Seasonally mediated niche partitioning in a vertically compressed pelagic predator guild

Abstract

Niche partitioning among closely related, sympatric species is a fundamental concept in ecology, and its mechanisms are of broad interest for understanding ecosystem functioning and predicting the impacts of human-driven environmental change. However, identifying mechanisms by which top marine predators partition available resources has been especially challenging given the difficulty of quantifying resource use of large pelagic animals. In the eastern tropical Pacific (ETP), three large, highly mobile and ecologically similar pelagic predators (blue marlin (Makaira nigricans), black marlin (Istiompax indica) and sailfish (Istiophorus platypterus)) coexist in a vertically compressed habitat. To evaluate each species' ecological niche, we leveraged a decade of recreational fisheries data, multi-year satellite tracking with high-resolution dive data, and stable isotope analysis. Fishery interaction and telemetry-based three-dimensional seasonal utilization distributions suggested high spatial and temporal overlap among species; however, seasonal and diel variability in diving behaviour produced spatial partitioning, leading to low trophic overlap among species. Expanding oxygen minimum zones will reduce the available vertical habitat within predator guilds, likely leading to increases in interspecific competition. Thus, understanding the mechanisms of habitat partitioning among predators in the vertically compressed ETP can provide insight into how predators in other ocean regions may respond to vertically limited habitats.

1. Introduction

The principle of competitive exclusion [1,2] predicts that only one species can occupy the same ecological niche, defined as a hypervolume in n-dimensional (nD) space with environmental variables as axes [3]. Thus, within ecosystems, sympatric predators with similar ecological requirements must segregate along one or more axis of this hypervolume to coexist (i.e. niche partitioning; [4,5]). Niche partitioning occurs in a variety of forms, and is often observed along spatial [6] or temporal shifts in habitat use [7]. Competitive interactions may be further reduced among sympatric predators via separation of their trophic axes (i.e. having different diets) [8,9]. Consequently, competitive interactions and the resulting mechanisms of partitioning among species fulfilling specialized functional roles have significant influence on community structure and ecosystem functioning [10,11].

Billfishes (family Istiophoridae) are highly migratory top predators in the pelagic environment, yet seasonally co-occur in high abundances in coastal regions with narrow continental shelves [12,13]. However, owing to their large size, inability to be held captive, and their elusive nature, information on billfish ecology and competitive interactions remains limited. One region where multiple billfishes co-occur is the Pacific coast of Central America, known for high catch rates of sailfish (Istiophorus platypterus), blue marlin (Makaira nigricans), striped marlin (Kajikia audax) and black marlin (Istiompax indica) [14,15]. In Panama, recreational billfish catch is high during the dry season (December–April), when the Intertropical Convergence Zone (ITCZ) is pushed south by northerly trade winds, resulting in upwelling of cool, nutrient-rich waters [16]. Surface chlorophyll concentration peaks during this time, creating a surge in surface productivity in the region [17]. During the wet season (May–November), the trade winds lessen, upwelling decreases, surface waters return to a warm, nutrient-poor, low-productivity state [17] and coastal billfish occurrence declines [18].

In addition to seasonal atmospheric variability, the eastern tropical Pacific (ETP) contains several interacting oceanographic features leading to a year-round strong and shallow thermocline, and the world's largest naturally occurring oxygen minimum zone (OMZ) beneath the thermocline [16]. Seasonal upwelling in the dry season causes thermoclines and hypoxic boundaries to shoal in this region to as shallow as 25–30 m [19]. Because cold temperatures and hypoxic conditions below the thermocline cause a reduction of the available habitat for high-oxygen-demand species such as tunas and billfishes [20], these predators become restricted to the mixed layer, a condition known as hypoxia-based habitat compression [21,22].

While many studies have examined either movement patterns or trophic positions of various billfishes [23–26], none has integrated spatial tracking and isotopic analysis concurrently to compare the ecological niches of this predator guild. Blue marlin (hereafter BUM), black marlin (BAM) and sailfish (SFA), are phenotypically alike, prefer similar water temperatures, and are similarly vertically restricted by low temperature and oxygen at depth [21,27]. In addition, the stomach contents of BUM and SFA landed in the ETP indicate a dominance (97 and 96.2%, respectively) of surface-oriented teleost prey (e.g. scombrids; [25]), and, where examined, BAM show similar prey preferences [28]. Comparable blood hematocrit and haemoglobin concentrations, and similar gill morphologies indicate that physiologically, limitations among billfishes are likely similar [29–31]. As such, interspecific interactions among these similar billfishes are likely high while occupying the vertically compressed ETP.

Understanding how this predator guild responds to low-oxygen environments, environmental variability, and periods of high and low prey abundance has important implications for how this guild may impact the pelagic ecosystem, but also for predicting how populations and communities in other regions will respond to larger-scale climatic change [32,33]. For instance, global OMZs are predicted to grow both vertically and horizontally, and hypoxia-based habitat compression will begin to affect new regions and ecosystems [34,35]. For example, the greatest and most certain decreases in oxygen concentrations are projected to occur between 200 and 700 m, depths to which billfishes and many other pelagic predators regularly descend to forage [35,36]. Thus, understanding the mechanisms of coexistence among the billfish guild in the ETP provides an opportunity to study how sympatric marine predators may use and partition vertically compressed habitat.

Taken together, the seasonal resource availability, similar habitat preferences, diet and physiological limitations of billfish suggest that BUM, BAM and SFA have the potential for high ecological niche overlap in the vertically compressed ETP. Here, we employ a multi-disciplinary approach using long-term recreational fishery catch data, multi-year horizontal and vertical movement data, and stable isotope analysis to unravel the spatial, temporal and trophic mechanisms that facilitate coexistence of these closely related sympatric top predators.

2. Methods

(a) . Tagging and sample collection

BUM, BAM and SFA were caught off the Pacific coast of southeast Panama (electronic supplementary material, table S1), based out of Tropic Star Lodge, Piñas Bay, from November 2018 to February 2022, via rod-and-reel, trolling lures or natural bait. Each fish was brought alongside the vessel, and assessed for physical trauma associated with capture, and its weight estimated by an experienced captain. A subset of fish were tagged with a pop-up satellite archival transmitting tag (miniPAT, Wildlife Computers, Redmond WA, USA) anchored into the dorsal musculature using either an umbrella or a titanium dart. Tags were painted with an anti-fouling coating and tethered to the dart with 30 cm of 1.8 mm diameter fluorocarbon monofilament (136 kg tensile strength). Fish were kept in the water during tagging. Prior to release, a small fin clip (less than 2 cm) was taken from the distal end of the pectoral fin for stable isotope analysis. All samples were rinsed with fresh seawater, placed in vials, immediately stored on ice while at sea and then frozen until further processing.

(b) . Data and sample processing

(i) . Horizontal tracks

Only tags attached for more than 10 days were used in analyses. Horizontal tracks of billfish were reconstructed using the Global Position Estimator 3 (GPE3, Wildlife Computers), which uses the tag records of light, temperature and depth, and reference data on sea surface temperature and bathymetry with a user-defined movement speed. Blue marlin has been observed to sustain swimming speeds of 0.80–1.20 m s⁻¹, with bursts of up to 2.25 m s⁻¹ [37]. As such, the speed parameter is commonly set around 1.5–2 m s⁻¹ for pelagic predators like tuna and billfish [38,39]. Therefore, we ran several iterations of the model for each fish with 0.5 m s⁻¹ increments from 0.5 to 2.5 m s⁻¹, and selected the most probable track that converged, but limited large spikes and gaps between successive locations (S. Kohin (Wildlife Computers) 2023, personal communication). Estimated tracks were further filtered in R [40] using a speed–distance–angle algorithm [41] to remove improbable locations. Tracks and depth time-series were trimmed to remove locations or data that indicated the tag was no longer attached to the fish or that the individual had died.

(ii) . Depth and temperature

Tags were initially programmed to detach from the fish after 365 days in 2018, then 240 days in the 2019 and 2021 tagging seasons. Tags archived depth (±0.5 m) and temperature (±0.5°C) at 3 or 5 s intervals every other day for 240 and 365 day programmed attachment durations, respectively. Once released from the fish, tags transmitted summaries of the archived data via satellite uplink through the Argos system. Archived data were summarized into 24 h temperature–depth profiles across eight different depths distributed between the minimum and maximum depths recorded for the 24 h period. Tags also transmitted depth data as a time-series of 150 s intervals.

(iii) . Stable isotope analysis

Frozen fin samples were dried in a food dehydrator at 60°C for 24–48 h, and then powdered, and 0.5–1.5 mg of each sample was packed into a tin capsule. Isotopic analysis was performed at the University of California Davis stable isotope facility (Davis, CA, USA). δ¹³C and δ¹⁵N values are reported as per mille (‰) relative to international standards Vienna Pee Dee Belemnite (V-PDB) and atmospheric nitrogen, respectively.

(c) . Data analysis

See the electronic supplementary material for additional detail of model specifications.

(i) . Presence and habitat use

To examine long-term seasonal presence of billfish near the tagging location, daily recreational billfish catch records from Tropic Star Lodge were obtained for 2010–2020, including days in which fish were targeted but not caught. Because a catch depends on the angler's ability to bring the fish to the boat, we standardized fish raises (or sightings) by the number of boats fishing to determine daily sightings per unit effort (SPUE) [18], averaged per month. Then, using daily location estimates from the satellite tracking data, distance to the tagging location and distance to the nearest coast for each day were compared among species using a general linear mixed model (GLMM) to determine how fish movements compared with seasonal SPUE trends.

Vertical movement patterns were compared at a range of temporal resolutions among species. Using transmitted time-series data, hourly mean depth was calculated for each fish. To determine how mean depth varied throughout the 24 h period, seasonally, and annually among species, we used generalized additive mixed models (GAMM) in the R package mgcv [42], with mean depth as the response variable and various combinations of species, hour of day, season (wet or dry), the interaction between species and season, and tagging year as explanatory variables, with individual set as a random factor.

To determine how individual dive characteristics varied among species, we used linear mixed effects models (LMEs). Dive characteristics were obtained from the transmitted depth time-series data using the diveMove package in R [43]. This package identifies dives below a specified depth (we used 10 m) in a time-series, and for each dive identifies dive phases (i.e. descent, bottom and ascent). While ‘dives’ are not always present in pelagic fishes that are not required to return to the surface [44], in general the marlin and sailfish here displayed directed movements away from the surface (less than 5 m), and return to the surface after spending time at depth (electronic supplementary material, figures S1–S3). As such, the marlin and sailfish diving behaviour analysed here was well suited to the diveMove package, which was originally designed for marine mammal diving behaviour. Specific behaviours of interest (response variables, modelled separately) were number of dives per day, mean depth of the bottom phase, maximum dive depth, bottom phase duration, total dive duration and the ‘distance’ covered during the bottom phase of dives (calculated as the sum of absolute depth differences while at the bottom of each dive). Each response variable was summarized (e.g. total, mean, maximum, depending on the variable) for each fish for each day. LMEs were fitted in R using the nlme package [45] with restricted maximum likelihood, with multiple comparisons of means performed via Tukey contrasts.

Finally, we examined the horizontal and vertical space use concurrently among the three species using three-dimensional utilization distributions (3D UDs) calculated with the ks R package [46]. Core (50%) and extent (95%) 3D UDs were calculated both seasonally and overall for each species and compared to determine the amount of spatiotemporal overlap in 3D habitat use. To account for biases in spatial location density associated with variable track lengths and shorter tracks near the tagging location (via tag shedding), we followed a time weighting procedure outlined in Queiroz et al. [47]. Under this weighting scheme, individual location estimates early in the track received a lower weight than later locations. Calculated 3D UDs were therefore more representative of the actual distributions, and less affected by tag loss or a spatial bias towards deployment location. To determine a multiplier for the smoothing factor matrices, we used methods analogous to Simpfendorfer et al. [48] and Bubley et al. [49], and the plug-in bandwidth selector was used with a multiplier of 5 being applied to all smoothing factor matrices. The total volume of overlap between the core and extent 3D UDs was then calculated, and then divided by the volume of the respective species' probability contours to obtain a proportion of overlapping UDs for each species.

(ii) . Stable isotope analysis

Despite being captured over multiple years, all individuals within a species were grouped to assure robust sample sizes for subsequent statistical analyses, and the long isotopic integration time of fin tissue likely incorporates trophic behaviour across multiple seasons [50]. We first examined differences in the distribu­tion of δ¹³C and δ¹⁵N values among billfishes using Kruskal–Wallis tests owing to unequal sample sizes and variances. Post hoc multiple comparisons between species’ means were evaluated using a Dunn test with p-values adjusted using the Holm method. We then calculated δ¹³C range (CR), δ¹⁵N range (NR) [51] and standard ellipse area (SEA; [52]). SEA estimates were calculated using the SIBER R package [52] using a maximum likelihood approach and represent a bivar­iate estimation of trophic niche width. We also calculated small sample size corrected (SEA_C), and Bayesian estimates of SEA (SEA_B (median values)). Bayesian estimates were run over 10 000 iterations trimmed by the first 1000 [52].

Trophic niche overlap between species was determined using the nicheROVER R package [53], which uses Bayesian methods to calculate the probability of the size of the isotopic niche area of species A inside that of species B, and vice versa, and is not sensitive to variations in sample size [53]. Overlap estimates were run for 10 000 iterations and incorporated 95% of the niche region size.

3. Results

In total, 47 BUM, 55 BAM and 75 SFA were caught and either sampled, tagged or both. Of these, 30 BUM, 46 BAM and 30 SFA were tagged with miniPATs. Seventeen tags (16%) did not report, and after discarding tags with short-duration deployments (less than 10 days) or that exhibited patterns of post-release mortality (sinking to the seafloor within 3 days after fish release and remaining on the seafloor until the mortality function of the miniPAT was activated), data were retained for 24 BUM, 23 BAM and 19 SFA (electronic supplementary material, table S1). Cumulatively, there were 7602 days of tracking data (5970 dry season, 1632 wet season), with mean (±s.d.) days at liberty of 120 ± 69, 115 ± 65 and 114 ± 53 days for BUM, BAM and SFA, respectively (electronic supplementary material, table S1). Over the course of the deployments, the average straight-line distance (the distance from tagging to pop-up location) travelled was 609 ± 587, 327 ± 228 and 581 ± 402 km for BUM, BAM and SFA, respectively. Length of the reconstructed tracks averaged 2681 ± 1754 for BUM, 2097 ± 1077 for BAM and 2384 ± 1318 km for SFA. No difference was detected among the days at liberty (one-way ANOVA; F_2,63 = 0.05, p = 0.95), straight-line distance travelled (F_2,63 = 2.87, p = 0.06), or the cumulative distance travelled (F_2,63 = 0.99, p = 0.37) among species. However, species differed in their daily distance to the coast (GLMM, F_2,63 = 8, p < 0.001). BAM remained closest to the coast throughout their tracking periods (121 ± 114 km; BUM & BAM, est. = 18.9, s.e. = 4.8, Z = 3.9, p < 0.001; SFA & BAM, est. = 12.4, s.e. = 5.2, Z = 2.4, p = 0.03), but BUM and SFA were not different (BUM 228 ± 256 km, SFA 173 ± 132 km; p = 0.19).

All three species remained in the ETP across all years and show cyclical patterns in recreational catch rates, suggesting high horizontal overlap among species over long time periods (figure 1; electronic supplementary material, figure S4). BAM showed fidelity to the ETP throughout the year, with the mean distance from the tagging location remaining below 500 km for all tagging years (electronic supplementary material, figure S4). BUM and SFA mean distance from tagging location showed an increase followed by a decrease, indicating a return to the tagging region after approximately 125 days at liberty (electronic supplementary material, figure S4a). All three species have high SPUE in January and February, followed by a decrease in March, and subsequent increase beginning in June and July. BUM and BAM show a steady increase in SPUE from July to December, whereas SFA peak in July, but remain high through the end of the year and into the next (electronic supplementary material, figure S4b).

Figure 1.

Reconstructed locations of (a,d) black marlin, (b,e) blue marlin and (c,f) sailfish in the wet and dry seasons, along with the mean annual meridional (g) and zonal (h) sections of percentage oxygen saturation at the red-outlined latitudinal and longitudinal bands indicated in the inset plots, and the seasonal (wet versus dry) mean (± s.e.) temperature (i) and percentage oxygen saturation (j) at 5 m depth intervals at the tagging site. Data for (g–j) were obtained from NOAA's National Centers for Environmental Information World ocean atlas 2018 [54]. Note that the vertical black line in (h) is due to the presence of an oceanic island.

(a) . Vertical and horizontal habitat use

Average depth of each species was best explained by hour of day and season. The inclusion of tagging year did not improve model fit, so it was removed (electronic supplementary material, table S4). The overall best-fit model included an interaction between species and season within the hour of day smoother (65.5% deviance explained; electronic supplementary material, table S4). All species displayed patterns consistent with diel vertical migration during both seasons, where fish were shallow at night and deeper during the day (figure 2). During the wet season, all species showed substantial overlap in their predicted mean depth throughout the 24 h cycle (figure 2a), with only slight differences in hourly mean depth response curves (figure 2c–e). However, during the dry season, predicted mean daytime depths were non-overlapping (figure 2b). During the dry season, the relative timing and strength of this diel pattern was different for each species (electronic supplementary material, figure S5). Whereas SFA reached their peak (were deepest) in the early morning hours (approx. 6.00–8.00), BUM and BAM did not peak until the late afternoon (approx. 15.00–17.00; figure 2b; electronic supplementary material, figure S5). BUM and SFA were significantly deeper than BAM during all daylight hours (figure 2b,f,g). SFA were deeper than BUM in the morning from sunrise until approximately 10.00, and then at approximately 16.00 BUM average depth became deeper while SFA got shallower (figure 2b,h; electronic supplementary material, figure S5). As such, there became a large difference in mean depth of BUM and SFA from 16.00–20.00 (figure 2h), when BUM were deepest until sunset, and then mean depth was not different between BUM and SFA at night (figure 2h).

Figure 2.

Seasonal diel vertical behaviour of billfish in the eastern tropical Pacific (ETP). Grouped boxplots of black marlin (BAM), blue marlin (BUM) and sailfish (SFA) hourly depth overlaid with the predicted hourly mean depth (solid black line) and 95% confidence intervals (coloured shading) from the GAMM output during the wet season (a) and dry season (b). The difference between species of the GAMM partial effects response curves during the wet season (c–e) and the dry season (f–h). Mean depth was determined to be significantly different between species where the confidence intervals do not overlap 0 (horizontal dashed line) in (c–h).

Variable diving behaviour among species during periods of increased prey abundance (i.e. dry season, owing to increased upwelling) indicates that each species used the water column in a different way, which reflects interspecific differences in foraging behaviour and resource use. This is in contrast to diving behaviour in the wet season, when there was no difference between species (electronic supplementary material, table S3), a time when the thermocline and OMZ become deeper (figure 1i,j), and productivity in the region declines. To investigate differences in diving during the dry season, 18 889 BUM, 17 356 BAM and 34 229 SFA dives were analysed. Linear mixed effects model predictions showed that BAM dived to shallower depths than either BUM or SFA (figure 3a,b; electronic supplementary material, table S4). While BAM spent roughly the same amount of time in the bottom phase of a dive as SFA, they covered less distance than either BUM or SFA (figure 3c,d). BAM performed a similar number of dives every day to BUM; however, each dive was on average shorter in duration than BUM dives, but similar to SFA dives (figure 3e,f). SFA mean bottom depth and mean maximum depth were similar to BUM, but SFA performed roughly twice as many dives per day as either BUM or BAM, with dives being shorter in duration (figure 3e,f). Although SFA dive duration and bottom time were shortest, SFA bottom ‘distance’ was not different from BUM, and greater than BAM, indicating SFA were more vertically active during their brief time at the bottom of a dive (figure 3). While SFA performed the greatest number of dives throughout the day, more dives were performed and bottom distance covered was greatest in the early morning hours, and generally decreased until sunset (figure 3i). By contrast, number of dives and bottom distance covered during BUM dives was low in the morning and peaked in the late afternoon (figure 3h).

Figure 3.

Best-fit linear mixed effect model predictions (± 95% CI) of black marlin (BAM), blue marlin (BUM) and sailfish (SFA) daily dive behaviours for all dives extracted during the dry season from 150 s time-series depth data (a–f), paired with diel diving activity of bottom distance covered during a dive, coloured by the number of dives performed in each hour of the day (g–i). Only dives >10 m were considered. In a–f, statistical significance from other species is indicated by ‘*’.

In general, BUM used the largest volumes of water as they moved furthest away from the tagging location (table 1; figures 1 and 4; electronic supplementary material, figure S4). While BAM horizontal movements were greater than BUM movements in the wet season, BUM average maximum depth (102.1 ± 59.7 m) was nearly double BAM maximum depth (59.3 ± 26.8 m), leading to a larger volume of water used. Similarly, although BUM core and extent UDs extend further offshore than those of SFA, the volume of water used is similar for the core habitat and is comparable between the extent UDs owing to SFA diving to deeper maximum depths than BUM (figure 4).

Table 1.

Volumes of three-dimensional utilization distributions for blue marlin (BUM), black marlin (BAM) and sailfish (SFA) during the wet season (April–November), dry season (December–March) and overall. Volume is given in km³. 50%, Core use area; 95%, extent of volume use.

	wet		dry		overall
	50%	95%	50%	95%	50%	95%
BUM	1.12 × 104	8.75 × 104	2.58 × 104	3.30 × 105	2.57 × 104	2.87 × 105
BAM	4.11 × 103	4.61 × 104	4.93 × 103	5.26 × 104	4.93 × 103	5.77 × 104
SFA	9.08 × 103	5.24 × 104	1.89 × 104	1.88 × 105	1.89 × 104	1.93 × 105

Figure 4.

Overlap between two-dimensional 50 and 95% utilization distributions (2D UDs; (a–c)) and 3D core (50%; (d–f)) and extent (95%; g–i) UDs for blue marlin (BUM, blue), black marlin (BAM, black) and sailfish (SFA, orange) in the eastern tropical Pacific (ETP). UDs are shown for the wet season (a,d,g), dry season (b,e,h) and overall (c,f,i). The depth dimension used was maximum daily depth in (d–i). All figures are plotted on the same 3D spatial scale (x,y,z). 2D UDs are shown for spatial reference.

Core (50%) and extent (95%) volume use for all species showed a large amount of 3D habitat overlap both seasonally and overall (figure 4; electronic supplementary material, figure S6). In the wet season, core habitat overlap ranged from 28 to 71%, while extent habitat overlap ranged from 37 to 81% (electronic supplementary material, figure S6). Although the extent UD of each species was generally larger in the dry season (likely an artefact of more tracking days in the dry season), core UD sizes were similar between wet and dry seasons (table 1). Maximum proportion of habitat overlap tended to be larger in the dry season for both the core and extent UDs, ranging from 14 to 87% overlap, and 16 to 100% overlap, respectively (electronic supplementary material, figure S6). Overall, BAM 3D core and extent habitat use were nearly completely overlapped by both BUM and SFA habitat use areas, but BAM overlapped relatively little with either BUM or SFA 3D habitat use (20–32%; figure 4). However, SFA and BUM had high 3D core habitat overlap in the wet (44–69%) and dry (64–80%) seasons, and overall (70–77%; figure 4; electronic supplementary material, figure S6).

(b) . Stable isotope analysis

Stable isotope analysis showed a significant difference and low trophic overlap between the isotopic niches among this predator guild. Carbon isotope values ranged from −16.0 to −12.7‰ and nitrogen isotope values ranged from 10.8 to 14.9‰ (figure 5a). Carbon and nitrogen isotope values differed among species (Kruskal–Wallis test: carbon, χ² = 61.3, d.f = 2, p < 0.001; nitrogen, χ² = 46.5, d.f = 2, p < 0.001). Specifically, mean δ¹³C values for SFA were lower than BUM (Dunn test, Z = 5.8, p < 0.001) and BAM values (Z = 6.7, p < 0.001). δ¹⁵N values differed between all three species; SFA were lowest (SFA & BUM, Z = 2.8, p = 0.005; SFA & BAM, Z = 6.7, p < 0.001) and BAM were highest (BAM & BUM, Z = 3.6, p < 0.001).

Figure 5.

(a) Core trophic niche (40% of data, solid lines) and total trophic niche (95% of data, dashed lines) estimates for black marlin (black), blue marlin (blue) and sailfish (orange). (b) Standard ellipse area, SEA_B, estimates with 50, 75 and 95% credible intervals (the X represents maximum-likelihood estimated SEA_C).

Carbon range estimates were greatest for SFA, and the same for BUM and BAM (table 2). However, nitrogen range esti­mates were the lowest for SFA, and greatest for BAM (table 2). SEA estimates for BAM and BUM were greater than those for SFA (table 2). SEA_C and SEA_B esti­mates exhibited similar values (table 2 and figure 5b). Low total trophic niche overlap was observed between the three species (less than or equal to 30.8% across all species comparisons), the lowest overlap occurred between SFA and BAM (less than 1%) and the greatest overlap occurred between BUM and BAM (23.4–30.8%; table 3; electronic supplementary material, figure S7).

Table 2.

Summary statistics (mean (s.d.)) of black marlin (BAM), blue marlin (BUM) and sailfish (SFA) estimated weight (kg) and stable isotope values, with isotopic niche metrics including δ¹³C and δ¹⁵N ranges (‰); standard ellipse area (SEA), small sample size corrected SEA (SEA_C) and Bayesian estimates of SEA (SEA_B; median values shown).

species	n	est. weight	δ13C (‰)	δ15N (‰)	δ13C range	δ15N range	SEA (‰2)	SEAC (‰2)	SEAB (‰2)
BAM	17	175 (65)	−13.5 (0.49)	13.3 (0.57)	1.5	2.5	0.85	0.91	0.88
BUM	21	135 (65)	−13.9 (0.38)	12.1 (0.52)	1.5	2.1	0.61	0.64	0.62
SFA	45	35 (5)	−15.1 (0.44)	11.8 (0.23)	1.9	0.8	0.29	0.3	0.3

Table 3.

Total trophic niche overlap (%) between black marlin (BAM), blue marlin (BUM) and sailfish (SFA) in the ETP. Two different overlap estimates are presented for each species comparison based on whether the total trophic niche of species A is being compared to species B, or vice versa.

		species B
		BAM	BUM	SFA
species A	BAM	—	23.4	0.1
	BUM	30.8	—	15.9
	SFA	0.4	22.8	—

4. Discussion

Using a decade of recreational fishery catch data in combination with several years of horizontal tracking, high-resolution vertical movements and stable isotope data, we provide support for the ecological tenet of niche partitioning in this difficult-to-study large pelagic predator guild. We reveal a high amount of long-term and 3D habitat overlap but conclude that billfishes of the ETP temporally partition (both diel and seasonal) the available vertical habitat, potentially leading to the low trophic overlap observed here. While differences in life history, body size, morphology and physiological constraints are important factors in facilitating habitat partitioning in many pelagic species and could be playing a role in the patterns observed here [20,55,56], the hypoxia-based habitat compression in the ETP restricts these three billfishes to the same core habitats and seasonal prey productivity. As such, we believe the seasonal changes in diving behaviour and trophic niche separation described here are not likely driven by physiological or morphological constraints, but more likely represent behavioural mechanisms used to alleviate competition among this predator guild in a vertically compressed habitat.

Recreational fishing records demonstrate a strong, long-term seasonal pattern of co-occurrence among these billfishes in the ETP, and we documented a high proportion of 3D spatial overlap between them with multi-year tracking data. Core use areas were relatively small and showed high overlap among species (20–91%) compared with the 4 and 38% of 3D core habitat overlap between BUM and SFA, and SFA and white marlin (Kajikia albida), respectively, in the western North Atlantic [49]. Where vertical habitat was not limited by a shallow OMZ, such as in Bubley et al. [49], BUM used deeper habitat than SFA throughout the tracking period, and the authors suggest this is due to BUM being less constrained by temperature or distance from shore than sailfish. However, in the ETP, where a shallow OMZ exists, depth use is highly similar among these species, leading to a higher proportion of overlapping habitat ([27]; this study).

The tendency for tagged fish to remain near Panamanian waters after long periods of time is unusual for these highly mobile species. For example, BUM tagged in the North Pacific were 1276 ± 2191 km from the tagging location after 112 ± 75 (median ± inter-quartile range; n = 69) days at liberty [57]. Similarly, BAM tagged near the Great Barrier Reef were 2146 ± 1226 km (mean ± s.d.; n = 42, maximum displacement of 10 623 km) from the tag location after 8–180 days [26,58]. These distances, despite a similar number of or fewer days at liberty than our tagged fish, are roughly twice and seven times the mean displacement distances we found for BUM and BAM (609 ± 587 and 327 ± 228 km, respectively). SFA were the most site-attached of the 10 billfishes reviewed by Braun et al. [59], and their displacement distance reported here (581 ± 402 km) was similar to or slightly greater than those of other SFA studies in the Atlantic [24,60]. The low overall displacement distances reported here for BUM, BAM and SFA suggest fidelity to Panamanian and adjacent waters and thus high potential for spatial and temporal overlap.

Pelagic species' vertical movements are often hypothesized to be driven by foraging behaviour [36,61]. In the ETP, shoaling of the thermocline and oxycline during the dry season further compresses the already restricted (i.e. year-round) vertical habitat available to these predators, likely evoking increased interspecific competition for available prey. As such, the seasonal partitioning in vertical habitat documented here would serve to reduce competitive overlap within this pelagic predator guild. Indeed, during the dry season, when coastal productivity is at its peak, fishery SPUE is high for all species, and core 3D spatial overlap is high, we documented a substantial separation of mean depth use that was consistent across years. This vertical behaviour is in stark contrast to many previous studies of these species. For example, SFA are generally the most surface-oriented billfish, spending a greater proportion of time in water <10 m than BUM [27,49,59]. Additionally, similarly sized BAM (approx. 100–250 kg) had a mean hourly daytime depth of approximately 70–80 m in the western Pacific [26], compared with a mean hourly daytime depth of 12 ± 15 m for BAM in our study. During the wet season, BUM, BAM and SFA used roughly equivalent depths throughout the day, similar to that observed for BUM, SFA and white marlin in the western North Atlantic [49], where seasonal changes in the vertical dimension are not as stark and vertical habitat compression does not exist. The lack of vertical partitioning when coastal productivity is lower and vertical habitat compression is reduced or not a factor [49] suggests sympatric billfish predators are able to homogenize their diving behaviour and use similar depths throughout the day.

We identified both seasonal and diel differences in vertical habitat use and found key differences in the diving behaviour driving the differences in mean depth use observed here. For example, although SFA predicted hourly daytime depth was deeper than BUM during the dry season, both species dived to the same mean and maximum depths. However, SFA's deeper average depths were driven by SFA performing roughly twice as many dives per day as, but much shorter-duration dives than BUM. Variation in dive duration and frequency may be explained by morphological differences between BUM and SFA, as SFA are smaller and more laterally compressed, limiting their capacity to retain body heat when exploring colder depths for extended periods [62]. However, BUM and BAM tagged in this study were of similar body size (table 2; electronic supplementary material, table S1); thus differences in depth use cannot be explained by morphological differences between them, although differences in their physiology cannot be ruled out as marlin physiology is poorly understood. In contrast to the concept of thermal inertia allowing larger fish to remain in colder temperatures for extended periods of time [62], it has been suggested that vertical movements of tunas and billfishes are limited by the temperature of the heart rather than the size of the individual fish, because the temperature of the heart will closely follow ambient water temperature [63]. Under this scenario, one would expect that all individuals experiencing similar conditions, regardless of size, would be restricted to a similar depth. However, Williams et al. [26] documented significant differences in depth use in relation to body size among black marlin in the southwest Pacific, suggesting that differences in diving behaviour are more likely to be a function of their ecology rather than differences in physiology, as would be suggested by the findings presented here. Furthermore, SFA performed most of their dives in the early morning hours, during which they covered more bottom distance (a proxy for prey searching; see [22,64]), whereas BUM performed more dives and covered more bottom distance in the late afternoon. Predators may adjust their activity schedules to match periods when their prey are most active or vulnerable [65]. As such, crepuscular periods provide optimal foraging conditions for both BUM and SFA, because they coincide with periods of movement of vertically migrating prey [66]. While BAM were using a shallower portion of the water column, BUM and SFA were exploiting the same depths, but were potentially alleviating interaction with one another via diel partitioning (sunrise versus sunset) of foraging activity. These findings suggest that average depth by itself should not be used as an indication of habitat partitioning, and provide new insights into the mechanisms underpinning resource partitioning among large diving pelagic predators.

Not only did we identify seasonal partitioning of vertical habitat, but over long time-scales we documented strong interspecific differences in foraging as well. This is in stark contrast to Guillemin et al. [67], who documented no difference in mean δ¹³C and δ¹⁵N values, and high trophic niche overlap (50–90%, compared with 0–30% in our study) between BUM, BAM and striped marlin off southeast Australia, a region without a shallow OMZ. The lack of trophic separation in billfishes of the western Pacific leads the authors to conclude that resource competition is not a major factor driving the demography of these species in the western Pacific [67], and provides support for the influence of hypoxia-based habitat compression on the diet partitioning observed here in the ETP.

Although SFA are reported to be the most coastal of the billfishes [59], our tracking data demonstrate that BAM were the most coastally inclined, while SFA used more offshore habitats, further indicated by SFA being the most depleted in δ¹³C of the three species. However, SFA also had the smallest nitrogen range and SEA, indicating a small amount of trophic diversity and foraging on functionally similar prey groups [51]. The higher δ¹³C values of BUM and BAM suggest that they forage differently, potentially consuming prey from more coastal regions associated with the strong seasonal upwelling events of the ETP, which favour the enrichment of primary production [68]. Furthermore, the increased δ¹⁵N values, larger nitrogen ranges and greater SEAs of BUM and BAM indicate feeding on higher trophic levels and more generalist foraging behaviour than SFA, as BUM and BAM are much larger and have a larger mouth gape, and are able to consume a wider variety of prey items than SFA [69].

5. Conclusion

In contrast to other multi-billfish studies done in regions without vertical habitat compression, we document substantial horizontal and 3D space overlap, behavioural and temporal variation in diving behaviour, and separation of trophic niches among this predator guild of the ETP. Thus, these results highlight the power of complementary methodologies to gain insights into movements and behaviour in mediating niche partitioning among sympatric predators. In this era of unprecedented environmental change and habitat destruction, shifts in the geographic distribution of many species are already evident and are only predicted to increase [70], with oxygen limitation and altered oxygen distribution suggested to be major factors contributing to further changes in large pelagic predator horizontal and vertical distributions [35,71]. The data we have presented here demonstrate that the billfish complex in the ETP uses a suite of behavioural mechanisms to partition a vertically restricted habitat. While predator responses to vertical habitat compression in other ecosystems will likely be species-specific, partitioning within this ETP predator guild suggests that sympatric predators not currently experiencing OMZs may be able to behaviorally adapt and successfully respond to predicted widespread habitat compression in the future.

Ethics

All fish were tagged under permit from the Ministerio de Ambiente, Republica de Panamá (SE/A-64-19). All procedures were approved by Nova Southeastern University Institutional Animal Care and Use Committee (2019.04.MS1).

Data accessibility

Data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.tdz08kq4f [72].

Supplementary material is available online [73].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

R.K.L.: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing—original draft, writing—review and editing; J.J.V.: conceptualization, project administration, supervision, writing—review and editing; B.M.W.: conceptualization, project administration, supervision, writing—review and editing; M.S.S.: conceptualization, funding acquisition, project administration, resources, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This study was supported by the Guy Harvey Foundation, the Guardians of the Eastern Tropical Pacific Seascape donor group, R. Vergnolle, Nova Southeastern University Halmos College of Arts & Sciences, the Gallo-Dubois Scholarship, Fish Florida Scholarship, and a Batchelor Foundation Scholarship to R.K.L.