Biological responses to liming in boreal lakes: an assessment using plankton, macroinvertebrate and fish communities


*Correspondence author. E-mail:


1. Biological responses to liming of acidified surface waters are equivocal and limit the overall assessment of food web responses. In this study, we analysed community structure in limed, acidified and circumneutral lakes, based on the analyses of phytoplankton, zooplankton, macroinvertebrates (littoral, sublittoral, profundal) and fish between 2000 and 2004. We also studied associations between functional feeding groups in food webs.

2. Most univariate metrics of structure and function revealed similar community attributes among lake types, suggesting that community responses to natural recovery from acidification and liming management converge with those observed in circumneutral lakes. These trends were less clear in the multivariate analyses which showed significant community differences among lake types. For phytoplankton, these patterns were partly mediated by the invasive raphidophycean flagellate Gonyostomum semen.

3. The associations between functional feeding groups indicated less connectivity and food web complexity in limed lakes relative to the other lake types. We speculate that repeated lime applications comprise frequent pulse disturbances which offset the establishment of stable trophic relationships in the food webs of limed lakes.

4.Synthesis and applications. The limited structural and functional food web similarity among lake types supports the argument that liming constitutes an ecosystem-level disturbance. Managers should be aware of the ecosystem impacts of altered disturbance regimes when designing their management schemes because this can influence the success of restoration programmes. Furthermore, the lack of recovery, mediated in part by species invasions, suggests that impacts derived from global change are likely to lead to novel environmental situations. This calls for adaptive management strategies where managers are challenged to tackle multiple forms of anthropogenic stress simultaneously.


Anthropogenic acidification of surface waters has been a major environmental problem in northern Europe and eastern North America during the epoch of flourishing industrial activity. Acid rain impacted aquatic ecosystems by lowering pH and increasing aluminium concentrations beyond lethal toxic thresholds for organisms, leading to a loss of biodiversity and profound alteration of community structure and ecosystem processes (Schindler 1988). Although aquatic ecosystems show signs of recovery due to reduced emissions of acidifying compounds (Ormerod & Durance 2009), many countries continue to implement large-scale mitigation programmes based on lime application to surface waters and catchments (Henriksson & Brodin 1995; Sandøy & Romunstad 1995). For example, in Sweden, some 7000 lakes and 11 000 km of watercourses are limed at a yearly cost of c. 1·8 million €, in order to restore biodiversity (i.e. facilitate the recovery of acid-sensitive biota) and create conditions for recreational fishing (i.e. protect and enhance existing fish populations; Appelberg & Svensson 2001; SEPA 2007).

Liming has increased pH and alkalinity in many acidified waters resulting in improved water quality for aquatic biota. However, studies from Europe and North America have reported mixed results considering the biological responses to liming (Clair & Hindar 2005). In lakes, liming has often, but not always, induced improvements in fish (Gunn et al. 1990; Appelberg & Degerman 1991), phytoplankton (Renberg & Hultberg 1992; Järvinen, Kuoppamaki & Rask 1995), zooplankton (Stenson & Svensson 1995; Svensson & Stenson 2002) and benthic macroinvertebrates (Carbone, Keller & Griffiths 1998; Persson & Appelberg 2001). Inconsistencies of results among studies may not be surprising, however, given that abiotic and biotic constraints affect biological recovery in context-dependent ways (Yan et al. 1996, 2003; Binks, Arnott & Sprules 2005). These include fluctuations in water chemistry caused by repeated liming and re-acidification events, dispersal capacities of organisms, the characteristics of their habitats and taxon-specific time lags. Despite these inconsistencies, some generalities arise from liming. Almost all studies have shown that liming-induced community changes are not stable. Strong temporal variability, mediated by a return to an acidified state of the communities when liming was discontinued, characterize biological responses to liming (Clair & Hindar 2005). Therefore, the potential of liming seems limited to partial remediation of acidification impacts, rather than serving as an integral ecological restoration tool that favours the long-term recovery of desired ecosystem structural and functional aspects.

Furthermore, the need to continue to apply lime to maintain communities in an improved state comprises in itself a substantial ecosystem-level disturbance (Weatherly 1988), particularly in regions where acidity is attributable to natural causes (Bishop et al. 2001; McKie, Petrin & Malmqvist 2006) and/or where acidified systems show signs of natural recovery (Schindler 1997). Several lines of evidence support this conjecture. Inflow of acid water from runoff or tributaries into limed systems can create a chemical aluminium speciation that is potentially more toxic than that of untreated acid water (Rosseland et al. 1992; Teien et al. 2004). Liming can also increase turbidity and the inorganic content of particulate matter consumed by many invertebrates (Kullberg 1987), and may precipitate dissolved organic carbon, nutrients and trace metals (Kullberg et al. 1993; Wällstedt et al. 2008). Thus, liming can impact on biogeochemical and biological processes in lakes and may substantially alter the interactions between the biotic and abiotic environment.

Despite an increased knowledge of liming effects on aquatic communities and ecosystem processes, the impact on trophic relationships have been summarized only conceptually (Appelberg et al. 1993; Stenson, Svensson & Cronberg 1993). Empirical assessments of trophic relationships in pelagic and benthic food webs across lakes are largely lacking. Thus, the generality of response patterns of lake communities to liming and their effects on trophic relationships cannot be evaluated reliably. Critical to the evaluation of food web responses to liming is the ability to compare results with that of appropriate reference lakes. However, as conditions prior to anthropogenic acidification and/or liming are difficult to quantify and pre-acidification data are limited or non-existent (Guhrén, Bigler & Renberg 2007; Norberg, Bigler & Renberg 2008), a useful alternative is to benchmark the ‘success’ of liming by comparing community trajectories and food web structures between limed, acidified and circumneutral lakes. This reference lake approach (Yan et al. 1996) allows quantification of the trends and magnitudes by which communities and food web structure in limed lakes converge with targeted conditions in circumneutral lakes. Moreover, this approach will also show the differences between communities of limed lakes and those that undergo natural recovery from anthropogenic acidification. Natural recovery of the acidified lakes used in this study has been demonstrated recently (Stendera & Johnson 2008).

We studied biological responses to repeated liming across trophic levels (i.e. phytoplankton, zooplankton, invertebrate consumers, fish) and habitats (i.e. pelagic, benthic) using univariate and multivariate statistics. We use data from the Swedish Integrated Liming Effect Studies (IKEU) programme to assess community recovery patterns in limed lakes relative to acidified and circumneutral lakes over a 5-year period (2000–2004) using structural metrics, functional feeding categories and multivariate ordinations. In a next step, we determine how individual patterns of community dynamics collectively affect the direct and indirect trophic associations between functional feeding groups within the food webs of limed, acidified and circumneutral lakes. Specifically, we test the following hypotheses:

  • 1 Ecological change in response to repeated liming and natural recovery from acidification varies among groups of organisms, habitats and lake types.
  • 2 The food web structure of limed lakes differs from those in acidified and circumneutral lakes.

Materials and methods

Study lakes

Twenty-two lakes were used for this study (Appendix S1). Eleven lakes were integrated in the national liming programme launched in 1989 by the Swedish Environmental Protection Agency (Appelberg & Svensson 2001). Liming of all lakes was carried out prior to the start of the liming programme (between 1974 and 1985) consisting of point-source application of limestone powder at 2- to 3-year intervals. Seven untreated circumneutral lakes (mean annual pH 6·0–7·0) in areas with higher neutralization capacity or lower acid deposition were chosen to reflect target reference conditions for liming interventions. Four anthropogenically acidified lakes, which lacked alkalinity and had mean annual pH values <5·7, were selected to benchmark acidic conditions. Basic data for the lake groups showed that limed lakes were directly influenced by the treatment, i.e. an increase in pH, as well as in calcium concentrations and alkalinity (Appendix S1).

Sampling procedures

We evaluated phytoplankton, zooplankton and fish as well as macroinvertebrate community data in three benthic habitat types (littoral, sublittoral and profundal). Samples were collected in August (phytoplankton and zooplankton), in October (macroinvertebrates) and in July or August (fish) over a 5-year period (2000–2004). Analyses were based on biomass data for phytoplankton (mml−1), zooplankton (mm3 m−3), sublittoral and profundal macroinvertebrates (g m−2) and fish [g catch per unit effort (CPUE)] as well as semi-quantitative abundance data (individuals per CPUE) for littoral macroinvertebrates. Although the use of different units limits quantitative comparisons between organism groups, it allowed for the evaluation of efficiency of liming as a mitigation measure from a qualitative perspective (e.g. approximation of community structure and trophic relationships in limed lakes to target circumneutral conditions). Unless otherwise noted we refer to biomass for each group for simplicity.

Mid-lake water samples were taken in August in the epilimnion (0–2 m) using a Ruttner sampler. Samples were kept cool during transport to the laboratory where they were analysed for alkalinity, and concentrations of Ca, Mg, Na, K, SO4, Cl, F, NH4-N, NO2-N + NO3-N, total N, PO4-P, total P, remaining (organic) P (total P − PO4-P), Si, total organic carbon (TOC) and Chlorophyll a. Secchi depth, water temperature, dissolved oxygen concentration, conductivity and pH were measured in the lakes. All physicochemical analyses were performed at the Department of Aquatic Sciences and Assessment following international (ISO) or European (EN) standards when available (Wilander, Johnson & Goedkoop 2003).

Integrated epilimnetic phytoplankton samples (0–4 m) were also collected using a Ruttner sampler. Samples were taken from five sites in small lakes (<1 km2), pooled and a subsample was preserved in Lugol’s solution; for large lakes (>1 km2) only one sample was taken. Taxonomic analyses and enumerations were carried out under an inverted microscope using the Utermöhl technique (Olrik et al. 1989). Biovolumes were calculated from geometric shapes following Blomqvist & Herlitz (1998).

Zooplankton was sampled quantitatively using a 55-cm Plexiglas tube (ø 10 cm) equipped with a manually triggered closing mechanism. Samples were collected in the upper 8 m of the water column at 2-m intervals. Samples were pooled, screened (40 μm) and preserved in acid Lugol’s solution. Taxonomic analyses, enumeration and length measurements were carried out using an inverted microscope. Biovolumes were calculated from length measurements and known relationships for different taxa, life stages and/or size classes (Persson 2008).

Littoral macroinvertebrate samples were collected using standardized kick sampling with a hand net (0·5-mm mesh). A composite sample of five standardized kick samples (20-s duration × 1 m long × c. 0·5–1 m depth) was taken from stony, vegetation-free sites in each lake. Sublittoral macroinvertebrates were evaluated from five Ekman samples (247 cm2) taken within a 50 × 100 m rectangular area at 4–6 m depth. The sublittoral was defined as being located above the late-summer thermocline in stratified lakes. Profundal macroinvertebrate collections consisted of five Ekman samples taken within a 150 × 150 m quadrate in the deepest area of the lake. Samples were preserved in 70% ethanol in the field. All macroinvertebrate samples were sorted under 10× magnification and invertebrates were counted using dissecting and light microscopes. Organisms were identified generally to the species level, except some chironomid larvae and immature Oligochaeta.

Fish communities were sampled with Nordic benthic standard multi-mesh gillnets (Appelberg et al. 1995). A specific number of nets (= 8–48) with random distribution were deployed within specified depth strata, depending on lake area and maximum depth (Degerman, Nyberg & Appelberg 1988; Appelberg et al. 1995). Length measurements were taken for all individuals, and individual mass was estimated by using species-specific length–mass relationships. No fish data were available for one of the acidified lakes (Lake Härsvatten).

Structural community metrics and functional groups

We calculated structural metrics (total biomass and rarefied species richness) and functional groups for each community. Species richness was rarefied with EcoSim 7 (Gotelli & Entsminger 2009) to carry out standardized comparisons among lakes. Phytoplankton was divided into autotrophic, mixotrophic and heterotrophic groups following the classification of Jansson et al. (1996). Bacillariophyceae, Conjugatophyceae, Cryptophyceae, Cyanophyceae, Loxophyceae, Prasinophyceae, Xanthophyceae and Chlorophyceae were classified autotrophic, except the green algae Polytoma and Polytomella (mixotrophic taxa). Chrysophyceae, Craspedophyceae, Dinophyceae, Euglenophyceae, Haptophyceae and Raphidophyceae were considered to be mixotrophic. Taxa were classified as heterotrophic when they contained no photosynthetic apparatus (e.g. euglenozoan flagellates).

Zooplankton was divided into predators and filter-feeders (Gliwicz 1969a,b). The latter group does not necessarily ingest only bacteria and phytoplantkton but also preys on microinvertebrates. However, as filter-feeders ‘passively prey’ on microinvertebrates, they are not considered as predators in the strict sense. We considered as predators all cyclopoid copepods, the calanoids Heterocope and Eurytemora, the cladocerans Bythotrephes and Leptodora and the rotifer Asplanchna. The remaining taxa were assigned as filter-feeders.

Functional guilds of macroinvertebrates (detritivores, filter-feeders, grazers, predators) were classified according to Moog (1995). Due to the lack of size distribution data for fish it was not possible to unambiguously determine ontogenetic feeding modes (planktivory, benthivory, piscivory), and therefore no analyses of trophic guilds were carried out.

Statistical analyses

Repeated measures analysis of variance (rm-anova) was carried out in statistica v.5 (Statsoft Inc, Tulsa, OK, USA) to test for differences in physical and chemical variables, community metrics and functional groups for each of the studied communities between lake type (acidified, circumneutral, limed) and over the study years. The rm-anova models were calculated on the basis of type-III sums of squares to account for the unbalanced design. Likewise, the fairly conservative Scheffé test, which is useful for unbalanced designs, was used for pairwise comparisons. All dependent variables were log (x + 1) transformed to fulfill the requirements of parametric tests.

Non-metric multidimensional scaling (NMDS) was performed in Primer v.6 (Primer-E Ltd, Plymouth, UK) to explore the similarity of trends in community structure and function over the study period across lake types. As a nonlinear technique, NMDS ranks points in ordination space in a way that the distance between sampling points reflects community similarity. The ordination is based on a Bray-Curtis dissimilarity matrix derived from average values of all replicate lakes and square-root transformation of the sample by species matrix. In addition, a NMDS analysis was carried out for water chemistry, where the ordination is based on a Euclidean distance matrix derived from standardized and log (x + 1)-transformed water chemistry data. The final solutions for each community and the water chemistry analysis are based on 999 permutations.

Analysis of similarity (ANOSIM; 999 permutations) was also run in Primer v. 6 to test if significant differences in biomass of communities occurred between lake types. This analysis is an approximate non-parametric analogue of anovas, and it uses the R statistic to test differences between groups (= 0, no differences; = 1, all dissimilarities between groups are larger than the dissimilarity within groups). In the present study, ANOSIM was used to supplement the NMDS analyses, using the same samples. We first calculated the yearly biomass average for each lake type. This resulted in five replicates (i.e. 5 study years) × 3 lake types (acidified, circumneutral, limed) = 15 samples for the analysis. Similarity Percentage routine (SIMPER; also included in Primer v.6) was used to reveal which taxa contributed to dissimilarity between lake types.

Pearson correlation analyses between the log (x + 1)-transformed biomass data of functional groups were carried out in Statistica to reveal trends in trophic associations in acidified, circumneutral, and limed lakes as a function of habitat type and trophic position of communities in the food webs. This analysis provides insight into how liming directly or indirectly affects the trophic associations between food web components relative to acidified and circumneutral lakes. Because of the low sample size (= 4 acidified, = 7 neutral, = 11 limed lakes), the correlation analyses were carried out using the temporal replicates for each lake after having confirmed by means of a time series analyses that no temporal autocorrelation existed for a given variable (Box-Ljung statistic: <4·27, > 0·05). Trophic groups were eliminated from the analyses when they had significant temporal autocorrelation (Box-Ljung statistic range: 5·08–12·15; P level range: 0·007–0·042). This ensured that only temporally independent variables were included in the correlation analyses, while assuring that a representative sample number was available for analyses. To facilitate interpretation of the results, we focused on correlations between trophic levels and habitat types. Food web associations were quantified through the number of significant correlations from the possible number of all correlations, which was identical across lake types.


Lake characteristics and water quality

Most of the lakes had a surface area <1 km2, but a few lakes were up to five times larger. Lake Brunnsjön was the smallest lake (0·11 km2) while Lake Stengårdshultasjön was the largest (4·98 km2) (Appendix S1). The lakes also showed a depth gradient, with circumneutral Lake Stensjön being the shallowest (Zmax = 8·5 m) and the limed Lake Stora Härsjön being the deepest (Zmax = 42 m). Total P concentrations averaged 9·5, 7·4 and 8·2 μg l−1 and total N concentrations 400, 357 and 384 μg l−1, respectively, in acidified, circumneutral and limed lakes. The concentrations of these nutrients were not significantly different between lake types (Table S1). Not surprisingly, however, substantial among-lake type differences in water chemistry were observed for variables that are affected by acidification and liming. For example, the mean annual pH of acid lakes was always below 6, while circumneutral lakes had pH values between 6·4 and 7·0, and limed lakes showed a pH between 6·0 and 7·9. These pH values, but also Ca and Mg concentrations and alkalinity, were significantly different between lake types (Fig. 1a, Table S1). The differences in water chemistry among lake types were well captured in multivariate ordination space (Fig. 1b), and an ANOSIM showed significant differences in water quality between lake categories (ANOSIM: global = 0·996, < 0·001).

Figure 1.

 Water quality characteristics in the study lakes. (a) Univariate comparisons of selected water quality variables in limed, acidified and circumneutral lakes showing the means ± standard errors of 11 (limed), 7 (circumneutral) and 4 (acidified) lakes. Significant differences or similarities (revealed by the Scheffé test) between means of lake types are denoted by different letters. Units for alkalinity, Ca and Mg are milliequivalents per litre (meq l−1). (b) Non-metric multidimensional scaling (NMDS) ordination plot showing integral water quality dynamics over the study period in acidified, circumneutral and limed lakes between 2000 [00] and 2004 [04].

Univariate analyses of community metrics

Functional feeding groups and structural community metrics revealed four categories of response types to liming (Table 1). The first category consists of metrics that did not differ among lake types (both acidified and limed lakes are similar to circumneutral lakes). The second category comprises response types that indicated that limed lakes converged with circumneutral lakes and differed significantly from acid conditions (‘desirable’ outcome of liming). The third category shows response types that were similar in limed and acidified lakes, but differed significantly from circumneutral lakes (‘unsuccessful’ or ‘undesirable’ outcome of liming). Finally, the fourth category shows responses that indicate that limed lakes deviate from both acidified and circumneutral conditions (‘idiosyncratic’ responses).

Table 1.   Summary of lake type effects revealed by the rm-anova and post hoc comparisons (Scheffé test) (detailed statistics in Table S1)
  1. Ns, no significant difference between lake types. A, acidified lakes; N, circumneutral lakes; L, limed lakes; <, significantly lower; =, no significant difference between groups according to Scheffé test; –, metric not applicable. Font types used for comparisons: bold, convergence of liming with circumneutral conditions (‘desirable effects’); italics, limed lakes are within the range of acidified conditions (‘undesirable effects’); underlined, idiosyncratic conditions.

 Total biomassNsNsNsNsNsL<A = N
 Spp. Richness (rarefied)A<N = LNsNsNsNsNs
 Autotrophic biomassNs
 Heterotrophic biomassL<A = N
 Mixotrophic biomassNs
 Predator biomassNsNsNsNs
 Filter-feeder biomassL<NNsNsNs
 Grazer biomassA<N = LNsNs
 Detritivore biomassNsL = N<ANs

From the 29 metrics summarized in Table 1, 23 (79%) showed no response, 3 (10%) indicated desirable outcomes of liming, 1 (4%) showed undesirable effects and 2 (7%) highlighted idiosyncratic responses. Desirable outcomes of liming were indicated by higher phytoplankton species richness (rarefied), as well as by increases in littoral macroinvertebrate grazers and sublittoral macroinvertebrate detritivores. The undesirable outcomes of liming were indicated only by zooplankton filter-feeders, while idiosyncratic responses of liming were observed with heterotrophic phytoplankton and total fish biomass.

Multivariate ordinations

The NMDS analyses showed a clear separation of communities of acidified, circumneutral and limed lakes (Fig. 2), and the ANOSIM indicated a significant among-lake difference in community similarity (marginally different for littoral macroinvertebrates; = 0·06) (Table 2). NMDS analyses of functional feeding groups show a good overall agreement with the structural analyses, except that zooplankton and littoral macroinvertebrates in limed and circumneutral lakes were functionally similar (Appendix S2).

Figure 2.

 Non-metric multidimensional scaling (NMDS) ordinations showing temporal trends of phytoplankton, zooplankton, fish and macroinvertebrate (in three habitats) communities in acidified, circumneutral and limed lakes between 2000 [00] and 2004 [04]. Grey full arrows indicate desired recovery trajectories from acid towards circumneutral conditions. Grey dotted arrows indicate observed community trajectories resulting from liming.

Table 2.   Results of analysis of similarity (ANOSIM) tests showing the R statistic and P levels of global models and pairwise comparisons of different communities across lake types (acidified, circumneutral and limed lakes)
RP valueRP valueRP valueRP valueRP valueRP value
  1. Significant terms are highlighted in bold.

Global model0·830·0010·670·0010·680·0010·800·0010·820·0011·000·001
Acid × Neutral1·000·0080·900·0010·940·0080·990·0081·000·0081·000·008
Acid × Limed0·470·0080·670·0011·000·0080·990·0080·740·0161·000·008
Neutral × Limed0·930·0080·570·0010·350·060·410·0160·740·0081·000·008

The NMDS ordinations based on community structure showed two types of patterns. The first shows zooplankton and macroinvertebrate communities of limed lakes occupying intermediate positions between acidified and circumneutral lakes in the ordinations (Fig. 2), resulting from an incomplete species gain during the recovery process (Table S2). The second pattern shows that the phytoplankton and fish communities in limed lakes were positioned off the gradient between acidified and circumneutral conditions. This pattern was due to a relatively high proportion of species unique to each lake type and the relative contribution of species shared by lake types. For example, the raphidophycean flagellate Gonyostomum semen had a high contribution in limed (c. 42%) and acidified lakes (c. 54%) while this species occurred marginally in the circumneutral lakes. Similarly, the fish communities of limed lakes were characterized by the presence of Salmo trutta and Salvelinus alpinus and the absence of Tinca tinca relative to circumneutral lakes, leading to c. 28% dissimilarity between both lake types (Table S2).

Trophic associations in the food webs

The correlation analyses revealed differences in the associations between trophic groups among the lake categories (Fig. 3). Limed lakes had a simplified food web, as indicated by the lower number of significant correlations (= 9) between trophic levels than circumneutral (15) and acidified lakes (18). Limed lakes and circumneutral lakes deviated from acidified lakes by the fewer connections between fish and plankton, and between fish and macroinvertebrates. In contrast, limed lakes shared with acidified lakes fewer correlations between plankton and macroinvertebrates. Limed lakes also showed the lowest connectivity between macroinvertebrates in the different benthic habitat types (four significant correlations) followed by acidified (7) and circumneutral lakes (15).

Figure 3.

 Schematic food webs showing significant (< 0·05) relationships between functional groups in limed, acidified and circumneutral lakes revealed through correlation analyses. The negative signs on the arrows indicate negative correlations while positive correlations are not highlighted. The strength of correlations is indicated by the arrow dimensions: faint lines, Pearson correlation coefficients 0·27–0·50; intermediate lines, 0·51–0·74; thick lines, >0·75. Phyto (phytoplankton), – A (autotrophs), H (heterotrophs), M (mixotrophs), zoo (zooplankton), inverts (macroinvertebrates) – G (grazers), P (predators), D (detritivores), F (filter-feeders).


Our comparative study of multiple communities in limed, acidified and circumneutral lakes facilitated an important assessment of ecological responses of boreal lakes to management practices. As has been demonstrated for stream ecosystems (McKie et al. 2006; McClurg et al. 2007; Ormerod & Durance 2009), our results show that liming has a limited restoration potential in lakes.

Communities in limed, acidified and circumneutral lakes

Regarding univariate tests, the desirable outcomes of liming were only reflected in 10% of the studied metrics, while most (c. 80%) tests were not significantly different between lake types. This suggests that limed, acidified and circumneutral lakes share to a great extent similar community attributes across different trophic levels and habitat types, which may be due to an ecological compensation within trophic levels (Klug et al. 2000). This finding contributes to the debate about whether liming is required when acidified systems show signs of natural recovery. Natural recovery from acidification has been documented for European and North American waters (Stoddard et al. 1999; Davies et al. 2005). Stendera & Johnson (2008) have analysed decadal trends in community structure of the acidified lakes we have studied here, and for most communities found clear signs of recovery. Our rm-anova results suggest that natural recovery largely fulfils the desired goals without any need for management intervention. Moreover, some metrics (heterotrophic phytoplankton and total fish biomass) even indicate that liming may induce community responses that are not representative of circumneutral conditions, while natural recovery alone may not lead to such idiosyncratic responses. For example, liming programmes put much emphasis on ameliorating conditions for fish (Appelberg & Svensson 2001; SEPA 2007). While our results suggest that liming does not achieve fish standing stocks comparable with those in acidified and circumneutral lakes in Sweden (see also Persson & Appelberg 2001), liming management at least achieves the conservation goal of the Swedish Environmental Protection Agency to favour the persistence of acid-sensitive fish populations, including the salmonids Salvelinus alpinus and Salmo trutta.

Our multivariate analyses support the findings from univariate tests that liming has limited management potential from an ecosystem perspective. However, the main differences are that univariate measures indicate convergence of both acidified and limed lakes with circumneutral conditions, while the multivariate analyses do not. Both types of analyses thus provide complementary results as the multivariate statistics allow for an integral analysis of community composition that integrates species distribution patterns and their relative biomass, while the univariate tests permit a comparison of measures that emphasize different aspects of community structure. Despite the observed differences, both types of analysis showed that recovery patterns differed between communities and habitat types, thus supporting our first hypothesis.

In the NMDS, these differences were indicated by the positions of limed lakes in ordination space relative to acidified and circumneutral lakes. Phytoplankton and fish communities in limed lakes were positioned away from acidified and circumneutral lakes, whereas the other communities occupied intermediate positions between acidified and circumneutral conditions. The intermediate position of the macroinvertebrate and zooplankton communities of limed lakes illustrate a limited gain of species and different individual biomass patterns relative to circumneutral lakes (Table S2), while the phytoplankton ordination highlights additional factors that can confound liming effects. Both limed and acidified lakes showed mass occurrences of the invasive flagellate Gonyostomum semen (Raphidophyceae), suggesting that liming does not counteract other undesired effects resulting from global change (see below).

Trophic associations in the food webs

The results from this study clearly support our second hypothesis that food webs of limed lakes are different relative to those of acidified and circumneutral lakes. Limed lakes showed less complexity, i.e. fewer linkages between trophic guilds relative to acidified and circumneutral lakes. Several examples in the literature show the inverse relationship between the importance of environmental harshness and biological interaction in regulating ecological communities (Menge & Sutherland 1987; Peckarsky, Horn & Statzner 1990; Angeler et al. 2000). Based on this evidence, acidified lakes would be expected to be the most disturbed with their food webs being structurally and functionally the most altered (Stenson et al. 1993). Several arguments, including disturbance and recovery regimes, may help interpret our contradictory findings.

The acidified lakes studied here are undergoing a gradual recovery process (Stendera & Johnson 2008), corresponding to a simplified ramp-type response (Lake 2000), to the reduced deposition of acidifying compounds. By contrast, liming implies recurrent management interventions to avoid re-acidification of lakes, reflecting recurrent pulse disturbances. The results suggest that biological responses to liming, which can be also affected by intermittent re-acidification events (Clair & Hindar 2005), are also of pulse type. Our results suggest that ramp-type community responses could potentially lead to a gradual build up of trophic relationships in acidified lakes, while repetitive disturbance in the form of lime applications could disrupt the establishment of stable trophic relationships in limed lakes, thereby reducing their overall food web complexity. The observed decrease of phytoplankton biomass and associated changes in community structure immediately after liming events, mediated by a co-precipitation of P with lime addition, supports this argument (Hörnström et al. 1993; Persson & Appelberg 2001). The altered disturbance regime in limed lakes may also provide temporal windows where the mass development of Gonyostomum semen is favoured. This supports the conjecture that disturbed systems are more sensitive to the colonization and successful establishment of invasive species (Altman & Whitlatch 2007).

Management implications

Ecosystem-level responses are described, based on the trophic associations that emerge from recovering communities, among managed (limed) lakes, unmanaged lakes that show natural recovery (acidified lakes) and least-impacted sites (circumneutral lakes). Acidified lakes that undergo natural recovery may gradually build up trophic relationships, while in limed lakes the establishment of such relationships is limited by the nature of the disturbance regime imposed by management. We acknowledge that the liming method (lake vs. catchment application) probably has a strong influence on the disturbance regime, affecting both the frequency and magnitude by which combined re-acidification events and liming impacts collectively affect ecosystems, but we could not consider this aspect in our study. We emphasize that managers should be aware of the ecosystem impacts of altered disturbance regimes when designing management schemes because biological responses, and therefore the success of restoration programmes, depend to a great extent on levels of disturbance (Lake 2000).

Our results also suggest that other forms of anthropogenic stress, including species invasions, will probably lead to novel environmental situations, which increase the uncertainty and predictability of management schemes (Harris et al. 2006). Global climate change is likely to add further confounding factors (Skjelkvåle et al. 2003; Angeler 2007). Thus, managing boreal lakes to address a single environmental problem is likely to become obsolete. Managers are challenged to re-evaluate the benefits of liming and its contribution to broader management schemes that tackle several forms of anthropogenic stress simultaneously.


We thank Cristina Trigal, Kerstin Holmgren, Marcus Sundbom, Richard K. Johnson and the referees for helpful comments on an earlier draft of the manuscript. This work was supported by the Swedish Environmental Protection Agency through the IKEU programme.