Coloured ornamental traits could be effective and non-invasive indicators of pollution exposure for wildlife

Sub-lethal impacts of environmental pollution are difficult to detect and usually require invasive methods. We review the evidence that changes in ornamental colour in fish and birds can be associated with a variety of pollutants. Although these ornaments evolved for mate choice, they could also provide a non-invasive indicator of pollution exposure.


Introduction
In the field of conservation biology, anthropogenic changes to the environment are widely known to contribute to population declines and extinctions of vulnerable species (reviewed by Gibbons et al., 2000). In the field of behavioural ecology, ornamental traits are widely understood to be reliable indicators of individual condition and quality (Zahavi, 1975;Hamilton and Zuk, 1982) that consistently reflect environmental conditions (Griffith et al., 1999; . 1 Monaghan, 2008). Despite extensive independent development and some appreciation of their combined explanatory potential (Hill, 1995;Sutherland, 1998), these two bodies of theory are almost never combined by conservation biologists and wildlife managers. As suggested by Hill (1995), better integration of these fields could provide a general, proactive, non-invasive and effective diagnostic tool for detecting subtler anthropogenic effects on individuals before they are signalled by decreased reproduction or higher probability of mortality, which ultimate cause declining populations. In turn, those advantages could reduce the cost and increase the efficacy of mitigation efforts (Clout et al., 2002;Buchholz, 2007).
Estimation of anthropogenic effects using ornamental traits requires mechanistic understanding of both anthropogenic effects and the expression of ornamental traits in wildlife. Separately, these mechanisms are particularly developed for one category of each component; chemical pollution as a type of anthropogenic effect, and the pigment-based colouration of the integument as a type of ornamental trait. In this review, we integrate these two specific categories to illustrate how they could be used by conservation biologists now, describe some limitations for expanding that use, and identify profitable directions for future research.
Chemical pollution can have acute negative effects on wildlife via development (Colborn et al., 1993), physiology (Ross et al., 1994), behaviour (Zala and Penn, 2004), reproductive success (Goutte et al., 2014) and survival (Martínez-Abraín et al., 2006). It can also increase rates of hybridization (Ward and Blum, 2012) and decrease genetic diversity (Bickham et al., 2000), ultimately reducing population viability and leading to extinction (Kidd et al., 2007). All of these effects may be foretold by changes to the conspicuously coloured ornamental traits that are prevalent in species with strong mate selection, such as some birds and fish. As proposed by Darwin (1871), ornamental traits are morphological or behavioural traits that evolved via sexual selection to confer reproductive, rather than survival, advantages to the bearer through enhanced ability to attract mates. An example of this is the flamboyant tail of the male peacock. Zahavi (1975) expanded Darwin's theory to propose that such traits are so costly to their bearers that they can be produced and maintained as handicaps only by individuals of greater quality and condition. As such, ornamental traits are now widely recognized to reveal good genes, developmental conditions, local environments at the time when such traits are developed, or all three (von Schantz et al., 1999).
Ornamental traits reveal the quality and condition of their bearers partly because they have high phenotypic plasticity (Cotton et al., 2004), meaning that their expression is particularly sensitive to the environment and to the cascade of physiological mechanisms produced by stressful events (Hill, 1995;Buchanan, 2000). Environmental conditions can be canalized during development (e.g. Naguib and Nemitz, 2007), but they can also be reflected in a more recent or current fashion (e.g. Velando et al., 2006). This plasticity and honesty make ornamental traits especially useful for detecting both temporal and spatial variation in environmental conditions. A prevalent type of ornamental trait in vertebrates is integument colouration (Hill and McGraw, 2006b), which includes skin (e.g. Velando et al., 2006), scales (e.g. Plasman et al., 2015), hair (e.g. West and Packer, 2002) and feathers (e.g. Safran and McGraw, 2004). The appearance of the integument is dependent on its structure and the pigments deposited inside it as well as by the dirt, waxes and abrasion applied to or acquired by it over time (Hill and McGraw, 2006a). At least four kinds of costs can be revealed by these traits. First, the pigments involved in trait colouration may be physiologically costly to produce, such as melanin (Jawor and Breitwisch, 2003). Second, the pigments may be available only in some kinds of high-quality foods, such as carotenoids (McGraw, 2006a). Third, the pigments may be needed for other functions, such as immune or antioxidant system support, and available for ornaments only when those functions have been met (Faivre et al., 2003;Galván and Alonso-Alvarez, 2008). Finally, each of these costs may cooccur as additive or multiplicative effects. By any of these routes, only those individuals in good condition can afford to produce, acquire or allocate pigments for trait colouration without compromising survival, making colouration an honest signal to observers (Hamilton and Zuk, 1982;Grafen, 1990).
The ubiquity, plasticity and honesty of integument colouration make it a powerful indicator of the competing costs of environmental stressors, such as anthropogenic pollutants. The short temporal scale of environmental effects on the integument suggests the potential of using integument colouration as a tool to detect negative environmental impacts on individuals long before they cascade through populations, communities and ecosystems with effects that are more difficult and costly to reverse. These features give ornamental integument colouration enormous but largely untapped potential to diagnose many conservation problems at their most proximate stages, in order to support solutions that are more proactive (e.g. Baruch-Mordo et al., 2013), generalizable (sensu Caughley, 1994) and holistic (Caro and Sherman, 2013).
In the following sections, we describe the two main types of pigment-based colouration found in ornamental integuments of vertebrates, melanins and carotenoids, and review the known effects on them of a variety of chemical pollutants ( Fig. 1 and Table 1). Existing research in this area has focused only on fish and birds, groups that frequently express quantifiable coloured integuments and so have welldeveloped literatures. The Latin names for the species we refer to in the text are provided in Table 1. Following this review, we discuss the limitations to the use of colouration as an indicator of pollution exposure and propose some recommendations for future research.

Overview of pigment and pollutant types
Melanins are the most prevalent pigments in vertebrates, producing many yellow-brownish (pheomelanin) and greyblack (eumelanin) traits (reviewed by McGraw, 2005). They can also combine with keratin and air to produce structural colours, such as blue, violet, green and ultraviolet hues, as well as iridescent colours (Prum, 2006). As vertebrates can synthesize melanins de novo from amino acid precursors (Lin and Fisher, 2007), they are frequently assumed to be an unlimited resource for ornamental trait building, strictly controlled by genes (e.g. Badyaev and Hill, 2000;McGraw, 2006b). However, more recent studies have demonstrated that the expression of melanin-based ornaments can also be influenced by environmental factors, such as rearing conditions, parasite infestation and diet quality, to give these ornaments a plastic and honest quality in vertebrates (Fargallo et al., 2007;reviewed by McGraw, 2008;Guindre-Parker and Love, 2014).
Despite the ubiquity of melanin in vertebrates, a second kind of pigment, namely the carotenoids, are more widely studied. They produce many yellow, orange and red traits, but cannot be synthesized by vertebrates (Schiedt, 1989), so their acquisition through the diet makes them a limited resource. Moreover, carotenoids are thought to play key physiological functions, accepting free radicals to protect cells and tissues from oxidative damage and acting as immune system enhancers (reviewed by Lozano, 1994;von Schantz et al., 1999; but see Hartley and Kennedy, 2004;Pérez-Rodríguez, 2009). In combination, these properties provide information about animal foraging, carotenoid uptake and allocation efficiency that is both accurate and visible (McGraw, 2006a).  (Bortolotti et al., 2003). Photographs by Gary Bortolotti and Russell D. Dawson, courtesy of Kim Fernie. (C) Iridescent blue-green plumage of tree swallows (McCarty and Secord, 2000). Photograph by Natalia Lifshitz. (D) Yellow colouration of fins and tale of amarillo fish (Arellano-Aguilar and Macías Garcia, 2008). Photograph by Shane Webb, courtesy of Constantino Macías García. (E) Body yellow-orange spots of guppies (Toft and Baatrup, 2001;Kristensen et al., 2005;Shenoy, 2012). Photograph by Erik Baatrup. Lower panel provides examples of changes in ornamental colouration of fish and birds exposed to anthropogenic pollutants. (F) Ornamental body colouration of adult male red shiner (Cyprinella lutrensis) exposed to the oestrogen 17β-estradiol (i) and control water (ii). (McGree et al., 2010). (G) Gape and tongue colouration of female black-legged kittiwakes (Rissa tridactyla) with low (i) or high concentrations of pesticides and PCBs in blood (ii; Blévin et al., 2014). Photographs by Olivier Chastel. (H) Red bill spot (controlled by bill size) of yellow legged male gulls (Larus michahellis) experimentally fed with oil from the Prestige oil spill (i) or control sunflower oil (ii; Pérez et al., 2010). Photographs by Cristobal Pérez.

4
Both types of pigments are pertinent to the large literature addressing the negative physiological effects of pollution on wildlife either by disrupting the endocrine system (Tyler et al., 1998) and/or by altering the oxidative status of individuals (Isaksson, 2010). Endocrine-disrupting compounds interfere with the normal function of the endocrine system in several ways, such as mimicking natural hormones, blocking or altering the binding of natural hormones to hormone receptors, altering production and breakdown of natural hormones and altering the production and function of hormone receptors (Tyler et al., 1998). These effects have the potential to impact ornamental colouration if its expression is controlled by or related to sexual hormones (e.g. Bókony et al., 2008).
The effect of pollution on the oxidative status of individuals is subtler, but it may also be more ubiquitously and generally expressed. Metabolism produces reactive oxygen species as a byproduct, and when this production exceeds the capacity of the antioxidant defense and repair mechanisms, an imbalance called oxidative stress arises, which leads to oxidative damage to biomolecules (i.e. proteins, lipids and DNA; Halliwell and Gutteridge, 2007). This imbalance is known to be a key component in the life-history trade-offs between growth, reproduction and self-maintenance or survival (Monaghan et al., 2009). Therefore, animals exposed to pollutants with pro-oxidant properties and whose ornamental colouration depends upon pigments related to antioxidant defenses are faced with a trade-off between using antioxidants to combat oxidative stress or to signalling functions and, ultimately, may have to sacrifice their investment in signalling to support survival.
The categorization of pollutants is confusing because it sometimes refers to generalized effects on natural systems (e.g. endocrine-disrupting compounds), sometimes to the chemical nature of compounds [e.g. polycyclic aromatic hydrocarbons (PAHs)] and sometimes to the originating industries (e.g. petroleum products). In this review, we simplify that categorization by referring consistently to the anthropogenic context of pollutants and review examples of effects on ornament colouration for each of the following: pharmaceuticals and active ingredients in personal care products (PPCPs); pesticides; industrial pollutants that are prevalent but not restricted to petroleum production; and heavy metals that frequently result from mining and manufacturing. Within each of these source categories, we review the known effects on the production and expression of melanins and on the uptake and use of carotenoids for ornament colouration ( Fig. 1 and Table 1).

Pharmaceuticals and active ingredients in personal care products
This category conventionally includes prescription and nonprescription drugs, oral contraceptives, fragrances and cosmetics (Daughton and Ternes, 1999). These ubiquitous products are usually disposed of or discharged into the environment on a continual basis via domestic and industrial sewage systems and wet-weather runoff, where they enter water courses and come into direct contact with aquatic organisms. Relative to other pollutants, the emission of pharmaceutical products tends to be chronic and concentrated in areas with high human density (Kolpin et al., 2002). Given that some PPCPs (such as oral contraceptives) are specifically designed to target and modulate the endocrine system, they are likely to affect any form of ornamental colouration that is controlled by it.
Despite the high potential for PPCPs to influence ornamental colouration, this has been addressed only in fish, and no study has targeted birds. To date, there is no evidence of negative effects of PPCPs on melanin-based colouration, where likely effects include ornamental traits that are mediated and stimulated by hormones, such as androgens or thyroid hormones (reviewed by McGraw, 2006b).

5
The anticipated effects of exogenous oestrogens on aquatic species have been robustly demonstrated with fish. In a laboratory study, exposure to the artificial oestrogen 17αethinylestradiol (EE2) reduced the area of carotenoid-orange colouration as a percentage of the total body area of male guppies (Kristensen et al., 2005), a characteristic known to function as a sexually selected signal to females (Kodric-Brown and Nicoletto, 2001). In addition, exposed male guppies showed a significant reduction in courtship behaviour, sperm count and paternity when competing for fertilizations with unexposed males (Kristensen et al., 2005). Similar effects occurred in male guppies, zebra fish and red shiners that were exposed in the laboratory to the natural oestrogens 17β-estradiol (E2), EE2 and E2, respectively, and then expressed dampened mating colouration (Toft and Baatrup, 2001;Larsen et al., 2008;McGree et al., 2010). Treated male guppies produced fewer offspring and displayed paler ornaments even after 3 months of recovery in clean water (Toft and Baatrup, 2001). Male zebra fish also showed a significant reduction in courtship behaviours, failing to induce spawning in females (Larsen et al., 2008). Likewise, courtship behaviours of male shiners were significantly reduced, as was their fertilization success, causing null hatching success of the fertilized eggs. Importantly, when exposure ceased, other reproductive end points, but not colouration, of the shiners improved significantly, demonstrating that the dulling effects were longer lasting (McGree et al., 2010).

Pesticides
Pesticides are a broad category of compounds that are commonly used to protect crops, livestock, domestic animals and humans from damage and diseases caused by fungi (fungicides), insects (insecticides), rodents (rodenticides), competition with unwanted plants (herbicides) and other so-called pests. Pesticides can also be categorized according to their chemical structure into inorganic (compounds that contain arsenic, copper, lead or mercury) and organic chemicals [artificially synthesized chemicals, such as dichlorodiphenyltrichloroethane (DDT); reviewed by Freedman, 2001]. Animals may be exposed to pesticides through consumption of contaminated food or water, as well as through inhalation of contaminated air. Although pesticides have been used for millennia, their use has increased greatly during the past half-century because of their economic benefits and increased worldwide availability (Pimentel et al., 1992). This has increased the geographical and temporal risk of exposure for wide-ranging and migratory birds (Stutchbury, 2009) among other species. Pesticides primarily cause damage to organisms by producing free radicals that overwhelm the antioxidant system (Abdollahi et al., 2004), but some compounds also alter the endocrine system (Colborn et al., 1993).
Effects on melanin-based colouration can be complex and counterintuitive, as has been shown in red-legged partridges exposed to the contact herbicide diquat. In these birds, the eumelanic (black) plumage of adults was unexpectedly enlarged following exposure to the pesticide during development, to produce larger black-spotted bibs and black flank bands (Galván and Alonso-Alvarez, 2009), which signal higher-quality individuals of both sexes (Bortolotti et al., 2006). However, exposure to the herbicide simultaneously reduced expression of pheomelanin to cause the brown flank bands to be smaller (Galván and Alonso-Alvarez, 2009). The authors speculated that the pesticide increased oxidation, which depleted the intracellular antioxidant glutathione and reduced the amount available to produce brown pigment in the flank bands. Given that circulating glutathione blocks eumelanin synthesis, its depletion also caused the larger black bibs and flank bands (Galván and Alonso-Alvarez, 2009). This study underscores the need to understand underlying physiological processes to interpret the effects on colouration of exposure to pollution.
The negative effects of pesticides on carotenoid-based ornamental colouration have been robustly demonstrated in fish. For example, the consistent preference by female guppies for males with larger and more intense orange spots (described above) clearly predicts a disadvantage to males exposed to the fungicide vinclozolin and to the principal metabolite of the insecticide DDT [p,p′-dichlorodiphenyldichloroethylene (DDE)] because they reduced both the size and intensity of the spots (Baatrup and Junge, 2001). These morphological changes also occurred when guppies were exposed to the herbicide atrazine, with corresponding reductions in courtship displays and aggressive behaviour toward other males during competition for mates (Shenoy, 2012).
The effects of pollutants on adult colouration can also result from exposure during development. The sexually selected yellow colouration of adult male Amarillo fish (Macías Garcia, 1991) was duller in males exposed as embryos to the insecticide methyl parathion and resulted in lower rates of female visitation and copulation attempts during courtship (Arellano-Aguilar and Macías Garcia, 2008). The authors speculated that these effects in adults could have been caused by one or both of long-lasting damage to the physiological systems that process and deposit carotenoids, or permanent damage to their antioxidant system with resulting increases in the use of dietary carotenoids to combat oxidative stress during adulthood.
In some cases, the interacting and opposing effects of pollutants on different colourful ornaments may help to identify basic physiological pathways and reveal the detrimental mechanisms of pollution exposure. An example of this potential is provided by the red-legged partridges (described above), in which the diquat-exposed birds with enhanced eumelanic traits also had paler carotenoid-based ornaments in the form of red beak and eye rings (Alonso-Alvarez and Galván, 2011). Given that the red colouration in the head integument is positively correlated with the health status of individuals (Mougeot et al., 2009) and positively affects female reproductive investment (Pérez-Rodríguez and Viñuela, 2008) pesticide exposure were negative, despite the increase in eumelanin expression (Alonso-Alvarez and Galván, 2011). These negative effects on the carotenoid-based colouration of red-legged partridges were later confirmed for two fungicides (thiram and difenoconazole) and an insecticide (imidacloprid) via reductions in the percentage of carotenoid pigmentation in the eye ring. As additional evidence of net negative effects, all three pesticides reduced the size of eggs, imidacloprid and difenoconazole reduced the fertilization rate, and thiram and imidacloprid reduced chick survival (Lopez-Antia et al., 2013). Similar effects have been detected in free-living birds. Female black-legged kittiwakes with higher levels of various pesticides and polychlorinated biphenyls (PCBs) in their blood samples exhibited duller orangered labile integuments (i.e. eye ring, gapes and tongue), and these ornaments are believed to reflect individual quality in both sexes also (Blévin et al., 2014).

Industry-related compounds
Many kinds of industrial pollutants potentially affect the colouration of ornamental traits in vertebrates, but only three have been studied in this context, namely PCBs, phenols (particularly bisphenol A and octylphenol) and PAHs.
Polychlorinated biphenyls are stable, human-made organic compounds that were commonly used in electrical applications, hydraulic equipment, plasticizers, paints, plastics, rubber production and as an insulating agent (reviewed by Blocker and Ophir, 2013). Production of PCBs ceased in 1972 (Dunlap, 1981) after their long-lasting toxic effects were realized, but the compounds persist in the environment and are capable of bio-accumulating with trophic position in lipid tissues. The PCBs are well known to disrupt endocrine function by agonizing and antagonizing natural oestrogens (Colborn et al., 1993) and negatively affecting thyroid function (Boas et al., 2006). Similar endocrine-disrupting and oestrogenic effects occur from exposure to bisphenol A and octylphenol (Bonefeld-Jørgensen et al., 2007). These industrial pollutants typically reach wildlife through discharge to water courses, often via sewage treatment effluent (Kolpin et al., 2002). Despite its detrimental effects, production of bisphenol A is increasing because of its desirable commercial qualities for the manufacture of polycarbonate, epoxy and polyester resins (Crain et al., 2007). Octylphenol is a degradation byproduct of chemicals used in the manufacture of detergents and in agricultural and industrial products (Ying et al., 2002).
A third class of industrial pollutants that has been studied in the context of colouration is PAHs, which are the most toxic components of liquid petroleum products, but also occur as airborne pollutants (Eisler, 1987). Airborne PAHs typically result from combustion involving both anthropogenic sources (e.g. motor vehicles and many industrial processes) and natural ones (e.g. forest and prairie fires). Polycyclic aromatic hydrocarbons can reach aquatic environments via condensation in the atmosphere as well as via discharge of liquid waste through domestic and industrial sewage effluents, surface runoff from land, and spillage of petroleum products into water bodies (reviewed by Eisler, 1987). The acute toxicity of PAHs is mainly attributed to the oxidative stress they generate in exposed organisms, which adaptively activate their antioxidant system in order to survive in polluted environments (Pérez et al., 2010).
Among the few studies that have addressed the negative effects of PCBs on vertebrate colouration, only one examined melanic pigmentation. McCarty and Secord (2000) showed that subadult female tree swallows breeding along a river with high levels of PCB pollution displayed the blue-green iridescent plumage that is characteristic of adult birds of both sexes, rather than the brown upperparts that usually characterize subadult females. They speculated that the early expression of adult traits was caused by the oestrogenic properties of PCBs (McCarty and Secord, 2000). If delayed plumage maturation in females of this species is an adaptation to signal subordinate status, thereby reducing conspecific aggression (Coady and Dawson, 2013), earlier development of adult plumage may disadvantage these birds.
Surprisingly few studies have addressed the effects of PCBs on carotenoid-based colouration in fish and birds, given the enormous amount of literature that addresses their effects on more conventional physiological metrics (e.g. Eisler, 1987;Colborn et al., 1993;Flint et al., 2012), but they suggest consistently negative effects. For example, breeding male American kestrels that were exposed to a mix of PCB Aroclors lost brightness of the facial skin with yellow-orange pigments, which otherwise signals individual quality (Bortolotti et al., 2003). Likewise, black-legged kittiwakes showed dampening of the orange-red colouration in their faces with increasing concentrations of pesticides and PCBs in their blood (Blévin et al., 2014).
As a class of industrial chemicals, the effects of phenols on colouration have been studied only in fish. Guppies exposed to octylphenol produced fewer offspring, while displaying orange spots that were smaller (owing to inhibited growth of the spot) and less brightly coloured (Toft and Baatrup, 2001). Interestingly, cessation of exposure caused the colour but not size of the orange spots to recover, which might reflect functional differences in information content of the two colour signals. A similar effect occurred when male red shiners were exposed to the phenol bisphenol A for a period as short as 14 days, wherein there was a loss in the intensity of red breeding colouration in heads and fins and of blue iridescence in bodies (Ward and Blum, 2012). As an additional consequence of these colour changes, both sexes were less able to discriminate between conspecific and heterospecific partners during mate choice trials (Ward and Blum, 2012).  (Eisler, 1987), but their effects on ornament colouration have been explored only recently and in relation to the Prestige oil spill (González et al., 2006). Researchers working with yellow-legged gulls breeding in colonies located in the pathway of the spill focused on the carotenoid-based red bill spot, which is exhibited by both sexes and appears to reflect reliably the antioxidant status and capacity for parental investment of the bearer Morales et al., 2009). They found that the size of the red spot was negatively correlated with blood-based measures of aspartate aminotransferase, an enzyme that is commonly used as an indicator of hepatic damage in birds . Additionally, gulls that were experimentally fed heavy fuel oil from the same spill had higher concentrations of PAHs, vitamin E and carotenoids in blood plasma, suggesting that the latter two had been mobilized for antioxidant defenses to cause the observed reductions in the size of the red bill spot (Pérez et al., 2010). Like the studies examining orange spots in guppies after exposure to pesticides (above), the work after the Prestige spill suggests that: (i) sexually selected, carotenoid-based traits respond very rapidly to environmental pollutants; and (ii) these traits might be particularly suitable as indicators after acute events.

Metals
Several heavy metals, metalloids and trace elements are natural constituents of the Earth's crust, but they can also accumulate as a function of anthropogenic activity to become persistent environmental contaminants (Freedman, 2001). As metals cannot usually be degraded or destroyed, organisms that are exposed to metals via inhalation, absorption and ingestion often experience cumulative effects via one or both of bioaccumulation over time and biomagnification over trophic levels (Duruibe et al., 2007). Some elements, such as lead, mercury and arsenic, are toxic even at very low concentrations, but others, such as copper, selenium, iron and zinc, are essential to many biological processes and become toxic only at elevated concentrations (Ercal et al., 2001). Exposure to detrimental concentrations of either type can cause a cascade of effects that include decreased immune responsiveness (Kakuschke and Prange, 2007), increased oxidative stress (Valko et al., 2005) and, ultimately, lower survival or reproductive performance (Dauwe et al., 2004). The effects of metals on ornamental traits have been demonstrated for both melanins and carotenoids, but the direction of these effects differs.
For melanins, exposure to pollution from a lead smelter that included cadmium, zinc and copper resulted in larger eumelanic (black) breast stripes in great tits (Dauwe and Eens, 2008). This effect was surprising because larger stripes are preferred by females during mate choice (Norris, 1990). A similar positive trend was apparent in the brighter melanin-based plumage of belted kingfishers nesting close to a mercury-contaminated river and with higher levels of mercury in feathers and blood (White and Cristol, 2014). Like pesticides, metals seem to increase the expression of eumelanin-based ornamental traits, potentially disrupting the honesty of information they convey about bearer condition.
Also aligning with the evidence on ornamental traits for pesticides, metals appear to have a dampening effect on carotenoid-based ornaments. In great tits, pollution from metal smelters has been robustly demonstrated to reduce the carotenoid-based, yellow colouration in the breast feathers of both nestlings and adults to signal a loss in individual quality and condition (Eeva et al., 1998;Dauwe and Eens, 2008;Geens et al., 2009). The authors of these correlative studies emphasized different proximate effects of metal pollution on ornaments, which could include reduced access to carotenoid sources (Eeva et al., 1998), metal-induced oxidative stress (Geens et al., 2009), or both effects (Dauwe and Eens, 2008). It is already known that metal pollution reduces carotenoid synthesis in plants (Rai et al., 2005) which, in turn, could also reduce carotenoid concentrations in herbivores (e.g. caterpillars; Isaksson and Andersson, 2007). Whether they are predators or prey, individuals living in polluted sites appear generally to have reduced access to carotenoids, which may increase levels of oxidative stress to reduce condition in organisms with, as well as those without, carotenoid-based ornaments.
Another important challenge for the use of ornaments as indicators of environmental pollutants will be to unravel the separate effects on multiple traits and their associated impacts on individual quality and fitness. Many species exhibit more than one ornamental trait that may function as redundancy (e.g. back-up signals; Møller and Pomiankowski, 1993), as multiple, or reinforcing, messages (Johnstone, 1996;Candolin, 2003) or even as evidence of trade-offs with each other (Andersson et al., 2002). Our review of the currently available literature suggests that ornaments coloured by different pigments may respond to a given pollutant in opposite directions [e.g. red-legged partridge (Alonso-Alvarez and Galván, 2011) and great tit (Dauwe and Eens, 2008)], but also that different ornaments formed with the same pigment respond in a similar manner (e.g. various carotenoid-based ornaments in the amarillo fish; Arellano-Aguilar and Macías Garcia, 2008).
A third challenge for the use of ornamental traits as indicators of pollution exposure is the inability of humans to see ornaments as other species might. A well-known example of this limitation is provided by the tetrachromatic vision of both birds (reviewed by Cuthill, 2006) and fish (Bowmaker, 1990), which affords detection of the ultraviolet portion of the colour spectrum. Ultraviolet colouration is already known to be used in mate choice and to reflect the condition and quality of the bearer (reviewed by Prum, 2006), and could provide another indicator of pollution exposure if it could be measured accurately. Although conventional digital cameras can be fitted with ultraviolet sensors (e.g. Lifepixel, 2016), their use in the field has been restricted thus far to tame and accessible species (e.g. Meyer-Rochow and Shimoyama, 2008). Even the melanic and carotenoid-based traits we measure more confidently could look very different to other species (reviewed by Cuthill, 2006).

Recommendations for future research
Despite the limitations we acknowledge above, we see enormous potential for using ornamental traits that resulted from sexual selection to signal present-day exposure to anthropogenic pollutants. We suggest that in order to achieve this potential we will require the following: (i) the study and establishment of baselines of trait colour in the populations of interest; (ii) conducting broad-scale colour monitoring along different gradients of exposure to pollution; and (iii) an expansion in the existing techniques used to measure colouration, which are not much less invasive than the physiological metrics they might replace. Typically, quantification of colour still involves the capture and extensive handling of individuals, and those requirements potentially exclude many of the species that are most threatened by anthropogenic pollutants.
The first recommendation, the establishment of colour baselines, is necessary because of the enormous variability in the responses of ornamental traits to the wide variety of anthropogenic pollutants. Until more is known about how trait responses can be generalized among traits, species, pollutants, locations, etc., baseline metrics of trait colouration should be limited to small temporal and spatial scales with comparable environments. For example, bird feathers might be measured annually for a set of species in relation to standardized moult or breeding schedules and within a designated area. Once appropriate baselines are established, more robust comparisons could occur over space, time and anthropogenic conditions. Establishment of such baseline levels could create opportunities to assess predicted effects in such changes as land use practices, water treatment standards, industrial development and even the cumulative effects of adjacent human populations.
A second recommendation that could complement the development of the aforementioned baselines, is a broadscale monitoring of ornamental colour. For this, the colour of ornamental traits of many species could be measured along known gradients of exposure to pollution to obtain a general picture of the sensitivity of such traits. This could be carried out repeatedly along several similar gradients to generalize the conclusions and along gradients of different sources of pollution. In addition to this, researchers or managers already monitoring populations and capturing individuals for other purposes could take colour measurements and/or photographs to expand our knowledge on colour variation in wild populations.
Given that the difficulty and invasiveness of capturing and taking samples from wild animals is a core motivator for our review paper, we see a primary research need as the development of remote sensing tools for measuring the colouration of ornamental traits. By remote sensing, we mean the use of cameras that can collect multispectral imagery while operating autonomously, in a similar manner to the way wildlife cameras are increasingly used for animal detection and surveillance (Turner et al., 2003;Swann et al., 2004;Bolton et al., 2007) and to the way vegetation type and phenological stage have long been assessed via multispectral sensors contained in satellites or aeroplanes. Already it is possible to combine these techniques via multispectral sensors in unmanned vehicles that can operate in each of air, land and water as media (reviewed by Linchant et al., 2015). The revolutionary potential of these vehicles has already been appreciated in many other conservation domains, and more applications are revealed steadily (reviewed by Conservation Drones, 2016). So far, the use of these tools for monitoring wildlife populations has been restricted mostly to detecting the presence of individuals or specific behaviours (e.g. Swann et al., 2004;Claridge et al., 2005; .  Grenzdörffer, 2013), but they are already being used to monitor the health of plants through changes in colouration that include the infrared portion of the colour spectrum (Fletcher et al., 2001;Newete et al., 2014).
Even if drones can be brought rapidly to the service of quantifying ornament colour for conservationist purposes, a related challenge is to create methods for calibrating photographs taken outdoors in varying conditions. Just as it is important to standardize assessments of coloured ornaments based on human observation (Montgomerie, 2006) and digital photography , it will be important to develop methods to standardize images collected remotely. Although the same principles apply, much greater variation in light intensity and colour balance will apply to remote imagery collected outdoors. Variation in distance to subjects, resolution and orientation will make it harder to quantify relatively simple traits, such as ornament size. Overcoming the challenge of standardization will not be a trivial task, and it is probably best approached for this purpose within multidisciplinary teams that could include engineers, computing scientists and others who have solved similar problems for other purposes.
Equally important to standardizing the appearance of sexually selected ornaments by remotely collected photographs will be to continue the excellent pioneering work others have done to relate ornament colouration to physiological condition (e.g. Hamilton and Zuk, 1982;Kodric-Brown, 1985) and then to relate both metrics to environmental pollutants (e.g. Hill et al., 2002; Table 1). Although we focused our review on birds and fish because they have received all the attention to date, other taxonomic groups may offer advantages for the extension of these metrics via remote sensing. For example, amphibians are well known to be highly sensitive to environmental pollutants (Gibbons et al., 2000). Both amphibians and reptiles are typically terrestrial, which may make study via hand-held sensors in the wild easier than it would be for either birds or fish. The subjects of this work are not restricted to vertebrates; any taxonomic group that exhibits colourful integument, such as insects and crustaceans, is potentially relevant to advancing a general understanding of the way colouration could signal the detrimental presence or effects of anthropogenic pollution. With such general knowledge, conservation practitioners might be able both to detect and to mitigate the effects of pollution long before they cause local population declines or extirpation.
A final target of future research is to identify generalizations from studies of sexually selected indicators and use them to predict which species, populations and individuals are most likely to adapt to rather than deteriorate from changing environments. One example of this proactive approach would be to target the closely related, sympatric species that are more likely to use colourful ornamental traits for premating reproductive barriers (Ritchie, 2007;Price, 2008). Without those barriers, hybridization and genetic introgression could cause reductions in offspring survival (Pryke and Griffith, 2009) or the loss of local adaptations (Bourret et al., 2011), ultimately contributing to extinctions (Rhymer and Simberloff, 1996). A well-known example of this sequence occurred with cichlid fishes in Lake Victoria. There, intensified deforestation and agricultural practices increased water turbidity and reduced colour perception by fish, which increased rates of hybridization to cause cascading losses of fish diversity (Seehausen et al., 1997).
Another proactive research target would be to determine the extent and implications of increased expression of melanic traits in individuals that have been exposed to pollution. The ability of melanin to bind with metal ions means that animals with melanic ornaments can eliminate excess metals, which are essential in small quantities but toxic in higher concentrations, by sequestering them in hair of feathers (McGraw, 2003). Recent work suggests that this mechanism can cause directional selection for more melanic phenotypes in polluted environments (Chatelain et al., 2014), but the reverse may also occur (Senar et al., 2014). If melanin-based sequestration of toxicants is widespread in animals, it could have big implications for changing patterns of biodiversity, especially in urban areas. For example, McKinney's (2002) categorization of bird species as urban avoiders, adapters and exploiters was originally based on correlations between life-history traits and abundance, but a systematic review might reveal that the same categories correlate with the prevalence of melanic traits. The intriguing relationships among metal pollution and melanins have been most explored in birds, but they deserve investigation in mammals, especially in invasive urban-adapting species that exhibit variation in both black and reddish colouration caused by melanins (e.g. coyotes, Canis latrans; Anderson et al., 2009). Clearly, much more research is needed to address how melanins interact with a variety of anthropogenic effects.

Conclusions
Exponential growth in human populations and associated habitat degradation increasingly threaten the retention of biodiversity on Earth. Stemming this loss will require efficient conservation action that identifies the mechanisms of population decline, acts proactively to reduce the cost and increase the efficacy of mitigation, and operates holistically to generalize efforts across ecological units and scales. When these three things can be achieved, the results are heartening. For example, the identification of volatile chlorofluorocarbons in aerosol sprays as the cause of ozone depletion resulted in rapid regulatory change, widespread compliance, the recovery of an essential ecosystem service, and benefits to hundreds of species (Noakes, 1995).
Unfortunately, that critical first step-identifying the mechanism of population decline-is surprisingly difficult, especially when it involves anthropogenic pollutants with nebulous components, sources and effects. Millions of chemicals are produced annually and released into the environment (Postel, 1987), but only a small fraction of those are tested for their environmental impacts (Tolba, 1992). There is an enormous need to develop and apply more methods to monitor the health of wildlife exposed to new and existing anthropogenic pollutants, but current methods are invasive, sometimes lethal (e.g. Farombi et al., 2007), often limited to blood-based or reproductive parameters (e.g. Fernie et al., 2001) or are reliant on opportunistic observations of mortality (e.g. Piatt et al., 1990). Additionally, none of these methods is suitable for estimating effects of pollutants in wild but sensitive populations that inhabit locations that are remote, vulnerable to disturbance, difficult to access or endangered. Alas, those are precisely the species, populations and locations where such assessments are most needed to support proactive and effective conservation action! In this review, we champion an idea presented by Hill (1995), more than 20 years ago, that the ornamental traits of many organisms could be used to fill this void, revealing the quality and condition of environments that are also occupied by many other species. Since then, many researchers have explored species-and trait-specific effects of particular pollutants to accumulate a substantial literature. We synthesized and reviewed that literature in relation to two types of pigments that are prevalent in ornaments in vertebrate classes, i.e. melanins and carotenoids. We also assessed their known interactions with several environmental pollutants. Despite the number of relevant studies, most of this literature describes laboratory or case studies and is usually specific to taxonomic groups or particular toxicological effects. These characteristics impede the emergence of generalizable mechanisms that could reveal wider conservation problems and suggest appropriate mitigations. Consequently, the potential for ornamental traits to comprise a versatile tool for conservation biology remains unmet. We see seminal roles for development of this theory using pollution as an anthropogenic effect, ornamented vertebrates as study subjects, and pigments as mechanistic intermediaries. These targets are illustrative because pollution often interferes with the biochemical creation and physiological maintenance of ornamental traits, which are prevalently expressed with pigments. Moreover, pigments are costly to synthesize or acquire, are fairly well understood in the contexts of biochemistry, physiology, toxicology and sexual selection, and should be readily visible to both human observers and their automated devices. With time, we expect that ornamental traits will become familiar tools for diagnosing and addressing a wide variety of anthropogenic effects, with tractable applications for hundreds of species and ecosystems worldwide.
We attempted to advance Hill's (1995) insight by reviewing that literature and using it to pose ideas for future research that could help generalize the use of this method, while acknowledging some important limitations. We hope this review encourages more researchers and wildlife managers to include measures of ornamental colour whenever possible and, especially, when it can easily be added to existing monitoring or research protocols. We see enormous potential for these traits to provide a much-needed, non-invasive tool for detecting subtle and early effects of pollution long before they can be seen as more catastrophic effects at the level of populations. Species that exhibit carotenoid-based ornaments appear to offer particular promise as highly responsive indicators that might reveal pollution exposure that is not easily detected, but actually occurs, in other species and even whole communities.
Despite the great potential we see in the utility of ornamental traits as environmental indicators, we also emphasize the importance of unravelling the specific physiological mechanisms of pollutants on colouration before generalizing their interpretations, particularly when anthropogenic stressors cause opposing or multiplicative effects. A striking contradiction is evident in the enhancing effects of both pesticides and heavy metals on eumelanic (dark) traits, with simultaneous negative effects on pheomelanic (lighter) and carotenoid-based traits that are undoubtedly detrimental for organisms.
More systematic work on the specific sources, mechanisms and mitigations of pollutant effects on sexually selected ornaments could provide conservation biologists with a suite of context-specific canaries for the diverse coal mines of anthropogenic effects. At the same time, such work may provide unanticipated insights relevant to more basic questions in physiology, ecology and evolution.

Acknowledgements
We thank M. Murray and A. Gangadharan for their valuable comments on an early version of the manuscript and R. Beamonte-Barrientos for his valuable comments on the final version of the manuscript. The impetus for this review was provided by the Research on Avian Protection Project, which revealed that tens of thousands of water birds land annually on the ponds containing process-affected water in the mineable oil sands region of Alberta, but <1% appear to die as a result. This information revealed the urgent need to understand the sublethal health effects on wild birds of those landings, for which no other toxicological information exists. Conventional forms of this information will be difficult to collect owing to the remote location and on-site security and safety restrictions for data collection by people. Funding for that project resulted from a creative sentence in the ruling of R. v. Syncrude Canada Ltd. 2010 ABPC 229.