Melanistic coloration does not influence thermoregulation in the crepuscular gecko Eublepharis macularius

Melanistic body coloration has several different functions in vertebrates, such as being involved in visual signaling and organism communication, camouflage, and thermoregulation (reviewed in McNamara et al., 2021). Thermoregulation, the maintenance of an organism's optimal body temperature, is essential for animals for mating, foraging, predator escape, and other biological functions (Dunham et al., 1989; Seebacher and Franklin, 2005). In ectotherms, such as non-avian reptiles, the maintenance of their body temperature, and thus of the temperature-dependent basic life functions, relies on a thermal exchange with their environment (e.g. Kearney et al., 2009; Kiefer et al., 2007). Although moving between microhabitats and posture changes are the most common thermoregulatory behaviors observed in ectothermic vertebrates (e.g. Aubret and Shine, 2010 and references within; Bauwens et al., 1996; Kearney and Predavec, 2000), skin coloration may also influence thermoregulation. Higher concentrations of dermal melanin caused by the concentration of melanophores may result in darker coloration. Since darker coloration absorbs more radiation (Gates, 1980), some ectothermic vertebrates use darker coloration to increase heat absorption, the so-called Thermal Melanism Hypothesis (Clusella Trullas et al., 2007; Clusella-Trullas et al., 2008; Forsman, 1995; Gibson and Falls, 1979; Walton and Bennett, 1993). Vertebrate ectotherms living in areas with low levels of solar radiation – i.e. areas further from the equator – are more uniformly dark in color due to the melanophores distributed relatively equally throughout the dermis to create an overall lower skin reflectance and higher rate of heat absorption. Ectotherms occurring in areas of higher solar radiation, however, are less constrained in increasing heat absorption from solar radiation and can have clusters within their dermis that have higher concentrations of melanophores than other areas, causing for example a spotted pattern (Martínez-Freiría et al., 2020; Moreno Azócar et al., 2015; Szydłowski et al., 2020). Similarly, ectotherms that are mostly active during hours of low solar radiation – i.e. crepuscular and nocturnal species – may rely on other strategies to absorb heat (Garrick, 2008), and melanism in these species may be used for other functions, including for cooling rather than for heating (Geen and Johnston, 2014). Together with stable melanistic coloration (i.e. the melanistic coloration present in a pattern, regardless of whether or not that coloration experiences color change), melanosomes may move within the dermal melanophores, producing a skin darkening when moving towards the surface of the skin, or skin lightening when moving away from it, a phenomenon called rapid physiological color change (Sherbrooke and Frost, 1989). Rapid physiological color change may be used for thermoregulation by darkening or lightening the skin surface of ectotherms and increasing or decreasing the level of solar radiation absorption and/or body heat exchange with the environment, thus altering the internal temperature of the organism (Sherbrooke et al., 1994; Sherbrooke and Frost, 1989; Smith et al., 2016a). Although thermal melanism has been studied within and among species of ectothermic vertebrates, thermal melanism has been measured as overall darker or lighter skin colors, without inferring how the proportion of melanistic patterns may influence the rates of heating and cooling and the internal body temperature of the organism (though see Forsman, 1995 for snakes). Furthermore, to our knowledge, physiological color change in response to temperature changes influenced by the proportion of melanistic pattern (i.e. melanistic and non-melanistic coloration becoming lighter or darker at different rates) has not been examined.

The leopard gecko, Eublepharis macularius, a species with widespread melanistic pattern (Glimm et al., 2021), is characterized by melanistic spots in adults or stripes in juveniles on a lighter, often yellow, background of base dermal coloration. It is also commonly bred for pet-trade resulting in a variety of melanistic patterns of different types and quantity (Glimm et al., 2021; Kiskowski et al., 2019; Szydłowski et al., 2020) that is ideal for testing the influence of melanistic proportion on biological processes. To our knowledge, no study has assessed and quantified rapid physiological color change in this species. Eublepharis macularius occurs naturally in Pakistan, India, Iran, and surrounding regions (Agarwal et al., 2022) and is a crepuscular lizard primarily active in low light conditions. As such, we expect that heat absorption in this organism does not strongly depend on solar radiation – as it is active during times of low solar radiation – and that dorsal melanistic proportion would not affect heating rates in this species. On the other hand, the proportion of melanistic body coloration may have an influence on increasing cooling rates through increased rates of heat transfer during especially hot days (Geen and Johnston, 2014), while the organism is hiding. We expect E. macularius to be thigmothermic, i.e. to absorb heat by direct contact with warm substrates while hiding during the day or moving on surfaces that have absorbed heat during the day at night.

In this work, we used multispectral objective photography (Troscianko and Stevens, 2015) and a newly developed data extraction methodology to extract the melanistic coloration from the rest of the animal body coloration on live and free to move E. macularius. Although other algorithms and pipelines are available to extract color data, including segmenting color patterns from background colors (Abramjan et al., 2020; Glimm et al., 2021; Troscianko and Stevens, 2015), our approach allows the extraction of the same type of color data over time – i.e. every 30 min at which the images of the gecko were taken – on the same freely moving individual allowing coloration and color pattern to be compared over time. To test the hypothesis that melanistic coloration in animals active in low light conditions is not involved in thermoregulation, we used captive-bred individuals of E. macularius available to us and having different amounts of melanistic patterns (Glimm et al., 2021). Specifically, we tested if: 1) individuals with a greater proportion of stable dorsal melanistic pattern across the body experience increase heating or cooling rates than individuals with a lower proportion; 2) exposing individuals to lower suboptimal temperatures elicits a physiological color change in their skin; 3) individuals with a greater proportion of stable melanistic coloration experience a lower amount of overall physiological color change. The advantage of using captive bred individuals of E. macularius to test these hypotheses is that through captive breeding distinct individuals with different amounts of melanistic body coloration have been created that do not exist in nature. E. macularius is rapidly establishing as a model species to address different biological questions (Kiskowski et al., 2019; Kundey and Phillips, 2021; Pinto et al., 2023) including for studies on the function of coloration (Katlein et al., 2022; Szydłowski et al., 2020). As such, the goal of our study is not to investigate what would happen in nature to this species under the temperature conditions tested in this study, but to uncover the relationship between melanistic coloration and thermoregulation in reptiles active in dim light. Experimental results on this model are therefore expected to further our understanding on the role of melanistic coloration on thermoregulation versus other functions such as camouflage or communication in crepuscular and nocturnal species.

For each of the 12 tested geckos, 15 IR and 15 visible images were used for the analyses, with the exception of one IR image each for two geckos being of too low quality for data extraction, giving 178 IR data points and 180 visible data points (15×12 individuals). Similarly, 179 temperature measurements were used for each temperature type (IR images or average of the three dataloggers inside the terrarium) for the analyses as well. All temperatures taken from the experiments were evaluated against the geckos' native temperature range based on WorldClim data (Fig. 1). Experimental temperatures stayed within the quartiles of native temperature as planned by the experimental design, with the starting and ending temperature of 25°C corresponding to the median of native temperatures (Fig. 1). As a consequence, blocks 1 and 5 are closest in temperature to the interquartile range of the native temperature (Fig. 1), while block 3 had the greatest temperature difference from the median native temperature. Specifically, in block 3 all the experimental temperatures have a median temperature lower than the lower quartile of the native temperatures (warm side of the terrarium=17.25°C, cold side=16.5°C, hide spot=17°C, native lowest quartile=19.04°C; P<0.001 based on Wilcox tests for all data logger temperatures compared to native low quartile of native temperature), supporting that the temperatures we selected as suboptimal in this study are in fact suboptimal for this species in its native environment.

The geckos' body temperature was strongly correlated with the terrarium substrate temperature (r²= 0.97, P=0.33), both obtained from IR imaging. Correlation between the geckos' body temperature and the atmospheric or terrarium temperatures based on data loggers were also strong (r²= 0.92 or 0.93 depending on the comparison, P<0.001 for all correlations). Although, correlation between body temperature and each of the three dataloggers placed in the terrarium (cold side, warm, and hide) and between body temperature and atmospheric have identical r², the median atmospheric temperature is generally lower than the body temperature and the terrarium temperature based on data loggers (Fig. 1). Finally, correlation between the average temperature of all the three dataloggers within the terrarium and the substrate temperature estimated from the IR images were also high (r²= 0.85, P=0.40), although the temperature estimated by the IR images was relatively higher for each block than the one based on the datalogger (Fig. 1). We also found that relative humidity was not correlated with body temperature (r²=-0.23, P=0.82). To notice that while the median gecko temperature at the beginning and end of the experiment is lower than the median substrate temperature estimated by IR imaging, once the temperature of the experiment decreases (blocks 2-3), the median body temperature is higher than the median substrate temperature and closer to the lower quartile of the native temperatures (Fig. 1).

Significant differences in body temperature were detected during cooling (blocks 1 and 3, t-test t=41.2, P=2.08×10⁻¹³), heating (blocks 3 and 5, Wilcox test V=0, P= 0.0006), and between beginning and end of the experiment (blocks 1 and 5, Wilcox test V=78, P=0.0006) (Table 2). The heating and cooling rates were independent of body mass (r²=-0.003 P=0.35 for heating and r²=0.20 P=0.08 for cooling), SVL (r²=-0.09 P=0.78 for heating and r²=0.09 P=0.18 for cooling), or sex (r²=-0.04 P=0.45 for heating and r²=-0.06 P=0.57 for cooling) (Table 3).

Of the 12 geckos tested, nine had a melanistic proportion between 0% and 11% across the entire dorsal part of the body, while the other three had a melanistic proportion of 16%, 22%, and 41%. We found no influence of SVL (r²=-0.27, t=-0.70, P=0.50), sex (r²=-0.27, t=-0.73, P=0.49), or their interaction (t=0.71, P=0.49) on melanistic proportion (Table 3). Using the log of the variables to take into account any non-linear behavior, we also found that the melanistic proportion had no influence on heating rates (blocks 3 and 5, r²= −0.08, t=0.426, P=0.68) or cooling rates (blocks 1 and 3, r²=0.26, t=2.188, P=0.05) (Table 3, Fig. 2).

Using t-tests to evaluate differences in luminance between blocks, we found no significant differences between blocks 1 and 3 (cooling), blocks 3 and 5 (heating), and beginning and end of the experiment (blocks 1 and 5) for the melanistic and non-melanistic areas of the body (Table 2b and c for P-values). Tests on the influence of the proportion of stable melanistic coloration and log body temperature changes on changes in luminance for the melanistic and non-melanistic dorsal areas of the body indicate that during heating (blocks 3-5), the melanistic proportion has an influence on change in luminance for the non-melanistic area of the body (r²=0.47, t=-3.099, P=0.01, Table 4). Specifically, individuals with a greater melanistic proportion had higher luminance (luminance increases, thus the animal becomes lighter in color) in the non-melanistic areas of the body, while the melanistic part of the body did not experience any change in luminance (Figs 3A and 4A). Furthermore, we found an interaction between melanistic proportion and body temperature on changes in luminance during heating (blocks 3 and 5) for the non-melanistic areas of the body (r²=0.47, t=3.105, P=0.01) (Table 4, Figs 3A and 4A). Diagnostic tests revealed that one individual exerted significant leverage on the results of the model and influenced the significant results (Cook's distance >1). Melanistic proportion and body temperature had no influence on changes in luminance during cooling (blocks 1-3) or beginning and end of the experiment (blocks 1-5) for the non-melanistic area or the melanistic area of the body and for heating and the melanistic part of the body (Table 4).

Crepuscular and nocturnal reptiles experience low exposure to solar radiation during their active times and have been suggested to primarily rely on thigmothermy. Thigmothermy refers to the absorption of heat from the surrounding environment, rather than directly from solar radiation (Garrick, 2008; Kearney and Predavec, 2000). Despite relying mostly on substrate and surrounding temperatures, crepuscular and nocturnal reptiles can also bask for thermoregulation when needed (Angilletta et al., 1999).

In this study, we directly compare how substrate and atmospheric temperatures correlate with the internal body temperature in E. macularius and the role of melanistic pattern on physiological color change and thermoregulation in these animals. In our study, the main sources of heat during the cooling down of our experiments (blocks 1-3) were provided by the heat pad placed at one side of the terrarium and the lamp placed above the terrarium. We found a stronger correlation between body temperatures and terrarium substrate temperatures than for atmospheric temperatures, suggesting that E. macularius relies on absorbing heat from the ground and is thigmothermic. To notice that while the substrate temperature of the spot at which the gecko was located and the body temperatures were both estimated from IR photographs, the median atmospheric temperature – as estimated from data loggers – was generally lower than the gecko body temperature and the substrate temperature estimated from data loggers, suggesting that the higher correlation observed between substrate temperatures and body temperatures is not due to how the temperatures were measured. Our findings also support what has been previously suggested (Craioveanu et al., 2017; Garrick, 2008) based on measures of body surface temperature and ambient temperature differentials.

Although E. macularius mostly rely on the substrate temperature for thermoregulation, melanistic coloration – both as a stable coloration and as a physiological darkening of the skin – can be used for thermoregulation to increase or decrease the animal's body temperature during basking (Angilletta et al., 1999; Clusella Trullas et al., 2007; Smith et al., 2016a,b). Specifically, melanin can be used to absorb UV radiation and convert it into heat by photon-phonon transformation (McNamara et al., 2021) thus contributing to increased heating rates, but can also increase the rate of heat transfer for cooling in reptiles (Geen and Johnston, 2014). Compared to what we know on the role of melanistic coloration on diurnal ectotherms (Belliure and Carrascal, 2002; Garrick, 2008), the function(s) of melanistic coloration and especially of melanistic pattern in crepuscular and nocturnal reptiles are largely unknown. Our results based on 12 individuals of E. macularius with different proportions of melanistic pattern indicate that melanistic proportion is independent from body size (SVL), body mass, or sex, and that body size (SVL), body mass, or sex do not influence the heating or cooling rates. We found that the proportion of melanistic pattern does not influence cooling or heating rates in E. macularius. As E. macularius is crepuscular and only active for a few hours of daylight and mostly relies on thigmothermy, the characteristic spotted melanistic pattern of E. macularius in nature may have other functions than thermoregulation.

However, exposure to prolonged suboptimal low temperatures, as tested here, may trigger a physiological response to darken the skin color to increase heat absorbance. This phenomenon to date has only been observed in heliothermic reptiles – reptiles that regulate body temperature through solar radiation (Cowles, 1940). As E. macularius is crepuscular and shelters during daylight hours (Angilletta et al., 1999), physiological color change may be used more for cooling purposes, for example to increase heat transfer to the environment if shelters exceed optimal temperatures during daylight. In our study, we found that luminance (changes in luminance corresponds to physiological color change) does not change significantly between blocks during cooling temperatures, but it does during the heating phases of the experiment for the non-melanistic part of the body and depending on the melanistic proportion of the animals. We found that for the non-melanistic areas of the body, individuals with higher proportions of melanistic pattern experienced less darkening of the skin, and vice-versa. Physiological color changes during heating – along with other mechanisms such as changes in peripheral blood flow (Bartholomew et al., 1965; Rice and Bradshaw, 1980) – may provide protection against overheating, as the rate of heat gain has been observed to increase at higher environmental temperatures (Belliure and Carrascal, 2002). A caveat of these results is, however, that the significant relationship between melanistic proportion, heating, and physiological color change are strongly influenced by one individual with high melanistic proportion.

Taken together, our results suggest that in E. macularius melanistic pattern may not be used for thermoregulation, while physiological color change may occur to prevent overheating. Further research is needed to understand the role of melanistic coloration and melanistic pattern in geckos and the extent to which physiological color change occurs, especially in crepuscular and nocturnal species. In geckos, spotted patterns – such as the one observed in E. macularius – have been suggested to represent a more specialized type of camouflage, although bands and not spots have been proposed to be associated with nocturnal activity (Allen et al., 2020). Previous studies on crepuscular and nocturnal geckos have also indicated that physiological color change may be used more for background matching and camouflage than for thermoregulation (Vroonen et al., 2012; Zaidan and Wiebusch, 2007). The few studies investigating physiological color change for thermoregulation and camouflage in other nocturnal geckos propose that these functions are mutually exclusive phenomena (Vroonen et al., 2012; Zaidan and Wiebusch, 2007), but based on our results, this may depend on the occurrence and proportion of melanistic pattern. Future studies should therefore investigate the functional trade-off between melanistic coloration, including melanistic pattern and physiological color change, for thermoregulation versus its use in camouflage or signaling in crepuscular and nocturnal reptiles in captivity and in the wild. Finally, although our study is based on captive bred animals and wild E. macularius do not show the same extent of variation in melanistic proportion as the captive bred animals, our study is noteworthy mostly from a methodological and theoretical point of view. First, we developed a freely available software package that can be used to extract color pattern information from digital images of freely moving organisms with soft bodies, as in the case of geckos. Secondly, our results highlight the importance of melanistic pattern for studies on thermoregulation and coloration and suggest that melanistic pattern and melanistic coloration may be used for multiple non-exclusive functions. The conclusions of our study could help with further understanding the function of coloration and color pattern development in nocturnal and crepuscular reptiles, and how their function may differ from those of diurnal reptiles.

All capture, handling, and experimental protocols were approved by George Mason University IACUC committee (Permit number 1718778). Experiments were carried out to minimize stress and disturbance to the animals and in accordance with relevant guidelines and regulations.

All geckos were housed in the same room at George Mason University, with one gecko per terrarium (61×30×20 cm) and exposed to a 12-12 h day-night cycle. The housing containers for E. macularius were plastic boxes with newspaper bedding. The room temperature was kept stable between 25-28°C, with the average temperature generally being close to 25.3°C +/-0.2°C, and each terrarium contained moist and dry hides and a heat pad for thermoregulation. Humidity (41±10%) and temperature in the room were checked daily using a digital thermometer/hygrometer (ThermoPro TP50, Duluth, Georgia, USA). The health condition of each individual was checked daily by visual inspection, and no gecko was tested during the shedding process. Geckos were fed three times a week with a combination of crickets and mealworms dusted with calcium and vitamin powder. Feeding was withheld 3 days prior to testing, following the procedures of Angilletta et al. (1999). This was done to avoid interference in digestion due to exposure of animals to low temperature and to also avoid confounding effects due to potential food digestion. Feeding was resumed after testing. Drinking water was always available except during testing. Testing was carried out on 12 adult geckos (n=12, 7 males, 5 females). None of the females used in this experiment were gravid, and none of the tested individuals had been engaging in mating or any social interaction as each individual was housed and tested in isolation from the others.

Experiments were conducted in a temperature-controlled room of 4×2 m at George Mason University. This room has a 1×3 m open area in the middle, where the 51×25×30 cm glass terrarium used as a testing enclosure was placed on a rubber mat on the floor. The walls and floor of the terrarium were covered with white Teflon (SS Shovan, Shenzhen, Guangdong Province, PRC) to remove any potential effect of background coloration, and the lid was removed for the entirety of the experiment to facilitate obtaining the data. The use of Teflon prevents the geckos from trying to climb the wall (although leopard geckos do not climb, they may be on the two hindlimbs with the forelimbs on the wall of the terrarium), ensuring that the entire body of the gecko is always on the bottom of the terrarium. The Teflon on the bottom of the terrarium consisted of three layers, while the one on the side was a single layer. A 2 cm wide strip of black electrical tape (3 M, Hutchinson, Minnesota, USA) was used around the perimeter of the terrarium to adhere the bottom sheet of Teflon to the side sheets to prevent geckos from hiding underneath of the Teflon. The black electrical tape was used as the substrate temperature during calibration of the thermal camera measurements following guidelines on thermal camera calibration (F. Barroso, personal communication). A clean cardboard egg carton of 15×15 cm was placed in the middle of the terrarium as a hiding spot for the gecko, because reptiles behaviorally thermoregulate by seeking shelter in suboptimal temperatures (Aubret and Shine, 2010; Kearney et al., 2009; Woods et al., 2015). Except for the cardboard egg and three iButtons (DS1921G Thermochron, precision=0.5°C) to record the temperature (see below), nothing else was placed in the terrarium and the gecko was free to move and use the entire space available in the terrarium. A heating pad was placed underneath one side 15 cm from the end of the terrarium and set to 18°C (17.4°C±1.1°C) during the transitional and cold phase of the experiment (to have a warm spot throughout the experiments). A broad-spectrum UV-VIS light (Zoo Med PowerSun H. I. D Metal Halide UVB Lamp, 6500 K, 70 W, 95CRI) was placed 160 cm above the center of the testing terrarium to ensure photographs were taken under proper lighting (Troscianko and Stevens, 2015), as melanin strongly absorbs light in the UV-Vis spectrum (McNamara et al., 2021). This light was turned on 45 min prior to the start of each experiment and remained on for the length of the experiment (Fig. S1).

A thermometer/hygrometer (ThermoPro TP50 Deluth, Georgia, USA, precision=0.1°C, 1% humidity) was placed on the floor adjacent to the terrarium to measure the temperature and humidity of the room before and during the experiments. These temperature measurements were used to confirm proper operation of the cold room during the experiments but were excluded from the analyses. Four iButtons were placed in the terrarium and in the temperature-controlled room to monitor the temperature throughout the experiments and to ensure similar temperatures across experiments for the room and terrarium. Specifically, one iButton was placed on the floor outside of the middle section of the terrarium 20 cm away. Of the three iButtons placed in the terrarium, two were placed on the opposite ends (one on top of the heating pad and the second on the opposite end of the terrarium) and one under the cardboard egg carton. The iButtons were programmed to start collecting temperature data 30 min prior to the start of the experiment and continued to collect temperature data every 5 min for the entire duration of the experiment. Following each experiment, iButtons were sanitized with isopropyl alcohol and the terrarium was cleaned with soap and hot water to remove any potential scent or residue left from the previously tested individual. The top layer of the Teflon at the bottom of the terrarium was replaced after each experiment, while the Teflon sheets on the sides were sanitized with isopropyl alcohol after each experiment. A new cardboard egg carton was used for each tested individual.

While many studies have tested the thermal preference of E. macularius in laboratory conditions, none have tested native environmental temperatures in the field. Therefore, we estimated the native environmental temperatures for E. macularius and compared them with the temperatures tested in this study. Occurrences for this species were downloaded from the Global Biodiversity Information Facility (GBIF, October 2022; www.gbif.org) and imported into Rstudio (V4.1.2, R Core Team 2021) using the ‘occ_download_get’ function from the rgbif package (Chamberlain et al., 2022). Species occurrences for E. macularius were filtered for species' scientific name mismatches as well as NA values for latitude, longitude, species' scientific names, and country codes. Species occurrences were also cleaned and cross-checked for coordinate validity using the ‘clean_coordinates’ function from the package, Coordinate Cleaner (Zizka et al., 2019). Species occurrences which resulted in at least one flagged test labeled as, ‘FALSE,’ were removed from the dataset. Microclimate temperatures were extracted using the global model from the NicheMapR package (Kearney and Porter, 2017). Microclimate temperatures represent temperatures at 3 cm above ground with full sun (no shade) using 23 native coordinates for leopard geckos from GBIF. The model was run over 365 days for 10 years. Microclimatic temperatures for each day were taken as the average of each 60-min time interval across the day.

The range of experimental temperatures used in this experiment was 15-25°C (Table 1). The highest temperature was chosen as 25°C because this is the overall most frequent temperature at which these geckos are exposed in their housing environment. Additionally, this is within the preferred temperature range of E. macularius reported in previous studies, 25.8±1.0C (Werner et al., 2005), 28.2±0.6C (Angilletta et al., 1999). Differences could be due to seasonality. 15°C was used as the lower temperature to resemble natural low temperature experienced by this species in its natural habitat (Fig. 1), without eliciting hibernation (Khan, 2009), and because this temperature falls outside of the preferred temperature ranges as reported above. Fasted E. macularius also exhibited higher preferred temperature ranges than 15°C (Autumn and De Nardo, 1995). For the purposes of analysis, time periods in the experiment were separated into five ‘blocks’ to measure differences in variables between different phases of the experiment (Table 1). Briefly, blocks 1 and 5 correspond to the beginning and end of the temperature experiment, block 2 corresponds to the temperature going down, block 4 corresponds to the temperature going up back to 25°C, and block 3 corresponds to the lower temperature used in this study (Table 1). Experiments focused on studying the effects of lower temperatures as E. macularius is a crepuscular species that is less active during the warmer parts of the day.

To set up the maximum and minimum temperatures for this experiment, the temperature control for the experimental room was set at 25°C and then lowered by setting the room temperature control at 15°C when needed for the experiments (Table 1); temperature lowering between 25 and 15°C (or vice versa) took 1 h. To bring the temperature back to 25°C from 15°C a space heater was placed 1 m from the warm end of the terrarium and turned on after the temperature control of the room was adjusted to 25°C (block 4, Table 1). At the beginning of each experiment, we used the room thermometer (ThermoPro) to confirm that the room temperature was at 25°C. The temperature of the terrarium was confirmed for each end of it by pointing an infrared thermometer (Etekcity Corporation, Shenzhen, Guangdong Province, PRC) held 30 cm from the surface pointing perpendicular towards the bottom of the terrarium. These temperature measurements were used to confirm proper experimental setup but were excluded from the analyses. Temperature checks of the room and terrarium were repeated every 30 min during the experiment using the in-room thermometer and the infrared thermometer, respectively. Temperature readings from iButtons were used to confirm air temperature readings after each experiment.

Geckos were tested in a random order. Individuals were weighed to the nearest 0.01 g. using a digital scale before the start of the experiment and snout vent length (SVL) was measured to the nearest 1 mm. Only one gecko per day was tested for 7 h, starting at 11:00am each day. While time of day does influence thermoregulation (Angilletta et al., 1999; Sievert and Hutchison, 1988), lighting was necessary for proper photography of dorsal color patterns as this study focused on the use of melanistic pattern and color change during thermoregulation. Because of this, the experiments were carried out during the day as to not significantly interfere with the 12 h Light/Dark photoperiod in the housing environment. Although this species is crepuscular, experiments were carried out during the day to replicate the conditions of the housing room, where geckos are exposed to light conditions during the day. At the end of the experiment, each gecko was returned to its housing terrarium. Geckos were visually monitored after the experiments to check for any health concern, such as dermal irritation caused by overexposure to ultraviolet radiation or throwing up. No geckos had any issues during or after the experiments.

To extract temperature data from multiple body parts of the gecko, as different body parts may have different temperatures, a CAT S62 smartphone Pro camera (Caterpillar Inc., resolution=12MP, emissivity=0.95) was used to take an infrared (IR) image (Barroso et al., 2016). The camera was held approximately 30 cm directly above the individual in order to maintain the same effective pixel size (i.e. the actual area each pixel represents in the photographed subject) across IR images, regardless of the body size of the animal or its position in the terrarium. To standardize IR images and determine reflective temperatures, an 8×8 cm square piece of wrinkled aluminum foil was placed next to the gecko in the terrarium when capturing the image of a gecko each time an IR image was taken (Usamentiaga et al., 2014). The square piece of aluminum foil in each IR image was used to extract average reflective temperature from IR images only and was not used in color analysis. Average reflective temperature is required in each IR image to standardize temperatures for the gecko and the terrarium. After taking the IR image of the gecko, the relative humidity and temperature of the temperature-controlled room were recorded.

After taking the IR image of the gecko, to obtain the color data for each gecko, visible images were taken using a full spectrum converted Canon 1300D with a Kolari Vision UV/IR cut filter (410-700 nm transmission). Images were taken approximately 40 cm directly above the gecko to ensure a good resolution across images and with a grayscale standard built from Teflon following the methods of Abramjan et al. (2020) in the frame of the image. Visible images were obtained only for the dorsal part of the geckos, as melanistic patterns are generally absent from the ventral side of the animals and as such the ventral side was not relevant to the study questions (Glimm et al., 2021). Two people (A.M. and E.W.) took the IR images making sure to standardize the method between them and another person (B.H.) always took the visible images. If the gecko was under the cardboard hide, the hide was lifted prior to taking the images and replaced afterwards. The time at which the temperature of the room was changed, the thermostat/thermometer readings, as well as the times at which visible and thermal photos were taken were manually recorded.

Temperature measurements from the IR images were extracted in FLIR Tools (Teledyne FLIR 2022) from five body parts of each gecko (head – base of parietal scales, left knee, left foot, central dorsum, eyes – right and left, snout, and tail – above the cloaca), as body temperature varies across the body (Barroso et al., 2016). Temperature data averaged from both eyes were used as the internal body temperature for the analyses, as suggested for lizards by Barroso et al. (2016). Overall, the average of the temperature between two eyes and other body regions showed strong correlation, with a Spearman correlation coefficient above 0.98 (P<0.001, Table S1). Furthermore, the average temperature from both eyes was found to be highly correlated with the temperature of the snout (r_s =0.96), which has also been suggested to be a good proxy for internal body temperature (Tabh et al., 2021). For each IR image, relative humidity, atmospheric temperature, distance (0 m), and emissivity (1) were first entered into FLIR Thermal Studio following Barroso et al. (2016) and Barroso (personal communication) to calibrate the temperature readings of the thermal camera. Reflective temperature was obtained as the average reflective temperature of the aluminum foil standard. This value was extracted from the IR image of the aluminum foil by overlaying a box entirely over the aluminum foil using the ‘Rectangle’ function in FLIR Tools. After calibrating and entering the average reflective temperature in FLIR, distance and emissivity were re-entered as 0.3 m and 0.96, respectively (Barroso et al., 2016). The substrate temperature of the terrarium was also measured by using the ‘Rectangle’ function to overlay a small box over the black electrical tape at the bottom of the terrarium for each image taken, as the electrical tape more accurately reflects the temperature of the terrarium (F. Barroso, personal communication).

Digital images obtained with the Canon camera were processed using custom image processing software written in the Python programming language; color space conversions and luminance calculations were done using the OpenCV package in Python (Bradski, 2008). Luminance was used as a measure of color change as this is the most important color component that affects solar radiation absorption (Smith et al., 2016a). First, images were normalized for potential changes in lighting conditions across images of the same individual for the different blocks by converting each image to Hue-Lightness-Saturation (HLS) color space and using the grayscale reference in the first image of that individual as the baseline. The Lightness parameter (HLS) of following images were then standardized so that the greyscale reference matched that of the first image of that individual. This standardization step therefore allows comparisons among images taken for the same individual, but not across individual. All images were then converted back to the RGB color space. Next, because the limbs were sometimes obscured from the camera view due to the posture of the gecko in the image, images were cropped to only include the head, trunk, and tail of each individual. For image color segmentation – to segment the studied areas into color regions – hierarchical k-means clustering was run on each image (the specific parameters used can be found in the available codes hosted on GitHub), as manual object segmentation can be inaccurate and time consuming, especially when dealing with a large number of images. In this process every pixel is assumed to be a datapoint in the RGB color space, then pixels are grouped into a predefined number of clusters based on their distances to each other in the three-dimensional RGB color space. Because the result of this method is sensitive to its initialization, cluster centers were initiated following a K-means++ algorithm to account for this (Arthur and Vassilvitskii, 2007). While the K-means algorithm is computationally inexpensive and fast, there were some limitations caused by lighting conditions with a high incidence of shadows. It was common for coloration in areas of discoloration due to shadowing to be incorrectly assigned to a cluster, resulting in a misrepresentation of the pattern for that image. Because of the unsupervised nature of k-means, there is no way to correct for this error once the segmentation step is started. To account for this effect, a visual confirmation step by the user was implemented before segmentation to ensure that the color clusters would accurately represent the pattern to be segmented. For the present study, k=2 was used for clustering. While this is conservative, it avoided overestimation of the amount of melanistic coloration (Fig. S2).

The results of K-means clustering for visible images was a Boolean mask representing stable melanistic and non-melanistic coloration of the entire body for each image. Melanistic proportion was calculated as the area of stable melanistic coloration relative to the entire dorsal area (head, trunk, and tail), averaged across all images (15) of a single individual. The average standard deviation of melanistic proportion across all 12 geckos – calculated as average of the standard deviation for each gecko based on the 15 images taken for each individual's melanistic proportion – was +/-3.5%. Variation in melanistic proportion could be attributed to variation in lighting conditions, different positioning of the animal across the 15 images, imaging cropping, and the fact that K-means is an iterative, unsupervised algorithm (Lloyd, 1982). The Boolean mask was applied to the color corrected image to extract the mean luminance value for melanistic and non-melanistic coloration and the proportion of the coloration that was either melanistic or non-melanistic relative to the total coloration of the gecko. Luminance values were derived from a weighted calculation of RGB color channels (specific information are included in the data extraction pipeline on GitHub). All Python codes used to extract color data are available on GitHub. The full dataset is publicly available on Dryad.

Pearson's correlation coefficients evaluated the relationship between the gecko body temperature and the terrarium substrate temperature, both calculated from the same IR image, or between the gecko body temperature and the average of the three datalogger temperatures placed inside the terrarium, or between the gecko body temperature and the atmospheric temperature as estimated on the data logger outside the terrarium. To confirm the accuracy of IR substrate readings, a Pearson's correlation test was also run on the average data logger temperatures from within the terrarium and the substrate temperature taken with IR imaging. Because one temperature value from the data loggers had a reading of zero, that value was removed from the analyses. Previous studies have found that humidity may influence body temperature (Galliard et al., 2021); as such, we also tested the potential correlation between humidity and body temperature using a Pearson's correlation test.

To test for overall differences in body temperature throughout the experiment, we performed paired t-tests or Wilcoxon-tests (based on normality of the data) between the average body temperatures of each gecko in each block. The rate at which geckos heat and cool down may be influenced by body mass or by the animal size; as such, a linear model was fit between body mass and the log of heating and cooling rates separately and the analyses were then repeated for SVL. The log of the response variable was used to account for a potentially exponential behavior of heating and cooling rates (Winterton, 1999). Furthermore, as different sexes may respond differently to heating and cooling, we also used a linear model to test the influence of sex effects on the log of heating and cooling rates separately. Heating and cooling rates were calculated taking the change in body temperature between blocks 1 and 3 (cooling) and blocks 3 and 5 (heating).

To ensure that there was no correlation between the proportion of dorsal melanism (taken for each gecko as the average log of melanistic proportion over the 15 images) and SVL or sex, a linear model was fit between melanistic proportion and SVL with sex as a factor. Melanistic proportion was transformed to log of melanistic proportion to obtain normality of the data. To investigate the influence of melanistic proportion on heating and cooling rates, we ran a linear model based on data obtained on all 12 geckos between the log of the average melanistic proportion taken for each gecko and the log of heating/cooling rates as defined above.

Paired t-tests or Wilcoxon-tests (depending on normality of the data) were used to evaluate any changes in luminance values (physiological color change) between blocks. To assess the influence of body temperature variation and log of the average melanistic proportion on physiological color changes (luminance) on the entire dorsal area (head, trunk, and tail) of the gecko, we used linear models with the luminance change between two blocks as the dependent variable with body temperature change between two blocks or log of averaged melanistic proportion of each individual as the independent variable. We also ran more general models where these two independent variables (melanistic proportion and body temperature) were jointly tested with their interaction. As the results do not vary between the simpler and more complex model, we only report the results of the complex model to also assess the influence of the interaction between melanistic proportion and body temperature. Analyses were run independently for changes between blocks 1 and 3 (cooling), 3 and 5 (heating), and 1 and 5 (initial and final body temperatures). Luminance for melanistic and non-melanistic parts of the body of each gecko was tested in separate models. Any statistically significant linear model results were investigated further with diagnostic tests, specifically the Cook's distance to evaluate how much leverage each individual exerted on the model. All statistical analyses were run in R (V4.1.2, R Core Team 2021).

We are thankful to Emanuele Scanarini, Tony Gamble, Andran Abramjan, Alyssa Stark, Scott Glaberman, and Pat Gillevet for helping with the experimental setup. Andrea Weeks and Daniel Hanley provided helpful comments on an earlier version of this paper.