Linking thermal adaptation and life-history theory explains latitudinal patterns of voltinism

Abstract

Insect life cycles are adapted to a seasonal climate by expressing alternative voltinism phenotypes—the number of generations in a year. Variation in voltinism phenotypes along latitudinal gradients may be generated by developmental traits at critical life stages, such as eggs. Both voltinism and egg development are thermally determined traits, yet independently derived models of voltinism and thermal adaptation refer to the evolution of dormancy and thermal sensitivity of development rate, respectively, as independent influences on life history. To reconcile these models and test their respective predictions, we characterized patterns of voltinism and thermal response of egg development rate along a latitudinal temperature gradient using the matchstick grasshopper genus Warramaba. We found remarkably strong variation in voltinism patterns, as well as corresponding egg dormancy patterns and thermal responses of egg development. Our results show that the switch in voltinism along the latitudinal gradient was explained by the combined predictions of the evolution of voltinism and of thermal adaptation. We suggest that latitudinal patterns in thermal responses and corresponding life histories need to consider the evolution of thermal response curves within the context of seasonal temperature cycles rather than based solely on optimality and trade-offs in performance.

This article is part of the theme issue ‘Physiological diversity, biodiversity patterns and global climate change: testing key hypotheses involving temperature and oxygen’.

1. Introduction

Understanding how organisms are adapted to a variable climate is a fundamental problem in ecology and evolutionary biology. Temperature is a major component of climate that has a pervasive effect on biological rates [1]. Species have adapted their phenology and life cycle to seasonal and latitudinal gradients in temperature [2–4]. The seasonal timing of phenological events in turn influences subsequent biological interactions [5,6] and these responses are predicted to be modified by climate change [7–9].

To predict how species will respond to future changes in temperature, we need to understand the genetic, phenotypic and environmental sources of variation that generate adaptive responses [10–13]. These sources of variation can be inferred from latitudinal patterns in thermally determined traits and thermal performance curves (TPCs), particularly when accompanied by breeding studies in controlled environments [14–16]. One key thermally determined trait that varies along a latitudinal gradient is voltinism—the number of generations each year [17]. Variation in voltinism is generated by variation in climatic factors or by variation in the life-history traits that underpin it, such as egg hatching phenology, or their interaction [18,19]. Egg hatching phenology is also a thermally determined trait that encompasses the thermal sensitivity of development rate and dormancy responses [20,21]. Overall, hatching phenology can represent a developmental bottleneck in the life cycle and can strongly influence how life cycles are synchronized with current and future changes in climate [22].

Theoretical models of voltinism and thermal adaptation provide conceptual frameworks to understand patterns of variation along environmental gradients. However, models of thermal adaptation and the evolution of voltinism have developed independently and approached the problem of latitudinal gradients differently [7,23]. Higher latitudes are typically associated with univoltine life cycles whereas lower latitudes are associated with multivoltine life cycles, although exceptions to the general pattern exist [24–26]. Lower latitudes are also associated with longer growing seasons and warmer climates, and vice versa at higher latitudes [27,28]. Season length, as determined by photoperiod, is the main evolutionary driver of voltinism in many phenological models of voltinism because photoperiod is a reliable cue of changes in season [29,30]. These models overlook the influence of temperature on development although both photoperiod and temperature covary along latitudinal gradients [31]. Models of voltinism patterns do not make explicit predictions about the evolution of TPCs. This is in contrast to models of thermal adaptation, which make several predictions for latitudinal gradients in TPCs of development rate depending on the selection pressures involved and whether traits have evolved to compensate for latitudinal gradients in temperature [16,32].

There are two general models of latitudinal compensation in TPCs [33]. The first is the optimality model of a TPC, where performance is maximized at the expense of thermal breadth of performance. In this model the temperature where performance is optimal is predicted to shift according to local temperatures in cases when temperature is the main driver of selection. We would then predict that higher-latitude populations will have a lower thermal optimum of performance and a shift in temperature limits of performance towards cooler temperatures (figure 1a) [23]. Few empirical studies support this model of thermal adaptation along a latitudinal gradient [34–36]. The mixed support is likely because optimality models assume that thermal limits are free to vary and that there is a trade-off between performance and thermal breadth [37,38]. Instead, thermal limits are conserved or have low plasticity along latitudinal gradients, and the thermal sensitivity of rates varies [32,39].

Figure 1.

Theoretical models of latitudinal variation in thermal reaction norms between high-latitude populations (blue, dashed) and low-latitude populations (orange, solid). Arrows indicate the direction of selection. (a) Optimality model: reaction norm is horizontally shifted to adapt optimal performance to environmental temperature. (b) Counter-gradient variation: high-latitude populations compensate for a shorter season length by increasing thermal sensitivity without increasing thermal breadth. (c) Co-gradient variation: thermal sensitivity of high-latitude populations decreases without increasing thermal breadth with or without compensation in other traits. (Online version in colour.)

TPCs that vary along the vertical axis, i.e. in thermal sensitivity, are predicted to arise when the length of the growing season is the main driver of selection. In this case, the second model of latitudinal compensation predicts that higher-latitude populations will develop faster than lower-latitude populations to compensate for a shorter growing season (counter-gradient variation, figure 1b) [40,41]. In other words, genotype by environment interactions are negative [33]. Counter-gradient variation along temperature gradients is the most commonly reported latitudinal pattern of TPCs [32] and has been described in fish [27], frogs [42], butterflies [43] and damselflies [40].

A second case of seasonal constraints on TPCs is co-gradient variation, which predicts that higher-latitude populations will develop slower than lower-latitude populations (figure 1c). Co-gradient variation is predicted to occur when genotype by environmental interactions are positive, and species may or may not compensate for latitudinal changes in temperature in other covarying life-history traits [33]. In these cases, other genetic or environmental selection pressures constrain the evolution of the TPC to compensate for temperature gradients [44,45]. Co-gradient variation is not as widely reported as counter-gradient variation [41] but has been observed in damselflies [46], grasshoppers [18], butterflies [45,47], beetles [48], snails [36] and crickets [10].

Intermediate responses between the two general thermal adaptation models of latitudinal compensation may also be generated because selection pressures acting on the thermal breadth and thermal sensitivity are non-mutually exclusive [33]. We predict shifts in both thermal breadth and thermal sensitivity in these cases [49–51]. Co-gradient and intermediate responses of TPCs imply that other life-history traits may also be influenced by the same selection pressures to compensate for these TPCs. The association between TPCs and life-history traits is particularly relevant for traits related to development, such as development rate, as these are important traits that determine voltinism and are thermally determined in insects. Covarying life-history traits, such as size and growth rate, have been proposed to explain differences in TPCs of closely related species [45,52]. The evolution of both voltinism and latitudinal compensation can thus shape TPCs of development but the role of voltinism patterns, or life history more generally, is absent in theoretical models of thermal adaptation [46,53,54].

Models of insect voltinism also infer the evolution of diapause and the expression of alternative developmental pathways. These models predict that a temperate insect with a switch in voltinism controlled by diapause should have a shorter development time in multivoltine populations than in univoltine ones [55,56], as observed in many insects [10,18,57–59]. Levins [60] described a model that predicted that switches in developmental pathways, such as by the induction of diapause, should evolve under greater environmental heterogeneity in space and time. Models of insect phenology based on temperature show that hatching phenology can be generated by seasonal changes in temperature alone [19,61,62] and that voltinism patterns can be generated using dormancy responses other than diapause [63].

For a widely distributed insect where temperature is a major determinant of development or voltinism, the latitudinal pattern of TPCs will depend on the mechanism that generates voltinism and whether species have compensated for changes in temperature along a latitudinal gradient [46,64]. If species express latitudinal compensation, we may predict a counter-gradient of TPCs. If voltinism imposes a genetic and environmental constraint on the evolution of the whole life cycle and species have not compensated for latitude in other life-history traits, such as diapause, we would predict a co-gradient pattern of TPCs of development rate whereby multivoltine populations at low latitudes develop faster than univoltine populations at high latitudes. At the extreme end, a diapause response can be considered an absence of temperature-dependent development.

Here we test these patterns in four species of matchstick grasshoppers (Orthoptera: Morabidae, sub-family Morabinae): Warramaba picta, W. whitei, W. flavolineata and W. grandis [65]. The genus Warramaba is an excellent system with which to examine the evolution of latitudinal adaptation of insect life cycles. Sexually reproducing Warramaba are widely distributed across more than 1000 km along a latitudinal gradient in both aridity and temperature in arid and semi-arid Western Australia and distributions overlap in some localities (figure 2) [66–69]. We hypothesize that the egg is the primary stage for determining voltinism, as represented by hatching phenology. Because the Morabinae are flightless, have limited dispersal and have strong genetic structuring within taxa, we predict there is a high potential for life cycle adaptation to local climates among lineages and species [65,66,70]. In this study, we provide the first comparative description of the life history, phenology and biology of Warramaba species from unpublished field diaries, as well as laboratory measurements on the thermal dependences of egg development and dormancy.

Figure 2.

Map of sampling localities for each species of Warramaba used in this study and their relative distributions. Distribution limits (dashed lines) were estimated using minimum convex polygons. Grasshoppers were sampled between 1967 and 2018. Northern populations (open circles) of W. whitei and W. flavolineata were located at latitudes lower than −29.18° based on a major divergence in mitochondrial DNA. Region was not assigned to W. grandis. (Online version in colour.)

2. Material and methods

(a). Inferring phenological timing of hatching

We define voltinism based on the phenological timing of hatching, as indicated by the presence of nymphs. Univoltine populations will have nymphs only during summer (December–February) and multivoltine populations will have nymphs present in summer and in winter (June–August). To infer the voltinism of Warramaba, we collated field observations made by M. J. D. White between 1967 and 1981, and by M. R. Kearney and colleagues between 2003 and 2018. These observations varied in their nature, ranging from presence/absence to breakdowns of instars. Most observations were made in spring (September–November) and summer and a targeted survey was undertaken in the winter of June 2017. Observations with no species or indication of stages recorded were excluded from the analysis for a total of 444 observations (electronic supplementary material, tables S1 and S2). To calculate sampling bias across different months of the year, we first determined whether a sampling locality had been surveyed for each month of the year. To compare geographical variation in phenology we then assigned collection localities of W. picta, W. whitei and W. flavolineata to either a northern or southern region. Northern sites of W. picta were located at latitudes lower than −25°, which was arbitrarily chosen because there was a large gap between sampling localities (figure 2). Because sampling sites of W. whitei and W. flavolineata were continuously distributed from north to south, a boundary at −29.18° latitude was chosen based on strong separation in mitochondrial DNA [66] (figure 2). Geographical region was not assigned to W. grandis collection localities because we sampled along a latitudinal transect and there is no genetic evidence for lineage separation. To account for different sampling intensities across years at different sites, we calculated the proportion of years where grasshoppers were present, with 0 indicating no grasshoppers were observed in all visits and 1 indicating that grasshoppers were observed at each visit. The distribution of grasshoppers at the nymphal and adult stages was calculated separately for each geographical region.

(b). Quantifying variation in developmental rates

To characterize the thermal dependence of egg development in Warramaba, we collected individuals of W. picta, W. whitei, W. flavolineata and W. grandis from Western Australia in the summer (December–February) of 2016 (electronic supplementary material, table S3). To test for intraspecific variation in the thermal dependence of egg development and dormancy in W. picta, W. whitei and W. flavolineata, we sampled replicate populations in the northern and southern regions of their respective distribution. These geographical regions followed known mitochondrial lineages in W. whitei and W. flavolineata [66]. To test for latitudinal variation in developmental traits in W. grandis, we sampled 14 populations along a latitudinal northern to southern transect spanning 551 km.

Initial egg mass was measured at the start of incubation to account for the effects of egg size on development rate (XS205DU, Mettler Toledo, Port Melbourne, Australia). In 2016, we incubated eggs from all populations, except southern W. flavolineata populations, at 16, 20–40°C in increments of 2°C, and 45°C (electronic supplementary material, table S4). Eggs incubated at 30°C were held in cabinet incubators (Scientific Equipment Manufacturers Pty Ltd, Magill, South Australia, Australia) and eggs at all other temperatures in thermocyclers (Biometra GmBH, Göttingen, Germany). The range of incubation temperatures chosen in 2016 captured the range that they are likely to experience in the field [71]. Eggs from W. flavolineata were excluded from these experiments because unpublished observations of W. flavolineata egg development in 2005 indicated southern mitochondrial lineages of this species have an obligate diapause that requires chilling for its termination (electronic supplementary material). In 2016, we investigated the conditions that would break the diapause in W. flavolineata eggs from southern populations (electronic supplementary material).

Viable eggs, as determined by the opacity and colour of eggs after dry storage, were randomly allocated to experimental temperatures in the thermocycler. Clutches from individuals were randomized and eggs from individuals were split among incubation temperatures. All remaining viable eggs from the cohort were incubated in the cabinet incubator at 30°C. Eggs incubated within the cabinet incubator were placed on filter paper on top of moist sand within Petri dishes (see also electronic supplementary material) [70].

The rate of egg development (% day⁻¹) was calculated from the time between the start of incubation and the time the first instar nymph emerged in the glass vial. We identified eggs that did not hatch after at least 54 days as putatively dormant. Any such eggs were externally inspected for signs of embryonic development. To test whether these eggs had died or were dormant, we incubated the putatively dormant eggs at 34°C in thermocyclers until they hatched. The lowest and highest temperatures at which continuous development and hatching were observed were taken as the lower and upper thresholds for development, respectively.

Data from the 30°C incubator was pooled with the data from the thermocyclers for analysis. We are confident that there is no significant variation in the thermal response of egg development between eggs incubated in thermocyclers and in cabinet incubators [72]. Dormant and non-dormant eggs were analysed separately. We assessed variation in development rate across the range of experimental temperatures using analysis of covariance (ANCOVA), with incubation temperature, a quadratic term for temperature and log₁₀ initial egg mass as continuous predictor variables, species and geographical region as categorical predictor variables, as well as an interaction term between species and region to test for regional variation within species. Statistical analysis was conducted in R v. 3.4.4 [73].

3. Results

(a). Field phenology of Warramaba in Western Australia

The hatching phenology of Warramaba observed in the field varied among species and populations. Nymphs and adults of southern populations of W. picta were found in all sampling months across their range, indicative of a multivoltine phenology (figure 3). We cannot conclude that northern W. picta were multivoltine based on our definition because we did not observe nymphs in a northern population of W. picta during winter, and only a single locality was surveyed where adults were observed. Late instar nymphs and adults were found in northern and southern populations of W. whitei during summer (figure 3). In winter, nymphs of W. whitei were found in northern populations and no nymphs were observed in southern populations (figure 3). Only one adult male grasshopper was found in one southern population (Arrow Lake, WA; electronic supplementary material, table S2). Adults and nymphs of W. flavolineata were observed in northern and southern populations during summer but no individuals of any stage were found in the winter, implying a univoltine phenology (figure 3). Populations of W. grandis reliably yielded individuals in summer and individuals were not found in winter in a southern population, suggesting W. grandis has a univoltine phenology (figure 3). A latitudinal gradient of W. grandis instars was observed during summer 2016 with earlier stage nymphs found in northern populations, and more mature, or adult, individuals found in southern populations (electronic supplementary material, figure S3).

Figure 3.

Proportion of years with observed presence of adult (inner ring) and nymphs (outer ring) for Warramaba picta, W. whitei and W. flavolineata in northern and southern regions of their respective distributions and for W. grandis across their distribution. Region was not assigned to W. grandis. Grasshoppers were surveyed at intermittent periods in the field between 1967 and 2018 (444 observations). 0 (yellow) indicates that localities were sampled, and species were absent. Blank segments were not surveyed that month for any year. Most surveys were conducted during spring (Sep–Nov), summer (Dec–Feb) and a targeted survey in winter (June–Aug).

(b). Egg development and dormancy responses to temperature

Northern populations of W. picta did not develop below 24°C, whereas southern populations did not hatch below 20°C (figure 4; electronic supplementary material, table S5). No eggs of W. grandis and northern W. flavolineata hatched below 28°C. Northern populations of W. whitei did not hatch below 22°C, whereas southern populations did not hatch below 26°C. The lower thresholds for development (i.e. no development observed after 54–55 days) varied with species and region, ranging in increasing order from 20°C in southern W. picta, to 22°C in northern W. whitei, 24°C in northern W. picta, 26°C in southern W. whitei and 28°C for all populations of W. grandis and W. flavolineata. The upper thermal limit of egg development for all species and populations was 36°C. Above 36°C, eggs appeared to be dormant, but we were unable to hatch any eggs that had not already started visible embryonic development. Most eggs hatched between these thermal limits and were dormant below the lower thermal limit (electronic supplementary material, figure S5). Non-dormant eggs (n = 496) hatched between 12 and 76 days (median 21 days) across all temperatures. Development rate of non-dormant eggs across all species and regions was best described with a quadratic term for temperature (figure 4, ANCOVA F_1,487 = 106.400, p < 0.001). The interaction between species and region was a significant source of variation in development rate (figure 4, F_1,487 = 17.950, p < 0.001). A winter diapause was observed in eggs from southern populations of W. flavolineata which required chilling to be broken (electronic supplementary material, figure S8).

Figure 4.

Mean development rate (% day⁻¹ ± s.e.) of non-dormant eggs for each species of Warramaba at constant temperatures between 16 and 45°C (n = 496). Thermal response curves were characterized for northern (solid lines) and southern (dashed lines) populations of Warramaba in 2016 experiments. Eggs were kept in dry storage (22–24°C, 30% RH) until used in experiments. (Online version in colour.)

4. Discussion

Insect life cycles have evolved under both spatial and temporal changes in temperature along latitudinal gradients. In studies of Northern Hemisphere insects along latitudinal gradients, switches from multivoltine life cycles to univoltine life cycles are typically generated by varying the expression of diapause between lower latitudinal populations and higher latitudinal populations [28,57,74]. Voltinism patterns in the Southern Hemisphere are similar but fewer studies have explored the mechanisms that generate these patterns [18,25,75]. There have also been large scale studies of thermal response and adaptation along latitudinal gradients in the Southern Hemisphere [76–79]; however, these studies have focused on mesic habitats.

Along a latitudinal gradient in arid and semi-arid Western Australia, the widely distributed matchstick grasshopper genus Warramaba showed an interspecific switch in voltinism (figure 3). Southern populations of W. picta and northern populations of W. whitei at lower latitudes were multivoltine and W. flavolineata and southern populations of W. whitei at higher latitudes were univoltine. The exception to this pattern was W. grandis, which was univoltine despite having the second widest latitudinal distribution of the Warramaba and a parapatric distribution with the multivoltine W. picta (figure 3). Conserved univoltine responses have also been observed in a high-latitude damselfly [58]. The most striking latitudinal pattern was the difference in the TPC of egg development rate between voltinism phenotypes (figure 5). Divergences in temperature-dependent developmental traits have been reported for different voltinism phenotypes in other insects [45,46,80,81]. There were three characteristics that generated the divergent pattern of Warramaba TPCs through interactions between voltinism, temperature and species identity. None of these characteristics was predicted by models of voltinism or thermal adaptation alone [33,55,82].

Figure 5.

(a) Voltinism-shaped variation: multivoltine populations (solid lines) at low latitudes develop faster and over a wider temperature range than univoltine populations (dashed lines) at low temperatures. Dormancy is represented by the absence of a thermal response curve. Arrows indicate the direction of selection by seasonal and latitudinal gradients in temperature. (b) Mean development rate (% day⁻¹ ± s.e.) of non-dormant eggs from multivoltine populations (solid line) and univoltine populations (dashed line) of Warramaba spp. at constant temperatures between 16 and 45°C. Data collected in 2016 from species and regions were pooled into their respective voltinism pattern inferred from hatching phenology in the field between 1967 and 2018. Eggs were kept in dry storage (22–24°C, 30% RH) until used in experiments.

First, there was a co-gradient pattern in development rate where multivoltine W. picta and W. whitei developed faster than univoltine W. flavolineata and W. grandis (figure 5). The same pattern was observed between geographical regions of W. whitei, with populations in the north developing faster than univoltine populations in the south (figure 4). Co-gradient patterns of variation indicate that there are positive genetic by environmental interactions shaping the evolution of TPCs [41]. A potential environmental constraint on the evolution of egg TPCs is the length of the growing season along the latitudinal gradient [57]. If spatial changes in season length impose a greater selection pressure than spatial changes in temperature per se, then the TPC for development would be expected to change more along the y-axis (development rate) than on the x-axis (position of thermal limits) [83].

The divergence of voltinism in Warramaba also follows patterns of genetic divergence in Warramaba based on the mtCO1 gene [66]. Warramaba are inferred to have rapidly expanded their distribution from climate refugia at low latitudes to higher latitudes during the glacial cycles of the Pleistocene [66]. This expansion is associated with patterns of chromosomal rearrangements and a latitudinal gradient in fecundity in W. whitei [67,84]. The most ancestral karyotype in the genus is exhibited by the multivoltine W. picta [66], suggesting that this is the ancestral state and that univoltinism in southern W. whitei, W. grandis and W. flavolineata is a derived condition.

Second, there was a narrowing of the thermal breadth of egg development in univoltine populations that biased development towards higher temperatures (figure 5). The narrow thermal breadths for development are unusual compared with other insects [21]. The variation of thermal breadth in Warramaba was primarily generated by the lower thermal limits of development, which are typically more plastic than upper thermal limits [28,85]. Below the lower thermal limit, Warramaba eggs were dormant. We characterized two types of dormancy in Warramaba below the lower thermal limit: a non-diapause dormancy in all species and a ‘winter’ diapause in W. flavolineata typical of Northern Hemisphere insects. One W. whitei egg hatched at 34°C after remaining dormant at 38°C (electronic supplementary material, table S5) but we could not confirm whether sexually reproducing Warramaba had a ‘summer’ diapause like W. virgo [71]. This is likely because egg development in Warramaba is also dependent on soil moisture, and the incubation conditions within the thermocycler at these temperatures were confounded by a desiccating environment when water in the PCR tube evaporated between daily checks [71,72].

Although most eggs from southern W. flavolineata did not hatch when we incubated them immediately after oviposition in 2005 and 2006, eggs incubated in 2016 developed continuously at 30°C after they had been kept in storage at 22–24°C for a considerable amount of time. This suggested that the threshold for diapause development is high enough to be broken at the storage temperature. Diapause in W. flavolineata is hypothesized to represent the opposite end of a spectrum of potential TPCs in response to a cooler and more seasonal climate [86]. In this case, the TPC can be viewed as a flat reaction norm of zero rate, in contrast to a high development rate in a multivoltine species like W. picta. Among the sexually reproducing species of Warramaba, diapause was only expressed in W. flavolineata, and most strongly in the highest-latitude populations where seasonal variation in temperature is greatest. This pattern of greater diapause expression at higher latitudes is consistent with the predictions of Levins [60].

The non-diapause dormancy below the lower thermal limit was characterized by an increase in egg mass but an absence of visible embryonic development when examined after 54–55 days of incubation. These eggs developed immediately after being transferred to 34°C [87]. We also found an intermediate proportion of non-dormant and dormant eggs at the threshold temperature, which suggested a degree of developmental plasticity near this threshold; i.e. some eggs will continuously develop, whereas other eggs will remain dormant (electronic supplementary material, table S5). There are fewer well-studied examples of insects whose life cycles are not regulated by a diapause than of species with distinct diapause patterns, and these patterns are likely under-represented in the literature [19,56,62,88].

The expression of both types of dormancy at low temperatures in Warramaba likely evolved to prevent hatching during winter. Eggs are small and unable to behaviourally avoid extreme environments, although soil microenvironments where they are oviposited may provide a degree of buffering against extremes [71,89]. Dormancy synchronizes life stages with conditions favourable for growth and reproduction, or by avoiding detrimental conditions [20]. Dormancy can provide increased desiccation tolerance during summer and prevent early emergence [87,90]. This selection pressure would also explain the bias of egg TPCs in univoltine populations towards warmer temperatures as eggs from univoltine populations will enter dormancy when temperatures cool down, whereas multivoltine populations will continue to develop (figure 5).

Third, the patterns of egg development TPCs showed mixed support for latitudinal compensation in nymphal mass and growth rate. We did not detect an overall change in hatching mass across incubation temperatures indicative of a temperature size rule, but there were differences among species (electronic supplementary material, figure S5A). As observed in other insects at high latitudes, univoltine W. flavolineata may compensate for a lower egg development rate by hatching at larger sizes. However, the small hatching sizes we observed in W. grandis were inconsistent with this prediction (electronic supplementary material, figure S5A). Size may also be regulated by maternal investment or differential allocation of resources during development, which were not considered in this study [91,92]. Warramaba may have compensated for latitude in their first instar growth rate (electronic supplementary material, figure S5B); when nymphs were held at the same temperature until the second instar, the magnitude of the difference between species in their growth rate decreased (electronic supplementary material, figure S5B). The pattern of egg development in W. grandis does not explain the more mature demography of instars observed in higher-latitude populations of W. grandis than in lower-latitude populations (electronic supplementary material, figure S3). This trend in demography could be generated by responses in other life stages, such as a prolonged adult stage or slower growth rate at higher latitudes. However, we did not find evidence that W. grandis from different populations grow slower at 30°C in the laboratory (electronic supplementary material, figure S4A). Overall, the variation in egg TPCs among species and region may be latitudinally compensated to various degrees in traits not considered here.

Our results highlight the significance of the egg stage for generating an adaptive variation for insect life cycles in variable climates. In Warramaba, we could infer phenology based on field observations and characterization of egg TPCs in the laboratory. We have used hatching phenology as a proxy for voltinism. However, our field observations were limited during winter; further intensive sampling during autumn, winter and spring is needed to refine our understanding of the phenology of Warramaba. Our results also demonstrated the utility of assessing thermal performance across the entire range of ecologically relevant temperatures. We did not detect a latitudinal gradient in development rate at 30°C (electronic supplementary material, figure S4A). When comparing development rates at 30°C with mean soil temperatures simulated using a microclimate model, there was a counter-gradient pattern of variation; populations of W. picta were not associated with the highest soil temperature despite having the fastest rate (electronic supplementary material, figure S4B). In this case, plastic or acclimatory responses may have masked the co-gradient variation we observed in the TPC across all experimental temperatures [59,93–95]. Overall, the latitudinal patterns of egg TPCs in Warramaba were only explained by considering both the evolutionary constraints of voltinism and latitudinal changes in temperature [45,46].

A growing number of studies have concluded that a holistic assessment of environmental and life-history variation along latitudinal gradients is needed to understand the thermal adaptation of species and the potential evolutionary responses to climate change [16,77,96,97]. Theoretical models can improve our understanding of the evolution of TPCs under variable environments [60]. Climate change has generated renewed interest in thermal adaptation and evolutionary models of TPCs under variable environments [14,98]. Continuous development and testing of TPC models will aid in our understanding and our ability to predict potential climate change responses and explain exceptions to the patterns of thermal adaptation under current evolutionary frameworks.

Ethics

Grasshoppers were collected under a Licence to Take Fauna for Scientific Purposes (SF010681 and 08-00112-1) issued by the Department of Parks and Wildlife, Government of Western Australia to M.R.K.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material.

Authors' contributions

M.R.K. and A.A.H. conceived the study and helped edit the manuscript. J.D.K. and M.R.K. designed the study. M.R.K. collated the phenology data. J.D.K. conducted the experimental work, analysed the data and drafted the manuscript. All authors conducted the fieldwork and gave final approval for publication.

Competing interests

We declare we have no competing interests.

Funding

This research was supported by an ARC LIEF grant (LEI150100083) and an ARC Discovery Project (DP160100279) awarded to M.R.K. and A.A.H., as well as by an Australian Government Research Training Program Scholarship and a Holsworth Wildlife Research Endowment—Equity Trustees Charitable Foundation and the Ecological Society of Australia awarded to J.D.K.