Jordán, F. 2001. Ecosystem flow networks and community control. In: Villacampa, Y., Brebbia, C. A. and Usó, J.-L. (eds.), Ecosystems and Sustainable Development III, WIT Press, Southampton, pp. 771-780. Ecosystem flow networks and community control F. Jordán Collegium Budapest, Institute for Advanced Study, Hungary Abstract It is not easy to construct the food web graph of a community: both points and links are strongly questionable. Although recently described modern food webs are finely resoluted (have many points), two serious problems still remain. First, these are generally binary webs: links give no information on the importance of interactions (the number of large weighted webs is small). Second, their properties are very sensitive to methodology (i.e. data collection and aggregation procedure). Thus, high resolution in itself cannot guarantee reality. It is suggested that the mostly valuable conclusions on community organisation can be drawn by comparing weighted webs depicting the same community under different conditions. In this case, the comparison of similarly described webs may partly diminish methodical problems. Seasonal trophic flow networks of the Chesapeake Bay ecosystem are compared and differences are discussed from the viewpoint of community control. Changes in the functional importance of trophic links and the structural importance of trophic components are analysed. The main conclusion is that trophic control in the Bay community seems to be the strongest during winter. 1 Introduction Classical food web collections [1, 2] were criticised (e.g. Paine [3]) and recently, following methodical consensus [4], the number of modern webs is increasing (see one of the best: Martinez [5]). However, until we understand the problems of aggregation much better (see Solow and Beet [6]) and can estimate the level of redundancy in ecological systems more exactly [7], these webs are not trivially much better. Besides the careful determination of trophic components, two other problems appear. First, these are generally binary webs but information on the quality of links (e.g. flux magnitude [8], interaction strength [9]) would be extremely important for understanding community organisation [8]. Second, even if both points and links of the food web graph are described with large efforts, web properties still remain very sensitive to methodology. One possibility to diminish methodical problems is to compare webs describing the same community under different conditions. For example, both binary [10] and weighted [11] webs presenting temporal variation in trophic interactions are available. If similarly produced (i.e. described and aggregated) webs are compared, some of the methodical biases and artefacts will surely disappear. It is the change of food web structure what may shed a new light on community control. Thus, it is suggested that staring at a set of comparable webs can be more useful than staring at single, however highly resoluted webs (see Lawton [12]). The scarcity of data obviously limits statistical analyses of pairs (e.g. stressed/control, Ulanowicz [13]) or larger sets (e.g. four seasons, Baird and Ulanowicz [11]) of webs, and this is especially true for quantitatively weighted trophic networks. Nevertheless, as the number of weighted and well aggregated modern webs increases, our picture will be improved. In this paper, I present an analysis of one of the available comparable sets of weighted trophic flow networks. My aim is to explain the changes of food web structure from the viewpoint of community control [14]. 2 Methods and results The Chesapeake Bay is one of the best studied coastal regions of North America. Baird and Ulanowicz [11] published the seasonal carbon flow network diagrams, based on data on the mesohaline community of the Bay. 29 25 WINTER 28 26 29 30 SPRING 32 28 26 19 23 33 24 9 21 31 17 27 13 18 22 15 16 19 23 9 21 17 13 8 12 18 22 15 16 11 8 12 11 14 14 7 7 6 3 6 3 5 2 5 34 2 1 34 4 1 4 29 30 25 SUMMER 32 28 33 31 27 10 19 23 20 33 31 13 15 FALL 32 28 26 27 17 18 25 24 9 21 22 29 30 26 19 23 9 21 17 13 16 22 18 15 16 8 12 8 12 11 11 14 14 7 7 6 3 6 3 5 5 2 2 34 34 1 4 1 4 Figure 1: Seasonal carbon flow networks of the Bay. Table 1. Seasonal magnitudes of carbon flows for each source/sink pair (after Baird & Ulanowicz [11]). from to floww flowsp flowsu flowf from to floww flowsp flowsu flowf from to floww flowsp flowsu flowf 1 2 34 5 1 2 6 1 2 7 2 7 8 8 9 1 2 7 1 2 7 1 2 7 3 2 3 5 6 7 7 7 8 8 8 9 9 9 10 10 11 11 11 12 12 12 13 13 13 14 15300 67140 12600 6300 1210 50 2520 5092 4280 1328 525 525 1050 145 95 10 57 37 4 337 220 23 29895 29440 86799 50416 25208 0 0 6067 23920 31665 2162 692 691 1348 174 114 12 680 445 47 1795 1176 124 49545 23920 60536 95680 47840 26956 22724 19320 5667 5667 2834 1265 1265 2530 1159 552 3550 2325 245 1288 844 88 1834 1202 126 50400 43680 74438 18746 9373 3549 0 3731 2460 2460 1231 975 956 1914 330 217 23 250 163 17 449 294 31 31918 3 3 3 4 3 3 11 12 15 16 18 8 8 1 2 8 1 2 8 8 14 15 18 12 14 15 16 17 17 18 19 19 19 19 19 19 20 21 22 22 22 23 23 23 24 25 25 25 26 26 3667 4880 1120 560 3485 75 24 9 20 196 43 4,4 12 14 65 1,8 0 3,2 0,7 0,2 0,1 0 8 7410 16805 12350 6175 4059 270 84 34 68 702 152 17,4 55 63 300 7 41 90 3,9 2 13 9330 28000 17333 8667 3360 940 300 120 160 2560 540 4,9 1,9 117 136 652 8 46 102 1,3 4,5 1,3 0,2 5 25 4800 7980 5366 2684 3252 415 130 52 104 1080 232 2 93 108 517 4 24 53 2 0,6 0 2 13 15 18 14 15 16 22 14 15 22 14 15 18 22 23 27 22 18 22 23 31 19 21 22 23 1 26 26 27 27 27 27 28 28 28 29 29 29 30 30 30 31 32 32 32 32 33 33 33 33 34 2 2 3 8 1 3,8 0,6 1,1 12083 3 3 2 7 24 1,5 68 34 11 75 11 21 0,8 0,7 3 7,4 0,4 6,9 3,7 1 1,1 0,1 7 4,4 18630 6 6 265 76 26 11 0 13 1,6 54 8 15 1 1 4 30 0,4 4,6 3,7 2,8 1,3 0,1 8,1 5 26956 3 3 49 14 5 2 0 9 1 20 3 6 0,9 0,9 3,2 54 0,1 0,8 0,6 0,5 0 0 2,1 1,2 16585 The original diagrams written in energy language [15] are redrawn as conventional food webs, where consumers are always in higher position (Figure 1). For simplicity, undirected links are shown: arrowheads would always point at higher points. Detrital cycling links and the original source components "energy and nutrients" and "exogenous input" are excluded from this analysis (explained below). Thus, only producers, consumers and storages are considered. 2.1 Flow data and important links Links are characterised by the magnitude of carbon flux from prey to consumer (expressed in mg C/m2 per season). Table 1 contains seasonal flow data (published in [11]; w = winter; sp = springtime; su = summer; f = fall). Seasonal carbon flows between each pair of prey (so=source) and consumer (si=sink) can be ranked according to the ratio of that flow in the "carbon menu" of the consumer in question [16]. For example, sink #25 gets carbon from sources #14, #15, and #18 in winter. Based on feeding ratios, in the menu of consumer #25, they are ranked in this order. It will be assumed that the magnitude of fluxes characterise, in some sense, the importance of trophic links. Table 2. Components of the trophic flow network (after Baird & Ulanowicz [9]). 1. phytoplankton 2. suspended bacteria 3. sediment bacteria 4. benthic diatoms 5. free bacteria 6. heteromicroflagellatae 7. microzooplankton 8. zooplankton 9. ctenophore 10. sea nettle 11. other suspension feeders 12. Mya 13. oysters 14. other polychaetes 15. Nereis 16. Macoma spp. 17. meiofauna 18. crustacean deposit feeders 19. blue crab 20. fish larvae 21. alewife and blue herring 22. bay anchovy 23. menhaden 24. shad 25. croaker 26. hog choker 27. spot 28. white perch 29. catfish 30. blue fish 31. weak fish 32. summer flounder 33. striped bass 34. dissolved organic carbon 2.2 Keystone indices and important points The names of the 34 major trophic components of the Bay are given in Table 2. Points representing these groups can be characterised by their position in the flow network. A traditional measure of network position is "status" (with "contrastatus" and "net status", all introduced by Harary [17, 18]). I have slightly modified these measures and introduced "keystone index" [19], which is ecologically more reasonable. Considering a simple species deletion model (similar to Pimm [20]), we may calculate, on purely topological grounds, how many species will be secondarily extinct, following the removal of the xth species: extinction is caused by the disconnection of network flows. Whereas carbon flows are strictly directed in the trophic network, population dynamical effects spread in both bottom-up and top-down directions [21], and in both direct and indirect ways [22, 23]: this calls for considering undirected edges if community control is to be analysed. Table 3. Seasonal keystone indices of components. COMP KW KSP KSU KF COMP KW KSP KSU KF 1 23,5 28,5 31,5 28,5 18 1,36 1,23 1,57 1,19 2 14,11 15,36 18,34 15,77 19 6,13 4,71 4,62 4,42 3 8,97 9,14 9,95 9,81 20 - - 0,3 - 4 0,5 0,5 0,5 0,5 21 0,53 0,67 0,55 0,52 5 3,29 6,5 3,81 4,11 22 1,11 3,43 3,46 3,94 6 3,29 6,5 3,81 4,11 23 0,64 1,51 1,37 1,85 7 5,9 6,5 7,46 6,22 24 - 0,42 0,3 - 8 3,4 5,53 7,27 4,51 25 0,83 - 0,73 0,5 9 1,54 1,38 1,79 1,51 26 0,83 1,88 1,78 1,84 10 - - 2,59 - 27 - 1,74 1,67 1,72 11 1,29 1,3 1,31 1,26 28 2,32 0,82 0,45 0,49 12 1,29 1,55 1,56 1,51 29 0,83 0,83 0,73 0,79 13 1,12 1,09 1,1 1,09 30 - 3,25 3,11 3,27 14 1,52 1,44 1,44 1,61 31 - 0,53 0,44 0,55 15 1,69 1,65 2,15 2,27 32 - 2,42 2,2 2,48 16 0,36 0,73 0,73 0,69 33 - 7,72 7,44 0,89 17 1,19 1,19 1,19 1,19 34 3,29 6,5 3,81 4,11 Table 4. Seasonal congruency values (C) for non-specialist consumers. consumer Cw Csp Csu Cf 7 8 9 10 11 12 13 17 19 22 23 25 26 27 28 29 30 32 33 33 % 100 % 17 % 100 % 100 % 100 % 100 % 53 % 0% 33 % 66 % 75 % 100 % 33 % - 17 % 66 % 17 % 100 % 100 % 100 % 100 % 33 % 0% 0% 50 % 17 % 0% 33 % 66 % 83 % 66 % 100 % 83 % 17 % 100 % 100 % 100 % 100 % 100 % 47 % 0% 0% 33 % 42 % 33 % 66 % 0% 50 % 66 % 66 % 33 % 83 % 17 % 100 % 100 % 100 % 100 % 33 % 0% 0% 66 % 58 % 33 % 66 % 33 % 17 % 83 % 58 % average 65 % 50 % 58 % 54% The bottom-up keystone index (Kb) of species x measures how many species extinctions will follow the deletion of the xth species, caused by the disconnection of flows from the bottom up: K b c 1 n 1 1 K bc . dc dc (1) Here, n is the number of species x's predators, dc is the number of preys eaten by its cth predator, and Kbc is the bottom-up keystone index of its cth predator. Kb has to be calculated first for higher points in the web. The top-down keystone index (Kt) can be calculated similarly with the web upside down. Thus, this index measures how many species will go extinct, following the removal of species x, caused by the disconnection of the flows from the top down. In the simplest case, K Kb Kt . (2) The possibility of weighting, according to the relative importance of bottom-up and top-down forces, is open (e.g. mKb + nKt). The overall keystone index (K) reflects the positional importance of a point in maintaining any kinds of network flows. This index, although simple and represents only a certain aspect of "species importance", can give a quantitative prediction of keystones (which is a major task in community ecology, see [24, 25]). The seasonal keystone indices of each component are given in Table 3. 2.3 Congruency analysis The question is how important points are related to important links. In other words, if we analyse the prey choice of a non-specialist consumer, how large carbon intakes do belong to positionally important preys. Because we are interested in bi-directional effects mediated by trophic interactions, and detrital links can well be considered as unidirectional (detritus pool affects mostly detritivores but rarely feeds back directly to higher-level organisms), cycling links and source components ("energy and nutrients" and "exogenous input") given in [11] have not been taken into account. Since being in a key position means both to be a main energy gate [26] and to have far-reaching top-down effects [27], it is suggested here that if large flows fit well to positionally important network points, trophic control can be considered stronger. Large nutrient flows give also a strong top-down control ability to main gates. The congruency between important flows (Table 1) and important positions (Table 3) predicts the strength of trophic community control (versus nontrophic control, i.e. mutualism, interference competition, etc.). Congruency values calculated by a combinatorial stepwise method and expressed in percents are shown at Table 4. The method can be illustrated with the following example. In springtime, component #32 intakes carbon from components #18, #22, #23, and #31 (see Figure 1). The magnitudes of carbon flows are 0.4, 6.9, 3.7, and 1, respectively (Table 1). The corresponding prey ranking is #22 - #23 - #31 - #18. The springtime overall keystone indices (Ksp) are 3.43, 1.51, 0.53, and 1.23, respectively (Table 3). These can be ranked as #22 #23 - #18 - #31. Perfect congruency (100%) with the previous rank would be identical to that, while the worst congruency (0%) would correspond to the reversed rank order (#18 - #31 - #23 - #22). These can be transformed into each other in six or more steps (where each step means the exchange of two neighbours in the sequence). Thus, each sequence can be characterised by percentage values according to six steps (0%, 17%, 34%, 50%, 67%, 84%, and 100%; the minimal number of steps, of course, depends on the length of a sequence). The rank based on positional keystone indices (#22 - #23 - #18 - #31) corresponds to a 83% level congruency with the rank of flows (#22 - #23 - #31 #18). Congruency values averaged for each season inform about the dynamics of prey choice and seasonal variation in the significance of trophic community control. It has to be emphasised that if a biological system as complex as an ecosystem is characterised by a single index, much information is lost. Nevertheless, higher level phenomena may be revealed only by this way [28,29]. 3 Discussion The seasonal dynamics of the Chesapeake Bay ecosystem comprise both dynamical and structural changes in the trophic flow network of carbon. Structural changes involve the presence/absence of both trophic components (component #24 is not present during winter) and trophic links (component #33 eats component #19 only in summer and springtime). Dynamical changes involve both seasonally different magnitudes of carbon flows (component #7 gains carbon mainly from component #1 in summer but from component #6 in winter) and changes in the relative positional importance of network points (component #7 is in more important position in summer than in other seasons). The relationship between important links and important points also changes: larger carbon flows come mainly from positionally more important preys during winter, and from positionally less important preys, for example, in springtime. A single macroscopic community index (average congruency) was used to express the nature of community-level regulation. Seasonal values of this index inform about community control dynamics during the year. In conclusion, winter is the season when trophic control is the most powerful. It is unavoidable to list three main shortcomings of the analysis presented. First, as a theoretical study, it is sensitive to the quality of data (published in [11]). Even if this weighted trophic flow network is one of the best available food webs, the quality of data can always be criticised. Second, material flows were represented by only a single element (carbon). There are attempts to take other elements into account [30] but the carbon data base is the richest up to date. Third, conclusions are not general because results have no statistical value: a single community was analysed. It is questionable how these results (65% versus 50%) are significant but, as far as I think, statistical analyses could only be evaluated if we knew much more about constraints and limits on food web organisation. The same difference tells different stories depending on the realisable range of a particular community index. An approach and a method have been suggested to reveal some aspects of community control. Even if an excellent data base is needed for performing this kind of analysis, the predictive power of quantitative approaches may be helpful. As a number of community-level patterns change systematically during ecosystem development [31], community indices may help in understanding the properties of, for example, strongly perturbed ecosystems [32]. Ecosystem studies focusing on the (changing) relationships between functional components and interactions between them strongly emphasise the necessity and importance of the network perspective in ecology [33]. Acknowledgements I thank István Molnár and Professor Gábor Vida for continuous help and discussions, Professors Robert E. Ulanowicz and Stuart L. Pimm for comments on earlier work, and Professor Eörs Szathmáry for a fruitful stay in Collegium Budapest. Viktor Müller is kindly acknowledged for checking English. My work was supported by a grant of the Hungarian Scientific Research Fund (OTKA F 029800) and the Bolyai Award of the Hungarian Academy of Sciences. References [1] Cohen, J.E. Food Webs and Niche Space, Princeton University Press: Princeton, pp. 133-165, 1978. [2] Briand, F. Environmental control of food web structure. Ecology, 64, pp. 253-263, 1983. [3] Paine, R.T. Food webs: road maps of interactions or grist for theoretical development? Ecology, 69, pp. 1648-1654, 1988. [4] Cohen, J.E., Beaver, R.A., Cousins, S.H., De Angelis, D.L., Goldwasser, L., Heong, K.L., Holt, R.D., Kohn, A.J., Lawton, J.H., Martinez, N.D., O'Malley, R., Page, L.M., Patten, B.C., Pimm, S.L., Polis, G.A., Rejmánek, M., Schoener, T.W., Schoenly, K., Sprules, W.G., Teal, J.M., Ulanowicz, R.E., Warren, P.H., Wilbur, H.M. & Yodzis, P. Improving food webs. Ecology, 74, pp. 252-258, 1993. [5] Martinez, N.D. Artefacts or attributes? Effects of resolution on the Little Rock Lake food web. Ecological Monographs, 61, pp. 367-392, 1991. [6] Solow, A.R. & Beet, A.R. On lumping species in food webs. Ecology, 79, pp. 2013-2018, 1998. [7] Lawton, J.H. & Brown, V.K. Redundancy in ecosystems (Chapter 12). Biodiversity and Ecosystem Function, eds. E.D. Schulze & H.A. Mooney, Springer Verlag: Berlin, pp. 255-270, 1994. [8] Platt, T., Mann, K.H., & Ulanowicz, R.E. (eds). Mathematical Models in Biological Oceanography, The UNESCO Press: Paris, 1981. [9] Paine, R.T. Food-web analysis through field measurement of per capita interaction strength. Nature, 355, pp. 73-75, 1992. [10] Warren, P.H. Spatial and temporal variation in the structure of a freshwater food web. Oikos, 55, pp. 299-311, 1989. [11] Baird, D. & Ulanowicz, R.E. The seasonal dynamics of the Chesapeake Bay ecosystem. Ecological Monographs, 59, pp. 329-364, 1989. [12] Lawton, J.H. Food webs (Chapter 3). Ecological Concepts, ed. J.M. Cherrett, Blackwell Science: London, pp. 43-78, 1989. [13] Ulanowicz, R.E. Trophic flow networks as indicators of ecosystem stress (Chapter 34). Food Webs: Integration of Patterns and Dynamics, eds. G.A. Polis & K.O. Winemiller, Chapman and Hall: London, pp. 358-368, 1996. [14] Jordán, F. Seasonal changes in the positional importance of components in the trophic flow network of the Chesapeake Bay. Journal of Marine Systems, 27, pp. 289-300, 2001. [15] Odum, H.T. Ecological and General Systems, University Press of Colorado: Colorado, p. V, 1994. [16] Levine, S. Several measures of trophic structure applicable to complex food webs. Journal of theoretical Biology, 83, pp. 195-207, 1980. [17] Harary, F. Status and contrastatus. Sociometry, 22, pp. 23-43, 1959. [18] Harary, F. Who eats whom? General Systems, 6, pp. 41-44, 1961. [19] Jordán, F., Takács-Sánta, A. & Molnár, I. A reliability theoretical quest for keystones. Oikos, 86, pp. 453-462, 1999. [20] Pimm, S.L. Food web design and the effect of species deletion. Oikos, 35, pp. 139-149, 1980. [21] Hunter, M.D. & Price, P.W. Playing chutes and ladders: heterogeneity and the relative roles of bottom-up and top-down forces in natural communities. Ecology, 73, pp. 724-732, 1992. [22] Menge, B.A. Indirect effects in marine rocky intertidal interaction webs: patterns and importance. Ecological Monographs, 65, pp. 21-74, 1995. [23] Abrams, P.A., Menge, B.A., Mittelbach, G.G., Spiller, D.A. & Yodzis, P. The role of indirect effects in food webs (Chapter 35). Food Webs: Integration of Patterns and Dynamics, eds. G.A. Polis & K.O. Winemiller, Chapman and Hall: London, pp. 371-395, 1996. [24] Bond, W.J. Keystone species (Chapter 11). Biodiversity and Ecosystem Function, eds. E.D. Schulze & H.A. Mooney, Springer Verlag: Berlin, pp. 237-253, 1994.2 [25] Power, M.E., Tilman, D., Estes, J.A., Menge, B.A., Bond, W.J., Mills, L.S., Daily, G., Castilla, J.C., Lubchenco, J. & Paine, R.T. Challenges in the quest for keystones. BioScience, 46, pp. 609-620, 1996. [26] Margalef, R. Perspectives in Ecological Theory, University of Chicago Press: Chicago, pp. 9-11, 1968. [27] Paine, R.T. A note on trophic complexity and community stability. American Naturalist, 103, pp. 91-93, 1969. [28] Ulanowicz, R.E. On the importance of higher-level models in ecology. Ecological Modelling, 43, pp. 45-56, 1988. [29] Ulanowicz, R.E. A phenomenology of evolving networks. Systems Research, 6, pp. 209-217, 1989. [30] Ulanowicz, R.E. & Baird, D. Nutrient controls on ecosystem dynamics: the Chesapeake mesohaline community. Journal of Marine Systems, 19, pp. 159172, 1999. [31] Odum, E.P. The strategy of ecosystem development. Science, 164, pp. 262270, 1969. [32] Jordán, F. Is the role of trophic control larger in a stressed ecosystem? Community Ecology, in press. [33] Patten, B.C. Network ecology: indirect determination of the life-environment relationship in ecosystems (Concluding Remarks). Theoretical Studies of Ecosystems - the Network Perspective, eds. M. Higashi & T.P. Burns, Cambridge University Press: Cambridge, pp. 288-351, 1991.