Pelagic fish camouflage in shallow waters from the humboldt current system through intracellular structures and reflectance mechanisms

Abstract

Pelagic fish have evolved specialized biogenic multilayer reflectors composed of stacks of intracellular anhydrous guanine crystals separated by cytoplasm, giving notorious silvery appearance to their skin. While the reflective properties of guanine crystals and their utility for fish camouflage have been shown in other fish species, this is the first evaluation on fish species from the southern hemisphere, and from the Humboldt current system. This is one of the most productive systems on earth, having particular oceanographic conditions such as upwelling, and thus under strong selection pressures. In this study, we conducted a comparative analysis of four pelagic species, Sardine, Anchovy, and Snoek, known for their silvery characteristics, and Mote sculpin, which lacks silvery features. We aimed to explore the biological mechanisms underlying light reflectivity in fish species and to understand how fish skin microstructures affect whole fish light reflectance and intensity in the visible spectrum. We measured the reflectance of individual fish using hyperspectral imaging and characterized the guanine crystal/cytoplasm layers within the skin of each fish using high-resolution scanning electron microscopy. These Scanning Electron Microscopy (SEM) images were analyzed using the 2D discrete Fourier transform to extract the spatial patterns that govern the light interaction with the guanine crystal structures. A novel spatial frequency analysis approach applied to SEM images explained reflectance differences between species with similar spectral behavior. Furthermore, this study presents the first fish classifiers based on the analysis of spatial frequency features, achieving up to 92.14% accuracy using a K-Nearest Neighbors classifier, highlighting the functional and taxonomic relevance of guanine microstructure organization. Our findings confirm, on pelagic fish species from the Humboldt current system, that silvery species have a chaotic distribution/arrangement of guanine crystals, whereas non-silvery species have a more organized arrangement. Accordingly, Fourier analysis indicated that silvery fish are capable of scattering light uniformly across the visible spectrum. In contrast, the Mote sculpin shows a stronger scattering of red light, distinguishing it from silvery fish.

Introduction

Camouflage is a crucial strategy in nature, and under high evolutionary pressure to thwart predators. Most animals rely on their acute color perception for hunting, highlighting the significance of camouflage patterns that incorporate color synchronization alongside features such as shape, contrast, and texture¹. Understanding the effectiveness of camouflage is a multifaceted endeavor, requiring a nuanced comprehension of biological mechanisms and advancements in optical imaging technology. Within Earth’s diverse environments, the open ocean emerges as a particularly challenging arena for camouflaging due to the unique constraint of lacking a background. The absence of objects for concealment means that creatures employing camouflage techniques must instead blend seamlessly into a perpetually shifting backdrop. Visual backgrounds in the open ocean stem from light interactions with water molecules, particles and organisms such as phytoplankton. Early characterizations indicated that this backdrop was subject to temporal fluctuations owing to changes in water composition^2,3. It exhibit spatial uniformity, with symmetrical properties along the horizontal plane and a predictable intensity gradient in the vertical direction⁴. Within the sunlit expanse of the open ocean, finding a hiding spot becomes an actual challenge and pelagic creatures have blended seamlessly into their surroundings, evolving a “silvery” mirror like camouflage⁵.

Fish with shimmering, mirror-like skin are masters of camouflage. Reflecting sunlight from above, they seamlessly blend with the surrounding light, becoming nearly invisible to predators⁶. Early theories suggested the silvery sides of open-ocean fish evolved to mimic the constant light from above. By reflecting or emitting light, fish reduce shadows on their bodies. This adaptation, called counter-illumination, helps them blend in with the surrounding light, making them harder to see against the open water. Iridescent scales and skin with their broad reflective properties, are believed to play a key role in this camouflage technique^7,8. Reflected light from fish, however, displays a degree of linear polarization differing to that of underwater light, which might be a disadvantage. To overcome this, fish developed a ”multilayer interference”, where alternating layers of materials with high and low refractive indexes within fish act as a perfect mirror^7,9. A prime example of this architecture involves guanine crystals intricately arranged alongside cytoplasmic layers. Collectively, these elements combine to construct intricate multilayer arrays, together conferring the whole fish light reflection properties^10,11. Guanine is the most prevalent purine employed in biological structural color systems. Within fish, this reflective quality emerges through a sequence of alternating layers: slim plate-shaped guanine crystals boasting high refractive indices (with n = 1.83 along the axis perpendicular to the biogenic plate axis) and low refractive index cytoplasm, primarily governed by that of water (n = 1.33). In fish, guanine crystals are on the scales and intracellularly in specialized cells known as iridophores¹¹. These iridophores are situated beneath the scales and in the stratum argentum, a subdermal skin layer. Then, each crystal forms within a crystal chamber constructed from a lipid bilayer. Recent investigations have demonstrated that these crystals originate from an amorphous precursor phase of guanine. In optical terms, the prevailing consensus has been the arrangement of these crystals in quarter-wavelength stacks, often referred to as ’ideal’ stacks. This indicates that the optical thickness of both the crystals and the cytoplasmic region is similar, with the maximum reflected wavelength being 4 times the thickness¹². The fundamental reflective unit was characterized as a collection of 4–5 crystals, separated by cytoplasm¹¹. Denton and Nicol¹³ in 1966 inferred that the reflective properties of stacks within a given fish are not uniform, allowing the fish skin to reflect light across a broad spectrum of incident angles and wavelengths. However, the notion of an ideal reflector within fish skin was challenged by McKenzie et al.¹⁴, who proposed that the crystals are positioned randomly. Their work, using transmission electron microscopy (TEM), revealed that the crystals within fish skin are aligned in parallel orientations but lack any systematic arrangement in stacks. The foundational concept of utilizing multilayer reflectors to generate structural colors is firmly established within biological systems, with a specific focus on fish skin color. This principle while somehow well understood in a handful of fish species, have never been explored in fish from the southern hemisphere, precluding a more general and in-deep understanding of this adaption among other fish species. Furthermore, different oceanographic conditions (upwelling), and strong predation pressures in the highly productive Humboldt Current System (HCS), are distinctive enough to assess the extent of the silvery camouflage on key HCS species.

This study also introduces a novel approach to examining camouflage by linking the internal structure of multilayer reflectors (i.e., guanine crystals) with the optical reflectance of fish skin. Using the continuous-space Fourier transform (CSFT), we analyze the spatial frequencies of guanine crystal arrangements and correlate them with electromagnetic reflectance in the visible-to-near-infrared (VIS–NIR) spectrum. To our knowledge, this is the first study to classify pelagic fish species using CSFT-based structural features within a supervised machine learning framework K-Nearest Neighbors (KNN). These methods provide new insight into how optical structures may have been selectively tuned by natural predation pressures to optimize camouflage.

Through spatial frequency and electromagnetic reflectance analysis, this research aimed to explore the mechanisms regulating reflectivity in pelagic fish species off the coast of Chile, on the HCS, including both silvery and non-silvery species. We hypothesized that differences in the density and arrangement of multilayer reflectors, such as guanine crystals and cytoplasm, within the skin of Chilean pelagic fish determine their silvery appearance and reflectance properties, which is extremely important in camouflage in open-ocean. Specifically, we hypothesize silvery species will show a chaotic or random arrangement of guanine crystals within the cytoplasm, conferring a flat reflectance in the VIS spectrum (whole fish), while the non-silvery species will show a more regular arrangement of guanine crystals within the cytoplasm, leading to an enhanced reflection at some wavelengths. The silvery species selected for this study were: the Sardine (Strangomera bentincki), Anchovy (Engraulis ringens), and Snoek (Thyrsites atun), and the non-silvery species was the Mote sculpin (Normanichthys crockeri) (Fig. 1). From these, sardines, anchovies and mote are preys of the snoek, all sharing the same Humboldt current system habitat. Sardines and anchovies are heavily targeted by fisheries in Chile, play a crucial role in the food webs of the Humboldt Current System (HCS), sustaining one of the biggest fisheries on earth. These fish inhabit open water environments and exhibit a metallic sheen and pigmented hues on their skin. In contrast, the Mote sculpin is often caught as bycatch in both industrial and artisanal fisheries and lacks the characteristic ”silvery” light reflectance. Thus, all the fish species studied here are part of the same fish assemblage and inhabits similar water column conditions. To date, there is no information on the optical properties of any Chilean species, nor on the mechanisms conferring their silvery features on fish from the HCS. More importantly, this is the first assessment of the silvery camouflage on fish species under a strong predatory pressure, and thus likely under a stronger selection for optimum camouflage. Understanding these differences could provide valuable insights into the biological mechanisms of camouflage and the ecological behaviors of these species, promoting the conservation and maintenance of important aquatic resources.

Fig. 1

Selected species for this study: (a) Sardine (Strangomera bentincki); (b) Anchovy (Engraulis ringens); (c) Snoek (Thyrsites atun); (d) Mote sculpin (Normanichthys crockeri).

Results

This investigation is the first to evaluate the reflective properties of silvery Chilean pelagic fish. It accomplishes this by employing a semi-quantitative analysis that integrates hyperspectral imaging with the microscopic examination of guanine crystal structures, utilizing both morphological and spatial frequency analysis.

Spectral analysis

Fig. 2

Average spectral curves of the selected species in this study: (a) Sardine (Strangomera bentincki); (b) Anchovy (Engraulis ringens); (c) Snoek (Thyrsites atun); (d) Mote sculpin (Normanichthys crockeri).

Figure 2a–d show the characteristic spectral curves and their associated standard deviations for the species Sardine, Anchovy, Snoek, and Mote sculpin, respectively. Due to the molecular properties of seawater, infrared (IR) radiation (wavelengths longer than 750 nm) is highly absorbed¹⁵; as a result, IR light penetrates only a few centimeters into seawater. In contrast, visible light (wavelengths within the range 400-750 nm) penetrates deeper (around 200 meters), making it the range of interest for our analysis. As a consequence, we focus our discussion to the range of wavelengths below the dashed line shown in the Fig. 2.

All spectra exhibit a valley in the 400-440 nm wavelength range. This behavior could be due to the absorption of light by melanophores, which contain melanin, a dark pigment that strongly absorbs in this wavelength range, contributing to the darker tones in certain areas of the fish’s body¹⁶. In the case of Sardine, Anchovy, and Snoek, the spectra show a relatively flat behavior in the 450-750 nm range. This suggests a significant contribution of xanthophores and iridophores in light reflection, as carotenoids in xanthophores absorb light in the 450-550 nm region, and iridophores scatter light, producing a constant brightness across this visible spectral region¹⁷. The average reflectance values of Sardine, Anchovy, and Snoek are $0.27\pm 0.05$, $0.19\pm 0.04$, and $0.29\pm 0.05$, respectively. It is observed that the three species exhibit a flat reflectance behavior in the visible range, with sardine and anchovy showing a notably similar flattened shape across the VIS-NIR spectrum. However, the overall reflectance of the anchovy is, on average, 30% lower than that of the sardine. This difference can be explained by the arrangement of guanine crystals, as discussed in the spatial frequency analysis in Sect. 2.3. Since the crystal arrangement in anchovy is more chaotic than in sardine, the latter exhibits higher reflectance due to reduced destructive interference.

The flat spectral behavior observed in these three species is further evidenced by modeling their spectral signatures using zero- and first-order polynomial fits as described in Sect. 4.1. Thus, the overall reflectance (zero-order approximation parameter) and the slope (first-order approximation parameter) were analyzed. Figure 3a,b presents box plots with individual data points and the statistical outcomes of a one way ANOVA analysis comparing the overall reflectance and slope across the four fish species, respectively. For overall reflectance, the F-value is 67.20 with a p-value of $5.69\times 10^{-24}$ ($df = 4$, $\hat{\omega ^2_P} = 0.65$), indicating a highly significant difference among the species. Post-hoc tests show significant differences: Anchovy vs. Mote Sculpin ($p = 0.086$), Anchovy vs. Sardine ($p = 0.036$), Anchovy vs. Snoek ($p = 5.30\times 10^{-10}$), Mote Sculpin vs. Sardine ($p = 0.016$), Mote Sculpin vs. Snoek ($p = 2.00\times 10^{-4}$), and Sardine vs. Snoek ($p = 0.014$). For slope, the F-value is 84.27 with a p-value of $2.20\times 10^{-27}$ ($df = 4$, $\hat{\omega ^2_P} = 0.70$), also indicating a highly significant difference. Post-hoc tests reveal significant differences: Anchovy vs. Mote Sculpin ($p = 4.39\times 10^{-21}$), Anchovy vs. Sardine ($p = 1.84\times 10^{-4}$), Anchovy vs. Snoek ($p = 1.82\times 10^{-27}$), Mote Sculpin vs. Sardine ($p = 0.004$), Mote Sculpin vs. Snoek ($p = 6.6\times 10^{-4}$), and Sardine vs. Snoek ($p = 0.027$). These results suggest significant variations in both overall reflectance and slope among the fish species, with Mote consistently showing distinct values compared to the others. The ”Silvery” group (Sardine, Anchovy, Snoek) would be expected to have high Overall Reflectance values. The figure shows Anchovy, Sardine, and Snoek generally having higher reflectance than Mote Sculpin, supporting their classification as silvery. The differences among them might relate to subtle variations in their specific structural characteristics. For example, slight variations in the density, orientation, or types of guanine crystals could lead to measurable differences in overall reflectance. Regarding the Slope, p-values of $p<0.05$ were observed between Anchovy vs. Sardine and Anchovy vs. Snoek. However, these values are relatively high compared to the p-values between Mote and any of the silvery species. The slope values for the three silvery species are close to zero, which supports the characterization of these species as having a flat reflectance spectrum. In contrast, the Mote sculpin shows an increasing reflectance from 420 nm onward, starting at a value of 0.08 and reaching 0.19 at 720 nm. This increase in reflectance at longer wavelengths may be attributed to the presence of guanine in leucophores and purines in iridophores, which enhance light scattering in the red region, suggesting that the Mote sculpin reflects more light at wavelengths corresponding to this color^16,17, as shown in Fig. 1d.

Fig. 3

Descriptive characteristics of the spectral behavior of the studied species: (a) Features extracted from the zero-order approximation; (b) Features extracted from the first-order approximation.

Morphological Guanine analysis

Table 1 Summary of thicknesses of guanine crystals and the multilayer stacks measured from SEM of the analyzed fish. The mean value and the standard deviation from the measurements are reported for each fish sample.

Fig. 4

Characteristics of the guanine crystal arrangement among the four studied species. (a) Cytoplasm thickness. (b) Multilayer thickness.

Fig. 5

SEM images of a dissected area of the fish skin. (a, b) Sardine, showing partially chaotic crystals. (c, d) Anchovy showing more chaotic crystal stacks that often interdigitate. (e, f) Snoek, showing even more chaotic crystals. (g, h) Mote sculpin presenting an organized crystal stacks.

It was observed in Scanning Electron Microscopy (SEM) images of the skin of Sardines, Anchovies, and Snoeks, that the guanine crystal multilayers display a disordered/chaotic arrangement, appearing as thick plates with crystal thicknesses ranging from 100 to 150 $\mu m$ ($\pm 46$). Upon analyzing the measurements in Table 1, it is notable that there was no significant difference in crystal thickness between Sardine and Anchovy species. In the case of the Snoek species, the cytoplasm thickness ranges from 0.40 to 0.80 $\mu m$, which is close to the values observed in Sardine and Anchovy. However, its multilayer thickness is greater than that of all the other species, at approximately 51 $\mu m$. Analyzing the measures of the crystals of the Mote sculpin images, the results diverge. The multilayers display a more ordered arrangement and are notably thicker, ranging from 75 to 90 $\mu m$ ($\pm 53$). The same trend is can be observed in the average cytoplasm thicknesses, where Sardines and Anchovies exhibit values ranging from 1 to 2 micrometers while Mote sculpin showcases notably thicker values ranging from 3 to 4 micrometers. Consequently, a significant difference exists between these two groups. The difference in cytoplasm thickness of the analized species can be observed in the violin plot depicted in Fig. 4a.

However, when we examine the thickness of the multilayers (i.e., all the grouped crystal layers), we notice a distinct trend among the groups. As can be seen in Fig. 4b, it is highlighted that the range of multilayer thicknesses in Anchovies is significantly larger compared to that of Sardines and Mote sculpin. Considering that there is not absolute control over the orientation of the fracture plane observed in the SEM images, it is possible that the measured crystal and cytoplasm thicknesses may be slightly affected by an oblique fracture. Nevertheless, such influence, if any, is relatively small and falls within the standard deviation of the thickness measurements. Analysis of the SEM images reveals several significant differences in the crystal organization within the skins of Sardines, Snoek, and Anchovies compared to those of Mote sculpin. A sample of SEM images is presented in Fig. 5; a remarkable difference is the orientation of the crystal arrangements relative to the scale surface and, consequently, to the light source. See Supplementary Fig. S1 for overall crystal distribution.

In the case of Mote sculpin, the majority of guanine crystals exhibit a higher degree of organization as shown in Fig. 5g,h when compared withe the SEM images of the other species. Conversely, in Sardines and Anchovies, the crystals tend to be more disordered (see Fig. 5a,b and c,d, respectively). Furthermore, Snoek also differs with respect to the other silvery species, showing a much larger number of alternating crystal-cytoplasm layers in the skin, as shown in Fig. 5e,f. Inside there are multiple crystal stacks leading to interdigitation between the crystals of neighboring stacks. The interdigitation has several consequences: locally reducing the average thickness of the cytoplasm spacing between the crystals (measuring around 60 $\mu m$), increasing the spacing between multilayers (measuring around 140 $\mu m$) and introducing more disorder by the variation in cytoplasm spacing and in the angles between adjacent crystals, factors which may have a profound effect on the intensity of reflectance, as discussed below.

Guanine crystal structure as a diffraction grating

It was observed from analyzing the magnitude spectrum of 2D discrete Fourier transform (DFT) of the SEM images that the guanine crystal structures mimic a diffraction grating, which may explain the fish interaction with light and color.

The computation of the 2D-DFT of the SEM images enables the identification of the dominant spatial frequency components and the computation of the spatial wavenumber, k that is directly related to the spacing between the guanine crystal structures. The wavenumber of the spatial crystal structures can be correlated with the spacing of a diffraction grating, d¹⁸ as we discuss here. In this work, we utilize k to refer to the spatial wavenumber, which is a measure of how often sinusoidal components (as determined by the 2D-DFT) of the spatial structure repeat per unit of distance, i.e., $k = 1000/d[cycles/1000nm]$, where d is the distance between the divisions of the grid (i.e., the spacing between the guanine crystal structures). This differs from its usage in electromagnetism, where the wavenumber is related to the wavefront’s propagation velocity and angular frequency.

Fig. 6

Sample of characteristic spatial frequency components from the guanine crystal structures of the studied species. (a) Sardine, (b) Anchovy, (c) Snoek and (d) Mote sculpin. On the left is a SEM image, in the center its Fourier magnitude spectrum, and on the right the highest magnitude frequency components present in the spatial frequency ring of the visible range.

Figure 6a–d presents the spatial frequency analysis of the SEM images of Sardine, Anchovy, Snoek, and Mote species, respectively. The left column display a SEM image of each species; the center column the corresponding DFT magnitude spectrum; and the right column displays the highest magnitude frequency components present in the spatial frequency ring of the visible range. The DFT magnitude spectrum for all these images where calculated by first rotating the input SEM image so that the longest structure appears vertical, such that for ordered structures, which present a pattern of vertical lines, the horizontal component contains most of the spatial structural information of the crystals.

From the Fourier magnitude spectrum of each image, the components with the highest magnitudes were selected within the wavenumber range of 1.33 to 2.63 [cycles/1000 nm], corresponding to the spatial frequency domain associated with visible light. For each image, the top 10% of the highest magnitude components within a spatial frequency ring—defined by an inner radius of 1.33 and an outer radius of 2.63 [cycles/1000 nm] were considered. Each spectral component is characterized by a vertical and a horizontal frequency, whose angle defines the dominant spatial orientation of the underlying structures.

For each image, three features were extracted to characterize the spatial organization of guanine crystal structures and their potential contribution to optical reflectance. The average vertical and horizontal spatial frequencies from high-energy spectral components describe the predominant orientation and periodicity of the platelets. Additionally, the skewness of the angular distribution of frequency vectors quantifies the degree of anisotropy in crystal alignment. Higher skewness values indicate more ordered structures, enhancing Bragg reflector efficiency and contributing to a more intense silver-like appearance.

In general, the angle values of the frequency components ranged from 0 to $\pi$ radians. A positive skewness indicates that the guanine crystal structures are dominated by organized patterns, such as vertical lines or fibers. In contrast, a skewness near zero suggests a more uniform angular distribution, implying a more random or isotropic structure.

For this study, each image was divided into square sub-images of size $L\times L$, where L corresponds to half the length of the shortest side of the original image. Sub-images were extracted with a horizontal and vertical displacement of 0.1L pixels. A total of 594 sub-images of anchovy, 561 of mote sculpin, 705 of sardine, and 407 of snoek were analyzed, and the three aforementioned features were computed for each.

In terms of frequency characteristics, vertical, horizontal and angle skewness differed between species (Kruskal-Wallis main effects all P-values lower than 0.001). Dunn´s post hoc test showed showed all species differed from each other (Dunn´s post hoc test, all P-values lower than 0.05). Anchovy showed an average vertical magnitude of $0.80\pm 0.53$, horizontal of $1.53\pm 0.59$, and skewness of $-0.63\pm 0.20$. Mote sculpin presented $0.89\pm 0.60$ (vertical), $1.36\pm 0.63$ (horizontal), and $0.33\pm 0.25$ (skewness). Sardine exhibited $0.59\pm 0.47$ (vertical), $1.71\pm 0.57$ (horizontal), and $0.08\pm 0.31$ (skewness), while snoek showed $1.08\pm 0.65$ (vertical), $1.17\pm 0.62$ (horizontal), and $-0.13\pm 0.21$ (skewness).

Analysis of the average frequency components by species suggests that anchovy and mote sculpin exhibit elliptical patterns in their high-energy spatial distributions. In anchovy, these frequencies are more widely dispersed in angular orientation, while in mote sculpin they are more aligned with the horizontal frequency component. Sardine showed less dispersion in the vertical component but a broader horizontal frequency distribution. Snoek displayed a more symmetric pattern, closer to a circular distribution within the selected frequency band.

These three features were used to train a KNN classifier performing 1000 bootstrap iterations with an 80/20 train-validation split. The resulting overall accuracy was 92.14% (95% CI: 90.11–94.07), indicating high accuracy in identifying fish species. Class-wise sensitivity and specificity values were consistently high: Sardine reached 97.10% sensitivity (95% CI: 94.33–99.29) and 96.79% specificity (95% CI: 94.90–98.41); Anchovy achieved 90.80% sensitivity (95% CI: 85.71–94.96) and 98.42% specificity (95% CI: 97.02–99.70); Snoek yielded 94.03% sensitivity (95% CI: 89.02–98.78) and 97.22% specificity (95% CI: 95.71–98.66); and Mote sculpin showed 86.00% sensitivity (95% CI: 79.65–92.04) and 97.08% specificity (95% CI: 95.32–98.55). Greater confusion was observed between the sardine and mote sculpin classes, which may be associated with similarities in their structural orientation patterns, particularly the predominance of vertical structures in both species.

Discussion

This study provides a comprehensive analysis of the optical properties and internal morphological structures of some pelagic fish species from the HCS, off the coast of Chile. The assessment of the guanine crystals arrangement in shaping a silvery camouflage on fish evolving under strong predatory pressures is of great significance, advancing our understanding of their camouflage strategies in the open ocean. The species analyzed, Sardine (Strangomera bentincki), Anchovy (Engraulis ringens), Snoek (Thyrsites atun), and Mote sculpin (Normanichthys crockeri), exhibit distinct reflective properties at whole fish level, and are tightly related to the arrangement of the guanine crystals within fish skin.

Optical properties and reflectance analysis

The results demonstrated that the reflectance spectra of the HCS silvery species (sardine, anchovy and snoek) are relatively flat in the visible range (450-750 nm), which aligns with the expected behavior of a mirror-like camouflage described by Johnsen and Sosik¹⁹. Their investigation and the present results supports the concept that mirror-like surfaces with flat reflectance spectra are an effective camouflage strategy in the open ocean. Also, Jordan et al.²⁰, explored the reflective properties of silvery fish species and their role in camouflage. Using hyperspectral imaging, they found that the flat reflectance spectra in the visible range contribute to effective background matching in open-water environments. In contrast, the Mote sculpin displayed an upward (in wavelengths) trend in reflectance, indicating a distinct mechanism of light interaction. The spectral analysis revealed that the silvery species exhibit a low degree of variation in reflectance across a broad range of wavelengths, supporting the hypothesis that their camouflage relies on minimizing detectability by blending into the open ocean environment. This flat spectral behavior is not only consistent with the presence of biogenic guanine structures in the skin, but also with their arrangement, which are responsible for broad-spectrum reflection and reduced polarization differences with the surrounding light. In contrast, the reflectance profile of the Mote sculpin suggests that its skin is less adapted for open-ocean camouflage, which may be due to its different ecological niche or behavior. Mote, however, is a common by-catch species in the Sardine and Anchovy fishery, and so they occupy a similar position in the water column at least during the fishing season. The increased light reflection at higher wavelengths in Mote, also suggest that occupies deeper waters than sardine and anchovy, as wavelengths above 700 nm are soon absorbed by the first meters of the water column.

It seems that not only coating variations influence the spectral signature of a particular species, but also the characteristics of their ecological habitats can significantly influence spectral signature divergence. Various selective pressures seem to be at play within the visual ecology of fish residing in different environments. The convergence observed across micro-structures and whole fish optical characteristics in the fish species studied here seems a promising line of research for those exploring camouflage, crypsis, and underwater vision.

Recent research has revealed an additional aspect of light-scattering processes in open oceans, which creates spatially heterogeneous backgrounds. Specifically, polarization (the directional vibration of light waves) alters the light environment in a way that varies with the position of the sun in the sky. This angular variability in the polarized light field means that fish encounter different polarization patterns depending on the direction they face. In such conditions, mirrors, which typically reflect light uniformly, may become highly detectable to organisms with polarization-sensitive vision. Given that the ability to detect polarized light is common across diverse fish families, natural selection likely plays a role in shaping the evolution of fish reflectance properties (multi layers structure) to adapt to this dynamic and variable environment.

Morphological basis of reflectance

The structural basis of these optical properties lies in the organization of guanine crystals within specialized skin cells, known as iridophores or leucophores, which are responsible for the silvery, reflective appearance of many fish²¹. These biogenic structures are composed of guanine, a purine base, and are organized into highly reflective stacks. The arrangement of guanine crystals in these cells influences how light interacts with the fish’s skin, leading to various optical effects. The present results confirm that a chaotic arrangement results on a whole fish silvery reflectance response (sardines, anchovies and snoek), while a more ordered/structured arrangement (mote) leads to less silvery reflective response. Structurally, guanine crystals are layered in a way that creates interference patterns, which can enhance or alter light reflection depending on the angle of light. This organization allows fish to reflect light in a broad spectrum, contributing to their silvery or iridescent appearance²². The specific arrangement of these crystals can vary between species, resulting in different reflective properties and camouflage abilities camouflage in their respective environments¹¹. This finding aligns with research by Gur et al.²³. In this study, the silvery species exhibited a disordered arrangement of guanine crystals, with cytoplasmic thickness and multilayer spacings consistent with the creation of broad-spectrum reflectors. This disordered structure is crucial for reflecting light uniformly across various incident angles, enhancing camouflage effectiveness in the open ocean. In contrast, the Mote sculpin displayed more ordered and thicker multilayer structures, suggesting a different optical strategy potentially related to a deeper lifestyle in the water column, where distinct environmental pressures influence the evolution of skin reflectance properties. The more chaotic arrangement of guanine crystals in silvery species, minimizing the polarization of reflected light, is a critical adaptation for remaining inconspicuous in the open ocean¹⁴. This aligns with the idea that different fish species employ varied structural strategies to achieve effective camouflage, influenced by their specific ecological pressures²⁴. Additionally, the results support the importance of multilayer interference in creating effective camouflage in pelagic fish. The notable concordance between the measured reflectance spectrum and the chaotic guanine multilayer structures in the silvery species supports their role in enhancing reflectance and contributing to the characteristic silvery appearance. Moreover, the study highlights the strategic approach adopted by Sardine and Anchovy, which involves deliberate ”controlled” disorder in guanine crystal stacks to enhance reflectivity while maintaining dermal layer thickness. The interdigitation of crystal stacks in these species creates regions of varying crystal spacing, contributing to overall structural disorder and improved reflection efficiency. The optical thickness of alternating layers in these fish matches closely with that of an ideal reflector, optimizing their camouflage abilities by enhancing reflection efficiency and reducing polarization effects. So, the differences in guanine crystal organization between the studied species underscore the evolutionary adaptations that have occurred in response to specific ecological pressures, with each species developing unique structural strategies to optimize camouflage in their respective environments. Ren et al.²⁵, exploring how guanine crystal stacks produce structural coloration in silvery fish, also concluded the importance of these crystal arrangements in achieving efficient light reflection, which is essential for the fish’s ability to blend into their surroundings and avoid detection by predators. The current study further supports this idea by demonstrating that the chaotic structure of guanine crystals in silvery species not only enhances reflectivity but also reduces polarization, thus optimizing their camouflage capabilities. In contrast, species like the Mote sculpin, with more ordered guanine crystal arrangements, may utilize a different optical strategy that is better suited to their deeper water column lifestyle, where different environmental pressures are at play.

Comparative analysis and ecological implications

The study’s findings highlight the adaptations that different species have developed in response to their specific ecological pressures. The silvery species’ camouflage strategy is a clear adaptation to the pelagic environment, where uniform background matching is essential for avoiding detection by predators. In contrast, the Mote sculpin appears to have evolved a different set of optical properties that are likely more suited to its ecological niche, possibly reflecting different predation pressures, or simply, a deeper distribution as red wavelengths are absorbed in the first 10 meters. The results also highlight the role of guanine crystal organization in determining the optical properties of fish skin. The disordered structure observed in silvery species aligns with the findings of Addadi et al.²⁶, who emphasized the importance of guanine crystal arrangements in achieving efficient light reflection. This study extends that understanding by demonstrating how variations in crystal structure can lead to different camouflage strategies across species. The controlled disorder in guanine crystal stacks observed in Sardines, Anchovies and Snoeks, for instance, is a strategic adaptation that enhances reflectivity while maintaining dermal layer thickness, optimizing their camouflage capabilities. This is the first evaluation of silvery camouflage in a fish assemblage subjected to a strong predatory pressure and thus where camouflage is particularly under a strong selection.

Fourier frequency analysis and structural characteristics

By means of a Fourier frequency analysis of the guanine crystal structure of the different species, we have confirmed the SEM analysis. The guanine arrangements present in anchovy, snoek, and sardine exhibit spatial frequency components of similar magnitude along the wavenumbers corresponding to the visible spectrum in multiple directions, according to the 2D Fourier magnitude spectrum. This causes these structures to disperses the entire range of visible light similarly and in all directions, so that the light reflected by the spatial structure cancel out appearing silver in color. Through hyperspectral analysis, it was observed that the spectral signatures of anchovy and sardine exhibit a similar spectral pattern; however, anchovy shows lower reflectance than sardine. This can be explained by the Fourier spatial frequency analysis, which reveals that the guanine crystal structure in anchovy is more chaotic than in sardine, resulting in greater destructive light interference in anchovy. Whereas by analyzing the structure of the Mote sculpin species, it presents a magnitude spectrum with a high horizontal energy component, which justifies the presence of the horizontal pattern of vertical lines. The structure of this species has higher energy in the horizontal component than in the vertical component, with a higher concentration of energy around wavenumber 1.4, so this structure scatters red light better, which justifies the reddish hue of this species. This novel approach to determine whereas a spatial crystal structure correlates with the color of a fish by considering the guanine-cytoplasm structure as a diffraction grating, pave the way to analyze further fish skin structures to determine the weight that light diffraction of the skin have among all the other light-matter interactions.

In this work, in addition to analyzing the spatial frequency components, three features were proposed to measure the orientation of guanine crystals, which reflects how chaotic the structure is. Using these features, a KNN classifier was implemented to distinguish the four species, achieving an average overall accuracy of 92.14% (95% CI: 90.11–94.07). This demonstrates that, although silvery fish exhibit similar reflectance behavior, they present structural differences in the arrangement, size, and shape of guanine crystals, which generate distinct spatial patterns that affect their reflectance due to the destructive interference caused by a disordered crystal arrangement.

Denton and Land²⁷ hypothesized that the silvery sheen observed in certain fish scales and skin could arise from two potential mechanisms: (1) a random distribution of thicknesses across different layers within the scales and skin, or (2) a systematic variation in thickness throughout these layers. In a subsequent examination of the silvery reflectance in fish, McKenzie et al.¹⁴, characterized fish skin as a chaotic system where the thicknesses of both high and low refractive index layers are randomly distributed. They also proposed that the neutralization of reflection polarization plays a crucial role in determining the reflective properties of fish skin. In this study we have found that both ordered and disordered features can be determinant in the reflectance enhancement since organic crystals are used to produce a remarkable array of optical phenomena based on the reflection and dispersion of light. The thickness of the cytoplasm spacings is highly variable, whereas the guanine crystals are much thinner and have a much narrower thickness distribution compared to the cytoplasm layers. While we acknowledge that dehydration and re-hydration procedures applied to samples prior microscopy may have caused some degree of distortion at that fine scale, as the same methodology was used, comparison among species are still valid. Furthermore, the measured dimentions of the microstructures are well in agreement with our Fourier frequency analysis (see below). The same narrow distribution of crystal thicknesses was found for the different scales and skins examined, Anchovy or Sardine. Various organisms that incorporate guanine into their reflective systems, like the spider Phoroncidia rubroargentea²⁸ in their abdomen and in skin of Koi fish²³ featuring a silvery integument, exhibit a consistent pattern in the distribution of guanine crystal thickness and cytoplasm spacings.

Guanine crystals ideally sized for peak reflection are around 60-70 nm, but sardines and anchovies have slightly lower dimensions suggesting potential limitations in crystal formation. For instance, the spider Tetragnatha montana generates thicker crystals, which consist of roughly 20 nanometer-thick crystals separated by a 30-nanometer layer of amorphous guanine²⁹. This unique arrangement allows the crystals to achieve a thickness close to the quarter wavelength optical thickness (approximately 130 nanometers). The broader distribution of cytoplasm thickness is incongruent with systems designed for producing precise structural colors, such as the carapace of the male copepod Sapphirina spp.³⁰ and the dynamic lateral stripe of the Neon Tetra fish¹⁰. In these intricate systems, both guanine crystal and cytoplasm spacing thicknesses are anticipated to exhibit a more consistent and regular pattern, ensuring peak reflectivity at specific wavelengths. It is noteworthy that many early pioneering studies investigating the reflective properties in fish with silver scales, which involved measuring guanine crystal thickness, consistently observed widely scattered and irregular thickness distributions. It should be noted, however, that the methods employed led to the complete dissolution or loss of the crystals, resulting in the measurement of the cavities that remained after crystal dissolution. These cavities underwent distortion during the drying and cutting processes, which may have also perturbed the thickness of the cytoplasmic layer. These factors likely account for the considerably larger thickness values observed in these studies for guanine crystals. While we still lack a comprehensive understanding of the mechanisms governing the formation of these intracellular guanine crystals and the precise determinants of their size, the fact that different systems yield crystals with similar shapes and sizes suggests the existence of a common formation mechanism. Worth mentioning that the dehydration process during SEM preparation may alter tissue structure, potentially affecting crystal arrangement. While this is a potential limitation, our findings remain insightful for comparisons among the species tested and also in general, given the agreement with the Fourier analysis. Furthermore, previous authors have described these structures with the same technique and also with optical analysis^31,32,33. The elevated refractive index, coupled with the variance in refractive indices across distinct crystallographic orientations (known as birefringence), arises from the anisotropic crystal structure of guanine. This dependency on the efficiency of electronic transitions, whether parallel or perpendicular to the H-bonded layers, plays a pivotal role. Consequently, the disparities within the fish scale system enable the reflection of light across a considerably wider range of angles. This effect is critical for camouflage in water. Unlike air, water scatters light, causing it to hit the fish’s skin and scales from various angles. Guanine crystals excel in this environment because they reflect light broadly and efficiently, minimizing the contrast between the fish and the downwelling sunlight. This is achieved by a delicate balance within the crystal structure: an ordered arrangement of guanine crystals with a hint of disorder creates the striking, broad-angle silvery iridescence we observe.

Implications for future research

Here, by using a powerful combination of hyperspectral imaging and morphological analysis, the camouflage of four fish species (3 silvery and one non silvery) was studied. This study provides a foundation for further research into the optical and morphological adaptations of fish from various habitats and ecological contexts, providing insights into the evolutionary pressures that have shaped their development. Future studies should include more species and varying environmental conditions could also yield valuable information on the complex interplay between structure, function, and ecology in the evolution of camouflage. Moreover, the study highlights the need to explore the genetic and developmental mechanisms that control guanine crystal formation and organization. There is also the need to further understand how this crystal organization plays a role (and its relevance) among all the interactions the light has with the fish skin. Understanding these processes could provide a deeper understanding of how different camouflage strategies have evolved and how they might continue to evolve in response to changing environmental conditions. In conclusion, this study demonstrates that Sardines, Anchovies and Snoeks exhibit a superior camouflage strategy compared to Mote sculpin fish, thanks to a two-pronged approach: 1 - Increased Reflective Surface: These fish have a larger area covered by stacks of guanine crystals, essentially multiplying the number of tiny mirrors reflecting light. 2- Optimized Crystal Packing: They achieve a thinner crystal stack by strategically adjusting the crystals’ orientation within the skin and promoting a more interdigitated arrangement (think interlocking fingers). This clever packing, while introducing some disorder in spacing, significantly enhances the overall reflectance. The study’s findings contribute to a deeper understanding of the biological mechanisms underlying camouflage and emphasize the importance of considering both optical and structural factors in the study of marine species. Future research could build on these findings to explore the broader ecological and evolutionary implications of these adaptations.

Methods

Fish samples and image acquisition

Fish measurements and imaging were conducted at the Comparative Animal Physiology Laboratory at the Universidad de Concepción, Chile. Fish were sourced from the Chilean fisheries control authority in accordance with national regulations and transported to the Comparative Animal Physiology Laboratory. All procedures involving animals were conducted under the institutional animal ethics permit CEBB-1501 2023, approved by the Institutional Ethical and Animal Use committee of the Universidad de Concepción. All methods were carried out in accordance with relevant guidelines and regulations, and are reported in compliance with the ARRIVE guidelines.

Adult fish, recently caught off the Bio-Bío coast, arrived within three hours via SERNAPESCA (Chilean National Fisheries and Aquaculture Service). This rapid transportation ensured the fish were fresh upon arrival. Previous research indicates that the freshness of fish can affect spectral data, and our methodology accounted for this by analyzing the fish promptly after arrival. This approach minimizes the potential impact of freshness on the reflectance spectrum, ensuring that our data accurately reflects the natural optical properties of the fish. This methodology is supported by previous studies on fish spectral analysis³⁴. For this study, two datasets were acquired: one consisting of hyperspectral images and the other of microscopic images. Therefore, we refer to each dataset as the HSI-dataset and SEM-dataset, respectively.

Both datasets contain images of the four fish species selected for this study, as shown in Fig. 1a–d: Sardine (Strangomera bentincki), Anchovy (Engraulis ringens), Snoek (Thyrsites atun), and Mote sculpin (Normanichthys crockeri), respectively.

Hyperspectral images dataset and its analysis

Fig. 7

Setup of the hyperspectral imaging system for reflectance measurements, composed of: (a) Pika L hyperspectral camera, (b) camera mounting tower, (c) halogen illumination, (d) sample board, (e) linear translation stage, (f) power supply/control interface and (g) laptop.

The HSI-dataset was captured using an indoor hyperspectral imaging system (benchtop-reflectance), as shown in Fig. 7, with its components labeled: (a) a Pika L hyperspectral camera (Resonon Inc., USA), mounted on the tower (b), and (d) the sample board positioned on the linear translation stage (e), which is illuminated from above by halogen lamps (c). During data acquisition, the stage moves, translating the sample beneath the camera. The movement and image acquisition are controlled by a laptop (g) using Spectronon Pro software (Version 3.4.11)³⁵ and the control interface (f).

The camera was calibrated before each use with both a white and a black calibration boards, to eliminate the effects of camera dark current and uneven light source distribution³⁶. The corrected diffuse reflectance spectral image is computed as follows: $R_{Cal}=\left( I_{Raw}- I_{Dark}\right) /\left( I_{White}- I_{Dark}\right)$, where $I_{Raw}$ represents the original diffuse reflectance spectral image of the sample, $I_{White}$ is the diffuse reflectance image of the whiteboard, and $I_{Dark}$ is the dark image (acquired with the black calibration board).

Using this setup and calibration procedure, we captured 25 images of Sardine, 28 of Anchovy, 34 of Snoek, and 18 of Mote culpin, obtained from 16, 23, 16, and 9 individual specimens, respectively. Baseline reflectance and the angle fish hyperspectral curve formed along wavelengths was extracted and assessed for potential differences among species. Data met paramemetric asssumtions and so a one way ANOVA using species as factor was performed.

From each hyperspectral image, a region of interest was selected, corresponding to the lateral tissue of the fish, excluding areas with saturated reflectance. Each pixel has an associated spectrum at that point. In this study, each image was analyzed based on its representative spectrum, corresponding to the average spectrum.

To characterize the spectral behavior of each species, each spectrum was approximated using zero- and first-order polynomials. In the case of the zero-order polynomial, the constant value corresponds to the average reflectance of the curve. The first-order polynomial was approximated using least squares, and the second coefficient, with the first and second coefficients referred to as bias and slope, respectively.

The variables analyzed were average reflectance and slope, each assessed under four experimental conditions (n = 105 total observations). Prior to the main analyses, assumptions of normality and homogeneity of variances were tested using the Shapiro–Wilk and Levene’s tests, respectively. As both assumptions were satisfied, a one-way analysis of variance (ANOVA) was applied separately to each variable to determine whether there were significant differences among the experimental groups. The ANOVA results included the F statistic, associated p-values, and the partial omega squared ($\hat{\omega ^2_P}$) as a measure of effect size. Partial $\hat{\omega ^2_P}$ was used to estimate the proportion of variance in the dependent variable attributable to the treatment effect, with values above 0.14 considered large. Post hoc pairwise comparisons between groups were conducted using Student’s t-tests, with Holm’s adjustment applied to control for the family-wise error rate. Adjusted p-values ($P_{Holm-adj.}$) were reported to determine statistically significant differences among treatment groups ($\alpha = 0.05$). Data visualization was performed using violin plots overlaid with boxplots to display the distribution, central tendency, and dispersion of the data for each group. Individual data points were included to illustrate sample variability, and statistically significant pairwise differences were highlighted with horizontal brackets and adjusted p-values.

Scanning electron microscopy dataset

The SEM dataset comprises images of 6 Sardines, 6 Anchovies, 5 Snoeks, and 7 of Mote sculpins. This dataset is smaller than the HSI-dataset due to both the higher cost of the technique used and the additional laboratory procedure required to perform the cuts in the fish skin we detail next.

Fish skin was excised using a scalp ($3\times 3\times 3 mm$), and fixed in 4% Paraformaldehyde in phosphate buffer with pH 7.4 at 4$^\circ$C overnight. Post fixation was done using 1–2% osmium tetraoxide for 3–4 hours. Graded dehydration was done using ethanol, followed by critical point drying using carbon dioxide³⁷. The specimens were then coated with gold using JEOL JEC- 3000 FC sputter coater. Ultrastructural analysis was done using the scanning electron microscope JEOL JSM- IT 300. Within each photograph, morphometric analyses were conducted to quantify the dimensions of the guanine plaques. Additionally, the distances between the cytoplasm and each plaque were measured, along with the overall thickness of the multilayer reflector system. To address the scale discrepancy between microscopic observations and the macroscopic appearance of the fish, we employed a combined approach of SEM and hyperspectral imaging. SEM was used to characterize the microstructure of guanine platelets within iridophores, revealing their arrangement in multiple layers that function as Bragg reflectors. This microscopic structure is crucial for broadband reflectance. Concurrently, hyperspectral imaging was performed on a defined region of interest (ROI) on the intact fish, allowing for the mapping of the spatial distribution and spectral properties of reflected light at the macroscopic level. By focusing both SEM and hyperspectral analysis on the same ROI, we were able to directly correlate the microscopic arrangement of the guanine platelets with the macroscopic optical properties. The observed multi-layered structure of the platelets in SEM images explains the broadband reflectance detected by hyperspectral imaging, ultimately contributing to the overall silvery appearance of the fish. Furthermore, morphological analysis of skin samples provided additional insights into the organization of these reflective structures.

Diffraction grating and spatial frequency analysis

Diffraction grating

A diffraction grating is a collection of reflecting (or transmitting) elements separated by a distance comparable to the wavelength of light under study. It may be thought of as a collection of diffracting elements, such as a pattern of transparent slits (or apertures) in an opaque screen, or a collection of reflecting grooves on a substrate. It provides angular dispersion, i.e., the ability to separate wavelengths based on the angle that they emerge from the grating. Gratings can be transmissive, like the multi-slit aperture, but they can also be reflective where the grooved surface is overcoated with a reflecting material such as aluminum.

The diffraction of monochromatic light incident on a grating is governed by the equation $m\lambda =d(\sin \alpha + \sin \beta )$, where m is an integer value describing the diffraction (spectral) order, $\lambda$ is the light’s wavelength, d is the spacing between grooves on the grating, $\alpha$ is the incident angle and $\beta$ is the diffracted angle of light (leaving the grating). The directions are discrete because constructive interference occur at integer multiples of the light’s wavelength¹⁸.

The primary purpose of a diffraction grating is to disperse light spatially by wavelength. A beam of white light incident on a grating will be separated into its component wavelengths upon diffraction from the grating, with each wavelength diffracted along a different direction. Dispersion is a measure of the separation (either angular or spatial) between diffracted light of different wavelengths. Angular dispersion expresses the spectral range per unit angle, and linear resolution expresses the spectral range per unit length.

The angular spread $\Delta \beta$ of a spectrum of order m between the wavelength $\lambda$ and $\lambda +\Delta \lambda$ can be obtained by differentiating the grating equation, assuming the incidence angle $\alpha$ to be constant. The change D in diffraction angle per unit wavelength is therefore

$$\begin{aligned} D=\frac{\partial \beta }{\partial \lambda }=\frac{m}{dcos\beta }. \end{aligned}$$

(1)

Our rationale is that we can extract the spatial patterns from the crystal SEM images using the image’s Fourier transform. This pattern will give us the grating spacing parameter, d, so we can use the grating equation to determine which wavelength of incident light would interacts with the guanine-cytoplasm structure the most.

Fourier transform

The Fourier transform is a mathematical operation that transforms a function of time (or space) into a function of frequency (or spatial frequency). For our purposes, it decomposes a spatial-domain signal (image) into its constituent spatial frequencies.

The discrete-space image f(m, n) of dimensions $M\times N$ that can be represented as the matrix $\varvec{f} = [f(m,n), 0\le m \le M-1,0\le n\le N-1]$ has a 2D discrete Fourier transform (DFT) defined by

$$\begin{aligned} F(u,v) \triangleq \sum _{m=0}^{M-1} \sum _{n=0}^{N-1} f(m,n) \exp \left[ -j2\pi \left( \frac{u}{M}m + \frac{v}{N}n\right) \right] \end{aligned}$$

(2)

where (u, v) are the integer frecuencies $0\le u \le M-1$, $0\le v\le N-1$, with units of cycles per image. Hence, the DFT is also a $M\times N$ image (generally complex-valued) that can be represented as a matrix $\varvec{F} = [F(u,v), 0\le u \le M-1,0\le v\le N-1]$.

As usual in Fourier analysis, our interest is to examine the DFT magnitude, i.e., $|F(u,v)| = \sqrt{F(u,v)F^*(u,v)}$, where $*$ is the complex-conjugate. In this study, we normalize it by its maximum value and we expressed it in decibels, i.e., $|\varvec{F}|_{dB}=10\log _{10} \left( \frac{\varvec{F}\varvec{F}^*}{\max (\varvec{F}\varvec{F}^*)}\right) .$

Diffraction grating, spatial frequency and visible light dispersion

For the dispersion phenomenon to occur, the grid spacing must be comparable to the wavelength of the light incident on the grid. Therefore, With a grid whose wavenumber falls within the range of 1.33 to 1.61, there will be greater dispersion of red light. In the range of 1.75 to 2, greater dispersion of green light will occur; between 2 and 2.22, the dispersion will be greater for blue light; and from 2.22 to 2.5, greater dispersion of violet light will be observed.

Now, to utilize this whole information, for each SEM image $\varvec{f}_i$ we computed its corresponding DFT image $\varvec{F}_i$ using (2). The frequency components (u, v) of each complex-valued DFT gave us the spatial frequencies (in cycles per image) of the guanine-cytoplasm structures. The radial frequency defined as $\Omega = \sqrt{u^2 + v^2}$ (cycles per image) and its corresponding direction $\theta = \tan ^{-1}(v/u)$ (in radians) can be calculated to determine the direction $\theta$ of maximum radial frequency. This direction defines the most relevant frequency component of the spatial structure of the crystals. For simplicity we termed this direction the Horizontal component and the least relevant frequency as the vertical component in Fig. 6. Potential differences between species on vertical and horizontal magnitudes and angle skewness, were tested by a one way ANOVA. Data did not meet parametric assumptions and so a Kruskal Wallis test was applied, followed by a Dunn´s post hoc test to find differences between species. Differences were considered significant with a P value lower than 0.05.

The obtained spatial frequencies of interest (cycles per image) where then converted to the wavenumber (cycles per 1000nm) simply using the instantaneous field-of-view (iFoV) of the SEM images. This wavenumber, k, was then related with a diffraction grating whose grooves are repeated each $k=1000/d[cycles/1000nm]$, and the corresponding color that such a grating would diffract the most.