Trophic Structure and Community Stability in an Overfished Ecosystem

See allHide authors and affiliations

Science  16 Jul 2010:
Vol. 329, Issue 5989, pp. 333-336
DOI: 10.1126/science.1190708


Since the collapse of the pelagic fisheries off southwest Africa in the late 1960s, jellyfish biomass has increased and the structure of the Benguelan fish community has shifted, making the bearded goby (Sufflogobius bibarbatus) the new predominant prey species. Despite increased predation pressure and a harsh environment, the gobies are thriving. Here we show that physiological adaptations and antipredator and foraging behaviors underpin the success of these fish. In particular, body-tissue isotope signatures reveal that gobies consume jellyfish and sulphidic diatomaceous mud, transferring “dead-end” resources back into the food chain.

The northern Benguela upwelling system on the southwest coast of Africa has historically been one of the world’s most productive ocean areas, supporting industrial-scale fisheries (1, 2). However, overexploitation coupled with unfavorable environmental conditions drove the collapse of the sardine fishery in the late 1960s (3, 4). Loss of the filter-feeding sardine (Sardinops sagax) changed the structure and function of the ecosystem, and it became dominated by jellyfish, bearded goby (Sufflogobius bibarbatus), horse mackerel (Trachurus trachurus capensis), and hake (Merluccius capensis) (5). The massive increase in jellyfish biomass after the collapse (2, 6) has been regarded as a trophic dead end (7). The loss of the sardine has forced many higher trophic animals, including predatory seabirds, mammals, and fish, to switch to feeding almost exclusively on the bearded goby (3, 8, 9). Thus, the bearded goby has come to play a key role in the food web of northern Benguela [(8, 5) and see also figures 8 and 9 of (5)]. Paradoxically, despite increased predation pressure, the goby population thrives, but the reasons for this have remained obscure.

The gobies live in a hypoxic environment. As in other nutrient-rich upwelling areas, high productivity along the inner shelf area has created extensive diatomaceous mud belts, where intense decay processes create hypoxic conditions, with high concentrations of hydrogen sulphide (H2S) and methane in the sediment surface layers (10). Sediments in these hypoxic conditions (<1 μM dissolved oxygen, which corresponds to <0.4% oxygen saturation at 12C°) cover over 8944 km2 and represent 50% of the shelf area (11). 7000 km2 are sulphidic (12) and can be dominated by white mats of large sulphide-oxidizing bacteria (13) and chemolithotrophic bacteria (12). Diffusion and methane-driven ebullition of H2S gas from the sediment have been involved in massive killing events of fish and invertebrates (10, 14, 15). For the Benguela upwelling system, sulphidic water column conditions can be dated back to the early 1900s (15, 16), but their frequency seems to have increased in the past three decades (17). One of the unresolved questions is how the inhospitable benthic conditions, ranging from hypoxic to anoxic, link with and affect the pelagic ecosystem, and what role hypoxic-tolerant fish play in the transfer of energy from the biogeochemical cycle in the bottom sediments to the pelagic food web [(18), p. 12].

We conducted a cross-shelf survey off Namibia in April 2008 (23°20′S, 14°12′E to 23°40′S, 13°15′E). We combined direct ecosystem observations made with acoustic methods, trawl sampling, and environmental data from the water column and sediments with onboard controlled experiments on the behavior and physiology of S. bibarbatus (19). We further analyzed diurnal and depth variations in the stomach contents of bearded gobies and used stable isotope analysis on goby muscle and its potential prey species (19). This multidisciplinary approach allowed us to determine diel movement and feeding patterns for the bearded goby and its trophic links with other organisms, and to obtain evidence for its behavioral and physiological adaptations.

An acoustic transect across the shelf revealed a layer apparently void of life, extending from the bottom up to 20 to 60 m above the bottom (fig. S1). This empty pocket corresponds with water masses having oxygen levels below 10%. Environmental data showed a gradually decreasing temperature and oxygen level from the surface to the bottom (fig. S2; O2 is also indicated in Fig. 1, A and B). The highest densities (average catch, 40 kg/hour) of S. bibarbatus occurred over the inner shelf, where the seabed is a thick layer of diatomaceous mud characterized by millimolar concentrations of total sulphide (1 to 2 mmol) in the topmost 3 cm [(11) and fig. S3]. Here, oxygen levels were very low (4.7% oxygen saturation), and no other vertebrates were present. Acoustic measurements (Fig. 1, A and B) and trawls, however, confirmed that S. bibarbatus were on the seabed during daylight hours, as has previously been shown by video photography [(20) and fig. S4].

Fig. 1

Acoustic record (at 38 kHz) (A) showing gobies ascending from the sediment in the afternoon (time interval on the x axis is UTC 16:30 to 18:00 time), joining an acoustic SL of jellyfish, and (B) returning to the sediment in the morning (time interval on the x axis is UTC 03:50 to 05:20). Sequential near-bottom pelagic trawling produced no catches before the ascent and a monospecific catch of gobies during the ascent. The vessel was stationary during the acoustic observations; therefore, many echoes were returned from each fish, depicted by lines. The average oxygen level is given at 10-m depth intervals. Temperature decreased gradually from 18°C at the surface to 13°C above the bottom.

Acoustic and trawl data showed that gobies ascend from the bottom in the evening to join an existing acoustic scattering layer (SL) in the water column (Fig. 1A) and then return to the bottom in the early morning (Fig. 1B) (19). The pelagic SLs were dominated by two species of large jellyfish [Aequorea forskalea and Chrysaora fulgida (= hysoscella)] whose biomass is currently estimated to exceed that of finfish off Namibia (2). S. bibarbatus are found significantly more often with jellyfish than other fish species are [data from >11,000 samples of pelagic fish landings at Walvis Bay, 1991–2006 (Kruskal-Wallis test, H = 18.18, P = 0.001, n = 40 observations)], and they are six times more often associated with jellyfish in commercial catches than is their predator, the horse mackerel (T. trachurus capensis) (Table 1), suggesting that S. bibarbatus may choose to associate with jellyfish when in the water column.

Table 1

Temporal changes in the frequency at which the dominant species of pelagic fish were caught with jellyfish, expressed as monthly variation in total number of catches of each fish species together with jellyfish, divided by the total number of catches of the same species when jellyfish were not present. Data were collected from randomly selected samples of the landings of the pelagic fleet in Walvis Bay, Namibia, for the period 1990–2007. Only months where the total number of samples (N) was greater than 100 are shown. Information for juvenile hake (M. capensis) was not available, as this species is not a routine part of the pelagic fishery.

View this table:

Goby and horse mackerel (a goby predator) aversion to jellyfish (C. fulgida) was examined by allowing individual fish to swim between two chambers, one with and one without jellyfish during 300-s trials (19). Horse mackerel strongly avoided jellyfish and either fled the jellyfish chamber within 18 ± 11 s (mean ± SEM) or never moved across the mesh divider to associate with jellyfish. In contrast, gobies took significantly (two-sided t test, n = 14 horse mackerel and n = 13 gobies, P < 0.001; df = 25) longer (210 ± 37 s, mean ± SEM) to leave the jellyfish chamber and frequently moved between the two chambers.

Habitat choice experiments showed that gobies preferred to stay on diatomaceous mud as compared to a sand alternative (t test on normalized arc-sine–transformed data: t = 3.29; P = 0.0017, n = 30 individuals) (19). We also observed fish burrowing into the hypoxic diatomaceous mud, especially if threatened or disturbed. These observations support the acoustic and trawl data showing that gobies associate with jellyfish but their predators do not, and also that gobies choose to rest on, and hide within, mud during the day.

S. bibarbatus was found to have an extremely low critical oxygen level {[O2]crit = 5.3 ± 0.3% (mean ± SD) of air saturation, which is the lowest [O2] at which resting oxygen uptake can be maintained, marked with a dotted line in Fig. 2A}, allowing it to sustain aerobic metabolism even in deeper waters below the main SL where oxygen saturation is ~10% (Fig. 1, A and B). Although S. bibarbatus built up an oxygen debt during anoxia (Fig. 2C), probably related to lactate oxidation, the rate of lactate accumulation in the blood declined markedly after 1 hour (Fig. 2D), suggesting metabolic depression. Metabolic depression is also indicated by a suppressed ventilation rate at [O2] below [O2]crit (Fig. 2B) (19).

Fig. 2

Physiological studies on S. bibarbatus [(A) to (D) and (F)] and M. capensis (E) (19). (A) Gobies maintained a constant oxygen consumption rate until [O2]crit of 5.3 ± 0.3% (mean ± SD) of air saturation (~0.3 ml of O2 liter−1) was reached. [O2]crit is the lowest [O2] in which the animal is able to maintain its resting rate of O2 consumption (35) (n = 7 individuals). Different symbols represent individual fish. (B) Ventilation rate increased in response to falling water [O2] until [O2]crit, then ceased (the regression line was obtained by locally weighted scatterplot smoothing) (n = 7 individuals). (C) Representative trace showing that 3 hours of exposure to anoxia caused an oxygen debt, because post-anoxia oxygen consumption was increased by ~35% (n = 6 individuals). (D) Accumulation of blood lactate during 3 hours of anoxia (O2 < 0.5% air saturation). The rate of increase slowed after 1 hour, indicating metabolic depression. Significant differences (P < 0.05; one-way analysis of variance; Student-Newman-Keuls post-test) between time points are indicated by dissimilar letters. n = 33 fish in total and 3 to 6 fish at each time point. Values are means ± SEM. (E) Oxygen consumption of gobies exposed to different sulphide concentrations at a normoxic oxygen level (>50% of air saturation) (n = 5 individuals). (F) Isolated, spontaneously contracting heart preparations of S. bibarbatus (n = 6 hearts) successfully recovered from 20 min of anoxia, whereas hearts of its predator M. capensis (n = 6 hearts) were irreversibly damaged. Asterisks indicate a statistically significant difference (P < 0.05; one-way repeated measures analysis of variance performed on non-normalized data; Student-Newman-Keuls post-tests) from the control normoxic level (0%). Values are means ± SEM.

H2S is a respiratory poison that blocks cytochrome c (21). The rate of oxygen consumption by S. bibarbatus was virtually unaffected by 100 to 200 μM total sulphide [corresponding to 6 to 12 μM H2S (21, 22)], although it became 98% suppressed at a total sulphide level of 500 μM (= 30 μM H2S) (Fig. 2E). These physiological data for S. bibarbatus correspond with known effects on isolated mitochondria (from relatively sulphide-tolerant fish and mammals), in which [H2S] below 6 μM are found to stimulate oxygen consumption (probably because mitochondria use oxygen to detoxify H2S to thiosulphate), whereas H2S levels above 11 to 14 μM inhibit mitochondrial respiration (21, 23). Thus, S. bibarbatus possesses a H2S-sensitive cytochrome c, and its H2S tolerance probably relies on anoxia tolerance, allowing fish to be temporarily independent of mitochondrial respiration through a sufficiently high capacity for anaerobic (glycolytic) adenosine triphosphate production, combined with metabolic depression. Thus. the anoxia tolerance of the bearded goby is extreme, surpassing that of other marine teleosts surveyed in a recent meta study (24), and allows S. bibarbatus to survive the very high [H2S] in the mud where these fish forage and hide (fig. S3).

Hypoxia may impair escape responses, making gobies more vulnerable to predators (25). This was not the case. Touching them with a lever mounted through the lid of the sealed aquaria caused an escape response even after 7 to 9 hours of exposure to oxygen levels below their [O2]crit and a subsequent 4 to 5 hours in complete anoxia (n = 7 individuals) (19). Thus, anoxic gobies remain alert.

Anoxia tolerance in S. bibarbatus and one of its predators, the hake M. capensis, was compared by measuring the performance of excised hearts (19). These experiments showed that the pumping capacity [heart rate times contraction force (26)] of both species’ hearts was reduced by ~80% after 20 min of anoxia. After 40 min of subsequent reoxygenation, only goby hearts recovered to pre-anoxic values, suggesting that the hake hearts had sustained permanent damage (Fig. 2F). Thus, hake are unable to tolerate the hypoxic seafloor, thereby providing the gobies with a refuge from these predators.

Comparative analyses of the stable isotope composition of carbon and nitrogen of S. bibarbatus muscle, A. forskalea, C. fulgida, euphausiids (krill), and sediment show that jellyfish contributed a minimum of 17 to 37% and a maximum of 40 to 60% to the diet of S. bibarbatus (Fig. 3 and fig. S6), significantly more than did euphausiids and sediment (Kruskal-Wallis test, H = 65.34, P < 0.001) (19). Whether gobies scavenge on moribund jellyfish on the sea floor or feed on them while swimming in the SL at night is uncertain. However, in support of a benthic feeding hypothesis, the low stable sulfur isotope values of gobies as compared to those of zooplankton and jellyfish indicated that the gobies derive part of their diet from the sulphidic benthos (fig. S5 and table S1) (19). The isotope results indicated that the gobies may derive 34.2 ± 6.7% (n = 7 individuals) of their diet from the sulphidic benthos (fig. S7) and the associated sulfur-containing bacterial mats of Thiomargarita namibiensis and Beggiatoa spp. present on the benthos (10). The benthophagy interpretation is supported by stomach contents analysis showing that gobies feed on benthic polychaetes (fig. S8) (19).

Fig. 3

Isotope analyses showing maximum and minimum contribution of euphausiids, jellyfish, and sediment to the diet of S. bibarbatus, as determined by a four-endpoint Isosource model (based on carbon and nitrogen) (19). Bars denote mean ± SEM. Bars with identical letters were not significantly different from each other at an α level of 5% (Tukey test, P < 0.05; A. forskalea, n = 11; euphausiids, n = 6; C. fulgida n = 25; mud, n = 5 samples; S. bibarbatus, n = 41).

At night, S. bibarbatus enter pelagic waters. This may be to reoxygenate, as recently shown for sprat (Sprattus sprattus) in hypoxic Norwegian fjords (27), but it may also be for digestion. Gut fullness was significantly higher and stomach contents were less digested in fish ascending from, than in those returning to, the sea floor (fig. S9 and table S2, proportion odds logistic regression, P < 0.001, n = 75 individuals). Delayed digestion due to lack of oxygen is, to our knowledge, a previously unknown phenomenon.

Besides polychaetes, diatoms were the most common item in S. bibarbatus stomachs (fig. S8). Because S. bibarbatus is not a filter feeder [their branchiospines are merely stumps and too far apart to filter diatoms (fig. S10)], our data suggest that they feed partly on the benthos taken directly from the sulphide-rich diatomaceous mud (figs. S7 and S8), similar to Gobionellus sagittula in the Pacific (28). Collectively, our results indicate that, owing to digestion suppression in hypoxic situations (29), gobies feed on the benthos within hypoxic water during the day, and then move to oxygen-richer pelagic waters at night, when predation pressure is lower, to digest, and possibly feed on live jellies.

The marine ecosystem off Namibia has witnessed a number of profound ecological and environmental changes since the collapse of the commercial pelagic fisheries at the end of the 1960s, including a proliferation of jellyfish (2), a change in the fish community structure (9), a possible increase in hypoxia and toxic gas eruptions (17), and consequently a change in the food web dynamics (3). The fact that gobies feed on jellyfish means that the commercially harvested fish species off Namibia are now feeding from a higher trophic level than they were earlier and that here, at least, there is evidence of fishing up the food chain (30). We would expect such a shift in trophic relations to result in lower production and harvests at higher trophic levels, as has indeed been witnessed in the northern Benguela system after the late 1960s (31).

Climate change will probably increase the inflow of hypoxic waters from the tropics and increase coastal upwelling (4), which in an area such as the Benguela could lead to an increase in the frequency of sulphide eruptions and anoxic water masses (17), events that are expected to put harvested species and thus local fisheries under severe stress (32). Such events will relax the predation pressure on the bearded goby population. Owing to the goby’s tolerance of low oxygen and high H2S levels, and their antipredator and feeding strategies, these fish have the possibility of increasing their biomass. The population may act as a stabilizing production unit of the Benguela and between their predator populations and the ecosystem. An increase in goby biomass produced during severely hypoxic years will probably be a mechanism whereby food can be stored within the system for harvested fish populations that return in years with favorable environmental conditions (33).

Organisms able to tolerate such extreme conditions may be successful in many rapidly changed environments (34). The bearded goby, with its remarkable suite of behavioral, physiological, and ecological adaptations, is already playing a critical role in the ecosystem off Namibia, and it is likely to continue to do so into the future.

Supporting Online Material

Materials and Methods

Figs. S1 to S10

Tables S1 and S2


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank the crew of the G. O. Sars; F. Midtøy for assistance; and P. Ellitson, M. Hordnes, R. Jones, R. Amundsen and the rest of the scientific crew. We thank the National Research Foundation of South Africa, the Research Council of Norway, and our home institutions for funding and support. We thank BENEFIT (Benguela Environment Fisheries Interaction and Training), S. Sundby, D. C. Boyer, J. Otto Krakstad, and the crew of the research vessel Dr. Fridtjof Nansen for support with earlier goby cruises, laying the basis for the present study. We thank K. Helge Jensen for statistical support. We appreciate the comments on this manuscript by J. Giske, C. Jørgensen, M. P. Heino, and the anonymous reviewers. Care and handling of experimental animals were performed in accordance with institutional guidelines. J.A.W.S. was a postdoctoral researcher funded by the Natural Sciences and Engineering Reserach Council of Canada at the time when the research was conducted.
View Abstract

Navigate This Article