sign in sign up
<<
Home , Planktivore

R EPORTS

was measured at 37°C with phosphatidylcholine hy- 20. F. Bauche´, M.-H. Fouchard, B. Je´gou, FEBS Lett. 349, subunit Yb-2 (accession number 121719) and endo- droperoxide at 3 mM GSH according to (16). Control 392 (1994); R. Shalgi, J. Seligman, N. S. Kosower, Biol. thelin converting enzyme (NCBI accession number samples were treated identically but with 5 mM Reprod. 40, 1037 (1989); J. Seligman, N. S. Kosower, 1706564) could be identified by MALDI-TOF or pep- 2-mercaptoethanol. R. Shalgi, ibid. 46, 301 (1992); H. M. Fisher and R. J. tide sequencing. 18. Other examples of “gene sharing” or “moonlighting Aitken, J. Exp. Zool. 277, 390 (1997). 23. Supported by the German Ministry of Education, proteins,” respectively, are reviewed by J. Piatigorsky, 21. F. Weitzel and A. Wendel, J. Biol. Chem. 268, 6288 Science and Technology, the Italian Ministry of Uni- Prog. Ret. Eye Res. 17, 145 (1998); C. J. Jefferey, (1993); R. Brigelius-Flohe´, B. Friedrichs, S. Maurer, M. versity and Scientific Research, National Research Trends Biol. Sci. 24, 8 (1999). Schultz, R. Streicher, Biochem. J. 328, 199 (1997); Council, Italy, and the BIOMED 2 program of the European . 19. M. Maiorino et al., Biol. Chem. Hoppe Seyler 376, 651 P. A. Sandstrom, J. Murray, T. M. Folks, A. M. Diamond, (1995); F. Ursini et al., Methods Enzymol. 252,38 Free Radical Biol. Med. 24, 1485 (1998). (1995). 22. In other gels, mitochondrial glutathione S-transferase 2 April 1999; accepted 29 July 1999

tions by using the natural logarithm of the , Fisheries, and ratio between the mean value of the variable in mesocosms with (zooplanktivo- - Dynamics rous fish or invertebrates) or nutrients (inor- ganic N compounds) added and in unmanip- ulated, control mesocosms (7). Zooplankti- in Marine Pelagic vores caused significant decreases in meso- Fiorenza Micheli* , both in mesocosms with no N added (Fig. 1A) and in mesocosms Anthropogenic nutrient enrichment and fishing influence marine ecosystems enriched with N (Fig. 1B). Zooplanktivores worldwide by altering resource availability and food-web structure. Meta- caused an increase in biomass, analyses of 47 marine mesocosm experiments manipulating nutrients and but this trend was statistically significant only consumers, and of time series data of nutrients, , and fishes from 20 in systems that were also enriched with N natural marine systems, revealed that nutrients generally enhance phytoplank- (Fig. 1, A and B). Nitrogen addition caused ton biomass and carnivores depress biomass. However, resource and similar and significant increases in phyto- consumer effects attenuate through marine pelagic food webs, resulting in a plankton biomass in mesocosms containing weak coupling between phytoplankton and . Despite substantial two (phytoplankton and zooplankton; Fig. 1C) physical and biological variability in marine pelagic ecosystems, alterations of or three trophic levels (phytoplankton, zoo- resource availability and consumers result in general patterns of community plankton, and zooplanktivores; Fig. 1D). Un- change. der either food-web configuration, nutrient addition did not affect mesozooplankton bio- Increased nutrient loadings and fisheries ex- lower trophic levels, and whether changes in mass (Fig. 1, C and D). The effects of the ploitation are major human perturbations to nutrient availability and primary manipulations were not significantly correlat- marine ecosystems worldwide (1). Alteration cascade up marine food webs to affect higher ed with either experiment duration or meso- of resource availability represents a “bottom- trophic levels. cosm size in zooplanktivore-manipulation ex- up” perturbation of marine ecosystems, where- To address these questions, I assembled periments (8), and the effects were only weak- as removal of consumer biomass through data from experimental manipulations con- ly correlated with duration but not with size in fishing represents a “top-down” . ducted in marine mesocosms and from long- nutrient-manipulation experiments (9). There- An understanding of how bottom-up and top- term monitoring of open marine ecosystems. fore, these results are unlikely to be biased by down processes influence the dynamics of Experiments conducted in mesocosms elimi- the short duration or small mesocosm sizes marine communities is necessary for effec- nate open-system dynamics but represent used in most experiments. tive management of marine ecosystems in the controlled alterations of nutrient availability For the 20 open marine ecosystems, I face of environmental variability and multi- and food-web structure. In contrast, long- examined the cross-correlation between time ple human impacts. However, it is difficult to term monitoring of open marine systems doc- series of nutrients, productivity, and biomass determine the effects of resource availability uments patterns at realistic spatial and tem- of different trophic levels using Spearman and food-web interactions in open (pelagic), poral scales. The first data set comprised rank correlation (10). Theoretical models ex- highly variable marine systems; most propo- phytoplankton and mesozooplankton (mostly ploring the relations among resource avail- sitions are based on anecdotal evidence from herbivorous crustaceans larger than ability, food-web structure, and biomass of catastrophic events such as El Nin˜o years (2), 150 to 300 ␮m) data from marine mesocosm different trophic levels predict patterns of fishery collapses (3), and the introduction of experiments where nutrient availability was biomass accrual along productivity gradients exotic species (4). To determine how marine manipulated by adding N compounds, or at equilibrium, that is, after transient effects pelagic ecosystems respond to variation in where food-web structure was manipulated have disappeared (11, 12). Because seasonal the quantity of resources and consumers, I by adding or removing zooplanktivorous fish events such as upwelling and sudden increas- conducted meta-analyses of data from a va- or invertebrates (5). The second data set con- es in fish density from immigration or spring riety of experimental and natural systems and sisted of time series (7 to 45 years) of N reproduction are transient effects, I used year- examined whether changes in the availability (measured as the annual loading ly values of productivity and biomass to of consumers (pelagic zooplanktivorous fish) or as the average N concentration during winter approximate equilibrium conditions. Year-to- cascade down marine food webs to affect months), primary productivity, and the bio- year fluctuations in mesozooplankton bio- mass of phytoplankton, mesozooplankton, mass were negatively correlated with zoo- and pelagic zooplanktivorous fish for 20 open planktivorous fish (r ϭϪ0.22; 95% confi- National Center for Ecological Analysis and Synthesis, ϭϪ Ϫ ϭ Santa Barbara, CA 93101, USA. marine ecosystems (6). dence limits 0.31 and 0.12; N 19), For the mesocosm experiments, I quanti- indicating that fish may control *Present address: Dipartimento di Scienze dell’Uomo e dell’Ambiente, Universita’ di Pisa, 56126 Pisa, Italy. fied responses of phytoplankton and mesozoo- mesozooplankton biomass. In contrast, the E-mail: [email protected] plankton to nutrient and food-web manipula- correlation between mesozooplankton and

1396 27 AUGUST 1999 VOL 285 SCIENCE www.sciencemag.org R EPORTS phytoplankton was not significant (r ϭ characterized by bottom-up control of prima- down effects on the phytoplankton (12). The Ϫ0.07; 95% confidence limits ϭϪ0.15 and ry producers (phytoplankton) through N avail- abstraction used in many theo- 0.01; N ϭ 19). This result may indicate that ability and top-down control of herbivores retical and empirical studies ignores the com- mesozooplankton does not control phyto- (mesozooplankton) through predation by car- plexity of species interactions and thus may plankton biomass, although a nonsignificant nivores (zooplanktivorous fish). Both analy- inadequately describe real food webs. Sec- correlation could arise through mechanisms ses indicated a weak coupling between pri- ond, the efficiency of the transfer of primary other than uncoupling between trophic levels. mary producers and herbivores. Zooplankti- productivity to higher trophic levels and the Negative correlations between zooplanktivo- vores tend to decrease mesozooplankton impact of herbivores on primary producers rous fish and mesozooplankton and between abundance, but the mesozooplankton com- may depend on food quality, particularly the mesozooplankton and phytoplankton were monly has no effect on the phytoplankton proportion of edible and inedible algae within found in six systems, but they were not sta- (Fig. 1). Conversely, increased N availability the phytoplankton (16). Increased proportions tistically significant (significance level ␣ϭ enhances primary producers but does not en- of inedible algae frequently accompany in- 0.05) except for the correlation between hance the mesozooplankton (Figs. 1 and 2). creased productivity caused by anthropogenic zooplanktivores and mesozooplankton in one In general, the effects of consumer-resource nutrient enrichment (17). Finally, in open ma- system, the subarctic Pacific (13). Thus, in interactions do not cascade upward or down- rine systems, advection or loss of nutrients pelagic marine ecosystems alterations of con- ward through marine pelagic food webs. and individuals from the focal system may sumer abundance can cascade down food The effects of carnivores (zooplanktivo- dampen effects of local biological interac- webs to affect phytoplankton biomass, but rous fish) on herbivores (mesozooplankton) tions and lead to an uncoupling between ad- this effect is uncommon. Similarly, effects of and of nutrients on plants (phytoplankton), jacent trophic levels (18). These mechanisms changes in N availability and primary pro- and the loose coupling between herbivores might act jointly to weaken primary producer- ductivity rarely cascade upward to affect bio- and plants, are pervasive. These patterns were herbivore coupling in marine pelagic food mass of marine pelagic consumers. In gener- observed at vastly different spatial (meso- webs. al, N availability and primary productivity cosms to open ocean systems) and temporal These results have implications for man- were positively correlated with phytoplank- scales (days to decades) and are similar to agement of marine ecosystems. First, the ton biomass (Fig. 2). Correlations of nutrients those found in syntheses of data from fresh- generality of a weak coupling of N loading and productivity with mesozooplankton and water systems (15). The generality of these and phytoplankton productivity with higher zooplanktivore biomass were not significant patterns indicates that similar mechanisms trophic levels (Figs. 1 and 2) implies that and showed no overall trend (Fig. 2). Posi- may underlie the dynamics of closed (fresh- anthropogenic nutrient loading to coastal tive, although nonsignificant, correlations be- water) and open (marine) aquatic systems. waters is unlikely to result in increased fish tween primary productivity and biomass of Open, highly variable systems such as marine biomass, regardless of local physical and all trophic levels were found only in two pelagic ecosystems may be regulated by local biological conditions and of the magnitude systems (14). biological interactions similar to those occur- of nutrient enrichment. Phytoplankton pro- Meta-analyses of data from mesocosm ex- ring within naturally closed lake ecosystems duction resulting from increased nutrient periments and natural marine ecosystems in- or experimentally enclosed marine and fresh- loading may be recycled within the plank- dicated that pelagic marine food webs are water systems. ton by (19) or be lost from There are at least three biological mecha- pelagic marine food webs when nisms that might account for the observed settles to the ocean floor (20). Second, weak coupling between primary producers fluctuations in stocks of planktivorous pe- and herbivores. First, coupling between tro- lagic fishes commonly affect zooplankton phic levels may be dampened by species communities but rarely cascade through interactions within the zooplankton; interfer- marine pelagic food webs to affect phyto- ence among zooplankton species may limit plankton biomass. Thus, pelagic fisheries their population growth and hinder their top- are expected to influence other components, not directly targeted by the fishery, by affecting zooplankton biomass and food availability for other carnivores. Fig. 1. Responses of phytoplankton and meso- However, it is unlikely that manipulations zooplankton to the addition of (A and B) zoo- of marine food webs similar to those pro- planktivorous fish or invertebrates and (C and posed for lakes (21) could be effective in D) inorganic N compounds to mesocosms con- controlling the response of primary produc- taining pelagic marine communities. (A) Zoo- planktivore addition was the only manipulation ers to nutrient enrichment in coastal waters. conducted in these experiments; (B) in addition Improved understanding of consumer-re- to manipulating zooplanktivores, nutrients were source dynamics is critical both to predict added in identical amounts to both control and the consequences of multiple anthropogenic zooplanktivore mesocosms; (C) both control perturbations to aquatic ecosystems and to de- and nutrient-enriched mesocosms contained velop sustainable management practices. only phytoplankton and zooplankton; (D) both Fig. 2. Correlation of (A) annual N availability control and nutrient-enriched mesocosms con- (winter concentrations or loadings of inorganic tained phytoplankton, zooplankton, and zoo- N) and (B) mean annual primary productivity planktivorous fish or invertebrates. Means are with (i) phytoplankton, (ii) mesozooplankton, References and Notes averages of the log-transformed ratios of the and (iii) zooplanktivorous fish biomass in ma- 1. P. M. Vitousek, H. A. Mooney, J. Lubchenco, J. M. mean treatment biomass divided by the mean rine pelagic food webs. Means are averages of Melillo, Science 277, 494 (1997); L. W. Botsford, J. C. Castilla, C. H. Peterson, ibid., p. 509. in the controls, weighted by sampling variances. Spearman rank correlations between time se- 2. R. T. Barber and F. B. Chavez, ibid. 222, 1203 (1983); Bars are 95% confidence intervals. The number ries, weighted by sampling variances. Bars are P. Lehodey et al., Nature 389, 715 (1997). of experiments used to calculate each average 95% confidence intervals. The number of cor- 3. G. Murphy, in Fish , J. A. Gullard, log response ratio is indicated to the right of relation coefficients averaged is indicated near Ed. (Wiley, Chichester, UK, 1977), pp. 283–308; M. J. each mean. each mean. Fogarty and S. A. Murawski, Ecol. Appl. 8, S6 (1998).

www.sciencemag.org SCIENCE VOL 285 27 AUGUST 1999 1397 R EPORTS

4. A. E. Alpine and J. E. Cloern, Limnol. Oceanogr. 37, 52, 611 (1995); R. R. Dickson, P. M. Kelly, J. M. tween 7 to 45 pairs of data points. Because of 946 (1992); Y. P. Zaitsev, Fish. Oceanogr. 1, 180 Colebrook, W. S. Wooster, D. H. Cushing, J. Plankton temporal autocorrelation within time series, the as- (1992). Res. 10, 151 (1988); Food and Agriculture Organiza- sumption of independence between years is violated 5. Only experiments conducted in marine or estuarine tion of the United Nations, Gen. Fish. Counc. Medi- and cross-correlation estimates may be biased. Bias waters [salinity 4 to 35 practical salinity units (psu)] terr. No. 63 (1990); Baltic Marine Environment Pro- due to the autocorrelation within each data series and including both treatment and control mesocosms tection Commission (Helsinki Commission), Baltic was corrected by adjusting the degrees of freedom of were included. Experiments ranged from 4 to 365 Sea Environmental Proceedings No. 35B (1990); W. the cross-correlation with the formula proposed by days and were conducted in containers ranging from Hickel, J. Berg, K. Treutner, ICES Mar. Sci. Symp. 195, M. S. Bartlett [ J. Res. Stat. Soc. Suppl. 8, 24 (1946)]. 3 ϫ 106 to 1.3 ϫ 106 liters in volume. Of the 47 249 (1992); “Reports of the ICES Advisory Committee 11. L. Oksanen et al., Am. Nat. 118, 240 (1981). experiments included in the analyses, 14 were unrep- on Fishery Management,” ICES (Int. Counc. Explor. 12. E. McCauley, W. W. Murdoch, S. Watson, ibid. 134, licated and 33 used two to four replicate mesocosms. Sea) Coop. Res. Rep. No. 196 (1993); J. Jakobsson, 288 (1988); G. G. Mittelbach, C. W. Osenberg, M. A. The nutrients added were nitrite, nitrate, or ammo- ICES Mar. Sci. Symp. 195, 291 (1992); K. Kononen, H. Leibold, in Size Structured Populations, B. Ebenman nia, alone or in combination with phosphate and Theede, W. Schramm, Kiel. Meeresforsch. 6, 281 and L. Persson, Eds. (Springer-Verlag, Berlin, Germa- silica. Zooplanktivores were various species of mysid (1988); P. Muck, in The Peruvian Upwelling System: ny, 1988), pp. 217–235; R. Arditi and L. R. Ginzburg, J. shrimp, coelenterates, chaetognates, or planktivorous Dynamics and Interactions, D. Pauly et al., Eds. Theor. Biol. 139, 311 (1989); G. A. Polis and R. D. fish. When time series of experimental results were (ICLARM, Manila, Philippines, 1989); E. Ojaveer, Ed., Holt, Trends Ecol. Evol. 7, 151 (1992); G. A. Polis and given, I averaged the data over the whole duration of Ecosystem of the Gulf of Riga Between 1920 and 1990 D. R. Strong, Am. Nat. 147, 813 (1996); K. S. McCann, the experiments for meta-analysis. Data were ex- (Estonian Academic Publisher, Tallinn, 1995); S. A. Hastings, D. R. Strong, Proc. R. Soc. London Ser. B tracted from tables or digitized from figures reported Schulz, W. Kaiser, G. Breuel, Int. Rev. Gesamt Hydro- 265, 205 (1998). in the following papers: P. C. Abreu et al., Estuaries 17, biol. 76, 351 (1991); K. Sherman and L. M. Alexander, 13. Zooplanktivores versus mesozooplankton: subarctic 575 (1994); J. G. Baretta-Bekker, B. Riemann, J. W. Eds., Variability and Management of Large Marine Pacific, r ϭϪ0.81, P Ͻ 0.01, N ϭ 10; Gotland Sea, r ϭϪ Ͼ ϭ ϭϪ Baretta, E. Koch-Rasmussen, Mar. Ecol. Prog. Ser. Ecosystems (AAAS, Washington, DC, 1985), pp. 33– 0.01, P 0.10, N 12; Arkona Sea, r 0.23, Ͼ ϭ ϭϪ 106, 187 (1994); D. L. Breitburg et al., Limnol. Ocean- 54; A. Shiomoto, K. Tadokoro, K. Nagasawa, Y. Ishida, P 0.10, N 12; middle Adriatic Sea, r 0.03, P Ͼ ϭ ϭϪ Ͼ ogr. 44, 837 (1999); E. E. Deason and T. J. Smayda,J. Mar. Ecol. Prog. Ser. 150, 75 (1997); G. Sinovcic and 0.10, N 9.3; German Bight, r 0.51, P 0.10, ϭ ϭϪ Ͼ ϭ Plankton Res. 4, 219 (1982); P. H. Doering et al., Mar. V. Alegria-Hernandez, Oceanol. Acta 20, 201 (1997); N 10.2; Gulf of Alaska, r 0.09, P 0.10, N Ecol. Prog. Ser. 52, 287 (1989); R. S. Fulton, J. Exp. 23.6. Mesozooplankton versus phytoplankton: sub- P. E. Smith and R. W. Eppley, Limnol. Oceanogr. 27,1 ϭϪ ϭ ϭ Mar. Biol. Ecol. 72, 67 (1983); Oecologia 62,97 (1982); A. J. Southward and G. T. Boalch, in Aspects of arctic Pacific, r 0.57, P 0.09, N 10; Gotland Sea, r ϭϪ0.47, P Ͼ 0.10, N ϭ 12; Arkona Sea, r ϭ (1984); J. Plankton Res. 6, 399 (1984); E. Graneli and Long-Term Changes in the Ecosystem of the Western Ϫ0.01, P Ͼ 0.10, N ϭ 12; middle Adriatic Sea, r ϭ K. Sundback, J. Exp. Mar. Biol. Ecol. 85, 253 (1985); E. English Channel in Relation to Fish Populations, T. Ϫ0.34, P Ͼ 0.10, N ϭ 8.6; German Bight, r ϭϪ0.28, Graneli et al., J. Plankton Res. 15, 213 (1993); R. P. Wyatt and M. G. Larraneta, Eds. (Instituto Investiga- P Ͼ 0.10, N ϭ 19.5; Gulf of Alaska, r ϭϪ0.26, P Ͼ Harris et al., in Marine Mesocosms, G. D. Grice and ciones Marinas, Vigo, Spain, 1988), pp. 415–447; S. 0.10, N ϭ 19. M. R. Reeve, Eds. (Springer-Verlag, New York, 1982), Suvapepun, Mar. Pollut. Bull. 23, 213 (1991); M. 14. Primary productivity versus phytoplankton: Kattegat, pp. 353–388; M. Hein and B. Riemann, J. Exp. Mar. Viitasalo, thesis, Finnish Institute of Marine Research, r ϭ 0.42, P Ͼ 0.10, N ϭ 10; Middle Adriatic Sea, r ϭ Biol. Ecol. 188, 167 (1995); A. S. Heiskanen, T. Tam- Helsinki, 1994; D. M. Ware and G. A. McFarlane, Can. 0.38, P Ͼ 0.10, N ϭ 6.3; Primary productivity versus minen, K. Gundersen, Mar. Ecol. Prog. Ser. 145, 195 Spec. Publ. Fish. Aquat. Sci. No. 121 (1995), p. 509; D. mesozooplankton: Kattegat, r ϭ 0.2, P Ͼ 0.10, N ϭ (1996); S. J. Horsted, T. G. Nielsen, B. Riemann, J. Woehrling and G. Le Fe`vre-Lehoe¨rff, Oceanol. Acta 5.3; Middle Adriatic Sea, r ϭ 0.41, P Ͼ 0.10, N ϭ 9.4. Pock-Steen, P. K. Bjørnsen, ibid. 48, 217 (1988); A. 21, 113 (1998). Primary productivity versus zooplanktivores: Katte- Jacobsen, J. K. Egge, B. R. Heimdal, J. Exp. Mar. Biol. 7. L. V. Hedges and I. Olkin, Statistical Methods for gat, r ϭ 0.19, P Ͼ 0.10, N ϭ 7.6; Middle Adriatic Sea, Ecol. 187, 239 (1995); K. Kivi, H. Kuosa, S. Tanskanen, Meta-Analysis (Academic Press, Orlando, FL, 1985); J. r ϭ 0.55, P Ͼ 0.10, N ϭ 5.4. Mar. Ecol. Prog. Ser. 136, 59 (1996); J. Kuiper, U. H. Gurevitch and L. V. Hedges, in Design and Analysis of 15. D. J. McQueen, Freshw. Biol. 23, 613 (1990); M. T. Brockmann, H. van het Groenewoud, G. Hoornsman, Ecological Experiments, S. Scheiner and J. Gurevitch, Brett and C. R. Goldman, Proc. Natl. Acad. Sci. U.S.A. K. D. Hammer, ibid. 14, 9 (1983); P. Kuuppo-Leinikki Eds. (Chapman & Hall, New York, 1993); C. W. Os- 93, 7723 (1996); M. T. Brett and C. R. Goldman, et al., ibid. 107, 89 (1994); P. Kuuppo, R. Autio, H. enberg, O. Sarnelle, S. Cooper, Am. Nat. 150, 798 Science 275, 384 (1997). Kuosa, O. Seta¨la¨, S. Tanskanen, East. Coast. Shelf Sci. (1997). Averages of the mean response ratios across 16. M. Leibold, Am. Nat. 134, 922 (1989); J. P. Grover, 46, 65 (1998); P. Olsson, E. Graneli, P. Carlsson, P. all studies, weighted by the sampling variance, are ibid. 145, 746 (1995); M. T. Brett and D. C. Mueller- Abreu, J. Exp. Mar. Biol. Ecol. 158, 249 (1992); H. W. considered significantly different from zero (that is, Navarra, Freshw. Biol. 38, 483 (1997). Paerl, J. Rudek, M. A. Mallin, Mar. Biol. 107, 247 there is a significant effect of experimental treat- 17. H. W. Paerl, Limnol. Oceanogr. 33, 823 (1988); E. (1990); J. L. Pinckney, H. W. Paerl, E. Haugen, P. A. ment) when the 95% confidence limits around the McCauley, J. A. Downing, S. Watson, Can. J. Fish. Tester, Mar. Ecol. Prog. Ser., in press; B. Riemann, mean do not overlap zero. Aquat. Sci. 46, 1171 (1989). T. G. Nielsen, S. J. Horsted, P. K. Bjørnsen, J. Pock- 8. Log of the response ratio versus duration of experi- 18. G. A. Polis, W. B. Anderson, R. D. Holt, Annu. Rev. Steen, ibid. 48, 205 (1988); B. Riemann et al., ibid. 65, ments (in days), r ϭϪ0.21, P Ͼ 0.10, N ϭ 18 Ecol. Syst. 28, 289 (1997); G. R. Huxel and K. S. 159 (1990); S. Schulz, G. Bruel, A. Irmisch, Limno- (without nutrients added), and r ϭϪ0.38, P Ͼ 0.10, McCann, Am. Nat. 152, 460 (1998). logica 20, 89 (1990); N. C. Sonntag and T. R. Parsons, N ϭ 17 (with nutrients added) for phytoplankton; r ϭ 19. L. R. Pomeroy, Biol. Sci. 24, 242 (1974); T. Fenchel, J. Plankton Res. 1, 85 (1979); A. Uitto, S. Kaitala, H. 0.29, P Ͼ 0.10, N ϭ 13 (without nutrients added), Annu. Rev. Ecol. Syst. 19, 19 (1988). Kuosa, R. Pajuniemi, Aqua Fenn. 25, 23 (1995). and r ϭ 0.07, P Ͼ 0.10, N ϭ 10 (with nutrients 20. G. T. Rowe, in Coastal Upwelling, F. A. Richards, Ed. 6. The time series data sets consisted of yearly or added) for mesozooplankton. Log of the response (American Geophysical Union, Washington, DC, summer averages of nutrients, productivity, or bio- ratio versus mesocosm volume (in liters), r ϭϪ0.20, 1981); L. Legendre, J. Plankton Res. 12, 681 (1990); P. mass. Time series ranged from 7 to 45 years and had P Ͼ 0.10, N ϭ 18 (without nutrients added), and r ϭ Wassmann and M. Barnes, Oceanogr. Mar. Biol. been gathered between 1948 and 1994 in 16 coastal Ϫ0.30, P Ͼ 0.10, N ϭ 17 (with nutrients added) for Annu. Rev. 29, 87 (1991). areas from the Baltic Sea (nine areas: Arkona Sea, phytoplankton; r ϭ 0.38, P Ͼ 0.10, N ϭ 13 (without 21. J. Shapiro, V. Lamarra, M. Lynch, in Proceedings of a Great Belt, Bornholm Sea, Gotland Sea, Archipelago nutrients added), and r ϭϪ0.001, P Ͼ 0.10, N ϭ 10 Symposium on Water Quality Management Through Sea, Gulf of Riga, Kattegat, Mecklenburg Bay, and (with nutrients added) for mesozooplankton. Biological Control, P. L. Brezonik and J. L. Fox, Eds. Oresund), the North Sea (four areas: Skagerrak, Ger- 9. Log of the response ratio versus duration of experi- (Univ. of Florida, Gainesville, 1975), pp. 85–96; R. D. man Bight, Southern Bight, and Northumberland ments (in days), r ϭ 0.23, P ϭ 0.09, N ϭ 54 (without Gulati et al., Eds., Biomanipulation—Tool for Water Coast), the English Channel (off Plymouth, UK), the zooplanktivores), and r ϭ 0.067, P ϭ 0.01, N ϭ 14 Management (Kluwer, Dordrecht, The Netherlands, middle Adriatic Sea, and the Gulf of Thailand and four (with zooplanktivores) for phytoplankton; r ϭ 0.57, 1990); S. R. Carpenter and J. F. Kitchell, Limnol. offshore areas from the Peruvian and the California P ϭ 0.08, N ϭ 10 (without zooplanktivores), and r ϭ Oceanogr. 37, 208 (1992). upwelling systems, the Gulf of Alaska (ocean station 0.08, P Ͼ 0.10, N ϭ 10 (with zooplanktivores) for 22. I thank P. Amarasekare, J. Bascompte, L. Benedetti- P), and the subarctic Pacific (south of the Aleutian mesozooplankton. Log of the response ratio versus Cecchi, O. Bjørnstad, D. Breitburg, M. Brett, S. Car- Islands). All systems are subject to intense human mesocosm volume (in liters), r ϭϪ0.12, P Ͼ 0.10, N penter, K. Cottingham, G. Englund, B. Kendall, J. disturbance through fishing and anthropogenic nutri- ϭ 54 (without zooplanktivores), and r ϭϪ0.35, P Ͼ Kitchell, H. Lenihan, K. McCann, E. McCauley, G. ent loadings to the coastal systems. Data were ex- 0.10, N ϭ 14 (with zooplanktivores) for phytoplank- Mittelbach, W. Murdoch, C. Parmesan, C. H. Peterson, tracted from tables or digitized from figures pub- ton; r ϭϪ0.09, P Ͼ 0.10, N ϭ 10 (without zoo- J. Pinckney, O. Sarnelle, D. Schindler, A. Sih, and two lished in the following papers and reports: L. Anders- planktivores), and r ϭ–0.14, P Ͼ 0.10, N ϭ 10 (with anonymous reviewers for helpful comments and S. son and L. Rydberg, East. Coast. Shelf Sci. 26, 559 zooplanktivores) for mesozooplankton. For all analy- Glaholt for helping in assembling the data used in (1988); M. C. Austen et al., J. Mar. Biol. Assoc. UK 71, ses, qualitatively similar results were obtained when these analyses. This study was conducted at the 179 (1991); G. T. Boalch, D. S. Harbour, E. I. Butler, short- and long-duration experiments were excluded. National Center for Ecological Analysis and Synthe- ibid. 58, 943 (1978); E. Bonsdorff, E. M. Blomqvist, J. 10. I combined correlation coefficients (r) using standard sis, a Center funded by NSF (grant DEB-94-21535), Mattila, A. Norkko, Oceanol. Acta 20, 319 (1997); meta-analytical techniques described by W. R. Shad- the University of California–Santa Barbara, the Cali- R. D. Brodeur et al., Calif. Coop. Ocean. Fish. Investig. ish and C. K. Haddock [in The Handbook of Research fornia Resources Agency, and the California Environ- Rep. 37, 80 (1996); R. Milla´n-Nu´n˜ez, S. Alvarez-Bor- Synthesis, H. Cooper and L. V. Hedges, Eds. (Russel mental Protection Agency. rego, C. C. Trees, ibid., p. 241; A. Corten, Neth. J. Sea Sage Foundation, New York, 1994), pp. 261–281]. Res. 25, 227 (1990); D. H. Cushing, ICES J. Mar. Sci. Each coefficient was obtained from correlation be- 23 March 1999; accepted 7 July 1999

1398 27 AUGUST 1999 VOL 285 SCIENCE www.sciencemag.org