The art of otolith chemistry: interpreting patterns by integrating perspectives

Additional keywords: barium, carbonates, increments, membranes, stable isotopes, strontium.

Observers actually know that despite the apparent simplicity of the laws governing their formation, the works of nature are infinitely varied, from the most important to the least, no matter what their species or family … the quarters of an orange, the leaves of a tree, the petals of a flower are never identical; it thus seems that every kind of beauty draws its charm from this diversity. [Pierre-Auguste Renoir in 1884; Nochlin 1966, p. 46]

A robust understanding of migration, life history strategies and growth dynamics is critical for effective management of fishes in marine, coastal and freshwater systems worldwide. The practice of otolithology, or the broad use of otoliths to investigate fish dynamics (sensu Gaemers 1978), has yielded an explosion of data and subsequent insight into the dynamics of fishes, their communities and ecosystems. The field has proliferated in the types of questions that can be asked with these remarkable structures, spurred on by monumental discoveries such as the identification of annual (Reibisch 1899) and daily (Pannella 1971) increments. A particularly fruitful expansion of this field derives from the increasingly detailed and high-throughput chemical analyses of otolith increments. Much of this growth has been tightly coupled to technological advances, such as the development and refinement of laser technology allowing multi-element high-resolution and high-precision measurements that are, importantly, reasonably affordable for a wide array of fish ecologists. The expansion and maturation of the otolith chemistry subfield is evident in the published proceedings of the International Otolith Symposia (IOS) over the years. For instance, in the book published after the Hilton Head, North Carolina, IOS, only 10% (4 of 40) of contributed papers focused on the elemental or isotopic composition of otoliths (Secor et al. 1995), whereas in the two special issues published after the Mallorca, Spain, symposium, 48% of contributed papers incorporated chemical analyses of otoliths or analogous structures such as vertebrae (Morales-Nin and Geffen 2015; Geffen et al. 2016). A review of the titles of contributed talks and posters at the IOS in Keelung, Taiwan, in 2018 showed an estimated 43% of talks and 34% of posters focused on otolith chemistry techniques (see http://isis.cmima.csic.es/aforo/presentations/IOS2018Taiwan.pdf, accessed 22 November 2018). Beyond the official IOS and their proceedings, other special sessions devoted to otoliths and comparable structures have become regular features at marine and aquatic science conferences, with their own associated publications (e.g. Walther et al. 2017; Hunter et al. 2018). Clearly, a growing number of investigators is turning towards chemical tools to unravel key information about fish life histories.

The increasing prominence of otolith chemistry can be seen in the broader body of published literature. To assess the overall publication trends in otolith chemistry, a literature search was conducted in ISI Web of Science (conducted on 3 October 2018 with search terms: otolith AND (chem* OR microchem* OR elem* OR isotop*)) from 1967 through 2017 to quantify annual publications of elemental and isotopic analyses of fish otoliths since the work of Devereux (1967) on oxygen isotope ratios. Results were culled to remove non-fish biomedical human and animal model research, exclude fluorescent marking research and retain only peer-reviewed journal articles (excluding book chapters or conference abstracts), leaving a total of 1505 papers published over five decades (Fig. 1a). Although this search is undoubtedly an undercount, it illustrates the non-linear trajectory and rapid growth in published work in recent decades, driven in large part by advances in instrumentation, experimental validation and high-throughput approaches. The critical role of the periodic symposia can be seen in upticks following IOS years (1993 in Hilton Head, USA; 1998 in Bergen, Norway; 2004 in Townsville, Australia; 2009 in Monterey, USA; and 2014 in Mallorca, Spain), with publications of associated books or special issues, such as this one. Given that publication rates in all subject areas have also grown significantly over the same time period, the annual percentage of otolith chemistry papers relative to total otolith publications as indexed in the Aquatic Sciences and Fisheries Abstracts (ASFA; search term: otolith*) was calculated (Fig. 1b). Although percentages were initially high in the early years due to low numbers of total otolith publications each year before 1990, the proportion increased steadily throughout the 1990s and subsequently stabilised in the past decade. These trends are similar to those observed in prior assessments of annual publication rates (Campana and Thorrold 2001; Secor 2010; Starrs et al. 2016; Tanner et al. 2016). Clearly, the field is robust, with an average of 91 publications annually constituting an average of 19% of all otolith publications since 2013.

Fig. 1.  (a) Annual publications of otolith chemistry papers from 1967 to 2017 as indexed in ISI Web of Science (search terms: otolith AND chem* OR microchem* OR elem* OR isotop*). Results were culled to retain peer-reviewed journal articles on fish otolith chemistry only (n = 1505). Asterisks indicate years of previous International Otolith Symposia. (b) Annual percentage of otolith chemistry papers relative to all otolith papers as indexed in the Aquatic Sciences and Fisheries Abstracts (AFSA) database (search term: otolith*). AFSA retrievals were limited to peer-reviewed publications in scholarly journals (n = 8944).

Although the applications of otolith chemistry are increasingly diverse, the primary questions investigated can be broadly grouped into six major categories: (1) identifying natal origins and connectivity patterns; (2) assessing individual variability in migration patterns; (3) reconstructing environmental exposure histories (e.g. temperature, dissolved oxygen, pollution); (4) quantifying growth and physiological dynamics (e.g. metabolism, stress, reproduction); (5) validating age estimates; and (6) determining dietary histories.

Investigating any of these questions requires the appropriate selection of chemical proxies that reveal the targeted variable of interest, an extractive or in situ analysis of sufficient material to yield quantifiable measurements of the analyte and appropriate placement and resolution of analyses in relation to increment spacing. A given element may yield information for multiple questions, given the possibility of both extrinsic and intrinsic control on elemental incorporation. Still, otoliths hold incredible promise for revealing comprehensive life history information about the lives of fishes. This paper reviews developments in otolith chemistry research by using a metaphor to structure the discussion of critical frontiers in the field. Specifically, it is argued that an otolith is like a painting. Much like a painting is influenced by multiple ‘filters’ that influence the final image, including the studio, the canvas, the media and the artist, an otolith is influenced by analogous filters, including the environment, the constituent elements and isotopes, the biogenic structure composition and the physiology of the fish itself. Our challenge as otolith chemists is to interpret the observed chemical patterns much like a museum visitor interprets the painting. Before this metaphor is used, the paper begins by briefly considering the historical foundations of this field and the technological advancements that have led to a proliferation of otolith chemistry data that require interpretation.

The growth of the field prompts us to reflect on the origins of otolith chemistry. Far from springing from a single source of inspiration, otolith chemistry can be viewed as a natural evolution of the broader advances in biogeochemistry across systems and taxa. The case of strontium is emblematic of this evolution. The element was originally discovered in 1790 by Adair Crawford and William Cruickshank and named for the small village of Strontian, Scotland, where unidentified minerals were unearthed in local lead mines (Mellor 1961). Strontium was first isolated in 1808 by electrolysis by the great chemist Sir Humphry Davy^A, who also isolated calcium, magnesium and barium (Knight 1998). The potential for strontium to reveal biogeochemical processes and elemental cycling was quickly realised. Forchhammer (1865) first discovered strontium in seawater and living tissues of Fucus algae, indicating the possibility of active biotic uptake. Dieulafait (1877) identified strontium in fossilised and modern brachiopods and suggested that shell composition reflected contemporary ocean chemical composition. Papillon (1870) made one of the first assessments of strontium in living vertebrates and, through experimental diet manipulations, discovered that strontium in pigeon bones was derived at least in part from dietary sources. These early investigations prompted a wide-ranging number of publications quantifying strontium in water, soil and biological tissues throughout the late 19th and early 20th centuries (for a review, see Odum 1950).

The first systematic approach to understanding strontium cycling came with the ground-breaking work of the ecologist Howard T. Odum. A graduate student at Yale, Odum was encouraged to investigate what he called ‘the strontium problem’ by his supervisor G. Evelyn Hutchinson. Hutchinson and Odum recognised that a comprehensive model of strontium cycling was not merely a chemical curiosity, but a window into global ecosystem processes and the biotic and abiotic connections that link diverse systems (Limburg 2004). For his thesis, Odum (1950) analysed 1100 samples of Sr/Ca in water, sediment, rocks and biota to quantitatively understand the magnitudes of elemental reservoirs and transfers among them. He used these measurements to propose large-scale cycling models that addressed weathering, riverine transport, atmospheric deposition and ocean cycling, in addition to uptake and incorporation into living tissues and calcified hard parts (Odum 1950). This remarkable body of work was the basis for a series of papers that addressed specific aspects of cycling dynamics (Odum 1951a, 1951b, 1957a, 1957b). Notably, Odum (1951b) experimentally measured the relationship between water and Physa gastropod shells and calculated a distribution factor, or partition coefficient, to quantify elemental incorporation rates for this species, much like otolith chemists continue to calculate today. Although the effect of ambient water composition on elemental uptake was clear, Odum was keenly aware of the possibility of taxonomic and physiological factors to regulate uptake. He hypothesised that membrane transport capacity, indicated by the thickness of the tissue separating seawater from the calcifying surface, and the complexity of the circulatory system could explain taxonomic differences in strontium uptake from algae to vertebrates (Odum 1957a; Fig. 2). Note that Fig. 2 includes fish otolith measurements, which is the first reported quantitative measurement of otolith strontium I have found in the literature. Odum’s work was foundational for the field of elemental tracer ecology in general, and emphasised many of the same issues otolith chemists grapple with today, including geological controls on water composition, temporal and spatial variability in ambient composition and physiological mediation of elemental uptake dynamics. These questions continue to drive important field-based and experimental work in otolith chemistry.

Fig. 2.  Semiquantitative model proposed by Odum (1957a) hypothesising the role of the strength of a circulatory system (y-axis) and the thickness of the calcifying tissue (x-axis) on incorporation of strontium. Lines are isopleths approximating equal Sr/Ca ratios. Note the inclusion of fish otoliths in the upper left quadrant. Figure reprinted with permission from Odum (1957a).

Currently, otolith chemistry as a discipline is diverse in both analytical approaches and the types of questions investigated using these remarkable biogenic structures. Probe-based instrumentation, such as laser ablation inductively coupled plasma mass spectrometry (ICP-MS), the current workhorse of the field, has improved in sensitivity and precision for many analytes. Laser wavelengths of 213 or 193 nm are common in many ICP-MS facilities, and femtosecond lasers hold promise for further improvements in analyses with single- and multiple-collector ICP-MS (Lord et al. 2011; Yang et al. 2011). Concordantly, the improvement in spatial resolution has allowed more targeted analyses of specified life history periods. Although whole otolith dissolutions were common in the 1990s, increasingly precise probe-based methods have allowed targeting of natal cores or post-settlement periods or complete life history transects across all growth increments. Furthermore, two-dimensional spatial mapping of elemental or isotopic distributions are possible with ICP-MS (Woodhead et al. 2007; Wang et al. 2013; McGowan et al. 2014; Petrus et al. 2017) or techniques such as proton-induced energy emission spectroscopy and scanning X-ray fluorescence microscopy (Limburg and Elfman 2017). These mapping approaches yield valuable visualisations of chemical heterogeneity across otoliths, allowing researchers to identify concentric fluctuations in an analyte, such as strontium, as well as radial differences in elemental patterns that would not be evident from a single transect alone (Fig. 3). Although these topographic quantifications of otolith elements have been explored before (e.g. Gauldie et al. 1991; Tzeng et al. 1997), they are becoming more readily available with improvements in instrumentation and mapping software.

Fig. 3.  Two-dimensional map of strontium across an otolith of an Atlantic cod Gadus morhua quantified with scanning X-ray fluorescence microscopy (Limburg and Elfman 2017). The colour scale bar indicates mass fractions. © 2016 The Fisheries Society of the British Isles. Reprinted with permission.

With an expanding scope of analytical possibilities comes a persistent challenge for the otolith chemist. Elemental maps provide unprecedented insight into chemical patterns across structures, but have these advances translated into increased understanding about the life histories of these fishes? Put simply, how do we interpret these maps? A common tool that otolith chemists have used to translate chemical patterns into meaningful life history information is a partition coefficient, or discrimination coefficient, which is simply the ratio of the concentration of element E ([E]) in an otolith to the concentration in water, as follows:

This simple metric is most directly applicable when determining partitioning between an accreting solid and the immediately surrounding fluid. Although used extensively for calcifying organisms and the water they lived in, this metric has long been recognised to be a simplification that omits important intermediary transitions that can have dramatic effects on elemental uptake. If ultimately derived from seawater, elements must pass through multiple membranes and barriers before incorporation into an otolith, with possibility of discrimination or concentration at every step (Campana 1999). These barrier transitions may not be equilibrium processes, and many factors, including temperature, precipitation rate and solution composition, can affect partition coefficients in both abiogenic and biogenic precipitates. For these reasons, Morse and Bender (1990) urged workers to ‘use considerable caution in interpreting data on the composition of natural carbonates’. For fishes, a more explicit decomposition of the participation coefficient was suggested by Walther et al. (2010) to recognise the strong roles that each barrier transition may play in mediating overall uptake:

Indeed, this equation itself is insufficient given that elements may derive partially or predominantly from dietary sources. Thus, a companion equation to quantify elemental uptake from the diet would be:

Clearly, a partition coefficient is a convenient shorthand, and the potential for physiological regulation or decoupling poses considerable challenges for attempts to interpret observed chemical patterns across otolith increments. These difficulties have been recognised for some time by otolith chemists, although practitioners must continuously remind themselves of the multifaceted intrinsic and extrinsic dynamics that influence otolith chemistry. The challenge remains how to appreciate these complexities yet continue to move the field of otolith chemistry forward without stagnation. To this end, a qualitative conceptual framework using an artistic metaphor is proposed below that may hold value in approaching the challenge of otolith chemistry interpretation.

Imagine you are at a museum, wandering through galleries and admiring the artwork. You notice a painting on a wall that stops you in your tracks with its beauty and power. You have never seen this image before, but something about its use of colour and form intrigues you. You wonder to yourself: when was this painted? Who was the artist? What were they trying to say? This prompts you to learn everything you can about the artist, their life and times, and the potential messages encoded in the image that so arrested you.

This metaphor of art appreciation has value when attempting to interpret the complex chemical patterns we observe in otoliths. Superficially, the ability to analyse and produce dramatic spatial maps of elemental variability in otoliths represented by shapes and colours has obvious parallels to a painted landscape. But more deeply, this metaphor gives us a framework for appreciating the many interpretive layers that can be brought to bear when attempting to decode an image, either artistic or ichthyological. Consider the 19th century artistic movement of Impressionism, which was practiced by artists including Claude Monet, Pierre-Auguste Renoir and Camille Pissaro, among others. Evolving out of the prior paradigms of Naturalism and Realism, which emphasised faithful and even idealised representations of nature, Impressionism used novel combinations of brushstrokes and colour to not only represent the world around the artists, but also to incorporate the observer’s subjective experience (Thompson 2000). Mood, perspective, experience and the artist’s interpretation of the world around them were integrated into images that communicated more than a simple image of their environment. One of the first paintings to be considered Impressionist and an inspiration for the label itself was the painting Impression, Sunrise (1872) by Claude Monet (Fig. 4). A depiction of rowboats in the port of La Havre, the work uses bold brushstrokes that are visible swathes of paint when viewed up close but meld at distance to illustrate a familiar scene. The painting also preserves a moment in time that captures the particular light, colour and shadows unique to an early sunrise, and evokes the experience of the painter that morning. The art critic Jules-Antoine Castagnary wrote in 1874 that these types of paintings were Impressionist ‘in the sense that they render not the landscape, but the sensation produced by the landscape’ (Rubin 1999). The artist is inserting themselves as a filter through which the landscape is observed, recorded and interpreted. Impressionism explicitly integrates images of the natural world with the subjective experience of the observer and the artist, and these paintings can be analysed and appreciated through multiple lenses. This perspective has metaphorical parallels to the challenges otolith chemists face when interpreting the often complex chemical patterns that are influenced by physiological and environmental factors alike.

Fig. 4.  Impression, Sunrise [Impression, soleil levant], by Claude Monet (1872).

Critical analysis of paintings can occur at many levels, as can interpretation of otolith chemistry patterns. The first level is that of the studio, or the environment in which the painting was made. Did the artist paint outdoors or inside, from a model or a memory? More broadly, when was this painting made in time, and what were the social and political settings that influenced the artist? The second level is that of the media, or the materials used to create the image. Did the artist use oils or gouaches, a few colours or a wide palette? How did the choice of media inform the image and its representation of the subject? The third level is that of the canvas, or the surface on which the image was painted. Was the canvas primed or bare, made of linen or cotton, or perhaps made of something entirely different, like a brick wall or the inside of a limestone cave? How did the canvas composition influence the choice of media and subject matter, and vice versa? Finally, the fourth level is that of the artist. Who was this person, and what was their motivation? Were they old or young, healthy or sickly, a loner or sociable? How did their life influence the way they saw the world and the way they made paintings? Each of these levels can be thought of as a filter that stands between reality, such as a landscape, and the one that eventually ends up in the final painting. Instead of a faithful snapshot of reality, the painting is filtered through the perspective of the artist and the materials they use to represent their environments.

Thus, a painting holds more information than simply what the landscape actually looked like because it simultaneously provides insight about the artist themselves. This is precisely the same challenge that we face when interpreting chemical patterns in otoliths. The ‘studio’ is the environment in which the fish lives. This includes the geographic and physical setting, as well as the spatial variability in the elements and isotopes that become proxies for the environment. The ‘media’ are the elements themselves. Each element records its own unique set of information and, together, a suite of elements provides the palette that can be observed. The ‘canvas’ is the otolith itself, with its specific organic and inorganic composition that mediates elemental incorporation dynamics. The canvas may change significantly when investigating otolith analogues, such as scales, fin rays or vertebrae, which are increasingly popular alternative structures that have their own growth dynamics and compositional properties. Finally, but most importantly, the fish is the Artist that ultimately integrates these filters and creates an otolith reflecting its individual experience. Internal characteristics such as physiological status, maturity and stress can all act as filters that modify the eventual chemical pattern, much like an artist’s subjective perception influences the paintings they make. For the remainder of this paper, I will use this metaphorical framework to highlight a few specific advances in each of these four areas, and I conclude with a proposed conceptual approach to integrate these perspectives.

A large proportion of otolith research seeks to identify natal origins or reconstruct movements through distinct habitats. Central to most of these efforts is the assumption that ambient variability in chemical constituents will be reflected in the accreted increments of the life history stage of interest, with uptake derived primarily through a water-derived pathway. For elements or isotope ratios in otoliths that have been validated to strongly or at least partially reflect the ambient dissolved composition, geographic variability in such constituents is an ultimate control on the spatial resolution of this tool (Elsdon et al. 2008). The ‘studio’, therefore, is the chemical background through which a fish travels naturally or experiences during experimental manipulations, and the degree to which the myriad environments a fish inhabits are chemically distinct determines whether such reconstructions from otoliths are even possible.

Considerable progress in geospatial mapping of chemical variability has been essential for many fields that use composition of animal tissues to determine migration pathways. Much of this work has been driven by researchers investigating terrestrial fauna and isotopic markers such as δ¹⁸O, δD and δ¹³C in hard or soft tissues (West et al. 2010; Hobson and Koehler 2015). The creation of high-resolution maps of isotope variability, or isoscapes, requires a combination of field-collected samples and models that interpolate and predict isotope composition for unsampled locations. Animals of unknown origin can then be assayed for the marker of interest and matched to the isoscape to quantitatively estimate putative locations of origin or migration. A principle concern in these efforts is parameterising temporal variability of an isotope marker in a given location (Wunder 2010; Trueman et al. 2012). The accuracy of locational assignments requires temporal variability to be smaller than differences between locations; with large overlap in signature values between locations, the ability to confidently determine which location was the true origin decreases. For isotope ratios that are especially labile and sensitive to climatic conditions, such as δ¹⁸O, it is important to remember that these isoscapes are not static but dynamic, and time is a crucial variable (Dutton et al. 2005; Walther and Thorrold 2009). Although much of this isoscape mapping work has been targeted towards terrestrial landscapes, similar tools have been used to construct marine isoscapes, although limited spatial coverage of sampling in some ocean basins remain an issue (McMahon et al. 2013a, 2013b).

A powerful and increasingly popular isotope system for migration studies is ⁸⁷Sr/⁸⁶Sr ratios. The utility of this system is due to several attractive features, including strong geological control that leads to significant nested variation at local to continental scales (Banner 2004) and the lack of apparent fractionation during uptake, meaning water values are directly recorded in biogenic structures without physiological decoupling (Beard and Johnson 2000). Because of the dominant influence of geological composition and age on this radiogenic ratio, predictive maps based on rock age and type can be constructed for use to assign natal origins or habitat use histories in fishes (Kennedy et al. 2000, 2002; Hegg et al. 2013). A primary value of these models and maps is their ability to provide at least first-order estimations about whether certain habitats may be chemically distinguishable. Thus, these maps provide guidance on the feasibility of testing hypotheses about migration between specific locations (e.g. Humston et al. 2017). This type of a priori prediction of chemical variability is incredibly valuable given limited budgets and difficulties in sampling logistics that restrict the ability to sample water from all putative habitats before a project begins. These maps, to put it simply, reveal what questions we might ask.

Models that predict large-scale variation in ⁸⁷Sr/⁸⁶Sr ratios have become increasingly sophisticated over the past decade. Barnett-Johnson et al. (2008) predicted >90% of the variability in ratios recorded in salmonid otoliths in the western US with a model based on the proportion of the watershed containing granitic bedrock. Subsequently, Bataille and Bowen (2012) developed a model of bedrock lithology for silicates and carbonates as well as age to estimate rock-specific radiogenic decay of ⁸⁷Rb to ⁸⁷Sr from parent rock compositions, which were combined with geographic information system (GIS) geology maps to estimate ⁸⁷Sr/⁸⁶Sr variation across the contiguous US. This model then incorporated weathering coefficients for specific rock types combined with hydrological flow measurements to estimate local catchment compositions for specific regions. The flux-weighted models that incorporated bedrock type, age and weathering had a high degree of accuracy (~70%) in predicting observed ⁸⁷Sr/⁸⁶Sr ratios in specific regions. This model was further improved by Bataille et al. (2014) through refinements of model components including siliciclastic sedimentary rock submodel improvements to account for grain recycling, and an updated weathering model for Alaska that incorporated permafrost and glacial processes that can affect Sr flux at higher latitudes. These models point to the need to incorporate both general isotope systematics as well as regional hydrodynamic processes to more accurately estimate the dissolved ⁸⁷Sr/⁸⁶Sr present in streams and rivers that fish inhabit, thereby allowing highly accurate estimation of natal origins and migration patterns of freshwater and diadromous fishes (Brennan and Schindler 2017).

For fish that transit salinity gradients, a central concern is understanding how targeted elements vary across an estuary. The theory and practice of hydrological mixing dynamics are of paramount concern to discern whether movements between salinity regimes are even detectable. Although the pertinent concepts about mixing have been discussed previously (e.g. Kraus and Secor 2004; Gillanders 2005; Milton and Chenery 2005; Walther and Limburg 2012), they are worth revisiting in the context of understanding the ‘studio’ in which otoliths are formed. First, when considering differences between fresh and marine chemical signatures, the chemical compositions of both water masses on either end of a mixing dynamic, or endmembers, must be known. For example, the larger the difference between fresh and marine Sr concentrations, the greater the likelihood of detecting a movement between those habitats. For an element like Sr, where marine concentrations are relatively homogeneous and constant on ecological time scales compared to fresh waters (de Villiers 1999), it is the freshwater endmember that is of most concern for otolith chemists. The strong geological control on freshwater Sr composition therefore requires one to inspect a geological map of the catchment to investigate the bedrock composition of the drainages. Given significant variation in Sr composition among lithologies, some systems, such as those dominated by marine carbonates, may have very similar or nearly identical compositions to that of marine waters (Brown and Severin 2009). With limited differences in the selected chemical composition between fresh and marine endmembers, there is little scope for detecting movement in those systems. This ultimate environmental control on the utility of chemical tracking for a particular system thus requires the researcher to carefully consider the regional geological setting before selecting otolith chemistry as a viable tool.

Beyond considering the endmembers themselves, the mixing dynamics across the salinity gradient are also of great interest. A dissolved element is considered conservative if it exhibits linear mixing between two water masses, such as fresh and marine endmembers. Non-conservative mixing is indicated by positive or negative curvature in the relationship between the element and salinity, which results when there are internal kinetics that result in net addition or removal of the constituent at an intermediate salinity (Ward and Montague 1996). Note that a linear relationship with salinity does not mean that there is an absence of kinetics for the constituent in the estuary, only that there is no net kinetic addition or removal at mid-salinities. A relevant example of this is dissolved Ba, which exhibits net addition at low salinities where riverine particles containing adsorbed Ba first hit the salt wedge and Ba ions are desorbed and released into the water column (Coffey et al. 1997). This low salinity peak in Ba is observed in most estuaries worldwide, with the magnitude of the peak depending on the particulate load and geological composition of the catchment (Sinclair and McCulloch 2004).

However, the ability to determine whether an element is conservative or non-conservative based on linear or non-linear behaviour requires that specific set of assumptions are met. If one or more of those assumptions are violated, non-linear relationships between the element and salinity may be observed even if the element itself is still behaving conservatively. Critical assumptions include: (1) there are only two endmembers (violated if there are multiple streams or even groundwater contributions with distinct chemical compositions all contributing to the estuary); (2) that salinity itself is conservative (violated if strong evaporative processes cause alterations in salinity that are not due to mixing); and (3) that the endmember compositions are temporally stable with respect to the residence time of the estuary (violated if short-term flood or drought regimes alter upstream weathering of unique lithologies, for example). Thus, the observation of non-linear behaviour can, on the one hand, indicate non-conservative processes, but may indicate on the other hand that one or more assumptions have been violated. Why is this of concern for otolith chemists? The shape and scope of the mixing curves for these tracer elements are again the ultimate control dictating whether movements are discernible. The hydrodynamics and elemental kinetics of the system are useful to identify whether movement between specific salinity regimes is resolvable, and they may also indicate important ecological considerations, such as the contribution of multiple catchments (and therefore potential habitats for migration) or the dominance of evaporative processes that could induce salinity stress for stationary fishes (Barnett-Johnson et al. 2008; Gillanders and Munro 2012; Hegg et al. 2015; Mohan and Walther 2015). Careful consideration of the environmental processes and their controls on constituent behaviour is key to the successful use and interpretation of otolith chemistry patterns as a movement tracking tool.

Turning to the media, or the chemical constituents that may capture some aspect of the life of a fish, we must identify the palette of elements and isotopes available for observation. A small handful of elements has dominated the literature, in part because certain elements are readily incorporated into the aragonitic crystal lattice, they may pass through Ca channels and substitute directly for Ca, thereby more effectively representing the ambient water composition, and because of their relative abundance in otoliths, rendering them reliably assayed with common instrumentation. Thus, the alkaline earth metals isolated by Sir Humphry Davy, such as Sr and Ba, are of great utility. In addition, the isotopic ratios of C and O are readily quantified in any carbonate structure and useful given the great deal of work on fractionation and isotope systematics in other aragonitic skeletons, such as those of scleractinian corals. The other key to a useful ‘colour’ is that the element or isotope ratio varies environmentally with a gradient such as salinity or responds to another variable of interest like temperature or dissolved oxygen (Radtke 1989; Thorrold et al. 1997; Elsdon and Gillanders 2002; Limburg et al. 2015). These required characteristics naturally winnow to the spectrum of potentially useful elements down to a handful, which are typically sufficient for many questions and systems.

However, with advances in instrumental sensitivities and precisions and expanding experimental validations to understand uptake dynamics of other elements, our palette may expand. Less frequently investigated elements such as Fe, Pb, Cu, Zn and S (bulk and isotope ratios) are detectable in otoliths, although analytical challenges for quantifying their concentrations persist (Spencer et al. 2000; Limburg and Elfman 2010; Daverat et al. 2012; Di Franco et al. 2014; Hüssy et al. 2016; Doubleday et al. 2018). The degree to which some of these elements reflect ambient water compositions or are highly regulated by physiological discrimination is poorly known (but see Geffen et al. (1998) and Milton and Chenery (2001b), among others). Experimental validations of these alternative elements are rare, due in part to the difficulty in rearing fish in carefully controlled elemental environments with minimal contamination of trace metals such as Fe and Pb during rearing or analysis (Arslan and Secor 2008; Selleslagh et al. 2016). Yet, we must encourage new otolith chemists to devise ways to experimentally manipulate water and diets of these alternative elements in an attempt to widen the available palette beyond the now-traditional elements.

A key feature of an otolith is that it is essentially a mixed-media artwork. Rather than simply an abiotic accretion, the structure is biogenic and created with a complex protein lattice that forms the scaffolding on which the carbonate crystals accrete (Murayama et al. 2005; Miller et al. 2006; Weigele et al. 2016). The nature and composition of the organic fraction of an otolith is still being revealed, posing both challenges and opportunities for interpretations of chemical patterns. The degree to which proteins bind and alter elemental availability for transport across membranes or incorporation into the crystal structure itself contributes greatly to the ability of an element to accurately reflect ambient water compositions (Payan et al. 1999, 2004a, 2004b; Borelli et al. 2001, 2003a, 2003b). A careful consideration of protein–element interactions at all stages along the relevant physiological pathways is important. For example, Thomas et al. (2017) found that although the dissolved fraction of Ca, Sr and Ba in endolymph is large, a significant proportion of those same elements is protein bound and thus potentially unavailable for incorporation into the otolith. Within the otolith itself, the assumption that elements substitute for Ca in the carbonate crystal may not be uniformly accurate, and the protein lattice itself may facilitate the uptake of some elements (McFadden et al. 2016). Izzo et al. (2016) found that the fraction of protein-bound elements within otoliths varied, with minimal proportions of Ba bound to proteins but larger proportions of protein-bound Mn and Cu. These element-specific protein interactions contribute, in part, to their respective partition coefficients and illustrate the complexities of attempting to interpret elemental patterns in these mixed-media structures.

In addition to affecting uptake dynamics of traditional elements, the protein matrix itself can be assayed for vital information about the life of a fish. Although low in total concentration, the proteins offer the potential for insight into dietary dynamics of fishes that cannot be probed effectively with elements such as Sr and Ba. The primary limitation for assaying the organic fraction has been purifying sufficient material for analysis. However, new methods for isolating and quantifying proteins show great promise. For example, Lueders-Dumont et al. (2018) found that otolith-bound δ¹⁵N distinguished between farmed and wild salmonids, and Sirot et al. (2017) used otolith δ¹⁵N and δ¹³C together to investigate the diets of both modern and archived otoliths. Beyond just bulk analysis, the ability to quantify compound-specific amino acid stable isotope ratios is a frontier with much potential. The advantage of this compound-specific approach is the use of certain amino acids that undergo minimal trophic fractionation (‘essential amino acids’) to estimate baseline carbon fixation sources and other highly fractionated amino acids to estimate trophic position (McClelland and Montoya 2002; Larsen et al. 2013; Nielsen et al. 2015). The ground-breaking work by McMahon et al. (2011a, 2011b, 2016) showed that this approach was possible in otoliths, and δ¹³C values of amino acids effectively distinguished fish who fed on mangrove or seagrass food webs. More recently, Vane et al. (2018) used δ¹⁵N of ‘trophic’ and ‘source’ amino acids to estimate ontogenetic trophic shifts of mobile fishes. This compound-specific approach is likely to become more frequently used as methods and instrumentation become more accessible. Truly, the organic fraction of otoliths is a fruitful area of study not only to increase our understanding of elemental incorporation dynamics, but also to open the door to additional questions about dietary and metabolic histories contained in the proteins themselves.

The canvas refers to the structure on which the chemical patterns are created. As described above, an otolith is composed primarily of calcium carbonate but with a small but significant fraction of organic matter, thereby influencing the choice of elemental and isotopic proxies that can be investigated. The properties of otoliths, including their acellular structure and lack of metabolic reworking, are fundamental to our ability to interpret patterns as time-stamped reflections of a particular period in the life of a fish (Campana and Thorrold 2001). Although the primary form of most otoliths is aragonite, other crystal forms may be present in particular species, such as vaterite and calcite, which have important implications for the relative abundance and incorporation dynamics of particular elements (Tzeng et al. 2007; Pracheil et al. 2017). Thus, the properties of an otolith are paramount for dictating how and which chemical proxies are incorporated and effectively interpreted. Parallel with the growth in the field of otolith chemistry has been the rapid development of alternative and complementary chemical records in other structures, such as scales, fin spines and rays, eye lenses and vertebrae. These ‘otolith analogues’ have their own set of challenges and opportunities, largely dictated by the relative proportion of calcified and uncalcified material, the degree of metabolic stability and the periodicity of increment formation (for a review, see Tzadik et al. 2017). Workers investigating these alternative structures tend to come from a background of either soft tissue stable isotope research or otolith chemistry, with concomitant focus on one or the other set of proxies in these structures, although work using combinations of proxies may become more frequent.

Scales offer the potential for paired organic isotope assays with Sr/Ca and Ba/Ca measurements with less limitations of sample size as encountered with otoliths (Woodcock and Walther 2014; Seeley et al. 2015). However, their overlain bipartite structure renders the interpretation of interior increments difficult for many species (Hutchinson and Trueman 2006; Trueman and Moore 2007), resorption may occur and the periodicity of scale increment formation may be irregular or decoupled from that in otoliths for some species (Campana 2001; Abecasis et al. 2008; Upton et al. 2012). Furthermore, workers must be cautious to exclude regenerated scales if they wish to obtain a more complete life history record (Seeley et al. 2017), and scales only begin formation at squamation, which means they are typically useless for determining larval origins. Fin spines and rays offer similar potential for assays of organic stable isotope ratios, although with comparable uncertainties about increment periodicity, metabolic reworking and record completeness. A primary reason to use structures such as fin spines or scales is the non-lethality of sampling, which may be required for work with imperilled species or catch-and-release fisheries.

Other structures that are sampled lethally include eye lenses and vertebrae. Vertebrae have been targeted by those investigating migrations and the life history of sharks, where otoliths are not available for analysis (Raoult et al. 2016; Smith et al. 2016; McMillan et al. 2017). One intriguing and exciting reason to investigate vertebrae in these organisms is the presence of material accreted before birth, potentially providing information about in utero growth, and therefore maternal diets and movements (Carlisle et al. 2015). Although, as with the other structures, the potential for metabolic instability and lack of experimental validation of uptake and turnover for many of these long-lived and large species is a continued impediment (but see Werry et al. 2011; Smith et al. 2013). Eye lens composition is an emerging technique that uses separated laminae for individual analyses of stable isotopes or elemental composition (Wallace et al. 2014; Quaeck-Davies et al. 2018). For all these structures, continued investigation into transport, incorporation and stability of chemical signatures is key. One potential value-added way to rapidly advance this field is for researchers who conduct experiments on otolith chemical uptake dynamics to retain as many tissues as possible (vertebrae, eye lenses, scales etc.) from the fish at the conclusion of the experiment. Although costly to analyse multiple structures, I encourage researchers to reach out to one another to attempt to coordinate analyses among multiple laboratories to maximise outputs. Importantly, this multilaboratory effort could allow comparisons of patterns in the same element across multiple structures, which is sorely needed in this field.

Finally, for otolith chemists attempting to understand elemental uptake dynamics into these carbonate structures, we must stay familiar with the advances in other biogenic carbonate research, including work on scleractinian corals, foraminifera and sclerosponges. Certainly, the taxonomic and physiological constraints are quite different in these disparate groups, but theoretical and methodological advances in coral sclerochronology, for example, could provide a framework for modelling and experimental approaches to understanding comparable processes in fishes. Researchers in those fields are engaged in similar attempts to understand how physiological processes are regulating elemental movement across membranes to the calcification surface and to what extent those transport dynamics may influence a proxy such as the Sr/Ca ratio and temperature relationship. Current debate revolves around the extent to which Rayleigh fractionation in partially closed calcification systems is also mediated by Ca²⁺-ATPase activity, which itself responds to environmental conditions such as temperature, pH, aragonite saturation state and elemental concentrations (Gaetani et al. 2011; Thien et al. 2014; DeCarlo et al. 2015; Tanaka et al. 2015; Giri et al. 2018). Combined models that incorporate direct environmental and vital effects on elemental incorporation have been developed (e.g. DeCarlo et al. 2016), which could be usefully adapted and modified for fish physiologies and provide a framework for future investigation and experimentation. Thus, keeping pace with advances in analogous fields that are investigating alternative ‘canvases’ could allow otolith chemists to make great strides in understanding the interplay between biogenic and abiogenic factors without starting from scratch.

Finally, we come to the ultimate integrator of extrinsic and intrinsic drivers and the creator of the otolith itself: the fish. As the artist responsible for the patterns we observe in these structures, the ‘perspective’ of each fish must be considered when interpreting chemical signatures recorded in otoliths. Although it is objectively obvious that the physiology and experience of the fish is important, it is all too easy to think of an otolith as a simple recorder of the environment with no filtering of the ambient signal. A survey of the otolith chemistry literature will find numerous statements to the effect that ‘elements in water are recorded in otoliths’ as a justification for using the approach to study movement dynamics. This language is imprecise in a few ways. First, it implies passive and direct incorporation of elements without any physiological modification (either concentration or discrimination) during the various pathways from water to otolith. Second, it assumes water is the only source of otolith elements, which may be incorrect for some constituents. Third, it collectively lumps the uptake of all elements into a single monolithic group, implying that transport dynamics are similar for any element found in an otolith. Although some elements do reflect ambient water composition, as modified by physiology, temperature and other factors, other elements (e.g. Mg) have not yet been found to accurately reflect water composition and may more closely reflect metabolism (Woodcock et al. 2012; Limburg et al. 2018). This language is of course a shorthand and could be forgiven because not every publication on otolith chemistry requires a full exposition of the complexities of uptake in the Introduction of a paper. But we must be cautious to not let literary shorthand devolve into faulty assumptions. Students and new practitioners of otolith chemistry must be especially careful to remind themselves of these assumptions and do their best to use accurate phraseology.

As mentioned throughout this paper, the myriad intrinsic factors that can potentially regulate or even decouple otolith elemental composition from that in the environment are complex and, in some cases, poorly understood. Metaphorically, the fish as an artist creates its artwork that is both constrained and enabled by its health, condition, behaviour and environment. Taxonomic groups with divergent physiologies could be viewed as analogous to artistic movements with recognisable styles. Impressionists, Surrealists and Modernists produce identifiably different artworks, much like Istiophoridae, Salmonidae and Megalopidae produce distinct otoliths. Although these issues have been recognised throughout the history of this field, much work remains to be done to disentangle the relative effects of these processes. A focus on ionic transport of elements across membranes is critical (Hirose et al. 2003). Many of the central elements in our palette, such as Sr and Ba, are assumed to mimic the behaviour of Ca and move through Ca ion channels. If these ions do follow such transport pathways, then calcium homeostasis and the hormonal and other pertinent regulatory processes are of critical concern for understanding physiological mediation of ion behaviour. As is well known in the physiological literature, maintenance of homeostasis will be affected by a wide variety of physiological and life history perturbations, including stress, circadian rhythms, vitellogenesis, reproduction, osmoregulation and metamorphosis (Marshall and Grosell 2006). Sex and reproductive status are potentially quite influential, particularly for marine fish, which live in more chemically homogeneous environments (Sturrock et al. 2014, 2015), but these parameters are not always recorded in wild-collected fish samples (exceptions include Edmonds et al. 1991; Campana et al. 1994; Gauldie 1996; Secor and Piccoli 1996; Milton and Chenery 2001a; Ashford et al. 2006; Brenkman et al. 2007; Han and Tzeng 2007; Jónsdóttir et al. 2007; Arkhipkin et al. 2009; Albuquerque et al. 2012). Of particular relevance to those tracking diadromous movements is the altered ion regulation requirements for fish inhabiting fresh or marine environments where they are hyperosmotic or hypo-osmotic respectively. Elemental partition coefficients are known to vary with salinity (Elsdon and Gillanders 2003; Gillanders and Munro 2012; Panfili et al. 2015), and ion regulation may be a partial explanation for this phenomenon.

Together, the importance of ion transport processes points to a continued need for experimentation that explicitly integrates physiological and environmental perspectives to understand uptake. Here, otolith chemists may find useful collaboration with physiologists already heavily engaged in ion transport dynamics, particularly because that field has grappled with the effects of aquatic acidification on stress, homeostasis, ion transport and carbonate accretion responses (Pörtner 2008; Heuer and Grosell 2014; Esbaugh 2018). Complementary approaches to quantifying environmental experiences, such as archival tagging technologies, can also be usefully coupled with otolith chemistry analyses to disentangle extrinsic forcing on observed chemical patterns (Darnaude et al. 2014; Darnaude and Hunter 2018). Further, meta-analyses of the existing wide range of experimental evaluations of factors influencing elemental uptake in otoliths will be critical for highlighting gaps where future efforts can be focused (Izzo et al. 2018). Much as a painting is created by an artist with an idiosyncratic set of experiences and perspectives, an otolith does not form spontaneously in water, but rather is created by a fish with its own physiological dynamics and experiential history that mediate and influence the observed patterns incorporated into the structure itself.

The continued challenge for practitioners of otolith chemistry is to bear these multifarious and interactive extrinsic and intrinsic factors in mind when attempting to interpret a chemical pattern for a specific question. How to reconcile these issues while still deriving meaningful information from these structures? One potential approach, at least conceptually, borrows terminology from the field of signal processing. Specifically, we may consider an otolith chemical pattern to be similar to an acoustic waveform or captured image, with the target signal of interest embedded with additional information, or noise, that obscures its resolution. Of central interest, then, is the signal-to-noise ratio (SNR), which dictates whether our signal has a chance of being resolved from the ‘background’ of other dynamic factors.

Consider the case of attempting to identify a fish movement across a salinity gradient with otolith chemistry (Fig. 5). The expectation is that a tracer such as Sr/Ca varies substantially across the salinity gradient with the freshwater endmember identifiably lower than the marine endmember (Fig. 5a). However, as discussed above, this assumption must be validated and a catchment with a limestone-dominated bedrock may have elevated freshwater Sr/Ca and therefore a minimal difference between endmembers (Fig. 5b). These two scenarios impose very different constraints on the ability to detect a movement from fresh to marine waters, particularly given other factors that include additional variation on the final otolith pattern (depicted here as random noise, but which could also manifest as systematic or non-random alterations). If the magnitude of these alternative influences is moderate, then a movement across an environmental backdrop of a large difference between endmembers should still be resolvable. In that scenario, the SNR is high (Fig. 5c). However, with reduced endmember difference, the noise may be equivalent in magnitude to the signal, the SNR is low and the movement is more difficult to detect (Fig. 5d). Of course, the categorisation of particular factors as signal or noise depends on the question being asked. An investigator may identify a physiological transition, such as reproductive status, as the desired signal, in which case chemical variability induced by movements across environmental gradients would be the noise for that study.

Fig. 5.  Signal to noise example. Freshwater endmembers of Sr/Ca may be (a) significantly lower (2.5 mmol mol⁻¹) or (b) more similar (6.5 mmol mol⁻¹) to the marine endmember (8.5 mmol mol⁻¹). In these scenarios, a movement between habitats may be readily distinguished despite physiological ‘noise’ when the signal-to-noise ratio (SNR) is high (c) or obscured when the SNR is low (d). In both scenarios, a movement occurs at 1000 µm from the otolith core.

In order to implement this type of SNR approach, the relative magnitudes of pertinent factors must be known at least to a first-order approximation. If magnitudes of the effects of intrinsic and non-target extrinsic factors are known, then the magnitude of the required chemical gradient to impart detectable signals of movement could be estimated. This estimation may be possible for some taxonomic groups given current experimental results. Species-specific or stock-specific differences in uptake dynamics may challenge our ability to extrapolate these results to species where experimental work is limited (Clarke et al. 2011; Barnes and Gillanders 2013; Chang and Geffen 2013). However, multi-elemental comparisons among proxies that have varying degrees of physiological control is a promising way to identify the relative effects of intrinsic and extrinsic factors (Grammer et al. 2017). This SNR framework could be useful in pointing towards gaps in experimental knowledge and guide new practitioners towards essential preliminary steps when evaluating whether otolith chemistry will be a viable tool.

Returning to the artistic metaphor, we must consider the museum patron who enjoys and analyses the art in front of them. This patron is the scientist attempting to understand something about the life and times of fishes. Much like the art historians that now use advanced image processing and even elemental analyses to establish provenance and creation of important artworks (Marin et al. 2015; Cucci et al. 2016), otolith chemistry tools have grown in sophistication and accessibility as instrumentation has advanced. Yet, technological advances do not eliminate the possibility that the observer is misled when interpreting a pattern. Observer bias is an ever-present issue in biological studies (Holman et al. 2015; Kardish et al. 2015), and otolith chemists must caution themselves about the possibility that they are biased towards observing particular types of patterns, or they may simply not recognise patterns that are present. Scientists are the final filter through which the data are passed before an interpretation is produced.

Despite these many challenges inherent to the interpretation of otolith chemistry patterns, the field has grown dramatically over recent decades and afforded great insight into the life histories, population dynamics and habitat requirements of a diverse range of species. An otolith is truly an integrated portrait of the life of a fish, including information about the environment, its physiology and growth patterns in one singular structure. Rather than turning a blind eye to these complicated issues, our field has flourished by embracing this complexity and applying innovative experimentation and adoption of technological improvements to disentangle these dynamics. By reminding ourselves that an otolith reflects not just the external environment, but also the singular point of view of each fish, we can continue to interpret these complicated patterns. We use otoliths to appreciate not just the landscape imprinted in the image, but also the perspective of the artist. As visitors to these aquatic galleries, we are truly lucky to have such rich and rewarding images to contemplate.

The author declares that he has no conflicts of interest.

This research did not receive any specific funding.