Butterflies are excellent study models for examining melanism and animal coloration because such features can be used as markers to understand which selective forces have been in action (Endler, 1982; Andersson, 1994). For instance, comparable color patterns that are seen in distinct species might result from convergent evolution driven by similar environmental pressures (Hoekstra, 2006). The color patterns of butterfly wings are formed through a set of physical mechanisms and pigments, such as melanin (Watt, 1964), and might play a role in mate recognition (Wiernasz, 1995), thermoregulation (Clusella-Trullas et al., 2007), protection against predators (Kettlewell, 1955) and parasites (True, 2003), as well as in desiccation and UV damage prevention (Majerus, 1998).

Organisms are expected to have darker colors in environments with lower temperatures because the concentration of melanin in the cuticle increases the absorption of thermal energy (Clusella-Trullas et al., 2007). As ectothermic and diurnal individuals, butterflies need heat to carry out their activities, such as foraging and looking for mates (Brakefield and Willmer, 1985; Wiernasz, 1995). So, the presence of darker patches might favor butterflies by quickly increasing their body temperature to carry out their activities (Kingsolver, 1983). Dark melanic butterflies also seem to be associated with humid environments (Stelbrink et al., 2019) as a form of protection against parasites. Alternatively, in warmer, highly sun-exposed, environments, insects are expected to be brighter to avoid overheating (Schmitz, 1994; Heidrich et al., 2018). However, melanism may not always be expressed homogeneously throughout the animal's body, as is the case for melanic differences found between dorsal and ventral surfaces of butterfly wings, a thermoregulatory strategy (Zeuss et al., 2014; Stelbrink et al., 2019) that captures more thermal energy when exposing dark melanic dorsal surfaces and dissipates more heat when exposing less pigmented ventral surfaces (Tsai et al., 2020). In addition to affecting lightness, melanin might also influence other spectral parameters of animal coloration, such as saturation (i.e., the purity of the color of an area judged in proportion to its brightness). For instance, melanic colors mostly result in patterns with lower saturation (i.e., less vivid colors), such as black, gray, and brown (McGraw, 2006), although the opposite relation has also been shown to be true for monarch butterflies (Hanley et al., 2013).

Whereas, most studies of lepidopteran coloration assemblages have addressed species and environmental conditions from the Palearctic (Zeuss et al., 2014; Heidrich et al., 2018; Stelbrink et al., 2019), Nearctic (Zeuss et al., 2014; Stelbrink et al., 2019), or Australian (Xing et al., 2016, 2018; Dalrymple et al., 2018) regions, the Neotropical species have been considerably understudied. In a country of continental dimensions, which holds an astonishing biodiversity and has seven different morphoclimatic and phytogeographic domains (Ab'Saber, 1967), it is surprising that only one study, conducted with Amazonian species, aimed to investigate lepidopteran color assemblages in Brazil (Spaniol et al., 2019).

Considering that environmental conditions can drive behaviors and select different coloration patterns (Endler, 1992) and that the spectral composition of sunlight can be modified by the physical characteristics of the environment (Endler, 1993), the present study aimed to compare the visual patterns displayed by assemblages of butterflies from two distinct Brazilian biomes, namely, the tropical rainforest and the xeric white forest. We expect butterfly assemblages from each biome to differ regarding their melanism and/or color saturation; our hypotheses are as follows: (i) butterflies that occur in the rainforest should be more melanic than those occurring exclusively in the white forest since in warmer, more sun-exposed environments, such as the xeric white forest, insects are expected to be brighter to avoid overheating (Schmitz, 1994; Heidrich et al., 2018), while in moist environments, such as the tropical rainforest, butterflies should tend to be darker (Stelbrink et al., 2019) as a form of protection against parasites; (ii) also due to thermoregulatory advantages (Zeuss et al., 2014; Stelbrink et al., 2019), differences in melanism between dorsal and ventral wing surfaces should be more pronounced in more heat-stressful environments (Tsai et al., 2020), such as the xeric white forest; finally, (iii) in highly shaded environments, such as rainforests, in opposition to considerably leafless ecosystems, such as the xeric white forest, individuals should have more saturated colors to enhance their contrast against the green dappled foliage background and compensate for the greenish illuminant created by sunlight filtering through the treetops (Endler, 1993).

This study was approved by the UFRN's Animal Ethics Committee (CEUA, protocol number 013/2020) and followed all legal requirements of Brazilian law (number 11.794/2008).

The comparison of butterfly assemblages from each biome did show differences regarding the melanism of their wings, confirming our first hypothesis and previous evidence showing that the lightness of butterflies' wings contrasts between different ecosystems, a trait that has been linked to thermoregulation (Tsai et al., 2020). Our expectation was that tropical rainforest butterflies should be more melanic (i.e., to present lower mean gray values) to counteract a higher parasitic rate found in humid environments (Stelbrink et al., 2019). Studies have shown that there is a pleiotropic effect between melanocortins and the immune function of animals (Ducrest et al., 2008) and that more melanic animals could defend themselves against parasites by strengthening their immunocompetence or directly preventing the proliferation of parasites due to the toxicity of melanin molecules (Côte et al., 2018). In addition, we also expected that white forest butterflies should be less melanic to avoid overheating (Schmitz, 1994; Forsman et al., 2002; Heidrich et al., 2018) since the xeric white forest is notoriously known for its high solar radiation, high average annual temperature, low cloud cover, and low relative humidity rates (Prado, 2003).

Our data also revealed that butterflies from the rainforest have wings that are significantly more melanic on their dorsal faces when compared to their own ventral wing surfaces, but the same was not true for white forest assemblies, which is the opposite of what we would expect according to our second hypothesis. Differences in melanism between dorsal and ventral wing surfaces have been associated with thermoregulatory advantages (Tsai et al., 2020), in such a way that heat-stressful environments, such as the xeric white forest, should have selected for more prominent differences, although it has also been observed that butterflies from regions with less sun exposure tend to be darker (Dalrymple et al., 2018). Here it is worth noting that the studies that proposed the relationship between melanism and thermoregulatory strategies were based on assemblages of butterflies from colder regions (Zeuss et al., 2014; Stelbrink et al., 2019; Tsai et al., 2020), a rule that might not fully apply to tropical environments. Having a darker dorsal surface and a lighter ventral area might also be a thermoregulatory adaptation related to more heterogenous biomes, which consist of different phytophysiognomies, as in the case of the tropical rainforest. The tropical moist broadleaf forests biome (Olson et al., 2001) presents two distinct phytophysiognomies that were considered in this study, the Coastal Atlantic Forest, a true tropical rainforest, with tall trees that form high and closed canopies, and the Atlantic Coast Restinga, a type of coastal tropical forest, characterized by medium-sized trees and shrubs adapted to drier conditions. Those tropical rainforest butterflies that frequently transition from a dryer and open phytophysiognomy (e.g., the Atlantic Coast Restinga) to another that is much more shaded and humid (e.g., Coastal Atlantic Forest) should benefit from the thermoregulatory advantages (Tsai et al., 2020) of having a darker dorsum and a lighter ventral surface. If these interpretations are correct, it would be expected that butterflies inhabiting both biomes (i.e., rainforest and white forest) would also show differences in melanism between their ventral and dorsal wing surfaces. Indeed, a posteriori analysis, following the same method described in this study, but using 75 butterfly species that occur in both biomes (Supplementary Table S2; the statistical analysis used the taxonomy as a proxy for the phylogenetic dependence), showed a significant difference in melanism between their dorsal and ventral wing surfaces (9.89 [4.37, 19.4]), with a darker dorsum and a lighter ventral surface (Figure 2 and Supplementary Figure S2). In contrast, for a considerably dryer, open, and leafless biome, like what is found in the Xeric Shrublands (Olson et al., 2001), known in Brazil as the Caatinga or white forest, a tendency to show a similar degree of melanism on both sides of their wings (Figure 1) could be understood as a strategy to avoid desiccation and UV damage (Majerus, 1998).

An alternative explanation for the difference in melanism between the dorsal and the ventral faces of rainforest butterflies' wings would be that it works as an anti-predatory strategy (Murali, 2018), making it difficult for predators to capture them (Pinheiro et al., 2016). When butterflies present a different pattern of melanism between ventral and dorsal areas they can produce an effect called dynamic flash when flapping their wings during flight, which would help to confuse predators about their exact location (Murali and Kodandaramaiah, 2020). Although literature still lacks studies comparing the predation pressure of butterflies in different tropical environments, we believe is reasonable to guess that the abundance of butterfly predators is greater in the rainforest than in the white forest, due to the greater general diversity of species found in the rainforest, which would be in line with our proposition.

Even though the latitude in which our specimens were captured did not vary much (Supplementary Tables S1, S2), and despite the fact that the solar incidence of different regions of the Brazilian northeast is comparably high (Lopo et al., 2014), differences in solar incidence between localities with diverse kinds of phytophysiognomies were expected (although the solar incidence was not directly accessed in the present study), since vegetation cover reduces the intensity of sunlight and considerably decreases the proportion of short wavelength light from the illuminant (Endler, 1993). However, our third hypothesis, stating that shaded environments (i.e., rainforest) should have butterflies with more saturated colors, was not corroborated.

Since the illuminant spectrum also influences the production of animal coloration (Endler, 1993), changes in the geometry of canopies, resulting from deforestation, could also influence the dynamics and spatial distribution of organisms of different colors (Spaniol et al., 2019). Spaniol et al. (2020) showed that degraded environments might be experiencing a loss in the diversity of butterflies' color patterns since in degraded and more exposed environments butterflies with more conspicuous colors may have been reduced by predator activity. In line with this, we have found that butterflies' dorsal wing sides present a higher variance between species (regardless of the environment) when compared to their ventral wing side. On the one hand, a lower color variance on ventral wing surfaces might be seen as an indication that some ecological factor (e.g., predation avoidance) is exerting a strong homogenizing force on ventral wing coloration. On the other hand, for dorsal wing surfaces, a greater variety of active selective forces (e.g., thermoregulation and socio-sexual signaling) could explain the existence of a higher color variance.

To our knowledge, this is the first study to compare the coloration of butterfly assemblages in humid and semi-arid tropical environments. Due to its occurrence in a variety of ecosystems, future research with Neotropical butterflies must be encouraged and should incorporate data on communities from savannahs, prairies, and floodlands, in addition to comparing butterfly populations from a specific biome (e.g., rainforests from the Northeast and the Southeast regions of Brazil), but that occur in very different latitudes (Zeuss et al., 2014). It would also be interesting to investigate the relationship that other body regions, such as butterflies' thorax and the basal region of their wings (Munro et al., 2019), have with melanism and color saturation, as well as how Neotropical butterflies reflect infrared radiation (see Kang et al., 2021). Furthermore, since most studies that assess macroecological patterns focus on the consequences of population dynamics and changes in coloration patterns (Zeuss et al., 2014; Pinkert et al., 2017; Xing et al., 2018), experiments that relate environmental degradation to the degree of melanism in butterfly populations and communities should also be pursued. Future studies could also include visual modeling techniques to assess how changes in the sensory environment of organisms (e.g., changes in vegetation structure and environmental light) may affect the trade-off between animal communication, predation, and thermoregulation.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

SS: conceptualization, methodology, software, formal analysis, investigation, data curation, writing—original draft, and visualization. FG: formal analysis and writing—review and editing. MC: conceptualization, resources, and writing—review and editing. DP: conceptualization, methodology, resources, writing—review and editing, visualization, supervision, project administration, and funding acquisition. All authors contributed to the article and approved the submitted version.