Introduction

This work focuses on the defining of reference conditions, which is one of the major keystones of ecological assessment of water body status required by the European Commission Water Framework Directive (WFD; EC 2000). Most biological assessment systems are based on the concept of comparing the current biological community to the “reference conditions”—a status of community observed in the absence of human disturbance or alteration (Bailey and others 2004). Therefore, reference conditions can serve as an important guide to set management aims, but it is important to emphasize that reference conditions are not the same as restoration goals (Egan and Howell 2001). However, the term “reference condition” has been used to refer to multiple, and often confusing and contradictory, concepts. For example, it can be used to refer to the condition of ecosystems at some point in the past; the best remaining conditions in a region heavily modified by human activity and so on (Stoddard and others 2006). In this article, the term “reference conditions” is used in the WFD sense which defined type-specific reference conditions as the biological conditions associated with no or very low human pressure (EC 2000).

Reference conditions can be defined in a number of different ways (Reynoldson and others 1997; Stoddard and others 2006). By far, the most common approach for estimating reference conditions is to quantify them at a set of sites relatively unexposed by human activity. This approach is widely known as the “reference site approach” (Hughes 1995; Bailey and others 2004) and presents a scientifically sound method for setting expectations, provided that the method of selecting reference sites is clearly defined. However, the spatial approach faces a practical obstacle because it is often difficult to find undisturbed sites. Establishment of reference sites is especially challenging in Europe (Moss and others 2003) for all but polar and mountain areas, and even there, climate warming and airborne pollutants make it unlikely that true reference sites exist. It may be especially difficult to find minimally impacted waters for some particular types, e.g., shallow lowland lakes (Bennion and others 2004).

The derivation of reference conditions from paleolimnological studies offers an alternative method in such cases. Paleolimnological research has a long tradition in Europe using various elements e.g., diatoms (Bennion and others 1995) and pigments (Kamenik and others 2000). Nevertheless, the reconstructions are often subject to variabilities of more than one order of magnitude (Sayer 2001), and this can obscure precise reconstruction of reference conditions.

Historical data probably give the best insight into how reference lakes looked. They often suffer, however, from poor quality, be it due to different sampling methods or different taxonomic resolution. The basic problems are the paucity of data available and the definition of the “reference period” which is usually considered the period before 1850 (Battarbee 1999) or before the Second World War if impacts from anthropogenic land use and urbanisation can be considered as negligible (Reichmann and Schulz 2004).

In situations without minimally disturbed sites and historical data, empirical models derived from associations between biological indicators and human-disturbance gradients can be extrapolated to infer conditions in the absence of human disturbance (Stoddard and others 2006). A promising approach to estimate the natural trophic state for lakes is the use of models that predict the trophic state of a lake from nutrient loading (e.g., Vollenweider 1976; OECD 1982), however, it must be borne in mind that any empirical model involving nutrient export coefficients bears various methodological errors (Ryding and Rust 1989).

It is acknowledged (Moss and others 2003) that, currently, the best approach is a combination of reference sites, modelling, paleolimnology and expert judgment. Integrating several approaches may lead to firmer, more defensible reference conditions, particularly if the conclusions derived from the different approaches are consistent (Stoddard and others 2006). The process depends on the accumulation of experience and integration from all of these lines of evidence rather than statistically rigorous procedures (Moss and others 2003). Such an approach was used for setting chlorophyll-a reference conditions in European water legislation and is presented in this article.

According to the Water Framework Directive, Member States are required to develop lake ecological status assessment systems based on various characteristics of phytoplankton, macrophytes, benthic invertebrates and fish fauna. An intercalibration exercise is foreseen to harmonise assessment systems of all EU and to ensure that the obligation to reach good status has the same meaning throughout Europe. The first stage of the intercalibration (2003–2008) aimed at setting of reference conditions and class boundaries for chlorophyll-a values for all lake types and all geographical regions of the EU. The work was carried out under the Common Implementation strategy of the Water Framework Directive (EC 2001) by joint effort of all European countries and greatly supported by several scientific studies of the European Framework 6 project REBECCA, e.g. setting of total phosphorus (TP) reference conditions (Cardoso and others 2007), preliminary analyses of chlorophyll-a values in reference lakes of Northern and Central regions (Carvalho and others 2008), development of chlorophyll-nutrient relationships (Phillips and others 2008) and phytoplankton responses to eutrophication (Ptacnik and others 2008).

The initial steps in development of reference conditions were (1) identification of anthropogenic pressures and selecting appropriate biological components and indicators to address their impact; (2) defining geographically homogenous regions and common types of lakes within them. This study has focused on the major pressure for lakes of Europe—eutrophication and the most relevant community detecting this pressure—phytoplankton. Chlorophyll-a was selected as a simple indicator for phytoplankton abundance with sufficient data availability across Europe. Five lake regions including all of the EU Member States and Norway were established based on Illies limnofaunistic division of Europe (1978)—Alpine, Atlantic, Central/Baltic, Mediterranean and Northern regions (EC 2005). Common lake typology with fifteen international lake types were selected for intercalibration (Table 1) based on natural abiotic characteristics—altitude, alkalinity and mean depth—which are important factors in determining the composition and abundance of biological communities (e.g., Kolada and others 2005). Additional factors were used in several regions, e.g., lake area (Alpine region), humic substances (Nordic region), climate factors (Mediterranean region), water retention time (Central region) the importance of which has been demonstrated in several studies (e.g., Miettinen and others 2005).

Table 1 Description of common international lake types included in analysis
Full size table

In short, this paper summarizes the evolution of type-specific chlorophyll-a reference conditions and describes the data and research used in their development. More specifically, we set the following objectives: (1) to describe and compare different approaches of selection of reference lakes and setting reference conditions and to evaluate consistency of the results; (2) to determine whether selected reference lakes may be considered representative of the whole lake population; (3) to explore relationships between chlorophyll-a values and environment variables and define the most important drivers of chlorophyll-a concentration in unimpacted lakes; (4) to develop empirical models for estimating site-specific reference chlorophyll-a values from a set of potential predictor variables.

Material and Methods

Data Sets

Altogether, data for approximately 1200 lakes and 2700 lake years were pooled from national datasets into intercalibration databases (see Table 2). These databases contained both basic data (altitude, surface area, mean depth, alkalinity), quality data (chlorophyll-a, nutrients, Secchi depth) and pressure data (land use, population density, other impacts).

Table 2 Description of lake data sets by region, type and country
Full size table

Data were collected from environment agencies and scientific institutes including data both from national monitoring programs and several research projects. Inevitably, with such a large dataset of lakes from many countries, there were questions regarding data quality which were solved before the data analyses:

  • Checking and correction of data were required before the data could be used: a common problem was erroneous units, values below detection limits coded in different ways, as well as numerous other irregularities;

  • Data quality was further checked by revealing outliers and testing well established relationships (e.g., between conductivity and alkalinity, chlorophyll-a and TP).

One problem was the heterogeneity of the data: due to different data origins, different sampling and analysis methods were used (except Mediterranean region, where sampling was carried out using an unified strategy). Despite the large heterogeneity of the data, some common patterns can be defined:

  • Samples were mostly from the vegetation season; also winter/spring season samples included in the Alpine region;

  • Approximately 4 sampling dates per season were used (ranging from 1–2 to 10);

  • Most samples were taken from the epilimnion/surface layer; Mediterranean region—samples from the euphotic zone, defined as 2.5 Secchi depth;

  • Spectrophotometry with ethanol/acetone extraction was used for chlorophyll-a measurements.

For our analyses, we used chlorophyll-a annual average values (Alpine region) or average values of vegetation season (all other regions).

Selection of Reference Lakes

A preferred approach to locate reference sites is to establish a set of criteria that, in total, describe the characteristics of sites in a region that are not or only minimally exposed to stressors (Stoddard and others 2006). It is advisable to use a set of reference criteria based on pressure data, not chemical and biological parameters which must be used only for confirmation of, not for selection of, reference sites because of the possibility of circularity and preconceived notions about the chemistry and biology at a “typical reference site” (Bailey and others 2004).

A list of criteria for the selection of reference sites was developed in every region (Table 3) based on criteria assessing the pressure from the catchment. Land use, point sources and population density were the main factors, but additional elements were included by several regions, such as the change of the natural regime, artificial modifications of the shoreline, introduction of fish and fish-farming activities, mass recreation, and invasive species. Some Member States (UK, Ireland, and Austria) additionally used paleolimnological data for confirmation of reference sites. For instance, Taylor and others (2006) found that diatom assemblages in the sediment core top samples from 11 of 34 candidate reference lakes in Ireland showed relatively little deviation from those in sediment core bottom samples. The core bottom samples appeared to pre-date ca. 1850, or the onset of agricultural intensification and major aforestation in the catchment, so those sites were selected as reference sites in the Atlantic region.

Table 3 Pressure criteria used for reference lake selection
Full size table

Historical quantitative data on phytoplankton were used only in the Alpine region, available from the 1930s for Carinthian lakes (Findenegg 1935; 1954) and for several lakes in the Northern Calcareous Alps (Ruttner 1937). Since these lakes were not affected by major anthropogenic pressure from industrialisation, intensive urbanisation or agriculture, the 1930s reflect reference conditions with insignificant anthropogenic impact.

A distinctive approach was taken in the Alpine region in which sites were accepted as reference sites if their actual trophic state did not deviate from the reference trophic state prior to industrialisation, intensive urbanisation or agriculture. From paleoreconstruction (e.g., Löffler 1972; Klee and others 1993) and theoretical considerations using the Vollenweider phosphorus loading model (Vollenweider 1976; OECD 1982), it was concluded that oligotrophy is the natural reference trophic state of deep Alpine lakes (AL3, mean depth >15 m). Lakes belonging to the lake type AL4 (mean depth 3–15 m), however, tend to have a higher reference trophic state at oligomesotrophic level. This is proved again by loading model calculations and paleoreconstruction (e.g., Lotter 2001; Schmidt and others 2002).

In summary, the reference lakes have been selected using the following approaches or a combination of these approaches:

  • Criteria assessing the pressure from the catchment, e.g., predominantly (90%) natural land cover, absence of major point sources, population density (e.g. <10 inhabitants/km²), were used in all regions for initial selection of candidate reference lakes;

  • TP concentration corresponding to the defined natural trophic state was used in the Alpine region;

  • Paleoreconstruction for selection or confirmation of reference sites was used in the Alpine, Atlantic and Northern regions;

  • Historical data reflecting the reference state were used in Alpine region;

  • Phosphorus loading model (Alpine and Nordic regions) was used for TP concentrations in the reference lakes;

  • Additional screening by quality criteria (nutrient, chlorophyll-a) and expert judgment was broadly used in the final review of reference lake lists.

According to the reference criteria, 359 reference lakes were selected across the EU. Most of the reference lakes (267 lakes) belong to the Northern region (Table 2), dominated by lakes in Norway (136 lakes) and Finland (95 lakes), while the lowest numbers occur in the Central Baltic (38 lakes) and Mediterranean region (10 reservoirs), which can be explained both by data availability and the level of anthropogenic pressure in those regions.

Setting of Reference Conditions and High/Good Status Boundary

In different regions chlorophyll-a reference values and the values corresponding to the boundary between the High and Good quality classes (H/G boundary) were established by following common principles (median of chlorophyll-a values in reference lake type specific populations was used for setting reference value, 75th–95th percentile—for setting H/G boundary), but slightly different methods.

In the Alpine region, class boundaries were set in two steps: (1) the reference conditions and boundaries were set for the annual mean total phytoplankton biovolume; (2) the reference value and boundaries for chlorophyll-a were derived from regression with total biovolume. The use of total biovolume is justified by the fact that historical data from the 1930s, which represent the best reference data, are available for total biovolume only, not for chlorophyll-a values. The reference value was calculated as the median of the values measured in the set of selected reference lakes, while H/G boundary—as 95th percentile of values in the reference lake population. The use of 95th percentile was justified by the strict criteria used for selection of reference sites and use of the arithmetic means of the lakes (1–19 years each) in the analyses, instead of single lake-years, in order to prevent a bias toward lakes with more data available.

In the Central-Baltic region, the median value of mean chlorophyll-a concentrations in reference lakes was used as the reference value for chlorophyll-a and the 75th percentile—as the H/G boundary. The 75th percentile was considered more appropriate for setting the H/G boundary than the 90th percentile which would result in a relatively high proportion of lakes that would be assessed to have high status but not assigned as reference lakes. To avoid the problem of insufficient data, the analysis was based on lake-years, not single lake averages. The results were compared with similar lake types from the Northern region and results of project REBECCA (Carvalho 2008) and found to be similar.

In the Mediterranean region, the median chlorophyll-a value in reference sites was taken for reference conditions (in fact, for Maximum Ecological Potential, as both Mediterranean types represented only reservoirs). The High/Good potential boundary is not required to be reported for heavily modified or artificial water bodies so it was not calculated. The reference lake number was small in this region, and future research is planned to revise the current results.

In the Northern region, the reference value was calculated as the median value of type-specific chlorophyll a of reference lakes, supplemented with expert judgments for types with insufficient data. H/G boundaries were set primarily at the 90th percentile of the distribution of the metric in reference lakes, thereafter, the values were compared with the response curves of phytoplankton taxonomic indicators (Ptacnik and others 2008) in conjunction with statistical analysis of Member State reference lake populations.

It is well known that no single value can represent reference conditions over all types of water bodies: ecosystems are complex and their characteristics mutually vary within large ranges, determined by external and internal factors (Moss and others 2003). Therefore, the final results of reference conditions and the H/G boundary were expressed as ranges, not fixed values for the following reasons: (1) a broad range of natural conditions within every common lake type (2) different monitoring practices in use, e.g., sampling depth, time and frequency, which have influences on chlorophyll-a data; (3) in the Mediterranean region, the main concern was interannual variability since the results were derived from one single sampling year dataset.

Statistical Analysis

Statistical analyses were carried out using the following methods:

  • To derive type-specific reference conditions, descriptive statistics were used for chlorophyll-a for each lake type (medians, quartiles and percentiles);

  • Cumulative frequency analyses were used for defining reference conditions (the cumulative distributions of the reference lake population were compared to the non-reference lake population);

  • Analysis of Variance (ANOVA) was used to compare the mean chlorophyll-a concentrations among regions, types and Member states;

  • Mann–Whitney U test was used to examine how representative the selected reference sites are of all lake populations. To address this issue, selected descriptors (altitude, depth, area, alkalinity, conductivity, TP, chlorophyll-a, Secchi depth) of type-specific reference lake populations were compared with impacted lake populations;

  • General Linear Model (GLM) was used to estimate the best model to predict mean chlorophyll-a from several predictor variables. 2 types of predictors were used to estimate chlorophyll-a values: (1) altitude, alkalinity, mean depth, surface area (log transformed in order to obtain all normally distributed variables); (2) humic type and region used as dummy variables—3 humic types (low, medium and high colour) and 5 regions (Alpine, Atlantic, Central/Baltic, Mediterranean and Northern).

Results and Discussion

Chlorophyll-a Concentrations in European Reference Lakes

The median type-specific chlorophyll-a concentrations in reference lake populations ranged from 1.4 to 7.8 μg/l (see basic statistics in Table 4) and in general followed the natural trophic gradient, influenced by depth, alkalinity and humic level (Fig. 1).

Table 4 Chlorophyll-a concentration in European reference lakes: mean, median, minimum and maximum values, lower (25%) and upper (75%) quartile, standard deviation and number of lakes (n)
Full size table
Fig. 1
figure 1

Boxplots comparing chlorophyll reference conditions by different lake IC common types. The middle square indicates the median value, the top and bottom of the box are the upper (75%) and lower (25%) percentiles, the upper and lower line extend to the limits of non-outlier range, and the values beyond the lines represent outliers and extreme values

Full size image

Depth was an important factor impacting chlorophyll-a reference values: the lowest values were found in deep Mediterranean reservoirs (1.4 and 1.8 μg/l), deep Alpine lake type AL3 (2.0 μg/l) and deep Northern lake type N2b (2.0 μg/l); there was no significant difference between the deep lake reference populations (ANOVA, P > 0.1). Conversely, the highest values were recorded for only one very shallow lake type, LCB2 (depth <3 m, median chlorophyll-a value 7.5 μg/l, which clearly differs from deep lakes (ANOVA, P < 0.001).

Surprisingly, the most important factor was water colour—all three humic lake types (colour 30–90 mg Pt/l) had significantly higher chlorophyll-a values compared to non-humic lake types, Mediterranean and Alpine lakes (Fig. 1, humic types N3a, N6a, N8a). The highest concentrations were found in shallow humic moderate alkalinity lake type (median value 7.8 μg/l) which was significantly higher compared with all other lake types (ANOVA, P < 0.01). Still, there were relatively few lakes in humic lake types N6a and N8a, so additional research is needed to draw clear conclusions.

Multiple Regression Model for Site-Specific Reference Chlorophyll-a

The results of the GLM model using independent predictors are presented in Table 5, revealing a significant effect of region, humic type, altitude, depth, and alkalinity, but not lake area. Overall, the variance explained by the model was 48.0% which is clearly acceptable and comparable with variability explained by TP model (51.4%, Cardoso and others 2007) or by diatom-inferred TP (47%, Bradshaw and Anderson 2001).

Table 5 General linear model result for chlorophyll-a variation in European reference lakes
Full size table

Our analyses has highlighted that a geographical region, humic type, altitude, depth and alkalinity all have a significant relationship with chlorophyll-a (Table 5) and several different regression models for predicting chlorophyll-a concentrations are required across Europe to take account these predictors (Table 6):

Table 6 Equations predicting chlorophyll reference conditions in European lakes, using humic type, altitude (alt), alkalinity (alk) and mean depth (Z) as independent predictors
Full size table

  • The strongest gradients were observed for humic type: chlorophyll-a concentrations generally increased with increasing humic content (P < 0.00001);

  • Depth (r = −0.14, P = 0.0073) and altitude (r = −0.09, P = 0.0038) had a negative correlation, while alkalinity was in positive correlation with chlorophyll-a values (0.11, P = 0.043);

  • As for the region effect, we found only two statistically significant classes: lakes in the Central Baltic region had on average more chlorophyll-a, but Mediterranean lakes less chlorophyll-a than lakes on the Alpine, Atlantic and Northern regions.

These findings are in accordance with well-established relationships between TP concentration and alkalinity (e.g., Vighi and Chiaudani 1985; Cardoso and others 2007), altitude (Müller and others 1998; Cardoso and others 2007) and depth (e.g., Ryder and others 1974). Our results only partly agree with Cardoso and others (2007) which found higher TP concentrations both in Central Baltic and Mediterranean region. The difference can be explained by the low number of Mediterranean lakes included in both studies and Mediterranean lake types which comprise reservoirs which biological characteristics differ from natural lakes. Higher chlorophyll-a values in the Central Baltic region can be explained by a type effect (lakes are shallower with higher alkalinity and lower altitude) and latitudinal effect related to temperature and its effects on mineralization in the catchment.

More controversial is the relationship between humic substances and chlorophyll-a values—for a long time, humic lakes were considered unproductive systems, as humic substances form complexes with phosphate ions and organophosphorus compounds, thereby reducing phosphorus availability to phytoplankton (Jones and others 1993). From the other side, it is now known from extensive recent experimental work that both natural ultraviolet as well as visible light induce major photolytic changes in complex organic molecules and generate large quantities of readily utilisable substrates for microbial metabolism (e.g., Winter and others 2007) which make available a huge reservoir of organic carbon and energy (e.g., Salonen and others 1992). Other possible mechanisms include light adaptation yielding higher Chl:biomas ratios (Phillips and others 2008) and the selective attenuation of ultraviolet light by humic substances that protects algae from photoinhibition in the surface layers (Moeller 1994) Several studies already have demonstrated higher chlorophyll-a values in humic lakes (Jasser 1997; Webster and others 2008), our data also confirm high productivity of humic ecosystems.

Representativeness of Reference Lakes

In general, the reference lake population represented all lake populations; there were no significant differences between reference and non-reference lakes by hydromorphological and physico-chemical (alkalinity, colour) parameters. Nevertheless, there were some exceptions where reference lake selection or type characteristics must be reconsidered:

  • CB1 type reference lakes were significantly deeper than CB1 non-reference lakes (median mean depth values, 7.7 and 5.9 m respectively), less alkaline (median alkalinity values, 2.0 and 2.5 meq/l respectively) and with lower humic content (median for reference lakes, 18 mg Pt/l, and for non-reference lakes, 38 mg Pt/l);

  • Also, N2a reference lakes were significantly deeper and less alkaline than non-reference lakes of this type, and N2b reference lakes possessed lower alkalinity than non-reference lakes;

  • There were also differences in CB2 and CB3 lake types, but the small number of available reference lakes hinders drawing of meaningful conclusions.

As expected, in most lake types reference lakes differed significantly from non-reference lakes in chlorophyll-a, TP and Secchi depth. Nevertheless, in several types (AL4, CB3, N2b, N6a, N8a), reference lake chlorophyll-a distribution did not differ significantly from impacted lake chlorophyll-a distribution (N2b median value for reference lakes 2.0 μg/l, for non-reference lakes 2.5 μg/l; N6a median value for reference lakes 3.8 μg/l, for non-reference lakes 3.3 μg/l). In fact, in some lake types, reference lake populations outnumber impacted lakes; e.g., there are 71 reference and 25 non-reference lakes within the N2b type population. Even if some sound reasons for such homogeneity could be supposed (e.g., the whole type is relatively unimpacted), it is necessary to review the reference lake selection criteria and the sensitivity of the selected indicators (TP, chlorophyll-a, Secchi) to pressure factors occurring in these lake types.

Chlorophyll-a Reference Conditions in European Water Legislation

The final outcome of our work consists of reference conditions and establishment of a High/Good status boundary based on chlorophyll-a for the common lake types selected within the Geographical Intercalibration Groups for lakes. The chlorophyll-a values and ranges are given in Table 7.

Table 7 Chlorophyll-a reference conditions, High/Good boundary values and ranges by IC lake type defined and agreed in the WFD Intercalibration process
Full size table

Recently, the outcome of the work formulated into the EU legislation as “Commission Decision establishing the values of the Member State monitoring system classifications as a result of the Intercalibration exercise” (EC 2008), therefore marking the first attempt in international water policy to move from physico-chemical quality standards to ecological quality targets.

The Member States shall use the chlorophyll-a ranges defined for the common types to set the most suitable boundaries for their national types according to the type characteristics, e.g., lakes types with low alkalinity, low humic matter, high altitude, high depth, and/or short retention time correspond with reference values close to the minimum of the range and vice versa.

Comparison with Chlorophyll-a Reference Values USA and Australia

Also, other nations have codified the concept of reference conditions in their legislation. For instance, in United States ecoregional reference conditions were established representing conditions minimally impacted by human activities (US EPA 2000). In general, the chlorophyll-a reference values are of the same magnitude as those for comparable lakes in Europe (1.9–4.9 μg/l). Although several approaches for setting reference conditions were proposed (historical data, predictive models, data of minimally impacted sites), in practice the 25th percentile of a sample distribution from the entire population of lakes was used to derive reference values. As a consequence, reference conditions are significantly higher for ecoregions with higher overall human impact (e.g., ecoregion VI, Western Corn Belt plains and ecoregion XIII, Southern Florida Coastal Plain).

Chlorophyll-a reference values for Australian freshwater lakes were derived using the statistical distribution of reference lake data collected within five geographical regions across Australia and New Zealand (ANZECC 2000). Although the term “reference condition” was used in a much less stringent way, allowing use of altered sites when it is not possible to find unimpacted sites, chlorophyll-a reference values for Australian lakes range from 3 to 5 μg/l and lie in the same magnitude as those found for lakes in Europe.

It can be concluded that lake chlorophyll-a reference conditions set in the European, USA and Australian legislation frameworks are broadly comparable, even if there are substantial differences both in theoretical background and practical application of the reference condition concept and setting environmental quality targets.

Methodological Concerns and Future Directions

There are several aspects which strongly influence the confidence of the chlorophyll-a reference values e.g., insufficient number of reference lakes, inherently large heterogeneity of data (different sampling and analyses methods); insufficient geographical coverage of the data; high natural variability within common lake types.

The main problem is that the present work has only focused on eutrophication pressure and only on quantitative part of phytoplankton while considering other pressures and taxonomic composition of phytoplankton are still the tasks for the nearest future.

We therefore believe that the work to be continued for the period of the next River Basin Management Plan, and that a longer period is needed to validate the present results and develop reference conditions both for phytoplankton biomass and taxonomic composition with higher confidence.

Conclusions

  1. 1.

    Mainly a “reference site” approach—selection of lakes with no or very minor human impact—was used for setting chlorophyll-a reference conditions in Europe. The selection of reference lakes based on criteria assessing the pressure from the catchment. (land-use, population density and absence of point sources). In some regions paleolimnological data, historical data and modelling of nutrient load were used to validate their choice of reference sites.

  2. 2.

    According to the reference criteria, 359 reference lakes were selected across the EU, representing fifteen common intercalibration types. In the majority of types, reference lakes may be considered type representative; nevertheless, in several cases reference lake selection or type characteristics must be reconsidered, for example, the Central-Baltic types and several Northern types. The dataset ideally needs further inclusion of lakes from the Central-Baltic (especially lake types LCB2 and LCB3) and Mediterranean region. Also, several Northern lake types have a low number of lakes and high variability of the data (LN6a and LN8a).

  3. 3.

    Reference conditions for 15 international lake types were calculated using a common principle: the median value of the measured metric at reference sites was used for reference conditions, while for the High/Good boundary, a percentile between the 75th and 95th was used.

  4. 4.

    Additionally, empirical models were developed for estimating site-specific reference chlorophyll-a concentrations from a set of potential predictor variables. Chlorophyll-a concentrations increased with humic level and alkalinity, decreased with lake depth and altitude, and varied with geographical region, while there was no clear relationship between chlorophyll-a and lake area.

  5. 5.

    A cross-region comparison of reference values and target value for chlorophyll-a shows a very good consistency between regions and types: high chlorophyll-a values are associated with low depth, high alkalinity, low altitude, and high water colour; conversely, the lowest values were defined for deep lakes with low content of humic matter.

  6. 6.

    The setting of ecological classification is included in the EU legislation (EC Decision), marking the first attempt in international water policy to move from physico-chemical quality standards to ecological quality targets.