Influence of Food Web Structure on Carbon Exchange Between Lakes and the Atmosphere

See allHide authors and affiliations

Science  11 Jul 1997:
Vol. 277, Issue 5323, pp. 248-251
DOI: 10.1126/science.277.5323.248


Top predators and nutrient loading in lakes were manipulated to assess the influence of food web structure on carbon flux between lakes and the atmosphere. Nutrient enrichment increased primary production, causing lakes to become net sinks for atmospheric carbon (Catm). Changes in top predators caused shifts in grazers. At identical nutrient loading, Catm invasion was greater to a lake with low grazing than to one with high grazing. Carbon stable-isotope distributions corroborated the drawdown of lake carbon dioxide and traced Catm transfer from algae to top predators. Thus, top predators altered ecosystem carbon fixation and linkages to the atmosphere.

In many lakes, carbon (C) inputs from terrestrial systems are sufficiently high that lakes are supersaturated with CO2, and there is net diffusion of CO2 out of surface water, making lakes conduits of C from the terrestrial environment to the atmosphere (1-3). In productive lakes, primary production by algae and C storage in biota and sediments are high, so that aqueous CO2 is depleted and Catm diffuses into surface waters (2, 4).

Primary production by algae in lakes is determined by interactions among a variety of factors that include nutrient loading (5) and food web structure (6). In general, primary production is high in lakes with large nutrient loads. Food web structure is often determined by the feeding characteristics of fishes. Lakes with planktivorous fishes are generally characterized by small bodied zooplankton grazers (7) that are less effective at suppressing algal abundance and growth than are communities dominated by large bodied grazers that often coexist with piscivorous fishes (6). Thus, food web structure as determined by the dominant feeding modes of predatory fishes has the potential to regulate aquatic primary production and C fluxes between lakes and the atmosphere.

We independently manipulated fish communities and nutrient loading rates in four lakes to test the interactive effects of nutrient loading and food web structure on lake productivity and C exchange with the atmosphere (Table 1). Two lakes were dominated by zooplanktivorous fishes (minnows), and two by piscivores (bass) (8). From 1993 to 1995, one lake from each food web configuration was enriched with nitrogen and phosphorus, and the two other (reference) lakes were monitored without fertilization (Table 1).

Table 1

Lake characteristics and design of whole lake experiments to test the interacting effects of nutrient loading and food web structure on primary production rates and C exchange between lakes and the atmosphere. All lakes are located at the University of Notre Dame Environmental Research Center near Land O' Lakes, Wisconsin (89°32′W, 46°13′N). Values of water transparency (Secchi depth), depth of the mixed layer, and mixed layer dissolved inorganic carbon (DIC) are long-term means taken from weekly samples during the stratified season from 1991 to 1995. The food web structures of Peter Lake and Tuesday Lake were manipulated in May 1991 and September 1991, respectively, by rotenone treatment to remove all piscivores, followed by restocking of planktivorous species. Liquid fertilizer containing PO4, NH4, and NO3 at an N:P atomic ratio of 25 was added from a central station in Peter Lake and West Long Lake from May to September in 1993, 1994, and 1995 (9). P loading rates for these 3 years were 3.06, 1.83, and 0.97 mg P · m−2 · day−1 for Peter Lake, and 3.19, 2.24, and 0.92 mg P · m−2 · day−1 for West Long Lake. Baseline P loading rates are about 0.1 to 0.15 mg P · m−2 · day−1. Primary production was measured biweekly, and CO2 fluxes weekly, from 1991 to 1995 in all lakes except Tuesday Lake. In Tuesday Lake, primary production was measured in 1989 when the lake was planktivore-dominated and not enriched with nutrients (6). Planktonic primary production was measured at six depths by the 14C method. Daily integrated C fixation was calculated by continuous measurements of surface irradiance and weekly profiles of light attenuation, temperature, chlorophyll, and dissolved inorganic C (6). C stable-isotope distributions in the pelagic food webs were monitored in 1994 and 1995. Lake characteristics are described in detail in (6).

View this table:

At ambient nutrient loads, mean summer primary production rates by pelagic algae ranged from 170 to 414 mg C · m−2 · day−1 among all lakes (Fig.1A). Nutrient enrichment increased primary production in both the piscivore lake and the planktivore lake. However, at enhanced nutrient loads, the planktivore lake always had higher primary production than the piscivore lake. At equivalent nutrient loads, the mean summer primary production rates in the planktivore lake were 1.4 to 2.9 times higher than in the piscivore lake (Fig. 1A).

Figure 1

(A) Effect of P input rate on primary production in four lakes with contrasting food web configurations. Each symbol represents a summer mean for one lake*year combination from 1991 to 1995. Lakes characterized by high planktivory and low grazing rates are shown by circles. Triangles represent lakes with piscivores and high grazing rates. Lakes that were experimentally fertilized are denoted by filled symbols: gray symbols are lake*years before enrichment, and black symbols are during enrichment. (B) Relation between calculated CO2 flux between lakes and overlying atmosphere (18), and the estimated primary production rate in 1992 to 1995 (correlation −0.84, P < 0.001). Positive values of CO2 flux represent net flow of C out of the lake, and negative values represent net flow of C into the lake. The dashed line represents lakes that are in equilibrium with the overlying atmosphere with respect to CO2 concentration. CO2 flux rates were only measured in 1994 and 1995 for Tuesday Lake (open circles). Symbols are as described in (A).

Before nutrient enrichment, all lakes were net sources of C to the atmosphere, and C efflux rates ranged from 24 to 113 mg C · m−2 · day−1 (Fig. 1B). After nutrient enrichment, C efflux decreased with increases in primary production such that enriched lakes were either approximately in equilibrium with atmospheric CO2 or were net sinks for atmospheric carbon (Catm). Summer mean influx of Catm at enhanced nutrient loads ranged from 101 to 140 mg C · m−2 · day−1 to the planktivore lake and from 0 to 30 mg C · m−2 · day−1 to the piscivore lake. These differences in Catm influx rates to lakes are due to regulation of primary production by food web structure. In the piscivore-dominated lake, planktivorous fishes were essentially eliminated, and a large-bodied herbivorous zooplankton community dominated by Daphnia pulex was present (9). Because these large grazers are able to suppress algal growth responses to nutrient enrichment (6, 9, 10), the effect of increased nutrient load on primary production is weak in piscivore-dominated lakes. In planktivore-dominated lakes, algae are not suppressed by grazers, and primary production is sensitive to increases in nutrient load (9). Thus, algae are more effective at depleting dissolved CO2 in planktivore-dominated lakes and allow Catm to diffuse down the concentration gradient from the atmosphere into surface waters.

Cole et al. (3) reported that the global variation in lake CO2 concentration ranges from about 16 times undersaturated to about 16 times supersaturated. We observed most of this range in relative CO2 concentration in our experiment. The fertilized zooplanktivore lake was about 16 times undersaturated with CO2, unfertilized lakes were about four times supersaturated with CO2, and the fertilized piscivore lake was close to equilibrium with the atmosphere. Thus, the potential magnitude of the effects of food web structure on C exchange between lakes and the atmosphere is substantial.

A brief perturbation to the food web of Peter Lake confirmed that food web dynamics can produce the shifts in primary production and dissolved CO2 observed between experimental lakes (Fig.2). The mean length of zooplankton grazers ranged from 0.2 to 1.0 mm in 1993 to 1995 as a result of a partial minnow die-off in 1994 and subsequent population recovery in 1995 (9, 11). In correspondence with the food web changes, the partial pressure of CO2(P CO 2) ranged from approximately in equilibrium with the atmosphere to greatly undersaturated relative to the atmosphere (12).

Figure 2

Comparison of crustacean grazer length, primary production, and ΔP CO 2[P CO 2 (lake)P CO 2 (air)] before, during, and after a period of low minnow abundance in Peter Lake (1994 to 1995). Data are shown as the mean ± SD. Sample sizes are given in parentheses. Dotted line in bottom panel represents dissolved CO2 in equilibrium with the atmosphere.

We used C stable isotope distributions in the biota to confirm in-lake CO2 drawdown and trace Catm through the pelagic food webs of the experimental lakes (13). Catmexhibits high 13C:12C ratios relative to aqueous CO2 produced from terrigenous C (14,15). This enriched isotopic composition of Catmtranslates into a relatively high δ13C in algae when their growth is supported by fixation of Catm(15). Drawdown of lake CO2 concentration also changes the C stable isotope distribution because algal fractionation of C isotopes is reduced when CO2 is diminished from water (16).

When net C flux was out of the lakes—that is, at ambient nutrient loads—C stable isotope distributions (δ13C) of planktonic algae, zooplankton, and fishes were between −30 and −36% (Fig. 3), indicating the importance of terrigenous C in supporting growth of aquatic biota (15). Enhancing primary production through nutrient enrichment caused substantial decreases in CO2fugacity that resulted in changes in the rates and direction of C flux between lakes and the atmosphere. The increased contribution of Catm in food webs is illustrated by higher values of δ13C in algae, zooplankton, and fishes. The greatest proportion of Catm was observed in the planktivore-dominated lake (Fig. 3), where grazers were unable to suppress primary production rates with nutrient enrichment (Fig. 1A). These results demonstrate that top predators can influence C flow between ecosystems and the atmosphere through their effects on primary producers. This finding elaborates upon earlier discoveries that nutrient enrichment increases algal productivity and decreases CO2 fugacity in lake (2, 4) and open ocean (17) ecosystems.

Figure 3

Carbon stable isotope distributions of three pelagic food web components [(A) Algae, (B) zooplankton, and (C) fish] as a function of CO2 flux rate between lakes and the atmosphere (correlation coefficients shown, all P < 0.001). Invasion of Catm into lakes due to CO2 drawdown results in relatively high values of δ13C. Lower values represent decreasing degree of CO2 drawdown and an increasing importance of terrigenous C in biota. Data are shown for 1994 and 1995 for each of the four lakes in the experiment. Negative values of CO2 flux represent net flow of C into lakes, and positive values represent net flow out of lakes. Symbols are as described in Fig. 1.

By independently manipulating food web structure and nutrient loading at the scale of whole ecosystems, we demonstrated that top predators alter fundamental biogeochemical processes that control internal ecosystem dynamics and interactions with the atmosphere. Shifts in top predators determined whether the experimentally enriched lakes operated as net sinks or net sources of Catm.

  • * Present address: Department of Zoology, University of Washington, Seattle, WA 98195, USA. E-mail: deschind{at}

  • To whom correspondence should be addressed. E-mail: srcarpen{at}


View Abstract

Navigate This Article