Impact of dissolved organic carbon, phosphorus, and grazing on phytoplankton biomass and production in experimental lakes

Phytoplankton biomass and production in lakes tend to be increased by phosphorus input and decreased by grazing or high levels of colored, dissolved organic carbon (DOC). We estimated and compared the effects of these three factors by using data from three lakes that were manipulated during 1991-1995, and data from a reference lake. Multivariate probability distributions of chlorophyll or primary production, as predicted by P input rate, DOC, and grazer length, were fit to the data. All three factors had substantial effects on chlorophyll, primary production, and their variability. Comparable reductions in the mean and variance of chlorophyll and primary production were achieved by reducing P input rate from 5 to 0.5 mg me2 d-l, increasing DOC from 5 to 17 mg C liter I, or increasing mean crustacean length from 0.2 to 0.85 mm. The negative effect of mean crustacean length (an index of size-selective predation) results from grazing by herbivorous zooplankton. The negative effect of DOC on primary producers could be explained by shading. The results suggest that natural variation in colored DOC concentrations is a major cause of variation in primary production. Understanding the patterns of phytoplankton biomass and primary production in lakes has been a central concern of limnologists for both theoretical and practical reasons. Among the various factors that influence primary production, phosphorus inputs are emphasized for their role in cultural eutrophication (Schindler 1977). Food-web structure, through its influence on grazing, also affects primary production. Size-selective predation by fishes alters planktonic food-web structure (Brooks and Dodson 1965) and was recognized early on as a potential cause of variance in phosphorus-chlorophyll relationships (Vollenweider 1976). Crustacean zooplankton body size is correlated with chlorophyll (Pace 1984; Carpenter et al. 1991). Food-web manipulations that alter zooplankton size cause changes in primary producers (Gulati et al. 1990; Carpenter and Kitchell 1993). Dissolved organic carbon (DOC), especially colored compounds, may have diverse and powerful effects on lake ecosystem metabolism (Wetzel 1990). DOC potentially limits primary production by shading (Jones 1992). Furthermore, Acknowledgments We thank T M. Frost, K. Reckhow, and an anonymous reviewer for helpful comments on the manuscript; D. L. Christensen, K. L. Cottingham, D. M. Post, D. E. Schindler, D. Thomas, and N. Voichick for assistance; R. A. Hellenthal and T. M. Frost for logistical support at the University of Notre Dame Environmental Research Center and the University of Wisconsin Trout Lake Station; and the National Science Foundation for support. DOC may interact chemically with iron and phosphorus to limit nutrients available to phytoplankton (Jackson and Hecky 1980; Francko 1986). Analyses of phosphorus<hlorophyll relationships hint that lakes with high DOC may not conform to regressions for unstained lakes (Vollenweider 1976). However, the net effects of DOC on chlorophyll and primary production are not as well known as are those of phosphorus and grazing. Although a diverse literature documents various effects of these three factors, their relative strengths and interactions are poorly understood. One reason for poor resolution is that it is difficult to obtain data in which the three factors vary independently at the whole-lake scale. We obtained such data by deliberate manipulation of P inputs and the food web in lakes with contrasting DOC concentrations. In this paper, we describe the joint effects of P, DOC, and grazing on chlorophyll and primary production. It is extremely difficult to predict the interlake and temporal patterns of primary production in our experimental lakes, or in lakes in general. Regressions using multilake data demonstrate strong correlations, but predictions from such models have high variances. Furthermore, the predictions apply to populations of lakes and may not be pertinent to a specific lake of interest. Deterministic simulation models can examine mechanisms in individual lakes, but require many structural assumptions and the estimation of many parameters. Prediction errors of such models are substantial, although they are estimated only rarely.