Introduction

Eutrophication of surface waters and degradation of aquatic ecosystems are major environmental concerns at local and global scales. The structure and function of aquatic ecosystems have been recognized as key characteristics for assessment of surface water quality (European Commission 2000). Lake sediments can provide continuous records of change in biological communities and have been used to reconstruct past variations in aquatic ecosystem quality and structure (Bennion et al. 2004a; Lepistö et al. 2006). To date, palaeolimnological assessment of past lake trophic state, including the application of transfer functions, has mainly utilised diatoms (Bennion et al. 2004b; Leira et al. 2006; Taylor et al. 2006). Reliance on a single taxonomic group can lead to erroneous ecological inferences, as a result of inconsistent biological responses to changes in water quality over time (Cameron 1995), and/or differential dispersal and preservation in sediments (Korhola and Rautio 2001; Sayer 2001). In addition, communities from different trophic levels can be shaped by distinct ecological processes in aquatic food webs. Phytoplankton communities may vary in response to nutrient availability (Tilman et al. 1982), whereas herbivores and higher trophic levels are sensitive to both nutrient enrichment and intensified biotic interaction, which can also vary with lake trophic state (Jeppesen et al. 2003). Furthermore, ecological communities in freshwater systems have mainly been examined in terms of their morphological and taxonomic identity. More recently, functional structure, or the distribution of species with different ecological roles, has been recognized as an important characteristic that can provide additional insight into the processes that shape lake communities (Heino 2008).

Invertebrates have been increasingly recognized as important biological elements for assessing the ecological status of lakes (European Commission 2000). Cladocera play an important role in the ecology and water quality of lakes, as they occupy an intermediate trophic level in lake food webs (Dodson and Frey 2001; Lampert and Sommer 2007). Hard-shelled forms of Cladocera, such as Chydoridae and Bosminidae, are well preserved in lake sediments, and their remains can provide reliable estimates of past living assemblages (Davidson et al. 2007; Korhola and Rautio 2001). The taxonomic structure of Cladocera remains preserved in lake sediment cores has been used to track past changes in the environment, including eutrophication (Hofmann 1996; Shumate et al. 2002), acidification (Jeziorski et al. 2008), water level (Korhola et al. 2000) and air temperature (Lotter et al. 1997). In addition, variations in taxonomic structure of cladocerans have been found to show strong response to biotic factors that can vary with lake trophic state, such as fish density (Jeppesen et al. 1996) and submerged macrophytes (Davidson et al. 2007). Despite mounting evidence for strong community responses to nutrient loading, few studies have attempted to infer the roles of different processes from a functional group perspective. Body length and habitat preference are important functional properties for cladocerans, and compositional shifts in functional groups possessing these ecological properties can be used to infer changes in size-selective predation and habitat (Barnett et al. 2007).

To date, most studies examining cladoceran community response to nutrient enrichment have come from either shallow lakes (Jeppesen et al. 1996; de Eyto et al. 2003; Davidson et al. 2007) or sites covering a large spatial scale and/or climatic gradient (Lotter et al. 1997; Bjerring et al. 2009). Shallow lakes are often characterized by abundant macrophytes, weak thermal stratification and high nutrient loading from sediments. Invertebrate communities from these systems can show patterns that are distinct from those of deep lakes (Scheffer 1998). Temperature and related variables often exert the strongest control on cladoceran assemblages over gradients in altitude or latitude (Lotter et al. 1997; de Eyto et al. 2003). In this study, we restricted our study area to the Irish Ecoregion, i.e. the island of Ireland. Over this narrow climate gradient, we examined the pattern of sedimented cladoceran communities in relatively deep lakes along a gradient of increasing TP availability.

We set out to assess the role of nutrient availability in structuring cladoceran communities, with respect to both taxonomic and functional differences. First, we expected to find a strong relationship between the taxonomic structure of the cladoceran community in surface sediments and lake trophic state. Second, to assess the functional impact of nutrient enrichment on community composition, we examined the distribution of cladoceran groups possessing different functional properties, i.e. habitat preference and body size. Third, we evaluated the role of nutrient availability in structuring communities from two trophic levels, i.e. primary producers and consumers, by identifying the degree of similarity in independent responses of sub-fossil assemblages of Cladocera and diatoms from the same set of lakes. Finally, the spatial patterns of cladoceran community variation were evaluated by comparing microfossil remains in uppermost and bottommost sediment samples from short cores taken in six Irish lakes impacted by intensive agriculture.

Materials and methods

Study sites

Thirty-three lakes across Ireland, but mainly in the western part of the island, were selected for palaeolimnological analysis along a TP gradient, ranging from 4.0 to 142.3 μg l−1 (Table 1). Ireland is located between 51° and 55° N latitude and between 5° and 10° W longitude in northwest Europe. The climate of Ireland is dominated by the moderating Gulf Stream and is generally relatively warm for its latitude, with an annual average temperature of around 9°C. Minimum air temperature falls below zero for about 40 days per year in inland areas and <10 days per year in most coastal areas.

Table 1 Characteristics of 33 Irish lakes that yielded surface sediment samples for Cladocera analyses (dash means no data available)
Full size table

Data and sample collection

A total of 17 physico-chemical and land use variables were used in this study (Table 1). Hydrochemical and physical data for the study sites were assembled from Irvine et al. (2001) and Wemaëre (2005). Water samples for chemical analysis were collected mainly during the summer season (May–October) in 1996–2001, with a sampling frequency of 1–10 per year (Table 1). CORINE land cover data includes 41 land cover classes categorized into five groups (urban, forestry, pasture, peatland and agriculture) based on a 25-ha minimum mapping unit (Bossard et al. 2000). Summary statistics of the physico-chemical and land cover data were determined for the 33 study lakes (Table 2). Most of the Cladocera training-set lakes are located in lowland areas, <150 m above sea level. The majority of the lakes are relatively small, with a median surface area of 21.4 ha, and 30 of the 33 lakes have maximum depths >4 m. Most lakes in the Cladocera training set have TP values <100 μg l−1, with a median TP value of 34.7 μg l−1. The lakes generally have pH values >7 (range 7–8.5) and conductivities between 100 and 400 μS cm−1. Pastureland is the main land cover type in the catchment of many of the lakes, reflecting the importance of livestock farming in Irish agriculture.

Table 2 Statistical summaries of 17 environmental variables from 33 Irish lakes
Full size table

A Renberg gravity corer (HTH Teknik, Vårvågen 37, SE-95149 Luleå) was used to collect sediments at or near the deepest point in 28 of the 33 lakes during the summers of 2003 and 2004. Deep-water surface sediments from the other five lakes (Ballycar, Castle, Cullaunyheeda, Dromore and Inchichronan) were sampled with an Ekman grab in 2002. Surface sediments (0–1 cm) were processed for Cladocera remains and for diatom frustules (Chen et al. 2008). No diatom data were available for Lakes Ballycar and Inchichronan, but both diatom and Cladocera data and associated environmental data were available for 31 of the 33 study sites.

Cladoceran analysis

Extraction of Cladocera from sediment samples mainly followed the method of Korhola and Rautio (2001). Approximately 5 g of wet sediment was deflocculated with 50 ml 10% KOH and 5 ml of 10% HCl was then added to remove carbonate. Samples were filtered through a 53-μm mesh sieve. Permanent slides were prepared by gently mixing a 0.05-ml aliquot of the well-shaken concentrate with glycerine-safranin jelly. The taxonomy of the Cladocera remains in this paper followed Alonso (1996), Duigan (1992), Frey (1962) and Goulden and Frey (1963). All Chydoridae and Bosminidae remains were identified to species level except when insufficient taxonomic features could be observed. Daphnia remains were identified to species groups (Daphnia longispina and D. pulex groups), based mainly on the postabdominal claws. Counts of Cladocera remains were adjusted to reflect the fact that individual organisms possess two shells, one headshield and one postabdomen. The most abundant type of remain was used to characterize the numbers represented by each taxon (Frey 1986). At least 100 individuals of Chydoridae were counted, except when Cladocera remains were scarce, in which case >70 individuals were enumerated. Ephippia were either absent or in low abundance in lake sediments and therefore were not included in subsequent data analysis. Cladoceran functional classes were grouped according to both habitat preference and body size (Table 3). Three habitat groups were defined according to Duigan (1992): (1) macrophyte-associated mainly, (2) both macrophyte- and sediment-associated or sediment-associated mainly, and (3) pelagic taxa. Three body size classes were identified: small (<0.5 mm), medium (0.5–1 mm) and large (>1 mm) cladocerans, using the average length of adult females (Alonso 1996).

Table 3 Names, authorities and codes of 31 common Cladocera taxa found in this study, with data summary on number of occurrence (Count), maximum abundance (Max) and Hill’s effective number of occurrence (Hill’s N2), predicted TP optimum and tolerance values (μg l−1) using weighted averaging methods
Full size table

Numerical analysis

Ordination was used to examine the structure of biological assemblages in sediments and their relationship with environmental gradients (ter Braak 1987). The relatively short length (1.83 SD) of the first axis of detrended correspondence analysis (DCA) of Cladocera data supported the use of redundancy analysis (RDA). RDA models variation within biological assemblage data in relation to measured environmental variables, and is based on the assumption that organisms respond linearly to changes in key environmental variables (ter Braak and Šmilauer 2002). Unrestricted Monte Carlo permutation tests were used to test the significance of each variable used in partial RDA and only variables with P < 0.05 under 999 permutations were accepted (Lepš and Šmilauer 2003). In addition, environmental variables with a variation inflation factor (VIF) of >20 were excluded due to their colinearity with other variables (ter Braak and Šmilauer 2002).

The unique and shared effects of environmental factors on species distribution of both Cladocera and diatoms were partitioned and quantified through the variance partitioning procedure (Borcard et al. 1992). Furthermore, the Procrustes randomization test (Jackson 1995) proved to be an effective statistical technique to assess the level of concordance between multivariate datasets and enabled direct comparison of the ecological responses of the Cladocera and diatom assemblages to the measured environmental variables in this study. The Procrustes test involves rotating a data matrix to maximize similarity with a target data matrix by minimizing the sum of squared differences (Jackson 1995). Besides providing a Procrustes correlation derived from the symmetric Procrustes residuals, the test can highlight visual patterns of correspondence between datasets. In the current research, the significance of correlation was assessed by 1000 permutation tests. Canonical correspondence analysis (CCA) was used for diatom data from the same set of 31 lakes, as a long gradient (3.57 SD) in the first DCA axis indicated that most diatom taxa showed unimodal-like responses.

Only Cladocera and diatom taxa with a maximum relative abundance ≥1%, in at least two and three sites, respectively, were used in data analysis. This resulted in a species pool of 31 and 120 taxa, respectively. Transformation of environmental data was employed to reduce the influence of extreme values and to ensure that the data approached a normal distribution (Table 2). The mean of the variable was substituted for the missing value prior to data analysis (Legendre and Legendre 1998). Square root transformation of ecological data was used to stabilize variance (Lepš and Šmilauer 2003). The ordination and Procrustes tests were carried out using the package Vegan (version 1.6–10) in the open source R program (version 2.2.1) (R Development Core Team 2006).

Calibration models for Cladocera were developed and compared using the standard methods of Partial Least Squares (PLS), Weighted Averaging (WA) and Weighted Averaging-Partial Least Squares (WA-PLS) (Birks 1998). Bootstrapping with 1000 permutations was used for cross-validation, as this technique can provide sample-specific prediction errors. To ensure optimum model performance, log transformed (log10(1 + x)) environmental data and square root transformed biological data were used. Optimal models with lowest prediction error (RMSEP) and high boot-strapped coefficient of determination \( \left( {r^{2}_{\text{boot}} } \right) \) were selected and used for reconstruction of TP in this study. Two outlier sites (Lisnahan and Sillan) were identified and therefore excluded from model development as used in Tibby (2004). Calibration models were constructed using the C2 program (version 1.4.2) (Juggins 2003).

Results

Distribution of cladocera in surface sediments of 33 lakes

In total, 38 Cladocera taxa, including eight planktonic taxa, were identified in this study. Table 3 lists the 31 common taxa, which occurred with abundances of ≥1% in at least two sites. The planktonic Daphnia longispina group and Bosmina longirostris were the most common taxa, with mean abundances of 30 and 17%, respectively. Among the 30 littoral taxa recovered, Acroperus harpae, Alona affinis, A. guttata/rectangula group, A. quadrangularis and Alonella nana occurred in all the lakes, with mean relative abundances of between 4 and 8%. This species diversity is comparable with that found in investigations of modern communities in lakes of Ireland (de Eyto and Irvine 2002). Species recorded as rare in contemporary ecological surveys (Duigan 1992) were often under-represented in surface sediments, although some, such as Alona intermedia, Alonella excisa and Graptoleberis testudinaria, were present in surface sediments from a majority of the lakes. A detailed comparison of the sub-fossil sediment assemblages and modern chydorids (de Eyto and Irvine 2002) indicated that the sediments captured a minimum of 66% and an average of 80% of the total number of chydorid species in the water samples from six Irish lakes. In addition, planktonic Bosmina and Daphnia were common in both the water column (Irvine et al. 2001) and surface sediment samples for the same set of Irish lakes.

Quantifying cladocera-environment relationships

Of the twelve environmental variables that were statistically significant at P < 0.05 for marginal effects (Table 4), seven (altitude, catchment area, mean depth, pasture coverage, alkalinity, chlorophyll-a (Chl-a) and TP) were statistically significant at P < 0.001. Variance inflation factors (VIF) for the 12 environmental variables were all <20, indicating that each variable made a unique contribution to the total variance. These 12 variables were therefore included in subsequent analyses. Of the measured environmental variables, Chl-a displayed the largest contribution (8.9%) to the total variance in the Cladocera data, followed by alkalinity and TP, with 8.4 and 8.1%, respectively (Table 4).

Table 4 Summary of partial RDAs constrained by individual environmental variables (ratio of eigenvalue of the first axis (λ1) to the second axis (λ2); significance levels also shown)
Full size table

The relatively short DCA gradient length (1.83 SD) for the 33-lake training set is comparable to other Cladocera datasets (Brodersen et al. 1998; Lotter et al. 1998) and suggests that most taxa responded linearly to underlying ecological gradients. The first four axes of RDA explained 13.5, 11.1, 8.6 and 5.7% of the total variance in the Cladocera data. Trophic state variables Chl-a and TP were strongly related to the first RDA axis (Fig. 1), which captured the maximum variation of the assemblages. Bosmina longirostris, Daphnia longispina group, D. pulex group, Leydigia leydigi and Oxyurella tenuicaudis were positively related with trophic status variables and occurred in high abundances at eutrophic sites such as Ballybeg, Crans and Morgans. Alona intermedia, Alonella excisa and Camptocercus rectirostris were negatively correlated with the nutrient vectors and were abundant at oligotrophic sites, e.g. Anascaul, Beaghcauneen and Tay. Alkalinity, pH and conductivity were strongly negatively correlated with the second RDA axis. Chydorus sphaericus and Pleuroxus trigonellus were positively linked with conductivity and acidity vectors, and had high abundances at Atedaun, Ballyallia and Ballycar. Several species, including Alona rustica, Bosmina longispina, Chydorus piger, Monospilus dispar and Rhynchotalona falcata, were negatively correlated with both nutrient and acidity gradients.

Fig. 1
figure 1

RDA biplots constrained by 12 significant environmental variables (inserted) showing the 33 sites and 31 Cladocera taxa (see Tables 1 and 3 for site and taxon codes, respectively)

Full size image

Variation of cladoceran functional groups from 30 lakes

Results from correlation analysis of the six functional classes (Fig. 2) show that taxa with habitat preference generally display strong responses to nutrient levels in the 30 deeper lakes (maximum depth >4 m). There is a significant decrease in the relative abundance of those taxa that are associated with sediment only or with both macrophyte and sediment, along the TP gradient (r = −0.49, P = 0.006, n = 30; Fig. 2b). A significant increase in the pelagic group was also associated with nutrient enrichment (r = 0.43, P = 0.017, n = 30; Fig. 2c). In contrast, relative abundances for all three size classes showed no significant relationship with nutrient levels (P > 0.1).

Fig. 2
figure 2

Scatterplots showing the relative abundances of six functional groups of cladocerans along the TP gradient in 30 relatively deeper lakes (maximum depth >4 m). Three classes with different habitat preference (macrophytes, both macrophytes and sediments, and pelagic) are shown in plots (a–c) and three body size classes are shown in plots (d–f). Details on the categorization of both functional properties (habitat and body size) are shown in Table 3

Full size image

Comparing cladocera and diatom assemblages from 31 lakes

In total, 206 diatom taxa were identified in the surface sediments of the 31 lakes with cladoceran data. Achnanthidium minutissimum was the most ubiquitous taxon present in 30 lakes, while Asterionella formosa, Cocconeis placentula, Gomphonema parvulum, Puncticulata radiosa, Staurosirella pinnata, and Stephanodiscus parvus were found in >20 lakes. Some taxa showed high abundances, including Staurosira construens var. venter (60.9% at Lisnahan) and Asterionella formosa, Aulacoseira granulata var. angustissima, A. subarctica and Stephanodiscus parvus (range 30–40%).

Figure 3 summarizes the variation in the Cladocera and diatom datasets drawn from the same 31 lakes, which is explained by the measured environmental variables. The 12 significant environmental variables accounted for a majority of the total variance in both datasets (64–65%). Water chemistry variables alone, and in combination with physical and land use variables, were most important in influencing distributions of Cladocera and diatoms, accounting for 46.7 and 43.1% of the total explained variance, respectively. When compared with the influences of other variables in isolation, physical variables accounted for the largest portion of the total explained variance in the Cladocera data (34.4%), while chemical variables alone explained the highest proportion for the diatom data (33.5%).

Fig. 3
figure 3

Partitioning of the variances in the 31-lake Cladocera and diatom data sets explained by 12 significant environmental variables: a partitioning of total variance, b and c total explained variance divided among three groups of variables and by chlorophyll-a and TP, respectively. Areas highlighted in black represent the shared variance explained by the neighbouring variables or groups

Full size image

A high degree of correspondence between the ordinations of Cladocera (RDA) and diatom (CCA) data is evident from the Procustean plot (Fig. 4), which has a significant Procrustes correlation of 0.565 (P < 0.001). More than half of the 31 sites, including Ballybeg, Effernan and Garvillaun, display short distances (i.e. residuals <1) between ordination scores for both datasets. Assemblages from Anascaul, Atedaun, Beaghcauneen and Crans have long distances and the highest Procrustes residuals (>2). Sites with high residuals are often associated with high peat coverage in the catchments, which is not shown here.

Fig. 4
figure 4

Procrustean superimposition plot of surface sediment Cladocera and diatoms in the 31 Irish lakes. Open circles represent the diatom assemblages and the arrows point to the cladoceran assemblages. Distance between symbols is proportional to the Procrustean residual for each site

Full size image

Development of cladocera-TP transfer function

Among all measured environmental variables, only TP and Chl-a produced eigenvalue ratios of the first axis (λ1) to the second axis (λ2) of >0.5 in the RDAs (Table 4), suggesting that TP is appropriate in the construction of a calibration model (Birks 1998). The WA-PLS-2 model, based on 31 lakes, resulted in an optimal performance with \( r^{2}_{\text{boot}} \) of 0.727 and RMSEP of 0.238 (log10 TP μg l−1) (Fig. 5), with better predictability and lower prediction error than the models using WA or PLS modelling methods, which are not shown here. The WA-PLS-2 model also tends to overestimate TP when compared to measured values at the low end of the TP gradient (i.e. <1.0 log10TP; Fig. 5). The WA method was used to estimate the TP optima and tolerance values for 31 common Cladocera taxa (Table 3). Among the 31 common Cladocera taxa, Rhynchotalona falcata and Bosmina longispina have the lowest TP optima of 9.8 and 14.3 μg l−1, respectively, while Oxyurella tenuicaudis and Leydigia leydigi have the highest TP optima of 95.0 and 56.9 μg l−1, respectively. All the other Cladocera taxa have an optimum TP in the range of 17–43 μg l−1. Most Cladocera taxa have TP tolerances <3 μg l−1.

Fig. 5
figure 5

Scatter plots showing the comparison of observed TP with predicted TP and TP residuals of the second component WA-PLS model from 31 lakes. TP data are log10-transformed and Cladocera data are square-root transformed. Boot-strapping was used for cross-validation

Full size image

Top–bottom comparison of cladoceran communities from six lakes

Table 5 summarizes the community changes between the bottom and top samples of sediment cores from six Irish lakes previously studied by Taylor et al. (2006). Across the six lakes, the planktonic group showed increased abundance towards the core top, while the increase was strongest in Egish (∼80%) and moderate in Mullagh and Inchiquin (∼30%), accompanied by decreased abundance of taxa associated with both sediments and macrophytes. Taxa that mainly live on macrophytes showed a unidirectional, although relatively small (<10%) decrease in relative abundance in all lakes. The TP values estimated by diatom (DI-TP; Chen et al. 2008) and cladoceran (CI-TP) assemblages showed the same trend of increasing nutrient levels between the core bottom and top samples for each of the six lakes. The greatest increase in estimated TP, above 70 μg l−1, was inferred using Cladocera from Egish and Crans, where large-sized cladocerans showed the greatest increase in relative abundance (71.4 and 77.6%, respectively) with a concurrent reduction in smallest cladocerans (65.9 and 72.6%, respectively). In contrast, the CI-TPs showed a smaller increase than the DI-TPs (i.e. <10 vs. >15 μg l−1) in the other four lakes.

Table 5 Changes in the relative abundances of six functional habitat and size classes for Cladocera and differences between values of TP estimated from diatom (DI-TP) and Cladocera (CI-TP) assemblages in the bottom and top samples of cores (top minus bottom) from six Irish lakes previously studied by Taylor et al. (2006)
Full size table

Discussion

Multivariate analyses showed that the variation in taxonomic structure of cladoceran remains was most strongly influenced by lake trophic state across the lakes included in this study. Data from the same lakes also revealed a strong similarity in the variation of taxonomic structure of both cladocerans and diatoms in tracking nutrient enrichment. Furthermore, the cladoceran functional structure showed a significant shift from taxa that live on both sediments and macrophytes or sediments only, to pelagic taxa along the gradient of nutrient availability. Similar patterns in community variation were also observed between core bottom and top samples from six lakes. Our palaeoecological evidence confirms the important role of lake trophic state in driving the taxonomic and functional structure of cladoceran communities.

In this study of Irish lakes, trophic state showed the strongest influence on taxonomic structure of cladoceran communities. Relatively high variance in the cladoceran community structure explained by TP has also been found in several other training sets (Brodersen et al. 1998; Lotter et al. 1998). The connection between lake trophic state and Cladocera community structure is most likely linked through nutrient assimilation from algae and detritus (Dodson and Frey 2001). The significant relationship between TP and Chl-a (r = 0.75, P < 0.001, n = 33) verified the role of nutrient availability in limiting primary production in our training set lakes. Earlier experimental work by de Eyto and Irvine (2001) showed that cladocerans such as Chydorus sphaericus grew quickly when algae were the only food source. In the current work, Chl-a had the greatest influence on the composition of Cladocera assemblages when compared with other independent variables. This indicates the importance of food quantity for cladoceran growth and reproduction (Lampert and Sommer 2007).

Data generated in the current study highlight the significance of resource availability in influencing the taxonomic structure of both primary producers and consumers. For example, trophic state variables played a similar role in shaping the taxonomic structure of both Cladocera and diatom assemblages (Fig. 3). This is in agreement with the results from modern ecological studies in the region (DeNicola et al. 2004; Duigan 1992; Irvine et al. 2001). These findings suggest that nutrient transfer within the lake food web strongly structures the communities of both diatoms and cladocerans in our study lakes. This occurs despite the fact that diatoms showed a higher level of species turnover (i.e. DCA axis 1 gradient length) than Cladocera in this study, most likely because of a much higher species diversity of diatoms compared to Cladocera (149 versus 31 taxa >1% in at least two sites of the 31 lakes in common). Over the ordination spaces, both communities also showed a strong similarity in configuration from the same set of lakes (Fig. 4) and this degree of conformity is comparable with that found between the modern and surface sediment zooplankton from 39 lakes (Davidson et al. 2007). This suggests that environmental controls of species composition are similar for diatom and Cladocera assemblages.

The transfer of nutrients from primary producers, including diatoms, to consumers cannot account for all the patterns in the variation of cladoceran communities, particularly functional groups. Here, lake trophic state was also found to affect the cladoceran community through changes in habitat quality (Fig. 2). Specifically there was a significant increase in planktonic abundance, but a decrease in taxa generally associated with two other habitats (macrophytes and sediments), with nutrient enrichment. Pelagic taxa like daphniids are known to out-compete most other taxa in nutrient-enriched environments (Lampert and Sommer 2007). In addition, changes in lake nutrient state have been found to affect the extent and structure of aquatic macrophytes directly (Thoms et al. 1999). However, only a limited response from mainly macrophyte-dwelling taxa was evident in this study, possibly due to the relatively low abundance (<15%) of this functional class compared with other types of taxa. In addition, chydorids may not respond to changes associated with increasing productivity until macrophyte habitats have been substantially affected (Hofmann 1987). For example, a large-scale investigation of lakes in Scotland only found a weak correlation between the littoral microcrustacea, dominated by chydorids, and macrophyte cover (Duigan and Kovach 1994). Recent progress using other sedimentary indicators to infer macrophyte cover may assist efforts to define past habitat changes (Vermaire and Gregory-Eaves 2008) and their role in structuring cladoceran communities.

Palaeoecological data in this study showed no simple relationship between size-selective predation pressure and lake trophic state across lakes. For example, we found no significant change in relative abundance of each of the three size classes with nutrient enrichment (Fig. 2). A large spatial-scale study by Jeppesen et al. (2003) showed a unimodal-like response model of % Daphnia, reflecting predation pressure along a TP gradient up to >800 μg l−1 across several eco-regions, and that the effect of fish predation was greater in shallow lakes. In contrast, the current study involved a much shorter TP gradient (<150 μg l−1), while the majority of the lakes included in the research are relatively deep. However, size-selective predation seems to have affected temporal variation in the cladoceran community from lakes in this study (Table 5). For example, among the six lakes from which sediment core top and bottom samples were compared, Crans showed the largest increase in CI-TP and DI-TP, though there was only a very moderate shift between habitat classes. This lake, however, recorded the largest shift from small-bodied to large-bodied cladoceran taxa across the six lakes observed, indicating decreased predation pressure through time.

Functional variation was found to provide additional insight into the processes that shape the community in this study. This may reflect the fact that functional properties of cladocerans are often independent of their taxonomic identity (Barnett et al. 2007; Heino 2008). This is evident in the sediment core top–bottom comparisons of community structure (Table 5). In Inchiquin and Mullagh, where a small increase in TP (<5 μg l−1) was inferred by cladoceran taxonomic structure, we observed a strong increase (>30%) in the relative abundance of planktonic taxa, with a decrease of 25–30% in sediment-/macrophyte-associated taxa. Similarly, a large deviation between DI-TP and CI-TP values was observed in Egish where there was a major shift in functional structure (Table 5). One advantage of exploring functional patterns in palaeolimnological studies is that they can be less sensitive to inconsistencies of taxonomic resolution, which can be problematic when dealing with cladoceran remains. Sedimentary records of planktonic taxa like daphniids are often characterized by lower taxonomic resolution compared to littoral cladocerans (Korhola and Rautio 2001). This may lead to greater variation in cladoceran taxonomic structure being explained by cladocerans with better taxonomic resolution and hence higher species diversity.

The model developed to determine Cladocera-inferred TP in this study showed a performance comparable with those from other eco-regions with relatively low prediction errors. A training set with 32 Danish lakes along a large TP gradient (16–765 μg l−1) has slightly stronger predictability \( \left( {r^{2}_{\text{boot}} = 0.79} \right) \) and similar prediction error (RMSEP = 0.24) (Brodersen et al. 1998) compared to the current study. High λ1 to λ2 ratios, i.e. >1, are generally required for development of predictive models. However, the cut-off point of >0.5 applied in this study compares favourably with those adopted in other studies, e.g. 0.42–0.44 using diatoms (Hall and Smol 1992) and 0.48 using Cladocera (Amsinck et al. 2005).

Although trophic state variables accounted for the largest part of the variance in cladoceran communities, other environmental factors were also found to be significant in structuring the distribution of this group of invertebrates. Some cladocerans were found to be sensitive to alkalinity; taxa like Alona rustica and Chydorus piger had strong preference for low alkalinity. This indicates the importance of alkalinity in influencing the distribution of Cladocera (Whiteside 1970), as well as other correlated variables such as acidity (Nilssen and Sandoy 1990) and conductivity (Bjerring et al. 2009). In the current study, Cladocera from lakes with peat catchments showed strong dissimilarity in ecological responses when compared with diatoms, as shown in the Procrustes test (Fig. 4). This could be due to their more indirect response to acidic waters, via accompanying changes in predation or macrophyte habitat (Korhola and Rautio 2001), as opposed to a more direct physiological response by diatoms (Battarbee et al. 2001). Such a link may also confound the relationship between lake trophic state and body size classes for cladocerans (Fig. 2d–e), but this remains to be verified in future studies. Cladocera are also known to vary along the gradient of climatic factors associated with altitude. In the Irish Ecoregion, pasturelands that contribute high nutrient loads to lakes are mainly located at lower altitudes, whereas more acidic peatland catchments associated with granitic and siliceous bedrock are often located in the highlands. In the current study, with a limited climate gradient, the extent of pastureland was found to co-vary with altitude (Fig. 1) and the significance of altitude as a predictor of the composition of Cladocera assemblages is likely related to land use.

In conclusion, our results indicate that Cladocera assemblages preserved in lake sediments were reliable and sensitive indicators of lake trophic state. Trophic state variables accounted for a large portion of the total variance in the taxonomic structure of cladocerans. The ecological responses of Cladocera and diatoms, represented by their remains in the same surface sediment samples, were in reasonable agreement in response to the changes in measured environmental variables, particularly lake trophic state. This suggests that the structure of cladoceran and diatom communities could have been similarly affected by nutrient transfer within lake food webs. Variation in nutrient-associated factors such as habitat, however, also had a significant impact on the functional structure of cladoceran communities. In contrast, there appears to be no simple link between predation pressure and lake trophic state across our training set lakes, but the change in predation pressure associated with nutrient loading might cause the variation in temporal patterns of the cladoceran community within lakes. Results described here provide further justification for using Cladocera remains in lake sediments as independent indicators of environmental conditions, reconstructing past variations in nutrient status, and defining reference conditions for impacted surface waters. This study highlights the fact that both taxonomic and functional variation in communities, as well as the use of multiple biotic indicators, can provide more complete evidence about variations in the structure and function of aquatic ecosystems.