Inferring recent changes in the ecological state of 21 Danish candidate reference lakes (EU Water Framework Directive) using palaeolimnology


  • Rikke Bjerring,

    Corresponding author
    1. Department of Freshwater Ecology, National Environmental Research Institute, University of Aarhus, Vejlsøvej 25, 8600 Silkeborg, Denmark;
    2. Department of Plant Biology, University of Aarhus, Ole Worms Allé, 8000 Aarhus C, Denmark;
      *Correspondence author. E-mail:
    Search for more papers by this author
  • Emily Gwyneth Bradshaw,

    1. Quaternary Geology, Geological Survey of Denmark and Greenland, Øster Voldgade 10, 1350 Copenhagen K, Denmark;
    2. Department of Geography, Loughborough University, Loughborough LE11 3TU, UK; and
    Search for more papers by this author
  • Susanne Lildal Amsinck,

    1. Department of Freshwater Ecology, National Environmental Research Institute, University of Aarhus, Vejlsøvej 25, 8600 Silkeborg, Denmark;
    Search for more papers by this author
  • Liselotte Sander Johansson,

    1. Department of Freshwater Ecology, National Environmental Research Institute, University of Aarhus, Vejlsøvej 25, 8600 Silkeborg, Denmark;
    Search for more papers by this author
  • Bent Vad Odgaard,

    1. Department of Earth Sciences, University of Aarhus, C.F. Møllers Allé 120, 8000 Aarhus C, Denmark
    Search for more papers by this author
  • Anne Birgitte Nielsen,

    1. Quaternary Geology, Geological Survey of Denmark and Greenland, Øster Voldgade 10, 1350 Copenhagen K, Denmark;
    Search for more papers by this author
  • Erik Jeppesen

    1. Department of Freshwater Ecology, National Environmental Research Institute, University of Aarhus, Vejlsøvej 25, 8600 Silkeborg, Denmark;
    2. Department of Plant Biology, University of Aarhus, Ole Worms Allé, 8000 Aarhus C, Denmark;
    Search for more papers by this author

*Correspondence author. E-mail:


  • 1The European Water Framework Directive (WFD) requires that all natural European waterbodies should be assigned to one of five ecological categories defining the degree to which present-day conditions deviate from those uninfluenced or only negligibly impacted by anthropogenic activities (the reference condition). By 2015, all relevant waterbodies must have obtained ‘good’ ecological quality.
  • 2We describe the changes in ecological state in 21 Danish lakes using ad 1850 as a benchmark for reference conditions. Sediment samples representing 1850, 1900, 1950 and 2000 were analysed for diatom and cladoceran subfossils. Ecological status since 1850 was evaluated using correspondence analysis and dissimilarity measures to assess assemblage changes, and existing transfer functions were applied to infer changes in total phosphorous concentrations from diatoms (DI-TP) and submerged macrophyte coverage (SUB-COV) and benthi-planktivorous fish catch per unit effort (BP-CPUE) from cladoceran subfossils.
  • 3Eighteen lakes underwent significant changes, most markedly during the past 50–100 years, in either or both diatom and cladoceran community structure. Low floristic and faunal alteration was found only in three lakes; these were, however, already nutrient-rich in 1850.
  • 4In 1850, most lakes were already characterized by high DI-TP (median of 17 lakes = 86 µg TP L−1), high inferred BP-CPUE and low inferred SUB-COV, and these eutrophic conditions still prevail. In addition, the accumulation rate of sediment and cladoceran subfossils and the pelagic dominance of diatoms and cladocerans have increased.
  • 5When applying the thresholds proposed by a recent WFD classification for Danish lakes to the DI-TP values, only one lake could be described as having a ‘good’ ecological state with a concurrent low community change since 1850, limited to the cladoceran community, however. This suggests that this lake alone may serve as a potential reference site.
  • 6Synthesis and applications. Our study, demonstrating the potential of a palaeolimnological approach to assess deviations from reference conditions, suggests that Danish reference lakes may be difficult to find, most probably due to the country's long history of cultural impact. Lake managers consequently face great challenges in their endeavour to ensure ‘good’ ecological state by 2015. Therefore, further restrictions on land-use and nutrient loading in lake catchments are needed as is the initiation of restoration activities to improve the ecological state of the lakes.


The European Water Framework Directive (WFD) requires that all relevant European waterbodies, including lakes, should obtain ‘good ecological quality’ by 2015 (European Union 2000). Many European lakes are eutrophic due to excessive external loading from sewage and agricultural run-off (Schindler 2006). The WFD demands an improvement of ecological status and prevention of further deterioration.

The WFD stipulates that ‘ecological quality’ needs to be determined for individual lake types and that all relevant lakes should be assigned to one of five categories: ‘high’, ‘good’, ‘moderate’, ‘poor’ or ‘bad’. These categories are defined primarily by chemical and biological indicators supplemented by the degree to which present-day conditions deviate from those uninfluenced by anthropogenic activities (the so-called reference conditions). The target categories ‘high’ and ‘good’ are defined as ‘no, or only very minor anthropogenic alterations’ and ‘deviate only slightly from undisturbed conditions’, respectively.

For Danish lakes, Søndergaard et al. (2005a) used data from 709 lakes to propose boundaries for the WFD ecological classification. However, determination of reference conditions for each individual lake type in Denmark requires further detailed studies. Lack of monitoring extending back to pre-disturbed times and few historical data limit our insight into reference conditions. Recently, however, palaeolimnological approaches were applied to define the WFD reference state in British, Irish and Finnish lakes (Bennion, Fluin & Simpson 2004; Simpson et al. 2005; Leira et al. 2006; Räsänen, Kauppila & Salonen 2006; Taylor et al. 2006; Bennion & Battarbee 2007). Only a few of these lakes were found to be in reference condition when compared against a baseline date of 1850, a period prior to major industrialization and agricultural intensification, either due to eutrophication (Finland, Scotland, Ireland) or acidification (UK, Ireland). Palaeolimnological studies conducted on Danish lakes indicate that also in Denmark minimally impacted sites may prove difficult to find, even over a scale of centuries to millennia (Amsinck et al. 2003; Bradshaw, Rasmussen & Odgaard 2005). However, those studies varied in their aims, experimental set-up and applied biological proxies and were not designed to specifically address 1850 as the reference state target.

In the present study, we examined diatom and cladoceran assemblages in 210Pb-dated sediment cores of 21 Danish lakes over the last 150 years. Using the pre-industrial period ad 1850 as the baseline date, our aim was to elucidate whether present-day potential reference sites could be identified by assessing the deviation in ecological state in 21 lakes. Correspondence analysis and dissimilarity measures were applied to assess the floristic and faunal changes occurring at each site since 1850, and existing transfer functions were used to infer changes in total phosphorous concentrations from diatoms (DI-TP) and in submerged macrophyte coverage (SUB-COV) and benthi-planktivorous fish catch per unit effort (BP-CPUE) from cladoceran subfossils. Any change in ecological classification was assessed by applying the proposed WFD thresholds for Danish lakes (Søndergaard et al. 2003, 2005a) to the inferred DI-TP and BP-CPUE values of the study lakes.

Materials and methods

study sites

The 21 study lakes were widely distributed geographically. They had no major inlets, were of uniform size, except from Hostrup Sø, had a long water residence time and were relatively deep (Table 1) (Nielsen 2003, 2004; Nielsen & Sugita 2005). As such, they were expected to be only negligibly affected by human disturbance in the landscape and by sewage input. We divided the lakes into three types: moderately to highly alkaline lakes: ALK (12 lakes), low alkaline clear water lakes: LACW (four lakes), and low alkaline coloured lakes: LAC (five lakes), based on contemporary data from the last 5–10 years (Table 1) and the thresholds given in the Danish proposal for WFD classification (Søndergaard et al. 2003, 2005a).

Table 1.  Selected characteristics of the 21 lakes divided into Alkaline lakes (ALK), Low Alkaline Clear Water lakes (LACW) and Low Alkaline Coloured lakes (LAC) sorted by ascending total P within each lake type. Total P and Chl a are expressed as summer mean values. WFD class 2000 based on thresholds of 3 to 5 contemporary variables (total P, total N, Chl a, Secchi, pH), the number of variables used are given in brackets (for further details see Materials and methods). * WFD classification precluded due to lack of lake-specific thresholds or limited availability of environmental variables
Lake no.Lake nameLake typeArea (ha)Mean depth (m)Total P (µg L−1)Chl a (µg L−1)Total alkalinity (mm L−1)pHWFD class 2000
1Nedenskov SøALK10·04·52060·98·22·4(5)
2HvidsøALK5·65·130142·68·43·8 (5)
4Avn SøALK6·53·878378·34·8(4)
5Huno SøALK5·93·580408·84·6(5)
7Søbo SøALK21·43·680502·88·44·8(5)
8Vallum SøALK11·02·187613·48·34·2(5)
10Helle SøALK26·71·52801092·08·85·0(5)
11Ormstrup SøALK13·02·0470381·27·93·8(5)
12Sønderby SøALK7·92·915001403·38·75·0(5)
13Sjørupgårde SøLACW5·050130·58·1*
14Vedsted SøLACW7·75·050200·37·0*
15Hostrup SøLACW210·02·170390·78·02·8(4)
17Velling IgelsøLAC8·815100·27·5*
21Løvenholm LangsøLAC20·84·0214370·014·3*

Historical data on land cover of catchments around ad 1800 for 18 of the 21 lakes were digitized from 1:20 000 scale parish maps (from 1770–1820) using the GIS software ‘ArcInfo’ (Nielsen 2003; Nielsen & Sugita 2005) and used as an approximation of the land cover for the 1850 samples. Percentages of land cover types were calculated on topographical catchment basis (Bradshaw, Nielsen & Anderson 2006). Modern land cover data of the lake catchments were derived from 1:25 000 digital map AIS (Aerial Information System) based on data collected during 1992–1999. Lake-specific percentages of change in heavily impacted areas (MAN: agricultural area (including dry grassland) + built-up areas) were estimated within an 1800 m radius around the lakes.

sampling and laboratory procedures

Coring was performed mid-lake between 1999 and 2001 using a HON Kajak corer (Renberg 1991) for the upper sediments and a Russian corer (Jowsey 1966) for longer cores. The cores were sliced at 2-cm intervals, and chronologies were established based on 210Pb and 137Cs dating of 5–9 samples per core. Errors on the earliest dates range from ad 1932 ± 9 years to ad 1898 ± 19 years (Nielsen & Sugita 2005). The 210Pb chronologies were extrapolated back to ad 1850 by assuming a constant sediment accumulation rate (SAR) below the base of the 210Pb record. Sediment samples from four periods were selected: 1850, 1900, 1950 and the present (~2000) for analysis of diatom and cladoceran subfossils. SAR was estimated by linear interpolation between dated samples. Further details on sediment sampling and dating can be found in Nielsen (2003).

The samples were prepared for diatom analysis following Renberg (1990) and slides were analysed under a microscope (phase contrast, 1000×). Taxonomy followed several sources including Krammer & Lange-Bertalot (1986–1991), and pelagic diatom taxa were defined as taxa known to spend at least part of their life span in the pelagic (e.g. Bradshaw & Anderson 2003). A minimum of 300 diatom valves were counted per sample and all taxa except unidentified valves were included in the data analysis.

For cladoceran subfossils (> 80 µm), approximately 5 g (wet weight) sediment was heated in 10% KOH for 20 min. To ensure reliability, total counts of rare subfossils were performed on the 140 µm fraction and the rest on subsamples (1–40% of total sample) from the 80–140 µm fractions. Subfossils were taxonomically identified in accordance with Frey (1959) and Flössner (2000) using a binocular microscope (100×) and an inverted light microscope (320×); the most representative subfossil of each taxon was used for the data analysis. The dry weight of each sample was measured to correct for water content, and accumulation of pelagic and benthic cladoceran taxa was expressed as number of subfossils cm−2 year−1 (counts g−1 dry weight multiplied by accumulation rate). Cladocerans were separated into pelagic and benthic taxa according to Flössner (2000).

data analysis

The diatom and cladoceran data were expressed as percentage relative abundances in all analyses, excepting paired t-tests of difference of means which were based on the cladoceran data expressed as subfossils cm−2 year−1.

Hill's N2 (Hill 1973) was used to estimate the change in taxa diversity of diatoms and cladocerans in the top (present day) and bottom (1850) core samples. Between-year differences (2000 minus 1850 values) in taxon diversity and SAR as well as in the accumulation rates of pelagic and benthic cladoceran subfossils were calculated and tested by paired t-tests of difference of means (P < 0·05) between the two periods using log-transformed data for each lake type separately. Bonferroni correction for multiple tests was applied.

Detrended correspondence analysis (DCA) was applied to determine whether linear or unimodal statistical techniques would be most appropriate for modelling the species’ records (values > 2 standard deviation units of the gradient length of DCA axis 1, indicating that most species respond unimodally along the gradient). Correspondence analysis (CA) based on species data from the 1850 and 2000 samples was used to assess the main direction and magnitude of floristic and faunal alteration. Developmental trends in species assemblage were identified based on the position of the 1850 sample relative to the 2000 sample. Down-weighting of rare species was applied for diatoms due to high taxon richness (a total of 149 taxa in 1850 and 2000), whereas for cladocerans (35 taxa in 1850 and 2000) taxa present in a minimum of three lakes were included. The CA for the diatom assemblage was based on all 21 study lakes, while Lake Sjørupgårde Sø was excluded from the cladoceran CA due to difficulties in identifying the abundant Bosmina (Eubosmina) to species level. All ordinations were performed using CANOCO 4·5 (ter Braak & Smilauer 2002).

Indicator taxa for the three lake types (ALK, LACW, LAC) in 1850 and 2000 were identified using an indicator species index (INDVAL) calculated by the product of relative abundance and the relative frequency of occurrence within lake types (Dufrene & Legendre 1997). Significance of taxa association for specific lake type and year was tested by permutations (250 random iterations). Taxa with an indicator value > 0·25 (Dufrene & Legendre 1997) and with P < 0·05 were considered indicator taxa. The analysis was performed in r (The r Foundation for Statistical Computing Version 2·2·0) using the labdsv package.

Squared chi-squared dissimilarity (SCD) coefficients for diatoms and cladocerans were calculated to quantify community changes between periods (1850, 1900, 1950, 2000) using the program ANALOG (version 1·6) (H.J.B. Birks & J.M. Line, unpublished). The SCD ranges from 0 to 2, indicating identical or completely different species compositions in the two samples, respectively. The critical limit for definition of sites with low community change, i.e. potential reference sites, was estimated based on the 5th percentile of the SCD distribution (see Flower, Juggins & Battarbee 1997; Bennion et al. 2004) between the 21 lakes within each year (1850, 1900, 1950 and 2000), using SCD < 0·69 for diatoms and SCD < 0·15 for cladocerans. Significant relationships between changes in assemblages and land cover were indicated by Pearson correlations (P < 0·05) between SCDs (1850–2000) and changes in MAN (1800–2000) within lake types.

Existing diatom and cladoceran transfer functions were applied to 1850 and 2000 core samples to quantify changes in nutrient and biological structure. For inference of TP, the Northern European training set of 152 shallow lakes (RMSEP of 0·21 log10 µg TP L−1) (Bennion, Juggins & Anderson 1996) was applied. The inferred values are estimated annual mean TP concentrations and thus not comparable with the measured summer mean values given in Table 1. Submerged macrophyte coverage (SUB-COV), expressed as percentage coverage (COV%), was inferred using a training set of 19 Danish lakes, including 13 macrophyte-associated and macrophyte-sediment associated cladoceran taxa (RMSEP = 0·57 log10 (COV% + 1)) for the simple weighted averaging model and inverse deshrinking (Jeppesen et al., unpublished data). Benthi-planktivorous fish abundance (BP-CPUE), expressed as catch per unit effort in multiple mesh-sized gill nets, was inferred using the training set of 31 Danish lakes, including six pelagic cladoceran taxa (RMSEP = 0·33 log10 (BP − CPUE + 1), fish net−1 night) (Jeppesen, Madsen & Jensen 1996 with minor modifications). Changes between the inferred values of DI-TP, SUB-COV and BP-CPUE in the 2000 and 1850 samples were considered significant if greater than the RMSEP. Relationships between changes in DI-TP and MAN were tested by paired t-tests of difference of means (P < 0·05).

WFD ecological classification of the lakes in 1850 and 2000 was assessed by applying inferred values of either DI-TP or BP-CPUE to the WFD classification thresholds for Danish lakes in accordance with Søndergaard et al. (2003, 2005a). Classification using SUB-COV was omitted as thresholds for macrophyte coverage were only available for the ‘high’ to ‘moderate’ ecological class (Søndergaard et al. 2003, 2005a), and classification using BP-CPUE was not conducted for LAC and LACW lakes since CPUE thresholds were only available for alkaline lakes (Søndergaard et al. 2003). In addition, WFD class in 2000 was estimated by the mean of the WFD classification (1–5: high, good, moderate, poor, bad) based on thresholds of 3 to 5 contemporary variables (total P, total N, Chl a, Secchi, pH) in accordance with Søndergaard et al. 2003, 2005a).


changes in tax a diversity and accumulation rates

In the samples dated to ad 1850, 141 diatom and 39 cladoceran taxa occurred, while 128 diatom and 40 cladoceran taxa were identified in the 2000 samples. Thirteen lakes showed a decrease in diatom taxa diversity and 11 lakes a decline in cladoceran diversity, seven of these lakes being identical (Fig. 1). The remaining lakes exhibited enhanced taxa diversity, with the exception of Huno Sø where no change in the diversity of either diatom or cladoceran taxa could be traced. The decreases in taxa diversities were about twice as high as the increases. However, for all three lake types, taxa diversity did not differ significantly between 1850 and 2000.

Figure 1.

Change in diversity (Hill's N2) in the diatom and cladoceran assemblages of the 21 study lakes. Lake number and type as in Table 1.

All lakes but three (Nedenskov Sø, Agsø, Ormstrup Sø) showed an increase in SAR (Fig. 2A), and 16 and 17 lakes showed an increase in the accumulation rates of pelagic (seven taxa) (Fig. 2B) and benthic (32 taxa) (Fig. 2C) cladocerans, respectively. The accumulation rates of cladoceran subfossils were generally highest in the ALK lakes and lowest in the acid (pH = 4·3, Table 1) LAC lake, Løvenholm Langsø. However, the increase in SAR (t = 4·43, P = 0·001, df = 11) and in the accumulation rate of pelagic cladocerans (t = 3·15, P = 0·009, df = 11) was only significant for ALK lakes – and for benthic cladocerans (t = 7·89, P = 0·0042, df = 3) only for LACW lakes.

Figure 2.

Accumulation rate (log-10 transformed) of sediment (A), pelagic cladoceran subfossils (B) and benthic cladoceran subfossils (C), and the relative distribution of pelagic diatoms (D) and pelagic cladocerans (E) of the 21 study lakes, comparing values for ad 1850 and 2000 samples. Lake number and type specified according to Table 1.

changes in community structures

The diatom community in all ALK lakes was dominated by pelagic taxa in 1850 and these remained dominant until the present day, excepting four lakes (Nedenskov Sø, Mølle Sø, Helle Sø, Ormstrup Sø) where a major shift occurred from benthic to pelagic dominance from 1850 to 2000 (Fig. 2D). Similarly, the diatom community in all but one LACW lakes (Skærsø) underwent a marked change from benthic to pelagic dominance (Fig. 2D). In the LAC lakes, the diatom community remained fairly stable and was dominated by either benthic (Skørsø, Sortesø, Løvenholm Langsø) or pelagic taxa (Velling Igelsø), whereas Agersø experienced a notable increase (from 54% to 95%) in pelagic abundance (Fig. 2D). Increasing pelagic dominance from 1850 until the present was found also for cladocerans in most study lakes (Fig. 2E). In three lakes, this pattern was particularly distinct – Møllesø, Skærsø and Velling Igelsø. One lake (Skørsø) showed a notable decrease in pelagic abundance (Fig. 2E).

Ordination by DCA based on diatom and cladoceran data, respectively, resulted in gradient lengths of the 1st DCA axis of 5·36 and 3·38 standard deviation units, suggesting use of unimodal ordination (CA). In 1850, both between and within the LAC and LACW lakes, the diatom and cladoceran species distribution in the CA ordinations showed conspicuous differences, and several of the lakes have maintained distinct assemblages characterized by acidic taxa (diatoms) or acidic/meso-oligotrophic taxa (cladocerans) until the present day (Fig. 3). However, four lakes exhibited assemblage changes in either diatoms or cladocerans resembling those of the ALK lakes, these being, respectively, Vedsted Sø and Hostrup Sø (Fig. 3A) and Skærsø and Velling Igelsø (Fig. 3B), the latter two primarily due to a major increase in Bosmina longirostris at the expense of Chydorus piger, B. longispina or Alonella excisa. In contrast to the LACW and LAC lakes, the species distribution was more similar among the ALK lakes (Fig. 3), being characterized by a change towards higher relative abundances of taxa associated with nutrient-rich conditions for both diatoms (e.g. Stephanodiscus parvus or Cyclostephanos dubius) and cladocerans (B. longirostris) (Supporting Information Fig. S1).

Figure 3.

CA ordination plots of the sample scores of ad 1850 and 2000 for diatoms (A) and cladocerans (B). The Entire CA plots shown to the right and expanded CA plots to the left. Lakes are connected by a line between the ad 1850 and 2000 samples with symbols indicating lake type and year (1850: unfilled symbols, 2000: filled symbols). Lakes numbered and lake types specified according to Table 1.

The indicator diatom species of the ALK lakes in both 1850 (S. parvus) and 2000 (S. parvus, Stephanodiscus medius) were planktonic and characteristic of alkaline eutrophic conditions, while the indicator species for the LAC lakes were associated with relatively low pH conditions in both 1850 (Eunotia pectinalis v. minor, Tabellaria flocculosa v. flocculosa, Frustulia rhomboides) and 2000 (E. pectinalis v. minor, Pinnularia microstauron v. microstauron, Eunotia curvata v. curvata) (Supporting Information Fig. S1A). Benthic diatoms were indicator species in the LACW lakes and mainly associated with multiple habitats in both 1850 (Navicula atomus, Navicula cocconeiformis, Navicula seminulum, Navicula spp., Eunotia spp.) and 2000 (Cocconeis placentula v. lineata, Navicula rhyncocephala v. rhyncocephala, Nitzschia fonticola, Cyclotella pseudostelligera), with the exception of the latter, C. pseudostelligera, which is planktonic and typical of moderate nutrient conditions. The plant-associated Ceriodaphnia was the only indicator species for cladocerans, and only in 1850 in the LAC lakes (Supporting Information Fig. S1B).

The squared chi-square distance dissimilarity scores exceeded the critical limit for either or both diatoms and cladocerans in 18 of the study lakes, indicating major floristic and faunal changes since 1850 to the present day (Fig. 4). For diatoms the change was most pronounced during the most recent 50-year period (1950–2000), while the cladoceran community changed most substantially during the last 100 years (1900–2000). Only six lakes (Huno Sø, Vedsø, Hvidsø, Agsø, Sortesø, Agersø) showed minor changes since 1850; in Agsø and Agersø only in the faunal assemblage and in Sortesø only in the floristic assemblage.

Figure 4.

Lake-specific community changes [squared chi-squared distance (SCD)] between 50-year periods and from 1850–2000 sorted after increasing total diatom community change (1850–2000) from left to right within each lake type. (A) Diatoms, (B) Cladocerans. * refers to a significantly higher SCD than the estimated critical level (horizontal dotted line). NS indicates non significant change for all time intervals. Lake number according to Table 1.

changes in environmental conditions

DI-TP concentrations were estimated for 17 lakes, SUB-COV for 14 lakes and BP-CPUE for only 12 lakes due to poor analogues between the assemblages of the training sets and core samples. DI-TP ranged from 11 to 166 µg TP L−1 in 1850 (median = 86 µg TP L−1) and from 3 to 178 µg TP L−1 (median = 82 µg TP L−1) for the modern samples (Fig. 5A). Nine lakes exhibited a significant increase (change in DI-TP > RMSEP) in DI-TP since 1850, the increase being particularly pronounced (> 30 µg TP L−1) in four lakes (Huno Sø, Helle Sø, Vedsted Sø, Skærsø). Eight lakes exhibited a significant decrease in DI-TP (five of these being ≤ 20 µg TP L−1), particularly notable for Vedsø and Sjørupgårde Sø (> 40 µg TP L−1) (Fig. 5A). Values of inferred SUB-COV were low (median = 5%) in 1850 and remained low until 2000 (median = 5%) in most of the lakes (Fig. 5B). However, three lakes (Nedenskov Sø, Velling Igelsø, Agersø) experienced a significant decrease in SUB-COV (≥ 10%), and only one lake (Hostrup Sø) exhibited a minor (< 10%) increase in SUB-COV since 1850 (Fig. 5B). Overall, inferred BP-CPUE showed high fish abundances (median = 68 fish net−1) in 1850 and these remained high or even increased significantly in six lakes (median = 77 fish net−1) (Fig. 5C) to the present day. Only four lakes (Agsø, Hvidsø, Søbo Sø, Sønderby Sø) exhibited a significant decrease in BP-CPUE.

Figure 5.

Changes of the 21 study lakes in diatom inferred total phosphorus (DI-TP) (A); cladoceran-inferred submerged macrophyte coverage (SUB-COV) (B); benthic-planktivorous fish abundance (BP-CPUE) (C); WFD classification states (1, high; 2, good; 3, moderate; 4, poor; 5, bad) estimated using inferred DI-TP (D) or BP-CPUE (E) values and the thresholds given by Søndergaard et al. (2003, 2005a) (current–ad 1850 inferred values) and human impact on land cover (MAN) (current–ad 1800 values). (F) Lake number and type specified according to Table 1.

For the majority of the study lakes, MAN increased from 1800 to 2000, the changes being most evident for Sjørupgårde Sø, Skærsø and Skørsø (Fig. 5F). No significant relation was found between changes in MAN (1800–2000) and diatom or cladoceran community changes (SCD from 1850–2000) for ALK and LAC lakes (LACW lakes was not tested due to insufficient land cover data); nor could any significant relation be traced between changes in DI-TP (2000–1850) and MAN (2000–1800).

change in wfd classification

Based on DI-TP values, most of the lakes were classified as in ‘moderate’ (four lakes), ‘poor’ (nine lakes) or ‘bad’ (one lake) state already in 1850 and, excepting Sjørupgårde Sø (Fig. 5D), did not change to ‘high’ or ‘good’ state over the years. Only one LACW lake (Skærsø) was classified as ‘high’ and only two LAC lakes as either ‘high’ (Agersø) or ‘good’ (Velling Igelsø) WFD state in 1850; in 2000 both LAC lakes were ‘high’, while Skærsø had deteriorated to ‘moderate’ state (Fig. 5D). As for the DI-TP-based WFD classification, most lakes were classified as ‘moderate’ or ‘bad’ state using BP-CPUE inferred values (Fig. 5E). Yet, when comparing the exact WFD states based on BP-CPUE to those estimated using DI-TP, there was a discrepancy, which was particularly pronounced for Hvidsø and Helle Sø (Fig. 6D,E). This is most probably due to higher uncertainties and thus lower accuracy of the BP-CPUE inference models compared to the DI-TP models (Jeppesen et al. 1996) as well as to the small size of the training set used to infer BP-CPUE. Also the training set used to infer SUB-COV was small, and BP-CPUE, SUB-COV and WFD status based on BP-CPUE inferred values should therefore be regarded with caution. Overall, the WFD class in 2000 estimated by the mean of the WFD classification based on thresholds of 3 to 5 contemporary variables (Table 1) agreed well with the WFD class estimated using DI-TP (Fig. 5D).


minimally impacted lakes

Only one study lake (Agersø) showed insignificant changes in community assemblages, restricted to the cladoceran community (Fig. 4), and concurrent assignment to ‘good’ or ‘high’ ecological WFD state since 1850 using DI-TP values and the thresholds proposed by the recent WFD classification for Danish lakes (Fig. 5D). Since 1850, the cladoceran assemblage of this current meso-eutrophic (< 40 µg TP L−1) LAC lake (Table 1) has comprised taxa typically associated with clear water conditions (A. excisa, Alonella nana, Rhynchotalona falcata), the latter two species also being indicative of acidic conditions. With the exception of a minor assemblage change between 1900 and 1950 (Fig. 4), mainly triggered by increased abundance of the pelagic Cyclotella stelligera, the diatom assemblage has shown little floristic change. The diatom assemblage included taxa typical of intermediate nutrient levels (e.g. Achnanthes minutissima, Asterionella formosa). Low DI-TP concentrations (≤ 11 µg TP L−1) (Fig. 5A) have been inferred in this lake since 1850 and a relatively low submerged macrophyte coverage (< 10%) was inferred for 2000 (Fig. 5B), reflecting the current dominance of submerged mosses in the lake's vegetation (Amsinck et al. 2003). Lack of documentary and contemporary data prevents validation of the relatively high inferred BP-CPUE values (> 70 fish net−1) (Fig. 5C). Furthermore, the BP-CPUE training set comprised mainly alkaline lakes, reducing the accuracy of the BP-CPUE inference for Agersø, and the estimates must therefore be interpreted with caution. Our study suggests that only Agersø can be regarded as minimally impacted and it is thus the only potential reference lake.

moderately to highly impacted lakes

The majority of the study lakes underwent major assemblage changes during the study period, while DI-TP (median = 86 µg TP L−1, 17 lakes) and BP-CPUE (68 fish net−1, 12 lakes) were high and SUB-COV (median = 5%, 14 lakes) was already low in 1850 (Fig. 5A–C). Thus, 18 lakes exhibited significant changes in either or both diatom and cladoceran community structure, most markedly during the past 50–100 years (Fig. 4). In ALK lakes, the assemblage changes were mainly driven by increased abundance of pelagic taxa (e.g. the diatoms S. parvus or C. dubius) and the cladoceran B. longirostris, indicative of nutrient-rich conditions. Excepting four lakes showing assemblage changes in either diatoms (Vedsted Sø, Hostrup Sø) (Fig. 3A) or cladocerans (Skærsø, Velling Igelsø) (Fig. 3B), resembling those recorded in the ALK lakes, LAC and LACW lakes have exhibited distinct assemblages typically characterized by benthic or mesotrophic pelagic diatom taxa (LACW) and/or acidic diatom taxa (LAC) or acidic/meso-oligotrophic cladoceran taxa until the present day. In addition, taxa diversity decreased in almost half of the lakes, the decrease being twice as high as in lakes experiencing an increase (Fig. 1). Furthermore, SAR and the accumulation rate of cladoceran subfossils rose (Fig. 2A–C) as did the ratio of pelagic to benthic taxa of diatoms (Fig. 2D) and cladocerans (Fig. 2E), suggesting enhanced productivity. Accordingly, most lakes belonged to the WFD ‘moderate’ to ‘bad’ ecological state already in 1850 and have shown either no improvement to the ‘good’ or ‘ high’ WFD states or have deteriorated even further (Fig. 5D).

The community structure of the diatom and cladoceran assemblages ad 1850 also reflected the early impact of eutrophication, especially in ALK lakes, which were dominated by centric diatoms such as C. dubius, Cyclotella radiosa and S. parvus, the latter functioning as an indicator taxa, and with the pelagic B. longirostris and Chydorus sphaericus being the most prevailing cladocerans. The LACW lakes were also dominated by taxa typical of mesotrophic conditions (e.g. the planktonic diatoms C. radiosa and Aulacoseira ambigua and the cladocerans Bosmina coregoni and A. excisa), in addition to the diatom Fragilaria brevistriata and the cladoceran A. nana, both associated with macrophytes. The benthic diatoms Navicula and Eunotia were identified as indicator taxa. Only few inferred values are available for LACW in 1850, but those that are obtainable indicate less eutrophic conditions than in ALK lakes (Fig. 5A). The LAC lakes were diverse in terms of diatom and cladoceran assemblages, although the most abundant species were the planktonic Cyclotella (e.g. C. comensis and C. ocellata) and plant-associated Cocconeis taxa (e.g. C. placentula and C. neodiminuta) as well as meso-oligotrophic/acidic cladoceran species such as Bosmina longispina, A. excisa, C. piger and R. falcata. Indicator taxa were acidophilous diatoms such as Eunotia pectinalis v. minor and the plant-associated cladoceran Ceriodaphnia spp. Accordingly, the few inferred values of the LAC lakes exhibited low TP regimes (Fig. 5A) and relatively high SUB-COV percentages in 1850 (Fig. 5B).

Low floristic and faunal alteration was found only in three ALK lakes (Huno Sø, Vedsø, Hvidsø), while negligible change in either diatoms or cladocerans occurred only in LAC lake Sortesø and ALK lake Agsø, respectively (Fig. 4). However, the ALK lakes were nutrient-rich (> 75 µg TP L−1) already in 1850. In agreement with this, the lakes were characterized by high abundance of the pelagic S. parvus (20–73%) and B. longirostris (51–91%). During the subsequent 150 years, Huno Sø, Hvidsø and Agsø experienced a further increase in DI-TP (> 90 µg TP L−1 in 2000), while Vedsø exhibited a decrease (from 124 to 82 µg TP L−1) (Fig. 5A). Contemporary data reveal that Sortesø is currently eutrophic (Table 1). In contrast, LAC lake Velling Igelsø exhibited clear community change and low DI-TP levels (< 20 µg TP L−1) (Fig. 5A), consequently fulfilling the ‘good’ criterion throughout the study period (Fig. 5D). However, as Velling Igelsø underwent significant floristic and faunal change (Fig. 4), the latter partly due to enhanced abundance of pelagic B. longirostris at the expense of C. piger and plant-associated cladoceran species such as Graptoleris testudinaria, Sida crystallina and A. excisa, it does not qualify as a reference site.


Despite the fact that 1850 is prior to the major industrialization and intensification of agriculture and that the catchments of the study lakes were deliberately selected to be minimally human-impacted [average agricultural area (including dry grassland) + built up areas (MAN): 47%, 1800; 54%, 2000, Fig. 5F] compared to the average Danish catchments (average MAN: 78%, 2000), our study suggests that the majority of the lakes were impacted by eutrophication already in 1850 and have further deteriorated, preventing their classification as reference sites. Our findings agree with previous palaeolimnological studies indicating early eutrophication on a centennial to millennial scale in Danish lakes (e.g. Odgaard & Rasmussen 2000; Bradshaw et al. 2005, 2006). In addition, the low proportion of lakes with insignificant community changes since 1850 concur with the findings of similar studies in Scotland and Ireland (Bennion et al. 2004; Leira et al. 2006). However, in the latter two studies, the lakes with minimal community change were oligotrophic and have remained so since 1850 whereas the Danish sites were already eutrophic in 1850.

The majority of the study lakes were in a WFD ‘moderate’ or ‘bad’ ecological state in 2000 (Table 1, Fig. 5D). Accordingly, contemporary studies document that eutrophication remains a major problem in Danish lakes (Jeppesen et al. 2005; Søndergaard, Jensen & Jeppesen 2005b) despite reduced nutrient loading to these waterbodies via implementation of improved wastewater treatment and other pollution-combating measures, especially in recent decades.

Future challenges

Enhanced temperatures (1·2 °C in yearly mean temperature since 1873, Cappelen 2002) and precipitation (109 mm increase during the last 180 years, 56 mm increase in run-off during the last 75 years, Larsen et al. 2005) may have mediated an increase in nutrient loading (Jeppesen, Søndergaard & Jensen 2003; McKee et al. 2003) and reinforced the eutrophication observed during the past century. However, the major changes in land-use and nutrient loading likely override the effect of climatic changes (Jeppesen et al. 2005). Consequently, lake managers in cultivated landscapes, such as Denmark, will face numerous challenges in their attempt to fulfil the principal goal of the WFD, the compliance of ‘good ecological quality’ in lakes by 2015 using measures such as strengthening of the restrictions on land-use and nutrient loading in lake catchments and application of differential management tools aiming to bring down the internal nutrient loading and improve the ecological state of the lakes. In addition to these, climate warming may counteract the long-term efforts made to reduce lake eutrophication (Jankowski et al. 2006).


We wish to thank John Birks for access to his program ANALOG, Anne Mette Poulsen and Tinna Christensen for manuscript editing and two anonymous reviewers for improving this manuscript. The project was funded by the Danish research projects CONWOY, AGRAR 2000 and CLEAR, the EU project EUROLIMPACS and SOAS, University of Aarhus, Denmark.