Trophic structure, species richness and biodiversity in Danish lakes: changes along a phosphorus gradient


Erik Jeppesen, National Environmental Research Institute, Department of Lake and Estuarine Ecology, PO Box 314, DK-8600 Silkeborg, Denmark.


1. Using data from 71, mainly shallow (an average mean depth of 3 m), Danish lakes with contrasting total phosphorus concentrations (summer mean 0.02–1.0 mg P L−l), we describe how species richness, biodiversity and trophic structure change along a total phosphorus (TP) gradient divided into five TP classes (class 1–5: <0.05, 0.05–0.1, 0.1–0.2, 0.2–0.4,> 0.4 mg P L−1).
2. With increasing TP, a significant decline was observed in the species richness of zooplankton and submerged macrophytes, while for fish, phytoplankton and floating-leaved macrophytes, species richness was unimodally related to TP, all peaking at 0.1–0.4 mg P L−1. The Shannon–Wiener and the Hurlbert probability of inter-specific encounter (PIE) diversity indices showed significant unimodal relationships to TP for zooplankton, phytoplankton and fish. Mean depth also contributed positively to the relationship for rotifers, phytoplankton and fish.
3. At low nutrient concentrations, piscivorous fish (particularly perch, Perca fluviatilis) were abundant and the biomass ratio of piscivores to plankti-benthivorous cyprinids was high and the density of cyprinids low. Concurrently, the zooplankton was dominated by large-bodied forms and the biomass ratio of zooplankton to phytoplankton and the calculated grazing pressure on phytoplankton were high. Phytoplankton biomass was low and submerged macrophyte abundance high.
4. With increasing TP, a major shift occurred in trophic structure. Catches of cyprinids in multiple mesh size gill nets increased 10-fold from class 1 to class 5 and the weight ratio of piscivores to planktivores decreased from 0.6 in class 1 to 0.10–0.15 in classes 3–5. In addition, the mean body weight of dominant cyprinids (roach, Rutilus rutilus, and bream, Abramis brama) decreased two–threefold. Simultaneously, small cladocerans gradually became more important, and among copepods, a shift occurred from calanoid to cyclopoids. Mean body weight of cladocerans decreased from 5.1 μg in class 1 to 1.5 μg in class 5, and the biomass ratio of zooplankton to phytoplankton from 0.46 in class 1 to 0.08–0.15 in classes 3–5. Conversely, phytoplankton biomass and chlorophyll a increased 15-fold from class 1 to 5 and submerged macrophytes disappeared from most lakes.
5. The suggestion that fish have a significant structuring role in eutrophic lakes is supported by data from three lakes in which major changes in the abundance of planktivorous fish occurred following fish kill or fish manipulation. In these lakes, studied for 8 years, a reduction in planktivores resulted in a major increase in cladoceran mean size and in the biomass ratio of zooplankton to phytoplankton, while chlorophyll a declined substantially. In comparison, no significant changes were observed in 33 ‘control’ lakes studied during the same period.


In recent years, food-web structure and interactions in the pelagic of lakes have been subject to intense debate. With a few exceptions (e.g. Hrbácek et al., 1961 ; Brooks & Dodson, 1965; Brooks, 1969), the prevailing view a decade or two ago was that food webs are primarily regulated via the available resources (‘bottom–up’ control). Since then it has become evident that food webs may be regulated via fish (called predatory control or ‘top–down’ control) ( Carpenter, Kitchell & Hodgson, 1985; Gulati et al., 1990 ; Carpenter & Kitchell, 1993). However, opinions vary on the relative importance of resource and predatory control along a nutrient gradient. On the basis of the studies of Fretwell (1977) and Oksanen et al. (1981) of terrestrial environments, Persson et al. (1988) claimed that herbivory on phytoplankton depends on the number of food web links and that the zooplankton grazing pressure is high in lakes with an even number of links (e.g. lakes with only zooplankton and phytoplankton or lakes with predatory fish, planktivorous fish, zooplankton and phytoplankton) and low in lakes with an odd number of links. Predatory control will, therefore, be strongest in food webs with an even number of links (two, four, etc.) and resource control highest in food webs with an odd number of links. There are several examples supporting the hypothesis ( Hansson, 1992; Persson et al., 1992 ; Wurtsbaugh, 1992), but also many exceptions among both lakes and streams ( Leibold, 1990; Flecker & Townsend, 1994; McIntosh & Townsend, 1994; Mazumder, 1994; Brett & Goldman, 1996). Among the reasons for the deviations are behavioural changes of prey, aiming at reducing the predation risk ( McIntosh & Townsend, 1994), and changes in the composition of primary producers towards grazing-tolerant/grazing-resistant forms at high grazer density. Moreover, species at each food-web level may differ in sensitivity towards potential predators. In addition, some species cover more than one trophic level or show ontogenetic shifts in food preference. In nature, food webs are therefore rarely simple.

Empirical analyses are alternative approaches for evaluating food-web interactions. Based on statistical analyses of experimental data, McQueen, Post & Mills (1986) and McQueen et al. (1989) concluded that resource control is highest at the bottom of the food web and predatory control strongest at the top of the web. They also suggested that predatory control is high in oligotrophic lakes and low in eutrophic lakes, particularly because eutrophic lakes are typically dominated by cyanobacteria that are difficult for grazing zooplankton to handle. Elser & Goldman (1990) and Carney & Elser (1990) then developed the intermediate control hypothesis, arguing that the zooplankton grazing pressure on phytoplankton is (a) low in oligotrophic lakes due to low nutrient availability and because the zooplankton is dominated by copepods that are less efficient grazers than large-bodied cladocerans; (b) high in mesotrophic lakes in which the zooplankton is dominated by the efficient grazer Daphnia spp.; and (c) low in eutrophic lakes in which the phytoplankton is dominated by grazing-resistant species, such as cyanobacteria. In contrast, Sarnelle (1992) argued that changes in the biomass of zooplanktivorous fish have their greatest impact on the phytoplankton (via the zooplankton) in eutrophic lakes. Likewise, Jeppesen et al. (1997, 1999) recorded dominance by small-bodied zooplankton and a low potential grazing pressure on phytoplankton in eutrophic lakes, irrespective of whether the phytoplankton was dominated by cyanobacteria or by edible green algae, thus suggesting major top–down control of zooplankton.

Changes in nutrient loading also result in changes in community structure at each trophic level or within different taxonomic groups. To quantify such changes, numerous indices have been developed (for a review see Washington, 1984). The most commonly used are species richness and various indices of diversity, and in particular the Shannon–Wiener Index. Contrasting results have been obtained for the different taxonomic units among studies. Some authors have found a unimodal relationship between species richness and trophic state, peaking at some intermediate level in the gradient ( Stockner & Benson, 1967), whereas others record a monotonical decline ( Patalas & Patalas, 1966) or even no significant changes ( Eckmann & Rösch, 1998). Diversity indices, and particularly species richness, are sensitive to sample size ( MacArthur & Wilson, 1967), invasion barriers, pH, latitude, lake area, heterogeneity, predation and disturbance ( Hall, Cooper & Werner, 1970; Magnuson, 1976; Fryer, 1985; Rørslett, 1991; Keller & Conlon, 1994) apart from ecological stress factors such as eutrophication. Variables such as heterogeneity and pH may, in turn, be influenced by eutrophication, reflecting a decline in submerged macrophyte coverage and enhanced phytoplankton production, respectively. The validity of some of the commonly used diversity indices has been questioned by several authors ( Hurlbert, 1971; Chutter, 1972; Goodman, 1975). Consequently, Hurlbert (1971) developed an alternative index, the probability of inter-specific encounter (PIE) index, incorporating both species richness and evenness components. The biological relevance of the PIE index is, therefore, more apparent than that of the commonly used indices ( Washington, 1984), but so far it has not been used for aquatic ecosystems. In addition, most studies of species richness and diversity have focused on one or two taxonomic groups, whereas only a few have compared the response of several trophic levels/taxonomic groups to changes in trophic state (e.g. Reed, 1978).

Using survey data from 71, mainly shallow, Danish lakes, we took an empirical approach to elucidate how trophic structure, species richness, biodiversity and the extent of top–down control are affected by changes in trophic state. We also included case studies from three lakes in which the biomass of cyprinids changed markedly as a consequence of fish kill or manipulation.


Samples for the analysis of water chemistry and phytoplankton and zooplankton communities were taken fortnightly during summer (1 May–1 Oct). Total phosphorus and chlorophyll a were analysed on a depth-integrated sample from the photic zone at a mid-lake station according to Søndergaard, Kristensen & Jeppesen (1992) and Jespersen & Christoffersen (1987), respectively.

Zooplankton densities were determined using depth-integrated water samples taken with a Patalas sampler and pooled from one to three stations. Depending on the total phosphorus (TP) level, between 4.5 and 9 L of the pooled sample were filtered through an 80-μm net and fixed in Lugol's iodine (1 mL, 100 mL tap water) and 0.5–1 L was settled overnight in Lugol's solution. Rotifers and small nauplii were counted from the settled samples, whilst all other zooplankton were counted from net samples. At least 100 individuals of the dominant zooplankton species were counted. Length–weight relationships, according to Dumont, Van De Velde & Dumont (1975) and Bottrell et al. (1976) , were used to estimate biomass. If possible up to 50 individuals were measured.

Phytoplankton was counted on Lugol-fixed sedimented water samples (pooled sample from a mid-lake station in the photic zone) using an inverted microscope. Biovolume was calculated by fitting the different species and genera to geometric forms ( Utermöhl, 1958; Edler, 1979; Rott, 1981). A factor of 0.29 was used to convert biovolume (mm3) to biomass (mg dw) ( Reynolds, 1984).

The composition and relative abundance of the pelagic fish stock in the lakes were determined by standardized fishing (J.P. Müller et al., unpublished) with multiple mesh (6.25, 8, 19, 12.5, 16.5, 22, 25, 30, 33, 38, 43, 50, 60, 75 mm) gill nets. The length and depth of each section of mesh was 3 and 1.5 m, respectively. Fishing was conducted in each lake between 15 August and 15 September, as previous trial fishing indicated that the distribution of the fish populations was most even during this period (Müller et al., unpublished). Moreover, young-of-the-year fish were also large enough to be included in the catch by this time. The nets were set in the late afternoon and retrieved the following morning. Catch per unit effort (CPUE) of planktivorous fish was calculated as mean catch per net.

Species richness of submerged (isoetids, angiosperms, mosses and charophytes) and floating-leaved macrophytes was determined in early August, more or less corresponding to peak biomass. We divided each lake into zones (typically 15–22). In each zone, we conducted observations from a boat using a water glass or by diving at 10 randomly selected locations at 25–50 cm depth intervals from the shore to the macrophyte boundary. At each point, we measured coverage divided into the following categories: <1, 1–5, 5–25, 25–50, 50–75, 75–95,> 95% and then interpolated these data to a coverage for the whole lake ( Jeppesen et al., 1998 ). For the sake of clarity, the data were divided amongst five TP categories: <0.05, 0.05–0.1, 0.1–0.2, 0.2–0.4 and> 0.4 mg P L−1. The data analysis comprised stepwise regression on loge-transformed data. In some cases a unimodal relationship to TP appeared. We then included [loge(TP)]2 as an independent TP variable. To estimate changes in biodiversity, we used the Shannon–Wiener Index (H′, bits)


where ni and bi are the number and biomass of the ith species, n and b are the total number and biomass of all species, and s is the total number of species. We also used the Hurlbert ‘encounter’ index (PIE, bits)



The lakes studied were, overall, small and shallow ( Table 1). Spearman's correlation on loge-transformed data revealed a strong significant positive relationship between TP and total nitrogen (TN), making it difficult to discriminate between the effect of the two nutrients on trophic structure. We selected TP as the independent nutrient variable, as it has been shown most often to be the best indicator of trophic state in temperate lakes world-wide ( OECD, 1982) as well as in Danish lakes ( Kristensen, Jensen & Jeppesen, 1990). Total phosphorus was significantly negatively correlated with mean depth, but not related to lake surface area. As trophic structure and dynamics are substantially affected by lake depth ( Keller & Conlon, 1994; Jeppesen et al., 1997 ), we included depth as an independent variable in our regression analysis to avoid bias introduced by variations in depth along the TP gradient. We also included lake area, as species richness and biodiversity have been shown to be sensitive to the size of the area studied ( MacArthur & Wilson, 1967).

Table 1.  Frequency distribution of some morphometric and chemical data for the 71 study lakes
 MeanMedian25% Percentile75% PercentileMinimumMaximum
Lake area (km2)
Mean depth (m)
TP (mg P L−1)
TN (mg N L−1)2.081.891.172.830.395.89
Chlorophyll a (μg L−1) 8752301315399

Species richness and diversity

Multiple regressions revealed that fish species richness was significantly unimodally related to TP (P<0.0001) and positively related to mean depth (P<0.0001). Lake area contributed significantly (P<0.001) if mean depth was excluded, otherwise it did not (P>0.14). The mean number of fish species increased from six in class 1 to nine in classes 3–4, followed by a decrease to seven in class 5. The Shannon–Wiener Index for fish abundance and biomass was significantly (P<0.001 and P<0.02, respectively) unimodally related to TP ( Fig. 2), and for abundance only significantly (P<0.01) related to mean depth. Hurlbert's index (PIE) was significantly unimodally related to TP (P<0.003), and positively related to mean depth (P<0.008) and peaked (0.6 bits) in class 3. In contrast to species richness, lake area did not contribute significantly to the index variations in PIE when TP was included as the independent variable.

Figure 2.

Box-plot showing the Shannon–Wiener diversity index based on abundance (Hn) (middle) and biomass (Hw) (lower) and the Hurlbert PIE index (lower) for fish, zooplankton and phytoplankton. For further details see the Fig. 1 legend.

Zooplankton species richness declined considerably (P<0.0001) with increasing TP ( Fig. 1), from a mean of 44 in class 1 to 32 in class 5. The decline was particularly noticeable for the number of cladoceran species (14 to eight), although a pronounced decline was also observed in the species number of copepods and rotifers (P<0.0001 and P<0.007), respectively. Multiple regressions revealed that only for rotifers did mean depth contribute significantly (P<0.0001) to the variation in species richness, the contribution of TP then being insignificant (P>0.25). The Shannon–Wiener Index, based on both abundance and biomass, was significantly (P<0.001) unimodally related to TP ( Fig. 2) and positively related to mean depth (P<0.0001). The indices peaked in class 4 (1.9–2 bits). Likewise, PIE was significantly unimodally related to TP (P<0.0001). It rose with mean depth (P<0.0001) ( Fig. 2) and peaked in class 4 (0.7 bits). Lake area did not contribute significantly (P>0.2) to the variation in species richness or any of the diversity indices for the selected taxa when TP was included as an independent variable.

Figure 1.

Box-plot showing the species richness of phytoplankton, total zooplankton and cladocerans, fish, submerged macrophytes and floating-leaved macrophytes in five different TP classes. The full line represents median values. Also shown are 10, 25, 75 and 90% percentiles of the variables.

Phytoplankton species richness was significantly (P<0.002) unimodally related to TP, changing from an average of 84 in class 1 to 95 in class 4, followed by a decline to 81 in class 5. Mean depth contributed significantly (P<0.05) in a positive manner to the variation in species richness when TP was included, while lake area did not (P>0.3). The Shannon–Wiener Index, for both abundance and biomass, was significantly unimodally related to TP (P<0.002 and P<0.05, respectively). Lake area contributed significantly and positively to the variation in the diversity indices (P<0.0001 and P<0.02, respectively), whereas mean depth did not (P<0.5). PIE was, however, not related to TP (P>0.5).

Marked changes occurred in species richness of submerged and floating-leaved macrophytes according to TP. The number of submerged macrophyte species declined from an average of 11.7 species in class 1 to 0.5 in class 5. In addition, the maximum depth distribution at which submerged macrophytes were recorded declined significantly ( Fig. 3; P<0.001). Species richness of floating-leaved macrophytes tended to be unimodally (though not significantly) related to TP, ranging from an average of 2.1 in class 1 over 3.1 in class 3 to 2.5 in class 5.

Figure 3.

August biomass of zooplanktivorous fish (CPUE, catch in multiple mesh-size gill nets, 14 different mesh sizes 6.25–75 mm in late summer) versus summer mean lake water concentrations of total phosphorus. Also shown are the percentage of carnivorous fish, summer mean (1 May–1 Oct) of zooplankton:phytoplankton biomass ratio, epilimnion chlorophyll a concentration, Secchi depth and the maximum depth of submerged macrophytes versus total phosphorus. Mean±SD of the five total phosphorus groups is shown (from Jeppesen et al., 1999 ).

Trophic structure

The catch of planktivorous fish by weight (CPUEw) increased significantly (P<0.0001) with TP ( Fig. 3). From the lowest to the highest TP class, CPUEw increased 10-fold. The increase is mainly attributed to roach at low and bream at high TP ( Fig. 4). The CPUEw of both bream and roach was significantly positively related to TP (P<0.001). Mean depth did not contribute significantly when TP was included as an independent variable (P>0.2).

Figure 4.

Box-plot showing mean fresh weight of various fish species caught in multiple mesh-size gill nets in late summer (15 August–15 Sept). Note the different scale on the Y-axis. For further details see the Fig. 1 legend.

With increasing TP, the contribution of piscivorous fish to total CPUEw decreased substantially (P<0.0001), from 62% in class 1 to 10–15% in classes 3–5 ( Fig. 3). Amongst the piscivores, perch dominated in the nutrient-poor lakes, whereas northern pike (Esox lucius L.) and pike perch (Lucioperca lucioperca L.) were most abundant in nutrient-rich lakes ( Fig. 4). The latter was introduced into Denmark and occurs in only a few lakes. However, the CPUEw of perch was negatively related to TP and mean depth, while CPUEw of pike was unrelated to both variables. Benthivorous ruffe (Gymnocephalus cernuus L.) was significantly (P<0.0001) unimodally related to TP, being highest in class 3. Only for perch did mean depth contribute significantly to a multiple regression including TP (P<0.05) as perch biomass increased with depth.

Besides the changes in biomass, marked changes occurred in the size/age structure in populations within the fish community. For example, the average body weight of perch decreased considerably (P<0.01) from 56 g in class 1 to 27 g in class 5 ( Fig. 5). In contrast, pike mean weight increased significantly (P<0.02) with TP, from 650 g in class 1 to 1200–1500 g in classes 4 and 5. No changes were observed for pike–perch or ruffe (P>0.8 and P>0.2, respectively). Amongst the cyprinids, the body weight of both the dominant roach and bream declined significantly (P<0.004 and 0.03, respectively) with TP. For roach this was 132 g in class 1 to 35–56 g in classes 4 and 5, and for bream from 466 g to 188–267 g. No additional (P>0.15) effect of mean depth on body weight was found for any species (P>0.15) when TP was included as an independent variable.

Figure 5.

Box-plot showing average body weight of different fish species (caught in multiple mesh-size gill nets) in five different TP classes. For further details see the Fig. 1 legend.

In accordance with the major changes in fish community structure, the contribution of Daphnia spp. to the total biomass of cladocerans decreased significantly (P<0.0001) with TP ( Figs. 6 and 7) from 63–70% in classes 1 and 2 to 38% in class 5. The mean individual body weight of cladocerans also decreased accordingly from 5.1 μg in class 1 to 1.5 μg in class 5. The decline in body weight of cladocerans reflected not only the low relative abundance of Daphnia spp., but also a reduction in body weight of the remaining Daphnia spp. (P<0.01) and of small cladocerans (P<0.002). The contribution of cyclopoids to the total abundance and biomass of copepods increased ( Fig. 6) from an average of 49% in class 1 to 77% in class 5. Daphnia spp. and calanoid copepods contributed 70% of the total biomass in class 1 and 30% in class 5, while the contribution of rotifers to total biomass was rather low (10–15%) or constant between classes ( Fig. 6). Multiple regressions revealed that CPUE of plankti-benthivorous fish by numbers contributed positively and significantly (P<0.0001) to the variation in the percentage of Daphnia spp., to mean specimen biomass of cladocerans and to the biomass percentage of calanoids to total copepods. TP or TN did not add significantly to any of the regressions (P>0.2). In addition to TP, the average body weight of cladocerans and the percentage biomass of calanoids amongst copepods and the biomass percentage of Daphnia increased with mean depth (P<0.0001), suggesting lower fish predation pressure in deep rather than in shallow lakes.

Figure 6.

Time-weighted summer mean biomass and percentage contribution of zooplankton to total biomass in five different TP classes.

Figure 7.

Box-plot showing summer average specimen weight of various genera and groups of zooplankton in five different TP classes. For further details see the Fig. 1 legend.

The changes in the zooplankton appear to cascade to the phytoplankton. Consequenty, the mean zooplankton:phytoplankton biomass ratio during summer decreased significantly with increasing TP (P<0.0001) from an average of 0.46 in class 1 to 0.08–0.15 in classes 3–5 ( Fig. 3). Mean depth did not make a significant contribution (P<0.06). With increasing TP, changes occurred in the phytoplankton community. Dinophytes, chlorophytes, diatoms and chrysophytes dominated at low TP, whereas cyanophytes and diatoms were dominant at intermediate TP, with chlorophytes and cyanophytes being dominant at high TP ( Fig. 8).

Figure 8.

Time-weighted summer mean biovolume and percentage contribution of various phytoplankton in five different TP classes.

In three eutrophic lakes surveyed intensively for 8 years, we observed major changes in fish abundance during the study period due either to fish kills or fish manipulation. Data from these lakes may, therefore, help elucidate the structuring role of fish. In Lake Engelsholm, cyprinids were removed by netting from April 1992 to September 1994. In total, 19.2 t fresh weight, or 438 kg ha−1, were removed, of which bream constituted 85%. The highest proportion (11.5 t) was removed in 1992, 2 t in 1993 and 5.7 t in 1994. Cyprinids were calculated at 660 kg ha−1 in 1990 and at 138–260 kg ha−1 in 1992–96 ( Møller, 1998). In Lake Arreskov, a major fish kill occurred in autumn and winter 1991 ( Sandby, 1998). An additional 4 t of cyprinids were removed in 1995 and during 1993 and 1995 the lake was stocked with 0+ pike, amounting to a total of 141 individuals per ha. Cyprinid biomass was calculated at 172 kg ha−1 in 1987 and 71 kg ha−1 in 1995 ( Sandby, 1998). In Lake Hejrede, 72 kg ha−1 of cyprinids were removed during 1991–94. In addition, a major, but unquantified, fish kill occurred under the ice in winter 1995–96 ( County of Storstrøm, 1996). In all three lakes, the reduction in the planktivorous fish biomass resulted in an increase in mean body weight of cladocerans and the zooplankton:phytoplankton biomass ratio, while chlorophyll a decreased ( Fig. 9). In comparison, no significant (P>0.3) changes with time were observed in any of these variables in a reference set of 33 lakes studied during the same period, suggesting that the changes in the three lakes reflect variations in fish abundance rather than inter-annual fluctuations determined by variations in, for example, climate.

Figure 9.

Time-weighted summer mean body weight of cladocerans, zooplankton:phytoplankton ratio and chlorophyll a in three lakes during 8 years in which major changes have occurred in cyprinid biomass due to fish removal or fish kill, and in 33 other reference lakes (box-plot, see legend of Fig. 1). Arrows show years of major interventions/kills.


Species richness and diversity

We observed a different response in species richness and biodiversity for the various taxonomic groups. For most of the variables included, species richness and diversity showed a unimodal relationship with TP ( Figs. 1 and 2) and were also variably related to lake area and/or depth. A different pattern was found for species richness of zooplankton and submerged macrophytes, which declined monotonically with TP.

Numerous studies have shown that species richness of fish in lakes, in accordance with the theory of island biography and habitat diversity ( MacArthur & Wilson, 1967), is strongly linked to lake area ( Magnuson, 1976; Browne, 1981; Keller & Crisman, 1990; Bachman et al., 1996 ). We found fish species diversity to be independent of lake area, which may, however, be ascribed to the small variation in lake area characterizing our study lakes ( Table 2). How the species richness and diversity of the fish community change along a trophic gradient is not clear. While several studies indicate a decline in species richness with increasing eutrophication (for a review see Larkin & Northcote, 1969; Lee, Jones & Jones, 1991), others have not recorded any changes ( Bachman et al., 1996 ; Eckmann & Rösch, 1998). In our study of mainly shallow lakes, species richness and the three selected diversity indices were all unimodally related to TP. Depth added significantly to species richness and diversity when based on abundance, which might be explained by a greater number of available niches in deep lakes.

Table 2.  Correlation analyses of morphometric and nutrient data based on loge-transformed data
 Surface areaMean depthTN
  • *


  • **


  • ***


TPns−0.36 *0.64 ***
Surface area0.49 **ns
Mean depth0.48 **ns

Most studies have reported an increase in microcrustacean zooplankton species richness with lake size (e.g. Patalas, 1972; Fryer, 1985; Dodson, 1991, 1992). Others have found a closer correlation with lake depth ( Keller & Conlon, 1994), which may be explained by the larger heterogeneity ( Keller & Conlon, 1994) and perhaps also the lower fish predation pressure in deeper lakes ( Keller & Conlon, 1994; Jeppesen et al., 1997 ). In our study of mainly shallow lakes, however, lake depth contributed significantly only to the variation in the species richness of rotifers, but not of cladocerans or copepods. The higher sensitivity of rotifers to lake depth may reflect the fact that rotifer species, unlike most microcrustaceans, are adapted to life in the often oxygen-poor hypolimnion ( Hofmann, 1985), implying that the abundance of species is probably higher in stratified than in non-stratified lakes. We found that the species richness of all selected zooplankton taxa decreased monotonically with TP and was not independent of lake size. This contrasts with the findings of Dodson (1991, 1992) who, by compiling data from 32 European and 66 North American lakes, found that species richness in both sets of lakes increased with lake size. In addition, it was unimodally related to phytoplankton production, which in other studies has been shown to be linked with TP ( OECD, 1982). Furthermore, Dodson (1992) found that species richness was higher in areas rich in lakes. Our study includes only a few oligotrophic lakes so we cannot exclude the possibility that a unimodal pattern would emerge if more nutrient-poor lakes had been included. Another explanation of the high species richness in the lower TP classes, however, could be that the abundance of submerged macrophytes increases habitat heterogeneity and then, presumably, the species richness. In shallow mesotrophic lakes in which the area and volume occupied by submerged macrophytes may be high, the presence of macrophytes may have a great impact on overall structural complexity. This contrasts with deep lakes in which the macrophytes are restricted to near-shore areas. In contrast to species richness, all three selected diversity indices were unimodally and positively related to TP. The higher species richness in the lower TP classes does, therefore, not result in higher diversity.

For phytoplankton, both species richness and the Shannon–Wiener Index were unimodally related to TP, whereas the importance of lake depth and lake area varied: diversity increasing with lake area and species richness with lake depth. As for zooplankton, different patterns may be found in the literature ( Margalef, 1978). For example, studies of the historical development in species richness and diversity of diatoms in Lake Washington also revealed a unimodal relationship with increased eutrophication, while Margalef (1980) provided evidence of higher diversity in oligotrophic than in eutrophic lakes and marine waters. The Shannon–Wiener Index for phytoplankton, based on abundance, varies from 1.0 to 4.5 bits, but typically falls in the range 2.4–2.6 ( Harris, 1986). Our data are at the lower end of this range, which may reflect an overall lower diversity in lakes (and shallow coastal estuaries) compared with the open ocean ( Margalef, 1978, 1980).

Cross-analysis of data from 641 lakes revealed that lake area contributed most significantly to the variation in macrophyte species richness in Scandinavian lakes ( Rørslett, 1991). Residual analyses showed low species richness in lakes with low pH and a unimodal relationship to eutrophication, with mesotrophic–eutrophic lakes supporting more species than oligotrophic and hypertrophic lakes. In accordance with the latter, case studies of cultural eutrophication of mesotrophic lakes have typically shown a loss in species richness ( Ozimek & Kowalczewski, 1984; Kowalczewski & Ozimek, 1993, Sand-Jensen, 1997). We also found a significant decrease in macrophyte species richness in our study. The low contribution of oligotrophic lakes to our dataset may explain why we, unlike Rørslett (1991), found a monitonical decline with increasing TP. Hence, a recent study of Danish lakes including more oligotrophic lakes showed a tendency towards a unimodal relationship between species richness and TP (Vestergaard & Sand-Jensen, unpublished). Compared with the other taxa in our study, the percentage decline in species richness with increasing TP was particularly high for submerged macrophytes. This may reflect the concurrent significant decrease in maximum depth distribution and macrophyte-covered area with eutrophication and, accordingly, a loss of habitat heterogeneity for the plants ( Fig. 3). For floating-leaved plants, which are less sensitive to increased turbidity, we observed a unimodal relationship with TP.

The general unimodal response in species richness and/or diversity for the different taxonomic groups to increasing TP or trophic state is to be expected ( Dodson, 1992): only a few species occur in distilled water and hypertrophic conditions lead to loss of species due either to competitive exclusion ( Tilman, 1982) or adverse conditions (e.g. high pH, Hansen, Christensen & Sortkjær, 1991; and low oxygen in the hypolimnion during summer or under ice during winter, Magnuson, 1976). Various other factors also play a role, however, and may obscure clear-cut relationships. We have already mentioned morphometry, physical structure and biological complexity in general. Variation in top-down control also plays a role, and it is generally believed that high predation or grazing pressure result in loss of diversity of prey organisms ( Paine, 1969), though the effect seems to depend on nutrient state ( Proulx & Mazumder, 1998). Increased fish predation may have contributed to the decline in species richness and diversity of zooplankton at high TP, but it does not explain the decline in the same variables for phytoplankton as the grazing pressure on phytoplankton was low at high TP. The variation in disturbance of various kinds may also be important and it is generally accepted that species richness and diversity are highest at intermediate disturbances ( Connell, 1978).

Trophic structure

The increase in CPUEw and the decline in the percentage of piscivores with increasing TP agree with the results from other studies in both temperate and subtropical regions ( Bays & Crisman, 1983; Hanson & Peters, 1984; Persson et al., 1988 ; Quiros, 1990; Bachman et al., 1996 ). The change in the fish community from dominance by perch and pike in mesotrophic lakes to exclusive dominance by cyprinids in eutrophic lakes is well-documented in other studies of North European lakes ( Svärdson, 1976; Leach et al., 1977 ; Persson, 1983; Persson et al., 1988 ). The superiority of roach in eutrophic lakes is attributed to a high potential growth rate along with a higher predation efficiency on cladocerans ( Persson, 1983) and an ability to exploit smaller zooplankton prey ( Stenson, 1979; Lessmark, 1983). Moreover, cyprinids are omnivorous, while large perch are piscivores. Finally, the loss of habitat complexity with the disappearance of submerged macrophytes, and thus increased turbidity, disfavours percids and, furthermore, augments intra-specific competition among them ( Persson et al., 1988 ).

The enhanced dominance of cyprinids was accompanied by a decline in average size of both cyprinids and perch. The decline probably reflects enhanced competition for food, which in turn may be mediated by reduced predation of piscivores on young fish. Thus, the fraction of large piscivorous individuals among the perch population declined. In addition, the average weight of pike increased considerably, probably as a result of enhanced cannibalism due to reduced structural complexity (less macrophytes) and, consequently, the loss of refuges ( Grimm & Backx, 1990). As large pike control small prey fish less efficiently than small pike ( Grimm & Backx, 1990), the incapability to control young planktivores decreased.

It is generally believed that ruffe biomass increases with increasing TP ( Biro, 1977; Hartman & Nümann, 1977; Persson, 1983; Bergman, 1991). Our study covered a larger TP gradient than in previous work, however, and showed a significant unimodal relationship peaking in class 3 (0.1–0.2 mg P L−1). The ruffe is benthivorous ( Johnsen, 1965; Bergman, 1991) and a competitor to young benthivorous perch. It has been argued that an increase in ruffe abundance with increasing TP reflects the fact that the foraging ability of ruffe (unlike that of perch) is largely independent of light ( Bergman, 1991). Thus, ruffe may forage effectively in turbid eutrophic lakes ( Bergman, 1991; Bergman & Greenberg, 1994). However, in eutrophic lakes ruffe faces other competitors. The observed decline at high TP is probably caused by the increase in the abundance of bream ( Fig. 3). The bream is supposedly a superior competitor in these lakes, both because it is an efficient predator on benthic invertebrates and because it, unlike ruffe, feeds efficiently in the pelagic and therefore may alternate between the benthic and pelagic feeding mode. That ruffe is an inferior competitor at high cyprinid density is evident from several biomanipulation experiments showing high abundance of ruffe during the first 1–2 years after cyprinid removal, followed by a major decline when young perch, after having initially taken advantage of the ‘empty niche’ in the pelagic, reached the benthivorous state (E. Jeppesen and M. Søndergaard, unpublished). The enhanced importance of pike–perch in eutrophic lakes is well-known from other European lakes ( Svärdson, 1976) and in accordance with their adaptations (high light sensitivity of their eyes, Ali, Ryder & Anctil, 1977) for foraging efficiently in turbid water.

Confirming most other studies, zooplankton biomass increased with increasing TP ( McCauley & Kalff, 1981; Hanson & Peters, 1984). In contrast to a number of studies ( Brooks, 1969; Bays & Crisman, 1983), however, we did not find any differences in the proportion of major taxa to biomass. Rotifers typically constituted 10–15% of biomass in all TP classes, this being consistent with results from oligo-mesotrophic Norwegian lakes ( Hessen, Faafeng & Andersen, 1995). The results from the Danish and Norwegian lakes suggest that changes in the contribution of rotifers are generally not to be expected over a substantial TP gradient (3–1000 μg P L−1) in north temperate lakes, which contradicts studies from subtropical lakes, in which an increase in rotifer:zooplankton biomass was found with increasing trophic state ( Bays & Crisman, 1983). Some studies have found an increasing share of cladocerans with TP at the expense of copepods ( Patalas, 1972; Rognerud & Kjellberg, 1984; Straile & Geller, 1998). In our study, the contribution of cladocerans and copepods to total biomass in the five TP classes ranged from 50 to 58% and 34 to 42%, respectively, and the contribution of cladocerans tended to decrease and copepods to increase slightly, though not significantly, with increasing TP. Major changes occurred in the contribution of calanoids and cyclopoids to copepod biomass, however, as well as in the contribution and mean individual weight of large cladocerans. A shift from calanoids to cyclopoids with increasing trophic state has been suggested by several authors ( Gliwicz, 1969; Patalas, 1972; Bays & Crisman, 1983) and has been related to changes in food size spectra ( Pace, 1986) or predation ( Hessen et al., 1995 ; Jeppesen et al., 1997 ). Owing to their ability to make evasive jumps, cyclopoids are often less vulnerable to fish predation than calanoids ( Winfield et al., 1983 ), although there are exceptions ( Brooks, 1969). Moreover, juvenile cyclopoids might be superior to calanoids at high food concentrations ( Santer & Van den Bosch, 1994), and increased predation on juvenile calanoids by adult cyclopoids may further reduce calanoid abundance ( Straile & Geller, 1998). Several authors record an increased share of small cladocerans with increasing trophic state (e.g. McNaught, 1975; Patalas, 1972). This is often attributed to enhanced fish predation ( Brooks, 1969; Lyche, 1990). Accordingly, we found an increase in the contribution of Daphnia spp. among cladocerans and of cladoceran mean body weight after being released from fish predation after fish removal and fish kills ( Fig. 9). World-wide, biomanipulation experiments in eutrophic lakes have shown similar results provided that fish biomass was substantially reduced (e.g. Benndorf, 1990; Hansson et al., 1998 ).

The changes in the zooplankton community structure and biomass appeared to cascade down to phytoplankton. Assuming that during summer cladocerans ingest organic carbon corresponding to 100% of their biomass per day (compared with copepods 50% per day and rotifers 200% per day) ( Hansen et al., 1992 ) and assuming that they exclusively feed on phytoplankton, then grazing amounted to 59% of phytoplankton biomass per day in class 1 and 16–19% per day in classes 4 and 5. The latter is so low that zooplankton probably have little effect on phytoplankton growth, while in class 1 and 2 lakes it seems probable that zooplankton apply a considerable grazing pressure to phytoplankton. In addition, the observed reduction in the individual weight of cladocerans and Daphnia spp. presumably further adversely affected the grazing capacity of phytoplankton, because small-bodied cladocerans are less efficient grazers on large-sized phytoplankton than the large-bodied forms ( Gliwicz, 1977, 1990). How the alteration in fish community structure and biomass, mediated by the changes in trophic state, affects herbivory on phytoplankton has been the subject of extensive discussion ( DeMelo, France & McQueen, 1992; Carpenter & Kitchell, 1992). Our data, as well as those of Leibold (1990) and Sarnelle (1992), suggest a more significant effect in eutrophic lakes than in mesotrophic lakes.

Several authors have argued that filamentous cyanobacteria prevent grazer control by zooplankton in eutrophic lakes, which may explain a decrease in herbivory from mesotrophic to eutrophic lakes ( Elser & Goldman, 1990; Carney & Elser, 1990). Supporting this view, controlled laboratory experiments have shown that dense cyanobacterial assemblages can prevent the growth of large cladocerans (e.g. Lampert, 1981; Dawidovicz, Gliwicz & Gulati, 1988; Gliwicz, 1990). Furthermore, in field experiments Daphnia spp. have occasionally failed to respond to minor, and in a few cases even to major, reductions in planktivorous fish populations (e.g. Van Donk et al., 1990 ; Riemann et al., 1990 ; Moss, Stansfield & Irvine, 1991; DeMelo, France & McQueen, 1992). The results from Danish lakes indicate, however, that the role of cyanobacteria in the decline in herbivory from mesotrophic to eutrophic lakes is considerably less significant than that of fish. First, the zooplankton:phytoplankton ratio and the size of cladocerans are low not only in eutrophic lakes dominated by cyanobacteria but also in the most nutrient-rich lakes dominated by edible green algae. Second, a reduction in fish predation on zooplankton, due either to the removal of cyprinids or to fish kills, caused a major increase in the abundance and average size of cladocerans and a decrease in phytoplankton biomass in eutrophic lakes, irrespective of whether cyanobacteria or chlorophytes dominated previously ( Jeppesen et al., 1997 ; Søndergaard, Jeppesen & Jensen, 1998; Jeppesen et al., 1999 ; authors’ unpublished data). Finally, whereas in Danish lakes CPUEw contributed highly significantly to the variation in the zooplankton:phytoplankton ratio and cladoceran mean size, the percentage contribution of cyanobacteria to the total biomass of phytoplankton did not ( Jeppesen et al., 1997 ).

The strong evidence of a decrease in herbivory, mediated by planktivorous fish from mesotrophic to hypertrophic lakes implies that the cascading effect of fish manipulation on phytoplankton in shallow lakes will be greatest under hypertrophic conditions, though the benefit of such a manipulation may be transient as planktivorous fish are highly successful in such lakes ( Fig. 3; Kitchell et al., 1977 ; Leach et al., 1977 ; Persson et al., 1988 ; Jeppesen et al., 1990 ). Therefore, a return to a turbid state and to low zooplankton grazing seems likely, even though it may be somewhat delayed in shallow lakes with an extensive growth of submerged macrophytes ( Meijer et al., 1994 ). The data in Fig. 3 suggest that the likelihood of obtaining high zooplankton grazing in shallow lakes increases markedly when summer mean TP drops below about 0.1 mg P L−1. Therefore, the prospects of long-term success of biomanipulation in shallow lakes are most probably greatest below this threshold ( Jeppesen et al., 1990 ). The results shown in Figs. 3–8 may leave the impression that the shift from clear to turbid conditions occurred gradually with increasing TP. However, several case studies of shallow lakes have shown a step-wise shift and, within certain nutrient regimes (typically 0.5–0.15 mg P L−1; Jeppesen et al., 1990 ), lakes may shift between the two states ( Moss, 1990; Jeppesen et al., 1990 ; Scheffer, 1990), this being the case in Danish lakes ( Jeppesen et al., 1990 ). Lake-specific differences in nutrient thresholds, and the fact that we also included somewhat deep lakes, may explain the lack of step-wise shifts in the present study.

In summary, species richness and diversity of most of the selected taxonomic groupings were unimodally related to TP in the nutrient range from 0.02 to 1.0 mg P L−1 and, moreover, often related positively to mean depth and/or lake area. Only the species richness of zooplankton and submerged macrophytes deviated by showing a monotonical decline with increasing TP. Major changes occurred in the community structure and size distribution of fish, which apparently cascaded to the lower trophic levels. The share of piscivores declined markedly and, moreover, the size of the predatory fish changed towards those size classes less likely to control cyprinids (larger pike and smaller perch). Enhanced predation appeared to lead to a marked reduction of the zooplankton:phytoplankton biomass ratio and the grazing pressure upon phytoplankton. The latter may have been exacerbated by a decline in the size of grazing zooplankton.


The authors thank the counties for access to data from the Survey Programme of Danish Lakes and Martin R. Perrow for valuable comments. The study was supported by the Danish Natural Science Research Council (Grant 9601711) and the research programme ‘The Role of Fish in Ecosystems, 1999–2001’, funded by the Ministry of Agriculture, Fisheries and Food. They thank Anne Mette Poulsen, Kathe Møgelvang and Henrik F. Rasmussen for skillful technical assistance.