Ontogenetic changes in the body temperature of an insect herbivore

Authors


Summary

  1. Over ontogeny, many insect larvae grow substantially, through at least several orders of magnitude in body size. Increasing size can profoundly change how individuals interact with their environments, by altering the opportunities for, and constraints on, feeding, changing the relative risk and sources of predation and shifting the relative importance of physical factors in the environment.

  2. Here I use eggs and larvae of Manduca sexta, which are herbivores on solanaceous plants in the south-western US, to examine how body size affects body temperature. Larvae grow in excess of 10 000-fold by mass in a few weeks, from 0·001-g hatchlings to 12–15-g fifth-instar larvae.

  3. Using infrared thermography, I show that increasing body size leads to large changes in body temperature: over ontogeny, average larval temperature increased by 3–7 °C. The temperatures of eggs, hatchlings and early larval instars were coupled to leaf temperatures (Datura wrightii), which were much cooler than ambient air temperatures. The temperatures of larger larvae, by contrast, were similar to air temperatures, or somewhat higher.

  4. Changing body temperatures reflect that small and large larvae were immersed differentially in leaf boundary layers, received different amounts of incoming solar radiation, and used thermal heterogeneity on leaf surfaces in different ways.

  5. I develop a simple species distribution model that links maximum observed air temperatures in the south-western US with known thermal tolerances of eggs and larvae. This model predicts that eggs of M. sexta can occupy significantly larger fractions of the landscape than can large larvae.

  6. Large differences among stage-specific microclimates, such as those observed for M. sexta, are likely to be general features for insects and other organisms whose body sizes span large ranges, and stage-specific microclimates pose general and largely unrecognized problems for species distribution models.

Introduction

Significant theory has focused on estimating how body size affects body temperature and thermal ecology (Stevenson 1985; Huey 1992; Porter & Kearney 2009; Gardner et al. 2011). The effects of size, however, generally are analysed with respect to adults and at the level of species – some focal species have large adult bodies and others have small bodies, and one analyses systematic theoretical and empirical differences between them. Oddly, links between body size and thermal ecology are rarely applied to shifts in body size over ontogeny, perhaps because analyses focus on nongrowing adults.

For herbivorous insects, at least, this is a serious oversight for two reasons. Firstly, many insects grow by large factorial amounts; for example, caterpillars often pass through several orders of magnitude in body size. Increasing size can profoundly change how larvae interact with their environments (Reavey 1993), by altering opportunities for and constraints on feeding (Reavey 1993; Hochuli 2006), changing the relative risk and sources of predation (Cohen et al. 1993; Costa 1993) and shifting the relative importance of physical factors in the environment (Schmidt-Nielsen 1984; Casey 1992; Gardner et al. 2011). Secondly, insect herbivores are closely associated with their substrates (host plants), which can create microhabitats whose characteristics diverge markedly from conditions nearby (Bakken 1989). This effect is particularly pronounced for leaves. Because leaves exchange heat and water at high rates with their environments, they can sustain, in their boundary layers, high vapour pressures and temperatures that differ substantially from ambient (Willmer 1982; Pincebourde & Woods 2012). During ontogeny, many insects pass through a range of body sizes that spans the typical thickness of leaf boundary layers (Woods 2010), suggesting that early, small stages live in microenvironments dominated by host plant physiology, whereas later, larger stages grow out of those influences and integrate conditions at larger spatial scales.

Understanding ontogenetic patterns of thermal exposure, for insect herbivores in their natural environments, will be central to understanding ecological shifts during climate change (Parmesan et al. 1999; Bale et al. 2002; Parmesan 2006). The availability of better spatial and temporal data on climatic conditions has led recently to the development of species distribution models (SDMs), which aim to predict distributions and range limits from environmental data (Elith et al. 2006) and, increasingly, from mechanistic models of organismal physiology and biophysical exchange (Helmuth 2009; Kearney & Porter 2009; Sears, Raskin & Angilletta 2011). Current SDMs, however, tend to idealize organisms as having a mean body size, a fixed physiology, etc. [for an exception focusing on a different range of body sizes, see (Kearney, Matzelle & Helmuth 2012)]. In fact, species consist of individuals whose physical and ecological properties can differ radically – because of variation in phenotypes among individuals, differences in life stages across complex life cycles (Kingsolver et al. 2011) and increases in body size during growth. If these sources of variation are large, then which set of traits of all those exhibited does one build into an SDM? How should changing sets of organismal traits be weighted across life cycles?

Here, I illustrate the magnitude of this issue by measuring body temperatures of an insect herbivore across the egg and larval stages. Temperature plays a central role in the lives of insects (Bale et al. 2002; Harrison, Woods & Roberts 2012). It affects critical biological processes, including rates of movement, feeding and growth (Sherman & Watt 1973; Casey 1993). It also affects processes at the level of population and above – including ecological and evolutionary aspects of population growth rates (Frazier, Huey & Berrigan 2006), species interactions (O'Connor, Gilbert & Brown 2011) and the evolution of thermal performance curves (Asbury & Angilletta 2010). Despite the importance of temperature, larvae of insects in general do not regulate their body temperatures, although many exceptions are known (Heinrich 1993; Serratore, Zalucki, and Carter 2013).

Specifically, I examine the consequences of ontogenetic increases in body size for the body temperature of eggs and all larval instars of a model insect, Manduca sexta (Sphingidae), on one of its primary host plants, Datura wrightii (Solanaceae), in the field in south-east Arizona. M. sexta are widespread in North American south of ~43°N (Opler, Lotts & Naberhaus 2012), with populations throughout agricultural areas of the eastern US and in the deserts of the American Southwest. The biotic and abiotic conditions limiting local populations are largely unknown, although body temperatures of eggs and larvae in the south-western deserts frequently approach their upper thermal limits (Casey 1976; Potter, Davidowitz & Woods 2009). A prior study of wild fifth-instar M. sexta, in the Mojave Desert, concluded that larval body temperatures follow air temperatures (Casey 1976). However, smaller larvae and eggs may have very different thermal experiences – because they live in microhabitats that are more strongly influenced by leaf physiology (Potter, Davidowitz & Woods 2009; Pincebourde & Woods 2012). I tested this hypothesis by measuring temperatures of M. sexta and their host-leaf tissues using an infrared video camera. These data I then integrated into a simple model of geographic variation in thermal exposure to estimate fractional occupancy of the landscape in south-eastern Arizona by eggs and large larvae.

Materials and methods

Species, locality and weather

This study was carried out in July and August 2011 at field sites located within 20 km of Portal, Arizona (USA) [31·923843° lat, −109·130025° lon, 1425 m elevation] and at the Southwestern Research Station (American Museum of Natural History). Two species of Manduca (Lepidoptera: Sphingidae), M. sexta and a congener M. quinquemaculata, are common in this area. In over 3 years of work at these sites, including 2011, >90% of individuals observed in the field were M. sexta, and I therefore focused on this species during the infrared measurements. Nevertheless, eggs and early-stage larvae (L1 & L2) of the two species are difficult to distinguish, and some of the younger individuals in this study may have been M. quinquemaculata. This issue should have little bearing on the conclusions of this study, as the feeding ecology and growth trajectories of the two species are similar. Although M. sexta uses a variety of hosts in the south-western deserts of the United States, we filmed it only on its primary host Datura wrightii (Solanaceae). In Arizona, the major pulse of activity by M. sexta (adult flight and oviposition) is associated with summer monsoon rains, which usually start in mid-July and go for 6–8 weeks.

Weather was logged at a fixed weather station, with sensors ~ 3 m above the ground. Logged variables included solar radiation (S-LIB silicon pyranometer, Onset, 300–1100 nm), air temperature and relative humidity (S-THB-M, Onset) and wind speed (S-WCA-M, cup anemometer). Signals from the sensors were logged onto a computer (U-30, Onset) and retrieved by shuttle every few days. Weather data were logged every 5 min day and night, from the first day on which IR recordings were made (July 22, 2011) to the last day (August 8, 2011), for a total of 18 days.

Components of heat balance

In the methods below, I describe a set of measurements aimed at understanding egg and larval heat balance over ontogeny. The first section describes methods for quantifying patterns of body temperature as a function of stage and body size. The subsequent three sections describe major factors that affect egg and larval heat balance. These factors include, firstly, how far individuals project from the leaf surface, which is important because leaf and air temperatures can differ strongly; the distance away from the leaf that an individual project determines the relative weighting of these two sources of heat exchange. Secondly, I measure exposure to direct sunlight, which markedly affects incoming heat from solar radiation. Thirdly, I quantify the placement of eggs and larvae with respect to patterns of thermal variation on leaf surfaces, which will affect conductive heat exchange.

Ontogeny of body temperature

Short videos (1–5 min) of Manduca on D. wrightii were taken at each field site with a calibrated mid-wave infrared camera (3–5 μm sensitivity; ThermoVision SC4000, FLIR, Boston, MA, USA; frame rate set to 10 Hz). For portability, the camera was powered by a solar panel, and videos were recorded onto a laptop computer. All videos were taken during the daytime, between 9 am and 2 pm, except for a small set of nighttime videos. In addition, during each video, I recorded local air temperature and relative humidity 1 m above the ground (the height at which most leaves were filmed) using a Hobo U23 external temperature/relative humidity data logger. The data logger's sensor was protected from direct sunlight by a radiation shield.

Manduca were found by searching on leaves of D. wrightii, both around peripheral leaves and in the centres of plants. When found, eggs and larvae were filmed in place (if larvae were moving, I waited until they stopped and started to feed before filming them). For each Manduca filmed, I estimated the per cent of its body exposed to direct sunlight. While filming, I briefly turned over each leaf (for a few seconds), so that the egg or larva was visible to the camera. Just after turning over the leaf, I immediately shaded it with a cardboard square. Shading was important because reflected infrared radiation, especially from direct sunlight, could lead to errors in estimated leaf and larval temperatures (Leigh et al. 2006). Long-term shading would also cause problems by reducing heat inputs from solar insolation. Therefore, I estimated leaf and larval temperatures from still images extracted from the video sequences in the first 0·3 s after shading, before any significant changes in temperature occurred.

Temperatures were extracted in ExaminIR software (FLIR) using tools for examining specific regions of interests (ROIs). Egg ROIs consisted of a few pixels in the centre of the egg. Larval ROIs were lines drawn along the larval midsection from the head to the posterior segments, with temperatures averaged along the line (in general, larval body temperatures showed little longitudinal variation). To estimate surrounding leaf temperatures, I drew ROIs that sampled a doughnut-shaped region around each individual, with the width of the doughnut approximately twice the diameter of the Manduca body; these doughnut-shaped areas included leaf mesophyll and nearby veins if they were present. In all leaf ROIs, care was taken to exclude holes from prior feeding damage, as these sometimes let through views of objects in the background that had much different temperatures.

Information on the calibration of the camera and sensors, and on the emissivity of larval cuticle, can be found in the Supporting Information.

Instar-dependent projection from leaf surfaces

On a subset of eggs and larvae from the field, I measured body mass and the distance that they projected from the surface of their host leaf (D. wrightii). For L2–L5 larvae, I placed the end of a ruler against the leaf surface and determined how far the dorsal side of the caterpillar projected. For eggs and L1 larvae, I photographed them, with the leaf surface parallel to the field of view, through a stereomicroscope at magnifications of 10–20×. In all photographs, there was ruler in the background, which I used to calibrate distances. Images were subsequently analysed in ImageJ (version 1.45s) (Rasband 2012) to obtain projection distances.

Because larvae often cling to leaf veins, and because small larvae cling preferentially to higher-order veins, understanding the morphological characteristics of veins is important for assessing how far larvae project away from the leaf mesophyll. Methods for measuring leaf vein morphology are presented in the Supporting Information.

Instar-dependent exposure to direct sunlight

For most larvae that were filmed in IR (= 75), I also noted whether they occurred on the abaxial (lower) or adaxial (upper) side of the leaf and what proportion of the individual was exposed to direct sunlight.

Instar-dependent changes in larval location and temperature of leaf veins

Preliminary thermal images indicated that leaves had consistent patterns of temperature: petioles were hottest and leaf mesophyll coldest, and the thermal transition between them followed the branching pattern of veins, which are important footholds for larvae. To analyse the effect of this pattern, I quantified both the position of a subset of Manduca on their host leaves (= 109 Manduca; a few larvae were excluded because their positions could not be determined unambiguously) and used thermal images to estimate, for each of 68 leaves, the temperatures (abaxial side only) of its petiole, the midvein, a secondary vein and the leaf mesophyll (higher-order veins were generally indistinguishable from mesophyll).

Biogeography of heat stress

To illustrate the differential effects of life stage on exposure to extreme climatic conditions, I developed a simple species distribution model that couples information on critical thermal maximum of eggs and larvae with geographically distributed data on air temperatures in the south-western deserts of the US. Climatic data were downloaded from the Daymet archive (Thornton et al. 2012), which provides daily data for North America at a resolution of 1 km. Data for the years 1998–2008 were obtained for tiles 11 015, 11 016, 10 835 and 10 836, which are contiguous and cover a block bounded by 30–34°N latitude and 108–112°W longitude. These tiles cover the core of M. sexta's range in the northern Sonoran and Chihuahuan deserts. The total area of the tiles is 158 249 km2, and most of the pixels represent potential habitat for M. sexta. Using scripts written in R (R Development Core Team 2010) and several packages (ncdf, raster, rgdal), I picked out the hottest air temperature (tmax, 2 m off the ground) that occurred during the coverage time (for each pixel). I then used this map of hottest temperatures to estimate suitable thermal habitat for both eggs and fifth-instar (L5) larvae of M. sexta. I assumed that eggs reached maximum temperatures 4 °C cooler than maximum air temperature and that L5 larvae reached maximum temperatures 1 °C warmer than maximum air temperature (see Fig. 2). For eggs, I assumed that the critical thermal maximum was 43·3 °C. This value is the LD50 for eggs from these populations exposed to 2-h heat shocks (Potter & Woods 2012). For L5 larvae, the critical temperature is less well known, but likely is ~ 1 °C higher (here I assume 44·3 °C). Larval growth rates were close to zero at 42 °C (Kingsolver & Woods 1997), and third-instar larvae had upper thermal limits (30-min exposure) of 44–45 °C (Casey 1976).

Data and analyses

Data were analysed by linear modelling and linear mixed-effects models using scripts written in R (R Development Core Team 2010). In all cases, data were assessed for normality, and fitted models were verified by examining residuals for patterns. In some cases, to meet these criteria, data were log-transformed.

Results

Site characteristics

Data from the weather station are summarized in Table 1. Because IR recordings were taken only between 9 am and 2 pm, I first filtered the data, keeping records just from those periods. A figure showing full time series of the weather variables is available in the Supporting Information.

Table 1. Summary of weather at the study site, near Portal, Arizona, during the sampling period 22 July 2011–8 August 2011
 Solar radiation (W/m2)Temperature (°C)Relative humidity (%)Wind speed (m/s)Wind gusts (m/s)
Mean76228·548·12·033·87
95% CI of mean743–78028·3–28·747·3–48·91·97–2·103·77–3·98
Mean max115131·662·13·676·80
Mean min28625·137·00·591·53

Ontogeny of body temperature

Altogether I filmed 38 eggs and 76 larvae (= 15 L1, = 7 L2, = 14 L3, = 14 L4 and = 26 L5; see examples in Fig. 1) found on 106 different leaves distributed among 43 individual plants. During filming, mean air temperature at leaf height was 30·4 °C (range 24·3–36·3 °C) and mean leaf temperature was 27·7 °C (range 21·8–35·8 °C). Thus, leaves were in general cooler than ambient air. This effect has been observed before (Smith 1978; Potter, Davidowitz & Woods 2009) and likely arises because leaves of D. wrightii transpire at high rates (H.A. Woods & J.K. Wilson, unpublished data). The mean body temperature of all filmed Manduca (= 114) was 30·1 °C (range 22·9–38·3 °C). However, the single value for mean body temperature across all stages hides large, size-dependent shifts in thermal ecology (Fig. 2). Eggs and L1 larvae matched leaf temperature closely regardless of local air temperature, whereas the largest larvae (L5) more closely matched air temperature, and in many cases were warmer than air temperature. The intermediate larval stages (L2–L4) fell between these extremes (Fig. 2). Linear modelling in R, which included main effects of stage and difference between air and leaf temperatures, and their interaction, showed that all three terms were highly significant (Table 2). Separate video recordings of L5 larvae during nighttime (= 6) showed that air temperatures were about 1 °C warmer than leaf temperatures; during the night, body temperatures of L5 larvae matched air temperature to within 0·2 °C.

Figure 1.

Thermal images of Manduca sexta on Datura wrightii, its primary host plant in the Chihuahuan desert. Two eggs, weighing ~1·6 mg each (a), a first-instar larva weighing 8·7 mg (b), a third-instar larva weighing 205 mg (c) and a fifth-instar larva weighing 4·5 g (d). All temperatures are in °C, and white arrows point to M. sexta. All eggs and larvae were on the abaxial side of the leaves. Images were extracted from video series (10 frames per second) during which leaves were briefly turned so that their abaxial sides were towards the camera. A few seconds after turning the leaves, the scene was shaded briefly by a cardboard square held between the leaf and the sun; extracted images are from the first 0·3 s after shading. The high temperatures in the background were from exposed ground surrounding the host plant.

Figure 2.

Ontogenetic changes in body temperature of Manduca sexta. (a) Manduca and air temperatures scaled to leaf temperatures. The x-axis represents air temperature (Tair) scaled to leaf temperature (Tleaf); most of the values are >0, meaning that Tair was almost always greater than Tleaf. The y-axis represents Manduca body temperature (TManduca) scaled to leaf temperature; almost all of the values are positive, meaning that insects were in general warmer than their corresponding leaves. The horizontal dotted line indicates the relationship expected if Manduca and leaf temperatures were identical. The diagonal dotted line indicates identical Manduca and air temperatures. Temperatures of early stages (eggs and L1 larvae) closely matched leaf temperatures. Later stages were vertically elevated, meaning they were progressively warmer than leaf temperature. The body temperatures of L5 larvae were similar to, or elevated above, air temperature. (b) Box and whisker plots showing progressive elevation of Manduca body temperature above leaf temperature through ontogeny. L4 larvae were the exception, which represents by chance that they were filmed under particularly low air-to-leaf differentials in temperature.

Table 2. Analysis of data in Fig. 2. The dependent variable Y was (difference between Manduca and leaf temperature), and the independent variable X was (difference in temperature between air and leaf). The data were modelled in R, using the function lm() and model structure log10(1 + Y)~X*Stage, with Stage treated as a discrete numerical predictor. Stage refers to eggs and all five larval instars. Adjusted R2 = 0·87
 Estimated coefficientStandard error of coefficientt-value P
Intercept0·14240·01927·40<0·00001
X 0·03410·00585·88<0·00001
Stage0·07740·006911·2<0·00001
X × Stage0·00530·00173·030·0031

Instar-dependent projection from leaf surfaces

Compared with eggs, L1 larvae projected from leaves slightly less far – because they unroll and stretch out once they leave the egg. Subsequently, projection distances increased regularly with body mass (Fig. 3). For larvae (excluding eggs), the fitted scaling exponent (b) was 0·345 (std. error = 0·0046, R2 = 0·99, t = 74·8, < 0·00001), which is just above what we would expect if larval growth were isometric (⅓). A figure on the morphology of leaf veins can be found in the Supporting Information. Primary and secondary veins lifted larvae 0·5–2 mm away from the surface of the mesophyll [see Supporting Information].

Figure 3.

Scaling of the distance Manduca project from the leaf surface in relation to estimated maximum and minimum thicknesses of the boundary layer over the leaf. Maximum thickness (=3 mm) was calculated for large leaves (characteristic dimension = 10 cm and slight wind of <0·5 m/s) whereas minimum thickness was calculated for small leaves (characteristic dimension = 3 cm and wind speed = 10 m/s) (see (Woods 2010) for details).

Instar-dependent exposure to direct sunlight

Larvae in instars 1–3 were almost always completely shaded, at least during the middle of the day when I took measurements. In separate observations, we found that early-instar larvae placed experimentally on the adaxial side of the leaf, towards the sky, rapidly crawled back under the leaf into the shade (a typical response; see also Sherman & Watt 1973). By contrast, L4 and L5 larvae were often partially exposed to direct sunlight. For larger fifth instars, this effect reflected that their bodies outgrew the dimensions of their host leaves, especially when they had already consumed portions of it. Because they were most likely to be exposed to direct sunlight, I used fifth instars to examine the effects of direct sunlight on body temperature. I plotted residual body temperatures (L5 only) from the linear model fitted to the data in Fig. 2 against percentage of the larval body in direct sunlight. There was a clear positive relationship (Fig. 4b), and linear regression indicated that the relationship was highly significant (R2 = 0·33, slope = 0·0023, std. error = 0·0007, t = 3·55, = 0·0017).

Figure 4.

(a) Box and whisker plots of per cent of body exposed to direct sunlight, for each larval stage. (b) Residual body temperatures of L5 larvae (calculated from model fit to data shown in Fig. 2) plotted against% of the larval body that was in direct sun. Note that residuals are in units of log10(temperature difference + 1).

Instar-dependent changes in larval location and temperature of leaf veins

Manduca showed strong stage-specific shifts in their location (Fig. 5a; see also Fig. 1). Eggs occurred exclusively on mesophyll. A few L1 larvae held directly onto mesophyll but most held onto secondary veins. Later, larval instars were distributed increasingly on larger veins, and L5 larvae held almost exclusively onto leaf petioles. In parallel, larger veins were hotter (Fig. 5b): petioles were 4·34 °C (0·58 SEM) warmer than mesophyll; midveins were 2·45 °C (0·42) warmer; and secondary veins were 1·06 °C (0·25) warmer.

Figure 5.

(a) Distributions of eggs and all five larval instars across different leaf structures. Eggs occurred exclusively on leaf mesophyll; early-instar larvae were associated with mesophyll and secondary veins, whereas older larvae clung to lower-order veins (midveins and petioles). (b) Relative temperatures of petioles, midveins and secondary veins, each compared with its associated mesophyll temperature. All temperatures were extracted from infrared images of the abaxial side, within 0·3 s of shading. Lines connect measurements taken from the same leaf (= 68 leaves). (c) Scan of the abaxial side of a Datura leaf showing the locations at which infrared temperatures were estimated.

The data were structured in this analysis, because each set of four measurements (petiole, midvein, secondary vein and mesophyll) were grouped by the leaf from which they were extracted. I therefore analysed the data using linear mixed-effects (LME) models in R (Pinheiro & Bates 2009; R Core Development Team 2010), via the function lme() in the package nlme. Data were coded as 0 = petiole, 1 = midvein, 2 = secondary vein and 3 = mesophyll. Whereas the data in Fig. 5 show the temperature data scaled to mesophyll temperature, in the linear mixed-effects model, I used the raw temperature measurements. The model, specified as temperature~location on leaf surface and leaf ID as the random effect, showed that location was highly significant (coefficient = −1·44, std. error = 0·067, t = −21·4, < 0·00001) [see Supporting Information for an additional figure showing raw temperature data].

Biogeography of heat stress

During the period 1998–2008, maximum air temperatures (2 m off the ground) in the sampled block ranged from around 30 °C in the far northeast to >45 °C along the western edge (Fig. 6b). Based on the assumption that eggs were 4 °C colder than maximum air temperatures, eggs would have surpassed their critical thermal maximum (43·3 °C) in only 1928 km2 of the total sampled area (1·2%) (Fig. 6c). By contrast, based on the assumption that they were 1 °C warmer than maximum air temperatures, L5 larvae would have surpassed their critical thermal maximum (44·3 °C) in 46 258 km2 of the sampled area (29·2%).

Figure 6.

(a) Topographic and political representation of the area encompassing our study site (denoted by red circle), near Portal, AZ, and surrounding areas, most of which is suitable habitat for Manduca sexta. The area shown lies within 30 and 34°N latitude and 108 and 112°W longitude. (b) Maximum predicted air temperature during the period 1998–2008, during just the months of June–September, which is the main flight season for Manduca sexta. Data were obtained from the Daymet archive. (c) Area in which egg temperatures are predicted to have exceeded 43·3 °C (green shading), assuming that eggs reached temperatures 4 °C lower than air temperature (see Fig. 2). (d) Area in which L5 temperatures are predicted to have exceeded 44·3 °C (green shading), assuming their body temperatures exceed air temperature by 1 °C.

Data accessibility

Raw data used to construct Figs 2-5 can be accessed at http://dx.doi.org/10.5061/dryad.8sd03.

Discussion

Few organisms live directly in what we call ‘climate’. Rather, they live in and around other objects, which filter climate into the microclimates actually experienced. Collectively, the set of potential filters is large (Huey 1992; Helmuth 1998; Porter et al. 2002; Gedan et al. 2011; Sears, Raskin & Angilletta 2011). For herbivorous insects, the most important filters are leaves (Pincebourde & Woods 2012), which can generate temperatures and humidities that are substantially better for insects than those available nearby (Casey 1976; Willmer 1982).

Leaf-dominated microhabitats, however, extend from within the leaf tissues to at most a few millimetres away from the surface, suggesting that individual insects could grow out of them. The results presented above support this idea: the body temperatures of small stages of M. sexta – eggs and early-instar larvae – were strongly coupled to host-leaf temperature, whereas those of larger larvae came to resemble local air temperatures. Applying a simple version of boundary-layer theory (Gates 1980; Schuepp 1993; Boulard et al. 2002; Boulard 2004; Vogel 2009) to the Manduca-Datura system, Woods (2010) estimated that, for biologically realistic leaf sizes and wind speeds, boundary layers are 0·3–3 mm thick. Eggs and larvae of M. sexta clearly span this range (Fig. 3). Projecting out of the boundary layer exposes larvae more directly to ambient air, which in general is warmer than the leaf surface (Fig. 2). Two other mechanisms contributed to ontogenetic changes in body temperature: shifts in exposure to direct sunlight (Fig. 4a) and differential use of thermal heterogeneity on leaf surfaces (Fig. 5). Larger larvae were warmer because they were exposed more often to direct sunlight, and they tended to hold onto increasingly warm leaf veins. Because the area of direct contact between a larva and its leaf is quite small (tips of prolegs), the overall effect of the latter factor probably is small.

Consequences for eggs and larvae

Systematic changes in body temperature over ontogeny offer size-specific opportunities and risks. Compared with large larvae and eggs, small larvae may have greater scope for behavioural thermoregulation, because they can exploit fine-scale variation in leaf surface temperatures. However, the body temperatures of small insects also are closely coupled to the temperature of their host leaf; consequently, their thermal experience is driven by the physiology and energy balance of the leaves themselves, which are largely out of larval control [but see (Pincebourde & Casas 2006a,b)]. The body temperatures of large stages, by contrast, are decoupled from leaf temperature; their thermal experience is driven more by air temperature and relative exposure to sunlight. Larger larvae also are relatively more mobile and may be able to use larger-scale thermal heterogeneity for behavioural thermoregulation (Casey 1976). These results also imply that larger individuals face greater diurnal variability in temperatures, which may require more eurythermal physiology. A set of studies (Casey 1976, 1977; Reynolds & Nottingham 1985; Kingsolver & Woods 1997; Petersen, Woods & Kingsolver 2000; Woods & Bonnecaze 2006; Kingsolver & Nagle 2007; Potter, Davidowitz & Woods 2009, 2011) over the past 35 years, on the thermal physiology of M. sexta, suggests that later ontogenetic stages have higher critical thermal maxima, although no single study has carried out a rigorous ontogenetic comparison.

For other insect herbivores, how these changes play out will depend on the range of body sizes through which the insect passes, the physiology of their host leaves (Pincebourde & Woods 2012), and the thermoregulatory behaviours that larvae use (if any) when they are too hot or too cold (Serratore et al. 2012). For M. sexta, small stages were cooler because their host leaves were cooler than air. This shift probably exposes later stages to relatively greater risk of heat stress but also may increase rates of feeding and growth during periods of intermediate or low temperature. However, the physiology of Datura is unusual, especially for a desert plant; leaf temperatures of other plant species can be substantially higher than air temperature (Helliker and Richter 2008; Vogel 2009). In these cases, the relative thermal experience may shift so that the earliest stages are hottest or most variable.

Consequences for species distribution models

This study shows that the body temperatures of an insect herbivore, living in a desert ecosystem, increased by up to 7 °C during ontogeny. Late in ontogeny, therefore, individuals should be at the greatest risk of death from high temperatures, as the increasing maximum body temperatures more than offset any variation in stage-specific thermotolerance. I explored the potential magnitude of such an effect using a simple species distribution model. The model indicated that over the time period examined (1998–2008) eggs of M. sexta on D. wrightii were very unlikely to have exceeded their critical thermal maximum, whereas L5 larvae were quite likely to have done so (Fig. 6). These model predictions are consistent with observed abundances of M. sexta in south-eastern Arizona (H. A. Woods, pers. obs.): M. sexta is substantially less abundant along the western edge of the modelled geographic area, perhaps because it is too hot for large larvae.

This simplistic approach ignores many factors that could be modelled (such as cold nighttime temperatures, effects of moderate temperatures on feeding and growth rates, temperature-dependent interaction with parasitoids, etc.), focusing instead just on high air temperature and its interaction with known thermal tolerances of M. sexta. Note too that even sublethal exposure to high temperatures can significantly affect the performance of individuals. For example, eggs of M. sexta exposed to cycling temperatures, with a 2-h daytime plateau at the highest temperatures, hatched significantly later when the high temperature exceeded 39 °C (Potter, Davidowitz & Woods 2009), presumably because embryos were having to spend resources and time repairing damage from the heat shocks.

Nevertheless, the model illustrates the general problem for species distribution models (SDMs) of potentially large changes, across ontogeny, in microclimatic experience [see (Kingsolver et al. 2011)]. Especially for terrestrial insects, which often grow through several orders of magnitude in body size, the microclimates and body temperatures experienced at one size can differ strongly from those experienced at another. This effect significantly complicates the building of SDMs, because it entails either (i) matching stage-specific environmental experience with stage- or size-specific physiological and behavioural capabilities; or (ii) choosing a priori the size or stage that is expected to be most vulnerable to climatic or microclimatic change. In either case, the conclusion is that real species in real environments are substantially more complicated than most SDMs imply. Circumventing this problem would appear to require, at a minimum, allocating more resources and time to understanding the stage-specificity of both climate and physiology.

Acknowledgements

I thank the staff and director of the Southwestern Research Station for laboratory space and access to field sites, and to Chris and Bill Wilbur for providing housing. Thanks also to Joey Latsha, Keaton Wilson and Kristen Potter for discussions, help in the field and laboratory and comments on the manuscript, and to Pat Carter for access to a manuscript still in press. Joel Kingsolver and two anonymous reviewers gave additional comments on the manuscript, and Jared Oyler and Claudine Tobalske helped make Fig. 6. This work was supported by the National Science Foundation (IOS 0844916) and by the University of Montana.

Ancillary