The Mediterranean intercalibration exercise on soft-bottom benthic invertebrates with special emphasis on the Italian situation

Problem

The Water Framework Directive (WFD) requires that the water bodies of all Member States (MSs) must achieve good quality status by 2015. This Directive, designed in view of an integrated water management plan, represents a milestone in the evolution of water policies, not only for the area covered, which extends to all aquatic systems: surface waters (rivers, lakes, coastal and transitional waters) and groundwaters, but also for its ecosystem-based approach to water management. The European WFD substantially changes the concept of Water Quality, by assuming that a water body needs to be protected as an environmental resource and not as something to be exploited.

In this frame, the biological–ecological quality assumes a predominant role (Vighi et al. 2006). This innovative legislation has pointed out that an ecological evaluation of aquatic ecosystems is the most appropriate tool for the assessment of the ‘quality status’ of water bodies. Great importance is given to ‘Biological Quality Elements’ (BQEs), which need to be studied and analysed with appropriate scientific tools. Chemical and physico-chemical together with hydromorphological quality elements have to be analysed as well, for the final integrated evaluation of the ‘quality status’ of the water bodies (Borja 2005; Casazza et al. 2007).

An important step in the Directive development is the ‘Intercalibration’ exercise, requiring a comparison of data and information from different Member States (MSs) within the same ecoregion, in order to achieve a common and agreed evaluation of the ecological status. The aim of the intercalibration process is to achieve consistency and comparability of the results obtained by the different classification systems developed by European MSs for the biological quality elements (Borja et al. 2007; 2009).

The first step in the process was the identification of a set of Intercalibration Sites (IS), selected on the basis of expert judgment and other available information, as pressures, impacts and biological data, in order to build an intercalibration network. The system requirements were: for each MS at least two sites corresponding to the boundary between the normative definitions given in the Directive of ‘High’ and ‘Good’ status, and at least two sites corresponding to the boundary between the normative definitions of ‘Good’ and ‘Moderate’ status (Anonymous 2000).

Each MS classification system was then applied to the sites of the intercalibration network. The results were expressed by a standardized measure as the Ecological Quality Ratio (EQR). This ratio represents the relationship between the values of the biological parameters observed for a given typology within the water body and the values for these parameters in the Reference Conditions (RCs) for that typology. The ratio is expressed as a numerical value between zero and one, with ‘High’ Ecological Status (ES) represented by values close to one and ‘Bad’ ES by values close to zero (Anonymous 2003a). Due to the lack of pristine areas in many countries, the identification of RCs proved to be a very difficult task. RCs are therefore defined as the description of the biological quality elements that exist, or would exist, at ‘High’ status: that is, with no or very minor disturbance from human activities (Anonymous 2003b). The WFD identifies several processes for deriving RCs when an undisturbed site does not exist. RCs may be based on historical data and available information, or can be derived by models, or established by expert judgments (Anonymous 2000). Following these instructions, some researchers considered a combination of the different methods in order to define ‘virtual reference conditions’, which do not actually exist but are based upon the experience gained from the area and conceived as the potential quality elements that should be present (Borja et al. 2004; Bald et al. 2005).

Once the RCs were established and the EQR calculated, the EQR scale for each classification system was divided into five classes, ranging from ‘High’ to ‘Bad’ ES, by assigning a numerical value to each of the boundaries between the classes. The value for the boundary between the classes of ‘High’ and ‘Good’ status, and between ‘Good’ and ‘Moderate’ status, had to be established through the intercalibration exercise (Anonymous 2005): therefore the results of the intercalibration exercise determined the EQR values for the High–Good and the Good–Moderate boundaries in each MS’s classification system. Values for the other two class boundaries were established independently by the MSs themselves. The boundary between ‘Good’ and ‘Moderate’ status is especially important because it sets the target for restoration plans of water bodies that fail the environmental objective of achieving ‘Good’ ES (Heiskanen et al. 2004); if this boundary is not harmonized on a scientific basis, some MSs could be disadvantaged and others privileged, which would potentially lead to a severe disequilibrium in economic resources allocation (Buffagni et al. 2007). The intercalibration process was conducted for the different surface water and ecoregions across Europe. Taking into account latitude, longitude, tidal range and salinity, the European coastal waters were split (Anonymous 2003a) in four basic ecoregions or ecoregion complexes: the Atlantic/North Sea complex, the Baltic Sea, the Black Sea and the Mediterranean Sea. Within these ecoregions, different Geographical Intercalibration Groups (GIG) were set up. Experts for each biological quality element (phytoplankton, macroalgae, angiosperms and benthic invertebrates) attended their own GIG, forming four subgroups.

There are three different options for intercalibration depending of methods adopted by the MSs (Anonymous 2005): option 1, MS in a GIG use the same assessment method; option 2, MSs use a common metric method indentified specifically for the purposes of the intercalibration exercise; option 3, MSs compare different national methods at IS.

All these options require that data on sites cover the whole range of quality classes to secure statistical robustness of intercalibration results (Birk & Hering 2009). In the Mediterranean ecoregion the intercalibration process for the soft-bottom benthic invertebrates subgroup was established in the year 2004 by Cyprus, France, Greece, Italy and Spain, joined later by Slovenia. The above MSs agreed to follow option 3: they developed their own classification schemes, which were assessed against each other through the exchange and evaluation of data, and the boundaries in each classification scheme were harmonized to obtain an acceptable level of agreement. The exercise was concluded in June 2007.

A synthesis of the results produced by the benthic invertebrates subgroup within the Mediterranean GIG is presented in this paper. Particular attention is given to the difficulties encountered by Italy throughout the different steps of the process.

Material and Methods

In Table 1 the methods, referring to ES determination on the basis of soft-bottom macrobenthos, assessed by each MS, are shown, together with their current status. This latter is said to be ‘finalized’ if it is already part of the national legislation, and ‘officially accepted’ if it has not yet been adopted; the last column indicates whether the values corresponding to the boundaries between classes have been identified. The three methods adopted throughout the intercalibration exercise are described by formulas and referred to the relevant literature. They are considered sensitive to all kinds of pressure, particularly organic enrichment. A unique typology has been considered: soft sediment habitat (GIG 2008).

The methodology to assess the ES, the criteria for establishing biological RCs, and the boundary-setting procedure are briefly presented below for the countries that successfully concluded the intercalibration exercise (Cyprus, Greece, Slovenia and Spain), and discussed in more detail for Italy (Anonymous 2008; GIG 2008). The intercalibration sites are all located in coastal sedimentary habitats and do not comprise transitional waters, such as estuaries and coastal lagoons. Macrobenthos data were provided by each GIG member and datasets were exchanged by the participants to check the attribution of species to the different ecological classes; in some cases, when more data were needed, additional sampling was carried out by MSs. A complete list of macrobenthos data used in the Intercalibration exercise is available from the official national experts of MED–GIG for the soft-bottom benthic invertebrates quality element.

Greece and Cyprus followed the same steps in the intercalibration exercise. In Greece (total coastal length 15,147 km) three Intercalibration Sites (IS) were selected (103 samples). In Cyprus (coastal length 735 km) only one was selected (15 samples). For the assessment of ES, the Bentix Index (Simboura & Zenetos 2002) was used. This index is based on the relative percentage of sensitive and tolerant species in the benthic fauna, weighted to derive a single formula. These countries identified real reference sites, on the basis of the benthic invertebrate conditions: in these sites over 75% of the species were sensitive and the values of the Bentix Index were among the highest found (Bentix >5). The setting of the RCs was based on the autoecology of species; a species reference list was compiled to identify the species characterizing each type of community, habitat and water body, thus establishing RCs on an ecological basis (Simboura et al. 2005). Class boundary values (Table 2) were set by plotting the percentage of sensitive taxa (on the left y-axis) and of tolerant taxa (on the right one) against the decreasing Bentix Index values on the x-axis (GIG 2008,Fig. 1 page 6). The point where the two curves cross corresponds to the central value of the Good ecological class; here the two groups of sensitive and tolerant species are each 50% of the total fauna. At the High–Good class boundary (Bentix = 4.5), the percentage of sensitive taxa drops to <60% of the total fauna and the percentage of the tolerant taxa accounts for more than 40%. At the Good–Moderate class boundary (Bentix = 3.5), the percentage of tolerant species reaches over 60% (roughly 2/3 of the total fauna) and the sensitive taxa <40% (1/3 of the fauna). In muddy habitats, where the benthic fauna is normally dominated by tolerant species, the reference value of the Bentix Index and the boundary values between ‘Good’ and ‘High’ status was modified, taking into account the lower percentage of sensitive species in these habitats, and here the reference value of Bentix Index was set at 4 on the basis of tests conducted in coastal muddy habitats with known environmental pressures (Simboura & Zenetos 2002).

Spain. Experts from Catalonia and the Balearic Islands took part in the intercalibration exercise. Seven IS were selected along the 826-km of the Catalonia coast and four IS along the 1428-km coastline of the Balearic Islands; a total of 105 samples were used. A new method to assess the ES, named MEDOCC Index (from Mediterranéo Occidental) was proposed by Pinedo et al. (in GIG 2008). It is an adaptation of the AMBI Index (Borja et al. 2000) to the Western Mediterranean area: AMBI classifies species in five ecological groups depending on their sensitivity to pollution, whereas in MEDOCC the second- and first-order opportunistic species are considered together, thus reducing the number of ecological groups to four (instead of five). Another difference in comparison with AMBI is the assignment of some species to a different ecological group. RCs were set on models and expert judgment, creating a theoretical benthic community composed only of sensitive and indifferent species; the MEDOCC reference values were then calculated for this hypothetical community. Class boundary values were set by plotting the percentage of ecological groups against the value of the MEDOCC Index of the entire dataset. The sequence of quality classes and class boundaries (Table 2) was interpreted on the basis of the ecological group percentages, following the ecological model of Pearson & Rosenberg (1978) and Glémarec & Hily (1981), which refers to the benthic community succession along an increasing gradient of disturbance. At the High–Good class boundary (MEDOCC = 1.6), the sensitive ecological group accounts for more than 40% of total fauna composition. At the Good–Moderate class boundary (MEDOCC = 3.2), the tolerant ecological group accounts for 20–50% and sensitive taxa are also present (10–40%).

Slovenia joined the benthic invertebrate subgroup after the intercalibration network had already been established, using data from its monitoring program (24 samples). The Multimetric AMBI (M-AMBI) (Borja et al. 2004), which combines AMBI Index, richness and diversity in a multimetric approach, was adopted as a national method. Due to the absence of any proper reference sites along the 46-km of the Slovenian coast, RCs were set on expert judgment. Two sampling sites subject to minimal human impact were selected, and the average values of AMBI, richness and diversity calculated in these sites and increased by 15% to obtain a more confident evaluation, were considered as reference values. The boundary between ‘High’ and ‘Good’ ecological class was set taking into account the EQR values in the two sites used to calculate RCs (GIG 2008). Slovenian boundary values for the five quality classes are reported in Table 2.

Italy. Six IS (18 samples) were selected along the 8490-km of the Italian coastline (Table 3); samples were collected at these sites annually from 2002 to 2004. This dataset was not considered adequately representative for the entire national situation; hence a final decision on how to implement the classification of ecological status based on benthos has been postponed. At a later stage, the additional data from the six IS and more data made available by the National Monitoring Program of the Italian Ministry for the Environment (established by National Law 979/82), were analysed. A total of 193 samples could be processed. In the absence of an Italian national method and of reference sites, the three indices proposed within the subgroup were tested: Bentix, M-AMBI and MEDOCC, and different virtual RCs were derived using all the Italian data and tested.

The three indices were applied and the boundary values between classes were derived differently depending on the index used. The Bentix Index was applied using the best index value as reference, and those as boundaries values published by the Bentix Index authors for mixed sediments.

The M-AMBI was applied testing different combinations of virtual RCs and boundaries. The choice retained for the comparison has been to consider as reference samples those in which the percentage of individuals belonging to sensitive species was more than 70%. In these reference samples the median values of richness, diversity and AMBI Index were used as reference values. The 10th percentile of the EQR values in the reference samples was selected as the boundary between ‘High’ and ‘Good’ ES; the width of the four other classes was evenly spaced over the remaining interval.

The MEDOCC Index was applied testing different reference values obtained following the same approach described above for the application of M-AMBI. The class boundary values were obtained by plotting the percentage of ecological groups against the values of the MEDOCC index for the Italian dataset. All the methods tested on Italian data, and the related reference and class boundary values, are summarized in Table 3.

Results

Each member of the Mediterranean GIG subgroup provided a classification of the IS, using their own national method. The first step in the intercalibration exercise was to test the level of agreement among different classification systems using common criteria.

After the first review, the European Commission (EC) requested further clarification on the degree of comparability between some specific results of the intercalibration exercise. The main problem was related to the different criteria used to evaluate whether the assessment results were comparable, making it very difficult to judge whether the intercalibration exercise has achieved the same level of comparability for all the results. In response to the request by the EC, all GIGs have re-analysed their data, calculating a number of common comparability metrics according to the two criteria suggested in Document ENV-COM240108-5 by Van de Bund et al. (2008):

The average class difference was the main criterion recommended to be used. It shows to which extent the Member States’ methods may give different classification results. The possible criterion for sufficient comparability is proposed to be less than a half class (0.5) difference (Van de Bund et al. 2008).

Differences between classification systems can be due to systematic differences and/or random error. The smaller the difference, the better the comparability between the classification systems:

The second criterion proposed to be used in the evaluation is the percentage of agreement between Member States’ classification methods. This can be used only as a supporting tool, as these values are highly sensitive to the data distribution over the quality range, and thus need to be considered with caution. Following the judgment based on these two criteria, the different National assessment methods used in the Mediterranean benthic invertebrate subgroup proved to be sufficiently comparable.

In Fig. 1 the EQR values obtained by the finalized or officially accepted classification systems (Bentix for Greece and Cyprus, MEDOCC for Spain and M-AMBI for Slovenia) are compared two by two. Samples identically classified by the two methods lie in the grey area. The highest percentage of agreement (67%) is between MEDOCC and Bentix; it is about 62% between M-AMBI and Bentix, and only about 45% between MEDOCC and M-AMBI. Generally MEDOCC appears less strict than the other two methods: 15% of the Greek and Spanish samples are classified in ‘High’ ES by MEDOCC and in ‘Good’ ES by Bentix, whereas 40% of the Spanish and Slovenian samples are classified in ‘High’ ES by MEDOCC and in ‘Good’ ES by M-AMBI. Despite these differences, the absolute average class difference among all the classification systems is equal to 0.43, meaning that the classification systems on average give the same results in over 50% of cases (Van de Bund et al. 2008).

As far as the Italian data are concerned, the classifications obtained by applying the three methods to the data collected in five different years for the six IS have been compared: the results are shown in Table 4. The ES values from the benthos data were compared with an a priori classification of the six IS, obtained by expert judgment on the basis of the available water quality information. The application of M-AMBI corresponded with the a priori classification of the ES in four cases, whereas the other two methods corresponded only three times. A worse correspondence is generally shown in the sites defined a priori in Good–Moderate ES; for example, the three methods classify the Marinella site as ‘High’.

In Fig. 2 the different percentages of classification in the five classes of all the Italian samples obtained by applying the three methods is shown: M-AMBI appears to be the strictest index with only 33% of samples classified in ‘High’ ES, as opposed to the 58–60% obtained by both MEDOCC and Bentix. Very few samples are classified in ‘Moderate’, ‘Poor’ or ‘Bad’ ES. The percentage of agreement between the different methods, compared two by two, is 46%. The worst results are between M-AMBI and MEDOCC (38% agreement). The absolute average class difference is equal to 0.68.

Discussion

The intercalibration represented an exercise on a significant scale. As it was part of a common effort to reconcile the diversity of methods and approaches already in place across Europe, it has also promoted considerable research and monitoring effort to gather appropriate data and to develop indicators for the benthic invertebrate quality elements (Casazza et al. 2004, 2005; Borja et al. 2007; Simboura & Reizopoulou 2008).

The level of agreement obtained comparing the different classification systems proposed by Greece, Cyprus, Spain and Slovenia can be considered sufficient after the intercalibration exercise. The absolute average class difference among the classification systems was 0.43, which is within the acceptable range (lower than 0.5), although the absolute classification difference between the MEDOCC and M-AMBI indices on data from Spain and Slovenia was 0.57.

However, the Mediterranean exercise remains incomplete, calling for a provision allowing for updates as MSs complete the exercise for additional quality elements during the next 4–5 years. That the process remains open is not only a reflection of the dynamic nature of the European Union, whose membership has been growing, adding new national methods and boundaries that would need to be intercalibrated against existing ones. The selection of reference conditions itself, within the WFD, is a dynamic process: in this, all the MSs should propose and select their own RCs for each of the typologies. However, by changing the RCs the final result can be very different. The conditions and results can be revised and modified during the intercalibration process, in which the other MSs can provide further data for validation of these reference values (Muxika et al. 2007; Borja et al. 2009).

According to the papers of Dauvin et al. (2008a,b), the implementation of European Directives has drawn attention to the need to refine the classification system for marine habitats and to redefine the terminology used. The present lack of consensus can be historically explained by the existence of classification schemes adopted by different academic schools and corresponding national administrative regulations (among which the most commonly used are the French and UK systems). For the Mediterranean, the benthic system set by the Marseille school (Pérès & Picard 1964) has a long and glorious tradition, which encompasses many peculiarities of this regional sea and has yet to be fully harmonized with concepts and terms derived from the Atlantic experience.

As regards Italy, the results used in the intercalibration exercise and presented here have not been considered definitive, as a national method is still under development and an agreement on class boundaries has not yet been reached. In the WFD marine quality assessment, the following key steps can be identified (Muxika et al. 2007): the classification of water bodies and typologies; the selection of RCs for each of the typologies, representative of the absence of anthropogenic influence; the availability of good classification tools or metrics; the suitability of ecological class boundaries. Following our analysis of data, we underline the main problems that arose for Italy at different steps of the intercalibration process.

First of all, the dataset consisting of six Intercalibration Sites (four on the Adriatic coast and two on the west coast) was considered insufficient, especially considering the Italian coastal length. The availability of new data from more stations, more evenly distributed along the Italian coast, did not allow all the steps of the procedure to be re-developed.

In fact, in several benthic sampling stations, chemical-physical parameters (e.g. oxygen, organic matter, sediment texture pattern) had not been measured regularly, making it impossible to confront the overall Ecological Status (ES) from benthic fauna with additional information.

A major problem was related to the lack of reference conditions, which are considered the starting point for the classification requested by the WFD (Van de Bund 2004) but are also critical in the whole evaluation process (Borja et al. 2009). In the case of Italy based on the dataset, different virtual RCs were tested. The mean or median value of reference sites was retained for the subsequent analysis, being considered to be the most robust values for use as a reference (Wallin et al. 2003).

In the absence of an Italian national method, we tested the three methods proposed in the Mediterranean benthic subgroup, following the recommendation by Díaz et al. (2004) that the suitability of indices that already exist should be evaluated preferably. The methods were tested on the Intercalibration Sites to compare the results of ES with initial a priori classification given by expert judgment: an insufficient agreement was found. Lacking a good opportunity for validation in a condition showing a clear trend of water quality (similar, for example, to the cases shown in Muxika et al. 2005), we could not trust too much the a priori ES of the selected IS (which is one of the three options for index validation proposed by Borja & Dauer 2008), as their degree of impact was not adequately assessed. The comparison among the three methods was also carried out on the whole set of data collected in Italy. The three methods behaved in a different manner, with the M-AMBI proving to be more ‘severe’ than the remaining two. Furthermore, the Italian coast is extremely long and shows variable features, so the experience gained until now suggests that to select the same reference value for all the regions could be unrealistic.

Conclusions

Regarding the soft bottom benthic invertebrates MED–GIG intercalibration, the assessment methods proposed by Greece and Cyprus, Spain and Slovenia proved to be sufficiently comparable, although further harmonization would be needed. Italy and France have not yet completed the process and have not been able to provide a good comparability between their IS; moreover, some cases (i.e. comparison between Spanish and Slovenian results) show a rather low level of comparability. Future work should focus on a better harmonization especially to reduce the uncertainties in the intercalibration methods and to develop more harmonized and precise procedures for setting RCs.

Italy is working towards evaluating additional reference sites that take into account more of the Italian coastal variability in order to complete the entire process with the selection of a national method and related Reference Conditions.

Further work is also needed to cover the key pressures that have not been assessed in this first phase. Most of the methods which have already been agreed on a European scale, are known to respond to specific pressures (e.g. eutrophication), which might be relatively less important for the Mediterranean Sea. In the future, it will be important to look at how biological indicators respond to different combinations of pressures. Another issue to be considered in the new intercalibration work programme is the possibility of dividing the Mediterranean ecoregion into at least two separate sub-regions (e.g. Eastern and Western Mediterranean), due to the high diversity and richness of coastal habitats within the Mediterranean region, together with the extremely complex difference in its environmental conditions. During this first phase of the intercalibration exercise, from the common analysis of soft-bottom benthic community features across the Mediterranean, it is clear that richness and diversity values are higher in Greece and Cyprus than in Spain and Italy. Moreover, as the classification methods presented so far refer only to soft-bottom habitats, excluding transition waters, further work needs to be carried out to develop methods that can be used for hard-bottom substrates (Mangialajo et al. 2007; Pinedo et al. 2007) and lagoon habitats (Salas et al. 2004; Dauvin 2007; Blanchet et al. 2008; Mistri et al. 2008).