Diet analysis is one of the most common methods in trophodynamic study along with direct observation of predator prey-relationships (Choat et al., 2004; Fox et al., 2009; Kramer et al., 2013; Young et al., 2015; Wen et al., 2016). Stomach contents can directly inform what a consumer ingests by examining the frequency of occurrence of species in stomach samples (Cortés, 1999). This method can work well to uncover trophic interactions for commonly occurring and abundant species. However, there is often uncertainty in prey identification and metabolic requirements of a consumer (Young et al., 2015). For species of conservation concern, sample sizes for study can be low as there are limitations to lethal sampling. Non-lethal methods such as gastric lavage show what has been recently consumed (Cortés, 1997; Frisch et al., 2016), but may be unreliable for determining the full scope of the diet of an individual. For larger species such as sharks, angling may also induce gastric emptying before landing (Frisch et al., 2016), which can affect results. In fact, many studies using gastric lavage and baited capture report high proportions of empty stomachs in predators. Due to these limitations, biochemical tracers, and immunological testing through molecular identification of prey (Symondson, 2002) have been used to supplement stomach content studies. Taxonomic molecular barcoding can supplement gut content study and assist in characterization of diet (Paula et al., 2016). Molecular barcodes designed to identify diet specialization on invasive macroalgae by coral reef grazers has uncovered trophic links previously unknown in herbivorous fishes in Hawai'i (Stamoulis et al., 2017). Combination of multiple methods in diet analysis can identify trophic position and food web links for trophodynamics, but care is needed when using these methods to identify ecological roles. Habitat is known to be an important consideration in most ecological studies, and it is mentioned most frequently in studies examining stomach contents (98 percent, Table 2). Yet even recent studies have ignored potential effects of habitat by combining samples across regions without understanding community composition and available prey. Spatial scale and habitat variables are crucial elements to defining diet analysis in trophodynamics research. Seasonal (i.e., temporal) variability in prey abundance or availability must also be considered. While these aspects are present in some studies, they must be more widely applied to help define ecological processes and trophodynamics in coral reefs and other marine habitats.

Biochemical tracers identify dietary sources in a food web and what individuals directly consume (Post, 2002; Young et al., 2015). Tracers quantitatively measure assimilation through tissues to define the diet of an individual without the uncertainty of species identification of stomach content studies (Post, 2002; Layman et al., 2007). Nitrogen and carbon isotopic enrichment in bulk tissue are examined through stable isotope analysis to define the relationship between a predator and its prey (Post, 2002; Layman et al., 2007, 2012; McClanahan et al., 2011). For producers, carbon isotope proportions vary due to different methods of energy production through photosynthesis (Michener et al., 2007). Carbon (δ¹³C) is used to identify source pools as there is limited variation in values after trophic transfer, while nitrogen (δ¹⁵N) is used to identify trophic position as enrichment is progressive across trophic levels (Kolasinski et al., 2016). Spatiotemporal comparison of diet is also possible with fatty acid analysis. When a predator consumes a species, fatty acids are absorbed with little change in the unique signature, meaning that prey can be identified from adipose tissue and blood samples (Budge et al., 2006). Without a known database of fatty acid profiles, these studies can be limiting. For biochemical tracers, food web limitations mainly come from time and resources available to define an appropriate ecological baseline. Without understanding carbon sources, studies may lack context. For example, carbon flow is challenging to elucidate in coral reefs due to multiple end-members feeding from both benthic and pelagic sources. Application of compound specific amino acid analysis (AA-CSIA) and Bayesian mixing-model techniques can assist in defining these baselines (McMahon et al., 2015). AA-CSIA can provide better baselines by looking at specific amino acids that fractionate with each trophic step (Bradley et al., 2016). AA-CSIA can also reduce the number of samples for analysis by eliminating the need for an exhaustive baseline. But tissue turnover rates, functional redundancy, and complications in retention of nitrogen in certain organisms can also limit analysis (Post, 2002; Layman et al., 2007, 2012; McClanahan et al., 2011). Without species-specific study, improved techniques, and time-series baselines, it is difficult to get a true estimate of trophic position (Layman et al., 2012). Recently, Matley et al. (2016) showed that diet tissue determination factors and turnover rates measured for temperate species may not apply to tropical species, as δ¹⁵N enrichment was lower than published values for a slow-growing predatory reef fish. In larger predatory species such as sharks, care in sample preparation is also necessary as urea retention without extraction can severely affect values, which is often not a common method employed for stable isotope analysis (Li et al., 2016). The physical and biogeochemical nature of a habitat can also have large effects on trophic community structure even at a fine scale. For example, Kolasinski et al. (2016) studied macro-invertebrate communities of coral reefs and found significant temporal and spatial differences in food web lengths indicating a variety of energy pathways which complicate trophic level definition. Therefore, consideration should be given to understanding the data sources that may limit research to specific habitats or time periods. Recent studies use large-scale spatial and temporal differences in sample collection as caveats in the discussion (Frisch et al., 2016), as opposed to incorporation into the study design. From reviewed publications, only 42 percent of stable isotope analyses included a reference to carbon sources, and 21 percent for fatty acid analyses; 50 percent of studies used a combination of stable isotopes and stomach contents, while 25 percent used a combination of fatty acid and stomach contents (Table 2). Stable isotopes contributed to 58 percent of studies that assessed trophic levels of organisms and have been used in 39 percent of Ecopath studies. Refined analysis of trophic levels through biochemical tracers is likely to strengthen and improve trophodynamics research as applications of these methods expand.

Statistical models are widely used in ecological studies to understand energy flow and provide a whole of system interpretation. Due to the complexities of marine food webs, trophic relationships are often used to categorize interactions between trophic levels and among individuals within models (Bozec et al., 2005). While linear models often show patterns in specific relationships from quantitative study, they are not robust enough to handle complex food webs such as those on coral reefs (Evans et al., 2013). Dynamic models are required to describe ecosystem organization through networks of species that interact (Liu et al., 2009). In trophodynamic studies, species abundance and diversity are typically categorized by functional groups where diet and habitat similarities are shared by multiple species. This allows simplification of models to a tractable level where input parameters are determined by functional group and often include known values of productivity, biomass, and metabolic efficiency.

While there are multiple dynamic food web models available, Ecopath (Polovina, 1984) is the most commonly used for marine and aquatic study (Heymans et al., 2016) and has been applied to coral reefs. Under the assumption of energy equilibrium, Ecopath uses mass balance equations to determine trophic relationships of biomass and productivity (Polovina, 1984). Ecopath can be fitted for time series with Ecosim (EwE) (Gotelli and Entsminger, 2004) and also account for spatial differentiation of habitat with Ecospace (Opschoor, 1995). While Ecopath is a popular method, few studies using this model in marine research include time series and spatial data, and even fewer are applied to coral reef ecosystems (Heymans et al., 2016).

For Ecopath to be effective, thermodynamics and ecological knowledge behind fitting a model are required to prevent misuse (Evans et al., 2013; Young et al., 2015). Thirty-nine percent of reviewed Ecopath studies (Table 2) for coral reefs used Fishbase (Froese and Pauly, 2000) as a source of data for the model. While this resource is a reputable data repository with relevant diet and productivity information, there is considerable uncertainty in empirical data and regional species variation. Without understanding ontogenetic diet change, temporal and biological shifts in behavior and foraging of a species, responses to disturbance may be poorly interpreted (Young et al., 2015). Even if an organism has a highly specialized diet and small home range, external stressors and biogeography vary (Sale, 2002). Ecopath relies upon the strength of interactions between predators and prey, and requires extensive knowledge of each species fitted into the model. While these models take diet into account, predator avoidance is less well-described and parameters such as refuge availability and niche occupation are difficult to fit into a model. For mobile, large predators, effects on lower trophic groups are poorly understood as top-down forces often result in diffuse predation that is difficult to characterize with Ecopath parameters. Predator-prey structure in ecological models was recently examined and researchers found that more complex models do not necessarily invalidate predicted behaviors in simpler models, but care must be used when employing the latter (Walters et al., 2016). The authors showed how assumptions of feeding rates of predators, prey availability and other factors drastically change isocline patterns in predator-prey models. They note that uniform spatial assumptions should be avoided when using trophic models. Of the reviewed studies, habitat was referenced in 89 percent of Ecopath studies focused on coral reefs (Table 2), but many were not specific to a single area and assumed homogenous habitat across regional scales. Testing underlying assumptions is critical when employing predator-prey models (Heymans et al., 2016), and understanding interactions at an appropriate spatial scale will assist in increasing the efficacy of these models.

While some scientists have urged that more process-based ecological models be developed to identify the underlying mechanistic behavior of an ecosystem (Evans et al., 2013; Turney and Buddle, 2016), recent articles still often ignore dynamics of spatial and temporal variability as well as historical baselines (Lamy et al., 2016). Some authors have moved past dynamic models such as Ecopath and explored individual-based predator-prey population models to account for spatiotemporal heterogeneity (Thierry et al., 2015). Others have tested vulnerability of reef ecosystems by measuring overall entropy from bioenergetics (Arreguín-Sánchez and Ruiz-Barreiro, 2014), interaction strength between trophic levels (Bascompte et al., 2005) and applied network analysis (Navia et al., 2016). Integration of unified models may also be beneficial. While unified models often are used to explain theories of biodiversity, recent developments of metacommunity analysis between labrid fishes and corals have identified patterns of species associations (Connolly et al., 2017). Species associations are critical in integration of trophodynamics into management. Critics of food web models for coral reefs highlight that any model will be a simplified version of real-time processes and caution should be taken when using them to inform management. Although management is mentioned in 98 percent of reviewed publications (Table 2), few articles using Ecopath in marine environments have been used for management purposes (Heymans et al., 2016). To better inform management, research is needed on spatiotemporal variability to better fit models. As Ecopath provides an informed snapshot of behavior of a system, including variability over space and time may assist in reducing uncertainty.

Acoustic telemetry is widely used in aquatic ecology, but rarely incorporated into trophodynamic studies. Telemetry can be used to better understand the distribution, residency, and behavioral patterns of species and applied to understanding how predators interact with prey (McCauley et al., 2012; Young et al., 2015; Matley et al., 2016). Matley et al. (2016) determined that although two species of co-occurring reef fish had overlapping diets, space use differed between the two species suggesting niche separation. Studying movement can also show how behavior may be affected by environmental stressors and conditions. Telemetry can also be used to supplement other methods to identify behavioral adaptions to resources such as targeting invasive species (Bierwagen et al., 2017). Improved analytical methods for ecology such as network analysis (Espinoza et al., 2015), Bayesian statistics (Johnson et al., 2010) and state-space models (Jonsen et al., 2005) are assisting predictive capability of movement relative to environmental variability. For example, telemetry has helped define population dynamics through focused mark-recapture models (Dudgeon et al., 2015). Other applications such as identifying feeding patterns from movement and prey associations that may influence food web dynamics may play a role in future trophodynamic analyses. Despite the advantages of telemetry, the time and financial investment required may limit sample sizes and application to trophodynamics (Young et al., 2015). Telemetry was used in less than 10 percent of reviewed studies and less than 20 percent in combination with other empirical methods (Table 2).

Habitat and community structure are well-documented to be important variables in ecosystem processes, particularly when examining trophodynamics. Yet, scientists often take a “one size fits all approach” to food-web studies. Dornelas et al. (2006) demonstrated that coral reefs cannot be explained by widely-accepted theories of biodiversity, which suggests that spatiotemporal stochasticity observed in coral reefs is not well-understood. This validates the need to understand process based interactions. Most trophodynamic studies of coral reefs come from the Caribbean and Great Barrier Reef, which cover large spatial scales and latitudinal gradients. While biogeographic variability is constantly acknowledged, publications link conservation concern to broad topics such as management zones irrespective of reef variability or geographic position (Frisch et al., 2014; Rizzari et al., 2014; Boaden and Kingsford, 2015), and perform large scale analyses based on databases that do not account for fine-scale variation (Campbell and Pardede, 2006; Graham et al., 2008; Campbell et al., 2011; Barneche et al., 2014; Alonso et al., 2015; Aguilar-Medrano and Barber, 2016). While management zones are important to understanding human disturbance, many studies exploring fishing effects lack historical baselines, movement data, and diet relative to changing diversity and community structure (Greenwood et al., 2010; Edgar et al., 2011). Additionally, the influence of natural cycles is rarely considered (Kruse et al., 2016) although they can play an important role in ecosystem function. A review by Bijoux et al. (2013) discussed the effect of natural cycles (diurnal, tidal, lunar, and seasonal variation) on fish movement where the authors found that studies ignoring natural cycles increased unexplained variation in the data thereby reducing their effectiveness in defining ecosystem processes.

In addition to lacking spatiotemporal context, researchers may be misinterpreting controlling forces in coral reefs. Across all ecosystems, biomass pyramids and the role of trophic subsidies are not well-understood (Trebilco et al., 2013). Top-down vs. bottom-up organization of coral reefs is debated in the field, particularly in regard to the focus of conservation efforts (herbivore focused or predator focused). Pyramids are the primary structure used to argue resource control, but may oversimplify the unique and fine-scale interactions that occur at the reef level. Marine food webs have been described as “ecological road maps” and the varying networks that link different trophic levels in coral reefs could explain their resilience to cascading effects (Turney and Buddle, 2016). In coral reefs, the semi-open nature of communities increases the difficulty in mapping food webs from source to consumer and the influence mesopredators may have on lower trophic levels. Trophic cascades were mentioned in 24 percent of reviewed publications, yet there is little empirical evidence documenting cascading effects in coral reefs. Even in the few cases documenting trophic cascades, the results are inconclusive. Estes et al. (2011) used the Northern Line Islands as an example of trophic cascade, but the original research by Sandin et al. (2008) showed no evidence of trophic cascade through use of principle component analysis. The work by Sandin et al. (2008) reflects conclusions made by Bascompte et al. (2005) who showed strength of trophic interactions on coral reefs buffer cascading effects but become weaker under pressure of external perturbations such as overfishing. This does not mean that cascading effects do not exist, but evidence is limited and little is known outside of correlative associations regarding herbivore biomass increases due to predator removal or other effects of disturbance.

In trophic cascades, if a trophic level is added or removed, coral-algal phase shifts and changes in ecosystem stability can occur (Terborgh, 2015). The strongest evidence of cascades found in coral reef systems come from herbivory, where depletion of grazers such as echinoderms from overfishing or pathogens allowed for explosive population growth of macroalgae in the Caribbean (Hughes et al., 1987; Mumby et al., 2005). Rasher et al. (2017) also described a trophic cascade via reduced exposure of herbivores to predators which resulted in variation in feeding patterns based on fear effects. Though the expected negative effects of cascades in reefs aren't always clear. In cases where conservation efforts have maintained predator populations via reduced fishing, the expected cascading effect of macroalgal increase from a depletion of grazers as prey doesn't always occur (Mumby et al., 2007). Keystone species and functional redundancy are not always present in reefs with high biodiversity (Hoey and Bellwood, 2009), which makes comparison across spatial scales all the more difficult. Even if assumed cascading events are observed such as phase shifts, finding the cause of such patterns at large spatial scales is challenging (Dulvy et al., 2004). Attention should also be paid to temporal gradients over which cascades occur, considering life-history and growth capabilities of populations as well as lagged effects from perturbations that can occur over extended time periods (Dulvy et al., 2004; Grubbs et al., 2016). The difficulty in linking trophic levels to the same event that categorizes a trophic cascade is difficult to support empirically, particularly when most reefs are already in a degraded state from external stressors. The fact that cascades have not yet been effectively documented for coral reefs undermines the assumed organization of coral reef food webs. High biodiversity complicates ecosystem models and trophodynamics, but may be key to the functionality of reef ecosystems in a changing environment.

Although pyramid structure is evident in coral reefs, there are high numbers of mesopredators representing intermediate trophic levels that are thought to exhibit functional redundancy. Mesopredators are considered to have less of an effect on the trophic structure of a system (Paine, 1980; Estes et al., 2011), but there is little empirical evidence to support this due to difficulty in defining predator-prey relationships. A recent meta-analysis of food web studies found that aquatic models produced a strong pyramid pattern, suggesting scale variance in predator-prey ratios according to biomass power laws consistent with Hatton et al. (2015) (Turney and Buddle, 2016). The analysis also showed that on average aquatic communities have a higher diversity of mesopredator species than herbivores, with low abundance of top predators. Intermediate effects of mesopredators on aquatic systems are largely unknown and often grouped within many ecosystem types such as intertidal, pelagic, and reef (Hatton et al., 2015). Elasmobranchs and other fishes are known to feed at different trophic levels based on stage of maturity, where they exhibit high functional redundancy as mesopredators and limited redundancy as apex (Navia et al., 2017). Thus, assigning a single ecological role to a species is limiting and can affect the predictive nature of model capability. The role and effects of mesopredators requires further exploration and definition to refine where greatest predation and productivity sources occur in reef systems. Many researchers believe that coral reefs are influenced by both top-down and bottom-up processes, and it should not be a question of one vs. the other (Terborgh, 2015). Further exploration of middle-driven systems and consideration of the driving interactions between trophic groups should be considered.

There are many components of trophodynamics that we are beginning to understand which can be used to inform ecosystem function and changes over time. Many of these components are shared among marine systems, but discretion should be used when assuming similar trophodynamics processes that are not fully described for coral systems. Current research validates that coral reefs fundamentally have:

These concepts need to be explicitly considered in study designs of trophodynamic research and uncertainty should be acknowledged before drawing conclusions regarding the ecological role of any species. Additionally, the underlying trophodynamic processes describing these observations should incorporate long-term datasets that accurately reflect the scope of data used. Before describing ecological roles of reef organisms, it is necessary to determine which interactions may have the highest influence on the trophic structure of these complex systems. For researchers, this space is still largely under-studied, and collaborative efforts are likely needed to elucidate the mechanisms that contribute to the stability of these ecosystems.

There are still many ecological unknowns in coral reef systems and there will likely never be a perfect approach to fully describe coral reef trophodynamics, but we can strive for better empirical data collection and analysis of patterns. Standardizing the approach to applied questions may help create a more cohesive space for collaboration in future studies. There is a need to apply multiple methods and clear definitions of spatial and temporal scale to meet the needs of trophodynamic research (Figure 2). How a species contributes energetically and how they interact with other species within a community take different methodological approaches and clear synthesis between the two to identify ecological roles. Arguably, coral reef systems do not appear to energetically behave the same as other marine food webs, such as intertidal or pelagic systems, and trophodynamic study should consider different scenarios and models. Based on concepts in this paper, we suggest recent literature may not adequately acknowledge the unique differences in coral reef food webs against the broader literature in marine trophodynamics, particularly over varying spatiotemporal scales. While theories for coral reefs are constantly being modified, conclusions of many articles still resort to generic descriptions of standard pyramid structure to explain biodiversity. Predation and competition within predator-prey interactions should be further considered in addition to exploring the effects of both bottom-up and top-down approaches. Without a better understanding of essential reef processes that affect ecological roles of species over both space and time, caution should be used in applying results to management and conservation efforts.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.