J Paleolimnol (2011) 45:415–431 DOI 10.1007/s10933-010-9418-4
ORIGINAL PAPER
Defining ecological and chemical reference conditions and restoration targets for nine European lakes Helen Bennion • Gavin L. Simpson • N. John Anderson • Gina Clarke Xuhui Dong • Anders Hobæk • Piero Guilizzoni • Aldo Marchetto • Carl D. Sayer • Hansjo¨rg Thies • Monica Tolotti
•
Received: 25 February 2009 / Accepted: 3 July 2009 / Published online: 27 April 2010 Ó Springer Science+Business Media B.V. 2010
Abstract This paper aims to determine the ecological and chemical reference conditions (*1800–1850 AD) and degree of floristic change at nine enriched lakes, covering a range of types across Europe, using fossil diatom assemblages in dated sediment cores and application of total phosphorus (TP) transfer functions. Additionally the study assesses the potential of analogue matching as a technique for identifying reference sites and for estimating reference TP
concentrations for the study lakes using a training set of 347 European lakes and 719 diatom taxa. Oligotrophic, acidophilous to circumneutral taxa were predominant in the reference samples of several of the deep lakes, and benthic Fragilaria spp. dominated the reference samples of two high alkalinity shallow lakes. The degree of floristic change from the reference sample, assessed using the squared chord distance (SCD) dissimilarity coefficient, revealed that two sites
H. Bennion (&) G. L. Simpson G. Clarke X. Dong C. D. Sayer Environmental Change Research Centre, Department of Geography, University College London, Gower Street, London WC1E 6BT, UK e-mail:
[email protected]
A. Hobæk Department of Biology, University of Bergen, P.O. Box 7800, 5020 Bergen, Norway
G. L. Simpson e-mail:
[email protected] G. Clarke e-mail:
[email protected] X. Dong e-mail:
[email protected] C. D. Sayer e-mail:
[email protected] N. John Anderson Department of Geography, Loughborough University, Loughborough LE11 3TU, UK e-mail:
[email protected] A. Hobæk Norwegian Institute for Water Research (NIVA), P.O.Box 2026, 5817 Nordnes, Bergen, Norway e-mail:
[email protected]
P. Guilizzoni A. Marchetto CNR-Istituto per lo Studio degli Ecosistemi, Largo Tonolli 50, 28922 Verbania Pallanza, Italy e-mail:
[email protected] A. Marchetto e-mail:
[email protected] H. Thies Institute of Ecology, University of Innsbruck, Technikerstr. 25, 6020 Innsbruck, Austria e-mail:
[email protected] M. Tolotti Fondazione E. Mach, Istituto Agrario di S. Michele all’Adige, IASMA Research and Innovation Centre, Environment and Natural Resources Area, Via Mach 1, 38010 S. Michele all’Adige, Trento, Italy e-mail:
[email protected]
123
416
had experienced slight change (Lago Maggiore, Felbrigg Lake), five experienced moderate change (Mjoesa, Loch Davan, Loch Leven, White Lough, Esthwaite Water), and two showed evidence of major change (Groby Pool, Piburger See). For three lakes, there were no analogues in the diatom dataset owing to the uniqueness and diversity of the diatom reference assemblages. For the remaining six sites the number of analogues ranged from 2 to 44. For two deep lakes most of the analogues seemed appropriate as they were of the same type and had low TP concentrations. However, for two other deep lakes and two shallow lakes some of the analogues differed markedly in their depth and alkalinity from the lake in question or had TP concentrations seemingly too high to represent reference conditions suggesting that the analogues may not be suitable as reference sites. For the deep lakes, similar reference TP values were calculated using the EDDI Combined TP transfer function and the analogue matching technique with concentrations typically\20 lg L-1. However, for the shallow lakes, the analogue matching method produced inferred values considerably higher than those of the transfer function. The wide ecological tolerances of many of the diatom taxa found in the reference samples most likely explain the selection of inappropriate analogue sites. In summary, the study demonstrates that palaeoecological techniques can play a valuable role in determining reference conditions and indicates that the analogue matching technique has the potential to be a useful tool for identifying appropriate reference sites for lakes impacted by eutrophication. Keywords Analogue matching Diatoms Eutrophication Reference conditions Restoration targets Introduction Since the advent of the European Water Framework Directive (WFD) (European Union 2000) there has been an expansion of research on the ecological assessment of aquatic ecosystems. One of the key themes has been the establishment of reference conditions, defined by the WFD as those associated with no, or only very minor, anthropogenic impact, to anchor judgements of change. The WFD requires the definition of reference biological communities and supporting hydrochemical conditions for each
123
J Paleolimnol (2011) 45:415–431
waterbody type, the deviation from which is used to assess ecological status according to five bands (high, good, moderate, poor and bad). A number of approaches can be employed to establish reference conditions including spatial surveys, modelling, expert judgement and temporally based methods using historical data or palaeoreconstruction (e.g. Pollard and Huxham 1998; Anonymous 2003). However, since much of the European landscape has been subject to human impact for centuries, it is difficult to find minimally disturbed sites for the majority of ecosystem types. In regions where potential reference sites are few or lacking, palaeoecology provides arguably the only technique for establishing a reference condition with any confidence. In recent years, therefore, analyses of biological remains in sediment records, combined with transfer functions, have been employed at numerous lakes to define ecological and chemical reference conditions and assess deviation from the reference state (see review in Bennion and Battarbee 2007). A further method for establishing reference conditions from palaeo-data is to employ analogue matching. This is a form of space-for-time substitution, whereby fossil assemblages present in sediments deposited in a pre-disturbance period are matched with those assemblages found in modern surface sediments of pristine or less impacted lakes. The best modern analogues for the pre-disturbance assemblages are identified. These analogue lakes are then assumed to have similar community composition to those present in the pre-disturbance period of the impacted lake in biological groups other than those typically analysed in the palaeo-record (e.g. fish, benthic macroinvertebrates). The technique has been developed largely for identifying restoration targets for acidified lakes based on diatom analogues (Flower et al. 1997) or a combination of diatom and Cladocera assemblages (Simpson et al. 2005) and has proved to be a powerful technique for identifying site-specific ecological reference conditions in these waters. To date, with the exception of Sayer and Roberts (2001) for a small set of shallow English lakes, the method has not been applied to determine pre-disturbance communities in enriched lakes. Hence, in this paper we explore the potential of the method for identifying appropriate reference sites for a set of European lakes impacted by eutrophication, alongside the well established methods of deriving
Water depth in which cores were taken, where these differ from maximum water depth, is given in parentheses
Lake typology after Battarbee et al. (2010) b
45.9133N 8.5492E
a
20
63 1998 Medium alkalinity, low altitude, deep, large 10 194 370 (89)
46
Lago Maggiore
212.5
2004 Medium alkalinity, high altitude, deep, small 6 913 10.83E 47.17N
24
58
Piburger See
0.13
1995 High alkalinity, low altitude, shallow, small 200 95 52.6667N 1.2167W
1.8
30
Groby Pool
0.12
1998
2006 Medium alkalinity, low altitude, deep, large
High alkalinity, low altitude, shallow, small 107
32 65
65 1996 High alkalinity, low altitude, deep, small 60 100
50 1.2 0.03 52.9028N 1.25516E Felbrigg Lake
15.5 (14) 1 Esthwaite Water 54.3593N 2.9858W
10.8495E
6.91524W 0.07 54.416N White Lough
11 (10)
10 2006 Low alkalinity, low altitude, deep, large
Medium alkalinity, low altitude, shallow, small 1998 High alkalinity, low altitude, deep, large 1999 25 32
4 122
165 107 2.7 (2) 25.5 (4)
453 (236)
57.0944N 2.92302W 0.42 56.1981N 3.37586W 13.2 Loch Davan Loch Leven
362 60.766N Mjoesa
Longitude Area (km2) Max deptha (m) Alt (m) Current mean Lake typeb TP (lg L-1) Latitude
The nine study lakes were selected on the basis of their well documented eutrophication histories and the availability of existing dated sediment cores (Table 1; Fig. 1). Six are included within the EU project Eurolimpacs, covering a range of lake types and a broad gradient from Norway in Northern Europe to Austria and Italy in the South. The remaining three sites (Loch Davan, Felbrigg Lake and Groby Pool) are small, shallow lakes in the UK and represent a lake type for which reference sites are usually difficult to find (e.g. Bennion et al. 2004). Diatom analysis was undertaken on the full core sequence and the diatom assemblages in the pre-enrichment sample (typically *1800–1850 AD) were chosen to represent the reference samples (Table 1). All diatom samples were prepared and analysed using standard techniques (Battarbee et al. 2001a). For analogue matching, the ‘Combined total phosphorus (TP)’ dataset from the EDDI database is used as the modern training set (Battarbee et al. 2001b; http://craticula.ncl.ac.uk/Eddi/jsp/). This dataset is comprised of 347 lakes and 719 diatom taxa, spanning a broad range of lake types from several European countries including the UK, Denmark, France, Switzerland, Austria, Germany and Italy (Fig. 1). This training set contains a mixture of impacted and relatively unimpacted lakes and covers a wide range of TP concentrations (2–1,189 lg L-1, mean 99 lg L-1). All data are expressed as percentage relative abundances. The diatom names follow the old nomenclature, prior to the revisions of Round et al. (1990), as several of the datasets in the EDDI database pre-date the revisions and all taxonomy was harmonised with the EDDI Combined TP training set prior to data analysis. The degree of similarity (or dissimilarity) in the diatom communities between each fossil (reference) sample from the nine lakes and every sample in the modern training set was calculated using the squared chord distance (SCD) dissimilarity measure (Overpeck et al. 1985; Gavin et al. 2003; Wahl 2004), computed using R (Version 2.8.1; R Development Core Team 2008) and the analogue package version 0.6–3 (Simpson 2007; Simpson and Oksanen 2009). The SCD is a signal to noise coefficient (sensu.
Table 1 Basic characteristics of the nine study sites (ordered by latitude) with coring date and depth of reference sample
Methods
Site name
reference conditions and assessing degree of change from palaeoecological data.
70 89
417 Coring date Reference sample depth (cm)
J Paleolimnol (2011) 45:415–431
123
418
Fig. 1 Location map of the nine study sites and the sites in the EDDI Combined TP dataset
Overpeck et al. 1985) and attempts to emphasize the signal or pattern in the data at the expense of the noise or random variation in species abundances. Prentice (1980) and Overpeck et al. (1985) have shown theoretically that signal to noise coefficients have desirable properties for determining ecological resemblance amongst samples and that these performed better than other types of coefficients in their tests at distinguishing between similar and nonsimilar samples. This is not universally the case however (Faith et al. 1987). Having determined the SCD between each fossil sample and every sample in the modern training set, those lakes in the training set that are sufficiently similar to the fossil sample are then selected as modern analogues, i.e. those samples that have a SCD less than or equal to a critical value. There is no universally accepted means of determining a critical value where the samples can not be grouped a priori. As dissimilarity is a function of the number of taxa in the training
123
J Paleolimnol (2011) 45:415–431
set and the diversity of the individual samples, critical values need to be determined for each application. Here we used the distribution of the pairwise dissimilarity values for the training set to guide our choice of critical value. This distribution was strongly left skewed, with many large and few small pairwise dissimilarities (mean = 1.46, median = 1.53, mode = 1.72, inter-quartile range = 1.23–1.75). In such circumstances, low percentiles of this distribution tend to be pulled towards larger dissimilarities. This reflects the wide range of sites and nutrient conditions sampled in the EDDI training set, where only small groups of samples will be similar to one another. Therefore, we adopted the 5th percentile of this distribution as our critical value for determining analogue sites, yielding a value of 0.75. The 1st and 2.5th percentiles of the distribution correspond to SCDs of 0.45 and 0.60, which are used to identify very close and close matches, respectively. The degree of floristic change in the diatom assemblages between the reference sample and the surface sample of each core was also assessed using the SCD. The degree and direction of floristic change was plotted in ordination space along with the samples in the EDDI Combined TP dataset using principal components analysis (PCA). PCA plots were also constructed to illustrate the position of the reference sample in relation to the analogue samples for each of the study sites. Lower dissimilarity cutoffs than the critical value of 0.75 for analogue matching were used to assess the degree of floristic change between reference and surface samples. This is because we do not need to account for the inherent natural differences between lakes when determining if two samples are sufficiently similar within a lake. Observations from minimally impacted lakes in the UK, for example, have very low species turnover and consequently low dissimilarity between the 1850 core sample and the core top (in many cases SCD \ 0.1), reflecting natural, small scale variations in the species assemblages with time, whilst impacted lakes typically have high dissimilarity between core bottom and top samples with SCD values in excess of 1 (Bennion et al. 2003). Hence, SCDs less than the 1st percentile (0.45) were used to indicate minimal change over reference, those between 0.45 and 0.6 (2.5th percentile) slight change, those between 0.6 and 0.93 (10th percentile) moderate change and those greater than 0.93 were used to indicate major change.
J Paleolimnol (2011) 45:415–431
419
Table 2 Performance statistics of the diatom-TP transfer functions (log10TP lg L-1)
RMSEP (jack) 2
WA
LWWA
0.334
0.299
r (jack)
0.637
0.717
Average bias
0.003
0.013
Maximum bias
0.719
0.649
To define chemical reference conditions, TP concentrations were estimated for each of the nine lakes by application of the EDDI Combined TP training set to the fossil assemblages in the reference samples using the EDDI web site (http://craticula. ncl.ac.uk/Eddi/jsp/). The results of the simple weighted averaging (WA) and locally-weighted WA (LWWA) inverse regression models are presented. LWWA generates a dynamic training set optimised for each fossil sample by choosing the 50 closest (smallest chi-squared distance) modern samples to each fossil sample and uses these in a WA reconstruction (Battarbee et al. 2005). Comparisons suggest that this method is more effective when applied to the large merged EDDI datasets than simply fitting a single global transfer function (Juggins 2001). The models are based on log10-transformed TP data. The error associated with the models is expressed as the root mean squared error of prediction (RMSEP) calculated using the jack-knife cross-validation method (see Table 2 for performance statistics). The diatom-inferred TP (DI-TP) concentrations from the WA and LWWA models are compared with those estimated using the analogue matching technique. For the latter, for each site, two sets of values are calculated based on, (1) the mean of the current TP concentrations for the analogue sites and (2) a weighted mean with weights being the inverse of the dissimilarity so that closer analogue sites contribute more to the value.
Results Reference diatom assemblages and floristic change The reference sample of Mjoesa is dominated by taxa typically associated with nutrient-poor conditions (e.g. Achnanthes minutissima, Cyclotella comensis,
Aulacoseira islandica, Aulacoseira subarctica) (Fig. 2). Whilst many of these taxa are also present in the surface sample, the latter contains several species indicative of more productive waters namely Tabellaria flocculosa, Fragilaria crotonensis, and Stephanodiscus parvus. The SCD score of 0.76 reflects this moderate floristic change (Table 3). The reference sample of Loch Davan is dominated by non-planktonic taxa ([90%). The assemblage is very diverse with occurrence of a large number of taxa in relatively low percentages but the most abundant species are Achnanthes minutissima and benthic Fragilaria spp. (Fig. 2). The small planktonic component is comprised of Cyclotella taxa (C. comensis, C. pseudostelligera, C. radiosa) and Tabellaria flocculosa, indicative of waters with low to medium nutrient concentrations. Whilst the surface sample is still dominated by non-planktonic taxa, three taxa, Stephanodiscus parvus, Fragilaria capucina var. mesolepta and Synedra pulchella, not observed in the reference sample, are present and many of the formerly important non-planktonic taxa, in particular Achnanthes minutissima, decline. The SCD score of 0.76 reflects this moderate floristic change (Table 3). The reference diatom assemblage of Loch Leven is diverse, comprising both planktonic (40%) and nonplanktonic (60%) taxa. Aulacoseira subarctica is the dominant planktonic diatom (*20%), and Tabellaria flocculosa, Cyclotella comensis, Cyclotella radiosa and Stephanodiscus parvus are also present (Fig. 2). The non-planktonic flora is particularly diverse and includes the small, benthic Fragilaria spp. and the epiphytic taxa Cocconeis placentula, Achnanthes clevei and Amphora pediculus. These taxa are typically found in circumneutral to slightly alkaline waters of moderate nutrient concentrations. The diatom assemblage in the surface sample of Loch Leven is markedly different with a greater relative abundance of planktonic taxa usually associated with nutrient-rich waters, including Aulacoseira subarctica, Stephanodiscus hantzschii, Stephanodiscus parvus, Aulacoseira ambigua, and Asterionella formosa. Consequently, the relative abundances of other taxa decline, most notably the benthic Fragilaria taxa, Tabellaria flocculosa and Cyclotella comensis. The SCD score of 0.82 indicates moderate floristic change (Table 3). The reference sample of White Lough is comprised of Aulacoseira subarctica and Cyclotella
123
420
J Paleolimnol (2011) 45:415–431
Fig. 2 Summary diagram of the key diatom taxa in the reference (R) and surface (S) samples of the nine study lakes (% relative abundance)
Table 3 Summary of results including squared chord distance dissimilarity scores (SCD) between the reference (*1800– 1850 AD) and surface samples of the nine study sites (scores less than the critical value at the 2.5th and 10th percentile are indicated by * and **, respectively, and scores greater than the 10th percentile are indicated by ***, see text), number of analogues (N), range of SCD values of the analogue sites (SCD Lake
SCD
Degree of change
AM-TP (weighted)
AM-TP range
WA- WA-TP TP ±RMSEP
LWWA- LWWA-TP TP ±RMSEP
Mjoesa
0.76**
Moderate
2
18
11–25
26
12–56
17
9–32
Loch Davan
0.76**
Moderate
0 [0.75
Na
Na
Na
18
8–39
18
11–30
Loch Leven
0.82**
Moderate
0 [0.75
Na
Na
Na
47
21–103
38
19–76
White Lough
0.80**
Moderate
3
0.49–0.72 17
17
14–23
33
15–76
27
13–58
Esthwaite Water
0.88**
Moderate
9
0.49–0.75 24
24
11–63
37
17–82
23
11–47
Felbrigg Lake
0.48*
Slight
7
0.43–0.72 76
79
14–116
66
30–147
57
29–113
Groby Pool
1.32*** Major
8
0.58–0.75 145
149
7–452
67
30–149
63
30–132
Piburger See
1.12*** Major
0 [0.75
Na
Na
Na
13
6–30
14
7–28
0.12–0.75 18
18
2–98
18
8–40
14
7–26
Lago 0.52* Maggiore
Slight
N
44
SCD range AM-TP (simple)
range), inferred total phosphorus based on the analogue matching using simple (AM-TP simple) and weighted (AMTP weighed) means, the range of TP values for the analogue sites (AM-TP range), inferred total phosphorus based on weighted averaging (WA-TP) and locally-weighted weighted averaging (LWWA-TP) and the inferred values ±RMSEP for both models
0.63–0.74 18
All inferred TP values are expressed in lg L-1
123
J Paleolimnol (2011) 45:415–431
ocellata, the latter a species associated with relatively unimpacted alkaline systems in Northern Ireland and the Shropshire Meres (Anderson and Rippey unpublished), as well as numerous non-planktonic taxa. Whilst Aulacoseira subarctica is also abundant in the surface sample, several taxa indicative of more productive waters, namely Stephanodiscus parvus, Asterionella formosa, Aulacoseira islandica and Aulacoseira ambigua increase or are observed only in the surface sample (Fig. 2). The SCD score of 0.80 indicates moderate floristic change (Table 3). The reference sample of Esthwaite Water is comprised of a relatively diverse assemblage with a range of planktonic taxa generally associated with waters of low to moderate nutrient concentrations (e.g. Aulacoseira subarctica, Cyclotella comensis and Cyclotella radiosa) and non-planktonic taxa (e.g. Achnanthes minutissima) (Fig. 2). In contrast the surface sample is dominated by species more commonly associated with nutrient-rich waters, namely Asterionella formosa, Aulacoseira granulata, Fragilaria crotonensis and Stephanodiscus spp. The SCD score of 0.88 indicates moderate floristic change (Table 3). The reference diatom assemblage of Felbrigg Lake is relatively undiverse with only 19 taxa and is dominated by non-planktonic Fragilaria taxa (F. construens, F. construens var. venter, F. construens var. binodis, F. brevistriata, F. pinnata, and F. elliptica) (Fig. 2). Fragilaria spp. dominated assemblages are typical of alkaline, shallow lakes of intermediate nutrient status. The surface sediment sample is broadly comprised of the same taxa, the only notable difference being the appearance of the planktonic diatom Stephanodiscus parvus albeit in small amounts. The SCD score of 0.48 reflects this slight floristic change (Table 3). The reference diatom flora of Groby Pool is similar to that of Felbrigg Lake. It too is relatively undiverse with only 19 taxa and is dominated by non-planktonic Fragilaria spp. (F. construens var. venter, F. construens var. binodis, F. brevistriata, F. pinnata) (Fig. 2). Cyclostephanos dubius, a centric planktonic diatom and Aulacoseira granulata var. angustissima, another planktonic form, are also present. These latter taxa are commonly associated with alkaline, productive waters. In contrast to Felbrigg Lake, however, the surface sample of Groby Pool is markedly different from the reference assemblage as Stephanodiscus parvus appears and
421
constitutes *40% of the assemblage thus suppressing the non-planktonic flora. The high SCD score of 1.32 reflects this major floristic change (Table 3). The reference assemblage of Piburger See is diverse and dominated by benthic species characteristic of circumneutral to slightly acid mountain waters, namely Fragilaria spp., several Achnanthes spp., Cymbella spp. and Naviculaceae species, and Fragilaria capucina var. gracilis (Fig. 2). In contrast, the diatom assemblage in the surface sample has low diversity and is dominated by planktonic diatoms, particularly Asterionella formosa, Cyclotella spp. (C. comensis, C. radiosa), and Fragilaria crotonensis, characteristic of circumneutral waters with moderate nutrient levels. The high SCD score reflects this major floristic change from the reference assemblage (Table 3). The reference assemblage of Lago Maggiore is dominated by Cyclotella species, mainly C. comensis (*80%), a species typical of oligotrophic lakes (Fig. 2). Cyclotella comensis is also abundant in the surface sample (*30%) but other taxa indicative of more productive waters become increasingly important, notably Fragilaria crotonensis but also Asterionella formosa, Cyclotella pseudostelligera and Diatoma tenuis. The SCD score of 0.52 reflects this slight floristic change from reference condition (Table 3). The PCA results reveal two clear axes of variation in the species data, with 14.6 and 10.4% 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. Owing to the large number of taxa, the diatom taxon scores are not plotted but the key taxa occurring in each section of the plot are shown in Fig. 3 to facilitate description. The magnitude and direction of floristic change is shown by the arrows between the reference and surface samples and, with the exception of Piburger See, the arrows for all sites point towards the right of the diagram (Fig. 3). This reflects the shift away from the relatively nutrient-poor taxa located on the left of the plot towards taxa typically associated with more productive waters on the right. The shallow lakes (Felbrigg Lake, Loch Davan, Groby Pool) are positioned towards the lower part of the plot associated with largely non-planktonic taxa whilst the deep lakes are generally located in the upper part of the diagram reflecting the greater percentage of planktonic taxa in their assemblages compared with the
123
422
J Paleolimnol (2011) 45:415–431
Fig. 3 PCA plot of the reference and surface samples of the nine study sites set within the context of the EDDI Combined TP dataset. Arrows connect the reference and surface samples for each core. The direction of the arrow indicates the direction of floristic change and the length is a measure of species turnover
shallow waters. The arrow for Piburger See points towards the top of the figure reflecting a shift from a predominantly non-planktonic flora to one comprised of planktonic taxa. Axis 1 appears, therefore, to represent a productivity gradient and axis 2 represents shifts in diatom life form. Analogue matching Modern analogues (SCD \ 0.75) in the EDDI Combined TP dataset were found for six of the nine sites (Tables 3, 4; Fig. 4). No analogues were found for Loch Leven, Loch Davan and Piburger See. Mjoesa had only two analogues with SCD scores of 0.63 and 0.74, both located in Northern Ireland (Fig. 4a). One of the analogue lakes was medium alkalinity and the other high alkalinity and both were formed in kettle holes and were hence relatively deep ([10 m), supporting many taxa in common with Mjoesa including Asterionella formosa, oligotrophic Cyclotella taxa, Aulacoseira subarctica, Synedra rumpens and Achnanthes minutissima. White Lough had only
123
three analogues, the closest match being Chiemsee (SCD = 0.49), a deep lake in Germany, and the others being a medium alkalinity, relatively deep lake in Northern Ireland (SCD = 0.68), and a medium alkalinity, deep lake in France (SCD = 0.72) (Fig. 4b). The assemblages of these lakes were matched with White Lough principally due to the high percentages of Aulacoseira subarctica. Esthwaite Water had nine analogues with SCD scores ranging from 0.49 to 0.75, located in Northern Ireland (4 sites), Switzerland (2 sites), France, Germany and Austria (Fig. 4c). They covered a wide range of alkalinity values (56–1,420 lg L-1) but all were relatively deep lakes (Table 4) supporting a predominantly plankton dominated assemblage. The main taxa in common with the Esthwaite Water reference sample were Asterionella formosa, Aulacoseira subarctica, Achnanthes minutissima, Cyclotella comensis and Cyclotella radiosa. Felbrigg Lake had seven analogues with SCD scores ranging from 0.43 to 0.72 including sites in England (4 sites), Denmark (2 sites) and Austria
J Paleolimnol (2011) 45:415–431
423
Table 4 Summary characteristics of the analogue sites Lake
Mjoesa
White Lough
Esthwaite Water
Altitude (m)
Conductivity (lS cm-1)
TP (lg L-1)
170
10
7.09
300
93
11
190
13
7.46
1420
166
25
Median
180
12
7.28
860
130
18
Mean
180
12
7.28
860
130
18
Minimum
105
6
6.60
211
35
14
Maximum
1003
109
8.30
790
333
23
Median
518
73
7.65
501
217
15
Mean
542
63
7.52
501
195
17
Minimum
105
6
6.60
56
35
11
Maximum
1046
109
8.30
1420
494
63
432 495
10 21
7.74 7.69
300 503
217 255
22 24
Minimum
28
1
7.09
550
175
14
Maximum
540
4
8.36
3760
463
116
60
2
7.94
1825
389
89
125
2
7.90
1878
351
76
Minimum
30
1
6.56
446
46
7
Maximum
1200
12
8.13
4190
682
452
Median
258
4
7.85
2110
421
72
Mean
145
Mean
Lago Maggiore
Alkalinity (leq L-1)
Maximum
Median Groby Pool
pH (pH units)
Minimum
Median Mean Felbrigg Lake
Max depth (m)
445
4
7.64
2127
377
Minimum
65
6
7.60
29
106
2
Maximum
2339
370
8.65
2620
535
98
Median
551
19
8.20
56
302
13
Mean
669
49
8.15
293
293
18
(Fig. 4d). These were essentially high alkalinity, shallow lakes (\4 m) with assemblages dominated by non-planktonic Fragilaria spp. Groby Pool had eight analogues with SCD scores ranging from 0.58 to 0.75 located in England (4 sites), France (2 sites) and Austria (2 sites) (Fig. 4e). These were lakes with medium to high alkalinity values and were all shallow (\4 m) with the exception of one site (Table 4). Similarly to the analogues for Felbrigg Lake, the assemblages of the Groby Pool analogues (SCD scores 0.58–0.75) were comprised of nonplanktonic Fragilaria taxa but here lakes where Fragilaria construens var. venter was particularly abundant provided the closest matches. Lago Maggiore had by far the largest number of analogues with 44 matches located mostly in pre-Alpine regions of Austria, Switzerland, Italy and Germany (Fig. 4f). The SCD scores ranged from 0.12 to 0.75 with 20
sites having a very close match (SCD \ 0.45). The analogue lakes were of low to medium alkalinity and were all deep (Table 4). The assemblages of these lakes were matched with Lago Maggiore principally due to the high percentages of Cyclotella comensis. As expected, the analogue sites are located in close proximity to the reference sample in ordination space (Fig. 5) with the exception of Esthwaite Water where the nine analogues are widely dispersed in this low dimensional representation of the data. Diatom-inferred total phosphorus reference conditions Each of the methods employed to infer reference TP concentrations for the study sites produced slightly different values (Table 3). The transfer function based on the EDDI Combined TP dataset with simple
123
424
J Paleolimnol (2011) 45:415–431
Fig. 4 Maps showing the location of the analogue sites for a Mjoesa, b White Lough, c Esthwaite Water, d Felbrigg Lake, e Groby Pool, and f Lago Maggiore. The study site is indicated
by a cross and the analogue sites by filled circles. The number in parentheses indicates the number of analogues found
WA generally produced slightly higher DI-TP values than the LWWA model, with the former generating higher concentrations at seven sites and with both models producing the same values at two sites. However, the differences were always less than the RMSEP of the transfer functions. Both analogue
matching methods (simple mean and weighted mean of TP measured for the analogue sites) produced very similar results (Table 3). Indeed identical concentrations were inferred using the two methods at four of the six sites where analogues were found. When the transfer function and analogue matching outputs were
123
J Paleolimnol (2011) 45:415–431
425
Fig. 5 PCA plot as in Fig. 3 showing the location in ordination space of the analogue sites for a Mjoesa, b White Lough, c Esthwaite Water, d Felbrigg Lake, e Groby Pool, and f Lago Maggiore
compared, the approaches produced similar reference TP values for three lakes (Mjoesa, Esthwaite Water and Lago Maggiore). For White Lough, the analogue matching method produced lower inferred concentrations than the transfer functions and for Felbrigg Lake and Groby Pool the opposite was true. In the latter case, the analogue matching inferred values were more than double those produced by the transfer functions and fell outside the range of values for the LWWA model (Table 3). In the case of the deep lakes, Mjoesa, White Lough, Esthwaite Water and Lago Maggiore, the TP concentrations of each of the analogue lakes were all within a small range (Fig. 6). However, for the two shallow lakes, Felbrigg Lake and particularly Groby Pool, the TP values of the analogue lakes covered a broad range with ranges of 14–116 and 7–452 lg L-1, respectively (Table 3).
Fig. 6 The distribution of TP values in the analogue sites for each study site where analogues were found
123
426
Discussion Reference diatom assemblages and floristic change The reference diatom communities of the nine study sites are site-specific which is not surprising given that the lakes are geographically widespread and cover a broad range of lake types. Nevertheless there are some common features with oligotrophic, acidophilous to circumneutral taxa such as Achnanthes minutissima and the planktonic taxa Cyclotella comensis and Aulacoseira subarctica being predominant in the reference samples of several of the deep lakes. These same taxa were observed in the reference assemblages of deep oligotrophic lakes in Scotland (Bennion et al. 2004) and were associated with deep lakes of low to medium alkalinity in a study of reference samples from 106 lakes in the UK (Bennion and Simpson 2010). The reference samples of the three shallow lakes are all dominated by nonplanktonic taxa. However, whilst the medium alkalinity site, Loch Davan, had a very diverse flora, the high alkalinity lakes, Felbrigg Lake and Groby Pool, were both dominated by benthic Fragilaria spp. High relative abundances of these taxa are common in the assemblages of temperate, alkaline, shallow waters where a favourable light climate supports growth on the surface sediments or attached to plant surfaces (Bennion 1995) although they can also live in the tychoplankton of turbid, shallow lakes. Therefore, the reference diatom communities of our study lakes share some similarities with those observed in lakes of a similar type elsewhere, with factors such as alkalinity, productivity and water depth being important drivers of species composition. Of the nine study sites, two exhibited slight change (Lago Maggiore, Felbrigg Lake), five experienced moderate change (Mjoesa, Loch Davan, Loch Leven, White Lough, Esthwaite Water), and two showed evidence of major change (Groby Pool, Piburger See) in their diatom assemblages from reference condition. In all cases the shifts were caused by eutrophication over the last century. Whilst the diatom response to enrichment was site-specific there were several taxa that frequently appeared or increased in the surface sample relative to the reference sample, namely Asterionella formosa, Fragilaria crotonensis, Stephanodiscus parvus and Stephanodiscus hantzschii.
123
J Paleolimnol (2011) 45:415–431
Indeed one of either of the two latter taxa exhibited increases or were new arrivals in seven of the nine sites. These Stephanodiscus spp. have been observed in the upper sections of numerous lake sediment core sequences and are known to track eutrophication (e.g. Bradbury 1975; Battarbee 1978; Anderson 1997; Bennion et al. 2000). When interpreting the degree of floristic change between the reference and surface sample, it is important to note that for several lakes the most eutrophic phase has passed and the diatom assemblages have experienced some recovery following reduction of nutrient loads in recent decades. For example, at Mjoesa the P load to the lake was reduced in the mid-1970s and diatom analysis of a sediment core from the lake reveals that the diatom flora has shifted from one dominated by Fragilaria crotonensis and Stephanodiscus spp. at the peak of enrichment in the 1950s and 1960s to a more even assemblage comprised of taxa associated with oligotrophic and mesotrophic conditions (Clarke and Bennion unpublished). Similarly at Loch Leven, the diatom communities have responded in the last decade to the reduced P loading and are presently less dissimilar to the reference assemblages than they were at the peak of enrichment in the mid-1970s (Bennion et al. 2001b). Also at Lago Maggiore, over the last 15–20 years the eutrophic species such as Stephanodiscus spp., which appeared during the eutrophic phase from 1960 to 1980, have been replaced by taxa associated with lower trophic status, namely Asterionella formosa, Fragilaria crotonensis and Cyclotella comensis (Marchetto et al. 2004). Importantly, however, whilst many of the taxa observed in the reference samples of these lakes have increased again in the surface sediments in response to nutrient reduction, their relative abundance has not yet returned to values seen in the reference assemblages, and hence the SCD dissimilarity scores are still relatively high. Piburger See experienced major floristic change from the reference assemblage with a shift from a diverse community dominated by benthic species characteristic of circumneutral to slightly acid waters to an assemblage of low diversity dominated by planktonic taxa, notably Fragilaria crotonensis and Asterionella formosa, which are considered as indicators of nutrient enrichment in temperate lakes (e.g. Tilman 1982). This appears to be related to a recent
J Paleolimnol (2011) 45:415–431
increase in the supply of silica and nitrogen (Thies et al. in preparation), which are known to be the main factors controlling populations of both of these diatom taxa in alpine lakes (e.g. Saros et al. 2005). Major floristic change also occurred at Groby Pool, a high alkalinity shallow lake. Here the assemblage switched from one dominated by non-planktonic taxa, principally Fragilaria spp., to one with a high relative abundance of the planktonic diatom Stephanodiscus parvus. Similar shifts have been observed in palaeoecological records from numerous enriched shallow lakes (e.g. Battarbee 1978; Schelske et al. 1999; Bennion et al. 2001a). The expansion of planktonic taxa probably reflects an increased prevalence of pelagic relative to benthic primary production as is thought to be typical of shallow lakes experiencing eutrophication (Vadeboncoeur et al. 2003). In contrast, the other high alkalinity shallow lake, Felbrigg Lake, exhibited only minor floristic change with non-planktonic Fragilaria taxa occupying around 90% of both the reference and surface assemblages. Whilst there is a small increase in the representation of centric planktonic taxa (e.g. Stephanodiscus parvus) in the surface sample indicative of enrichment, this is swamped by the dominance of the Fragilaria spp. and hence the SCD score is relatively low. In this case the diatom data suggest minor change even though dynamic eutrophicationinduced changes are inferred from other fossil groups such as plant macrofossils, pollen and Mollusca (Sayer et al. 2010). Caution must therefore be exercised when interpreting ecological change from diatom records of shallow lakes dominated by Fragilaria taxa. The exclusion of such taxa and an increase in the counts of non-Fragilaria taxa and/or a multi-proxy approach is advised in these situations. Analogue matching Modern analogues in the EDDI Combined TP dataset were found for six of the nine sites. In some cases the analogue sites seemed to be appropriate but in others the lake chemistry and physical attributes of the analogues differed markedly from the target lake (compare Tables 1, 4). For example, for Mjoesa the two analogue lakes had considerably higher alkalinity and TP values than those found currently in this lake and were physically rather different being much smaller and shallower. For White Lough, one of the
427
analogue sites was in Northern Ireland and was a lake of similar type but with lower TP concentrations than those of White Lough today (23 compared with 60 lg L-1) and therefore its selection as a potential reference site seems appropriate. However, the other two analogue sites were of much lower alkalinity and greater depth ([70 m) than White Lough with somewhat lower TP values (15 lg L-1) than one might expect in high alkalinity waters in reference condition (Bennion and Simpson 2010). For Esthwaite Water, the majority of the analogues seem appropriate as they are medium alkalinity, deep waters (the same type as Esthwaite) with TP concentrations of \25 lg L-1. One of the Swiss sites seems less suitable, however, as it has a low alkalinity and a high TP concentration of 63 lg L-1, approximately double that of Esthwaite Water today. Lago Maggiore had by far the largest number of analogues (44), almost half of which had a very close match. The analogue lakes were of low to medium alkalinity and were deep (median 19 m) and therefore seem appropriate in terms of their typology. Whilst the TP values of many of these sites were relatively low (median 13 lg TP L-1), some were nevertheless higher than the current TP concentration of Lago Maggiore and in two cases the values exceeded 50 lg L-1 suggesting that these sites may be unsuitable as analogues. For the two shallow, high alkalinity lakes, Felbrigg Lake and Groby Pool, the lakes with the closest matches were lakes of a similar type suggesting that they may provide appropriate analogues. Owing to the intensity of land use in much of lowland Europe it is generally considered difficult to find minimally impacted waters to act as reference sites for shallow lake types that occur in low lying regions (e.g. Bennion et al. 2004), yet several analogue sites with relatively low TP concentrations (i.e. TP \ 40 lg L-1) were found for Groby Pool and Felbrigg Lake. However, several of the analogue sites had very high TP concentrations ([100 lg L-1) which intuitively seem too high to represent pre-enrichment conditions and are considerably higher than the TP values generated by the transfer functions, which were also high. The cosmopolitan nature of many of the diatom taxa found in the reference samples of our lakes and their relatively broad preferences for nutrient concentrations are the most likely explanations for the selection of inappropriate analogue sites. Clearly, a
123
428
similar diatom assemblage or at least an assemblage comprised of the same dominant taxa can occur across a relatively broad range of chemical conditions. This is well known for some diatom species, such as the non-planktonic Fragilaria taxa, as discussed below, but our results suggest that other taxa such as Achnanthes minutissima, Aulacoseira subarctica and Cyclotella comensis can also tolerate a broad range of nutrient conditions. Here, we are limited to the criteria of lake alkalinity, depth and TP concentrations for evaluating the appropriateness of the identified analogues. The two former variables were selected as they are the key criteria used in the Great Britain lake typology scheme and analysis of diatom reference samples from *100 lakes in the UK indicates that alkalinity and, to a lesser extent, lake depth are important in explaining the diatom distributions (Bennion and Simpson 2010).The suitability of analogue sites could be more rigorously assessed in future by exploring the ecological aspects of these sites. For example, Simpson et al. (2005) assessed suitability of 22 modern analogue sites for defining wider hydrochemical and biological reference conditions for acidified lakes in terms of hydrochemistry, aquatic macrophytes and macro-invertebrates. At present only basic chemical and physical data are available for the lakes in the EDDI training set and therefore comparisons with the wider fauna and flora of the analogues was not possible in this study. There is undoubtedly a trade off between employing large training sets to maximise the likelihood of identifying analogues but for which detailed data for evaluation purposes may be lacking and using smaller, local training sets where biological information may be available but which have limited applicability at a larger geographical scale. A further test of suitability could involve comparisons of contemporary macrophyte and zooplankton assemblages in the analogue sites with those recorded by remains of macrofossils and Cladocera in the reference samples of impacted sites. The reference diatom assemblages of Loch Davan, Loch Leven and Piburger See do not have any modern analogues within the EDDI training set. This can be explained by the uniqueness of the diatom assemblages in the samples from these three lakes. Loch Davan has a diverse assemblage (58 taxa) of mostly non-planktonic species with 50 taxa each representing less than 3% of the total assemblage. Similarly, Loch
123
J Paleolimnol (2011) 45:415–431
Leven has a diverse assemblage (59 taxa) comprised of both planktonic and non-planktonic species with 51 taxa each representing less than 3% of the total assemblage. Equally, Piburger See has a diverse (56 taxa) non-planktonic dominated assemblage with 48 taxa each representing less than 3% of the total assemblage. It is difficult to find matches for these highly diverse assemblages even though the EDDI Combined TP dataset has good representation of both non-planktonic and planktonic forms. Clearly a training set with a greater number and range of reference lakes is required to improve the diatom analogues. The EDDI training set used here contains relatively few unimpacted lakes as the dataset was not designed specifically for analogue matching purposes but rather for development of TP transfer functions that require sites to be distributed along wide environmental gradients. Consequently many of the lakes in the training set are productive and the mean TP of the dataset (99 lg L-1) is considerably higher than the current measured TP of six of the study sites and one assumes is higher still than the pre-enrichment reference conditions. Analogue matching has proved to be a useful method for identifying appropriate reference sites for lakes impacted by acidification (Flower et al. 1997; Simpson et al. 2005) but nevertheless in an application of the technique to 10 lakes in the UK, close analogues could not be found for two of the study sites as that lake type was underrepresented in the training set (Simpson et al. 2005). It was not our intention here to address critical aspects of the analogue matching approach itself. It could be argued, for example, that the choice of cutoffs to define analogue and non-analogue sites is ad hoc, and that different results could have been obtained if different percentiles of the distribution of observed dissimilarities in the training set had been used. As stated above, there is no statistical theory to guide selection of a particular dissimilarity value that best discriminates between similar and dissimilar sites. For this study we drew on prior knowledge from sites similar to those studied here and earlier research to guide our selection of the cut-offs (e.g. Bennion et al. 2003, 2004). Monte Carlo resampling of the observed dissimilarities has been suggested as one way of deriving a significance value for a selected cut-off or for choosing a cut-off at a certain level of Monte Carlo significance, for example the dissimilarity associated with the 0.05 level (Sawada
J Paleolimnol (2011) 45:415–431
et al. 2004). We applied this approach to the observed dissimilarities in the EDDI training set, but the results were unsatisfactory, providing unrealistically large cut-offs at high levels of Monte Carlo significance. This resulted from the strong left skew in the observed distribution of dissimilarity values, which was exacerbated following resampling. Receiver operator characteristic (ROC) curves (Gavin et al. 2003; Wahl 2004) are a recent development that provide a more rigorous statistical basis for selecting cut-offs, but this technique is only applicable when the training set can be a priori partitioned into types of sites on the basis of an external typology (i.e. one not derived from the species data themselves). A key issue is how to relate taxonomic and species abundance differences between samples to the computed dissimilarity; how do differences between samples in one or a few species, or differences across all species manifest themselves in the computed dissimilarities of such samples? Clearly much further theoretical work is required to address these important issues. Simulation studies where the magnitude of difference between samples can be controlled may be one way of deciphering the meaning of the computed dissimilarity and, potentially, also provide a means of deriving study specific cut-off values. Another avenue of progress may be to employ statistical distributions, such as the multinomial or the Dirichlet, appropriately parametrized using optima, tolerance and covariances observed from the training set, from which sampling may be performed to derive a reference distribution for determining cut-offs. Overpeck et al. (1985) observe, in an appendix to their study, that there are direct relationships between certain dissimilarity coefficients and the multinomial distribution, and other coefficients, such as the Hellinger distance, are related to statistical measures of the divergence of one probability function from another (e.g. the Kullback–Leibler distance or divergence). Drawing on the theoretical underpinnings of certain dissimilarity coefficients, therefore, may offer a firmer statistical basis for choosing cut-off values. Diatom-inferred total phosphorus reference conditions Both analogue matching methods (simple mean and weighted mean) produced very similar results
429
suggesting that there is little effect of weighting the mean according to the degree of dissimilarity. Diatom-inferred TP concentrations are traditionally derived using WA transfer functions (e.g. Hall and Smol 1999) and therefore a comparison was made of the values inferred using the analogue matching method with the well established transfer function approach. For three lakes, Mjoesa, Esthwaite Water and Lago Maggiore, the approaches produced similar reference TP values. For the two former lakes, the analogue matching values matched most closely with the DI-TP concentrations based on the LWWA model whereas for Lago Maggiore a closer match was made with the WA model. This is most likely due to the fact that only a small number of analogues were found for the former lakes whilst Lago Maggiore had 44 analogues in the training set. For White Lough, the analogue matching method produced slightly lower inferred concentrations than the transfer functions but this was based on only three analogues so it is difficult to draw any firm conclusions. For these four deep lakes, the TP concentrations of each of the analogue lakes were all within a small range suggesting that the inferred mean values are likely to be reliable. Nevertheless, in the case of Mjoesa and Lago Maggiore the DI-TP reference values are higher than the current measured TP suggesting that the models over-estimate concentrations of these two sites. This over-estimation occurs because there are too few large, deep, oligotrophic lakes in the EDDI training set and we recommend inclusion of more lakes of this type to enhance the performance of the method. For the shallow lakes, Felbrigg Lake and Groby Pool, the converse was true in that the values inferred from analogue matching were higher than those derived from the transfer functions. For Felbrigg Lake the analogue matching values were only slightly higher but for Groby Pool they were more than double those produced by the transfer functions. Notably, the differences between the concentrations inferred by the analogue matching method and the transfer functions were less than the RMSEP of the latter in all except Groby Pool (Table 3). In contrast to the deep lakes, the TP values of the analogues for these two shallow sites covered a broad range of 10 to [100 lg L-1. This is largely due to the abundance of Fragilaria taxa in the reference samples of these sites as these taxa are cosmopolitan and are widely
123
430
distributed across the TP gradient (Bennion et al. 2001a) rendering them poor indicators of lake TP concentrations. Transfer functions commonly overestimate TP values in high alkalinity, shallow lakes where these taxa dominate the assemblages (Bennion et al. 2001a; Sayer 2001) and it appears that the analogue matching technique may be equally susceptible to this problem. Nevertheless, where sufficient analogues are identified, a robust measure of central tendency (such as the trimmed mean or median) or the modal inferred value may be used instead of the mean to reduce this effect.
Concluding remarks The study demonstrates the valuable role that palaeoecological techniques can play in determining both chemical and ecological reference conditions and degree of floristic change for a range of lake types impacted by eutrophication. The findings suggest that the analogue matching technique has the potential to be a useful tool for identifying appropriate reference sites for enriched lakes but highlights that the method is limited by its reliance on the existence of unimpacted sites. Improvements to the training set employed in our study are evidently required before it can be applied to all lake types, the first step being the inclusion of a greater number of reference sites and lakes with lower TP concentrations, particularly for shallow lakes. Indeed, new initiatives that bring data together from the wider palaeolimnological community, such as the recently compiled LakeCores palaeolimnological meta-database (Battarbee et al. 2010) that includes information for 975 lakes in Europe, may be required to produce large training sets appropriate for analogue matching at a broad range of lake types. Analogue matching based on a combination of biological groups, rather than on diatoms alone, could be explored in future for identifying reference sites for lakes impacted by eutrophication. Acknowledgments This paper was written with support from the European Union (FP6 Integrated Project ‘Euro-limpacs: European project to evaluate impacts of global change on freshwater ecosystems’ GOCE-CT-2003-505540) and the Scotland and Northern Ireland Forum for Environmental Research (SNIFFER, project number WFD08). Thank you to two anonymous reviewers for their valuable comments.
123
J Paleolimnol (2011) 45:415–431
References Anderson NJ (1997) Reconstructing historical phosphorus concentrations in rural lakes using diatom models. In: Tunney H, Carton OT, Brookes PC, Johnstone AE (eds) Phosphorus losses from soil to water. Cab International, Wallingford, pp 95–118 Anonymous (2003) Common implementation strategy for the water framework directive (2000/60/EC). Guidance on establishing reference conditions and ecological status class boundaries for inland surface waters. Available at: http://forum.europa.eu.int/Public/irc/env/wfd/library Battarbee RW (1978) Observations on the recent history of Lough Neagh and its drainage basin. Philos Trans R Soc Lond 281:303–345 Battarbee RW, Jones VJ, Flower RJ, Cameron NG, Bennion H, Carvalho L, Juggins S (2001a) Diatoms. In: Smol JP, Birks HJB, Last WM (eds) Tracking environmental change using lake sediments. vol 3: terrestrial, algal, and siliceous indicators. Kluwer, Dordrecht, pp 155–202 Battarbee RW, Juggins S, Gasse F, Anderson NJ, Bennion H, Cameron NG, Ryves DB, Pailles C, Chalie F, Telford R (2001b) European Diatom Database (EDDI). An information system for palaeoenvironmental reconstruction. ECRC Research Report No 81, University College London, 94 pp Battarbee RW, Monteith DT, Juggins S, Evans CD, Jenkins A, Simpson GL (2005) Reconstructing pre-acidification pH for an acidified Scottish loch: a comparison of palaeolimnological and modelling approaches. Environ Pollut 137:135–149 Battarbee RW, Morley D, Bennion H, Simpson GL, Hughes M, Bauere VA (2010) A palaeolimnological meta-database for assessing the ecological status of lakes. J Paleolimnol (this issue). doi:10.1007/s10933-010-9417-5 Bennion H (1995) Surface sediment diatom assemblages in shallow, artificial, enriched ponds, and implications for reconstructing trophic status. Diatom Res 10:1–19 Bennion H, Battarbee R (2007) The European Union Water Framework Directive: opportunities for palaeolimnology. J Paleolimnol 38:285–295 Bennion H, Simpson GL (2010) The use of diatom records to establish reference conditions for UK lakes subject to eutrophication. J Paleolimnol (this issue). doi:10.1007/ s10933-010-9422-8 Bennion H, Monteith D, Appleby P (2000) Temporal and geographical variation in lake trophic status in the English Lake District: evidence from (sub)fossil diatoms and aquatic macrophytes. Freshwat Biol 45:394–412 Bennion H, Appleby PG, Phillips GL (2001a) Reconstructing nutrient histories in the Norfolk Broads, UK: implications for the role of diatom-total phosphorus transfer functions in shallow lake management. J Paleolimnol 26:181–204 Bennion H, Fluin J, Appleby P, Ferrier B (2001b) Palaeolimnological investigation of Scottish freshwater lochs. Final report to SNIFFER No. SR(00)02 by ENSIS Ltd, University College London, 424 pp Bennion H, Simpson G, Hughes M, Phillips G, Fozzard I (2003) The role of palaeolimnology in identifying reference conditions and assessing ecological status of lakes.
J Paleolimnol (2011) 45:415–431 In: Ruoppa M, Heinonen P, Pilke A, Rekolainen S, Toivo H, Vuoristo H (eds) How to assess and monitor ecological quality in freshwaters. Tema Nord 547. Nordic Council of Ministers, Copenhagen, pp 57–63 Bennion H, Fluin J, Simpson GL (2004) Assessing eutrophication and reference conditions for Scottish freshwater lochs using subfossil diatoms. J Appl Ecol 41:124–138 Bradbury JP (1975) Diatom stratigraphy and human settlement in Minnesota. Geol Soc Am Special Paper 171:1–74 European Union (2000) Directive 2000/60/EC of the European Parliament and of the Council of 23 October 20000 establishing a framework for Community action in the field of water policy. Off J Eur Communities L327:1–72 Faith DP, Minchin PR, Belbin L (1987) Compositional dissimilarity as a robust measure of ecological distance. Plant Ecol 69:57–68 Flower RJ, Juggins SJ, Battarbee RW (1997) Matching diatom assemblages in lake sediment cores and modern surface sediment samples: the implications for conservation and restoration with special reference to acidified systems. Hydrobiologia 344:27–40 Gavin DG, Oswald WW, Wahl ER, Williams JW (2003) A statistical approach to evaluating distance metrics and analog assignments for pollen records. Quatern Res 60:356–367 Hall RI, Smol JP (1999) Diatoms as indicators of lake eutrophication. In: Stoermer EF, Smol JP (eds) The diatoms: applications for the environmental and earth sciences. Cambridge University Press, Cambridge, pp 128–168 Juggins S (2001) The European diatom database user guide version 1.0. University of Newcastle, Newcastle upon Tyne 72 pp Marchetto A, Lami A, Musazzi S, Masaferro J, Langone L, Guilizzoni P (2004) Lake Maggiore (N. Italy) trophic history: fossil diatoms, plant pigments, chironomids and comparison with long-term limnological data. Quatern Int 113:97–110 Overpeck JT, Webb T, Prentice IC (1985) Quantitative interpretation of fossil pollen spectra—dissimilarity coefficients and the method of modern analogs. Quatern Res 23:87–108 Pollard P, Huxham M (1998) The European Water Framework Directive: a new era in the management of aquatic ecosystem health? Aquat Conserv Mar: Freshwat Ecosyst 8:773–792 Prentice IC (1980) Multidimensional scaling as a research tool in Quaternary palynology: a review of theory and methods. Rev Palaeobot Palynol 31:71–104 R Development Core Team (2008) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-070, URL http://www.R-project.org
431 Round FE, Crawford RM, Mann DG (1990) The diatoms: biology and morphology of the genera. Cambridge University Press, Cambridge, p 747 Saros JE, Michel TJ, Interlandi SJ, Wolfe AP (2005) Resource requirements of Asterionella formosa and Fragilaria crotonensis in oligotrophic alpine lakes: implications for recent phytoplankton community reorganizations. Can J Fish Aquat Sci 62:1681–1689 Sawada M, Via AE, Vettoretti G, Peltier WR, Gajewski K (2004) Comparison of North-American pollen-based temperature and global lake-status with CCCma AGCM2 output at 6 ka. Quatern Sci Rev 23:225–244 Sayer CD (2001) Problems with the application of diatom-total phosphorus transfer functions: examples from a shallow English lake. Freshwat Biol 46:743–757 Sayer CD, Burgess A, Kari K, Davidson TA, Peglar S, Yang H, Rose N (2010) Long-term dynamics of submerged macrophytes and algae in a small and shallow, eutrophic lake: implications for the stability of macrophyte-dominance. Freshwat Biol 55:565–583 Sayer CD, Roberts N (2001) Establishing realistic restoration targets for nutrient-enriched shallow lakes: linking diatom ecology and palaeoecology at the Attenborough Ponds, UK. Hydrobiologia 448:117–142 Schelske CL, Donar CM, Stoermer EF (1999) A test of paleolimnologic proxies for the planktonic/benthic ratio of microfossil diatoms in Lake Apopka. In: Mayama S, Idei M, Koizumi I (eds) Proceedings of the 14th international diatom symposium. Koeltz Scientific Books, Koenigstein, pp 367–382 Simpson GL (2007) Analogue methods in palaeoecology: using the analogue package. J Stat Softw 22:1–29 Simpson GL, Oksanen J (2009) analogue: analogue matching and Modern Analogue Technique transfer function models (R package version 0.6-22 ). http://cran.r-project.org/ package=analogue Simpson GL, Shilland EM, Winterbottom JM, Keay J (2005) Defining reference conditions for acidified waters using a modern analogue approach. Environ Pollut 137:119–133 Tilman D (1982) Resource competition and community structure. Monographs in Population Biology. Princeton University Press, Princeton, p 296 Vadeboncoeur Y, Jeppesen E, Vander Zanden MJ, Schierup H, Christoffersen K, Lodge DM (2003) From Greenland to green lakes: cultural eutrophication and the loss of benthic pathways in lakes. Limnol Oceanog 48:1118–1408 Wahl ER (2004) A general framework for determining cutoff values to select pollen analogs with dissimilarity metrics in the modern analog technique. Rev Palaeobot Palynol 128:263–280
123