Assessing eutrophication and reference conditions for Scottish freshwater lochs using subfossil diatoms


  • Helen Bennion,

    Corresponding author
    1. Environmental Change Research Centre, University College London, 26 Bedford Way, London WC1H 0AP, UK; and
      Dr Helen Bennion, Environmental Change Research Centre, University College London, 26 Bedford Way, London WC1H 0AP, UK (fax + 44 20 76797565; e-mail
    Search for more papers by this author
  • Jennie Fluin,

    1. Geographical and Environmental Studies, University of Adelaide, Adelaide, South Australia, Australia 5005
    Search for more papers by this author
  • Gavin L. Simpson

    1. Environmental Change Research Centre, University College London, 26 Bedford Way, London WC1H 0AP, UK; and
    Search for more papers by this author

Dr Helen Bennion, Environmental Change Research Centre, University College London, 26 Bedford Way, London WC1H 0AP, UK (fax + 44 20 76797565; e-mail


  • 1The European Council Water Framework Directive requires reference conditions to be determined for all water body types including lakes. We examined the role of palaeolimnology, specifically the diatom record, as a tool for assessing eutrophication and for defining lake reference conditions and ecological status.
  • 2Sediment cores (representing c. ad 1850 to present day) were taken from 26 Scottish freshwater loch basins. Radiometric dating techniques (210Pb and 137Cs) established a chronology for each core. Two levels of diatom analysis were employed: a relatively high resolution (15–20 samples) at 21 lochs considered of high interest, and a lower resolution (four to five samples) at the remaining sites.
  • 3Detrended correspondence analysis and dissimilarity measures were applied to the core top (present day) and bottom (reference state, c. ad 1850) samples to assess floristic change at each site. Significant floristic change, indicative of nutrient enrichment, occurred in 18 lochs along a broad trophic gradient.
  • 4Two-way indicator species analysis (twinspan) was applied to the bottom (c. ad 1850) samples to classify the ‘reference’ diatom assemblages and thereby characterize the reference floras of the different lake types. twinspan identified four site end-groups, each with a characteristic diatom assemblage, although there was some overlap in the taxa present in the four groups. Water depth and productivity were key factors that explained the groupings.
  • 5Diatom transfer functions that reconstructed total phosphorus (TP) concentrations were used to evaluate eutrophication. Nineteen lochs had increases in diatom-inferred (DI) TP of > 5 µg l−1 (five of these > 20 µg l−1), six lochs had no change or negligible increases in DI-TP (< 2 µg l−1), and there was evidence of a decline in DI-TP in one loch over the period represented by the sediment cores. The inferred increases were significant at 12 lochs.
  • 6Synthesis and applications. Our data indicate that it may be difficult to find minimally impacted waters to act as reference sites, particularly for shallow, lowland lake types, in the current population. The derivation of site-specific reference conditions from the sediment record is a particularly valuable approach in such cases. Ordination, clustering and dissimilarity measures applied to palaeodata, combined with transfer functions, offer powerful techniques for characterizing lake types, defining ecological and chemical reference conditions, and assessing deviation from the reference state.


The need to assess the current status of fresh waters relative to some baseline state in the past (Moss, Johnes & Phillips 1997; Battarbee 1999) is encompassed in recent water legislation such as the US Clean Water Act (CWA; Barbour et al. 2000) and the European Council Water Framework Directive (WFD; European Union 2000). Both require that biological, hydromorphological and chemical elements of water quality should be based on the degree to which present-day conditions deviate from those expected in the absence of significant anthropogenic influence, termed reference conditions. The WFD states that, in the absence of long-term data, reference conditions based on modelling may be derived using hindcasting methods, and palaeolimnology (the study of the lake sediment record) is given as one such technique (Pollard & Huxham 1998; European Union 2000).

Diatoms (Bacillariophyceae), unicellular, siliceous algae, are one of the most widely used biological groups in palaeolimnological studies (Stoermer & Smol 1999; Battarbee et al. 2001). Of the biological elements relevant to the WFD, diatoms represent components of both the phytoplankton and phytobenthos. However, shifts in the diatom community often correspond closely to changes in other biological groups (Kingston et al. 1992). Diatoms are sensitive to water quality and are therefore good indicators of past lake conditions, such as lake pH (Battarbee et al. 1999) and total phosphorus (TP) concentrations (Hall & Smol 1999). In recent years, transfer functions have been developed to model the relationship between diatom assemblage composition and water chemistry in a training set of lakes. Once calibrated, such functions are then applied to fossil diatom assemblages in sediment cores to infer past water chemistry. Weighted averaging (WA) regression and calibration (ter Braak & van Dam 1989) and its extension WA partial least squares (WA-PLS) (ter Braak & Juggins 1993) are the most widely used techniques for reconstructing past environmental variables in this way (Birks 1998). The diatom record is therefore a potentially useful tool for assessing water quality and defining lake reference conditions, both chemical and ecological (Kauppila, Moisio & Salonen 2002).

Fresh waters are a major feature of the Scottish landscape (approximately 30 000, with a surface area of at least 0·4 ha) and are an important resource for industry, recreation and conservation (Maitland, Boon & McLusky 1994). In the last few hundred years or so human activity has changed the ecology of Scottish lochs and it is likely that almost all surface waters in Scotland have been impacted to some extent (Bennion et al. 2002). A state-changed scheme for the classification of Scottish standing waters has recently been developed whereby lakes are classified according to degree of change in trophic status, acid neutralizing capacity and contamination with toxic substances (Fozzard et al. 1999). In a survey of 174 Scottish standing waters in 1995, and a subsequent survey in 2000, phosphorus (P) enrichment was assessed as the major cause of downgrading using the scheme (Fozzard et al. 1999; Doughty, Boon & Maitland 2002), with diffuse pollution being a major contributor to increases in TP concentrations (Ferrier & Edwards 2002). Historical TP concentrations were estimated by the land use/loss coefficient method for around ad 1850 (Ferrier et al. 1997). This is a simple approach that calculates the total load of P as a sum of the individual loads exported from each separate nutrient source in the catchment.

Palaeolimnological studies offer an alternative method for establishing reference conditions and assessing the degree of nutrient enrichment. In contrast to the loss coefficient method (Ferrier et al. 1997; Moss, Johnes & Phillips 1997), they give an indication of the ecological response to anthropogenic impacts and thus can provide an ecological target for management purposes, a concept fundamental to the WFD and CWA. While palaeolimnology has been key in establishing the extent of acidification in Scotland from acid deposition (Flower & Battarbee 1983; Battarbee et al. 1985), there have been very few such studies to assess eutrophication in Scotland (Battarbee & Allott 1994; Bennion et al. 2002). The focus of eutrophication research to date has been on productive lochs where the symptoms of enrichment, such as algal blooms, loss of submerged plants and deoxygenation, are most obvious, e.g. Loch Leven (Haworth 1972; Bailey-Watts 1994). It is now recognized, however, that factors such as fish farming, forestry fertilization, agriculture and sewage effluent disposal can potentially enrich the large, oligotrophic lochs, which are perceived as pristine waters. There is a need to assess further the extent of cultural eutrophication in Scottish fresh waters. This study examined the fossil diatom record in dated sediment cores from 26 Scottish loch basins. The role of palaeolimnology as a tool for assessing eutrophication, and for defining reference conditions and ecological status, was then evaluated.

Materials and methods

site descriptions

Scotland is diverse in its geology, soil, climate and topography and hence in its range of lake types and vegetation (Palmer, Bell & Butterfield 1992; Birks 1996). A total of 24 freshwater lochs of high environmental value and/or showing some signs of degradation, as indicated by the Scottish classification scheme (Fozzard et al. 1999), were selected for study. Lochs Lomond and Awe both have two distinct basins and these were analysed separately, resulting in a total of 26 study sites. The lochs were not purposely chosen to represent the full range of lake types found in Scotland or to cover all geographical areas. Nevertheless, the sites were distributed throughout Scotland (Fig. 1) and ranged from large, deep water bodies in glaciated valleys to small, shallow waters formed in kettle holes, and thus varied considerably in area, depth, and volume (Table 1). Further bathymetric details are given for most lochs in Murray & Pullar (1910). The lochs are generally at altitudes < 200 m a.s.l. with varying degrees of human activity in the catchments. Current mean TP concentrations range from c. 5 to c. 150 µg TP l−1 (Table 1), covering the full spectrum from oligotrophic to hypertrophic conditions (OECD 1982), and mean pH values range from c. 6·5 to c. 7·5.

Figure 1.

Location map of the study lochs showing the major topographic areas. See Table 1 for names.

Table 1.  Selected characteristics of the 26 basins
Site nameCodeUK grid referenceAltitude (m a.s.l.)Loch area (km2)Loch volume (m3 × 106)Maximum depth (m)Mean depth (m)Catchment area (km2)Mean TP (µg l−1)Analytical resolution
  • Mean TP is the best available estimate of current mean values based on Scottish Environment Protection Agency monitoring data. Analytical resolution: H = high, L = low (see text).

  • *

    Data given for whole water body, not individual basins.

1. Loch Awe NorthAWENNM 930 065 36*38*1230* 7032*816*  3H
2. Loch Awe SouthAWESNM 930 065 36*38*1230* 9432*816*  4H
3. Loch of ButterstoneBUTTNO 058 449 96 0·43   1·5  7·5 3·4 21·6 25H
4. Carlingwark LochCARLNX 765 615 45 0·40   0·87  5·5 2·1 13·8149H
5. Castle LochCASLNY 090 815 43 0·75   2·0  5·5 2·6  7·5120H
6. Castle Semple LochSEMPNS 365 590 30 0·79   0·62  2·5 0·7 77·6 59H
7. Loch DavanDAVANJ 442 007165 0·42   0·37  2·7 1·2 33·8 27H
8. Loch DoonDOONNX 495 985210 8·2  43 40 9130  4H
9. Loch EarnEARNNN 640 235100 9·5 433 8842139 11H
10. Loch EckECKNS 141 939 22 4·3  67 4515·3103  4L
11. Loch EyeEYENH 830 795 16 1·60   1·0  2·1 1·2 15·2 55H
12. Loch of HarrayHARYHY 295 155  1 9·8  27  4·3 2·8101 25H
13. Kilbirnie LochKILBNS 330 545 35 0·81   3  9·2 3·0 13·9 56H
14. Loch KinordKINONO 442 995165 0·77   1·2  3·5 1·5  8·2 20L
15. Loch LevenLEVENO 150 02510613·7  52·4 25·5 3·9159 53H
16. Loch Lomond NorthLOMONNS 365 945  8*71*2628*19037*766*  6H
17. Loch Lomond SouthLOMOSNS 365 945  8*71*2628* 2537*766*  7H
18. Loch of LowesLOWENO 049 439 99 0·88   5·5 12·5 6·2 15·5 25H
19. Loch LubnaigLUBNNN 585 130130 2·3  32·4 4213190  8H
20. Loch MareeMARENG 985 675  828·0109111238·2440  4L
21. Lake of MenteithMENTNN 580 005 20 2·5  15·9 23 6 17·9 19H
22. Mill LochMILLNY 077 833 55 0·11   1·0 16·8 7·7  1·7 92H
23. Loch RannochRANNNN 610 58020518·8 974134 51640  5L
24. Loch ShielSHIENM 866 771  4·519·9 79212840·5250  8H
25. Loch of SkeneSKENNJ 785 075 85 1·1   1·7  1·8 1·4 48·3104H
26. Loch UssieUSSINH 505 574125 0·82   1·9 12 2·4  4·7 19L

field and laboratory methods

A sediment core was taken between 1995 and 1999 from the deepest part of each loch using either a mini-Mackereth piston corer (Duncan & Associates, Grange-over-Sands, Cumbria, UK) (Mackereth 1969) at shallow sites or a Glew gravity corer (J.R. Glew, PEARL, Department of Biology, Queen's University, Kingston, Ontario, Canada) (Glew 1991) at deeper sites. All cores were extruded in the laboratory at 0·5-cm intervals from 0 to 50 cm and thereafter at 1·0-cm intervals. Selected sediment samples from each core were analysed for 210Pb, 226Ra, 137Cs and 241Am by direct gamma assay, using standard procedures, to establish a chronology (Appleby et al. 1986; Appleby, Richardson & Nolan 1992).

Subsamples from each core were prepared and analysed for diatoms using standard methods (Battarbee et al. 2001). At least 300 valves (siliceous component of the cell wall bearing the taxonomic features) were counted from each sample using a Leitz research microscope with a 100× oil immersion objective and phase contrast. Principal floras used in identification were Krammer & Lange-Bertalot (1986–91). Two levels of diatom analysis were employed (Table 1): a relatively high resolution of 15–20 samples per core at those lochs considered to be of high interest (21 sites), and a low resolution of four to five samples per core at the remaining lochs. For the former, this resolution was considered appropriate to enable the onset, rate and directions in water quality to be fully assessed and, for the latter, to allow the general trend in water quality to be determined. In all cases, samples spanning the complete length of the core were analysed.

numerical analyses

The diatom data were expressed as percentage relative abundances in all analyses. Detrended correspondence analysis (DCA; Hill & Gauch 1980) of the top and bottom samples of the cores was performed using canoco version 4.5 (ter Braak & Smilauer 2002) to assess the direction and magnitude of floristic change at each site. A total of 99 taxa (> 1% in at least two samples) was included in the analysis (see Supplementary material). The top and bottom approach can be applied to a large number of lakes as it involves the analysis of only two samples per site from a sediment core (Cumming et al. 1992). This methodology has been successfully applied by the US Environmental Protection Agency's (USEPA) Environmental Monitoring and Assessment Program for Surface Waters (EMAP-SW; Dixit et al. 1999) and in Canada to infer nutrient changes in south-eastern Ontario lakes (Reavie, Smol & Dillon 2002). The approach makes the assumption that the top and bottom samples represent the present day and reference conditions, respectively.

For the UK, it is generally agreed that approximately ad 1850 is a suitable date against which to assess impacts for lakes, as this represents a period prior to major industrialization and agricultural intensification (Battarbee 1999; Fozzard et al. 1999). Because aquatic systems have been subjected to anthropogenic impacts over much longer time scales, our reference conditions are unlikely to equate to a natural or pristine state. The radiometric measurements were used to define the core sample dated to c. ad 1850 for each loch, herein referred to as the reference sample. For some shallow loch cores it was difficult to derive accurate chronologies owing to irregular radionuclide profiles and hence there are reasonably large errors (c. ±20–30 years in the lower core sections) associated with the dates. A chronology could not be established for the Loch of Harray core. The absence of a good 210Pb record has been observed at other shallow, lowland water bodies in the UK (Bennion, Appleby & Phillips 2001) and Denmark (Anderson & Odgaard 1994). The length of the sediment record was insufficient to extend back to ad 1850 at a number of the lochs (Awe North Basin, Carlingwark, Kinord and Rannoch) and in these cases the lowermost core sample (i.e. the oldest) was taken to represent the reference conditions.

The degree of floristic change between the reference and surface sample for each of the 26 sites was assessed using a squared chord distance coefficient (Overpeck, Webb & Prentice 1985) implemented in the statistical software R (Ihaka & Gentleman 1996). This is preferred to other dissimilarity measures as it maximizes the signal to noise ratio, it performs well with percentage data and has sound mathematical properties (Overpeck, Webb & Prentice 1985). The scores range from 0 to 2, whereby 0 indicates that two samples are exactly the same and 2 that they are completely different. Scores less than 0·29, 0·39, 0·48 and 0·58 indicate insignificant floristic change at the 1st, 2·5th, 5th and 10th percentile, respectively (Simpson 2003). The 5th percentile critical limit is used here to define sites with low floristic change between the reference and surface sample.

Two-way indicator species analysis (twinspan) (Hill 1979) was employed to classify the reference samples according to their diatom assemblages and thereby characterize the reference floras of the different lake types. Relative percentage abundances of the taxa were apportioned into five classes using pseudospecies cut-off levels as follows: 0%, 2%, 5%, 10%, 20%. The resultant site end-groups were plotted as polygons on a correspondence analysis (CA) biplot (ter Braak 1987) of the reference samples, implemented using canoco version 4.5, as before.

Diatom transfer functions were applied to the diatom data for each core, following taxonomic harmonization between the training sets and the fossil data. Reconstructions of diatom-inferred TP (DI-TP) for the smaller, shallow, productive lochs were produced using a north-west European training set of 152 relatively small, shallow lakes (< 10 m maximum depth) with a median value for the data set of 104 µg TP l−1 and a root mean squared error of prediction (RMSEP) of 0·21 log10 µg TP l−1 for the weighted averaging partial least squares two-component (WA-PLS2) model (Bennion, Juggins & Anderson 1996). Reconstructions of DI-TP for the larger, deeper, less productive lochs were produced using a newly developed training set of 56 relatively large, deep lakes (> 10 m maximum depth) from Scotland, Northern Ireland, Cumbria, southern Norway and central Europe, with a median value for the data set of 22 µg TP l−1 (H. Bennion & N.J. Anderson, unpublished data). For this data set, the best model was generated with simple WA and inverse deshrinking. The RMSEP of 0·25 log10 µg TP l−1 was slightly higher than that of the former model, largely due to the relatively low number of lakes in the training set. The RMSEP values were calculated using the jack-knife, or ‘leave-one-out’, cross-validation method, which better estimates the true predictive ability of the model (ter Braak & Juggins 1993). All reconstructions were implemented using calibrate (Juggins & ter Braak 1993).

In the absence of long-term water chemistry data for model validation, the DI-TP values for the surface sample were compared with the current annual mean TP for each loch as a means of assessing model performance (Fig. 2). There was good agreement between the measured and DI-TP values (R2 = 0·91) but, in a small number of cases, the diatom model either underestimated (e.g. Loch of Skene by 34 µg TP l−1) or overestimated (e.g. Carlingwark Loch by 46 µg TP l−1) current measured values and may therefore also under- or overestimate historical values. Consequently, the difference between current DI-TP and baseline DI-TP rather than between current measured TP and baseline DI-TP was used to derive the most reliable estimate of degree of change. Change in DI-TP was considered to be significant where it was greater than the RMSEP.

Figure 2.

The relationship between measured current annual mean TP and current DI-TP based on the surface sample (µg l−1) for the study lochs. The regression line and R2 value are shown.


degree of floristic change: dca and dissimilarity measures

The DCA results revealed two clear axes of variation in the species data, with 62% and 49% of the variance explained by axis 1 and 2, respectively (Fig. 3). Sites with similar sample scores on the two axes occur in close proximity, reflecting similar diatom composition (Fig. 3a). To facilitate description, the major groups of taxa have been subjectively defined and are shown in Fig. 3b along with some of the key taxa.

Figure 3.

DCA biplot of (a) the sites and (b) the species distributions in the diatom assemblages of the top (circles) and reference (squares) samples of the 26 study sites. Arrows connect the reference and top samples for each core in (a). The direction of the arrow indicates the direction of floristic change and its length is a measure of species turnover (in Hill's standard deviation units; Hill & Gauch 1980). See Table 1 for site names. Groups of taxa (subjectively defined) are shown in (b) and taxa mentioned in the text are indicated by black triangles. Taxa that commonly occur together in the assemblages lie in close proximity.

The first DCA axis separated the deep, less productive lochs (> c. 10 m maximum depth), on the right of Fig. 3a, from the shallow, productive lochs (< c. 10 m maximum depth), on the left. The short arrows between the reference and top samples at several of the deep, oligotrophic lochs indicate that they have experienced little floristic change (e.g. Lochs Eck, Lubnaig, Maree, Rannoch and Shiel). These waters were dominated by Cyclotella comensis, Cyclotella kuetzingiana and Achnanthes minutissima in their reference and top samples (Fig. 3b), and hence had low squared chord distance dissimilarity scores (< 0·48; Table 2). A number of deep lochs, e.g. Lochs Awe, Doon, Earn, Lomond and Lake of Menteith, however, exhibited change. The changes followed similar trajectories from a CyclotellaAchnanthes assemblage to a planktonic assemblage typical of mesotrophic waters (e.g. Asterionella formosa, Aulacoseira subarctica, Fragilaria crotonensis), as indicated by the down- and leftward direction of the arrows (Fig. 3a,b). These sites had relatively high squared chord distance dissimilarity scores ranging from 0·65 to 1·63, with the exception of Loch Lomond North Basin which experienced only small floristic change (0·22; Table 2).

Table 2.  The squared chord distance dissimilarity scores between the reference and surface sample of the 26 study sites. Critical values based on the 1st, 2·5th, 5th and 10th percentiles of the distribution of randomly generated, normally distributed deviates of the range 0–2 are 0·29, 0·39, 0·48 and 0·58, respectively. Scores shown in bold are below the critical value at the 5th percentile (see text)
Site codeSquared chord distance dissimilarity score
SEMP1. 1793
LOWE1. 1771

In contrast to the deep lochs, the diatom assemblages in the reference samples of the shallow lochs largely comprised a non-planktonic community (e.g. Fragilaria, Cymbella, Cocconeis, Achnanthes, Navicula spp.). The arrows for Lochs Davan, Eye, Harray, Kinord and Ussie point directly to the left (Fig. 3a), reflecting some degree of floristic change, yet the assemblages remain largely comprised of non-planktonic taxa (Fig. 3b). The squared chord distance dissimilarity scores for these waters indicated low to moderate floristic change, ranging from 0·31 to 0·82 (Table 2). The arrows for Lochs Butterstone, Castle Semple, Kilbirnie, Leven and Lowes point in a downward direction in Fig. 3a and were generally longer than those of the former group, reflecting a greater degree of change. The squared chord distance dissimilarity scores exceeded 1·0 for some of these lochs, indicating large floristic changes between the reference and top samples (Table 2). The non-planktonic assemblage of these sites has been replaced by a plankton-dominated one (e.g. Asterionella formosa, Aulacoseira subarctica, Fragilaria crotonensis; Fig. 3b). The arrows for Lochs Carlingwark, Castle and Mill point towards the far left of Fig. 3a, reflecting a shift to an assemblage dominated by small, planktonic, highly eutrophic taxa (e.g. Stephanodiscus spp., Cyclostephanos spp., Aulacoseira granulata; Fig. 3b). The squared chord distance dissimilarity scores were particularly high (c. 1·1) for Lochs Carlingwark and Castle (Table 2).

characterization of the ‘reference’ diatom assemblages

twinspan of the reference samples resulted in four site end-groups at three levels of division (Fig. 4). The first division separated site end-groups 1 and 2, comprising largely the deep lochs, from groups 3 and 4, comprising shallow lochs. Division two further separated the deep lochs into group 1, characterized by oligotrophic, acidophilous-circumneutral taxa, particularly Cyclotella kuetzingiana, Cyclotella comensis, Tabellaria flocculosa, Achnanthes minutissima and Brachysira vitrea, and group 2, which contained taxa commonly found in slightly more productive, circumneutral conditions (e.g. Cyclotella radiosa and Aulacoseira subarctica). This division therefore appeared to be related to an acidity and nutrient gradient. Division three separated the shallow lochs into group 3, characterized by a largely non-plankton flora indicative of intermediate nutrient levels, such as Fragilaria pinnata, Fragilaria brevistriata, Fragilaria virescens var. exigua and Achnanthes minutissima, and group 4, in which a number of taxa associated with nutrient-rich waters, such as Stephanodiscus parvus, were present and taxa typically found at low nutrient concentrations, e.g. Brachysira vitrea and Cymbella microcephala, were absent. The latter two taxa, however, were found in the group 3 lochs. The group 3 and 4 separation appeared therefore to represent the difference in productivity between the two groups of lochs.

Figure 4.

(a) twinspan classification of the reference samples into four site end-groups, with summary descriptions of the site types and the associated dominant diatom taxa in each group. The number of sites and key indicator taxa (with abundance classes in parentheses; see text) at each division are shown. (b) CA biplot of the sites with the twinspan site end-groups labelled and enclosed with polygons. See Table 1 and Supplementary material (available on-line) for site and taxon full names, respectively. Also see Supplemetary material for the CA biplot of the species distributions (Fig. 4c).

chemical change: transfer function results

There was a strong relationship (R2 = 0·91) between the DI-TP value for the surface sample and the measured current annual mean TP for each loch, suggesting that the models performed well for this set of sites (Fig. 2). Several large, deep lochs experienced relatively stable conditions between c. ad 1850 and the present, with DI-TP concentrations remaining below 10 µg l−1  (Fig. 5). However, some of the large, formerly oligotrophic waters (i.e. < 10 µg l−1) exhibited marked increases in DI-TP following a period of relatively stable conditions (Fig. 5). All of the shallower waters that are currently mesotrophic (TP 10–30 µg l−1) experienced varying degrees of enrichment (Fig. 5). The currently eutrophic (TP 30–100 µg l−1) or hypertrophic (TP > 100 µg l−1) shallow lochs all exhibited increases in DI-TP, with a number of sites (e.g. Castle, Kilbirnie, Leven, Mill) showing signs of recovery (a reduction in DI-TP) over the last decade (Fig. 5). Nineteen of the 26 lochs exhibited increases in DI-TP concentrations of > 5 µg l−1 (five of these being > 20 µg l−1) and in 12 cases the change was significant (change in DI-TP > RMSEP). Six lochs experienced no change or negligible increases in DI-TP (< 2 µg l−1), and at one loch (Skene) there was a decline in DI-TP, although this was not significant (Fig. 6).

Figure 5.

DI-TP (µg l−1) reconstructions ad 1850-present day (solid lines). Values based on the large lakes model (H. Bennion & N.J. Anderson, unpublished data) are shown as circles and those based on the north-west European model (Bennion, Juggins & Anderson 1996) are shown as squares. Dashed lines indicate the RMSEP. The lochs are ordered according to increasing current measured TP. Note that in the absence of a chronology, HARY data are plotted on a depth (cm) scale.

Figure 6.

Change in TP (µg l−1) at the study lochs (diatom-inferred current TP minus diatom-inferred baseline TP). The lochs are ordered according to increasing current measured TP. Sites where the squared chord distance dissimilarity scores are below the critical value at the 5th percentile (< 0·48) are shown as white bars or have no change in DI-TP. An asterix indicates where the change in DI-TP is significant (change in DI-TP > RMSEP).


assessing eutrophication and ecological status

Minimally impacted lochs

A number of large, deep lochs in our data set appeared to have experienced very little floristic change and hence negligible change in DI-TP concentrations since c. ad 1850. Lochs Eck, Lubnaig, Maree, Rannoch and Shiel were the most stable, with DI-TP remaining at oligotrophic levels (< 10 µg l−1). The pairs of top and reference samples of these lochs were adjacent on the DCA biplot and had low dissimilarity (< 0·48). The assemblages comprised taxa typically associated with nutrient-poor, slightly acidic to circumneutral conditions. In comparison with other lochs in the data set, these waters have relatively little cultivated land (largely heather, coarse grassland and forestry) and low human populations in their catchments. The only potential nutrient inputs are sewage from seasonal tourism (campsites, small hotels) and rural dwellings, and minor diffuse inputs from agriculture and forestry fertilizer. The lochs have large volumes and appear to be able to sustain current nutrient loads without incurring high P concentrations or plankton blooms. The palaeolimnological data indicate that these lochs provide examples of minimally impacted, oligotrophic, deep waters that could be considered as reference sites for eutrophication pressures.

Enriched lochs

Nineteen lochs had increases in DI-TP concentrations, 12 of which were significant (exceeded the prediction error of the model). Seventeen of these lochs (that is all except Lochs Kinord and Ussie) experienced significant floristic change. Of the large, deep sites, marked species shifts indicative of enrichment were seen in Lochs Awe North and South Basins, Lomond South Basin, Doon and Earn, where dissimilarity scores between the reference and top samples all exceeded 0·65. The reference samples of these lochs contained nutrient-poor, circumneutral taxa, while the top samples were dominated by planktonic, mesotrophic taxa. The changes in the diatom assemblages of some of these lochs were sufficient to result in two- to threefold increases in DI-TP concentrations. The data provide evidence that while these lochs remain on the border of oligotrophic to mesotrophic, they have changed ecologically in response to recent enrichment and could not therefore be classed as reference sites. In contrast with the minimally impacted lochs, the sources of nutrients to these enriched waters are many. For example, Lochs Lomond, Awe and Earn receive diffuse agricultural and forestry inputs, and sewage effluent from villages and hotels. In the latter two cases, there are additional sources of nutrients from fish cages.

The floristic change in Loch Lomond North Basin was too subtle to cause a marked increase in DI-TP but the appearance of taxa indicative of enrichment in recent decades does give some cause for concern. Similar findings were obtained from a palaeolimnological study of Loch Ness (Jones et al. 1997), where marked changes in the composition of the planktonic diatom flora were observed over the last few decades, reflecting an increase in nutrient loading. Loch Ness is still oligotrophic, however, and, as yet, there has been no observed increase in TP concentration. Data such as these provide an early warning and indicate that an ecologically important threshold, probably related to human disturbance of the lake ecosystem, has now been crossed (Battarbee 1999).

The second group of lochs that showed signs of enrichment were the relatively shallow, currently mesotrophic waters, such as Lochs Butterstone, Davan, Harray, Lowes, Menteith and, to a lesser extent, Kinord and Ussie. The floristic change shown by the DCA biplot and the relatively high dissimilarity scores reflect an increase in diatoms associated with nutrient-rich conditions. This change resulted in a marked increase in DI-TP concentrations in these lochs of c. 10–20 µg l−1, although in all cases concentrations remained at less than 30 µg l−1. The data provide evidence of enrichment and indicate that none of the currently mesotrophic lochs in the study data set are in their reference state, although Lochs Kinord and Ussie are the least impacted in this group. The nutrient sources to these waters are site-specific but include agricultural intensification, afforestation, sewage effluent and fish cages.

The final group of enriched sites are those that are currently eutrophic or hypertrophic, such as Lochs Carlingwark, Castle, Castle Semple, Eye, Kilbirnie, Leven and Mill. The diatom assemblages of these lochs exhibited a shift from non-plankton to plankton dominance with taxa indicative of highly nutrient-rich waters, particularly small centric taxa from the genera Stephanodiscus and Cyclostephanos. The large floristic changes (dissimilarity scores > 1 in many cases) resulted in marked increases in DI-TP of between 15 and 40 µg l−1 and in the two hypertrophic sites, Castle and Carlingwark, of > 50 µg l−1. Relative to other lochs in this study these sites are in productive catchments, often close to small towns, with potential nutrient sources from sewage treatment and industrial works, rural septic tanks, agriculture and urban surface run-off (Naysmith 1999). The lochs are generally shallow and poorly flushed and thus unable to sustain high P loadings (Bailey-Watts 1994). Internal loading from the P stored in the sediments may also be a contributory factor. Floristic changes similar to those seen in this latter group of lochs occurred in the fossil diatom assemblages of the shallow Norfolk Broads in eastern England (Bennion, Appleby & Phillips 2001). The early assemblages were dominated by Fragilaria taxa and other epiphytes, coincident with documentary evidence of clear water conditions and presence of a diverse macrophyte flora. An increase in the small, centric planktonic diatom taxa was observed from the 1960s, the enrichment period during which submerged macrophytes were lost and turbid conditions prevailed.

At Loch of Skene the DI-TP concentrations were high and fluctuated between 70 and 100 µg l−1 with no clear direction of change over time. A comparison of DI-TP surface and baseline values suggested a decline of c. 15 µg l−1, although this was not significant. Despite marked changes in the diatom flora, the DI-TP results indicate no major change in TP concentrations during the last century or more. This inconsistency can be attributed to the dominance of non-planktonic Fragilaria spp. throughout the diatom record, which masks the enrichment trend revealed by the shifts in the other important taxa. Dominance of these taxa is common in shallow, alkaline, nutrient-rich lakes and they can cause problems with diatom reconstructions (Bennion 1995; Bennion, Appleby & Phillips 2001; Sayer 2001). Such taxa are poor indicators of lake trophic status as their distributions are not related directly to epilimnetic chemistry but rather to habitat availability.

In summary, the palaeolimnological data for lochs that are currently nutrient-rich show that all have been strongly impacted by anthropogenic activity since at least ad 1850 and none could be described as reference sites. Indeed, some have switched from relatively nutrient-poor systems with DI-TP baseline concentrations of < 30 µg l−1 to highly productive waters dominated by plankton.


The palaeolimnological data for Lochs Castle, Kilbirnie, Leven, and Mill suggest that these lochs may be showing recent signs of recovery from eutrophication. At Loch Leven our data are supported by reports of ecological change (e.g. lower algal abundance, increased water clarity, increased macrophyte abundance) since the reduction of point sources of P (Fozzard et al. 1999), and the DI-TP for the surface sediment sample is in agreement with the current measured mean TP concentration. Higher resolution diatom analysis for the period representing the last decade is required to ascertain whether these subtle species shifts represent a sustained recovery.

defining reference conditions

While definition of chemical reference conditions and shifts in factors such as lake TP are important for management targets, assessments of ecological reference conditions and ecological integrity are now key requirements of water legislation across Europe and North America (European Union 2000; Barbour et al. 2000). The WFD requires member states to assign all water bodies to a particular type and to define reference conditions for each. A provisional typology for lakes in Great Britain has been developed based on characteristics such as water depth and catchment geology (G. Phillips, personal communication). However, these typologies must be ecologically meaningful to the extent that particular lake types support distinct biological communities.

The combination of CA and twinspan analyses employed here facilitated characterization of reference diatom assemblages associated with different lake types. The reference floras of the deep lochs differ markedly from those of the shallow lochs. The diatom assemblages reflect not only different morphometry but also the greater range of habitats, and the higher productivity and alkalinity of the shallow waters. In this data set many of the large, deep lochs were similar ecologically in their unimpacted state. Nevertheless, there were two groups of reference assemblages for the deep lochs, one being associated with those sites on base-poor geology (group 1) and the other (group 2) associated with slightly more productive systems. The DI-TP results indicate that group 1 lochs had baseline TP concentrations of approximately 3–5 µg l−1 while group 2 lochs had slightly higher baseline TP concentrations of 6–11 µg l−1. The pre-disturbance assemblages of the shallow lochs were also divided into two groups and, as for the deep lochs, were apparently distinguished on the basis of their different productivity. Both groups were dominated by non-planktonic Fragilaria taxa but group 3 contained several typically mesotrophic taxa (e.g. Brachysira vitrea, Cymbella microcephala, Achnanthes minutissima) while group 4 included a number of taxa associated with nutrient-rich waters (e.g. Stephanodiscus parvus). Importantly, however, the small, centric taxa indicative of hypertrophic conditions (e.g. Stephanodiscus hantzschii, Cyclostephanos[cf. tholiformis], Cyclostephanos invisitatus) were found in only negligible amounts in the reference samples, demonstrating that these taxa are largely a feature of enriched (impacted) waters. The DI-TP results indicate that group 3 lochs had baseline TP concentrations in the range 10–40 µg l−1 while group 4 lochs had higher baseline TP concentrations of 30–45 µg l−1 and are therefore naturally productive systems. Given the larger errors associated with DI-TP values for the most productive waters (Figs 2 and 5), the baseline concentrations for some lochs in the latter two groups may be overestimated. For this data set, the criteria of water depth and alkalinity/productivity appear to result in ecologically meaningful typologies.

Reference conditions may refer to the best that are currently available in the current population (i.e. least impacted examples) using a spatial-state approach (Wright, Sutcliffe & Furse 2000) or can be site-specific based on historical data. Our study highlights the difficulty of finding minimally impacted examples of lowland lochs, especially the shallow systems in productive catchments, to act as reference sites because all have been impacted to some degree. It could be argued, therefore, that for such lake types the derivation of site-specific reference conditions is the more valuable approach as they can be defined without regard to the present-day status of the lakes (cf. Hughes, Paulsen & Stoddard 2000).

For most of the lochs, the cores represent conditions prior to about ad 1850 but in a few cases the core base represents only c. 1920–30 (e.g. Lochs Carlingwark and Kinord; Fig. 5). Nevertheless, the assemblages in the lower core samples represent a relatively stable phase in the loch ecosystems prior to the period of recent eutrophication (Bennion et al. 2001). In many of the enriched lochs, the major species shifts have occurred in the last two to three decades resulting from increased nutrient inputs from anthropogenic sources. Thus, the condition represented by the diatom assemblages in the reference sample acts as a scientifically derived reference state against which to assess and monitor change.

In the majority of cases, the transfer functions give reliable nutrient inferences in that DI-TP for the surface samples corresponds closely with current measured TP, and the inferred trends are supported by other historical data where available. Few models, however, are entirely error free and diatom transfer functions are no exception. Potential error sources in diatom P models include poor estimates of species optima for some taxa, inherent bias in inverse deshrinking techniques, sediment sampling and representativity problems, differential diatom dissolution, and natural high variability in TP concentrations, and these have been discussed elsewhere (Anderson, Rippey & Gibson 1993; Hall & Smol 1999; Bennion, Appleby & Phillips 2001; Sayer 2001). Nevertheless, DI-TP records have been validated against long-term monitoring data with encouraging results (Bennion, Wunsam & Schmidt 1995; Rippey, Anderson & Foy 1997; Bennion, Monteith & Appleby 2000). These studies demonstrate that DI-TP trends are not invalidated by prediction errors.


Ordination and clustering techniques and simple measures of floristic change, combined with the transfer function approach, provide valuable tools for characterizing lake types, defining ecological reference conditions and assessing degree of chemical and ecological change. The approach also lends itself to multi-indicator palaeolimnological studies, whereby other biological groups representing a range of food web components are analysed, offering great potential for defining ecological reference conditions in a more holistic way than can be achieved using the diatom record alone. Palaeolimnology can play an important role in supporting spatial-state approaches by confirming whether reference sites selected using expert knowledge, mapped information and existing data are indeed minimally impacted, and by providing site-specific reference conditions for lake types where reference sites cannot be found in the current population.


Thanks to: SEPA and SNH for provision of water chemistry and site data, ECRC colleagues for their field and laboratory support, Alex Kirika, CEH for field assistance, Peter Appleby, University of Liverpool for radiometric dating analyses, John Anderson for training set data, the Cartographic Unit at UCL for figure production, Ian Fozzard and Rick Battarbee for comments on an earlier version of the manuscript, and John Smol and John Anderson for their constructive reviews. This project was funded by the Scotland & Northern Ireland Forum for Environmental Research (SNIFFER) 1997–2000.

Supplementary material

The following material is available from /journals/suppmat/JPE/JPE874/JPE874sm.htm

Appendix. List of 99 diatom taxa used in data analysis with codes, full names and authorities.

Fig. 4. (c) CA biplot of the species distributions.