A reprint from American Scientist the magazine of Sigma Xi, The Scientific Research Society

This reprint is provided for personal and noncommercial use. For any other use, please send a request to Permissions, American Scientist, P.O. Box 13975, Research Triangle Park, NC, 27709, U.S.A., or by electronic mail to [email protected]. ©Sigma Xi, The Scientific Research Society and other rightsholders Down Aquatic Food Webs

Industrial fishing over the past half-century has noticeably depleted the topmost links in aquatic food chains

Daniel Pauly, , and Maria Lourdes Palomares

Roll on, thou deep and dark blue delicate and largely invisible web of mals being caught have been changing Ocean–roll! marine life. in a fundamental way: In essence, Ten thousand fleets sweep over thee If one could don magic spectacles— rather than being satisfied with the big in vain; with lenses that make the murky , people are now routinely going af- Man marks the earth with ruin–his depths of the ocean become transpar- ter many of the little ones too. Al- control ent—and look back several centuries to though the situation is not exactly anal- Stops with the shore an age before widespread abuse of the ogous to eating one’s seed corn, the oceans began, even the most casual ob- result may be equally catastrophic. hen Lord Byron wrote these server would quickly discover that fish Little fish serve the marine ecosystem Wlines, nearly 200 years ago, ex- were commonly much more abundant. in various ways, notably as prey for big- ploitation of the seas was already well And many now-depleted species of ger fish. One straightforward way to de- under way, and some marine species, marine mammals (a list that includes scribe the feeding hierarchy uses the no- such as right whales, were on the path not just right whales but Hector’s dol- tion of . By convention, to extinction. Yet the poet’s verse sup- phins, Caribbean monk seals and marine plants (such as the tiny, drifting poses that the sea is immune from en- Steller sea lions, among others) would, phytoplankton) and various forms of or- vironmental degradation, a common by comparison, appear plentiful. But ganic detritus make up the first trophic misconception then and now. People without such special glasses, the differ- level; herbivores and detritivores are as- can, in fact, ruin the sea as surely as ences between past and present oceans signed to the second trophic level; and they can ruin the land. The only differ- would indeed be hard to discern. the first and higher-order carnivores are ence is that ecological destruction in The opacity of the sea thus accounts, said to occupy trophic levels ranging the ocean is harder to see, particularly in part, for the disregard that most peo- from three to five. when the damage is inflicted on the ple display toward the warnings that More precisely, the trophic level of we and others have been trying to con- such predators can take on non-inte- vey. But the remoteness and impene- gral values, because the diet of these is a professor at the Centre trability of this habitat are not the only animals is commonly somewhat of the University of British Columbia. He received impediments to our efforts. We have mixed. For example, an adult jack his doctorate from the University of Kiel in 1979. also been frustrated by the very rich- swimming around the Caribbean His research focuses on the management of fish- ness of the ocean, in that it provides a might eat equal amounts of herbivo- eries and the modeling of aquatic ecosystems. Villy seemingly never-ending supply of rous zooplankton (tiny plant-eating an- Christensen leads the Fisheries Resources Assess- . Indeed, with only rare excep- imals, which have a trophic level of ment and Management Program at the Interna- tions, the total catch from global fish- two) and small fish that have a trophic tional Center for Living Aquatic Resources Man- eries justs keep going up. level of three (say, because they con- agement in Manila. He earned a Ph.D. in 1992 from the Danish Royal School of Pharmacy. Rain- The premier source for such statisti- sume only the local herbivorous zoo- er Froese is leader of the Fishbase Project at the In- cal information about world fisheries is plankton). This jack would then belong ternational Center for Living Aquatic Resources the Food and Agriculture Organization, to trophic level 3.5. Management. He received his doctorate from the one of the technical organizations of the Unfortunately, the trophic level of a University of Hamburg in 1990. Froese’s current United Nations. Surveys compiled by captured fish is not stamped on its side. work includes the design of large biological data- the FAO are perhaps not as accurate or So it is not an easy task to track the av- bases and the analysis of the information con- as detailed as many scientists would erage trophic level of all the fish caught tained therein. Maria Lourdes Palomares also like, but they are in most cases all that is each year. But after modeling the food works on the Fishbase Project. She earned her doc- available. Two years ago, we examined webs of many marine ecosystems, we torate from the Ecole Nationale Supérieure a large set of FAO statistics on fisheries realized that we had accumulated a Agronomique de Toulouse in 1991. Her research interests cover the comparative studies of the biol- and showed that they do in fact por- great deal of knowledge about the ap- ogy of . Address for Pauly: Fisheries Centre, tend disaster, despite the continuing proximate trophic levels of commonly University of British Columbia, 2204 Main Mall, growth in total amount being taken fished species. All that was needed was Vancouver, B.C., Canada V6T 1Z4. Internet: from the sea each year. The alarm rings to combine these estimates with FAO [email protected] when one realizes that the kinds of ani- catch statistics collected since the 1950s.

© 2000 Sigma Xi, The Scientific Research Society. Reproduction 46 American Scientist, Volume 88 with permission only. Contact [email protected]. James L. Stanfield/National Geographic Image Collection Figure 1. Overview of the Tsukiji in Tokyo hints at the variety of offerings available to seafood lovers. But the mix of species being caught and sold around the world will probably not remain as rich as it is today. The authors have determined that the output of global fisheries—and presumably the compositions of the aquatic ecosystems from which these fisheries draw—are slowly shifting away from top predators toward organisms positioned lower on the food chain. This trend signals that many of today’s fisheries will in time collapse.

Doing so uncovered a systematic pigeonholed—perhaps shoehorned is 120 shift in the composition of global cap- the better metaphor, given that com- 100 global total ture fisheries, with the average trophic plex feeding habits could not readily level showing a slow slide downward be accounted for. Fish that normally 80 over the past half-century. To judge the consume zooplankton were assigned a 60 significance of our result requires an trophic level of three, because zoo- 40 understanding of the procedures we plankton (with a trophic level of two) employed, of the statistics underlying normally consume phytoplankton 20 catch (millions of tons) our analysis and of the various con- (trophic level of one). This procedure 0 cerns that have been raised since our ignores the fact that some zooplankton 1950 1960 1970 1980 1990 study was first published. are carnivorous and thus should be as- year signed a trophic level somewhat higher invertebrates demersal fish On the Level than two, which implies that their pelagic fish discards Historically, the notion of assigning predators should in turn be placed at trophic levels in ecology passed trophic levels above three. Figure 2. Global totals for the amount of fish through two stages. During the first Such complications forced this over- caught during the past half-century provide a misleadingly reassuring view of the state phase, initiated by Charles Elton of the ly simplistic scheme to give way in of the world’s fisheries. The increasing catch University of Oxford and Raymond 1975, when the late Frank H. Rigler, a of small, free-swimming (pelagic) species Lindeman of Yale University, trophic zoologist at the University of Toronto, has masked stagnation in take of bottom- levels were seen as pre-defined cate- published an influential critique of the dwelling (demersal) fish. (Source: United gories into which organisms could be concept. It was then that William E. Nations Food and Agriculture Organization.)

© 2000 Sigma Xi, The Scientific Research Society. Reproduction 2000 January–February 47 with permission only. Contact [email protected]. marine trophic levels Odum (of the University of Virginia’s Department of Environmental Sci- ences) and Eric J. Heald (of the Univer- sity of Miami’s Rosenstiel School of Marine and Atmospheric Sciences) people pointed out that more precise estimates of trophic level could be obtained from actual observations of diet. Their ad- vance, as followed in current practice, treats trophic level as an empirically determinable property of a species— like average size or metabolic rate. So how exactly do marine biologists estimate trophic level for a given species? One way is first to determine the average trophic level of the various things that the organism eats and then add one to the value obtained. Follow- ing this formula, marine biologists trophic level 4 trophic level 5 studying Pacific bluefin tuna (Thunnus swordfish people 3.2 thynnus orientalis), a carnivore that con- sumes smaller fish and large inverte- 3.0 brates such as squid, typically assign the adults trophic levels ranging from 4.2 to 4.5, despite wide variation from place to place in the exact composition of their diets. Still, there can be sub- stantial shifts in the value of trophic level that apply to different points in development of this and other species, squid 2.2 because dramatic changes in size and anchovies behavior take place during the life of 2.0 most fishes. Another complication is that it is of- ten quite hard to determine what all goes into the stomach of a fish. This dif- ficulty can be overcome to a large de- gree by doing what we did: considering the organism as part of an integrated ecosystem and modeling the web of copepods connections between the animals and plants in enough detail to gain a reason- 1.0 able understanding of what each crea- ture is consuming, at least on average. A second method for estimating trophic level relies on a curious phe- nomenon of nitrogen biochemistry. Ni- trogen, like the other common build- ing blocks of life, comes in distinct forms. The most common isotope, ni- trophic level 1 trophic level 2 trophic level 3 trogen-14, has seven protons and seven phytoplankton neutrons in the nucleus. But a second 0.0 stable isotope with eight neutrons, ni- trogen-15, is also found in nature. In- Figure 3. Trophic levels were initially defined to include only discrete steps (left). Organic terestingly, biochemical reactions dis- detritus and microscopic plants (phytoplankton) occupy the first trophic level. Tiny zoo- criminate to a small degree between plankton, which feed on phytoplankton, reside at the second level. Creatures that eat zoo- these two different kinds of atoms. So plankton sit at the third level, and so forth. But many marine creatures feed from multiple the nitrogen incorporated into the tis- trophic levels and so could not be fit into this classic scheme. Thus the modern approach allows the assignment of trophic level to span a continuum rather than forcing it to take on sues of a sea creature does not have ex- integral values. Marine biologists would, for example, assign the anchovy, which supple- actly the same isotopic ratio as is pre- ments its main diet of phytoplankton with some zooplankton, to a trophic level of about 2.2; sent in its food. Rather, the flesh of the people fishing for anchovies (and eating a diet of only these small fish) would then be animal tends to collect the heavier iso- assigned a trophic level of 3.2 (right). tope. (The same enrichment takes place

© 2000 Sigma Xi, The Scientific Research Society. Reproduction 48 American Scientist, Volume 88 with permission only. Contact [email protected]. among terrestrial organisms. In this re- Mediterranean, the figure also dimin- Indeed, the rise and fall of the Peru- spect, you are not what you eat.) ished, but it did so more gradually vian anchoveta creates a promi- Measurements of the nitrogen from (slipping from about 3.1 to 3.0 during nent dip in the curve for the global ma- various aquatic creatures have shown the same period). rine catch, which otherwise shows a that the isotopic ratio shifts by a rough- largely steady decrease from a value of ly constant amount from one trophic Hold the Anchovies about 3.4 in the early 1950s to less than level to the next—no matter what Exceptions to this general pattern have 3.1 today. species are involved. So the analysis of certainly happened, but they are for Might this worrisome trend be nitrogen isotopes from an organism the most part easy to explain. For ex- merely some sort of statistical artifact? provides a convenient way to ascertain ample, average trophic levels of the One could imagine, for example, that its position in the food web without catches from the South Atlantic and the the decline we noted merely reflects having to know exactly what it con- Western Pacific have climbed some- changes in the way the FAO collected sumes. All that is necessary is to com- what during the past few decades—an its information, say, by using only pare the measured isotopic ratio with effect we ascribe to the development of coarse groupings at first and refining that obtained from whatever organism new fisheries in these regions. And the the categories later. To test how our re- sits at the bottom of the local food average trophic level of the fish taken sults change when many species are chain. The difference in these two val- from the South Pacific rose sharply lumped together, we regrouped the rel- ues (divided by the constant difference during the 1970s, from about 2.3 to 2.9. atively detailed totals available for the between adjacent levels) immediately This increase corresponds to the col- North Atlantic, so that the accounting shows how far above base level the lapse of a huge industrial fishery for distinguished different kinds of fish creature should be placed. Peruvian anchoveta, a species with a only at high taxonomic levels. The re- Recently, we compared values of markedly low trophic level (2.2), and sult of this exercise was to lessen the trophic level determined in this way to the growth of a more modest fishery decline seen. Hence the shift we com- those we obtained from the more tradi- for horse mackerel, which occupies a puted with the full statistics must be, if tional approach, using Prince William considerably higher position in the ma- anything, underestimating the true ex- Sound in Alaska as a sample ecosys- rine food web (trophic level 3.3). tent of the drop. tem. Happily, the two sets compared favorably (Figure 4). This result adds to tertiary consumers our certainty that the 220 estimates of Larus crassirostris trophic level that we used in our analy- sis of global fisheries, for which we em- secondary consumers Octopus ployed just one method (modeling the Sebastes tririttatus food web), were largely correct. Sebastes taczanoskii We had, perhaps, less confidence in Actiniaria the accuracy of the FAO statistics for the catch from various parts of the primary consumers Hemigrapsus sanguineus world—a perennial problem for scien- Henricia tists trying to understand the function- Septifer virgatus ing of global fisheries. Some of the con- Mytilus edulis cern arises from the breadth of the Polychaeta brush that most countries use in report- primary producers and detritus ing their catches to the FAO. Although detritus the surveys sometimes track individual Zostera Sarugussum thunbergii species (cod, for example), the statistics Hizikia fusifolia often lump multiple species together Alaria crassifolia and report values only at the level of Gloiopeltis furucata genera (such as for hake), families (her- 0 5 10 15 20 ring and sardines) or even higher taxo- deviation of 15 N / 14 N ratio (per mil) nomic groupings (all bony fishes). Figure 4. Nitrogen isotopes provide an indi- 5 Still, with no similarly comprehen- rect means for determining trophic level. sive alternatives, we had no choice but The ratio of the two stable isotopes of nitro- to employ the FAO statistics, despite 14 15 gen ( N and N) changes with position on 4 this and other shortcomings. The re- the food chain because organisms concen- sults were nevertheless clear and rea- trate the heavier isotope in their tissues. The difference in nitrogen isotope composition sonably consistent (Figure 5). In most 3 places we looked, the average trophic from level to level is approximately con- level of the catches has declined over stant, as Minagawa and Wada (1984) showed for the waters off the coast of Hokkaido, the years. In the northwest Atlantic, for 2

Japan (above). The authors have used this estimate conventional example, trophic level plunged from a method to check their more conventional peak of nearly 3.7 in 1965 to 2.8 in 1997 strategy for assigning trophic level (by mod- (the last year for which statistics are eling the relevant food webs on a computer) 1 available), while in the northeast At- for certain test cases, such as for Prince 123 4 5 lantic it fell from about 3.6 to 3.4. In the William Sound (right). nitrogen isotope estimate

© 2000 Sigma Xi, The Scientific Research Society. Reproduction 2000 January–February 49 with permission only. Contact [email protected]. Mediterranean northwest Atlantic two cases one can test the hypothesis 3.8 3.8 directly: in the Gulf of Thailand and 3.6 3.6 over the Cantabrian Shelf, offshore 3.4 3.4 from northern Spain. For the Gulf of 3.2 3.2 Thailand, we determined the change in average trophic level (during periods 3.0 3.0 of intense industrial fishing) using sci- 2.8 2.8 entific trawl surveys of fish popula- 2.6 2.6 tions, rather than relying on the catch

mean trophic level statistics. Francisco Sanchez of the In- 2.4 2.4 stituto Español de Oceanografia in 2.2 2.2 Santander and several of his Spanish 2.0 2.0 colleagues did the same for the Can- 1950 1960 1970 1980 1990 1950 1960 1970 1980 1990 tabrian Shelf. And both groups found year year that fishing pressure indeed altered the makeup of the ecosystem enough to de- global freshwater global marine 3.5 3.5 press the average trophic level. So we 3.4 3.4 firmly believe that the average trophic level elsewhere is also truly declining. 3.3 3.3 It takes very little to convince one- 3.2 3.2 self that this situation is alarming—for 3.1 3.1 seafood lovers as well as for environ- 3.0 3.0 mentalists. After all, the average troph- 2.9 2.9 ic level of the global catch has already 2.8 2.8 slipped from 3.4 to 3.1 in just a few mean trophic level decades, and there are not many more 2.7 2.7 appetizing species to be found below 2.6 2.6 this level. (Recall that 2.0 corresponds 2.5 2.5 to copepods and other tiny zooplank- 1950 1960 1970 1980 1990 1950 1960 1970 1980 1990 year year ton, creatures that are unlikely ever to be filling one’s dinner plate.) So if the Figure 5. Mean trophic level of the catch from many regions, including the Mediterranean Sea trend continues, more and more re- (upper left) and the northwest Atlantic Ocean (upper right) has declined over the past 50 years. gions are likely to experience complete The global mean for freshwater fisheries shows almost continuous decline (lower left), but the collapse of their fisheries. general trend for marine fisheries (lower right) is complicated by the rise of the fishery for Peru- Faultfinders can argued that our in- vian anchoveta in the late 1950s and its subsequent collapse in the early 1970s. Because an- chovies are assigned to a comparatively low trophic level, output from this productive fishery terpretation may be overly gloomy. creates a prominent dip in the record of global marine catch (blue line). Omitting statistics from Maybe the shift toward lower values of the area containing this fishery produces a smooth decline (black line). average trophic level simply reflects an increase in number of plankton feeders Admittedly, the procedure of assign- does not skew the results. The second in many places, perhaps species that ing a single trophic level to a species case creates the possibility of bias, be- are now thriving from the coastal rests on a rather shaky scientific foun- cause there is a tendency in many fish- plankton blooms that take place when dation, because the feeding habits of eries to catch smaller and smaller spec- fertilizer leached from farmers’ fields is many fishes change as they grow. imens over time. But the actual trophic carried to the ocean by rivers and Some begin life eating only plankton level of the catch from such a fishery streams. We have examined this possi- and hop up a full trophic level when would typically be decreasing, whereas bility by comparing the decreases in av- they begin devouring other fish. Other our analysis would show it to be con- erage trophic level with the increases in species maintain roughly the same diet stant. That is, our assigning of a single amount of seafood being caught. In (and trophic level) throughout life. Still trophic level to a species probably miss- some places and times (for example, other species slide backward a step on es some of the real change that is taking the Mediterranean Sea before 1986, or the ladder, as larvae feeding on zoo- place. Here again, we conclude that the the South Pacific during the 1960s), de- plankton and later turning to plants or trophic level of the global catch might, clining trophic levels are matched by detritus for sustenance when they in actuality, be declining faster than we appropriately large increases in the reach maturity. had documented. amount hauled in each year. But for the Fortunately, ignoring these compli- But what if this effect applies only to most part, diminishing trophic levels cations only makes our estimates of the fish being netted, not to those re- do not coincide with burgeoning popu- trophic-level decline more conserva- maining in the wild? Perhaps marine lations of plankton feeders. The overall tive. Why? Because some 86 percent of ecosystems are not changing at all, picture is indeed quite bleak. the worldwide marine catch is made only people’s choice of what they ex- up of species that either maintain a con- tract. Although such a selection bias is Medium and Message stant diet throughout life or increase in conceivable for lightly exploited fish- Our study has prompted many fish- trophic level as they grow. In the first eries, we do not consider it likely to be eries scientists to review the statistics case, assigning a single trophic level a widespread phenomenon. Indeed, in they have collected over the years for

© 2000 Sigma Xi, The Scientific Research Society. Reproduction 50 American Scientist, Volume 88 with permission only. Contact [email protected]. scope of the changes that have taken place in lakes, rivers, estuaries, coastal waters and the open ocean. Perhaps if more people accept this message, fewer will be lulled into thinking, with Lord Byron, that humankind’s ruinous ways have marked only the earth.

Bibliography Caddy, J. F., J. Csirke, S. M. Garcia and R. J. R. Grainger. 1998. How pervasive is “Fishing down marine food webs”? Science 282:183. Christensen, V. 1998. Fishery-induced changes in a marine ecosystem: Insights from mod- els of the Gulf of Thailand. Journal of Fish Bi- ology (supplement A) 53:128–142. Cushing, D. H. 1998. The Provident Sea. Cam- bridge: Cambridge University Press. Kline, T. C., Jr., and D. Pauly. 1998. Cross-vali- dation of trophic level estimates from a mass balance model of Prince William Sound us- ing 15N/14N data. In Proceedings of the Inter- national Symposium on Fishery Stock Assess- ment Models, ed. T. J. Quin II et al. Alaska Sea Grant College Program Report 98-01. Lindeman, R. L. 1942. The trophic-dynamic aspect of ecology. Ecology 23:399–418. Minagawa, M., and E. Wada. 1984. Stepwise Africa enrichment of 15N along food chains: Fur- δ15 South ther evidence and the relation between N America and animal age. Geochimica et Cosmochimica Acta 48:1135–1140. Figure 6. Fishing off the coast of Peru regularly provides a plentiful catch of anchovies (photo- Odum, E. P. 1969. The strategy of ecosystem graph) because nutrient-rich water rises from depth in this region. This coastal upwelling is so development. Science 164:262–270. pronounced that satellites carrying the proper sensors can chart this comparatively cold water Odum, W. E., and E. J. Heald. 1975. The detri- reaching the surface to the west of Peru and northern Chile (green on map). (Photograph courtesy tus-based food web of an estuarine man- of the Instituto del Mar del Peru, Proyecto Bitacoras de Pesca. Image of sea-surface temperature grove community. In Estuarine Research, for May 1999 courtesy of the National Oceanic and Atmospheric Administration.) (volume 1), ed. L. E. Cronin. New York: Academic Press. places of interest to them so that they out such work, we could never have Pauly, D. 1995. Anecdotes and the shifting can judge whether these ecosystems are mounted our study of global trophic baseline syndrome of fisheries. Trends in suffering from a decline in trophic level. levels. But with today’s inexpensive Ecology and Evolution 10:430. Our approach offers them a new tool personal computers and powerful soft- Pauly, D., and V. Christensen. 1995. Primary for assessing whether fishing in a par- ware tools, the exercise has become production required to sustain global ticular region is sustainable. But the rather straightforward. Indeed, the fisheries. Nature 374:255–257. publication of our results also engen- time was ripe to combine decades of Pauly, D. V. Christensen, J. Dalsgaard, R. Froese dered considerable critique from some statistical information about yearly and F. Torres, Jr. 1998. Fishing down marine members of the fisheries community. catches–the fodder of fisheries re- food webs. Science 279:860–863. We welcomed this scrutiny and have search—with the modeling of food Rigler, F. H. 1975. The concept of energy flows and nutrients flows between trophic levels. tried to provide others with the means webs, a branch of ecology that had In Unifying Concepts in Ecology, ed. U. W. H. to examine the foundation of our analy- now grown mature. We thus believe van Dobben and R. H. Lowe-McConnell. sis and to test the procedures we used. that the results of our investigation The Hague: Dr. W. Junk B. V. Publishers. As part of this same effort, we main- provide a robust assessment of the tain a site on the World Wide Web— shifts taking place in many regions of Fishbase.org—to aid communication the globe—and for the world as a among interested fisheries scientists whole. Documenting the systematic and managers. We have also taken full decline in trophic level exposes the im- advantage of the internet to dissemi- mense influence that industrial-scale nate “,” software we developed fishing has had on marine and fresh- for modeling the food webs of aquatic waters ecosystems. ecosystems and determining such im- Clearly, we have skipped lightly here portant quantities as the trophic level over many of the details of individual of various components. fisheries in an effort to give a compre- We thus find ourselves manipulat- hensive overview. Yet we see much val- ing computer spreadsheets more often ue in presenting such a broad summary, than we would sometimes like. With- one that should raise awareness of the

© 2000 Sigma Xi, The Scientific Research Society. Reproduction 2000 January–February 51 with permission only. Contact [email protected].