- Top of page
- Total phosphorus
1. This synthesis examines 35 long-term (5–35 years, mean: 16 years) lake re-oligotrophication studies. It covers lakes ranging from shallow (mean depth <5 m and/or polymictic) to deep (mean depth up to 177 m), oligotrophic to hypertrophic (summer mean total phosphorus concentration from 7.5 to 3500 μg L−1 before loading reduction), subtropical to temperate (latitude: 28–65°), and lowland to upland (altitude: 0–481 m). Shallow north-temperate lakes were most abundant.
2. Reduction of external total phosphorus (TP) loading resulted in lower in-lake TP concentration, lower chlorophyll a (chl a) concentration and higher Secchi depth in most lakes. Internal loading delayed the recovery, but in most lakes a new equilibrium for TP was reached after 10–15 years, which was only marginally influenced by the hydraulic retention time of the lakes. With decreasing TP concentration, the concentration of soluble reactive phosphorus (SRP) also declined substantially.
3. Decreases (if any) in total nitrogen (TN) loading were lower than for TP in most lakes. As a result, the TN : TP ratio in lake water increased in 80% of the lakes. In lakes where the TN loading was reduced, the annual mean in-lake TN concentration responded rapidly. Concentrations largely followed predictions derived from an empirical model developed earlier for Danish lakes, which includes external TN loading, hydraulic retention time and mean depth as explanatory variables.
4. Phytoplankton clearly responded to reduced nutrient loading, mainly reflecting declining TP concentrations. Declines in phytoplankton biomass were accompanied by shifts in community structure. In deep lakes, chrysophytes and dinophytes assumed greater importance at the expense of cyanobacteria. Diatoms, cryptophytes and chrysophytes became more dominant in shallow lakes, while no significant change was seen for cyanobacteria.
5. The observed declines in phytoplankton biomass and chl a may have been further augmented by enhanced zooplankton grazing, as indicated by increases in the zooplankton : phytoplankton biomass ratio and declines in the chl a : TP ratio at a summer mean TP concentration of <100–150 μg L−1. This effect was strongest in shallow lakes. This implies potentially higher rates of zooplankton grazing and may be ascribed to the observed large changes in fish community structure and biomass with decreasing TP contribution. In 82% of the lakes for which data on fish are available, fish biomass declined with TP. The percentage of piscivores increased in 80% of those lakes and often a shift occurred towards dominance by fish species characteristic of less eutrophic waters.
6. Data on macrophytes were available only for a small subsample of lakes. In several of those lakes, abundance, coverage, plant volume inhabited or depth distribution of submerged macrophytes increased during oligotrophication, but in others no changes were observed despite greater water clarity.
7. Recovery of lakes after nutrient loading reduction may be confounded by concomitant environmental changes such as global warming. However, effects of global change are likely to run counter to reductions in nutrient loading rather than reinforcing re-oligotrophication.
- Top of page
- Total phosphorus
During the past 10–30 years, major efforts have been made in many countries to improve the ecological quality of lakes by combating external nutrient loading (Marsden, 1989; Sas, 1989), sometimes in combination with additional restoration measures such as biomanipulation (Benndorf, 1990; Gulati et al., 1990) or physico-chemical methods (Cooke et al., 1993). The effects of biomanipulation have been described in several recent reviews (Perrow et al., 1997; Hansson et al., 1998; Drenner & Hambright, 1999; Meijer et al., 1999; Benndorf et al., 2002; Mehner et al., 2002). Since the extensive reviews in the 80s and early 90s (Marsden, 1989; Sas, 1989), there have also been several summaries of lake responses to reductions in nutrient loading without the confounding effect of biomanipulation (e.g. Cooke et al., 1993; Van der Moelen & Portielje, 1999; Willén, 2001a; Jeppesen, Jensen & Søndergaard, 2002; Søndergaard et al., 2002). Most have focused on nutrients and phytoplankton, whereas other biological components have been only briefly covered.
The aim of this paper is to evaluate 35 case studies on lake re-oligotrophication based on information given in questionnaires filled in by scientists and in follow-up communications. We focus on changes in water clarity, nutrients, fish, plankton, submerged macrophytes, as well as resource and top-down control of phytoplankton. Based on previous studies we hypothesized that reduction in P loading, N loading or both would result in:
1 A notable delay in the reduction of in-lake total phosphorus (TP) concentrations because at least three retention times are needed to wash out 95% of the excess P pool in the water column of fully mixed lakes, unless P is permanently lost to the sediment, (Sas, 1989), and because internal loading continuously replenishes the P pool in the water column (Søndergaard, Jensen & Jeppesen, 2003; Nürnberg & LaZerte, 2004).
2 A quick response of the total nitrogen (TN) concentration to reduction in N loading, because N loss by denitrification results in negligible internal N loading (Jensen et al., 1992). Any delays might be greater in deep than in shallow lakes because of often longer hydraulic retention times in deep lakes, and a reduced denitrification capacity arising from a lower ratio of sediment area to water volume.
3 An increase in the in-lake TN : TP ratio because of an often higher TN : TP ratio of inflowing water, a decrease in internal P loading and, ultimately, when low TP concentrations have reduced primary production, reduced denitrification as organic carbon becomes limiting for denitrification (Levine & Schindler, 1989).
4 An increase in particle-bound P because P limitation of phytoplankton will gradually replace limitation by light or N (Sas, 1989), resulting in lower soluble reactive phosphorus (SRP) : TP ratios. Moreover, the number of free inorganic binding sites for P may increase (e.g. because of a higher Fe : P ratio in inflowing water), thereby facilitating precipitation of inorganic P and thus lowering the SRP : organic P ratio and consequently the SRP : TP ratio.
5 A unimodal response of the chlorophyll a (chl a) : TP ratio (McCauley, Downing & Watson, 1989). The increase in the chl a : TP ratio from high to moderate TP concentrations occurs because light limitation of phytoplankton growth by self-shading is replaced by increased P limitation because of a lower SRP : TP ratio (Sas, 1989; Reynolds, 2002). In deep lakes, a high P concentration in the hypolimnion favours motile dinophytes, which may assimilate P in the hypolimnion and transport it to the epilimnion. This, in turn, may lead to a greater chl a : TP ratio in the epilimnion when TP concentrations decrease in the illuminated water layer (Anneville, Gammeter & Straile, 2005; Dokulil & Teubner, 2005). In contrast, increased grazing (Jeppesen et al., 2003) and higher water transparency may lead to a decline in the chl a : TP ratio at low TP concentrations (Portielje & Van der Moelen, 1998).
7 No consistent response of the fish fauna. There is a general perception, particularly amongst those studying shallow lakes (Scheffer et al., 1993), that the fish community responds only slowly to loading reductions because large organisms such as fish have slow growth rates and high longevity, especially key fish species in nutrient-rich turbid lakes, such as common bream (Abramis brama L.) and carp (Cyprinus carpio L. and Carassius auratus L.). This hypothesis is challenged, however, by the fast responses observed in several reservoirs (Yurk & Ney, 1989; Kalff, 2002) and natural lakes, both deep (Müller & Meng, 1992; Eckmann & Rösch, 1998) and shallow (Jeppesen et al., 2002).
9 A greater Secchi transparency and therefore a spread of submerged macrophytes. However, macrophyte recolonisation may be slow owing to, for instance, limited seed banks and/or grazing on macrophytes by waterfowl (Mitchell & Perrow, 1997).
- Top of page
- Total phosphorus
Our analysis is founded on an elaborate questionnaire. In addition, loading data and summer and annual mean values of various lake attributes were examined for the time of maximum external loading, and every 5 years thereafter. Data were averaged from typically 3 years around each 5-year period (−1, 0, +1) to reduce the effect of inter-annual variation. In about 20% of the cases, lack of data reduced the period covered to 1–2 years or larger than 3 years (to cover more variables). Some of the case studies used in this synthesis are described in greater detail in other papers published in this special issue of Freshwater Biology (Anneville et al., 2005; Coveney et al., 2005; Dokulil & Teubner, 2005; Jeppesen et al., 2005a; Köhler et al., 2005; Moss et al., 2005; Phillips et al., 2005; Romo et al., 2005; Søndergaard, Jensen & Jeppesen, 2005).
Basic information on the lakes is given in Table 1. Our data set includes both shallow (mean depth <5 m or polymictic) and deep (mean depth up to 177 m) lakes, with trophic status prior to restoration efforts ranging from mesotrophic to hypertrophic. Included are lakes from the subtropics to the temperate zone (latitude : 28–65 °) in North America and Europe. All warm-temperate and subtropical lakes were shallow. All lakes have been subjected to a reduction in P loading, with or without additional measures to reduce N loading. The loading reductions typically began in the 1970–1980s. For most lakes nutrient reduction in the lake inlets was the only restoration measure taken, but in three lakes it was followed by a major (Lake Balaton) or moderate (Veluwemeer, Apopka) removal of fish. The recovery periods studied range from 5 to 35 years (mean = 15.6 years) but typically varied between 10 and 20 years (Table 1).
Table 1. Basic information on the study lakes ordered by depth type and decreasing latitude
|Lake type||Lake no.||Lake name||Recovery period included (years)||Country||Longitude||Latitude||Altitude (m)|| Catchment area (km2)|| Lake area (km2)||Mean depth (m)||Max depth (m)||Retention time (year)||Summer stratification||Months stratified||TPsum at maximum nutrient loading (μg L−1)||TPsum last year included in the study (μg L−1)|
|Shallow||1||Little Mere||10||England||2°24′W||53°20′ N||36||3.5||0.03||0.7||2||0.2||No|| ||3500||135|
|Shallow||5||Veleuwemeer||25||Netherlands||5°4′E||52°25′N||0|| ||30.5||1.5||5||0.14||No|| ||528||44|
|Shallow||11||Bagsværd||10||Denmark||12°27′E||55°46′ N||20||7||1.21||1.9||3.2||1.7||No|| ||237||107|
|Shallow||16||Ørnsø||10||Denmark||9°31′E||56°9′ N||19||56||0.42||4||6.6||0.05||Temporary|| ||106||80|
|Shallow||22||Peipsi||15||Estonia/ Russia||25–30 E||56–59 N||30||44260||3555||7.1||15.3||2||No|| ||27||39|
|Deep||28||Constance||25||Germany/ Austria/ Switzerland||9°18′E||47°39′N||395||10500||472||101||252||4.4||Yes||7||87||13|
We have synthesised the questionnaire answers and data in Tables 1–4 and Appendices 1–3. The direction of change (Table 4, Appendices 1 and 3) is based in most cases on statistical tests and in a few cases on qualified evaluations given by the data supplier. In cases where the questionnaire respondent expressed reservations for the direction of change (for instance because of large inter-annual variations), the sign (plus or minus) is bracketed. We used a chi-square test for analysing whether the direction of the changes in selected variables was significantly different from the expectations (SAS Institute, 1989). As Danish lakes comprise a relatively large share of the data set (36% of the shallow and 31% of the deep lakes), and as some of our hypotheses are based on earlier observations from Danish lakes, we ran the tests with and without the data from the Danish lakes. Where percentage values were calculated, they were based on the full data set and on the reduced set excluding Danish lakes. The latter are shown in parentheses. When data were scarce or scattered, we used LOESS regression (SAS Institute, 1989) to give an indication of the direction of changes on the data set divided into shallow and deep lakes. Here the Danish data were included. The LOESS procedure allows smoothing of data but provides no firm statistical tests. In addition, we used multiple regressions (forward procedure) on log-transformed data, with data for deep and shallow lakes pooled. Although these relationships may include transient effects, we believe that they are robust because they cover a very large gradient in mean depth and nutrient status compared to the changes in nutrient loading and concentrations experienced by each lake during the study period.
Table 2. Results from multiple linear regression analyses of selected environmental variables (P always <0.0001). The relationships assume potential hysteresis after nutrient loading reduction to be small because of the gradient in nutrients covered is larger than the one the individual lakes have gone through during the study period.
|Response variable||Constant||Predictive variable 1||Predictive variable 2||Predictive variable 3||r2||n|
|log (TN : TPsum)||1.03 ± 0.30||+0.35 ± 0.09 log (TN : TPload)||+0.37 ± 0.05 log (Zmean)|| ||0.46||77|
|log (TN : TPann)||1.26 ± 0.31||+0.40 ± 0.08 log (TN : TPload)||+0.24 ± 0.10 log (Zmean)|| ||0.41||46|
|log (SRPsum)||−3.46 ± 0.43||+1.34 ± 0.08 log (TPsum)||+0.25 ± 0.07 log (Zmean)|| ||0.80||128|
|log (DINsum)||−7.05 ± 1.17||+1.55 ± 0.15 log (TNsum)||−0.24 ± 0.06 log (tw)||0.75 ± 0.08 log (Zmean)||0.63||102|
|log (DIN : TNsum)||−2.91 ± 0.20||−0.31 ± 0.06 log (tw)||+0.68 ± 0.09 log (Zmean)|| ||0.39||102|
|log (DIN : SRPsum)||6.95 ± 0.77||−1.07 ± 0.44 log (TPsum)||+0.30 ± 0.13 log (Zmean)||−0.39 ± 0.08 log (tw)||0.49||121|
|log (Phytosum)||−3.30 ± 0.28||+1.12 ± 0.07 log (TPsum)|| || ||0.75||98|
|log (Chl asum)||−2.30 ± 0.56||+0.93 ± 0.06 log (TPsum)||+0.20 ± 0.09 log (TNsum)|| ||0.85||107|
|log (Chl a : TPsum)||−2.14 ± 0.58||+0.76 ± 0.22 log (TPsum)||−0.10 ± 0.02 (log (TPsum))2||−0.12 ± 0.05 log (Zmean)||0.21||107|
|log (Zoosum)||4.68 ± 0.60||+0.41 ± 0.11 log (TPsum)||−0.28 ± 0.08 log (Zmean)|| ||0.62||80|
|log (ZooPhyt)||0.02 ± 0.037NS||−0.36 ± 0.08 log (TPsum)|| || ||0.19||79|
Table 3. Results of multiple linear regression analyses of the contribution of selected phytoplankton taxa versus total phosphorus concentration in summer (TPsum; μg L−1) and mean depth (Zmean; m). The relationships assume potential hysteresis after nutrient loading reduction to be small because the gradient in nutrients covered is larger than the one the individual lakes have gone through during the study period.
|Response variable||Constant||Predictive variable 1||Predictive variable 2||Predictive variable 3||r2||n|
|log (% Chlorophyta + 1)||+3.27 ± 1.01||−1.08 ± 0.50 log (TPsum)||+0.17 ± 0.06 [log (TPsum)]2|| ||0.16||103|
|log (% Cyanobacteria + 1)||−1.27 ± 0.89||+1.72 ± 0.44 log(TPsum)||−0.16 ± 0.05 [log (TPsum)]2|| ||0.27||103|
|log (% Diatoms + 1)||+4.45 ± 0.26||−0.31 ± 0.06 log (TPsum)|| || ||0.20||103|
|log (% Cryptophyta + 1)||+2.95 ± 0.19||−0.04 ± 0.01 log (TPsum)2|| || ||0.20||99|
|log (% Chrysophyta + 1)||+4.89 ± 0.64||−1.69 ± 0.32 log (TPsum)||+0.15 ± 0.04 [log (TPsum)]2|| ||0.44||103|
|log (% Dinophyta + 1)||−2.38 ± 1.24||+1.16 ± 0.51 log (TPsum)||−0.12 ± 0.05 [log (TPsum)]2||+0.54 ± 0.11 log (Zmean)||0.30||103|
Table 4. Answers to questions about responses of submerged macrophytes to reductions in nutrient loading. Lakes 1–22 are shallow, all others are deep.
|Question||Answer for lake group 1||Answer for lake group 2||Answer for lake group 3|
|Does submerged macrophyte abundance increase?||Yes: 1, 5, 13, (17), 26, 27, 31, 32||No change: 7, 23, 29, 33||Decrease: 2, 4, 25|
|Do the depth limits of submerged macrophytes (excluding mosses) increase?||Yes: 5, 13, 20, 26, 27, 28, 32||No change: 1, 2, 25, 26||Decrease: none|
|Does percentage coverage or plant volume inhabited by plants increase?||Yes: 5, 12, 14, 26, 28, 32||No change: 13||Decrease: 2, 4, 25|
|Does the re-establishment of submerged plants occur gradually or abruptly?||Gradually: 13, 14, 26, (27), 32||Abruptly: 1, 12, 17||Exponentially: 5|
|Are there any changes in species richness, Simpson evenness and Shannon–Wiener diversity of submerged and floating-leaved macrophytes at the species or genus level?||Increase: 4, 5, 13, 30||No change: 20||Decrease: 2, 31|
|Do changes in macrophyte variables follow patterns different from those observed at increasing nutrient loading?||Yes: 4, 12, 13, 28||No: 5, 20, 21, 31|| |
- Top of page
- Total phosphorus
For both deep and shallow lakes we found clear effects of reductions in nutrient loading (Table 5). For most lakes, lake TP concentrations, chl a concentration in the surface water and phytoplankton biomass were lower and Secchi depth was higher. Internal loading apparently delayed the recovery, but in most lakes a new equilibrium between P in the sediment and water column was reached after 10–15 years, thus confirming earlier findings by Ahlgren (1978) and Sas (1989). There was a slight tendency to faster recovery in lakes with a short retention time, despite high TP concentrations when loadings were greatest, and in lakes with a long hydraulic retention time and previously low TP loading (Fig. 2; Table 1). No clear effect of lake depth on recovery was detected (Fig. 2). The latter corresponds well with results from earlier comparisons across lakes by Sas (1989) and Jeppesen et al. (1991). A plausible explanation is that lakes with short hydraulic retention times often were more heavily impacted in the past, resulting in higher P accumulation in the sediment and, following nutrient reduction, higher and/or longer internal loading, whose effects overrode the effective P removal through high flushing rates (Jeppesen et al., 1991).
Table 5. General overview of key findings from the analysed case studies on lake re-oligotrophication. Note that the responses of shallow and deep lake cannot be fully compared as the starting TP level before loading reduction differed, being overall higher in shallow lakes (Table 1).
|Response variable||Shallow lakes (mean depth <5 m or polymictic)||Deep lakes (others)|
|P response time to TP loading reduction||Typically 10–15 years||Typically 10–15 years|
|N response time to TN loading reduction||Typically <5 years||Typically <5 years|
|TP summer and annual||Decreased in most lakes||Decreased in all lakes|
|TN summer||Decreased in most lakes||No clear pattern|
|TN : TP summer||Increased in most lakes even in some lakes with lower TN : TP in the inlet||Increased in most lakes|
|SRP summer||Decreased in all lakes when TP decreased||Decreased in all but one lakes when TP decreased|
|SRP : TP summer||Decreased in all lakes when TP decreased||Decreased in all but one lakes when TP decreased|
|DIN : SRP summer||Increased in most lakes||Increased in most lakes|
|Secchi depth summer||Increased in most lakes||Increased in most lakes|
|Chl a summer||Decreased in most lakes||Decreased in most lakes|
|Chl a : TP summer||Increased or no changes||Increased or no changes|
|Phytoplankton biovolume||Decreased in most lakes||Decreased in most lakes|
|Phytoplankton community changes||Higher importance of diatoms, cryptophytes and chrysophytes||Decline in cyanobacteria and greater importance of dinophytes and chrysophytes|
|Fish biomass, judged from surveys, commercial catches or angling reports||Decreased in most lakes||Decreased in most lakes|
|Percentage piscivorous fish||Increased in most lakes, thus likely resulting in enhanced top-down control of prey fish||Increased in most lakes thus likely resulting in enhanced top-down control of prey fish|
|Fish community changes in European lakes (examples)||Cyprinids to percids plus cyprinids||Cyprinids to percids plus cyprinids or Percids plus coregonids to coregonids or Coregonids to coregonids plus salmonids, depending on TP levels|
|Zooplankton biomass||Decreased||No clear pattern|
|Zooplankton : phytoplankton biomass ratio||Increased in many lakes, probably reflecting release from fish predation||No clear pattern|
|Submerged macrophytes||No clear pattern||No clear pattern|
|Indications of enhanced bottom-up control of phytoplankton||Nearly all lakes||Nearly all lakes|
|Indications of enhanced top-down control of phytoplankton||Many lakes||No clear pattern|
|Seasonality||Largest reduction in TP and chl a concentrations during spring and autumn, later in the recovery phase also in summer. Exceptions are some lakes with major reductions in N loading, showing major effects also in summer in the early recovery phase||Largest effect in summer, later in spring and autumn|
The quick (<5 years) response of lakes to N loading reductions compares well with the results of other studies (Jensen et al., 1992) and may be explained by the fact that surplus inorganic N is lost to the atmosphere via denitrification rather than accumulated in the sediment. For two deep and some shallow lakes with high TN and correspondingly high TP concentrations (see below), the observed TN concentrations were often lower than predicted by the relationship established previously for Danish lakes, but approached the predicted values after 5–10 years (Fig. 3). The general increase in the TN : TP and DIN : SRP ratios in the lake water suggests that P became a more likely limiting nutrient for phytoplankton growth than before reductions in nutrient loading.
The increase in N relative to P may in part reflect that loading reductions have mainly been directed towards P. While this is a wise strategy for deep lakes (Sas, 1989), recent studies show that nitrogen may play a more important role in shallow lakes than hitherto anticipated (Moss, 2001; Gonzáles Sagrario et al., 2005; James et al., 2005). Nitrogen negatively affects both submerged macrophyte species richness (James et al., 2005) and the chances of maintaining a macrophyte-dominated state at moderately high TP levels (Gonzáles Sagrario et al., 2005). At a summer mean TP concentration of 30–150 μg L−1, submerged plants tend to disappear in shallow Danish lakes when the summer TN levels is above 1–2 mg L−1 (Gonzáles Sagrario et al., 2005), a threshold exceeded in most of the shallow lakes included in the present study (Fig. 4). Thus, to further improve the ecological quality of shallow lakes, it is important to consider not only P but also N loading (Moss, 2001; Gonzáles Sagrario et al., 2005). However, although a rapid decline in N is to be expected at reduced N loading, it may be difficult to achieve in practice because N in lakes typically derives from diffuse sources (Sharpley, Foy & Whiters, 2000).
The large increase in the summer DIN : TN ratio (Fig. 5; Appendix 1) indicates that denitrification could not fully compensate for the reduced uptake of DIN by phytoplankton. The insignificant response of DIN to TN loading reduction (Appendix 1) points in the same direction. No firm conclusions can be drawn on the underlying mechanisms. However, a possible explanation for shallow lakes is that the redox potential in the sediment rose because of lower sedimentation rates of organic matter and the resulting lower sediment oxygen demand. For deep lakes, reduced sedimentation of N to the hypolimnion and higher redox potentials may also play a role, and organic matter may have become limiting for denitrification in some of the deep lakes with low TP concentrations. Finally, an increase in atmospheric nitrogen pollution during the study period may play a role as suggested by results from Lago Maggiore (Mosello et al., 2001).
Probable explanations for the increase in the chl a : TP ratio that we found, notably in deep lakes (Lakes Geneva and Constance), include (i) excess SRP in the early recovery phase, (ii) access to P at high concentrations in the hypolimnion, as suggested by the frequent increase in dinophytes, which may migrate vertically in the water column (Reynolds, 2002), (iii) enhanced mixotrophy and greater P affinity of phytoplankton taxa (Reynolds, 2002; Anneville et al., 2005; Dokulil & Teubner, 2005), (iv) reduced self-shading, and (v), for some lakes, higher water temperatures resulting from global warming. In three lakes, summer chl a concentrations even increased, with the increase in one or two of them (Lake Võrtsjärv and possibly also Lake Peipsi) having been attributed to warmer climate (Kangur et al., 2002).
The phytoplankton generally followed patterns observed in earlier studies of oligotrophication (Willén, 2001b; Reynolds, 2002). Diatom growth is dependent on supplies of available silica, which tends to decrease with phosphorus enrichment (Schelske et al., 1986). Therefore, the shift in phytoplankton community structure towards diatoms that we observed at reduced SRP concentrations in shallow lakes may be because of a relaxation of silica limitation in addition to improved competitive capacity for phosphorus (Schelske & Stoermer, 1971). Notable is, however, also the observed shift in the shallow hypertrophic Lakes Gundsømagle and Søgård (Appendix 3), from chlorophytes to cyanobacteria despite a general major increase in the TN : TP ratio in the inflowing water and in the in-lake DIN : TN and DIN : SRP ratios (Appendices 1 and 3). Moreover, when TP concentrations declined and TN : TP ratios increased, N-fixing cyanobacteria were often replaced by heterocystous species (Jeppesen et al., 2002, 2005b; Phillips et al., 2005), indicating that the N : P ratio may be of minor importance for the response of cyanobacteria in these lakes. This is contrary to results obtained in some earlier multi-lake surveys (Smith, 1983; Smith et al., 1995) that encompassed a much narrower range of TP concentrations.
Fish communities showed major responses to reductions in nutrient loading. Catch data suggest that the fish biomass declined in most of the lakes for which quantitative data were available and the proportion of piscivores increased, indicating higher piscivores control of prey fish. In the eutrophic shallow Danish and Dutch lakes, the proportion of percids increased at the expense of cyprinids, and the contribution of pike and pikeperch rose in some cases as well (Jeppesen et al., 2005a; E.H.H.R. Lammens, unpublished data; R. Portielje, unpublished data). Furthermore, in the warm-temperate Albufera, an increase in littoral fish species and a higher contribution of piscivorous species were found during oligotrophication (Romo et al., 2005). In the deep mesotrophic to slightly eutrophic lakes, perch-coregonid communities changed to coregonid dominance (e.g. Lakes Constance and Geneva; Eckmann & Rösch, 1998). However, when TP was reduced further as in Lake Vättern, salmonids increased at the expense of coregonids (Degerman et al., 2001). These results follow the established fish-trophic state relationships (Colby et al., 1972; Persson et al., 1988). Our results suggest, therefore, that fish often respond rapidly to a P loading reduction, and in most cases major changes appeared in both community structure and biomass after <10 years. In subtropical shallow lakes, fish composition differs from that in northern temperate lakes (Schulz, Hoyer & Canfield, 1999), but oligotrophication can result in a similar pattern with a change in composition towards a higher contribution of piscivorous or omnivorous species.
The fast response of the fish community may challenge the idea of using fish manipulation as a lake restoration tool. There are good examples (although also many failures) showing that substantial removal of fish or stocking of piscivores or sometimes continuous fish management have improved the ecological state of lakes and sped up the recovery after reductions in nutrient loading (e.g. Hansson et al., 1998; Lammens, 1999; Meijer et al., 1999; Mehner et al., 2002). Our data indicate that major changes in the fish community often occur <10–15 years after the loading reduction, even without manipulation of the fish stock. Moreover, in lakes with short hydraulic retention times, reduced turbidity of the water column as a result of fish manipulation can result in higher accumulation of P in the surface sediments of the lake than if the water had remained turbid because less particulate P is leaving the lakes and because of higher retention capacity in the sediment (Søndergaard et al., 2003). This would result in lower losses of P via the lake outlet (Hansson et al., 1998; Meijer et al., 1999) and may thus delay the recovery in the long term if the lakes do not remain in the clear-water state (Søndergaard et al., 2003). Although biomanipulation can be a useful tool not least for rapidly shifting turbid shallow lakes to their clear alternative states (Jeppesen et al., 1990; Moss, 1990; Scheffer et al., 1993), we suggest that careful deliberation is needed on the benefits of such measures in the long term.
The greater zooplankton : phytoplankton biomass ratio, smaller chl a : TP ratio at summer mean TP concentrations <100–150 μg L−1 and the increased contribution of Daphnia to zooplankton biomass in shallow lakes are all signs of decreased fish predation on zooplankton and of an enhanced top-down control of phytoplankton (Brooks & Dodson, 1965; Carpenter & Kitchell, 1993; Gliwicz, 2003; Jeppesen et al., 2003). Changes in fish communities have the highest impact on zooplankton in shallow lakes where the overall predation risk appears to be higher (Keller & Conlon, 1994; Jeppesen et al., 2003). Thus, we hypothesise that the often strong response to nutrient loading reduction is a consequence of both enhanced resource and predation control of phytoplankton. Further studies are needed to test this hypothesis. Research is especially needed in the subtropics and tropics, where the dominant fish taxa can often graze both phytoplankton and zooplankton (Lazzaro et al., 2003) and fish predation on zooplankton is likely high even at low TP concentrations (Meerhoff et al., 2003; Jeppesen et al., 2005a), so that the typical trophic cascade type of effects may be less strong (Bays & Crisman, 1983; Crisman & Beaver, 1990).
We found indications of delayed responses of submerged macrophytes to increased water clarity in some case studies (Table 4). It may reflect lack of available seeds and turions and/or hindrance of establishment or spread by waterfowl grazing (Søndergaard et al., 1996) or competition with benthic filamentous algae. Pronounced fluctuations in biomass and species dominance in the early phase of plant re-establishment (Lauridsen et al., 2003) and low species richness at high nutrient levels may reduce the buffering capacity of plants against changes in environmental factors and thus enhance the risk of loss of the plants (Moss, 2001). Shallow eutrophic lakes may also rapidly switch states from clear to turbid and vice versa if water levels vary markedly (Blindow et al., 1993; Havens et al., 2001; Romo et al., 2004). As plants play key roles in shallow lakes (Scheffer et al., 1993; Moss, 2001), it is important to gain better insight into plant responses during recovery from excessive nutrient loading.
Phenological changes were not examined in the present analysis, but are addressed in several papers in this special issue (Anneville et al., 2005; Jeppesen et al., 2005a; Köhler et al., 2005; Phillips et al., 2005; Søndergaard et al., 2005). There is clear evidence that in deep lakes the period in which P limits phytoplankton gradually increases from initially only summer towards inclusion of spring and autumn later in the recovery phase (Anneville et al., 2005). In contrast, in shallow lakes the effect on P and plankton was initially greatest in spring and autumn, progressing later to the summer concurrently with a gradual reduction in internal P loading (Jeppesen et al., 2005a; Phillips et al., 2005; Søndergaard et al., 2005). In shallow lakes experiencing major reductions in external loading of both N and P, the effect on the concentration of TP and chl a in the early recovery phase may also be high in summer, despite a continued high internal P loading, because nitrogen can become limiting (Hameed et al., 1999; Köhler et al., 2005). The significant response, particularly in spring and autumn in shallow lakes in the early recovery phase, emphasises the importance of year-round monitoring when evaluating the effect of reductions in nutrient loading rather than only during summer, as is present practice in many countries.
The conclusions derived in the present analysis are based on correlation evidence and interpretation in terms of cause-and-effect relationships may therefore be complicated by confounding factors. One of those is global warming. Although the data presented here are too infrequent (5 year intervals) to elucidate the potential confounding effects of global warming on response to reductions in nutrient loading, climatic effects have been examined for some of the lakes included in the present data set (e.g. Straile & Adrian, 2000; Anneville et al., 2002; Kangur et al., 2002; Nõges et al., 2004). These analyses indicate an earlier onset of the clear-water phase (if any), stratification (if any) and fish spawning, reduced mixing in stratified lakes, and higher surface water temperature promoting higher internal P loading from sediment portions exposed to warm surface water. Moreover, in shallow and some deep lakes cyanobacteria may be more abundant and blooms may persist longer. However, the strong re-oligotrophication signals revealed by our analysis suggest that the observed changes in the lakes included in our data set reflect primarily the impacts of lower nutrient loadings rather than climate change. This conclusion is supported by results from mesocosm experiments which likewise suggest a much stronger effect of changing nutrient loadings than of changing temperatures in shallow lakes (McKee et al., 2003; Moss et al., 2003).
The general re-oligotrophication response patterns described here can be regarded only as a guideline when discussing the response of a particular lake. Each lake is unique in many respects and may exhibit a specific trajectory, as reflected in the appendices and figures presented (see also Moss et al., 2005). Further, although this study covers a wide range of lake types and climate zones, most of the lakes are situated in northern Europe and are relatively shallow Danish lakes in particular contributed importantly to the data set and had also been used previously to generate some of the hypotheses we examined. However, exclusion of the Danish lakes did not radically alter the overall pattern of response reported here. Reduction in the significance of trends was the most obvious difference, but can generally be attributed to reduced sample size (Appendices 1 and 3), although it could also reflect the greater depth gradient and climatic variation among the remaining lakes compared to the Danish lakes. For example we expect that the responses may differ between shallow cold-temperate and tropical or subtropical lakes, for instance owing to faster nutrient cycling and retention, better growth potential for submerged macrophytes, more truncated food webs and probably stronger top-down control of the zooplankton grazers in the warmer lakes (Lazzaro, 1997; Lazzaro et al. 2003; Meerhoff et al., 2003; Jeppesen et al., 2005b; Romo et al., 2005). Too few data from warm lakes were available for the present analysis to elucidate such differences in more detail. Thus, different patterns may emerge when examining warm-temperate or tropical lakes and very deep lakes, which also were poorly represented in the present data set.