Colour polymorphisms correspond to the co-occurrence of several distinct colour morphs that vary in hue and/or brightness, independently of sex, age or any other state-dependent modifiers. Colour morphs could represent different life-history strategies, maximising their fitness locally in the trait space through trade-offs between all their biological functions. This mechanism could play a role in the maintenance of the green-brown polymorphism in Orthoptera.

Grasshoppers are characterized by a widespread green-brown polymorphism and continuous variability in brightness within colour morphs. It has previously been found that brown individuals are warmer in the field than green conspecifics, but it is unclear if these differences are related to thermal physiology and/or thermal preferences. Therefore, we experimentally tested the thermal physiology and thermal preferences of three green-brown polymorphic species of acridid grasshoppers.

We found no differences between green and brown grasshoppers, either in heat-up and equilibrium temperature patterns, or in temperature preferences. Nor did we find support that the brightness variation is involved in the thermal physiology of these species. Instead, we show that body mass shapes the thermal physiology, with heavier individuals heating more slowly, and that males heated up faster and reached higher equilibrium temperatures than females. As females are heavier than males, the sex differences might be largely explained by size differences.

Our results suggest that neither the thermal physiology nor the thermal preferences explain temperature differences in the field. However, green and brown individuals might still select different microhabitats in the field, which may indirectly lead to differences in body temperature. The persistence of the green-brown polymorphism may result from other mechanisms such as niche partitioning via microhabitat choice, mating preferences or frequency-dependent apostatic selection.

Sampling

We studied the colour morphs of Acrida ungarica, Pseudochorthippus parallelus, and Gomphocerus sibiricus (hereafter simply referred as to ungarica, sibiricus, and parallelus, respectively). Individuals were collected in the field during the 2023 summer season (in mid to late June in east-central Germany for parallelus; in mid-July and mid-august in northern Italy for ungarica; in mid-July in the French Alps for sibiricus). Ungarica were collected as nymphs, parallelus and sibiricus were collected mostly as nymphs. The first sampling of ungarica led to an over-representation of males because field sex ratios were highly unbalanced (17:2, males:females). We, therefore, oversampled females to compensate for missing individuals during the second sampling campaign.

All three species are green-brown polymorphic (Figure 1). Ungarica features two colour morphs: green and brown. Sibiricus presents green and brown individuals. Brown sibiricus can be subdivided into a plain-brown morph and a pied morph. Though, the difference between brown and pied is rather subtle and both are characterized by the lack of green (Schielzeth and Dieker, 2020). Due to limited sample sizes of pied morphs and since field results – that motivated this study – pooled pied and plain-brown (Köhler and Schielzeth, 2020), we considered as the ‘brown’ morph both pied and plain-brown individuals. Parallelus features four colour morphs characterised by dorsal-lateral variation in the distribution of green. Individuals can be uniform green (laterally and dorsally green), lateral green (laterally green and dorsally brown), dorsal green (laterally brown and dorsally green), or uniform brown (laterally and dorsally brown). All species show continuous colour variation in brightness within their discrete colour morphs, resulting in significant inter-individual differences. Body colours may range from light-yellowish green (and brown) to rich dark green (and dark brown, sometimes almost blackish).

Housing & weighing

Field-caught grasshoppers were kept and raised to adulthood under controlled conditions in the lab (70% humidity, temperature around 20 °C at night and around 35 °C during the day). Grasshoppers were fed ad libitum with freshly cut grass blades, provided with a water tube for humidity and exposed to visible and UV lights (visible lights on from 6 am and 10 pm, and UV lights on from 8 am to 8 pm). Only adults with fixed colour morph were used for the experiments and, therefore, the colour morph of ungarica individuals remained unchanged throughout the experiments. Prior to the experiments, all individuals were weighed with a microbalance (XS105 Mettler Toledo, ± 0.1 mg). Each individual was tested in the same day in both experimental set-ups presented below (thermal preferences tested between 10 am and 1 pm, thermal physiology tested between 1 pm and 7 pm).

Thermal preferences

We tested thermal preferences using thermal gradients. Each gradient setup was made of an 88 x 41.5 x 0.8 cm Aluminium plate mounted on 22 cm high Aluminium feet. The preferred body temperature of different acridid species is known to range from 22 to 41 °C (Harris et al., 2013; Springate and Thomas, 2005; Blanford and Thomas, 2000). The thermal amplitude of the gradients was setup to contain this range by bathing one foot in ice and the other in water heated to 60 °C (using an immersion heater [Rommelsbacher TS 2003, 2000W] and a thermostat [Schego thermostat TR2]). This yielded the runways to range from 10.0 ± 1.1 °C to 51.0 ± 2.3 °C (mean ± SD). Plexiglass dividers were mounted vertically on top to create 4 identical runways of 80 x 8 cm. The lower part of the Plexiglas dividers was coated with Fluon (PTFE) up to 12 cm height to prevent grasshoppers from climbing the walls. The gradients were covered with a thin plexiglass lid allowing the direct observation of the grasshoppers’ positions without disturbance. The setup was lit with fluorescent tubes. Using 4 identical setups, we were able to test 16 individuals in one session. All recording sessions took place between 10 am and 1 pm.

Before starting the experiment, the temperature of each runway was measured at 5 equally-spaced positions (distance from cold edge: 0, 20, 40, 60, and 80 cm) using a handheld infrared thermometer (Farnell, dual focus infrared thermometer, ± 0.1 °C). We interpolated temperatures at intermediate positions by fitting a degree 3 polynomial with the nls function from R. Since the drift in temperatures for the duration of the experiment was relatively low (0.9 ± 0.2 °C, mean ± SD), we measured the temperature of the gradient only once at the beginning of the session. Individuals were then released in the middle of the runway (27.1 ± 1.3 °C, mean ± SD) and were free to move for an hour. Their positions were recorded every 10 min leading to 6 measures per individual. Their positions were converted to temperatures for further analyses using the nls fits described above. When an individual was seen perching on the wall or the lid, it was gently directed towards the middle of the runway and missing data were recorded for that reading (1.6% of all sightings).

On each gradient setup, we ensured that morphs and as much as possible sexes were balanced to minimise confounding effects (see Thermal physiology since we used the same groups of animals in both experiments).

Thermal physiology

Adaptive variation in animal colouration, particularly in visible reflectance, represents a trade-off between various competing functions, including camouflage and thermoregulation. Alternatively, adaptive variation in infrared reflectance is likely driven predominantly by thermoregulatory needs, as most animal visual systems are largely insensitive to near-infrareds wavelengths (Stuart-Fox et al., 2017). Hence, we tested heat-up patterns and equilibrium temperatures under a radiant infrared regime. To monitor internal body temperature, we inserted the tip of a K-type thermocouple connected to a datalogger into the ventral side of the thorax of living grasshoppers piercing the intersegmental membrane between the thorax and the abdomen. The thorax was then glued (KIMTEC sekundenkleber speed 40, ethyl-2-cyanacrylat) onto a small pole of 7 mm in diameter. The pole was mounted onto a platform that held the thermocouple so that the setup stayed immobile throughout the experiment. Once glued, grasshoppers were placed in a freezer for 10 min to chill them to a temperature of 5.8 ± 3.0 °C (mean ± standard deviation, SD). The grasshoppers were then placed under an infrared lamp emitting near- and mid-infrared radiation (Elstein HTS/1, 230 V, 400 W, wavelength range: 2-10 µm) placed 41 cm above the grasshoppers. The K-type thermocouples recorded the internal temperature of the thorax every 5 seconds. At the same time, a thermal camera (HIKMICRO B20, ± 2 °C) positioned in front of the experimental setup at an angle of about 60° took pictures every 20 seconds. Using the spot function of the HIKMICRO Analyzer software (v.1.3.1.5), we measured the external temperature of the thorax.

One experimental session comprised 4 grasshoppers. We ensured that morphs were balanced (2 greens and 2 browns for ungarica and sibiricus, and 1 individual of each 4 morphs for parallelus) within sessions and that sexes were also balanced as much as possible (due to unbalanced sex ratios in ungarica, 4 sessions were sex-balanced while 15 were male-only, and 13 were female-only). The experiment lasted for 13.2 ± 0.7 min (mean ± SD), which allowed to reach equilibrium for most experimental sessions. For the analyses, we truncated the 20 first seconds, because internal temperatures kept going down during this period in 24% of the individuals, likely due to temperature homogenisation via circulating haemolymph and were not representative of the heat-up effect of the radiant lamp.

Brightness measurements

Individuals were placed in the refrigerator over the night following the experiments. The next day, body colour was measured using a handheld spectrophotometer (Avantes, AvaSpec-ULS2048x16; optic fibre: FCR-7UVIR200-2-1.5X100, 1.5 mm diameter) coupled with a deuterium-halogen light source (Avantes, AvaLight-DH-S). The device was calibrated with a commercial white standard (Avantes WS-2) on each day. The AvaSoft 7 software (Avantes, v.7.8) captured the reflectance spectra configured with an integration time of 100 ms and an automatic averaging of 5 readings. Reflectance between 300 and 1000 nm was measured in a dark room, holding the probe at a 45° angle while gently touching the cuticle surface. Our device always displays a narrow peak in reflectance between 654 and 659 nm. This peak apparently represents an artefact of the device and was replaced by the average across reflectance at 650-654 nm and 659-664 nm.

Reflectance spectra were captured for the dorsal side and the right lateral lobe of the pronotum. Grasshopper body parts display complex colouration patterns (Figure 1). To account for this variability, we took 5 independent measurements of both body parts. For each individual, we averaged reflectance across these 10 measurements using the aggspec function from the pavo package (version 2.9.0, on R) (Maia et al., 2019). This led to a total of 432 reflectance spectra. We then derived the brightness value from each spectrum using the summary.rspec function from pavo by extracting the mean brightness. Mean brightness, simply referred to as brightness hereafter, corresponds to the mean relative reflectance over the entire spectral range, i.e. 300 – 1000 nm here. We deem this measure of brightness to be physiologically relevant, but note that it is different from how animals perceive brightness, since perception depends on the specificities of the visual system.

Replication statement

For a description of the replication design we used for each of our three predictors (colour morph, brightness, and weight), see Table 1.

Statistical analyses

All descriptive statistics given in the results are presented as the mean ± standard deviation. All statistical analyses were performed on R version 4.2.2 (R Core Team, 2023).

We analysed the three species and two sexes separately by subsetting species and sex (six subsets in total). Within each subset we tested for morph differences as well as for effects of brightness and body weight as predictors. Confidence intervals (CI) were estimated by non-parametric bootstrapping with 1000 iterations. Bootstrap resamples were made out of complete time series of individuals that were resampled with replacement, while constraining each resample to have the same sample size for each colour morph as in the original subset. For each bootstrap iteration, we calculated the statistic of interest – differences in means between morphs or Spearman’s correlation coefficients between brightness (or weight) and the response variable (internal temperature, external temperature or substrate temperature). We derived the 95% CI from the distribution of the statistic of interest using the percentile method. Differences in means or correlation coefficients were treated as significant if the bootstrapped 95% CI did not contain zero. When testing the colour morph effect (e.g. mean_green - mean_brown), a CI above zero indicates that green individuals were hotter (when internal or external temperatures are tested) or that they were standing on hotter spots (when the substrate temperature is tested) than brown individuals. When testing the brightness (or weight) effect, a positive effect size means that brighter (or heavier) individuals were hotter or that they were standing on hotter spots than darker (or lighter) individuals.

We derived three summary statistics from the time series. From the thermal physiology time series, we derived the heat-up speed, which represents the rate of temperature change between the 20^th and the 240^th second of the experiment, and the equilibrium temperature, which averages the temperature reached by an individual between the 660^th and 760^th second of the experiment. Both of those statistics have been derived for the internal and external temperatures. From the thermal preferences time series, we derived the preferred temperature, which corresponds to the average temperature of the substrate where the individual was standing between the 30^th and the 60^th minutes.

Normality was checked from quantile-quantile plots. Homoscedasticity was checked using Fisher tests (two levels) or Bartlett tests (more than two levels). When heteroscedasticity was found, data were log-transformed to meet homoscedasticity when possible (it only concerned the heat-up speed for parallelus males tested against morph and for sibiricus tested against sex). The morph effect was tested using t tests (when normality and homoscedasticity were met), Welch tests (when normality only was met), and Mann-Whitney tests (when normality was not met) for ungarica and sibiricus, and using one-way ANOVAs for parallelus. For the Kruskal-Wallis tests featuring a p < 0.05, means were separated using pairwise Mann-Whitney tests. The brightness and weight effects were tested using Pearson’s correlation tests when bivariate normality was met or Spearman’s correlation tests otherwise. Confidence intervals (CI) were derived from correlation coefficients of 1000 bootstrap resamples (sampling with replacement while constraining the resample to have the same sample size as its original dataset). Correlation coefficients were treated as significant if the bootstrapped 95% CI did not contain zero. The sex effect on the summary statistics was tested using Mann-Whitney tests when normality was not met, and using t tests otherwise. Tests on equilibrium temperatures of ungarica females were not performed because no equilibrium was reached (Figure 3, Figure S1).