Giant appetites: exploring the trophic ecology of California’s largest kelp forest predator, the giant sea bass Stereolepis gigas

: The recovery of endangered predators has the potential to influence the ecosystems they inhabit. After suffering severe population declines due to fishing pressure, giant sea bass Stereolepis gigas in southern California, USA, are beginning to recover. As large-bodied predators often associated with the kelp forest and rocky reef environments of southern California and Baja California, Mexico, the local recovery of this species could influence trophic dynamics in these systems. Here we leverage stable isotope and gut content analysis to describe the trophic ecology of adult giant sea bass. We found that they are generalist predators, with larger individuals relying more heavily on macroalgae-derived carbon, occupying a larger trophic niche, and ob - taining higher trophic positions. Using these results, we speculate about the relationship between giant sea bass and kelp forest ecosystems, a vulnerable yet key habitat, including the impact of the return of these predators, as well as how contemporary threats to kelp forests might mediate their continued recovery.


INTRODUCTION
Many marine ecosystems have experienced de clines in predator populations in recent decades (Heithaus et al. 2008, Ritchie & Johnson 2009, Sadovy de Mitcheson et al. 2013. Loss of top predators can result in trophic cascades (Heithaus et al. 2008, Donohue et al. 2017, disruption of nutrient cycling (Schmitz et al. 2010), and broader loss of biodiversity (Ritchie & Johnson 2009). In some cases, conservation and management action have been effective at supporting the recovery of depressed predator populations (Waterhouse et al. 2020). While this is a positive outcome, it can be difficult to predict how predator recovery will influence ecosystems that have been operating in their absence for decades or longer (Ritchie & Johnson 2009). One key component in understanding the implications of predator recovery is the trophic ecology of the predator in question. Understanding how predators fit into the trophic system (what they eat, which primary producers they rely on, etc.) can help assess the potential impacts of their return, as well as how prey and producer communities might mediate their continued recovery.
Giant sea bass Stereolepis gigas are Critically Endangered (Cornish 2004) large-bodied predators that were once abundant in the kelp forests and rocky reefs of southern California, USA, and Baja California, Mexico (Dayton et al. 1998, Domeier 2001, Cornish 2004, Hawk & Allen 2014, Erauskin-Extramiana et al. 2017. Historically a sought after recreational and commercial fisheries species, giant sea bass experienced severe population declines in the USA due to overfishing in the early 20th century (Baldwin & Keiser 2008, Allen 2017, Ramírez-Valdez et al. 2021. Despite near extirpation in California, recent scientific evidence suggests that they are recovering after implementation of fishing regulations that de creased intentional and incidental catch of the species in the USA (Pondella & Allen 2008, Allen & Andrews 2012, House et al. 2016).
Due to their relative rarity, studies on giant sea bass are scarce and knowledge of their trophic ecology is largely limited to qualitative natural history reports (Young 1969, Feder et al. 1974, Domeier 2001. Giant sea bass are assumed to be generalist, high trophic level predators that feed on a wide array of primarily benthic nearshore rocky reef and kelp forest species, ranging from stingrays and small sharks to lobsters and octopuses (Domeier 2001, Allen & Andrews 2012, House et al. 2016. They are suction feeders, rapidly expanding their jaws to create a flow of water into their mouths that carries their prey along with it (Bishop et al. 2008). One of the only efforts to study giant sea bass diets compiled observations of individuals from the young-of-the-year age class and found that mysid shrimp are their dominant prey source (Benseman et al. 2019). Little else is known about their feeding ecology. For instance, little is known of the influence of ontogeny, different primary producers, or individual prey species on adult giant sea bass trophic dynamics. This study gathers baseline information that supports efforts aimed at anticipating broader ecosystem responses to the recent growth of giant sea bass populations in southern California.
Understanding how giant sea bass rely on different primary producers in their environment can help determine the extent to which shifts in production dynamics, whether from natural or anthropogenic drivers, might mediate population recovery. Kelp forests are key habitat for giant sea bass (Domeier 2001, House et al. 2016, Clevenstine & Lowe 2021. Kelp supports complex trophic systems by serving as an ecosystem engineer, creating structure and habitat in nearshore environments for diverse communities of organisms (Teagle et al. 2017, Layton et al. 2019. Primary production in kelp forest systems is derived chiefly from macroalgae (e.g. giant kelp Macrocystis pyrifera) and phytoplankton (Duggins et al. 1989, Fredriksen 2003, von Biela et al. 2016. Fluctuations in these producers can propagate throughout the food web, influencing the growth and production of higher trophic level species, including fishes (Koenigs et al. 2015, von Biela et al. 2016. Kelp forests globally are declining in patch size and kelp density, likely due to anthropogenic drivers of global change (Johnson et al. 2011, Steneck & Johnson 2014, Krumhansl et al. 2016, Layton et al. 2019). Threats to this critical habitat could hinder the ongoing recovery of giant sea bass, though our understanding of their reliance on kelp forests as producers is limited. While previous studies have demonstrated that, generally, higher trophic level fish tend to rely more on macroalgae-derived primary production in kelp forest ecosystems (Koenigs et al. 2015, von Biela et al. 2016, no studies have assessed the link between primary producers and giant sea bass trophic ecology. Two commonly used tools to investigate trophic ecology are gut content analysis and stable isotope analysis. Gut contents provide information on what an organism was eating just before being sampled and are useful because they provide direct observations of different prey. Stable isotopes constitute a representation of what an organism eats that is integrated through space and time, depending on how quickly their tissues turnover. Ratios of nitrogen isotopes ( 15 N/ 14 N) can help determine the relative trophic position of an organism, because they enrich at a predictable rate with each increasing trophic level (Finlay et al. 2002, Post 2002. Ratios of carbon isotopes ( 13 C/ 12 C) tend to reflect sources of primary productivity, because carbon isotopic signatures are generally well conserved through trophic transfer (Hobson 2005, Kurle et al. 2011. Bulk ratios of carbon and nitrogen isotopes can be used to test for size/agebased differences in diet, incorporated into stable isotope mixing models to determine the relative contribution of different primary producers to the diet of a predator, and compared against each other to determine the isotopic niche of an organism (as a proxy for trophic niche). Compound specific stable isotope analysis of amino acids (CSIA-AA) splits bulk nitrogen into target amino acids that show different enrichment properties (Whiteman et al. 2019). By comparing these different amino acids, it is possible to get an estimate of the trophic position of an animal from a single sample of their tissue.
In this study, we leveraged bulk and compound specific stable isotope analyses along with gut content data to better understand the trophic role of a recovering predator, the giant sea bass. In particular, we set out to determine (1) what giant sea bass eat, and how that changes as a function of size/ontogeny; (2) which primary production sources giant sea bass are most reliant on and if/how that relationship changes throughout their life history; (3) the relative trophic position of giant sea bass; and (4) what insight this information can give us about the role of giant sea bass in one of their key habitats, the kelp forest ecosystem.

Sample collection
We collected fin clip samples -approximately 1 to 2 cm of tissue clipped from the anal fin -to perform stable isotope analyses. Typically, stable isotope analyses of fish use muscle tissue; however, previous studies have demonstrated that fin clips are good analogs for muscle tissue when the latter is not available, as was the case in this study (Suzuki et al. 2005, Sanderson et al. 2009, Hanisch et al. 2010, Jardine et al. 2011. Sampling took place from 2017 to 2020, and spanned the core range of giant sea bass, which is south of Pt. Conception, California, along the Pacific coast of Baja California and the upper Gulf of California, Mexico (Domeier 2001, Ramírez-Valdez et al. 2021 (Fig. 1). Most of our samples (n = 56) came from individuals caught by fishing cooperatives involved in the finfish fishery in Baja California, through a collaboration with the biological monitoring program, Proyecto Mero Gigante, and the non-profit Comunidad y Bio diversidad A.C. (COBI). The Mexico sampling locations ranged from predominantly soft bottom to predominantly rocky reef habitats and varied in the presence of kelp (Table S1 in the Supplement at www.int-res.com/articles/ suppl/m695p157_supp.pdf). Additionally, we collected samples from La Jolla, California, as part of a separate effort permitted by the California Department of Fish and Wildlife (CDFW) that in volved tagging and releasing giant sea bass (n = 5). We opportunistically collected 2 samples from deceased individuals that washed up in Solana Beach and Carlsbad, Califor-nia. All USA sampling locations were near or within rocky reef and kelp forest areas with adjacent sandy bottom habitats. Stable isotope samples are typically preserved by freezing, but due to limited availability of freezer facilities for our samples from Mexico, we preserved all the fin clip samples in a 95% ethanol solution. Previous work investigating the impacts of different pre servation methods on stable isotope analysis indicates ethanol preservation of fish tissues has minimal to no impact on isotope values (Sarakinos et al. 2002, Hetherington et al. 2019). In addition to fin clips, we gathered morphometric information on all the fish sampled, including total length (TL).
We gathered information on the isotopic signatures of potential primary producers (macroalgae vs. phytoplankton) through 2 methods. First, we col-  (Richards et al. 2020), we used our pyrosome samples as a primary consumer proxy for offshore phytoplankton production throughout our study region. We preserved these samples by freezing them. Second, we conducted a literature review and recorded M. pyrifera and particulate organic matter (POM; a proxy for phytoplankton) isotope values from studies that previously measured primary producers in our study region (Table S2) (Page et al. 2008, Hamilton et al. 2011, Vega-García et al. 2015, Piñón-Gimate et al. 2016, Kurle & McWhorter 2017, Gabara 2020).

Gut content data
Gut content data are difficult to collect for giant sea bass, because they are largely protected from harvest in the USA and those that are caught in fisheries are generally gutted at sea. To overcome this challenge and gather information on the diet of giant sea bass, we drew from multiple data sources. First, we conducted a literature review, recording all prey taxa mentioned and whether the observations were from live feeding events, from inspecting gut contents, or uncredited anecdotal reports. Second, the Hubbs-Sea-World Research Institute (HSWRI) shared information recorded on the stomach contents of giant sea bass caught incidentally during their juvenile white seabass Atractoscion nobilis gill net surveys in the nearshore waters off San Diego county. Third, we conducted gut content analysis on 2 individuals that washed up in the San Diego region, 1 on Solana Beach and 1 on Mission Beach, California, as well as 10 individuals from Isla Natividad, Baja California that had their gut contents retained by fishing cooperative members on our request. And finally, we reached out to fellow researchers for any unpublished records of gut content analysis or observations of feeding interactions in the field (e.g. videos opportunistically taken by scuba divers). Most notably, Dr. Larry Allen provided information on an individual from La Jolla (L. G. Allen pers. comm.) that he had necropsied and gathered gut content from in 2015. After compiling the disparate re cords of stomach contents, we generated a list of potential prey items and grouped each based on our confidence in the source of the information.

Bulk stable isotope analysis
We prepared fin clip and primary producer samples for bulk stable isotope analysis in accordance with previously published works (Hanisch et al. 2010, Jardine et al. 2011, Hetherington et al. 2019. We removed each sample from its ethanol preservative (for fin clips) or the freezer (for primary producers), rinsed it with de-ionized water for at least 1 min, and freeze-dried for 48 h. Once dried, we homogenized the samples using a mortar and pestle and/or a scalpel and placed approximately 1 ± 0.2 mg (for fin clips) or 5 ± 0.2 mg (for primary producers) of material into pre-weighed tin capsules. The Isotope Biogeochemistry facility at Scripps Institution of Oceanography analyzed all samples to determine the bulk carbon and nitrogen isotope values. We ex pressed all results as δ values (parts per thousand differences from a standard or per mil; ‰) using the following equation: (1) where X is either 13 C or 15 N and R is the ratio 13 C/ 12 C or 15 N/ 14 N using acetanilide standards (Baker A068-03, Lot A15467).

Compound specific stable isotope analysis
We performed compound specific stable isotope analysis of amino acid (CSIA-AA) nitrogen on a subsample of 20 fin clips. The size of our subsample was dictated by the cost limitations of running CSIA-AA. We selected samples from Guerrero Negro, Mexico that represented a wide size range of fish (TL: 51 to 197 cm). We split a single fin clip from 1 individual into 3 subsamples that we processed separately to produce procedural reproducibility errors for each of 4 target AAs: alanine (Ala), glutamic acid (Glu), glycine (Gly), and phenylalanine (Phe). Ala and Glu are trophic AAs, meaning they enrich with rising trophic levels, while Gly and Phe are source AAs, meaning they do not enrich with rising trophic levels (Whiteman et al. 2019). The difference in enrichment between these AAs allows us to estimate the trophic position of an individual from a single sample (see Section 2.5.). The δ 15 N of the selected AAs is not significantly altered by ethanol and thus preservation effects on trophic position estimation are negligible (Swalethorp et al. 2020).
We removed each sample from the ethanol preservative, rinsed it with de-ionized water for at least 1 min, and then freeze-dried for 48 h. We employed a relatively new high-precision method using highpressure liquid chromatography and offline elemental analysis isotope ratio mass spectrometry (HPLC/ EA-IRMS) (Broek & McCarthy 2014, Swalethorp et al. 2020. All specifics of this method can be found in Swalethorp et al. (2020). Briefly, we hydrolyzed a minimum of 6 mg (dry weight) of each sample in 1 ml of 6 mol l −1 HCl in capped glass tubes for 24 h at 90°C. We then dried the samples on a centrifugal evaporator under vacuum at 60°C, re-dissolved them in 0.5 ml of 0.1 mol l −1 HCl, and filtered them through an IC Nillix -LG 0.2 μm hydrophilic polytetrafluoroethylene (PTFE) filter to remove particulates. Finally, we re-dried the samples before re-dissolving them in 100 μl of 0.1% trifluoroacetic acid (TFA) in Milli-Q water and transferring them to glass inserts in glass vials. We stored all samples at −80°C for 1 to 4 wk prior to AA purification. For each sample, we purified and collected the trophic AAs Ala and Glu, and the source AAs Gly and Phe. We dried these AAs in a centrifugal evaporator at 60°C, re-dissolved them in 40 μl of 0.1 mol l −1 HCl, then transferred them into small tin capsules and dried them under vacuum. We used the Stable Isotope Laboratory facility at the University of California, Santa Cruz to carry out analyses on a Nano-EA-IRMS designed for high precision analysis of low mass samples (≥ 0.6 μg N).

Data analysis
We performed all analyses using R statistical software, version 3.6.1 (R Core Team 2019). The code for our analyses can be found at https://github.com/ kmblincow/GSBIsotopeAnalysis.
We analyzed the bulk stable isotope data for relationships with fish body size using Bayesian linear models coded in R and JAGS software (Plummer 2003) with the package R2jags (Su & Yajima 2020). To account for possible influences of year and sample location, we tested multiple models incorporating different combinations of these variables, including a derived categorical variable of year and sample site combined (see Table 1). We created the derived variable to address potential confounding changes in sites across years. We treated year, sample site, and the combined year and sample site variable as random effects, and the TL of each fish sampled as a fixed effect. We ran each model using 3 parallel Monte Carlo Markov Chain (MCMC) chains, each obtaining 350 000 samples, the first 25 000 of which were discarded as burn-in. We retained every 25th iteration to reduce autocorrelation, resulting in an output of 39 000 samples of the posterior distribution for each chain. We confirmed model convergence by evaluating trace plots and the potential scale reduc-tion factor (R ) (Gelman & Rubin 1992). We used leave-one-out cross-validation (LOO) to determine which models best predicted the data using the package loo (Vehtari et al. 2017(Vehtari et al. , 2020. We evaluated the isotopic niche of different age classes of giant sea bass using the SIBER package (Stable Isotope Bayesian Ellipses in R) (Jackson et al. 2011). Giant sea bass are thought to mature between the ages of 7 and 13 (Domeier 2001, Hawk & Allen 2014. Using the age-growth equation derived by Hawk & Allen (2014), we calculated the expected age associated with each of our samples and classified them as either 'immature' (< 7, n = 18), 'transition' (7 to 12, n = 23), or 'mature' (>12, n = 22). We calculated the standard ellipse area (SEAc), which is a representation of bivariate standard deviation, for each of these groups. Additionally, we performed Bayesian estimation of the standard ellipses and compared across groups. SEAc is robust to small sample sizes, unlike other isotopic niche metrics such as convex hulls (Jackson et al. 2011). We ran our SIBER model using 3 parallel MCMC chains, each obtaining 50 000 samples, the first 25 000 of which were discarded as burn-in. We retained every 5th iteration. We confirmed model convergence by evaluating trace plots and the potential scale reduction factor (R ).
We incorporated our bulk isotope results into a Bayesian isotope mixing model to determine what proportion of giant sea bass diets are derived from either phytoplankton or macroalgae primary production sources using the MixSIAR package (Stock et al. 2018). MixSIAR requires the input of source isotope values (i.e. macroalgae and phytoplankton primary production), mixture isotope values (i.e. giant sea bass), estimates of the trophic discrimination factors (TDF) between the source and mixture for each isotope, and data on relevant covariates (i.e. TL). We ran our model using 3 parallel MCMC chains, each obtaining 100 000 samples, the first 50 000 of which were discarded as burn-in. We retained every 50th iteration. We used the same convergence criteria as in the prior analyses.
To determine our source isotopic signatures, we collected the mean estimated bulk isotope values from multiple studies (including our own), and then calculated the mean and variance of these values. We subsequently used these means and variances as fixed source values (not estimated based on sample data) in our mixing model (Stock et al. 2018). We estimated the relative proportion of macroalgae-derived carbon by relying on isotopic estimates of the dominant Pacific coastal macroalgae species in our sam-ple locations, M. pyrifera (Edwards & Hernández-Carmona 2005). However, we had 1 fin clip sample from the upper Gulf of California, which does not encompass the range of M. pyrifera. Despite the fact that other coastal macroalgae present in the upper Gulf of California likely have similar values to M. pyrifera, we chose to exclude this sample from the mixing model analysis because it was unreasonable to assume M. pyrifera would be contributing to its isotopic signature. To convert pyrosome isotope values into phytoplankton primary production (under the assumption that pyrosomes rely exclusively on a phytoplankton-derived food web), we corrected our pyrosome-derived δ 15 N estimates by −2 ‰ and our δ 13 C estimates by 0.5 ‰ based on previously published estimates of the trophic position of pelagic tunicates (Hetherington et al. 2018, Décima et al. 2019, Schram et al. 2020. To our knowledge, there are no experimental estimates of TDF for giant sea bass. We chose to use mean TDF values of 0.9 ‰ (SD = 0.5) and 3.4 ‰ (SD = 0.5) for δ 13 C and δ 15 N, respectively, based on previously published isotope research for similar species (Artero et al. 2015) and broad TDF estimates for carnivorous species (DeNiro & Epstein 1978, Post 2002. Since our source and mixture pop ulations were multiple trophic steps apart, it was necessary to correct our TDF values based on that difference (Phillips et al. 2014). To do so we used the trophic position estimates from our CSIA-AA analysis (described below) to calculate a mean trophic position for giant sea bass and multiplied our TDF values by that value minus 1.
We used the 2 trophic (Trp) and 2 source (Scr) AA δ 15 N values to calculate the trophic position of giant sea bass using β and TDF values from Bradley et al. (2015) and the following equation: (2) where β is the δ 15 N offset between Trp and Scr AAs in primary producers, while TDF AA is the average δ 15 N enrichment of Trp relative to Scr AAs in consumers for each increasing trophic step. To generate more robust trophic position estimates we used weighted means of both Ala and Glu, and Gly and Phe (Nielsen et al. 2015) and the following equation: where δ 15 N x is the value of a specific Trp or Scr AA and σ x is the procedural reproducibility error re -ported here as the SD from replicate analysis of the 3 subsamples of 1 of the fin clips. These SD values were 0.24, 0.68, 0.17, 1.57 for Ala, Glu, Gly, and Phe, respectively. We also calculated weighted means for the β and TDF values and associated SDs (Bradley et al. 2015).
After calculating the trophic position associated with our samples, we performed a simple Bayesian linear regression to look for a relationship with fish size (TL) using the same model specifications described for the linear models above. For the fin clip sample that was split into 3 and run separately, we calculated the mean of the associated results so as not to triple count them.

Bulk stable isotopes
We gathered a total of 63 tissue samples from fish with TLs ranging from 44 to 197 cm (94.57 ± 42.08; mean ± SD). δ 13 C values ranged from −16.50 to −12.39 ‰ (−14.53 ± 1.10; mean ± SD), and δ 15 N values ranged from 15.95 to 19.54 ‰ (17.91 ± 0.90) (Fig. S1). Due to our opportunistic sampling, the sample sites are not evenly represented (Table S1), with the majority of samples coming from Baja California, Mexico (n = 56), in particular Guerrero Negro (n = 31). Our Bayesian linear models indicated a strong positive relationship between TL and δ 13 C (Fig. 2a). LOO model comparison found that sample site and TL were the most important variables for predicting δ 13 C (Table 1). While LOO identified year and TL as the best predictors of δ 15 N out of the variables we tested (Table 1), the relationship between TL and δ 15 N was weak (Fig. 2b). We should note that some of the models we evaluated using LOO had a small number of Pareto k values that exceeded 0.5. We evaluated each of these points individually, and found that they rarely exceeded 0.7, the recommended cutoff for the utility of LOO methods in model comparison (Vehtari et al. 2017). Based on the guidance of Vehtari et al. (2017), we decided our models were robust for LOO comparison.
By plotting the SEAc derived from the SIBER analysis, we found overlap in the isotopic niches of the 3 different age groups (Fig. 3a). We found that the mature age class had the highest Bayesian SEAc estimate, followed by the transition, then the immature age classes (Fig. 3b).
Based on the literature values and our primary producer samples, we calculated δ 13 C values (mean ± SD) of −14.81 ± 0.58 and −21.55 ± 0.45, and δ 15 N values of 9.46 ± 0.47 and 6.71 ± 0.39, for macroalgae and phytoplankton, respectively. When using these as our source values, the mixing model showed that giant sea bass are chiefly reliant on macroalgae as a basal carbon source and that this trend increases with size (Fig. 4). The median proportion of the diet associated with macroalgae was 0.61 (95% CI: 0.55−0.66) at the smallest end of the size range and 0.91 (95% CI: 0.83−0.95) at the largest end of the size range.

CSIA-AA
Of the 20 samples we used for CSIA-AA, we found that 3 had unreasonably high Phe δ 15 N values that led to very low trophic position estimates, and none of these were supported by the associated bulk δ 15 N data. These samples had possibly degraded and were classified as outliers and subsequently removed from further analyses (Fig. S2). Additionally, we were un able to determine the Gly value for 1 sample, so we did not include that sample in our trophic position estimates.  Table 1. Leave-one-out cross-validation model comparison testing the relationship between bulk isotope estimates and total length (TL; fixed effect), sample site (random effect), year (random effect), and year and site combined (YearSite; random effect). Expected log pointwise predictive accuracy (elpd_loo) is a measure of the predictive accuracy of the model which can be compared across models using the same data. The difference between these values for different models is given by the elpd_diff column, and the SE of component-wise differences of the elpd_loo between models is shown in the se_diff column Fig. 2. Bayesian estimates of the linear relationships between the bulk isotope results and total length. Note that the y-axis displays the scaled bulk isotope values, which have a mean of 0 and SD of 1. Black lines: mean posterior hyperparameter estimate for the slope and intercept for each respective model; gray lines: 1000 random draws from the hyperparameter posterior distribution for the slope and intercept for each respective model; black dots: raw data points. (a) δ 13 C and total length (cm) using the model which incorporated total length as a fixed effect and sample site as a random effect. (b) δ 15 N and total length (cm) using the model which incorporated total length as a fixed effect and year as a random effect 1.45), Glu values ranged from 26.65 to 32.11 ‰ (29.81 ± 1.72), Gly values ranged from 7.15 to 11.51 ‰ (9.42 ± 1.00), and Phe values ranged from 10.01 to 13.16 ‰ (11.67 ± 0.90) (Fig. 5a). The trophic position estimates ranged from 2.7 to 4.1 (3.39 ± 0.35). We found a strong positive linear relationship between trophic position and TL (Fig. 5b).

Gut contents
Synthesizing information from the compiled gut content data and literature reports, we found a total of 36 prey items, 25 of which were associated with direct observations from giant sea bass ranging in size from 43.7 to 141.3 cm TL (77.93 ± 26.99, n = 23) ( Table 2). Most of our gut content data did not include records of prey size, though where it was recorded, we found that the largest prey items tended to be associated with larger giant sea bass. We should note that this relationship was not reciprocal, and larger giant sea bass also had smaller prey items in their stomachs, such as the Solana Beach individual (137.0 cm TL) that had 20 individual crabs in its stomach ranging in size from 4.2 to 9.0 cm in carapace width. The gut contents revealed prey groups that were wide ranging in terms of life history strategies, habitat preferences, and predator-avoidance mechanisms. For example, we found fast-moving pelagic prey, such as sardines and market squid, alongside benthic prey that rely heavily on camouflage, such as turbots and octopus. Directly observed prey items included both fish (60%) and invertebrates (40%). Among directly observed fish prey, 10 different orders were represented, with no clear dominant group (1 to 2 species represented in each order). Among directly observed invertebrate prey, Decapoda was the dominant order, representing 60% of the species. Most directly observed prey were associated with benthic habitats (87.5%), with

DISCUSSION
As Critically Endangered (Cornish 2004) predators whose populations are beginning to recover, giant sea bass present an opportunity to learn how the return of generalist predator populations can influence the broader ecosystems to which they belong.
Using stable isotopes, we demonstrated that adult giant sea bass continue to undergo shifts in their diet as they grow -relying more heavily on macroalgaederived carbon, occupying higher trophic positions, and expanding their trophic niche with increasing size. Compiling available gut content records, we found that they consume a remarkably wide range of prey, consisting of a diverse assortment of primarily benthic-associated fish and invertebrate species. These findings lay the groundwork for understand-ing how giant sea bass recovery will influence kelp forest ecosystems and vice versa.
In addition to confirming long-held assumptions regarding giant sea bass trophic ecology, our study also uncovers some of the nuances underlying those assumptions. The wide array of observed prey supports the previous characterization of giant sea bass as generalist predators (Domeier 2001, Allen & Andrews 2012, House et al. 2016). Our trophic position estimates were relatively high, though not as high as those predicted from standardized trophic models. For example, FishBase estimates the trophic position of giant sea bass to be 4.5, while our mean estimate was 3.39 (Froese et al. 2022). The fact that these predators feed throughout the food web, including on low trophic level species (e.g. razor clams), results in a lower mean trophic position than predicted by standardized models. We think this also accounts for the weak relationship we found between bulk nitrogen and total length.
We found a relatively high overlap of trophic niche among age classes, though the mature age class had the largest trophic niche width. This suggests that while giant sea bass consume similar prey across size 165 Fig. 5. Compound specific stable isotope analysis of amino acids (CSIA-AA) results. (a) The δ 15 N values for the amino acids we tested. Glutamic acid and alanine are trophic amino acids which enrich with trophic level increases, while glycine and phenylalanine are source amino acids which do not enrich with increasing trophic levels. We used the differences between the source and trophic amino acids to estimate the trophic position for each fish tested using CSIA-AA. (b) The Bayesian estimate of the linear relationship between the trophic position and total length of samples analyzed using CSIA-AA. Black lines: mean posterior hyperparameter estimate for the slope and intercept for each respective model; gray lines: 1000 random draws from the hyperparameter posterior distribution for the slope and intercept for each respective model; black dots: raw data points  Table 2. Prey types gathered from direct and anecdotal observations. The size of the giant sea bass (GSB) and associated prey items are given where available, along with the habitat associated with the prey type, and the source of the observation. Lengths of prey items are recorded in total length for fish, octopus, and squid, carapace width for crabs, and carapace length for other crustaceans. HSWRI: Hubbs-SeaWorld Research Institute; NA: not available. See Section 2.2 for details on the sources of data ranges, the mature age class is interacting with the prey base in a different way. Explanations for the expansion of trophic niche width among the mature age class include an increased diversity of prey items available, individual specialization due to prey selection, or individual specialization due to spatial or temporal variability in prey (Matich et al. 2011). It is possible that size limitation influences the diversity of prey available across age classes. For the gut content samples where prey size information was available, larger prey items tended to be found in stomachs of larger individuals. For example, one of the largest giant sea bass sampled, measuring 141 cm in TL, had 2 shovelnose guitarfish Rhinobatos productus measuring 60 and 30 cm in TL inside its stomach. It is unlikely giant sea bass at the smaller end of those analyzed in this study (~40 to 60 cm) would be able to consume these prey. However, a separate large individual (137 cm TL) had 20 relatively small crabs in its stomach, ranging in carapace width from 4.2 to 7.7 cm. This suggests that size limitation might not be the sole driver of the increase in trophic niche width, and further study is warranted to investigate the role of individual specialization in the trophic ecology of mature giant sea bass. One goal of this study was to determine which primary production sources giant sea bass rely on and whether that changes throughout their life history. Our mixing model results indicated that they feed chiefly in macroalgae-based food chains, and that this becomes increasingly true as they grow, shifting from ~60% macroalgae-derived carbon to ~90% across the size range studied. Anecdotal reports suggest smaller giant sea bass rely more heavily on pelagic species such as anchovies and sardines (Baldwin & Keiser 2008). Our limited records of pelagic prey items were associated with smaller individuals (76.5 and 79.0 cm in TL). This could account for the shifts we see in the relative proportion of primary producer sources in giant sea bass diets. Transient, pelagic-associated prey could serve as links be tween offshore and nearshore environments, increasing the relative proportion of phytoplanktonderived carbon in the tissue of smaller individuals. As giant sea bass grow and transition to a diet consisting of predominantly benthic organisms, there is an increase in the proportion of macroalgae-derived carbon in their diets. Previous studies tracing primary producer contributions to fish diets in nearshore environments in the eastern Pacific also document high proportions of macroalgae-derived carbon (Duggins et al. 1989, Koenigs et al. 2015, von Biela et al. 2016. The vast majority of the prey items we doc-umented lived in benthic habitats, and most of those benthic prey were associated with sandy bottom habitats in particular. Initially, this might seem to contradict the mixing model results; however, sandy bottom habitats adjacent to kelp forests have been shown to rely significantly on primary production from drift macroalgae (Polis et al. 1997, Harrold et al. 1998, Crawley et al. 2009, Filbee-Dexter et al. 2018. It is likely that giant sea bass feed in sandy bottom habitats that rely on macroalgae from adjacent rocky reefs to subsidize their carbon supply. Indeed, studies investigating giant sea bass movement and anecdotal reports from diving and fishing communities suggest that they frequent areas where reef transitions to sandy bottom habitat (Burns et al. 2020, Clevenstine & Lowe 2021. It is unclear how current threats to kelp forest systems will influence the trophic ecology of giant sea bass and by extension their continued recovery, but we can speculate based on the results of this study. While the individuals in our analysis showed that a large portion of their diet is derived from macroalgae-based food chains, their prey are not necessarily obligate kelp forest inhabitants and neither are the giant sea bass themselves. Assuming that the ecosystem can still support an adequate prey supply, the potential negative impacts of kelp forest declines on giant sea bass might not be as severe as initially thought. One study that looked at the impact of Macrocystis pyrifera deforestation caused by urchin grazing on kelp forest food web structure found that the negative impacts of deforestation were not as strong at higher trophic levels similar to those occupied by giant sea bass (Graham 2004). Furthermore, a portion of the geographic range of giant sea bass does not coincide with the range of kelp (Ramírez-Valdez et al. 2021), and there are anecdotal reports of the local extirpation of kelp from rocky reefs in the Channel Islands having no apparent influence on their presence (Domeier 2001). These findings suggest that threats to kelp forests do not necessarily equate to threats to giant sea bass.
It is likely that the reason giant sea bass are often associated with kelp forest environments is the ability of these systems to support a high biomass of diverse prey groups, both in the immediate rocky reef environment and in adjacent sandy bottom habitats (Polis et al. 1997, Graham 2004. As such, the impact of the potential loss of kelp forests on giant sea bass will depend, at least in part, on the extent of the impacts on their prey. Generally, resourcelimited conditions, like those brought on by the loss of kelp, lead to increased diet specialization among predator populations because there is an increased need for efficient resource exploitation (Bolnick et al. 2003, Matich et al. 2011. For example, sea otters in Central California, a relatively resource-limited environment experiencing ongoing declines in kelp biomass, have more specialized diets when compared to sea otters in Washington (Laidre & Jameson 2006, Matich et al. 2011. It is possible that we could see a similar shift in giant sea bass trophic ecology with the reduction of kelp. Overall, their ability to exploit a diverse array of prey items suggests that giant sea bass populations will be able to adapt relatively well to changing conditions when compared to other more specialist predators. Giant sea bass have been essentially absent from kelp forests in the USA for much of the 20th century, making it difficult to predict how their recovery will influence these ecosystems. It is very possible that ecosystem conditions have changed since the last time giant sea bass were a regular feature of these kelp forest environments. As generalists, their impact is spread across a wide variety of prey. They appear to feed on both fish and invertebrate species, but among invertebrate communities, species from the order Decapoda were most commonly targeted. These findings are similar to what has been found for goliath grouper Epinephelus itajara, an Atlantic large-bodied generalist predator that is also recovering from severely reduced population sizes due to overharvest (Koenig et al. 2020). Research on the diet of goliath grouper has found that they feed chiefly on crustaceans and benthic fish, targeting prey throughout the food web (Koenig et al. 2020). We found that giant sea bass also primarily feed on benthic species across different trophic levels. This suggests their recovery could have top down impacts on the structure of benthic communities in particular; however, the extent of these impacts is difficult to ascertain and further research and careful consideration is needed to fully characterize them. The dangers of misrepresenting these effects can be seen in the case of goliath grouper, where misconceptions regarding the trophic impacts of the species on fisheries species has been used to promote the cessation of conservation efforts (Koenig et al. 2020).
Further work expanding upon the foundation presented here would help inform our understanding of giant sea bass as well as the broader impacts of predator recovery. For example, measuring isotopic signatures from multiple tissue types with different turnover rates from the same individual would help tease apart the underlying mechanism of the growth in niche width we observed in mature age classes. Continuing to compile gut content data would support a more comprehensive understanding of giant sea bass prey communities, including primary prey species. Obtaining isotopic signatures of dominant prey species would help contextualize our results in terms of the broader food web and help confirm our assertion that giant sea bass rely on adjacent macroalgae-subsidized sandy bottom areas as feeding habitat. Pairing these suggested inquiries into the trophic ecology of giant sea bass with long-term monitoring of kelp forest communities would help assess the role of this recovering apex predator within the broader ecosystem, including potential top down effects on prey communities.