
P. canadensis (
Figure 8,
Figure 9,
Figure 10 and
Figure 11). The lack of increase in summer Degree-day accumulations at the major sites in the Upper Peninsula (Ontonagon Co.; Stambaugh in Iron Co.; Mackinac/Chippewa Cos.;
Figure 4) stand in contrast to the warming near the edge of this cold pocket (
Figure 12;
Table 1).
4.1. All Life Stages are Important When Considering Adaptations to Thermal Constraints
All developmental stages of insects will likely experience thermal constraints as selection pressures determine their success [
61,
62,
63,
64]. We have shown that pupae of
P. canadensis emerge earlier in the spring than
P. glaucus [
29]. Such early starts would be an ecological advantage where thermal constraints exist, such as throughout the entire hybrid zone from Minnesota to New England (
Figure 1). Although post-diapause developmental constraints of pupae during winter and in the Spring are not growth related, they are extremely important for seasonal phenology and reproductive isolation of populations and hybrids [
27].
However, such pupal development at lower temperatures for early emergers is not without physiological and ecological costs. Voltinism and diapause “strategies” are variable in insects, but nutrient reserves must be carried forward through winter for post-diapause processes of metamorphosis such as adult formation in pupae, flight, and reproduction [
64,
65,
66,
67,
68]. This means that winter metabolic stresses (warm and cold) can impact survival and adult size [
10,
69,
70].
Another cost incurred by diapausing
P. canadensis pupae is lowered survival resulting from short term temperature stress extremes (both warm and cold) during mid-winter or Fall [
10,
68]. Such short-term stress-induced metabolic expenditures and weight loss in diapausing
P. canadensis pupae (and also recombinant hybrids of the late flight) compared to
P. glaucus [
10,
70] are not simply associated with desiccation as may occur in some Lepidoptera [
69], since the percent body water in
P. canadensis pupae was the same for fall, mid-winter, and spring, at 73%–75% [
70]. While the Late-Flight hybrids experience large early metabolic expenditures at cooler temperatures, as did
P. canadensis [
10], they, in contrast, do not emerge early. The total degree-days required to complete post-diapause pupal development and adult emergences for LF hybrids were almost twice those needed for EF (Pc) at 14, 18, and 22° C, and more than twice at 26° C (619 D-days for LF males and 725 D-days for females; 289 D-days male EF, 319 for female EF [
27]).
With increasing climatic variation globally [
71,
72,
73,
74], fluctuating seasonal temperatures (means and extremes) are becoming more common [
75,
76,
77,
78], especially in the winter and spring seasons [
79,
80,
81,
82]. Snow cover for overwintering pupae will become less predictable and such variation in air temperature will likely have more severe impacts [
61,
65,
66,
71]. Increased daily variance in temperatures (high and low) may also have significant or subtle impacts on insect populations even though the mean daily temperature and the degree-day accumulations may otherwise be identical [
79]. Variable winter temperatures may suppress metabolic rates in Lepidoptera [
21], and this may generate selection for “deeper diapause” intensity seen in lower latitude
Papilio glaucus populations from Georgia compared to Pennsylvania and Michigan [
10]. Similar increased diapause intensity has been seen in Diptera and Orthoptera [
11,
12].
4.2. Growth Rates and Voltinism
The economic damages associated with increased generations of insect pests [
79] in agricultural, silvicultural, and human living environments are likely to be very serious [
80,
81]. Ambient temperatures affect biochemical reaction rates [
82] and thus in combination with the nutritional quality of their hosts, largely govern the growth rate potentials of ectotherms, including immature arthropods [
22,
23]. Recent reviews [
83,
84] address multivoltine, bivoltine, univoltine, semi-voltine (2 years per generation; as in most arctic Lepidoptera; [
85]) and parti-voltine populations (with 3 or more years per generation, e.g., [
63]. An increased number of generations with climate warming has been noted in more than 28 species of Heteropterans and other insects [
86,
87]. Such increases have also been seen in 275 species of aquatic Odonata [
88], geometrid moths [
89], dragonflies [
90], grape berry moths, [
91], spruce bark beetles [
92,
93], and general Lepidoptera [
94,
95]. In Coleoptera, heat accumulation associated with unusually warm summers caused a shift from the predominant two-year cycle (semi-voltine) to a one-year cycle (univoltine), thereby doubling the rate of increase and spread of populations of beetles [
96,
97,
98,
99].
It has frequently been assumed that insect growth rates are at the maximum that is physiologically possible given a specific ambient temperature and specific amount and quality of resource. However, many animals do not grow at their physiological maximum, even with unlimited food of good quality [
100,
101,
102,
103]. Thus, slower (“optimal”?) growth may be favored at certain times under certain conditions, depending on various trade-offs between the costs and advantages of growing rapidly [
18,
103]. For example, reduced resistance to cold stress [
104] or disease [
105] can occur at fast growth rates. Locally adapted insect populations may also become differentially adapted for fast growth rates with thermal specialization [
106,
107,
108] or host plant specialization [
109].
Parmesan
et al. [
48,
110] have pointed out that very few studies have analyzed relationships between climate change for long periods of time, or across the entire geographic range of any species. However, some extensive long term insect studies such as ours here, not only include climate, but also include host use patterns, voltinism, morphology, hybrid zones, and local cold pockets [
1,
50,
111,
112]. While
P. canadensis larvae from thermally-stressed Alaskan populations do grow faster compared to populations with relaxed thermal stress in Michigan [
3,
8], the adult
P. canadensis females in Alaska also select the plant species for oviposition which have the highest nutritional quality for the fastest larval growth. In contrast, Michigan females, with relaxed thermal constraints, distribute their eggs more widely across different species of host plants of varying degrees of quality for larval growth with lower quality plants possibly providing “enemyfree-space” for avoiding host specific enemies [
4,
113,
114]. The general result of this behavioral rankorder of
Papilio oviposition preferences are latitudinal gradients in breadth of host use with alternating bands of specialization and generalization that fit with the alternating “constrained” and relaxed patterns of voltinism potential of the area [
4,
6]. The complex trade-offs affecting adult size may include different host plant preferences by the different broods of the multivoltine
P. glaucus, and it is feasible that poorer quality hosts (resulting in smaller pupae/adults) are chosen in one of the summer generations “in order to” escape heavy natural enemy pressure that might focus in on larvae using the usual most nutritious host plants as suggested in the “Voltinism-suitability” model [
4,
6].
4.3. Locally Rapid Responses to Climate Change in “Cold Pockets”
In contrast to the generally relaxed thermal constraints of most Michigan
P. canadensis populations compared to Alaska (both are obligate diapausers and univoltine), Michigan and Wisconsin “cold pockets” represent very localized thermal constraints (similar to Alaska conditions [
5,
17]). The cold pocket females have traditionally emerged later than those in surrounding areas and, as in Alaska, they pupate early at smaller sizes than general Michigan
P. canadensis [
5,
17]. These cold areas have exerted strong natural selection pressures relative to surrounding populations outside cold pockets. The rapid size (physiological/morphological) responses of females in these warming cold pockets may simply represent adaptive plasticity, rather than a genetically-based size increase (
Figure 8,
Figure 9,
Figure 10 and
Figure 11). For example, 1992 was an exceptionally cold year (
Figure 5; see also [
115]), and it is clear that female size in the following 1993 summer was significantly lower at several locations, possibly as a direct result of “thermal time” constraints during the summer growing season (
Figure 8b,
Figure 9b and
Figure 10b).
However, at least three other potential explanations for increased females sizes in cold pockets exist, including, (1) warmer winters (
Figure 5) which may require lower metabolic expenditures for diapausing pupae [
10] resulting in more biomass to convert into adult tissue, or (2) perhaps warmer springtimes affected post-diapause development and allowing larger adults to result, as shown in both
P. canadensis and
P. glaucus [
2]. Also, in local climatic cold pockets of Michigan, the females have historically selected ash (
Fraxinus spp.) leaves as a favorite, because these tender leaves have the highest nutritional quality compared to fully-expanded leaves of other hosts such as cherry, aspen, poplar, birch, and others (due to delayed bud-break and delayed leafing of ash leaves by 3–6 weeks in the coldest parts; [
6]). (3) Recent warming may have allowed these cold pocket females to select host plants such as cherry which allow faster growth and larger pupae (however, oviposition preference behavior of adult females here has not been assayed recently; since 1996 [
5]). While LF hybrid oviposition preferences appear be controlled by Z-1inked factors [
1,
116,
117,
118], local ash host races in thermally-constrained mountains around the Battenkill River basin in Vermont appear to be a result of, rather than a cause of, evolutionary divergence [
119].
Overall, the “voltinism-suitability” concept integrates abiotic (especially thermal) as well as biotic factors such as host plant nutritional quality and natural enemies. As Stamp [
120] said: “In temperate regions temperature determines the activity of both herbivores and their enemies, with each having different thermal ranges and optima... The limitations imposed by thermal conditions and host plants influence developmental rate and consequently the number of generations per year. How the developmental periods of generations fit into the growing season may affect foraging patterns and adult size... Enemies can force herbivores into microclimates that are sub-optimal in terms of food quality and temperature, which may contribute to reduction in survivorship of the herbivores... We need models that incorporate the effects of temperature, food quality, and predators...”.
While community interactions between herbivorous insects, host plants, natural enemies, and pollinators may be closely integrated in time and space, it is not likely that climate changes will result in concordant, congruent, or concurrent shifts in the composition of such species [
121,
122,
123,
124]. Modelling climate change impacts must include both autecological and synecological aspects and we should strive to better understand the relative importance of biotic and abiotic factors [
80,
113,
125,
126,
127,
128,
129] across the entire range of the species of interest. Local adaptations, including genetic evolution and phenotypic flexibility [
130,
131] also need to be encorporated into climate change models and geographic range predictions for insects [
1,
29,
110,
127,
132,
133,
134]. Interactions of changes in plant phenology, nutritional quality, and herbivores will also involve precipitation and carbon dioxide increases as well as temperatures [
135].
Increased thermal unit accumulations leading to a potential extra generation at the warmer side of the hybrid zone may enhance northward genetic introgression from the larger, and facultatively diapausing
P. glaucus [
1,
136,
137,
138]. However, multiple matings of these
Papilio females and males (potentially inter-specifically; [
139,
140,
141]) can make gene flow complicated to assess. Nonetheless, such changes in voltinism and genetic introgression have been shown to impact evolutionary and speciation processes [
142,
143,
144,
145,
146,
147,
148,
149,
150,
151], and divergent selection on recombinant hybrids has facilitated such processes in other Lepidoptera [
152,
153,
154]. However, we were nevertheless able to determine that introgression from the southern (larger)
P. glaucus is unlikely to explain these rapid forewing size increases of female
P. canadensis locally in cold pockets, since the anal cell hindwing band widths do not get narrower (as in
P. glaucus and hybrids [
29,
52,
53]) during the same period for Otsego County (
Figure 14), and similarly for adjacent Charlevoix and Cheboygan Counties (data not shown). This suggests that phenotypic plasticity in developmental responses (increased size in warming cold pockets) rather than genetic introgression from
P. glaucus is primarily responsible for these rapid size responses locally in cold pockets.
Figure 14.
The mean thermal landscape in D-days for NY, PA, MA, and New England during a warm decade (1998–2008) compared to a cold year (1992) and a warm year (2010). The historical hybrid zone (with
P. canadensis on the northern side) is indicated by the heavy line and D-days indicated by color bands. The dark brown is 2,900–3,400 F and indicates the northern-most limits for bivoltine potential in
P. glaucus on most plant species. The dotted lines indicate where two generations are not possible and where July (LF
Papilio hybrids) were seen in the Massachusetts State survey (1986–1990). These “LF” hybrids were larger than the sympatric
P. canadensis (see
Section 4.1) due to recent genetic introgression with correspondingly narrowed hindwing black bands [
27,
29]. Historical northern limits of dark (mimetic) morph females are shown as dots, and this W(=Y)-linked trait is slow to move even with climate warming [
1]. Note the extensive recent warming 2010 (and 2012 was similar) permitting northward movement of the bivoltine potential in eastern Massachusetts, where two generations were basically impossible before 1998.
Figure 14.
The mean thermal landscape in D-days for NY, PA, MA, and New England during a warm decade (1998–2008) compared to a cold year (1992) and a warm year (2010). The historical hybrid zone (with
P. canadensis on the northern side) is indicated by the heavy line and D-days indicated by color bands. The dark brown is 2,900–3,400 F and indicates the northern-most limits for bivoltine potential in
P. glaucus on most plant species. The dotted lines indicate where two generations are not possible and where July (LF
Papilio hybrids) were seen in the Massachusetts State survey (1986–1990). These “LF” hybrids were larger than the sympatric
P. canadensis (see
Section 4.1) due to recent genetic introgression with correspondingly narrowed hindwing black bands [
27,
29]. Historical northern limits of dark (mimetic) morph females are shown as dots, and this W(=Y)-linked trait is slow to move even with climate warming [
1]. Note the extensive recent warming 2010 (and 2012 was similar) permitting northward movement of the bivoltine potential in eastern Massachusetts, where two generations were basically impossible before 1998.
