http://efi-plus.boku.ac.at/ Project no.: 0044096 Project acronym: EFI+

Improvement and spatial extension of the European Fish Index

Instrument: STREP Thematic Priority: Scientific Support to Policies (SSP) - POLICIES-1.5

D 3.4 - Report on development of new metrics for the assessment of all European rivers including European historical diadromous fish distribution

Due date of deliverable: 31.12.2007 Actual submission date: 07.05.2008

Start date of project: 01.01.2007 Duration: 24 Month

Organisation name of lead contractor for this deliverable: CEMAGREF (HYAX), 3275 Route de Cézanne, CS 40061, 13182 AIX EN PROVENCE Cedex 5, FRANCE

Final version

Project co-funded by the European Commission within the Sixth Framework Programme (2002-2006) Dissemination Level PU Public X PP Restricted to other programme participants (including the Commission Services) RE Restricted to a group specified by the consortium (including the Commission Services) CO Confidential, only for members of the consortium (including the Commission Services)

Contents

Task 3.4 Diadromous species distribution Responsible author: Gertrud Haidvogl, University of natural resources and applied life sciences, Vienna, Austria.

Task 3.5 Central/Eastern Rivers assessment Responsible authors: Klaus Battes and Karina Battes, Bacau University, Romania.

Task 3.6 Mediterranean Rivers assessment Responsible authors: Teresa Ferreira and Pedro Segurado, Instituto superior de Agronomia, Lisbon, Portugal.

Task 3.7 Large Floodplain Rivers assessment Responsible author: Christian Wolter, Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany.

Task 3.8 Low species diversity rivers assessment Responsible author: Didier Pont, CEMAGREF Aix en Provence, France.

2 Table of Contents

1. Objective of the task...... 3 2. Data...... 4 2.1. Selection of diadromous fish to be considered...... 4 2.2. Information sources for historical distribution ...... 5 2.2.1. Using historical data on fish species distribution...... 6 2.3. Collection of data on present distribution...... 7 3. Methods...... 7 3.1. Data collection...... 7 4. Results – historical distribution...... 9 5. Modelling of the potential distribution ...... 17 6. References ...... 18

3

1. Objective of the task The existing EFI developed in the EU-funded FAME-project showed only very weak response to continuum interruptions. This was true for the overall index as well as for the ten singular metrics of the EFI with the exception of the metric for migratory species. Thus, it was one of the aims of the EFI+ project, to improve the ability of the index to detect pressures on river continuum by enhancing the data basis and by calculating particular metrics. With respect to the improvement of the data basis data we integrated in EFI+ a larger number and more precise pressure variables for river continuum. While in FAME only the connectivity situation at the river basin scale and at the segment scale was considered in general (existence of a barrier preventing the upstream migration of diadromous and potamodromous fish), we integrated in EFI+ altogether seven pressure variables in the central database. One variable accounted again for the catchment scale (existence of barrier downstream to the sea), six further variables are related to migration barriers at the segment scale (existence of barriers up- /downstream, number of barriers in the segment up-/downstream, distance to next barrier up- /downstream). Apart from the more detailed information on connectivity pressures, we aimed further in improving the data on fish. It was a characteristic of the FAME data set that long-term impacts on migratory fish (both, diadromous as well as potamodromous) could not have been considered since even unimpaired or only minimally disturbed reference and calibration sites may have been impacted already by the absence of migratory fish due to migration barriers. In order to improve the information on fish our approach was therefore to reconstruct “reference conditions” for the distribution of migratory fish species based on historical data. Due to the enormous amount of work necessary for the preparation of historical data we limited our data collection to diadromous species, even if data sources and case studies are available for some potamodromous species from former projects performed especially in Austria, Germany and France. Hence, the objective of subtask “continuity disruption” in work-packages two (data collection) and three (data analyses and modelling) was to compile information on the historical distribution of diadromous fish species and to compute metrics which can be compared later on with the present situation. Due to the well known incompleteness of historical information (occurrence not registered; loss of records, log term human impacts on the occurrence of migratory fish etc.) we will also analyse a potential distribution of fish species. This is based on models of the historical presence of diadromous fish species as a factor of environmental characteristics. Due to the particularities of historical data special modelling techniques have to be used (see for this the subchapter on used methods). Further analyses will consider the existence of barriers if a species is absent at present or any other type of pressure that may impact the occurrence of a diadromous species.

4 2. Data

2.1. Selection of diadromous fish to be considered In a first step we selected the fish species to be considered for the data search. The prerequisites for species selection were a broad geographical distribution, (former) commercial interest leading to a better chance of exact species determination and to more frequent recordings in historical information sources (see below for the limits of historical data on fish). However, we also tried to take into account all important European catchments. Thus we also considered diadromous species endemic to the Danube catchment. Finally, we also agreed to select only fish species which are obligatory diadromous. This excluded fish species which are potamodromous or even resident fish in many catchments and undertake diadromous migrations in others. Such situations occur e.g. for fish species of the and its connected rivers (species such as Vimba vimba). Against this later rule we kept Acipenser nudiventris, Acipenser gueldensaedti and Acipenser naccari for data collection, since it seemed possible for these three Sturgeon species of the Danube catchment and Adriatic Sea, respectively, to distinguish between potamodromous and diadromous forms.

Based on these prerequisites we selected the following list of 17 species: Family /Species Lampreys: River lamprey (Lampetra fluviatilis) Sea lamprey (Petromyzon marinus) Sturgeons: European Atlantic sturgeon (A. sturio/A. oxyrinchus1) Adriatic sturgeon (A. naccari) Beluga (Huso huso) Stellate sturgeon (Acipenser stellatus) Russian sturgeon (Acipenser gueldenstädti) Ship sturgeon (Acipenser nudiventris) Shads: Allis shad (Alosa alosa) Twaite shad (Alosa fallax) Danube shad (Alosa immaculata) Salmonids: Atlantic salmon (Salmo salar) Sea trout (Salmo trutta trutta) Coregonids: Coregonus sp. diadr. (houting, c. whitefish) Eels: Eel (Anguilla Anguilla) Smelts: European smelt (Osmerus eperlanus) Flounders: Flounder (Platichthys flesus)

According to information about the distribution on native species, which was compiled for 400 European catchments during the FAME project (see Reyjol et al., 2007; trout was only considered on species and not on sub species level, Sea Trout was not represented in this dataset) the most frequent species is the Eel. It occurs in 316 of the 400 catchments and in 22 main river regions and marine areas, respectively. Also the Atlantic salmon has a wide biogeographical

1 Latest genetic studies show that historically also A. oxyrinchus occurred in Europe; however it is not possible to distinguish between the two species only based on written historical records; therefore the two species are mentioned in common

5 distribution. It is native in 155 catchments, however only in 15 main river regions or marine areas ranging from rivers of the Bay of Biscay to the Gulf of Riga. The two Lampreys (River and Sea Lamprey) as well as the two shad species (Alosa alosa and Alosa fallax) are less frequent in terms of catchments, but the have a broad biogeographical distribution ranging from the Aegean and Adriatic Sea (Alosa fallax) to the Gulf of Riga (all three species except Alosa alosa). The European Sturgeon occurs from the (however, rare and no longer migrations upstream to the Danube are recorded) to the Gulf of Riga and is the only Sturgeon species which is common over all European rivers. All other sturgeon species are endemic to the Danube catchment and/or to the Adriatic Sea (Huso huso, A. nudiventris, A. gueldenstaedti, A. naccari, A. stellatus). For the Danube catchment moreover an endemic diadromous shad species, the Danube Shad (Alosa immaculata/A. pontica), was included. The main focus of the data search was on these 14 species. Three further species were integrated, even if there was no broad and detailed knowledge about the availability and reliability of historical information. Diadromous Coregonids were integrated, even if it was quite unsure whether it will be possible to determine diadromous forms from residential or potamodromous in historical written records. Moreover, Flounder and European Smelt were selected.

2.2. Information sources for historical distribution Data which were used to define the historical distribution of the species mentioned above can be grouped into three different classes: 1. The Institute of Hydrobiology (BOKU Vienna), the Leibniz-Institute of Freshwater Ecology and Inland Fisheries (IGB Berlin), CEMAGREF and the Inland Fisheries Institute (IRS) based the data search on the collection of printed literature from the 19th and the first decades of the 20th century. As a consequence these data are not consistent. The long temporal period covered integrates also possible changes of species distribution due to a change of environmental conditions (e.g. long term climate change) but also impacts of migration barriers which started to be more numerous and severe in terms of continuum interruption in the 2nd half of the 19th century, as well as the risk for pollution and pressure on water quality. Also the possibility of fish stocking has to be respected, even if stocking may have occurred also before the 19th century. CEMAGREF and IRS limited the data search to France and , while IGB and BOKU made a European- wide literature search and integrated information an all possible rivers to the diadromous species database (thus covering also Switzerland and the whole Danube catchment including Romania; as far as possible also Italy). Data from Hungary were collected from printed literature from the first half of the 20th century. At this time an impact of weirs seems very likely, however data and information on the erection of dams were not compiled. 2. For Portugal and Spain the Institute of Agronomy, Lisbon and the University of Madrid based their search for historical data on fish distribution on systematic archival primary sources from the late 18th and first half of the 19th century. These data sources were topographical and socioeconomic descriptions of the country, including information on fish as well as cadastral inventories including also information on fish. Portugal performed the data collection together with the Aberta University, Lisbon.

6 3. For Finland, Sweden, United Kingdom, Lithuania and The Netherlands present literature on the historical distribution of fish was used as main information source. Even if this information can be considered as reliable as the basic documents it was split into an extra data group since we do not have a complete overview on the historical basic data sources.

2.2.1. Using historical data on fish species distribution Historical data were collected based on written documents and fish species maps. They have in comparison with present data several particularities, which must be accounted for. First, historical information has to be checked in terms of reliability of the data. Even, if we decided to focus on well known and commercially exploited species, there is the problem of misclassifications. In our search for diadromous species this can be demonstrated well e.g. for the Danube catchment. Here the occurrence of river lamprey was recorded in several publications from the 19th century. At this time the biological knowledge Lampreys was still very vague and the life cycle of the species was not well investigated. Later on, for the Danube catchment singular Lamprey species have been determined for the Danube as it is also the case for Sturgeon and Shad species. However, Lampreys occurring natively in the Danube basin have been described as single species only in the 20th century (Eudontomyzon danfordi Regan, 1911; Eudontomyzon mariae (Berg, 1931); Eudontomyzon vladykovi Oliva & Zanandrea, 1959) and they are not diadromous. Therefore, all information on the occurrence of Lamprey species in the Danube and it´s tributaries was omitted. The same was the case for information on the occurrence of Allis shad and Twaite Shad in the Danube. However, in such case we registered the indicated locations and rivers section as basis for comparison with the modelled potential distribution of the Black Sea Shad A. immaculata, since all are diadromous species and historical information about the occurrence of Allis or Twaite Shad may have been a misclassification of the Danube Shad. Another problem of historical written documents is that information is often general on a genus level. This can be demonstrated for the distribution also for Sturgeon species in the Danube catchment, for which many authors refer often only to “Stör”, which is in this case the general term for the genus (and not the German name for A. sturio). If there was no additional information allowing a definite determination of one of the Danube sturgeon species (Beluga, Stellate, Ship and Russian sturgeon) such kind of information was also excluded from the database. While it is quite simple to identify misclassifications when it relates to a species outside of it´s biogeographical range of distribution, it is usually more difficult to detect incorrect information when it refers to a location within the range of a species´ distribution. In such cases expert judgement was used to identify doubtful information and to exclude according data. This appeared e.g. for the information on the occurrence of Allis Shad in the Rhone catchment. Nevertheless, it can be concluded that in contrast to the use and interpretation of historical fish data on e.g. Cyprinids the risk for misclassification for diadromous species may be lower due to their commercial interest (was one of the prerequisites for species selection) and the generally quite good knowledge about the species (at least in comparison to other fish). Also, the distinction of Salmo salar and Salmo trutta is in written documents not always clear. While the distribution of Salmon could have been reconstructed quite well also due to the existence of precise maps, Sea Trout could have sometimes not be considered, like e.g. in France due to the uncertainty of data.

7 Finally, only very general information for species occurrence on river level without any further local precision was not integrated but these data were registered to compare the modelled potential distribution with such indications. Finally, it has to be stressed that historical fish distribution data based for a large part on sources from the 2nd half of the 19th century and even on data from the first decades of the 20th century are not showing a “pristine” or “natural” fish community. Pressures on the occurrence of diadromous fish species are already recorded for the later Middle Ages, e.g. as a result of large number of weirs for energy production in many rivers. While a complete interruption of migration may have been more exceptional at this period, river regulation works, deteriorating water quality and overfishing became a more frequent phenomenon from the 18th and especially from the 19th century onwards (see e.g. Hoffmann, 1996, 1999, for central Europe, Balon, 1968, for the Beluga Sturgeon in Hungary or Lenders, 2003, for the Rhine). As a consequence the migration distances of diadromous fish may have been already reduced in the 19th century, although traditional weirs were often constructed in a way that they enabled fish migration, not at least due to the importance of fisheries as commercial factor. Another reason for possible changes in the distribution of diadromous fish in the 19th century is stocking of species and transfer of fish between different catchments or sub-catchments via navigation channels. This fact has to be considered e.g. for Eel or Salmon, for which written historical source refer regularly to stocking or to the interruption of migration routes through weirs. The advantage for our data search was that these pressures were recognized and discussed intensely and in many cases historical sources refer to previous migration routes which enabled us to compensate the pressures at least for a part. Finally, we have to take into account that historical climate changes affected the distribution of migratory fish. In order to account for these facts we will use air temperature data from the beginning of the 20th century for modelling of the potential distribution (see Mitchell, et al., 2004).

2.3. Collection of data on present distribution Information on the present occurrence of the selected diadromous species was accounted for on the level of existing sampling sites and integrated directly during national data collection. The EFI+ partners used for this information not only the results from the sampling but also additional data from other samples/sampling methods and commercial fisheries. For the present distribution four modalities were possible: species is absent, species occurs at present naturally, species occurs at present mainly due to stocking, no data/status unknown.

3. Methods

3.1. Data collection Historical data were collected based on written sources or fish distribution maps. Information was registered in two different ways. Some partners (BOKU, CEMA, IGB, ISA, HU) compiled data completely independent from existing sampling sites and the data input sheets for the central data base. Information was registered into data files (mostly Excel). Even, if we used directly only information which refers to a precise location (such as e.g. “Beluga Sturgeon occurs in the Morawa river up to Rabensburg”), we recorded also information referring to the presence of a

8 species in a river in general (e.g. “Salmon occurs in river Aare”) for later validation of the potential distribution of modelled species. Precise information on locations or reaches was after geo-referenced to national river networks. Other partners were transferring the compiled historical information directly to the central database. Here is was possible to indicate the historical distribution on site scale in table “Diadromous species”. Finally, all data were integrated to a separate diadromous species database and into one common GIS-document. For the later all data geo-referenced on national river networks or Google Earth were adjusted to the European CCM river network in order to have a common basis for environmental variables. The results of this step were the historical distribution maps of the selected diadromous fish species. As it can be seen in the historical distribution maps in chapter four there are gaps in the distribution of fish species. This is on the one hand because not all fish species could have been collected by all partners due to very limited time and financial resources. On the other hand it is also well known that historical sources for the distribution of species (this is true for all species, not only for fish) are more or less incomplete. The advantage of this data collection on diadromous species was, that the relevant species are well known and of commercial interest. Thus, the distribution is probably at least for those areas for which maps or comprehensive written documents exist quite complete. Nevertheless, especially the distribution in smaller rivers is probably underestimated. Due to the data gaps we have to assume for the historical distribution we will in a next step model the potential distribution of fish species. For this procedure we will develop models which explain the presence of diadromous fish as a factor of environmental variables. The selection of environmental variables had to be based on the availability of data available on a European scale. As for the central database of the EFI+ project we used also for geo-referencing historical fish species the CCM river network. It was in particular possible to extract from version 2 of the CCM some important environmental variables, such as altitude at the upper and lower end of a segment, slope, distance to sea, drainage area or stream order (Strahler). We further integrated mean monthly air temperature from the period 1901-1921. These data were provided by the Tyndall Climate Research centre (see Mitchell, et al., 2004). In order to account for regional effects of fish species distribution, we integrated the catchment, the eco-regions after Illies & Botosaneanue (1963) as well as the country code. It was not possible to obtain data on natural barriers on European scale and consider them in our data set. Thus it might appear that a potential distribution of species will be predicted for segments upstream of natural barriers. In such cases the validation of the national data providers is the only possibility to exclude concerned river segments after modelling and preparing maps for the potential distribution. In terms of modelling the potential distribution we have to take into account particularities of historical data in comparison to present fish samples. For present sampling it can be assumed in principle that a species is absent form a particular river section when it is not represented in a sample (of course depending on sampling method and period). This is not at all the case for historical data. Here, one can only say that according to historical information (and after validation of the reliability of the data source) a species was present at a particular site. For all other river segments, there is no knowledge whether the species was really absent (“true

9 absence”) or whether it was present but there is no historical record about this (false or pseudo- absence). Thus we will have to work with particular modelling techniques, which are especially adapted to the treat “presence-only” data (see below, chapter 5).

4. Results – historical distribution An large amount of records on the historical distribution of the selected diadromous fish could have been compiled. This was mainly possible due to additional funds and co-operations some EFI+ partners were able to establish and organise, respectively. Depending on the possibility to obtain further support the regional coverage of historical records differs between the countries. Also, partners had to focus on those species for which historical data were easy to find and analyse. A general search for all species over all European countries was done by the University of Natural Resources and Applied Life Sciences and the Leibnitz Institute for Freshwater Ecology, who analysed printed material mainly from the late 18th to the first half of the 20th century. Cemagref and the Inland Fisheries Institute did a literature survey for all species for France and Poland, respectively. This resulted in general in a good coverage of all species in these four countries. In order to limit financial resources needed France did not compile the numerous materials on the historical distribution of Eel. Since there is a good coverage for Eel in other areas it was decided to use later on the modelled potential distribution after verification of plausibility. Portugal and Spain analysed archive material resulting in a good overview especially on Eel, and revealing also sufficient data on the two Lampreys, on Shad species, Sturgeon, Salmon and Sea trout. Hungary, Romania and Italy provided information on the mostly endemic diadromous fish species of the Danube catchment as well as for the Adriatic Sea. The Netherlands and Lithuania analysed present information about the historical occurrence of all species concerned. For Sweden, Finland and UK data search had to be limited to literature about particular species. UK was especially focusing on Eel. Further literature on Shads as well as on Lampreys will be used for validation of the potential distribution. Sweden and Finland focused mainly on literature about the historical presence of Salmon and Sea Trout. As already indicated before, literature information had to be validated in terms of reliability of information. Even, if there are some historical sources reporting the occurrence of Eel in the Middle and Upper Danube the species was finally excluded from the Danube catchment except for the lowest part close to the Black Sea due to the contrasting information in historical literature (see especially Siebold, 1863, but also Heckel & Kner, 1858, v. d. Borne, 1883, Wittmack, 1875 for the discussion of native occurrence of Eel in the Danube). The number of records found for the different species in historical sources and the number of resulting segments in the CCM river network is indicated in the table 1. Most written evidence was found for Eel, Salmon and Sea trout. The number of records is also remarkable for the two diadromous Lamprey species, for Allis Shad as well as the Atlantic Sturgeon. A lower number but probably still sufficient for modelling is available for Twaite Shad (well distributed in terms of river size, see below). A surprisingly high number of records was found for Flounder (Platichthys flesus), an estuarine species for which historically larger migrations to Rhine, Loire or Elbe are recorded.

10 As expected the number of records is limited for the Sturgeon species of the Danube and the Danube Shad as well. It will be tested if modelling these species separately is useful or whether the historical distribution is already quite complete. As it is shown it table 1 the number of segments is too low for more detailed analysis of Coregonidae and the Adriatic sturgeon. For the Po catchment in particular further solutions to account for diadromous species will be discussed.

Table 1: Number of historical records and related CCM segments for the historical distribution of diadromous fish species

On the following pages the historical distribution maps for different species are shown. It has to be emphasised that these maps do not reflect the complete historical distribution but the distribution as we have been able to identify it based on historical documents. Data gaps are for a big part the result of the incompleteness of historical data. However, it has to be mentioned in this respect that migration barriers and other pressures like habitat degradation and water pollution are in some catchments and rivers already existing on a longer term and gaps may be also a result of existing pressures. This has to be considered in later analyses, and discussed depending on the results of tests of metrics sensitivity in WP four. To get a more complete picture and to compensate missing historical information and possible effects of historical pressures we will afterwards model the potential distribution of fish species as a factor of environmental variables. As it was already mentioned, we considered for the historical distribution only references with a precise location or river section for the occurrence of a species. General indications that a species occurs in a particular river were registered but they are not shown in the maps. Nevertheless, general information will be used finally to validate modelling of the potential distribution.

11

Figure: Identified written records about the historical distribution of the European Atlantic Sturgeon

12

Figure: Identified written records on the historical distribution of Sturgeon Species in the Danube catchment

13

Figure: Identified written records on the historical distribution of the Eel

14

Figure: Identified written records on the historical distribution of the Sea Lamprey (P. marinus) and the River Lamprey (L. fluviatilis)

15

Figure: Identified written records on the historical distribution of the Atlantic Salmon (S. salar) and the Sea trout (S. trutta trutta);

16

Figure: Identified written records on the historical distribution of Shad species (Allis Shad A. alosa; Twaite Shad A. fallax; Danube Shad A. immaculata);

17 Since historical information does not reflect the complete picture of the former distribution of a fish species it is not possible to make any straight biological or ecological conclusions about the distribution of fish in terms of environmental conditions. However some descriptive comparisons were done, to clarify whether we will have a good representation of datasets for modelling. As it turned out the original concern that we will have not enough information for smaller rivers due to the fact that historical evidence is more frequent for large and medium sized rivers was not correct. Table 2 shows the frequency of fish species in terms of stream order. For the most frequent species in the dataset we have a good representation of smaller rivers even for stream order 2. This concerns in particular Eel, Salmon and Sea trout. Some segments in such river sections are also available for the two Lamprey and Shad species (n. b. that records for Atlantic sturgeon are related to small coastal rivers).

Table 2: Number of river segments per Strahler stream order for 15 diadromous species

5. Modelling of the potential distribution Mapping species distribution with historical data induced specific problems for the preparation of modelling routines. Indeed, generally, we have only the information on the species presences and the observations are spatially structured (e. g. Legendre et al. 2002, Schabenberger & Gotway 2005, Dormann et al. 2007). In the literature, we found several methods to estimate the potential species distribution such as generalized linear model (GLM, Nelder & Wedderburn 1972, McCullagh & Nelder 1989), generalised additive model (GAM, Hastie & Tibshirani 1989, Guisan & Zimmermann 2000), Ecological Niche Factorial Analysis (ENFA, Hirzel et al. 2002). The use of probabilistic models (GLM, GAM with binomial distribution) requires the establishment of pseudo-absence to complete the datasets (e.g. Pearce & Boyce 2005, Lütolf et al. 2006) and the integration of spatial dependence which is a complex step (e.g. Dormann et al. 2007, Keitt et al. 2002, Miller et al. 2007). For example, several authors proposed some corrections/penalization of likelihood to take into account the potential misclassification (e.g. Lancaster & Imbens 1996, Pearce & Boyce 2005). However, all these approaches are in experimental step and involve important statistical developments and programming. ENFA computes uncorrelated factors that explain the major part of the ecological distribution of the species: The first factor is the marginality factor, which describes how far the species

18 optimum is from the mean habitat in the study area. The others factors (specialisation factors) are sorted by decreasing amount of explained variance (Hirzel et al. 2002). They describe how specialised the species is by reference to the available range of habitat in the study area. The routines for modelling the potential distribution of diadromous species have been established and first tests are relatively encouraging and the necessary adaptations to networks are moderate. ENFA does not provide true probabilities of presence, but it is possible to compute indices based on the position of the niche defined by the analysis within the multidimensional space of environmental variables (Calenge 2007). As a conclusion, this approach seems more adapted to solve our question.

6. References Balon, E. (1968): Einfluss des Fischfangs auf die Fischgemeinschaften der Donau. Arch. Hydrobiol./Suppl. Suppl. 34, 3. 228-249 Calenge C. (2007) Exploring Habitat Selection by Wildlife with adehabitat, Journal of statistical Software, 22(6), 1-19, URL: http://www.jstatsoft.org/v22/i06. Dormann C. F., McPherson J. M., Araújo M. B., Bivand R., Bolliger J., Carl G., Davies R. G., Hirzel A., Jetz W., Kissling W. D., Kühn I., Ohlemüller R., Peres-Neto R. P., Reineking B., Schröder B., Schurr F. M. & Wilson R. (2007) Methods to account for spatial autocorrelation in the analysis of species distributional data: a review, Ecography, 30, 609-628. Graham, C.H. & Hijmans, R.J. (2006) A comparison of methods for mapping species ranges and species richness. Global Ecology and Biogeography, 15, 578-587. Guisan, A. & Zimmermann, N.E. (2000) Predictive habitat distribution models in ecology. Ecological Modelling, 135, 147-186. Hastie, Trevor, and Tibshirani, R. (1989) Generalized Additive Models (with discussion), Statistical Science, 1(3) , 297-318. Heckel, J. and Kner R. (1858): Die Süßwasserfische der österreichischen Monarchie mit Rücksicht auf die angränzenden Länder. Leipzig. Hirzel, A.H., Hausser, J., Chessel, D. & Perrin, N. (2002) Ecological-niche factor analysis: How to compute habitat-suitability maps without absence data?, Ecology, 83, 2027-2036. Hoffmann, R. C., Economic development and aquatic ecosystem in medieval Europe. The American historical review 101 (1996) 631-668 Hoffmann, R. C., Fish and Man: Changing relations in medieval Central Europe. Beiträge zur Mittelalterarchäologie in Österreich 15 (1999) 187-195 Keitt et al., 2002 T.H. Keitt, O.N. Bjornstad, P.M. Dixon and S. Citron-Pousty, Accounting for spatial pattern when modeling organism–environment interactions, Ecography 25 (2002), pp. 616–625. Lancaster, T. & Imbens, G. (1996). Case-control studies with contaminated controls, Journal of Econometrics, 71(1-2), 145-160. Legendre P., Dale M. R. T., Fortin, M.-J., Gurevitch J., Hohn M. & Myers D. (2002) The consequences of spatial structure for the design and analysis of ecological study, Ecography, 25, 601-615.

19 Lenders, R. (2003): Environmental rehabilitation of the river landscape in the Netherlands. A blend of five dimensions. phD-Thesis, University of Nijmegen. Lütolf, M., Kienast, F. & Guisan, A. (2006) The ghost of past species occurrences: improving species distribution models for presence-only data. Journal of Applied Ecology, 43, 802–815. McCullagh, P. & Nelder, J.A. (1989) Generalized Linear Models, second edition edn. Chapman & Hall/CRC, London. Miller J., Franklin J. & Aspinall R (2007) Incorporating spatial dependence in predictive vegetation models, Ecological Modelling, 202 (3-4), 225-242. Mitchell, T., Carter T., Jones P., Hulme M. & New M. (2004): A comprehensive set of high- resolution grids of monthly climate for Europe and the globe: the observed record (1901- 2000) and 16 scenarios (2001-2100). Working paper 55. Tyndall Centre for Climate Change Research. Nelder, J; & Wedderburn R. (1972). "Generalized Linear Models". Journal of the Royal Statistical Society. Series A (General), 135, 370-384. Pearce, J. L. & Boyce, M.S. (2006) Modelling distribution and abundance with presence-only data, Journal of Applied Ecology, 43, 405-412. Schabenberger, O. & Gotway, C.A. (2005) Statistical methods: for spatial data analysis Chapman & Hall/CRC, Florida. Siebold, C. (1863): Die Süßwasserfische von Mitteleuropa. Leipzig. Wittmack, L. (1875): Beiträge zur Fischereistatistik des Deutschen Reichs sowie eines Theiles von Oesterreich-Ungarn und der Schweiz, im Auftrage des Deutschen Fischereivereins. Berlin.

20

Project no. 044096

Project acronym EFI+

Project title IMPROVEMENT AND SPATIAL EXTENSION OF THE EUROPEAN FISH INDEX

Instrument: Specific Targeted Project

Thematic Priority: Integrating and strengthening the European Research Area

Central and Eastern European Rivers

Period covered: from 01.01.2007 to 31.12.2007 Date of preparation: 10.02.2008 Start date of project: 01.01.2007 Duration: 2 years

Project coordinator name: Dr. Stefan Schmutz Project coordinator organization name: University of Natural Resources and Applied Life Sciences

Persons responsible for the report: Klaus Battes and Karina Battes, Bacau University, Romania

21

TABLE OF CONTENTS: 1. General overview on the river basins included in the EFI+ project from each country page 02 1.1. General data concerning the catchment areas from each country page 02 1.2. Characteristics of fish communities from the considered catchment areas page 09 2. Main pressures in Central and Eastern European Rivers page 12 3. General analysis of pressures page 17

1. General overview on the river basins included in the EFI+ project from each country 1.1. General data concerning the catchment areas from each country

LITHUANIA: Nemunas River basin covers more than 72% of Lithuanian territory. The rest ~28% are shared between 5 river basins (Table 1). Small rivers (L – 3-10 km) predominate, covering ~ 50% of total river length ~83% of total number. There are only 17 rivers longer than 100 km, 13 of them are in the Nemunas River basin.

Table 1 River basins, number of rivers of different length (N) and their total length (∑L, km) Area in L = 3-10 L = 10.1-30 L = 30.1-60 L = 60.1-100 L >100 Total River basin Lithuania N ∑L N ∑L N ∑L N ∑L N ∑L N ∑L km2 Nemunas 46695 2377 12169 418 6535 60 2490 20 1405 13 2548 2888 25147 Coastal rivers 2132 126 628 30 466 3 139 2 131 - - 161 1364 Venta 5140 365 1921 67 1047 5 195 2 190 1 161 440 3514 Lielupe 8939 669 3585 109 1801 15 562 4 309 3 445 1245 6702 Daugava 1857 105 519 17 229 3 141 - - - - 125 889 Prieglius 65 4 20 - 4 20 Total 64828 3646 18842 641 10078 86 3527 28 2035 17 3154 4418 37636

The total length of the largest Nemunas River is 937 km, and the basin area constitutes 97863.5 km2 while the Lithuanian part of the basin covers an area of 46 695.4 km2. Its drainage area is located in the territories of Byelorussia, Lithuania, Russia (Kaliningrad region), Latvia and Poland. The total water surface area of the Nemunas equals 157.9 km2. The Nemunas River ends up in the Curonian Lagoon, a freshwater coastal lagoon of the southeastern Baltic Sea. The largest tributaries in the Lithuanian part of Nemunas river basin are Neris, Nevezis, Merkys, Dubysa, Sesupe, Jura and Minija (Table 2).

Table 2 Characteristics of the main rivers of the Nemunas river basin Length, km Catchment area, km² River Total In Lithuania Total In Lithuania Merkys 203 185.2 4415.7 3781 Neris 509.5 228 24942.3 13849.6 Dubysa 139 139 2033 2033 Sesupe 297.6 157.5 6104.8 4899 Jura 171.8 171.8 3994.4 3994.4 Nevezis 208.6 208.6 6140.5 6140.5 Minija 201.8 201.8 2942.1 2942.1 Sventoji 246 246 6889 6801 Zeimena 79.6 79.6 2792.7 2792.7

In the Nemunas river drainage area the soils are mainly represented by moraine and sandy loams, which in some places are covered with peat formations. Loams prevail in the northwest of the basin, covering almost completely the basins of the rivers Minija, Jura, Dubysa and Nevezis. Sand and sandy loams cover significant areas in the upper part of the basin and in the littoral region

22 (the Jura River Basin, eastern part of the Dubysa River Basin and part of the Minija River Basin). Sandy-mud deposits occur in river valleys and cover the whole mouth of the Nemunas. In the southeast of the Nemunas catchment area more sandy grounds dominate, which are porous and permeable. Here the rain and snow waters are absorbed fast. Average annual runoff modules in this region are 7 – 9 l/s/km2. Climate is changing between maritime and continental in the Nemunas river basin as in the whole Lithuania. The mean annual temperature is +6 – +7 0C, the average in January being -4.9oC and in July +17oC. Lithuania is in the zone of surplus of humidity. Average precipitation varies from 600 to 700 mm. Rainfall contains 75 % of the total precipitation. Evaporation contains 65 % approximately, and surface runoff – about 32 %. The long-term average annual discharge to the Curonian Lagoon by the Nemunas river is 667 m3/s which equals to 21.1 km3. The Curonian Lagoon discharge to the Baltic Sea comprises approximately 23 km3 per annum. For most of the Lithuanian rivers big spring floods are typical while in summer and winter the flow of rivers is significantly reduced. In the western part of the Nemunas river basin the highest water runoff coefficients of 0.40 – 0.55 dominate, that is, about 40-50 % of annual precipitation flows down to the rivers. Average annual runoff modules of 9 – 14 l/s/km2 are the highest in all the country. The distribution of the river run-off throughout the year is soundly impacted by flood events. Spring floods in the rivers of this region usually start in the middle of May and last for 25 – 40 days. During spring floods approximately 30 % of annual runoff flows down, while in summer – 10 % and in autumn-winter season – 60 %. Minimal winter runoff is approximately 3 times higher than the summer one. The division of annual runoff in mid region of the catchment is mostly uneven. The rivers in the mid of the river basin are subject to a less abundant runoff than in other regions. Average annual runoff modules there amount for 4.5 – 6 l/s/km2. The general discharge of the river contains 33-40 % of snowmelt water, 32-25 % of rain runoff, and 10-40 % of groundwater. Most of the discharge of the river basin comes from combined surface/sub-surface runoff, including snowmelt water (in average by 40 %) that cannot infiltrate into deeper layers because of the frozen soil.

POLAND: There are three main river basins in the Baltic Sea drainage area in Poland: - The River Basin - total area 194.4, from which in Poland – 168.7 thousands km2 (56% of Poland area), main tributaries: Skawa, Raba, , Wisłoka, , , and , Drweca, Przemsza, , , , ), - The River Basin – area: 118.9 thousands km2, in Poland 106.0 thousands km2 (34%), main tributaries: Barycz, , Bobr, Nysa Klodzka, Nysa Luzycka), - The Pomeranian Rivers - flowing directly to the Baltic Sea, area: 28.2 thousands km2 (9%), main rivers: , Parseta, , Slupia, Pasleka). - The remaining 1% of Poland area contains small parts of Nemunas River Basin (Baltic Sea drainage area) in North-East Poland, few small rivers of Danube Basin and Dnister Basin (Black Sea drainage area) in South Poland and small part of Elbe River basin (North Sea drainage area) in West Poland. The sampling sites (919) considered for the EFI+ project were located in 4 groups in various regions of Poland: North Poland - Pomeranian Rivers (551 sites), East Poland - Narew R. with tributaries (50 sites), Central Poland - middle course of the Vistula River with tributaries (111sites), South Poland - upper Vistula R. tributaries (116 sites) and West Poland - middle Oder R. tributaries (91 sites) (see Fig. 1) General environmental characteristics of the sites located in particular regions are presented in Table 3. Most sites are located in Ecoregion No 14 (Central Plains) and 16 (Eastern Plains), some sites are also situated in Ecoregion No 10 (The Carpathians). Climatic conditions are differentiated between regions, with privilege of Atlantic climate in North Poland, severe, continental Climate in East Poland, some mixture of this two in Central Poland, mountaneous Sub-Carpathian climate in South Poland and mild, Atlantic climate in West Poland. The difference in duration of vegetation season between West and East Poland reaches up to100 days. All sites, except one, are located in

23 Baltic Sea drainage area, only the site on Strwiaz R. belongs to Black Sea drainage area. As much as 60% of sites is located in Pomeranian Rivers catchments, 27% - in Vistula R. catchment and 13% in Oder R. catchment.

Fig. 1 The location of the 919 sampling sites considered for the EFI+ project

Table 3 Environmental characteristics of the sites available in Polish dataset, with respect to particular regions: North Poland, East Poland, Central Poland, South Poland and West Poland Distance Geological No. No. No. Size of No. Catchment from Altitude River slope typology - large Region of of Ecoregion catchment oxbow name source m a.s.l. ‰ organic river sites rivers km2 sites km % sites sites Vistula, 3-2960 0.1-25.0 North 551 215 14, 16 1-180 0-75 0 0 0 Baltic Sea median 88 median 1.6 137-15300 0.07-1.8 East 50 9 16 Vistula 20-284 95-119 96 19 23 median 2910 median 0.2 Vistula, 12-195200 0.1-11.0 Central 111 19 14, 16 3-988 9-316 0 32 4 Baltic Sea median 490 median 0.8 Vistula, 4-960 0.6-60.8 South 91 40 10, 16 2-83 185-843 0 0 0 Dnister median 54 median 7.0 8-5370 0.2-44.1 West 116 47 14 Oder 3-223 60-630 0 0 0 median 131 median 2.4 Vistula, Baltic Sea, 3-195200 0.07-60.8 Poland 919 330 10, 14, 16 1-988 0-843 5 51 27 Oder, median 117 median 1.7 Dnister

Largest rivers were sampled in Eastern and Central Poland (median size of catchment 2910 and 490 km2 respectively), medium-size rivers dominated in West Poland (131 km2), while sites located on small rivers prevailed in South and North Poland (54 and 88 km2 respectively). The altitude gradient in whole dataset was high (0-843 m.a.s.l.), with low altitudes in North, and East Poland (< 120 m.a.s.l.), medium - in Central Poland (up to 316 m.a.s.l.) and highest in South Poland (up. to 843 m.a.s.l). Considerable altitude gradient was represented by sites from West Poland (60- 630 m.a.s.l.). Slope values were diversified from less than 0.1 ‰ to over 60 ‰. Lowest median slope values were found in Eastern and Central Poland (0.2 and 0.8‰ respectively), while highest - in South and West Poland (7.0 and 2.4‰ respectively). Despite low absolute elevation (≤ 75 m.a.s.l.) Pomeranian Rivers showed quite high slope values (median 1. 6‰, maximal value 25‰).

24 This explains some similarities between fish communities in this region and mountain ones. 96 % of sites in Eastern Poland were located on rivers of organic geological typology, but such rivers occurred only in this region. In Central and Eastern Poland 51 sites located on large rivers were sampled, including 26 sites on middle and lower Vistula R., as well as 27 sites in oxbow lakes of Vistula, Narew and Biebrza rivers. Considerable differentiation of governing environmental factors, like: climate, drainage basin, altitude, actual river slope and geological typology between groups of sites located in particular regions of Poland causes need of separate analysis of each region in terms of fish community structure and pressures assessment.

HUNGARY The data considered for the EFI+ project were part of the results of the ECOSURV project (Ecological Survey of Surface waters, Hungary) that was carried out in 2005. The information was collected in 2005 from April to September and the main goals of the project for fish were: completion of sampling manual, drafting a recommendation for the future monitoring manual, based on field experiences; standardised sampling in 234 waterbodies; establish a fish database; estimate the ecological status of the sampled waterbodies using the EFI, and examine the validation of the method; statistical analysis to validate the water typology. From the total number of 234 sampling sites included in the ECOSURV project, 193 rivers (running water) sites were added to the EFI+ project. The sampling sites well represent the Hungarian river types. According to Table 4, there are 8 types of waters based on cluster analysis (Halasi-Kovács & Tóthmérész, 2006), as follows: 1. Middle-height mountainous streams: 16% 2. Lower hills streams and small rivers: 23% 3. Gravel bottomed epipotamal section of medium and large rivers of major slope: 9% 4. Sand bottomed epipotamal section of medium and large rivers of minor slope: 10% 5. Lowland small streams and brooks: 18% 6. Lowland small and medium rivers and channels: 7% 7. Lowland (metapotamal) section of large rivers: 9% 8. River Danube: 8%

Table 4 The main characteristics at the sites considered for the EFI+ in Hungary Water discharge Actual water velocity Altitude Dominating substrate (m3/s) (cm/s) No. Perc. No. Perc. No. Perc. No. Perc. Cat. Cat. Cat. Cat. (pcs) (%) (pcs) (%) (pcs) (%) (pcs) (%) >150 48 24,9 <1 99 51,3 0-5 21 10,9 stone 3 1,6 150- 96 49,7 10 39 20,2 6-35 67 34,7 pebble 20 10,4 100 <100 49 25,4 100 29 15,0 36-75 69 35,8 pebble/sand 47 24,3 1000 19 9,9 76-100 29 15,0 sand 25 13,0 >1000 7 3,6 >100 7 3,6 sand/clay 18 9,3 clay 33 17,1 organic 47 24,3 sediment

Two sampling methods were used (Halasi-Kovács et al, 2005): wading, back pack electric sampling equipment, for wadable running waters and sampling from boat, generate electric sampling equipment, for larger rivers.

ROMANIA The lower Danube catchment area includes several ecoregions, as follows: ecoregion no. 10 – The Carpathians; ecoregion no. 12 – The Pontic Plain and ecoregion no. 16 – The Eastern Plain.

25 Table 5 presents the main river courses in Romania and Bulgaria with an area of the catchment that exceeds 8000 km2.

Table 5 Physical and geographical data concerning the rivers from the lower Danube catchment area Length of the Area of the Catchment Discharge Altitude catchment catchment Comments Ecoregions area (mc/s) (km) (km2) Headwater Mouth Ukraine The Prut 252.9 2839.6 81.6 - 2 Republic of Moldavia Ecoregion no. 16 - The Siret 726 43933 254.0 1238 2 Romania The Eastern The Ialomita 414 10430 38.8 2395 8 PlainEcoregion The Arges 339.6 12521 64.0 2030 10 Romania no. 12 - The Jiu 348 10070 94.0 1760 22 The Pontic Plain The Iskei 410 - 57.0 - - Bulgaria The Yantra 285.5 7862 42.0 1340 10

The Eastern region (Ecoregion no. 16) includes the largest river basins (The Siret and The Prut). In the South (ecoregion no. 12), besides the Olt River, all running water catchment areas do not exceed 10000 km2. The Olt River however represents an exception, the river regularization regions exceeding 60-70% from its catchment area. Table 6 depicts the river types included in each ecoregion, next to information regarding the areas, slopes, altitude, annual mean flow etc., together with the main fish species characteristic to every region.

For the EFI+ project only sampling sites located on the Siret catchment area were considered – a total number of 263 sites (see Fig. 2)

Fig. 2 The considered sampling sites located in the Siret catchment area, Romania

6 26 Table 6 The running water types characteristic to Romania Parameters

Area Lithological Slope Altitude Precipitations Temperature q q95% Potential fish Type Symbol Geology km2 structure ‰ m a.s.l. mm/year 0C l/s/km2 l/s/km2 fauna Ecoregion 10 – The Carpathians blocs, Trout 10-1000 a- siliceous Mountainous water course RO01 boulders, 40-200 >800 700-1400 -2+8 >20 >1 Grayling b - calcareous pebble Undermouth Water course from the high plateau 10-1000 a - siliceous boulders, RO02 20-50 500-800 600-800 7-9 5-20 0.5-2 Chub regions b- calcareous pebble Trout a- siliceous Grayling Water course sector from the high pebble, RO03 1000-10000 b- calcareous 3-20 500-800 600-800 7-9 5-20 1-3 Undermouth plateau regions boulders

Water course sector from hilly or a- siliceous sand, Undermouth plateau regions RO04 1000-10000 b- calcareous 0.5 - 5 200-500 500-700 8-10 3-15 0.4 -2 pebble Barbel c-organic a-siliceous sand, Chub Water course sector from RO05 10-1000 b-calcareous pebble, 1-3 500-800 600-800 7-9 3-20 0.2-2 Undermouth depression regions c-organic boulders sand, Water course sector with wet lands a-siliceous Undermouth RO06 1000-10000 pebble, 1-2 200-500 500-700 8-10 3-15 0.4 -2 from hilly and plateau regions. c-organic Barbel silt Sub-ecoregion no. 10- The Transylvanian Plateau Water course from hilly or plateau a- siliceous Chub 10-1000 sand, regions RO07 b - calcareous 5-30 200-500 500-700 8-10 2-10 0.2 - 0.8 Undermouth pebble c - organic Water course sector from hilly or a-siliceous pebble, Chub plateau regions RO08 1000-10000 b- calcareous 3-20 200-500 500-700 8-10 2-10 0.2-0.8 boulders Undermouth

Ecoregion no. 11- The Hungarian Plain Water course from hilly or plateau a- siliceous Chub sand, regions RO09 10-1000 b - calcareous 5-20 200-500 500-700 8-10 2-8 0.2-0.6 Undermouth pebble c- organic a -siliceous Chub 10-1000 sand, Water course from plain regions RO10 b -calcareous <8 <200 400-500 9-11 <3 <0.3 Undermouth silt c- organic a-siliceous sand, Barbel Water course sector from plain RO11 >2000 b- calcareous silt, <1 200-250 400-600 9-11 2-10 0.1-1 Carp regions c-organic clay a-siliceous sand, Water course sector with wet lands Undermouth RO12 1000-10000 b- calcareous pebble, 0.5 - 5 200-500 500-700 8-10 3-15 0.4 -2 from hilly and plateau regions. Barbel c-organic silt a-siliceous sand, Barbel Water course sector with wet lands RO13 >10000 b- calcareous silt, <1 <200 400-600 9-11 2-10 0.1-1 Carp from plain regions c-organic clay Ecoregion no. 12- The Pontic Region Water course from hilly or plateau a- siliceous Chub sand, regions RO14 10-1000 b -calcareous 5-20 200-500 500-700 8-10 2-5 0.2-0.4 Undermouth pebble c- organic

27 Table 6 (continued) a -siliceous Water course from plain regions sand, muddy Chub RO15 10-2000 b- calcareous <8 <200 400-600 9-11 <2 0.2 clay Perch c-organic Water course sector from hilly or a-siliceous sand, Undermouth plateau regions RO16 1000-10000 b- calcareous 0.5 - 5 200-500 500-700 8-10 3-15 0.4 -2 pebble Barbel c-organic Water course sector from plain a-siliceous sand, Barbel regions RO17 >2000 b- calcareous silt, <1 <200 400-600 9-11 2-10 0.1-1 Carp c-organic clay a-siliceous sand, Barbel Water course sector with wet lands RO18 >10000 b- calcareous silt, <1 <200 400-600 9-11 2-10 0.1-1 Carp from plain regions. c-organic clay The Danube - Cazane 570.900- sand, pebble, Carp RO19 calcareous 0.07 100-200 600-800 8-10 9 3 574.850 boulders The low River Danube - 574.000- sand, clay, RO20 siliceous 0.05 5-70 500-600 9-11 8 2 Carp* Cazane - Calarasi 698.000 pebble 698.00- The Danube - Calarasi-Isaccea RO21 siliceous sand, clay 0.04 5 400-500 9-11 7 1.5 Carp* 780.650 sand, silt Carp** Delta Dunarii RO22 805.300 organic <0.01 <5 400-500 >11 Pontic shad Ecoregion 16 – The Eastern Plain Water course sector from hilly or a- siliceous Chub sand, plateau regions RO23 10-1000 b -calcareous 5-20 200-500 500-700 8-10 2-5 0.2-0.4 Undermouth pebble c- organic a-siliceous Chub Water course from plain regions sand, muddy RO24 10-2000 b- calcareous <8 <200 400-600 9-11 <2 0.2 Undermouth clay c-organic Water course sector from hilly or a-siliceous sand, Undermouth plateau regions RO25 1000-10000 b- calcareous 0.5 - 5 200-500 500-700 8-10 3-15 0.4 -2 pebble Barbel c-organic Water course sector from plain a-siliceous sand, Barbel regions RO26 >10000 b- calcareous silt, <1 <200 400-600 9-11 2-10 0.1-1 Undermouth c-organic clay a-siliceous sand, Barbel Water course sector with wet lands RO27 >10000 b- calcareous silt, <1 <200 400-600 9-11 2-10 0.1-1 Undermouth from plain regions. c-organic clay Water courses qualitatively influenced by natural factors and temporary water courses Water courses qualitatively RO28 10-1000 influenced by natural factors Non permanent mountainous water siliceous blocs, boulders, 2-16 RO29 10-1000 20-150 >800 700-1100 -2+8 0 course pebble Non permanent water course from boulders, RO30 10-1000 calcareous 25-45 500-800 600-800 7-9 5-17 0 the high plateau regions pebble Non permanent water course from a-siliceous RO31 10-1000 pebble, sand 5-30 200-500 450-550 8-10 1.5-7 0 hilly and plateau regions. b-calcareous Non permanent water course from a-siliceous RO32 10-2000 sand, silt <8 <200 400-500 9-11 <2 0 plain regions b-calcareous Other fish species present: * starlet, starry sturgeon, Russian sturgeon, Beluga, Pontic shad, pike, tench, roach, red eye, common bream, crucian carp, catfish, perch, pikeperch, , asp ** starry sturgeon, Russian sturgeon, Beluga, pike, roach, tench, red eye, common bream, crucian carp, catfish, perch, pikeperch, asp 8 28

1.2. Characteristics of fish communities from the considered catchment areas

LITHUANIA: Native fish and lamprey fauna consists of 50 species, however only 47 of them inhabit Lithuanian inland water bodies at present. Three species (Atlantic sturgeon, Allis shad (historical presence of the latter species is questionable) and Blue bream) got extinct. Native fishes and lampreys of the Lithuanian freshwaters belong to 17 families (see ANNEX 1). The most abundant is family – 22 fish species, – in 4, Salmonidae, Percidae, Petromyzontidae - in 3, Clupeidae, Coregonidae, Gasterosteida -, in 2, Cottidae Acipenseridae, Thymallidae, Osmeridae, Esocidae, Angullidae, Siluridae, Gadidae, Pleuronectidae - in 1 species. Six long- distance migrating species currently spawn in Lithuanian rivers: Sea lamprey, River lamprey, Eel, Twaite shad, Salmon and Sea trout. Vimba should also be considered as long distance migrating species (at least – in Lithuania; migrates up to 400-450 km from the Sea). Shad and sea lamprey, on the contrary, seem to spawn close to the sea, i.e. in the Nemunas River delta area. There were many attempts of introduction of various non-native fish species in Lithuania. First attempts date back to 16-17-th century, but the most intensive works of introduction took place in 20-the century. Occasional introductions of several fish species were observed too. Overall, 18 fish species were released into rivers and lakes or reared in fish farms. Luckily, only 3 of them had more or less acclimated to local conditions: Gibel, which is now widespread in various water bodies, Amur sleeper – dwells in several small dystrophic lakes and pools (however, there are records from rivers, too), and Carp. Information on the success of natural reproduction of the latter fish species is rather controversial: natural spawning events are being observed quite often, sometimes young-of-the-year juveniles are recorded. However, yearlings are extremely rare in the water bodies prevented from occasional access of artificially reared carp. Natural reproduction of Brook char was observed too, however it took place in a small brook linking two ponds of fish farm. All attempts of introduction of this species into rivers had failed. At present, 9 alien fish species occur in the inland water bodies of Lithuania, while others got extinct (Table 7).

Table 7 List of introduced and acclimated fish species which are currently present in Lithuanian inland waters (C- common, M – minor) Present Introduced Common Name Latin Name Comments on reproduction status into the country Gibel or Prussian Carp Carassius gibelio (Bloch) C before 1852 reproduces naturally 1985 (occasional Amur sleeper Percottus glenii (Dybowski) M reproduces naturally introduction) Common Carp Cyprinus carpio (L.) C 16-17-th cent. Occasional natural reproduction Brook Char Salvelinus fontinalis (Mitchill) M 1961 only 1 spawning place Northern Whitefish Coregonus peled (Gmelin) M 1960-1962 no natural reproduction Grass Carp Ctenopharyngodon idella (Valenc) M ~1960-1965 no natural reproduction Silver Carp Hypophthalmichthys molitrix (Valenc) M 1962 no natural reproduction Rainbow Trout Oncorynchus mykiss (Walbaum) M 1885 no natural reproduction 1963 (occasional introduction) reproduction was observed, but Stone moroio Pseudorasbora parva (Schlegel) M 2007 (?; till 1990 got extinct; occasional new records in 2007 introduction)

POLAND: Total number of fish species registered in Polish rivers amounted to 57, 10 of them were alien species and 5 - diadromous ones (Table 8). In Central and North Poland 46-47 species occurred (8 and 5 alien respectively), while in West Poland - 37 (3 alien) and in East and South Poland - 25-28 (3 and 1 alien respectively). Species number reflects mainly heterogeneity of sampled river habitats - greatest in Central and North Poland (rivers of different size, various altitude and slope values etc.) and lowest in South Poland (small mountain rivers) and East Poland

29 (lowland large rivers). Alien species number was highest in Central Poland, especially in the Vistula R., quite well connected to Baltic Sea and other catchments (like Dniapro River in Black Sea drainage area via canals and Bug R.). Occurrence of diadromous species was strictly correlated with connectivity of rivers with the Baltic Sea. In Pomeranian rivers all 5 registered diadromous species (eel, European river lamprey, salmon, sea trout and vimba) were found, while in other regions - up to 2 species (including eel originating almost exclusively from stocking) were noted. In upper-most region of South Poland no diadromous species were found, despite historical information on spawning grounds of many species (sturgeon, salmon, sea trout, vimba, European river lamprey) located in this region.

Table 8 Number of fish species and specimens caught in various regions of Poland. List of alien and diadromous species present in catches. No. of No. of No. of diadromous Region No. of species fish caught alien species species North 47 77881 5 5 East 28 30453 3 1 Central 46 54950 8 2 South 25 13668 1 0 West 37 44948 3 2 Poland 57 221900 10 5 Alien species: Abramis sapa, Ameiurus nebulosus, Carassius gibelio, Coregonus lavaretus, Cyprinus carpio, Neogobius fluviatilis, Neogobius gymnotrachelus, Neogobius melanostomus, Percottus glenii, Pseudorasbora parva Diadromous species: Anguilla anguilla, Lampetra fluviatilis, Salmo salar, Salmo trutta trutta, Vimba vimba

In Polish rivers taken altogether 4 fish species: gudgeon, brown trout, perch and roach dominated in terms of frequency (50-46% of fishing occasions with species present), accompanied by pike, stone loach, and three-spined stickleback (see ANNEX 2a Poland). In terms of fish number and total fish aboundance roach was a dominant (about 20%), while gudgeon and common minnow were co-dominants. This picture is of course not realistic - there is no existing real fish community with such a dominance structure. As it was stated previously only separate analysis of fish communities in particular regions of Poland can give accurate outcomes. Rivers of North Poland (mainly small streams and upper reaches of larger rivers) were dominated in terms of frequency by brown trout, three-spined stickleback, gudgeon and bullhead, co-dominants were: perch, pike, roach and European brook lamprey while in terms of abundance - gudgeon, bullhead and brown trout dominated, accompanied with common minnow, bleak and roach (see ANNEX 2a Northern Poland). Dominating species represent communities of small streams with quite steep slope, while co-dominants, like pike, perch, bleak and roach are typical of middle-size, slowly flowing rivers. This is in accordance with characteristic of environmental conditions described above for rivers sampled in this region. In Eastern Poland most frequent fish species were: roach, pike, perch and white bream (> 80%), accompanied by ide, burbot, rudd, tench, bleak, bitterling, gudgeon, crucian carp, spined loach, loach, and bream. In terms of abundance fish communities were highly dominated by roach (>45%), accompanied by pike, white bream, perch, loach and rudd (see ANNEX 2a Eastern Poland). Such fish community structure clearly corresponds to environmental conditions in rivers sampled in this region (mainly large, slowly flowing rivers and their oxbows). Values of river slope in this part of Poland ranged between 0.7 and 1.8‰ and half of sites were located in oxbows. This explains high share of roach, white bream, rudd, pike and perch in fish communities. Rivers of central Poland were dominated in terms of frequency by gudgeon, stone loach perch, roach and pike (> 70%), accompanied by dace, chub, ide, bleak, burbot and three-spined stickleback. In terms of abundance roach was a dominant (27%) with perch, bleak, gudgeon and stone loach as co-dominants (see ANNEX 2a Central Poland). Considerable diversity of

30 ichthyofauna assemblages of this region was clearly connected with high diversity of environmental conditions (high altitude amplitude, river slope range 1-11‰, sampling in very small streams and in Vistula River with some oxbows). However slowly flowing larger rivers prevailed in this region (median catchment size 490 km2, median slope 0.8 ‰), what explain domination of roach, perch and bleak in fish communities. South Poland is a considerably different region from all previously described ones. Sub- mountain and mountain streams and small rivers of high slope dominated there. This correspond well with dominance of brown trout, common minnow, eastern sculpin, stone loach and chub (frequency > 50%), accompanied by spotted barbell and gudgeon. In terms of abundance dominated common minnow and eastern sculpin, together with brown trout, stone loach, chub and spotted barbell (see ANNEX 2a Southern Poland). Attention should be paid to roach, which was present in this region, but in minor abundance and low frequency. Some similarities, like high share of brown trout and common minnow, occurred between ichthyofauna of South and North Poland, due to high percentage of small streams of high slope in both regions, despite great difference in absolute altitude and climatic conditions. West Poland differed from South Poland region mainly with different main catchment area (Oder River instead of Vistula), lower altitudes and higher share of medium-size lowland rivers. However small, steep streams had considerable share in sampled sites also in this region, what was reflected in dominance of stone loach and gudgeon in terms of frequency and gudgeon, common minnow, stone loach and brown trout in terms of abundance (see ANNEX 2a Western Poland). Considerable share of roach (both in frequency and abundance) in this region was connected to mentioned above share of larger, lowland stretches of rivers, but also to high human impact in this densely populated and industrialized area. When only large rivers (sites located in Central and Eastern Poland) are concerned, high dominance (>75%) of roach, perch, pike, ide, bleak white bream in terms of frequency is visible (see ANNEX 2b Large Rivers). Concerning total fish abundance roach strongly dominates in large rivers (31%), followed by bleak, perch and gudgeon (Table 4A). In the case of oxbow sites fish community structure is similar to that noted in stagnant water bodies, with strong dominance of roach, pike, perch, rudd, tench, and white bream in terms of frequency. The only riverine species with frequency higher than 75% is ide. Concerning total fish abundance roach is main dominant (47%), followed by white bream, pike, rudd, spined loach, perch and bitterling (see ANNEX 2b Oxbows). Considerably small share of bream in electrocatches in large rivers is caused by the fact that bream lives mainly in mid-channel habitats, which are very difficult to sample. This is confirmed by the results of additional net catches done in Vistula and Narew rivers, in which bream clearly dominated.

HUNGARY: The total number of species identified for the EFI+ project was 62 (including 1 new species for Hungary - Neogobius gymnotrachelus) (Halasi-Kovács et al, 2005). An average number of 1-25 species were caught per site (median: 11). The number of catches (records) was 2094 pcs. and the number of specimens 80336 pcs. An average number of 2-2754 individuals were caught per site (median: 299). No fish length was measured. The age groups recorded were 0+; >0+. For the complete list of fish species considered for the EFI+ project, see ANNEX 3.

ROMANIA: For the EFI+ project only data sampled from the Siret catchment area were considered. The Siret River basin is located in the Eastern Romania and it is the largest one from the country. The main fish associations from the Siret catchment area depending on habitat conditions were the following: The mountainous regions (> 800 m a.s.l.): - brown trout (ca. 40-60%) – leading species - bullhead (10-20%)

31 - minnow (<5%) High plateau regions (500-800 m a.s.l.): - minnow (40-60%) – leading species - bullhead (15-20%) - brown trout (10-15%) - Mediterranean barbel (10-20%) - spirlin (5-10%) - grayling (<5%) Medium plateau regions (400-600 m a.s.l.): - Mediterranean barbel (20-40%) - minnow (20-30%) - spirlin (10-15%) - stone loach (< 5%) - Danube longbarbel gudgeon (< 5%) - chub – invasive species in the area (10-20%) Hilly regions (200-400 m a.s.l.): - undermouth (20-40%) – replaced by the chub - barbel (15-20%) - gudgeon (5-10%) (G. obtusirostris) - bleak (5-10%) - stone loach (< 5%) - crucian carp (invasive species) - stone moroko (invasive species) - sp. - Sabanejewia sp. - tench (in large rivers – the Siret) - perch (in large rivers – the Siret) - catfish (in large rivers – the Siret) - red eye (in large rivers – the Siret) - common bream (in large rivers – the Siret) - vimba (in large rivers – the Siret) Plain regions (below 200 m a.s.l.): - gudgeon (40%) (G. obtusirostris) - bitterling (20%) (Rhodeus ammarus) - chub (15-20%) - stone moroko (10%) – invasive species - Sabanejewia valahica (2-7%) - Cobitis danubialis (5%) - perch (1-2%) - crucian carp (sub 1%) The complete freshwater fish species list from Romania, together with the list of introduced and invasive species are included in ANNEX 4.a and 4.b.

2. Main pressures in Central and Eastern European Rivers

LITHUANIA: In total three main significant pressures were identified: urban waste water treatment plants (point sources), agricultural activities (diffused sources) and the hydro-morphological changes of water bodies (canalization and water flow regulations). The biggest overall loads of pollution come from the agricultural sources, i.e. livestock and application of fertilizers. Point sources of pollution, which is mainly industrial and municipal waste water is the second source of the pollution. Non sewered and sewered inhabitants are also

32 responsible for serious loads of pollution. Therefore based on the calculation of loads, the main following drivers for pressures can be identified: - agriculture, - municipal and industrial wastewater, - diffused pollution from sewered households, without waste water treatment. Pollution with nutrients clearly predominates, accounting for ca. 93% of total pollution, causing rivers being at risk. Pollution with toxic substances accounts for ca 2% only. In addition to pollution, pressures related to the morphological changes of water bodies were identified. The biggest pressure identified is river straightening, followed by dam construction. River straightening. In 1910-1997 more than 4.4 mln ha of land was drained in Lithuania, or 46.6% of the territory. In total, 46 thou. km of river beds were straightened, from those – 24.3 thou. km of rivers, longer than 3 km. Only 13.4 thousands km or 29% of total length of river beds remain natural, majority of them being large rivers. In proportional scale, >80% of river beds are straightened in 3-10 km length rivers, ~40-50% in 10-30 km length rivers, and <20% in >30 km length rivers. Overall length of straightened rivers which catchment size is >10 km2 in the Nemunas River basin is ~9000 km. Dam construction. There are more than 1150 dams constructed on rivers in Lithuania, 858 of them – in the Nemunas River basin. From those, ~50 dams are constructed for energy production (with installed turbines), the rest – for different purposes (water reservoirs, recreation). More than 100 dams are constructed at the distance of >10 km from river source. Overall situation (with regard to river basins, accessible for migrating fish; in green) is schematically presented in Fig. 3 below.

Fig. 3 An overview of the dams built on the main river courses in Lithuania

POLAND: Hydromorphological pressures: The main hydromorphological pressure in Polish rivers is undoubtedly lack of connectivity on catchment scale (Barriers catchment down) – this pressure occurred in 84% of fishing occasions (FO) (Table 9). Lowest pressure was found in Northern (76%) and Central (83%) Poland, while highest – in East, South and West Poland (99-100%). Only some Pomeranian rivers, lower Vistula R. and lower Oder R. have connection to the Baltic Sea, all tributaries of the middle and upper Vistula and Oder are impacted. In the case of Vistula R.

33 catchment the main obstacles for fish migration are Wloclawek dam on Vistula R., and for East Poland – also Debe dam on lower Narew R. Barriers in a river segment scale (up and down) have less impact – 16 and 9% FO in whole dataset. This impact is highest in West Poland (27 and 21%) and lowest in East Poland (2 and 3 %). Impoundment has considerable impact only in West Poland (78% FO) and Central Poland (11%), while negligible in other regions (1-3%). High impact of instream habitat modification was found for the whole dataset (34% FO), with maximal values in Central and South Poland (56-57%) and minimal in East Poland (7%). Water abstraction has significant impact in Central and West Poland (mainly for irrigation and fish ponds purposes), while negligible in East Poland (1%). Hydrograph modifications are strongest in Central and West Poland (40 and 35% FO) and quite substantial in East Poland – 29% (due to influence of Siemianowka Reservoir on upper Narew R.). Velocity increase was found mainly in Central Poland, while riparian vegetation alterations – mainly in South Poland (85% FO) and to less extent in West and Central Poland (Table 9). Water quality-related pressures: Sedimentation, eutrophication and organic pollution were highest in Central Poland, mainly due to large share of sites located on Vistula R. and its tributaries, flowing through densely populated and industrialized regions with high share of intensively managed arable fields in the landscape. Considerable impact of eutrophication and sedimentation was also found in West Poland, due to the same reasons (Table 10). Water quality index was assessed mainly at 2-ond and 3-rd class (out of 5) - 74% of total FO number, with the exception for South Poland (31% in 1-rst class) and for Central Poland (56% in 4-th and 5-th class). Higher impact on water quality was also stated in Central and West Poland than in other regions, due to the same reasons as mentioned above. Lowest impact was found in upper river stretches sampled in mountain region of South Poland - 57% FO in 1-rst and 2-ond water quality class.

HUNGARY: The main human pressures in Hungary are related to the following aspects: - Water quality: ▪ bad: 6 sites 3.1% ▪ poor: 52 sites 26.9% ▪ moderate: 86 sites 44.6% ▪ good: 46 sites 23.8% ▪ high: 3 sites 1.6% - Organic pollution (only N, P forms): mainly in the lowland rivers and streams (not easy to clarify whether there are natural or human impact) ▪ No: 93 sites 48.2% ▪ Weak (medium): 77 sites 39.9% ▪ Strong: 23 sites 11.9% - River alteration: channelization (No: 91 sites; Intermediate: 70 sites; Straightened: 32 sites), bed and bank fixation, transverse constructions. - Connectivity: ▪ At larger scale: for example the Iron Gate on the River Danube, what inhibit the migration and spreading of the diadromous species. ▪ At smaller scale: the end barriers on the small running waters at the mouth what inhibit the species exchange in the given river system.

ROMANIA The most important pressures from the Siret catchment area were the following: hydro- technical works – river regularizations, dam reservoirs, canalization, dikes etc.; pollution of aquatic habitats, especially downstream the large cities (Suceava, Piatra Neamt, Bacau etc.), by means of chemical pollutants, domestic wastes, fertilizers etc. and rubble pits located on the main river courses

14 34

Table 9 Main hydromorphological pressures determined for sites available in Polish dataset, with respect to particular regions, % sites impacted Region Barriers Barriers Barriers Impound Instream Floodplain Water Main water Hydro- Velocity Riparian catchment segment segment ment habitat ** abstra use mod increase vegetation* down down up ction North 76 7 15 1 29 86 10 Fishponds 2 1 22 East 100 3 2 1 7 0 1 Fishponds 29 1 6 Central 83 12 27 11 56 39 24 Irrigation 40 30 31 South 100 12 8 3 30 100 8 Drinking 0 0 85 West 99 21 27 78 57 60 37 Fishponds 35 9 33 Poland 84 9 16 12 34 37 14 Fishponds 13 5 29 * “low” or “slight” impact was treated as „no impact” ** “No” and “Small” floodplain connection were treated as “impacted”, “large” and “medium” – as “not impacted”, “No data” cases (mainly due to lack of former floodplain) were excluded .

Table 10 Main water quality-related pressures determined for sites available in Polish dataset, with respect to particular regions; % sites impacted or % sites in particular water quality class. Region Sedimentation* Water quality index Eutrophic Organic ation* pollution* 1 2 3 4 5 North 8 4 37 41 14 4 5 2 East 2 0 27 72 1 0 8 0 Central 60 1 3 40 41 15 92 48 South 4 31 26 33 6 4 23 3 West 10 1 29 49 16 5 57 3 Poland 14 6 30 44 15 5 24 8 * “low” or “slight” impact was treated as „no impact”

35

The hydro-technical works on the Siret catchment area include numerous dam reservoirs built for hydropower, flood protection, irrigation or fishponds. In the Siret River basin alone there are ca. 20 dam reservoirs on its middle and lower course (including its major tributaries). On the Bistrita lower course there are 7 man-made lakes and 14 hydropower plants. On the Siret river course there are 7 lakes, downstream the Bacau locality its course being totally regularized. Moreover, large-scale canalization works can be seen on the Bistrita River course, from Bicaz to Bacau – about 80 km, as well as dikes on the middle and lower Barlad River course. Hydro-technical works can be found throughout Romania: there are over 250 dam reservoirs (see fig. 4).

Fig. 4 Overview of hydropower constructions in Romania

The pollution of aquatic habitats represents another important factor leading to major pressures on river ecosystems. Industrial pollution represents the main source for chemical pollutants (fertilizers, cellulose and paper factory wastes, oil products etc.). The main water courses affected by industrial pollution are the following: The Suceava (downstream Suceava locality), the Bistrita (downstream Piatra Neamt locality), the Siret (downstream its junction with the Suceava and downstream Bacau locality) and the Trotus (mainly due to oil products). However, after the year 1990, this pollution decreased with 60-80% due to the closure of numerous chemical factories. Domestic wastes affect water courses downstream large cities like Bacau, Suceava, Piatra Neamt, Focsani, Vaslui and Barlad, leading to eutrophication and organic pollution of the middle and lower Siret river basin and its main tributaries. Thus, from the 10 major water courses, only the Moldova and Putna catchment areas are not affected. The rubble pits located on the main river courses affect river natural habitats, changing the water course, eliminating fish refuge or breeding places. Even if this pressure is not so important compared to the first two, the increased number of rubble pits especially in the middle and lower river courses, strongly affect the natural habitats. Up to 5% from the length of some river segments are severely affected.

36 3. General analysis of pressures

Nearly 65% of watercourses are potentially at risk in the Nemunas River basin in the Lithuanian territory. From those due to non-point pollution - 37%, point sources – 12%, morphological changes – 35%, hydropower – 1%, due to several impacts – 14%. Straightened rivers are polluted in quite a few of cases, particularly those, flowing in the agricultural regions in the lowlands of Nemunas River basin. As concerns polluting substances (nutrients), good/moderate status threshold values are more or less established already (based on status of fish communities). Hydromorphological alterations and continuity pose much more problems, because we lack data on rivers, affected by those pressures alone. Data on impact on fish communities is controversial, too. Continuity. Two types of disruption in river continuity can be singled out, differing in a spatial scale and in their impact on fish communities of the rivers of different types. Artificial obstacles on a river basin scale. Artificial obstacle on the main river that is essential for migratory fish to reach spawning grounds results in total disappearance of fish in the whole catchment and sub-catchments above the obstacle. On the other hand, long distance migrating fish and lamprey species rarely spawn in the small tributaries with catchment size less than 100 km2. Therefore basin- scale artificial obstacle seems to have no impact on the status of fish communities in small rivers. Artificial obstacles on a river scale. Artificial obstacle on a river scale prevents free fish migration to spawning, feeding and wintering places. In this case high or even good status of fish communities is unachievable in the river stretches above the obstacles, even in those of small streams. But, again, this concerns only rivers (river stretches) with catchment size greater than 30-40 km2. Fish species, sensitive to disruptions in river continuity rarely occupy <30-40 km2 catchment size stream stretches. Hydrology. Available data is insufficient for well based assessment of impact of decrease in water yield on state of fish communities (the lack of sites with deviation of water yield being the main reason of poorer ecological status). However, current data indicates >30% decrease in water yield being considerable pressure, particularly if such decrease occurs during the low flow period. Based on published data, water reservoirs constructed on the rivers change significantly the annual pattern of the flow. Reservoirs diminish maximum flow quantity during the high flow periods. However, the impact of reservoirs on minimal water yields is controversial, being dependant on annual abundance of water. There are differences in different regions of Lithuania, too. Published data on impact of land reclamation and regulation of river beds on deviation from natural flow quantity and pattern of the rivers of Lithuania is also highly controversial. Straightening of river beds. There are great differences in the status of fish communities depending on the environmental characteristics. There are evidences that fish communities meet good status criteria in some of highland, higher slope canalized streams in Lithuania. Fish community status in the lowland, low slope canalized rivers is poor or even bad; community is represented by few eurytopic species (usually roach, perch, pike), in very low densities. Currently we have no possibility to clearly indicate potential new metrics (not covered by previous FAME project) which incorporate Lithuanian river particularities. However, metric of fish density (total density as well as density of separate species) seems promising for identification of hydrological, and particularly – morphological pressure (canalization). Metrics characterizing benthic species seem promising for assessment of low slope canalized streams (however, not valid for higher slope ones).

Due to substantial differences between particular regions of Poland, described above, pressure analysis should be done separately for each region (Table 5 and 6). Rivers in North Poland were affected by continuity disruption on catchments scale (76% FO), but to much less extent than in other regions. Also instream habitat changes and riparian vegetation degradation were important pressures in this region, as well as floodplain water bodies disconnection with the river. Water quality was good or moderate in this region, due to less dense

37 population, low share of intensively managed arable fields and lack of heavy industry. Main pressures in this region were: "barriers catchment down", "Floodplain", "instream habitat" and "riparian vegetation". Hydromorphological pressures prevailed over water quality-related ones. In East Poland most important pressure was lack of connectivity on catchment scale (100% FO) and hydrological modification, due to strong influence of Siemianowka reservoir on Narew R. Water quality was moderate (2-end and 3-rd class made 99% FO, no 5-th class was found), other pressures occurred in less than 10% FO. This region should be considered as relatively natural, with slight human impact and sites on Biebrza and Narew rivers, located mainly in National Parks may be treated as reference conditions for lowland, organic rivers, except for diadromous species, which are strongly impacted by disruption of connectivity on catchment scale. Almost every river in this region have floodplain and no impact on its connection with river channel was noted. Rivers of Central Poland were strongly impacted by barriers on catchment scale as well as on river segment scale. Also instream habitat was modified in 56% FO, while hydrological regime - in 40%. Important pressures in this region are also floodplain disconnection, hydrological modification, velocity increase, riparian vegetation alteration, water abstraction (mainly for irrigation and fish ponds) and impoundment. Water quality was usually poor (56% FO in 4-th and 5-th class), strong eutrophication, sedimentation and organic pollution levels were also stated. This results from high population density, intensive agriculture, considerable urbanization and industrialization of this region. Vistula R. (30 sites) and some smaller rivers, like Brok R. are still quite polluted and eutrophicated, although this situation improved much in last two decades (only 15% of FO in 5-th class of water quality). Sub-mountain and mountain region of South Poland was affected mainly by permanent lack of catchment scale connectivity. Considerable changes of instream habitat and strong modification of riparian vegetation were also found. High level of disconnection of floodplain was a misleading parameter here, because only few rivers in this region possessed former floodplain. Water quality was the best in this region (57% FO in 1-rst and 2-cond class), only eutrophication level was a bit higher (23 % FO). Rivers sampled in this region (mainly small streams and upper reaches of larger rivers) are slightly impacted with water quality-related pressures, while considerable impact on morphology and especially connectivity is visible. Nevertheless among these streams one can find sites close to reference conditions for those kinds of rivers, with a precaution about connectivity for diadromous species, similar to those in Eastern Poland. West region of Poland is most populated, industrialized and modified by human activity part of Poland. This is reflected in high values of all hydromorphological pressures listed in Table 5, the most important of which are: connectivity on catchment and river segment scale, impoundment, instream habitat alteration, floodplain disconnection, (> 50% FO) and also hydrological modification, and water abstraction (mainly for fish ponds). Water quality was moderate in this region, however 49% FO were in 3-dr and 21% - in 4-th and 5-ths class, what indicates considerable impact, together with high eutrophication rate. Some data from this region were however collected in mid-1990-ties, so the present state of water quality in this region may be much better than this registered in the previous years (parallel to fish sampling). High level of pressure in this region, including considerable impoundment may be responsible for higher share of roach in fish communities, than in South Poland. In case of small, sub-mountain rivers increased share of roach may indicate high hydromorphological alteration, as opposite to lowland rivers of Eastern and Central Poland, where roach strongly dominate in almost unimpacted environments. Main drivers found in the pressure analysis for Polish rivers were: - barriers segment up - hydrological modification - hydropeaking and reservoir flushing - barriers segment down - impoundment - channelization + crossection - instream habitat modification and velocity increase - embankment + flood protection - riparian vegetation alteration and floodplain connectivity For general pressure analysis, presented above, only key pressures (having more complex and direct influence on fish communities) from each chain were used, in order reduce the number of

38 variables compared. Pressures with low significance (affecting scarce percentage of sites) were also excluded from analysis. Conclusions and potential new metrics which incorporate the considered river particularities in Poland: Described above substantial differentiation of governing environmental factors, like: climate, drainage basin, altitude, actual river slope and geological typology between groups of sites located in particular regions of Poland causes need of separate analysis of each region in terms of fish community structure and pressures assessment. However for new metrics development geographical aspects are not good criteria – except for large scale regions – e.g. Mediterranean, Western Europe, Northern Europe, and Central-Eastern Europe. Taking into account the above analysis of Polish dataset it looks obvious, that river slope should be considered as the most important, governing environmental parameter. The “Actual river slope” median value is closely related to fish community structure in particular regions. Eastern and Central Poland shows some similarities in ichthyofauna composition and dominance structure. The median slope values in these regions are 0.2 and 0.8‰ respectively (that is below 1 ‰). Also fish assemblages found in North and West Poland show some similarities, despite great difference in geographical parameters, including absolute elevation. Median river slope in those regions is 1.6 and 2.4 ‰. Considerably different fish community structure was found in mountain region of South Poland (median slope 7.0‰). Combining the data on environmental characteristic, fish community structure and main pressures in particular regions and taking into account available literature data one can formulate several conclusions, leading to some metrics changes and new propositions for Central-Eastern Europe rivers: 1) It is necessary to exclude in rivers of slope ≤ 1‰ the metrics basing on presence and abundance of brown trout, bullhead (Cottus sp.) and other species of similar environmental requirements, as far as these species are rare or completely not occur in such habitats also at close to natural conditions. This change is obligatory for “Organic” rivers. 2) In rivers of slope ≤1‰) presence of roach, rudd, bleak, bream, white bream and tench should not be treated as a result of human impact (lowering the evaluation score), but as a natural fish community composition (as it can be seen in Eastern Poland at small pressure level). This change is obligatory for “Organic” rivers. 3) In case of tench, crucian carp and rudd there is even possibility to develop a new metric, applicable only for rivers of slope ≤ 1‰ with former floodplain (and especially to “organic” rivers). This metric will base on the fact, that presence (and considerable abundance) of that species in mid-channel habitats indicates good connectivity between the river and floodplain water-bodies (oxbows). 4) In rivers considered (slope ≤1‰) the presence and abundance of predators - mainly pike and perch, but in larger rivers also pikeperch, wells and asp, may serve as a new metric, indicating good (close to natural) fish community structure. 5) In rivers of slope higher than 1‰ metrics developed in former EFI project, scoring positively presence and abundance of brown trout, bullhead (Cottus sp.) and other species of similar environmental requirements are applicable in Poland. However a precaution should be made, that in Poland brown trout (and also grayling) originates in a number of rivers exclusively or mainly from introduction and stocking, so other species of this group should have more “weight” in this metrics calculation for Polish rivers. 6) In rivers with slope >1‰ the presence (or probably better the abundance) of roach may still be treated as a metric indicating human pressure (especially connected with the presence of impoundment). This can be seen in the data set from West Poland, where roach constituted considerable part of fish communities in rivers of higher slope, but strong hydromorphological impact.

39 As for Romania, the sampling sites located in the Siret catchment area were chosen in order to illustrate the pressures presented above, according to three criteria: altitude, slope and habitat nature. According to the altitude, the sampling sites from the Siret catchment area were distributed as follows: - mountainous region (over 800 m a.s.l.) – 23 sites – 8.70% - plateau region (500-800 m a.s.l.) – 91 sites – 34.5 % - hilly region (200-500 m a.s.l.) – 89 sites – 33.7% - plain region (under 200 m a.s.l.) – 61 sites – 23.1% Depending on the river slope, the sampling site distribution was balanced in the plateau, hilly and plain regions, as follows: - mountainous region (> 40‰) – 9 sites – 3.42% - high plateau region (20 - 40‰) – 29 sites – 10.98% - plateau region (10-20‰) – 54 sites – 20.45% - hilly region (5-10‰) – 85 sites – 32.2% - plain region (0,2-5‰) – 87 sites – 32.95% The distribution of the 263 sampling sites from the Siret catchment area according to the riverbed nature was made as follows: - rocks and boulders – 76 sites – 28.79% - pebble – 141 sites – 53.41% - sand – 18 sites – 6.82% - clay – 29 sites – 10.98% The sampling sites located on pebble riverbed dominated (exceeding 50%), together with those located on boulders and rocks. This is caused by the fact that 6 out of the 10 rivers considered flow through mountainous regions - the Eastern Carpathians. Hydrological pressures refer to the following: - Barriers, upstream and downstream, affect 6-10% from the habitats characteristic to the sampling sites. The barrier impact is less severe compared to the hydrotechnical works (20 large dam reservoirs, over 200 km of canals and dikes). - The other pressures (hydropeaking, water abstraction, channelization, and embankment) play a major role too, even if they only affect up to 10% from the habitats The percentage of sited affected by these pressures was as follows: - Barriers – upstream – 24 sites – 9.1%; downstream - 17 sites – 6.4% - Hydropeaking – 14 sites – 5.3% - Water abstraction – 11 sites – 4.17% - Channelization - intermediate – 13 sites – 4.92% - straightened – 23 sites – 8.71% - Embankment - local – 15 sites – 5.68% - continuous – 15 sites – 5.68% Water quality of aquatic habitats refers to toxic substances, eutrophication, organic pollution and organic siltation, together with a division into water quality classes. The toxic substances affecting the habitats in the considered sites from the Siret catchment area were relatively low (high – 3.41%, intermediate – ca. 9.5%). Over 85% from the river courses could be considered to be clean waters from this point of view. Even if toxic substances sources were numerous (oil products, fertilizers, cyanide products, phenols etc.), after the year 1990 the effects became less severe due to decreases of ca. 80% of Romanian industrial production, as depicted below: - Toxic substances – no – 220 sites – 83.33% - low – 10 sites – 3.79% - intermediate – 25 sites – 9.47% - high – 9 sites – 3.41%

40 On the other hand, the domestic wastes and polluted waters coming from agriculture and husbandry caused an increase in organic substances, affecting ca. 40% from the considered habitats, causing eutrophication of ca. 20% of the habitats, especially in the middle and lower river courses. This phenomenon led to the accumulation of organic pollutants in sediments (in ca. 13% of the sites). The percentage of sites affected by these pressures is presented below: - Organic pollution - no – 150 sites – 56.82% - weak – 101 sites – 38.26% - strong – 13 sites – 4.92% - Eutrophication - no – 212 sites – 80.3% - low – 47 sites – 17.8% - intermediate – 5 sites – 1.9% - high – 0 sites - Organic siltation – no – 230 sites – 87.12% - yes – 34 sites – 12.88% - Water quality - 1- excellent – 184 sites – 69.69% - 2 - good – 42 sites – 15.91% - 3 – mediocre – 27 sites – 10.23% - 4 – poor – 11 sites – 4.17% Thus, from this point of view, the strongest pressures in the Romanian river courses come from hydrotechnical works, followed by pollution and not the other way around. Romania entered a period of important political, social and economic changes. As a direct result, the construction of new access roads and buildings increased in the past years. Since more and more building and infrastructure material is needed, the rubble pits on the rivers throughout Romania increased proportionally. That is why we consider that it might be an illustrative and country-specific metric for the EFI+ project, relative to the decrease of benthic fish densities.

A general comparison between Western and Central/Eastern European rivers might reveal the fact that while the first have no problems with pollution, the last are still characterized by good hydromorphological conditions. However, the present report shows a different situation as concerns the Central and Eastern European Rivers. While in Hungarian and Lithuanian rivers pollution was identified as the main pressure (mainly organic pollution in both countries), in Poland and Romania the main pressures were the hydromorphological ones: lack of connectivity, impoundment, water abstraction etc. in Poland and hydrotechnical works together with other hydromorphological pressures (including instream habitat changes due to rubble pits) in Romania.

21 41 ANNEX 1: List of fish species according to the EFI+ project from Lithuanian freshwaters:

Fish families and species Current status* Latin English Petromyzontidae 1 Petromyzon marinus L. Sea lamprey M 2 Lampetra fluviatilis (L.) River lamprey C 3 Lampetra planeri (Bloch) Brook lamprey C Acipenseridae 4 Acipenser sturio (L.) Atlantic sturgeon Ex Clupeidae 5 Alosa alosa (L.) Allis shad Ex;(historical presence questionable) 6 Alosa fallax (Leceped) Twaite shad C Salmonidae 7 Salmo salar (L.) Atlantic salmon M-C 8 Salmo trutta trutta (L.) Sea trout C 9 Salmo trutta fario (L.) River trout C Coregonidae 10 Coregonus albula (L.) Vendace C 11 Coregonus lavaretus (L.) Whitefish (both, sea and lake C forms) Thymallidae 12 Thymallus thymallus (L.) Grayling C Osmeridae 13 Osmerus eperlanus (L.) Smelt (both, sea and lake forms) C Esocidae 14 Esox lucius (L.) Pike C Anguillidae 15 Anguilla anguilla (L.) Eel C Cyprinidae 16 rutilus (L.) Roach C 17 Leuciscus cephalus (L.) Chub C 18 Leuciscus leuciscus (L.) Dace C 19 Leuciscus idus (L.) Ide C 20 Phoxinus phoxinus (L.) Minnow C 21 Phoxinus percnurus Lake minnow M 22 Scardinius erythrophthalmus (L.) Rudd C 23 Aspius aspius (L.) Asp C 24 Leucaspius delineatus (Heck.) Moderlieschen C 25 Tinca tinca (L.) Tench C 26 Chondrostoma nasus (L.) Nase M 27 Gobio gobio (L.) Gudgeon C 28 Barbus barbus (L.) Barbel C 29 Alburnus alburnus ( L.) Bleak C 30 Alburnoides bipunctatus (Bloch.) Schneider C 31 Blicca bjoerkna (L.) Silver bream C 32 Abramis ballerus (L.) Zope or Blue Bream Ex 33 Abramis brama (L.) Bream C 34 Vimba vimba (L.) Zahrte or Vimba C 35 Pelecus cultratus (L.) Ziege C 36 Rhodeus sericeus (Bloch) Bitterling C 37 Carassius carassius (L.) Crucian carp C

42 Cobitidae 38 Barbatula barbatula (L.) Stone loach C 39 Cobitis taenia (L.) Spined loach C 40 (Filippi) M 41 Misgurnus fossilis (L.) Pond loach M Siluridae 42 Silurus glanis (L.) Wels M-C Gadidae 43 Lota lota (L.) Burbot C Percidae 44 Gymnocephalus cernuus (L) Ruff C 45 Perca fluviatilis (L.) Perch C 46 Sander lucioperca (L.) Zander or pikeperch C Cottidae 47 Cottus gobio (L.) Bullhead C Pleuronectidae 48 Platichthys flesus (Duncker) Flounder C Gasterosteidae 49 Pungitius pungitius (L.) Ten-spined stickleback M-C 50 Gasterosteus aculeatus (L.) Three-spined stickleback C * : C – common species, M – minor species, Ex - extinct

43 ANNEX 2.a: Dominant fish species – according to frequency (% FO – fishing occasions), number of fish caught (%N) and total fish abundance (%A) in Poland and its particular regions Poland Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Gobio gobio 50 Rutilus rutilus 20 Rutilus rutilus 19 Salmo trutta fario 49 Gobio gobio 13 Gobio gobio 14 Perca fluviatilis 46 Phoxinus phoxinus 10 Phoxinus phoxinus 9 Rutilus rutilus 46 Alburnus alburnus 8 Alburnus alburnus 6 Esox lucius 43 Barbatula barbatula 7 Barbatula barbatula 6 Barbatula barbatula 39 Salmo trutta fario 6 Salmo trutta fario 6 Gasterosteus aculeatus 33 Perca fluviatilis 5 Perca fluviatilis 5 Northern Poland Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Salmo trutta fario 60 Gobio gobio 14 Gobio gobio 13 Gasterosteus aculeatus 45 Salmo trutta fario 12 Cottus gobio 12 Gobio gobio 44 Phoxinus phoxinus 11 Salmo trutta fario 11 Cottus gobio 42 Rutilus rutilus 10 Gasterosteus aculeatus 9 Perca fluviatilis 38 Cottus gobio 7 Phoxinus phoxinus 9 Esox lucius 37 Alburnus alburnus 7 Pungitus pungitus 8 Rutilus rutilus 34 Gasterosteus aculeatus 7 Rutilus rutilus 8 Lampetra planeri 33 Pungitus pungitus 5 Eastern Poland Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Rutilus rutilus 96 Rutilus rutilus 48 Rutilus rutilus 46 Esox lucius 96 Blicca bjoerkna 8 Esox lucius 8 Blicca bjoerkna 81 Esox lucius 8 Blicca bjoerkna 7 Perca fluviatilis 87 Perca fluviatilis 7 Perca fluviatilis 6 Leuciscus idus 77 Scardinius erythrophthalmus 6 Lota lota 5 Lota lota 73 Rhodeus amarus 5 Scardinius erythrophthalmus 5 Scardinius erythrophthalmus 68 Tinca tinca 62 Alburnus alburnus 57 Rhodeus amarus 51 Gobio gobio 39 Carassius carassius 34 Cobitis taenia 34 Misgurnus fossilis 34 Abramis brama 33 Central Poland Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundances (≥ 5%) Species %FO Species %N Species %A Gobio gobio 78 Rutilus rutilus 32 Rutilus rutilus 27 Barbatula barbatula 76 Alburnus alburnus 22 Perca fluviatilis 19 Perca fluviatilis 74 Perca fluviatilis 10 Alburnus alburnus 18 Rutilus rutilus 71 Gobio gobio 8 Gobio gobio 12 Esox lucius 69 Barbatula barbatula 5 Leuciscus leuciscus 50

44 Central Poland (continued) Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundances (≥ 5%) Species %FO Species %N Species %A Leuciscus cephalus 47 Cobitis taenia 46 Leuciscus idus 43 Alburnus alburnus 42 Lota lota 41 Gasterosteus aculeatus 33 Southern Poland Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Salmo trutta fario 80 Phoxinus phoxinus 35 Phoxinus phoxinus 30 Phoxinus phoxinus 67 Leuciscus cephalus 12 Cottus poecilopus 25 Cottus poecilopus 60 Barbatula barbatula 10 Salmo trutta fario 8 Barbatula barbatula 56 Barbus peloponesius 9 Barbatula barbatula 7 Leuciscus cephalus 52 Cottus poecilopus 9 Leuciscus cephalus 7 Barbus peloponesius 47 Salmo trutta fario 5 Barbus peloponesius 5 Gobio gobio 38 Western Poland Frequency (≥ 33 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Barbatula barbatula 86 Gobio gobio 29 Gobio gobio 31 Gobio gobio 66 Barbatula barbatula 19 Phoxinus phoxinus 22 Rutilus rutilus 56 Phoxinus phoxinus 17 Barbatula barbatula 18 Salmo trutta fario 48 Rutilus rutilus 10 Salmo trutta fario 10 Phoxinus phoxinus 45 Salmo trutta fario 8 Rutilus rutilus 6 Gasterosteus aculeatus 5

45 ANNEX 2.b: Dominant fish species – according to frequency (% FO – fishing occasions), number of fish caught (%N) and total fish abundance (%A) in large rivers and oxbows (East and Central Poland) sampled with electrofishing:

Large rivers Frequency (≥ 75 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Rutilus rutilus 99 Rutilus rutilus 36 Rutilus rutilus 31 Perca fluviatilis 95 Alburnus alburnus 24 Alburnus alburnus 19 Esox lucius 91 Perca fluviatilis 11 Perca fluviatilis 13 Leuciscus idus 90 Gobio gobio 5 Gobio gobio 10 Alburnus alburnus 81 Blicca bjoerkna 76 Oxbows Frequency (≥ 75 %) Fish number (≥ 5%) Total fish abundance (≥ 5%) Species %FO Species %N Species %A Rutilus rutilus 100 Rutilus rutilus 46 Rutilus rutilus 47 Esox lucius 98 Blicca bjoerkna 8 Blicca bjoerkna 8 Perca fluviatilis 91 Esox lucius 7 Esox lucius 7 Scardinius erythrophthalmus 86 Perca fluviatilis 7 Scardinius erythrophthalmus 6 Tinca tinca 84 Scardinius erythrophthalmus 7 Cobitis taenia 5 Blicca bjoerkna 80 Rhodeus amarus 6 Perca fluviatilis 5 Leuciscus idus 77 Rhodeus amarus 5

46 ANNEX 3: List of fish species according to the EFI+ project from Hungarian water bodies:

Species 1 Eudontomyzon mariae (BERG, 1931) 2 Eudontomyzon danfordi (REGAN, 1911) 3 Acipenser ruthenus (LINNAEUS, 1758) 4 Anguilla anguilla (LINNAEUS, 1758) 5 Rutilus rutilus (LINNAEUS, 1758) 6 Rutilus pigus (HECKEL, 1852) 7 Ctenopharyngodon idella (VALENCIENNES, 1844) 8 Scardinius erythrophthalmus (LINNAEUS, 1758) 9 Leuciscus leuciscus (LINNAEUS, 1758) 10 Leuciscus cephalus (LINNAEUS, 1758) 11 Leuciscus idus (LINNAEUS, 1758) 12 Phoxinus phoxinus (LINNAEUS, 1758) 13 Aspius aspius (LINNAEUS, 1758) 14 Leucaspius delineatus (HECKEL, 1843) 15 Alburnus alburnus (LINNAEUS, 1758) 16 Alburnoides bipunctatus (BLOCH, 1782) 17 Blicca bjoerkna (LINNAEUS, 1758) 18 Abramis brama (LINNAEUS, 1758) 19 Abramis ballerus (LINNAEUS, 1758) 20 Abramis sapa (PALLAS, 1814) 21 Vimba vimba (LINNAEUS, 1758) 22 Pelecus cultratus (LINNAEUS, 1758) 23 Chondrostoma nasus (LINNAEUS, 1758) 24 Tinca tinca (LINNAEUS, 1758) 25 Barbus barbus (LINNAEUS, 1758) 26 Barbus peloponnesius petenyi (HECKEL, 1852) 27 Gobio gobio (LINNAEUS, 1758) 28 Gobio albipinnatus LUKASH, 1933 29 Gobio kessleri (DYBOWSKI, 1862) 30 Pseudorasbora parva (TEMMINCK & SCHLEGEL, 1842) 31 Rhodeus sericeus (PALLAS, 1776) 32 Carassius carassius (LINNAEUS, 1758) 33 Carassius gibelio (BLOCH, 1782) 34 Cyprinus carpio (LINNAEUS, 1758) 35 Hypophthalmichthys molitrix (VALENCIENNES, 1844) 36 Barbatula barbatula (LINNAEUS, 1758) 37 Misgurnus fossilis (LINNAEUS, 1758) 38 Cobitis elongatoides (BACESCU & MAIER, 1969) 39 Sabanejewia aurata (FILIPPI, 1865) 40 Silurus glanis (LINNAEUS, 1758) 41 Ameiurus melas (RAFINESQUE, 1820) 42 Ameiurus nebulosus (LESUEUR, 1819) 43 Salmo trutta m. fario (LINNAEUS, 1758) 44 Umbra krameri (WALBAUM, 1792) 45 Esox lucius (LINNAEUS, 1758) 46 Lota lota (LINNAEUS, 1758) 47 Gasterosteus aculeatus (LINNAEUS, 1758) 48 Lepomis gibbosus (LINNAEUS, 1758) 49 Perca fluviatilis (LINNAEUS, 1758) 50 Gymnocephalus cernuus (LINNAEUS, 1758)

47 Species (continued) 51 Gymnocephalus baloni (HOLČIK & HENSEL, 1974) 52 Gymnocephalus schraetser (LINNAEUS, 1758) 53 Sander lucioperca (LINNAEUS, 1758) 54 Sander volgensis (GMELIN, 1788) 55 Zingel zingel (LINNAEUS, 1758) 56 Zingel streber (SIEBOLD, 1863) 57 Proterorhinus marmoratus (PALLAS, 1814) 58 Neogobius fluviatilis (PALLAS, 1814) 59 Neogobius kessleri (GÜNTHER, 1861) 60 Neogobius melanostomus (PALLAS, 1814) 61 Neogobius gymnotrachelus (KESSLER, 1857) 62 Perccottus glenii (DYBOWSKI, 1877)

48 ANNEX 4.a: List of Romanian freshwater fish species according to Nalbant 2003 considered for the EFI+ project:

Species name according to No. EFI+ species name Romanian synonyms Kottelat & Freyhof 2007 1 Abramis ballerus Ballerus ballerus 2 Abramis brama Abramis brama 3 Abramis sapa Ballerus sapa 4 Acipenser gueldenstaedtii Acipenser gueldenstaedtii 5 Acipenser nudiventris Acipenser nudiventris 6 Acipenser ruthenus Acipenser ruthenus 7 Acipenser stellatus Acipenser stellatus 8 Acipenser sturio Acipenser sturio 9 Alburnoides bipunctatus Alburnoides bipunctatus 10 Alburnus alburnus Alburnus alburnus 11 Alosa maeotica Caspialosa maeotica Alosa maeotica 12 Anguilla anguilla Anguilla anguilla 13 Aspius aspius Aspius aspius 14 Atherina boyeri Atherina boyeri 15 Barbatula barbatula Orthrias barbatulus Barbatula barbatula 16 Barbus barbus Barbus barbus 17 Barbus petenyi Barbus petenyi 18 Blicca bjoerkna Blicca bjoerkna 19 Carassius carassius Carassius carassius 20 Carassius gibelio Carassius gibelio 21 Chalcalburnus chalcoides Alburnus chalcoides 22 Chondrostoma nasus Chondrostoma nasus 23 Cobitis elongata Cobitis elongata 24 Cobitis elongatoides Cobitis elongatoides 25 Cobitis megaspila Not listed 26 Cobitis taenia Cobitis danubialis Cobitis taenia 27 Coregonus lavaretus Coregonus lavaretus 28 Cottus gobio Cottus gobio 29 Cottus poecilopus Cottus poecilopus 30 Cyprinus carpio Cyprinus carpio 31 Esox lucius Esox lucius 32 Eudontomyzon danfordi Eudontomyzon danfordi 33 Eudontomyzon mariae Eudontomyzon mariae 34 Gambusia holbrooki Gambusia holbrooki 35 Gasterosteus aculeatus Gasterosteus aculeatus 36 Gobio gobio Gobio obtusirostris Gobio gobio 37 Gobio kesslerii Romanogobio kessleri Romanogobio kesslerii 38 Gobio uranoscopus Rheogobio uranoscopus Romanogobio uranoscopus 39 Gymnocephalus baloni Gymnocephalus baloni 40 Gymnocephalus cernuus Gymnocephalus cernuus 41 Gymnocephalus schraetser Gymnocephalus schraetser 42 Hucho hucho Hucho hucho 43 Huso huso Huso huso 44 Knipowitschia caucasica Knipowitschia caucasica 45 Lampetra planeri Lampetra planeri 46 Lepomis gibbosus Lepomis gibbosus 47 Leucaspius delineatus Leucaspius delineatus 48 Leuciscus borysthenicus borysthenicus Petroleuciscus borysthenicus

49 Species name according to No. EFI+ species name Romanian synonyms Kottelat & Freyhof 2007 49 Leuciscus cephalus Squalius cephalus Squalius cephalus 50 Leuciscus idus Idus idus Leuciscus idus 51 Leuciscus leuciscus Leuciscus leuciscus 52 Liza aurata Liza aurata 53 Liza ramada Liza ramada 54 Liza saliens Liza saliens 55 Lota lota Lota lota 56 Misgurnus fossilis Misgurnus fossilis 57 Mugil cephalus Mugil cephalus 58 Pelecus cultratus Pelecus cultratus 59 Perca fluviatilis Perca fluviatilis 60 Perccottus glenii Odontobutis glenii Perccottus glenii 61 Phoxinus phoxinus Phoxinus phoxinus 62 Proterorhinus marmoratus Proterorhinus semilunaris 63 Pseudorasbora parva Pseudorasbora parva 64 Pungitius platygaster Pungitius platygaster 65 Rhodeus amarus Rhodeus amarus 66 Romanichthys valsanicola Romanichthys valsanicola 67 Rutilus heckelii Rutilus heckelii 68 Rutilus rutilus Rutilus carpathorosicus Rutilus rutilus 69 Sabanejewia balcanica Sabanejewia balcanica 70 Sabanejewia bulgarica Sabanejewia bulgarica 71 Sabanejewia romanica 72 Salmo trutta fario Salmo fario Salmo trutta 73 Salvelinus fontinalis Salvelinus fontinalis 74 Sander lucioperca Stizostedion lucioperca Sander lucioperca 75 Sander volgensis Stizostedion volgensis Sander volgensis 76 Scardinius erythrophthalmus Scardinius erythrophthalmus 77 Scardinius racovitzai Scardinius racovitzai 78 Silurus glanis Silurus glanis 79 Syngnathus abaster Syngnathus abaster 80 Thymallus thymallus Thymallus thymallus 81 Tinca tinca Tinca tinca 82 Umbra krameri Umbra krameri 83 Vimba vimba Vimba carinata Vimba vimba 84 Zingel streber Zingel streber 85 Zingel zingel Zingel zingel 86 Zosterisessor ophiocephalus Zosterisessor ophiocephalus

50 ANNEX 4.b: List of introduced and invasive species from Romania: I. Fam. Salmonidae 1. Onchorhyncus mykiss – Rainbow trout 2. Salvelinus fontinalis – Brook trout 3. Salvelinus namayicus – Lake trout 4. Coregonus albula 5. Coregonus lavaretus II. Fam. Acipenseridae 6. Polyodon spatula III. Fam. Cyprinidae 7. Ctenopharhyngodon idella – gras carp 8. Hypophthalmychties molitrix – silver carp 9. Aristichtyes nobilis 10. Mylopharhyngodon piceus 11. Pseudorasbora parva IV. Fam. Catostomidae 12. Ictiobus niger 13. Ictiobus bubalus 14. Ictiobus cyprinellus V. Fam. Poecilidae 15. Gambusia holbroki VI. Fam. Siluridae 16. Clarias sp. VII. Fam. Ictaluridae 17. Ictalurus nebulosus 18. Ictalurus melas VIII. Fam. Centrarchidae 19. Lepomis gibbossus IX. Fam. Odonthobutidae 20. Percottus glenii

51

EFI + - Improvement and spatial extension of the European Fish Index

WP 3, Subtask 7 - Mediterranean River Assessment

Periodical report – Testing new responsive metrics

Teresa Ferreira, Pedro Segurado, Paulo Pinheiro & José Maria Santos

Instituto Superior de Agronomia, Portugal

March 2008

52 Table of contents

1. Introduction...... 3

2. Identification of Mediterranean-type sites...... 6

2.1 Background...... 6

2.2 The EFI+ classification...... 7

3. Testing new metrics for mediterranean rivers...... 11

3.1 Methodology ...... 11

3.1.1 Overview ...... 11

3.1.2 Data screening ...... 11

3.1.3 Pressure indexes...... 12

3.1.4 Selected metrics...... 15

3.1.5 Quantification of species tolerance...... 17

3.1.6 Testing metric responses ...... 19

3.2 Results...... 24

3.2.1 Geographical gradients ...... 24

3.2.2 Pressure analysis...... 25

3.2.3 Testing metrics’ responses to pressure...... 28

4. Conclusions and recommendations ...... 47

5. References...... 50

2

53 1. Introduction

Previously in FAME project and also in other studies, fish assemblage’ metric responses to perturbation across Mediterranean areas were poor, and weaker than those used at the European level, both using fish-based or spatially based models (Pont et al. 2006; Schmutz et al. 2007). Major bottlenecks for the development of a multimetric index in Mediterranean regions include i) the peculiar richness patterns displayed at different space scales, ii) the naturally harsh and fluctuating, warm climate-dependent, aquatic environment, and iii) a complex and hardly-predictable combination of hydrological variability with human pressures, either present or inherited throughout centuries of fluvial and landscape uses. Moreover, there are also considerable within-region differences related to the micro-scale fluctuating environments and macroscale landscape patchiness, shaped by a complex geological evolutionary background. Attempts to develop local metric indices for the Mediterranean regions dealt with these limitations (Ferreira et al., 1996; Oliveira & Ferreira 2002), with modest degrees of success and always at small regional scales, while taxa-based fish indices for quality assessment are virtually non-existent.

In Southern European areas, the primary freshwater fish fauna is dominated by cyprinids and is characterized by a low number of genera and a high number of species per genera (Doadrio, 2001). Low species richness per site, a high degree of endemicity and basin-specific taxa assemblages, are problematic for developing biotic indices (Miller et al. 1988; Moyle & Marchetti 1999). Often, Mediterranean fish species have restricted distributions, and characterize a small region, a small basin or even a group of sites. A non-spatially, modeling approach such as the one in FAME’s tried to overcome this problem, however, the taxonomic variability of fish assemblages found in Mediterranean systems can be so high that the number of available total and reference sites was too reduced to apply a site-based approach in southernmost areas of Iberia and Greece (Pont et al. 2006).

Mediterranean-type regions generally experience limited water availability during part of the year. For 6000 years now, Man has overcome this water shortage by water storage in reservoirs, water abstraction from ground and surface sources and water transfers (Davies et al., 1994). While in temperate European rivers, anthropogenic disturbance frequently focuses on water quality and physical habitat modification, and

3

54 hydrological alterations are minima, in Mediterranean ecosystems, water quality issues are determined and amplified by the amount (or lack) of water that flows in the river channel (sometimes represented only by sewage water…), and water quality is superimposed by the hydrological yearly evolution. Even a small quantity of sewage can represent a large impairment when the river flow is smaller than it should be, but likewise in can be masked if flow is artificially increased by dam or irrigation outflows. The water quantity-dependent nature of human pressures results in less predictable, antagonistic or cumulative effects. These effects have been taking place for centuries, though intensified mid-last century onwards with the up-scale of engineering expertise and materials. As a result, it is often difficult to determine whether a site is experiencing a natural or otherwise induced flow change situation, or to quantify such change.

Hydrological variability of Mediterranean-type regions profoundly determines the life forms and life cycles of aquatic organisms, as well as ecological processes (Gasith & Resh 1999). Fish fauna from these heterogeneous ecosystems must frequently survive under alternating scenarios of too much or too little water with a few intermediate but crucial periods of investment in recruitment and growth. Under these conditions, fishes tend to have short life spans, rapid growth rates, high fecundity and early sexual maturity and spawning, as well as generalist and opportunistic feeding strategies (Granado-Lorencio 1996; Pires, Cowx & Coelho 2001; Vila-Gispert, Moreno-Amich & Garcia-Berthou 2002). During low-flow season, biotic controls (e.g. predation, competition) may take over assemblage responses to other pressures (Matthews & Marsh-Matthews, 2003). The apparent tolerance of native species to naturally harsh environments and their obvious short-term resilience may actively mask man-made pressures, e.g. impede the distinction between a fortuitous series of natural low-flow years and the downstream water decrease through damming. Doubtless, separating natural and human-made pressures is a central problem in bioassessment (Fausch et al. 1990). Finally, affinity taxa, sometimes with a recent genesis in geological terms, are likely to have similar ecological requirements, but frequently there is a lack of evidence for such assumption. However, metric development strongly relies on accurate guild classification and reliable tolerance responses.

The objectives of the sub-task 3.7 Mediterranean River Assessment, which will be dealt with along this report, included:

4

55 a) For the improvement of the database

- To identify truly Mediterranean-type sites and increase the number of fishing sites available for data treatment;

- To increase the quality and decrease the spatial scale of the impairment drivers, especially those related to hydrological and geo-morphological changes;

- To increase the quality of the reference conditions, through ecological data screening

b) For the improvement of metric response

- To attempt the definition of synthetic variables for different types of pressure, including biotic pressure;

- To test key-species (either as presence or abundance) as potentially relevant metrics responding to different types of pressures and to test individual species’ indicator value for pressure response, taking into account the environmental background;

- To test the response to different types of pressure of widely spread, longtime established, target exotic species, assuming that a large part of the native species are quite tolerant to harsh physical-chemical environments and therefore poor indicators (c.f. Ross, 1991; Kennard et al, 2005);

- To test new ecotaxa guilds for different types of pressures, and for combined effects and response types, taking into account the environmental background;

- To study the response to pressure of length-age metrics, or size-class proportions of the population (juveniles or adults), either for key-species of potamodromous, namely to detect hydromorphological river alteration impacts (also in Task 3.8) and connectivity losses (also in Task 3.6).

c) For contributing to follow-up Task 4

- Improvement of metrics used before, on the basis of the tolerance indicator’ values obtained in this study, tolerant species and intolerant species;

- Recommendations of inclusion of Mediterranean-specific metrics and single or combined pressures, to be used in Task 4;

-Recommendations to be incorporated in the development of the EFI model.

5

56

2. Identification of Mediterranean-type sites

2.1 Background

The task of identifying Mediterranean-type rivers at the European scale is particularly challenging, as no unequivocal and consensual criteria are found in the literature, even for classifying Mediterranean climate zones (Hooke, 2006). According to early definitions, such as those of Köppen (Harding, 2006; Hooke, 2006), the Mediterranean region corresponds to the climatic zone in which there is at least three times as much rain in the wettest month of winter as in the driest month of summer, the latter having less than 30mm precipitation. However, this definition has the limitation of only considering the temporal distribution of precipitation, which is not the single factor influencing the hydrological regimes. Other climatic parameters such as temperature and evapotranspiration also play an important role on water availability along the year.

More recent bioclimatic classification criteria, mainly those developed by Rivas- Martinez (1999; 2005), take into account the annual distribution and relationships among several climatic parameters. One of the most important parameters are the Ombrothermic Indexes that, in broad terms, are given by the quotient between Precipitation and Temperature, though they may express slightly different conditions depending on how they are calculated (see Rivas-Martinez 1999 for further details on index calculations). According to the Rivas-Martinez ombrothermic criteria, the Mediterranean macrobioclimate is characterized by, at least, two consecutive dry months during the summer. A month is defined as dry if the precipitation (mm) is less than twice the temperature (centigrade degrees). Hence, if the ombrothermic bimonthly quotient of the two driest months is higher than two, the territory is not Mediterranean. However, if that quotient is less than two, the territory may or may not Mediterranean, as the bimonthly deficient hydrical balance may or may not compensated with the previous month’s precipitation. To account for this compensatory effect a table of Summer ombrothermic compensation values was defined (see Rivas-Martinez, 1999, for further details).

The task of defining Mediterranean regions is even more demanding when we consider the high variability of annual precipitation among Mediterranean areas. Rainfall usually ranges between 275 and 900 mm, but certain Mediterranean-climate regions may fall

6

57 into the category of semiarid regions, i.e., with annual precipitation ranging between 200 and 500 mm (Velasco et al., 2003).

Despite the absence of consistent and simple criteria, there are four basic characteristics of Mediterranean climate that are most often mentioned in the literature, namely i) low annual precipitation, ii) high precipitation seasonality, iii) mild winters and iv) hot and dry summers (e.g. Blondel & Aronson 1999; Gasith & Resh 1999; Hooker, 2006). As a consequence, streams on this climatic region have two important features that makes them diverge from other European rivers: i) the frequent occurrence of extreme flood or torrential events due to the concentration of annual precipitation in few months and ii) the occurrence of a dry period, during which the water flow is interrupted, due to very low rainfall and high temperatures on summer months (Romero et al., 1998; Gasith & Resh, 1999; Magalhães et al., 2002; Bonada et al., 2005; Ferreira et al., 2007).

Among the climatic attributes typically attributed to Mediterranean regions, we intentionally favoured those that affect more directly the extent of the dry season. In fact, it is here assumed that these attributes are the most closely related to a key feature of Mediterranean streams that have strong implications on bioassessment analyses: the increased role of spatial pattern and physical characteristics of summer refugia on structuring fish assemblages (Magalhães et al., 2002).

2.2 The EFI+ classification

In our identification of Mediterranean sites, given the great number of sites and the extent of territory to be classified, we have used exclusively climatic information, mainly for its availability and simplicity to process in a GIS environment.

With the purpose of express in a simple and straightforward way the probability of a given river stretch to show Mediterranean features we based our classification on relationships between temperature and precipitation only. Information on these two parameters has the advantage of being easily available over a vast territory and with adequate spatial and temporal resolutions.

We adopted a conservative criteria of mediterranity by considering the fourmonthly estival ombrothermic index (Ios4), that is, the sum of monthly precipitation divided by the sum of monthly mean temperature of the two driest months (July and August in Europe) plus the two previous months (May and June):

7

58

May PpPp June July +++ PpPp August Ios 4 = × 10 Eq. 1 May June July +++ TpTpTpTp August

, where Ppm and Tpm are, respectively, the yearly positive precipitation (in mm, total average precipitation of those months whose average temperature is higher than 0ºC) and the yearly positive temperature (in tenths of degrees Celsius, sum of the monthly average temperature of those months whose average temperature is higher than 0ºC) on month m. The two previous months to the dry period are included because it is assumed that summer aridity greatly depends on the rain that falls during May and June.

Since this criteria included many regions from the Atlantic climatic zone we further use a total annual precipitation (TAP) threshold of 1200 mm to separate Mediterranean from temperate regions. We considered two levels of mediterranity according to the following criteria:

Mediterranity level 1 - Ios4 < 1 AND TAP < 1200mm

Mediterranity level 2 - Ios4 < 2 AND TAP < 1200mm

As climatic variables we used 30 seconds (600 - 800 meters) resolution maps of monthly precipitation and monthly mean temperature that are freely available in the WORLDCLIM website (http://www.worldclim.org/).

According to the resulting map (Fig. 1) the level 1 Mediterranean zone, among the countries of the EFI+ consortium, include most of the Iberian Peninsula, the Southern France coastal strip and Southern Italy. The level 2 Mediterranean zone mainly represent an extension to more continental zones. The map of figure 1 also shows some isolated areas (e.g. in Nantes region of France, central Hungary and eastern Romania) classified as level 2 Mediterranean zone. Sites included in those areas were not included in the mediterranean river dataset.

8

59

Fig. 1 – EFI+ classification of Mediterranity across Europe.

According to the chosen classification, Spain is the country with the largest Mediterranean area, followed by Italy, Portugal and France (Table 1). Spain has also the highest total number of sites, followed by Portugal, Italy and France (Table 1, Fig. 2). However, the number of sites in Spain is not representative of the area covered by each bioclimatic zone. For example, the level 1 mediterranean zone is clearly under- represented in Spanish dataset.

9

60 Table 1 – Total area (106 ha) of each bioclimatic region and number of Mediterranean sites per country.

Bioclimatic zone Italy France Spain Portugal

Temperate 12.19 49.82 10.46 1.14

Area Mediterranean level 1 11.63 1.47 29.60 6.91

Mediterranean level2 18.11 4.96 38.86 7.76

Temperate 461 1051 1791 105

Mediterranean level 1 51 20 1092 721 Number of sites Mediterranean level2 191 94 2448 818

Total 652 1145 4239 923

Fig. 2 – Location of Mediterranean river EFI+ sites.

10

61 3. Testing new metrics for mediterranean rivers

3.1 Methodology

3.1.1 Overview

Data analysis was essentially based on the following four questions, which basically are alternative ways to address the same problem of quantifying the effect of pressures on metrics:

• What are the upper tolerance limits of taxon or ecotaxon-based metrics to each kind of human pressure and to global disturbance?

• After accounting for environmental variability, how much does each kind of pressure and global disturbance contribute to explain taxon or ecotaxon-based metrics distribution and abundance?

• Is the absence of taxon or ecotaxon from environmentally suitable sites the result of any kind of pressure?

• Are different kinds of pressure and global disturbance related to a smaller or greater expected abundance of taxon or ecotaxon at sites according to their environmental suitability?

3.1.2 Data screening

In the present report the analysis of Mediterranean rivers was restricted to the Iberian Peninsula, since this region corresponds to a well defined biogeographical unit and, furthermore, contains most of the available Mediterranean sites. A subset of sites was selected since many had missing data for some key pressure variables (Fig. 3). Selected sites had complete information for, at least, the 25 pressure variables that are listed on table 2. We eliminated those pressure variables that had too much missing values (e.g. sedimentation) or were highly unbalanced (e.g. reservoir flushing). After the site selection procedure, the resulting dataset included 2128 sites, which represented 65% of the 3266 original sites (Fig. 4). These sites were included in 22 main river catchment systems (Fig. 5).

11

62 3.1.3 Pressure indexes

The selected single pressure variables were integrated in synthetic pressure variables according both to pressure-type-specific combinations and a global pressure combination. We used the scores of the first component of Principal Component Analysis as combined variables, in order to account for colinearities among variables. We also considered the biotic pressure, obtained from five single variables related to the number, abundance and ecotype of exotic species. Five pressure-type-specific combinations were therefore obtained expressing, respectively, problems of connectivity, hydrology, morphology, water quality, and biotic pressures. Two global pressure combinations were also considered either including or excluding biotic pressures. The selected single pressure variables, their classification scheme and the correspondent pressure types are shown in table 2.

Fig. 3 – Number of missing data on pressure variables per site.

12

63

Fig. 4 – Selected sites for new metric evaluation.

Fig. 5 – Catchment names of selected sites

13

64 Table 2 – Selected pressure variables and classification scheme.

Pressure Single pressure variables Classification type

Presence of barriers downstream in the catchment No (1), partial (2), yes (3) Presence of barriers upstream in the river segment No (1), partial (3), yes (4) Presence of barriers downstream in the river segment No (1), partial (3), yes (4)

No barrier (0) Number of barriers upstream or downstream in the river 1km segments <=1 (3), >=1 (4) Connectivity segment (2 separate variables) 5 km segments <=2 (3), >=2 (4) 10 km segments <=3 (3), >=3 (4)

No barrier (0) Distance to barriers upstream or downstream (2 separate 1km segments >250 (3), <250 (4) variables) 5 km segments >1250 (3), <1250 (4) 10 km segments >2500 (3), <2500(4)

Impoundment No (1), weak (3), strong (5) Hydropeaking No (1), partial (3), yes (4) Water abstraction No (1), weak (3), strong (5) Hydrology Hydrological modifications No (1), yes (3) Temperature impact No (1), yes (3) Velocity increase No (1), yes (3)

Channelisation No (1), partial (3), strong (5) Cross section No (1), partial (3), strong (5) Instream habitat alterations No (1), partial (3), strong (5) Morphology Riparian vegetation alteration No (1), slight (2), partial (3), high (5) Embankment No (1), slight (2), partial (3), high (5) Floodprotection No (1), yes (3)

Toxic substances No (1), weak (3), high (5) Acidification No (1), yes (3) Eutrophication No (1), low (3), interm. (4), extreme (5) Water quality Organic pollution No (1), weak (3), strong (5) Organic siltation No (1), yes (3) Water Quality Index 1 (good quality) – 5 (poor quality)

Number of exotic species Total abundance of exotic species Classification with 5 levels based on Biotic Proportion of exotic species among total fish abundance quantiles of the first principal Total abundance of exotic insectivorous species component scores. Total abundance of exotic piscivorous species

14

65 3.1.4 Selected metrics

Analyses of metric responses to pressures were based either on taxa-based metrics - aiming at identifying potential sentinel species - and guild-based metrics. For the taxa- based metrics we considered the 18 Iberian endemic , and two widespread invasive fish species Lepomis gibbosus and Gambusia holbrooki. We used both data on taxon presence/absence and abundance as responsive metrics, and the analyses were restricted to each respective potential species’ geographical range. The selected taxa, the potential distribution and the number of records are shown on table 3.

For the guild-based metrics three ecotaxa were tested: small cyprinids, large cyprinids and salmonids. Presence/absence, proportional and abundance data were used as response values. The cyprinid length classification is given on table 4.

For large cyprinids size class-based guilds were also considered. Since length data was limited to a restricted number of sites, only data on Barbus bocagei, Barbus sclateri, Pseudochondrostoma duriense and Pseudochondrostoma polilepys was used. For the Barbus species three size-classes were considered: <100mm (juveniles), 100- 200 (small adults) and >200 (large adults). For the Pseudochondrostoma species only two age classes were considered: < 120mm (juveniles) and > 120mm (adults). For size-class-based guilds only proportional data was used as response values.

15

66 Table 3 – Species used as taxa-based metrics, total relative abundance (number per ha x 104), number of occurrence sites (total and calibration) and potential area of distribution.

Number of Total Number Species calibration Distribution (river basins) abundance of sites sites

Achondrostoma arcasii 19.30 155 25 • Ave, Cantabrica, Catalana, Cavado, Douro, Ebro, Galaica, Guadiana, Lima, Lis, Minho, Tejo, Valenciana, Vouga oligolepis 15.98 132 2 • Ave, Cavado, Douro, Lima, Lis, Mondego, Oeste, Tejo, Vouga Anaecypris hispanica 0.03 5 2 • Guadiana Barbus bocagei 39.12 408 27 • Ave, Cavado, Douro, Galaica, Lima, Lis, Minho, Mondego, Oeste, Sado, Tejo, Vouga Barbus comizo 9.25 91 1 • Guadiana, Tejo Barbus graellsii 16.79 85 3 • Cantabrica, Catalana, Ebro Barbus haasi 0.25 14 2 • Catalana, Ebro, Valenciana Barbus microcephalus 4.56 51 2 • Guadiana Barbus sclateri 15.34 78 22 • Algarve, Guadalquivir, Guadiana, Mira, Segura, Sur Chondrostoma miegii 5.75 66 3 • Ebro Cobitis calderoni 0.90 39 1 • Ave, Cavado, Douro, Ebro, Lima, Minho, Tejo, Cobitis paludica 9.07 245 23 • Algarve, Cantabrica, Catalana, Douro, Ebro, Galaica, Guadalquivir, Guadiana, Lis, Mira, Mondego, Oeste, Sado, Sur, Tejo, Valenciana, Vouga Iberochondrostoma almacai 0.53 18 7 • Algarve, Mira Iberochondrostoma lemmingii 1.83 54 4 • Algarve, Douro, Guadalquivir, Guadiana, Sado, Tejo Iberochondrostoma lusitanicum 2.40 28 0 • Oeste, Sado, Tejo Pseudochondrostoma duriense 54.53 433 87 • Ave, Cavado, Douro, Galaica, Lima, Minho Pseudochondrostoma polylepis 23.79 245 20 • Algarve, Lis, Mondego, Oeste, Sado, Segura, Tejo, Valenciana, Vouga Pseudochondrostoma willkommii 3.43 48 2 • Guadalquivir, Guadiana, Sur Squalius alburnoides 33.85 308 14 • Douro, Guadalquivir, Guadiana, Lis, Mondego, Sado, Tejo, Vouga Squalius aradensis 5.94 34 15 • Algarve Squalius carolitertii 27.74 336 43 • Ave, Cavado, Douro, Ebro, Galaica, Lima, Minho, Mondego, Vouga Squalius pyrenaicus 39.19 272 35 • Ebro, Guadalquivir, Guadiana, Oeste, Sado, Segura, Sur, Tejo, Valenciana Gambusia holbrooki 11.18 112 1 • Algarve, Ave, Catalana, Cavado, Douro, Ebro, Guadalquivir, Guadiana, Lima, Lis, Mira, Mondego, Oeste, Sado, Segura, Tejo, Valenciana, Vouga" Lepomis gibbosus 21.38 263 3 • Algarve, Catalana, Douro, Ebro, Guadalquivir, Guadiana, Lis, Minho, Mira, Mondego, Oeste, Sado, Segura, Sur, Tejo, Valenciana, Vouga"

16

67 Table 4 – Cyprinid guild classification according to the mean body lengths.

Species Guild Achondrostoma arcasii Small Achondrostoma oligolepis Small Anaecypris hispanica Small Barbus bocagei Large Barbus comizo Large Barbus graellsii Large Barbus haasi Large Barbus microcephalus Large Barbus sclateri Large Chondrostoma miegii Large Iberochondrostoma almacai Small Iberochondrostoma lemmingii Small Iberochondrostoma lusitanicum Small Pseudochondrostoma duriense Large Pseudochondrostoma polylepis Large Pseudochondrostoma willkommii Large Squalius alburnoides Small Squalius aradensis Small Squalius carolitertii Small Squalius pyrenaicus Small

3.1.5 Quantification of species tolerance

As a first approach to assess the indicator value of species we estimated species tolerance based on estimations of optimal conditions and niche breath with respect to each pressure-type combination and global pressures. Two simple alternative approaches were used: the quadratic logistic regression and the weighted averaging approaches (e.g. ter Braak & Looman 1986; Jongman et al. 1995).

The quadratic or Gaussian logistic regression estimates a species response curve from presence-absence data using a second-order polynomial in the environmental variable as linear predictor. Theoretically this curve assumes a Gaussian species response curve, i.e., a symmetric unimodal curve describing species probability of occurrence along the environmental gradient, for which it is possible to estimate optimum condition (or niche position) and tolerance (or niche breath) values. The quadratic logistic function is expressed by:

17

68 ⎡ xp )( ⎤ 2 log⎢ ⎥ ++= 210 xbxbb ⎣ − xp )(1 ⎦

where p(x) is the probability of a species to occur as a function of x. The optimum and tolerance can be obtained as follows:

Optimum −= (21 bbu 2 )

Tolerance −= bt 2 )2(1

In this study we considered the upper tolerance (u + t) value (sensu ecological literature) as an estimator of species tolerance (sensu biotic integrity assessment literature) to pressures.

The method of weighted averaging is a simpler alternative to regression methods that circumvents the problem of fitting a particular response curve. This method has been long used in ecology and recently, as in the present study, it was used to quantify species tolerance to pressures in the context of biotic integrity assessments of rivers (Whittier et al. 2007; Welsh & Hodgson in press). The species optimum is simply obtained by taking the average of the values of the environmental variables weighted by species abundance, over those sites where the species is present (Jongman et al. 1995). Species tolerance is given by one standard deviation of the optimum. Optimum and tolerance are expressed as follows:

n ∑ xy iik i=1 Optimum u = n ∑ yik i=1

n 2 ∑ ()iik − uxy i=1 Tolerance t = n ∑ yik i=1

18

69 Here we used the bootstrap approach recommended by Whittier et al. (2007) in order to obtain more robust estimates of upper tolerance (u + t) values, particularly of those species collected in few sites.

The main disadvantages of the weighted averaging method is that it disregards absences and can give misleading results if the sampling is too uneven distributed along the environmental gradient (ter Braak & Looman, 1986). On the other hand the quadratic logistic approach has the main disadvantage of requiring an unimodal response for optima and tolerance to be estimated. Nevertheless, the two methods gives similar results in case of species with low probability of occurrence and/or narrow tolerance (ter Braak & Looman, 1986).

For comparison purposes, the upper tolerance values estimated with both approaches were rescaled in order to vary between 1 and 10 using the expression: 10*(species upper tolerance – minimum score) / range of values.

3.1.6 Testing metric responses

We used two distinct procedures to test the effect of different types of pressure on biotic metrics. In a first approach we assessed the contribution of pressure variables to the improvement of models that related each metric to natural environmental conditions. This was accomplished by testing the inclusion each pressure-type combination into models already fitted with variables describing natural environmental variability. In case the pressure variables explained significantly the remaining variation after fitting environmental models it was assumed that the metric was responsive to pressure. Alternatively we also used a calibration method to assess the metric response to pressure. A calibration dataset of sites with minimal human alterations was used to fit models that related each metric to natural environmental variability. The model was then extrapolated for the whole dataset and relationships between the resulting regression residuals and each kind of combined pressure were tested.

Since the aim of this study was primarily exploratory we intentionally disregarded some statistical problems during model fitting, namely problems derived from the fact that no independent validation subset was used for model accuracy assessment. Since for many taxon there were not enough sites to select a validation subset, models were validated using resubstitution methods and therefore an overestimation of accuracy is expected. Nevertheless we assume that this problem will not strongly interfere with the

19

70 aim of identifying general trends on metric responses. All statistical analyses were performed using S-Plus version 2000 for Windows (Statistical Sciences, 1999) and MASS library (Venables and Ripley, 1997).

Assessing the contribution of pressure variables to environmental models

The quantification of metric responses to environmental and pressure variables was based on regression methods. Logistic regression analysis was used for modelling presence/absence responses while linear regression analysis was used for modelling abundance data. A metric was considered responsive if a given pressure type contributed significantly to the model fit, after controlling for geographic and environmental effects. The procedure included the following steps:

1. Production of synthetic geographic / environmental variables by means of a Principal Component Analysis (PCA);

2. Adjustment of statistical models (logistic or linear models) describing metric responses to natural environmental variables using the scores of the PCA as independent variables, following a stepwise procedure based on Akaike Information Criterion (AIC);

3. Assessment of the contribution of each pressure type to the model produced in step 2, using Akaike Information Criterion, i.e., whether AIC values decreased with the addition of each pressure variable.

In the case of taxa-based metrics, several PCA, based on the correlation matrix, were produced for the geographical range (subset of river basins) where each species potentially occur. For the geographic / environmental PCA three main sets of variables were selected: pure geographical variables (longitude and latitude), variables expressing longitudinal river gradients (distance to source, distance to sea, size of catchment’s area, altitude and river slope) and local climatic variables (total annual mean precipitation, annual mean temperature, mean temperature in January and mean temperature in July). In the regression models the quadratic terms of the PCA scores were also considered for inclusion in order to account for possible unimodal relationships. Only those species occurring at least on 30 sites were considered for analysis.

20

71

Calibration-based approach

Regression-based statistical models describing metric response to natural environmental variability were first fitted using a subset of minimum disturbed sites – the calibration dataset (Table 5, figure 6). Similarly to the previous approach, logistic regression analysis was used for modelling presence/absence data while linear regression analysis was used for modelling abundance data. For ecotaxa-based metrics logistic regression was also used for proportional data. The significance of quadratic terms of environmental variables was also assessed to account for possible unimodal relationships. Only species with at least 20 presence records on the calibration sites were considered in the analysis.

The basic procedure included the following steps:

1. Production of synthetic geographic / environmental variables by means of a Principal Component Analysis (this step is common to the first step of the previous described approach);

2. Selection of the calibration dataset with minimal human disturbances;

3. Adjustment of geographic / environmental models using the scores of the PCA at calibration sites as independent variables, following a stepwise procedure based on Akaike Information Criterion (AIC);

4. Extrapolation to the whole set of sites and extraction of residuals;

5. Correlation analysis between residuals and each pressure-type and global pressure combinations.

The absence of a species from suitable conditions can be the result of inefficient sampling, metapopulacional dynamics, natural barrier to dispersal and human induced pressures. Based on this line of thought, in a first approach, presence/absence data was used to test whether the absence from suitable sites of each taxon or ecotaxon, according to the presence/absence calibration model, could be attributed to a given pressure type or global pressure. For that purpose the analysis was restricted to absence sites, i.e., using negative deviance residuals. The absolute value of such residuals is proportional to the probability of occurrence at absence sites and it is a measure of site suitability where the taxon is absent. If the relationship between residuals and pressure variables is significantly negative then pressure may be

21

72 responsible for the species absence at environmentally suitable sites. However a positive relationship does not mean that the species is responding positively to pressure, since the analysis is based on absence sites only. We did not consider residuals at presence sites (that is proportional to the probability of the species being absence from presence sites) because it would be unreasonable to consider that a species is present at an environmentally unsuitable site only because the site is affected by a given pressure.

In a second alternative approach a calibration model was fitted using abundance data instead. In this case the whole set of residuals (Pearson residuals) were related to each pressure-type combination also using correlation analysis. A significant negative relationship could mean that species were either less abundant than expected at disturbed sites or more abundant than expected at non-disturbed sites, while a positive relationship would mean that species were either more abundant than expected at disturbed sites or less abundant than expected at non-disturbed sites.

Fig. 6 – Geographical (left) and environmental (right; first two components of the PCA on environmental variables) location of calibration sites.

22

73

Table 5 – criteria used for the selection of the calibration site subset.

Pressure type Single pressure variables Criteria Connectivity Presence of barriers downstream in the river segment No (1) Hydrology Impoundment No (1) Hydropeaking No (1) Water abstraction No (1), weak (3) Hydrological modifications No (1), yes (3) Temperature impact No (1) Velocity increase No (1) Morphology Channelisation No (1) Cross section No (1) Instream habitat alterations No (1), partial (3) Riparian vegetation alteration No (1), slight (2) Embankment No (1), slight (2) Floodprotection No (1) Water quality Toxic substances No (1), weak (3) Acidification No (1), yes (3) Eutrophication No (1), low (3) Organic pollution No (1), weak (3) Organic siltation No (1), yes (3) Water Quality Index 1 (good quality)- 3 (interm. quality)

23

74 3.2 Results

3.2.1 Geographical gradients

The main geographic gradients are essentially described by four components of the PCA, which explained 84% of the variation (Table 6): i) the first component essentially describing the temperature gradient, ii) the second and third components expressing in different ways the longitudinal gradient, and iii) the fourth component describing mainly the actual river slope (Table 7). In subsequent analyses the first four principal components were retained as covariables, in order to account for the main geographic, climatic and longitudinal gradients when testing metrics’ responses to pressures.

Table 6 – Summary table of global Principal Component Analysis describing the main geographic/climatic gradient.

Comp. 1 Comp. 2 Comp. 3 Comp. 4 Standard deviation 1.96 1.53 1.42 1.00 Proportion of Variance 0.35 0.21 0.18 0.09 Cumulative Proportion 0.35 0.56 0.74 0.84

Table 7 - Loadings of the global Principal Component Analysis describing the main geographic/climatic gradient.

Variables PC 1 PC 2 PC 3 PC 4

Latitude -0.28 -0.27 -0.37 -0.31

Longitude -0.23 0.39 0.06 -0.20

Distance from source 0.16 0.34 -0.57 0.08

Size of catchment 0.15 0.29 -0.62 0.19

Altitude -0.42 0.17 0.10 0.18

Actual river slope -0.17 -0.15 0.03 0.87

Distance from sea -0.31 0.36 0.09 -0.09

Annual Mean Precipitation -0.01 -0.48 -0.24 0.00

Annual Mean Temperature 0.48 0.06 0.14 0.01

Mean Temperature in January 0.46 -0.14 0.05 -0.04

Mean Temperature in July 0.29 0.36 0.25 0.11

24

75 3.2.2 Pressure analysis

The relationships among variables within and between different pressure–type groups are evidenced by the plots of PCA loadings on the first two principal components (Fig. 7) and by their geographical distribution (Fig. 8). Within the different pressure-type groups all variables are roughly related to each other along the axis that explains most variation. Connectivity-related pressures are mainly divided in two groups along the second component axis: those related to barriers upstream and those related to barriers downstream. Concerning hydrological-related pressures, variables impoundment, hydropeaking and hydrological modifications are the most related to each other and with a distinct distribution from water abstraction, velocity increase and temperature impact, according to the second component axis. Among the morphological-related pressures, the embankment and channelisation are strongly related to each other and, to a lesser extent, to floodprotection. Cross section, instream habitat and riparian vegetation alterations are separated from the previous group along the second component axis. Most pressure variables related to water quality show a strong relationship with each other, except the variable toxic substances and, in particular, water acidification, which are separated from the main group of variables according to the second component axis. Biotic-related pressures are mainly divided in two groups along the second component axis: (1) number and proportion of exotic species, and total relative abundance of exotic piscivorous fishes; (2) total relative abundance of exotic species and total relative abundance of exotic insectivorous fishes. Concerning the two global pressure variable combinations there are two important facts that should be mentioned: (1) connectivity-related variables are clearly unrelated to the remaining pressure variables according to the first to principal component axis; this is also evident from the comparison of the geographical distribution of connectivity-related pressures with those of other pressure types (Fig. 8); (2) biotic-related pressures are more associated to the main pressure variable group than to connectivity-related pressures.

25

76 Fig. 7 – Loadings of pressure variables of PCA ran separately for each pressure-type and using all pressure variables together.

26

77 Fig. 8 – Map of sites showing the scores of the first component of PCA ran separately for each pressure- type and using all pressure variables together.

27

78 3.2.3 Testing metrics’ responses to pressure

3.2.3.1 Quantification of species tolerances

Although the two approaches to species upper tolerance estimation produced distinct results, they evidenced similar general trends (Tables 8 and 9; figures 9 to 11), namely: (1) according to both methods, large cyprinids are more often tolerant to a greater number of pressure combined variables than small cyprinids (Tables 8 and 9); (2) only in very few cases there are high discrepancies between the two methods (e.g. I. duriense; Fig. 9 to 11); (3) many species were consistently ranked according to upper tolerance values (e.g. S. aradensis, I. almacai and A. hispanica as globally intolerant species and G. holbrookii and B. bocagei as globally tolerant species; Fig. 9 to 11).

An important difference between the results of the two approaches is that the quadratic logistic regression estimates species’ tolerances that may go beyond the conditions experienced by species (S. aradensis and P. duriense; Fig. 11). This is due to the fact that this method is based on the estimation of a theoretical response curve to pressures. In opposition, the weighted averaging method always predicts tolerances that are included within the range of conditions experienced by species (Fig. 11).

Overall, water quality and morphological changes were the pressure-types that yielded consistently higher number of significant tolerance values among both approaches, while fewer and more contradictory responses were found for the remaining pressure- types.

28

79

Table 8 – Rescaled upper species’ tolerance values according to the quadratic logistic regression approach (empty cells correspond to non-unimodal relationships between species presence and pressure).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids Achondrostoma arcasii 2.86 1.14 4.16 6.47 3.87 Achondrostoma oligolepis 10.00 3.43 3.28 0.10 4.86 3.08 Anaecypris hispanica 0.01 0.32 0.04 1.66 0.48 Iberochondrostoma almacai 0.00 4.96 1.85 1.04 Iberochondrostoma lemmingii 0.80 1.03 1.82 4.15 0.18 3.00 1.96 Iberochondrostoma lusitanicum 0.70 3.87 5.68 0.11 5.04 3.24 Squalius alburnoides 1.84 2.04 0.69 4.54 2.71 5.21 5.47 Squalius aradensis 0.03 0.06 0.37 0.00 Squalius carolitertii 4.65 1.87 2.56 3.56 0.11 3.37 1.87 Squalius pyrenaicus 1.53 4.34 1.30 Large cyprinids Barbus bocagei 1.99 2.20 4.44 4.58 1.35 6.29 6.12 Barbus comizo 10.00 1.19 1.60 5.01 0.41 4.08 4.49 Barbus graellsii 0.19 0.46 4.86 3.28 0.47 5.83 4.99 Barbus haasi 0.80 2.76 10.00 0.00 0.00 6.43 3.01 Barbus microcephalus 0.90 3.27 0.00 10.00 0.59 3.94 Barbus sclateri 2.74 0.71 1.00 2.15 0.15 0.86 0.00 Chondrostoma miegii 0.00 0.53 5.21 2.86 0.41 5.71 8.49 Pseudochondrostoma duriense 3.72 1.81 4.60 0.16 10.00 3.66 Pseudochondrostoma polylepis 3.17 2.14 3.95 10.00 8.18 8.36 Pseudochondrostoma willkommii 0.60 1.48 1.73 4.47 0.38 3.77 10.00 Cobitids Cobitis calderoni 1.98 1.79 4.03 3.22 0.41 4.63 3.41 Cobitis paludica 3.75 5.24 2.10 5.88 5.83 Exotic Lepomis gibbosus 2.48 2.70 4.61 5.71 * 6.53 * Gambusia holbrooki 6.20 2.91 6.72 6.99 * 8.86 *

29

80

Table 9 – Rescaled bootstrap estimates of upper species’ tolerance values according to the weighted averaging method.

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids Achondrostoma arcasii 4.57 4.73 4.73 2.74 4.69 3.38 2.41 Achondrostoma oligolepis 3.45 4.85 5.70 4.88 0.72 6.00 4.73 Anaecypris hispanica 0.17 0.94 0.15 4.49 1.65 0.21 0.83 Iberochondrostoma almacai 0.00 0.00 0.00 3.53 2.54 0.09 0.66 Iberochondrostoma lemmingii 1.74 3.04 1.62 7.80 3.41 3.99 4.85 Iberochondrostoma lusitanicum 1.30 1.96 4.90 10.00 2.08 7.46 6.85 Squalius alburnoides 4.26 8.81 2.98 6.55 6.58 5.46 7.18 Squalius aradensis 0.07 0.63 0.45 3.03 2.00 0.17 0.20 Squalius carolitertii 7.36 4.46 2.09 3.05 0.68 0.00 0.00 Squalius pyrenaicus 3.94 5.37 2.35 4.52 2.50 1.97 2.26 Large cyprinids Barbus bocagei 6.06 9.28 4.53 7.04 7.93 6.90 8.39 Barbus comizo 5.41 8.59 1.04 7.43 9.85 3.52 8.25 Barbus graellsii 0.36 1.01 10.00 2.88 8.89 10.00 9.91 Barbus haasi 2.11 3.09 8.52 0.00 0.00 7.89 4.90 Barbus microcephalus 2.86 10.00 1.22 4.88 7.38 3.61 6.67 Barbus sclateri 4.23 4.34 2.52 1.27 1.93 1.08 0.85 Chondrostoma miegii 1.04 2.16 9.78 2.70 10.00 9.79 10.00 Pseudochondrostoma duriense 10.00 5.58 3.05 4.16 0.44 0.99 0.86 Pseudochondrostoma polylepis 5.94 9.48 3.66 5.06 7.31 4.59 5.73 Pseudochondrostoma willkommii 4.20 7.05 2.69 4.28 7.57 3.48 6.09 Cobitids Cobitis calderoni 5.99 8.65 4.14 2.33 4.23 3.74 3.88 Cobitis paludica 2.60 4.15 4.58 6.86 9.92 6.19 7.75 Exotic Lepomis gibbosus 4.10 7.87 3.57 6.54 * 4.97 * Gambusia holbrooki 2.92 6.88 4.36 8.85 * 9.07 *

30

81

Fig. 9 – Species fitted response curves to global pressure according to the quadratic logistic regression approach. Upper tolerance values (vertical solid line) and the available range of conditions at the species potential area of distribution (two vertical dashed lines representing the 1st and the 99th percentiles) are also represented.

31

82

Fig. 10 – Scatter plot of species relative abundance along the global pressure gradient. Upper tolerance values estimated with the weighted averaging approach (vertical solid line) and the available range of conditions at the species potential area of distribution (two vertical dashed lines representing the 1st and the 99th percentiles) are also represented.

32

83 a)

15.00 Tolerance Range of conditions available (1st and 99th percentiles) 10.00

5.00

0.00

-5.00

Combined global pressure (PC1 scores)

-10.00

Barbus haasi Barbus comizo Barbus Barbus sclateri Barbus graellsii Barbus Barbus bocagei Cobitis paludica Cobitis calderoni Lepomis gibbosus Squalius carolitertii Squalius Squalius aradensis

Squalius pyrenaicus Gambusia holbrooki Squalius alburnoides Chondrostoma miegii Anaecypris hispanica Anaecypris Barbus microcephalus Achondrostoma arcasii Achondrostoma oligolepis

Iberochondrostoma almacai Iberochondrostoma lemmingii Pseudochondrostoma duriense Pseudochondrostoma polylepis Iberochondrostoma lusitanicum Pseudochondrostoma willkommii b)

10.00 Tolerance Range of conditions available 8.00 (1st and 99th percentiles)

6.00

4.00

2.00

0.00

-2.00

-4.00

-6.00 Combined globalpressure (PC1scores)

-8.00

illkommii Barbus haasi Barbus sclateri comizo Barbus Barbus graellsii Barbus bocagei Cobitis paludica Cobitis calderoni Lepomis gibbosus Squalius carolitertii Squalius Squalius aradensis Squalius pyrenaicus Gambusia holbrooki Squalius alburnoides Chondrostoma miegii Anaecypris hispanica Barbus microcephalus Achondrostoma arcasii Achondrostoma oligolepis Iberochondrostoma almacai Iberochondrostoma lemmingii Pseudochondrostoma duriense Pseudochondrostoma polylepis Iberochondrostoma lusitanicum Pseudochondrostoma w

Fig. 11 – Estimates of species tolerance to global pressure according to the quadratic logistic regression approach (a) and weighted averaging approach. Species are ordered by increasing upper tolerance. The available range of conditions at the potential distribution range of each species is also represented.

33

84 3.2.3.2 Taxa-based metric responses to pressure – endemic cypriniformes

The analysis of Iberian endemic cypriniformes responses to pressure using presence and abundance data and different modelling approaches produced distinct results (Tables 10 to 16). All analysis showed that different taxa tend to respond in many distinct ways and strengths to the several pressure-type combinations. A common result among the different approaches is that connectivity showed small overall effects on species presence and abundance.

Non-calibration methods

The approaches that assessed the contribution of combined pressure variables to each taxon occurrence, either considering species presence or abundance, produced overall consistent results (Tables 11 and 13). However, most models explained low percentages of variability (Tables 10 and 12) according to the R2 values, with most models explaining less than 30% of data variability. The best models were produced for S. aradensis, though most probably it is due to an overfitting problem since few samples were available and the potential distribution of the species is very restricted.

Surprisingly, there were more overall significant positive (38 for both presence/absence and abundance approaches) than negative responses (22 for the presence/absence approach and 30 for the abundance approach) (Table 11 and 13). However, small cyprinids showed a greater proportion of negative responses to pressure (10 out of 20 for presence/absence data and 17 out of 26 for abundance data), than large cyprinids (11 out of 35 for presence/absence data and 13 out of 35 for abundance data). The pressure-type combination that describes morphological alteration of rivers was the variable that yielded more negative responses by individual species (eight negative responses for presence/absence data and 10 out of 11 negative responses for abundance data). S. pyrenaicus and B. sclateri were the taxa that most consistently responded negatively to pressure-type and global pressure combinations. A. oligolepis and I. lemmingii also showed many significant negative responses to pressure.

The strongest consistent negative relationships with pressure-type combination variables were attained by B. sclateri on its response to overall water quality deterioration (44% of explained variation for presence/absence data and 27% explained variation for abundance data) and by B. comizo and S. pyrenaicus on their response to morphological river alterations (respectively, 21% and 16% of model

34

85 explained variation for presence/absence data and 23% and 20% of model explained variation for abundance data).

Table 10 – Deviance-based pseudo-R2 of logistic regression models (% of deviance decrease in relation to the deviance of the pure geographic/climatic model) and contributions of pressure variables to the explained variation (values in bold).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids 1.71 2.52 1.71 2.61 1.71 2.56 2.17 Achondrostoma arcasii 0.00 32.80 0.00 35.32 0.00 33.99 21.89 22.47 22.47 22.47 26.83 23.80 22.92 23.61 Achondrostoma oligolepis 0.00 0.00 0.00 20.98 7.18 2.53 6.21 5.06 5.28 5.34 5.06 5.36 5.42 5.54 Iberochondrostoma lemmingii 0.00 4.39 5.47 0.00 5.85 6.92 9.22 13.08 12.83 15.30 12.83 12.83 13.03 12.83 Squalius alburnoides 2.23 0.00 18.53 0.00 0.00 1.75 0.00 48.81 48.81 48.81 48.81 48.81 48.81 48.81 Squalius aradensis 0.00 0.00 0.00 0.00 0.00 0.00 0.00 26.67 24.94 25.66 24.94 24.94 25.19 25.06 Squalius carolitertii 8.63 0.00 3.76 0.00 0.00 1.33 0.65 24.09 23.15 26.48 25.49 23.40 24.81 24.54 Squalius pyrenaicus 5.08 0.00 16.38 11.97 1.40 8.70 7.35 Large cyprinids 24.75 31.72 24.74 24.58 28.17 26.03 27.98 Barbus bocagei 1.54 30.30 1.50 0.67 17.43 7.99 16.66 22.99 22.99 27.32 22.99 25.42 24.04 22.99 Barbus comizo 0.00 0.00 20.58 0.00 12.42 5.68 0.00 31.57 31.01 29.52 29.52 50.16 30.71 34.43 Barbus graellsii 9.20 6.84 0.00 0.00 58.39 5.52 20.22 10.82 10.82 12.27 10.82 10.82 10.82 10.82 Barbus microcephalus 0.00 0.00 13.22 0.00 0.00 0.00 0.00 10.29 10.29 11.80 17.09 10.29 13.35 12.12 Barbus sclateri 0.00 0.00 14.24 44.34 0.00 25.55 16.79 15.89 15.89 15.89 15.89 35.51 15.89 22.00 Chondrostoma miegii 0.00 0.00 0.00 0.00 65.69 0.00 33.04 24.04 21.92 19.85 20.24 19.85 23.35 22.84 Pseudochondrostoma duriense 21.75 11.82 0.00 2.43 0.00 18.73 16.38 19.78 19.90 16.96 16.03 17.14 17.17 17.74 Pseudochondrostoma polylepis 22.58 23.19 6.53 0.00 7.74 7.95 11.51 9.87 9.87 11.02 9.87 12.16 9.87 9.87 Pseudochondrostoma willkommii 0.00 0.00 11.57 0.00 20.89 0.00 0.00 Cobitids 1.60 2.66 1.60 1.60 1.91 1.87 2.07 Cobitis calderoni 0.00 40.45 0.00 0.00 16.45 14.40 23.09 16.73 16.61 16.94 16.61 16.88 16.61 16.61 Cobitis paludica 0.85 0.00 2.30 0.00 1.91 0.00 0.00

35

86

Table 11 - Coefficient estimates of each pressure type for each logistic regression model using Iberian endemic cypriniformes (negative effects of pressure types are indicated in bold; darker blue indicates stronger contributions of pressure variables to the model according to the values of table 8; empty cells correspond to non-selected variables).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids Achondrostoma.arcasii - 0.58 - 0.81 - 0.98 0.81 Achondrostoma.oligolepis - - - -0.55 -0.30 - - Iberochondrostoma.lemmingii - - - - -0.30 - -0.65 Squalius.alburnoides 0.25 - -0.66 - - - - Squalius.aradensis ------Squalius.carolitertii 0.66 0.34 -0.42 - - 0.69 0.61 Squalius.pyrenaicus 0.42 - -0.97 -0.47 -0.86 -0.33 Large cyprinids Barbus.bocagei 0.50 1.38 -0.45 0.22 0.62 1.18 1.62 Barbus.comizo - - -1.43 0.32 0.34 -0.69 - Barbus.graellsii -1.38 -0.65 - 0.90 - 1.27 Barbus.microcephalus - -0.88 0.46 - - - Barbus.sclateri - -0.80 -1.11 - -1.29 - Chondrostoma.miegii -0.98 - - 0.56 - 0.87 Pseudochondrostoma.duriense 0.64 0.75 - - - 1.24 1.47 Pseudochondrostoma.polylepis 1.05 0.96 - - 0.32 1.11 1.10 Pseudochondrostoma.willkommii - 0.49 -1.06 - 0.20 - - Cobitids Cobitis.calderoni - 1.11 - - 0.20 1.01 0.98 Cobitis.paludica - - - -0.30 - - -

36

87 Table 12 - R2 of linear regression models and contributions of pressure variables to the explained variation (values in bold).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids 7.39 8.59 7.39 9.34 7.39 8.77 8.26 Achondrostoma arcasii 0.00 15.16 0.00 22.62 0.00 17.02 11.43 47.55 47.55 47.55 48.15 48.11 47.55 47.55 Achondrostoma oligolepis 0.00 0.00 0.00 2.39 2.22 0.00 0.00 26.39 26.39 26.39 26.39 27.71 26.39 27.03 Iberochondrostoma lemmingii 0.00 0.00 0.00 0.00 6.49 0.00 3.26 13.22 13.01 15.06 13.01 13.01 13.01 13.01 Squalius alburnoides 1.86 0.00 15.69 0.00 0.00 0.00 0.00 44.48 44.48 44.48 44.48 44.48 44.48 44.48 Squalius aradensis 0.00 0.00 0.00 0.00 0.00 0.00 0.00 29.55 27.48 27.66 27.10 27.10 27.73 27.50 Squalius carolitertii 11.37 1.94 2.79 0.00 0.00 3.13 1.99 21.86 21.51 25.40 22.34 21.51 22.43 21.73 Squalius pyrenaicus 2.03 0.00 19.50 4.72 0.00 5.22 1.29 Large cyprinids 27.41 33.75 27.14 26.48 29.67 28.24 30.18 Barbus bocagei 5.38 29.86 4.07 0.77 15.30 9.18 17.32 28.82 28.82 34.41 29.27 31.39 29.55 28.82 Barbus comizo 0.00 0.00 22.82 2.14 11.51 3.45 0.00 36.14 34.03 33.36 33.36 42.32 33.36 35.78 Barbus graellsii 11.54 2.96 0.00 0.00 31.79 0.00 10.15 13.42 13.42 15.07 14.52 13.42 13.42 13.42 Barbus microcephalus 0.00 0.00 12.70 8.78 0.00 0.00 0.00 15.14 15.14 16.70 19.54 15.14 16.96 15.14 Barbus sclateri 0.00 0.00 11.03 26.56 0.00 12.64 0.00 9.16 7.78 7.78 7.78 16.16 7.78 9.65 Chondrostoma miegii 16.25 0.00 0.00 0.00 56.20 0.00 20.98 22.26 21.13 19.44 19.44 19.44 21.48 21.44 Pseudochondrostoma duriense 15.73 9.93 0.00 0.00 0.00 11.81 11.58 22.09 24.09 19.61 19.61 21.54 21.53 22.44 Pseudochondrostoma polylepis 13.96 23.16 0.00 0.00 11.15 11.11 15.71 14.39 15.36 17.55 14.39 15.59 14.39 14.39 Pseudochondrostoma willkommii 0.00 7.37 21.03 0.00 8.97 0.00 0.00 Cobitids 5.08 9.81 5.08 5.08 5.67 6.63 6.86 Cobitis calderoni 0.00 50.82 0.00 0.00 11.05 24.69 27.37 25.25 25.25 25.25 25.62 25.25 25.25 25.25 Cobitis paludica 0.00 0.00 0.00 1.90 0.00 0.00 0.00

37

88

Table 13 - Coefficient estimates of each pressure-type for each linear regression model using log transformed relative abundance data on Iberian endemic cypriniformes (negative effects of pressure types are indicated in bold; darker blue indicates stronger contributions of pressure variables to the model according to the values of table 8; empty cells correspond to non- selected variables).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids Achondrostoma.arcasii 0.26 0.28 0.41 0.29 Achondrostoma.oligolepis -0.92 -0.26 -0.42 -0.60 Iberochondrostoma.lemmingii -0.09 -0.10 -0.05 -0.19 -0.18 Squalius.alburnoides 0.30 -0.71 -0.38 Squalius.aradensis Squalius.carolitertii 0.48 -0.37 0.39 0.29 Squalius.pyrenaicus 0.76 -0.81 -0.86 -0.12 -1.15 -0.87 Large cyprinids Barbus.bocagei 0.22 1.31 -0.25 0.20 0.67 1.00 1.49 Barbus.comizo -0.77 0.28 -0.67 Barbus.graellsii -0.84 -1.10 1.38 -1.13 2.29 Barbus.microcephalus -0.62 Barbus.sclateri -0.69 -1.32 -1.70 -0.93 Chondrostoma.miegii 0.99 1.83 Pseudochondrostoma.duriense 0.91 0.93 0.44 1.87 2.00 Pseudochondrostoma.polylepis 1.26 0.95 -0.45 0.26 0.90 0.91 Pseudochondrostoma.willkommii -0.41 0.27 Cobitids Cobitis.calderoni 0.15 0.05 0.12 0.16 Cobitis.paludica -0.11 0.15 0.08

Calibration methods

The calibration approach of assessing taxa-based metrics response to pressure variables yielded distinct results from the previous approach (tables 14 and 15), although fewer taxa were analysed due to sample size constraints. The models explained reasonably well the taxa presence and abundance for the calibration dataset, but the agreement between observed and predicted values for the non-calibrated dataset was variable for presence data and very low for abundance data (Table 16).

38

89 According to the calibration model approach using presence/absence data the variable integrating water quality disturbances showed significant negative relationships with most single taxa (i.e. environmentally suitable sites where the species is absent are in average more affected by water quality disturbances) (Table 14). Morphological and hydrological disturbances also showed many significant negative relationships, especially for large cyprinids. No significant negative relationships of taxa-based metrics with connectivity disturbances were found.

The calibration model approach using abundance data yielded quite distinct results from the approach using presence/absence data (Table 15). However, for both methods water quality-related disturbances yielded the largest number of significant negative correlations and connectivity disturbances yielded no significant negative correlations with taxa-based metrics. Furthermore, one of the taxa that was found to be more frequently negatively related to combined pressure-type variables was common to both presence/absence and abundance approaches (P. duriense). The most evident inconsistency between presence/absence and abundance calibration model approaches was found for B. bocagei. According to the presence/absence approach the species is negatively related to five out of seven combined pressure variables (all except connectivity and biotic combined disturbance) but according to the abundance approach the species is positively related to six out of seven combined pressure variables (all except connectivity related combined disturbance). Nevertheless, this result is not necessarily incompatible, since the two approaches try to address distinct questions. While the presence/absence approach suggests that the species tend to be absent from environmentally suitable but disturbed sites, the abundance approach does not necessarily suggest that the species is more abundant than expected at more disturbed sites: it might suggest instead that it is less abundant than expected at less disturbed sites. Nevertheless, the positive response based on abundance data yielded stronger correlations than the negative response based on presence/absence data.

Considering the strength of the global response estimates, and its consistency throughout pressure types, Squalius pyrenaicus, Barbus comizo and B. sclateri could be potentially useful as metrics. Nonetheless, as expected, different species have distinctive responses across pressure types and methods, and though auto- ecologically useful, they do not favour the selection of robust metrics.

39

90

Table 14 – Pearson correlation between negative deviance residuals of logistic regression calibration models and each pressure-type synthetic variables (significant negative correlations are shown in bold).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids Achondrostoma arcasii 0.01 0.01 0.00 -0.01 -0.02 0.00 -0.01 Squalius alburnoides 0.14 0.16 -0.08 -0.24 -0.12 0.01 -0.06 Squalius aradensis -0.10 -0.17 NA -0.36 -0.13 -0.15 0.05 Squalius carolitertii 0.07 0.00 0.06 -0.19 0.05 0.01 0.02 Squalius pyrenaicus 0.05 -0.04 0.05 -0.23 -0.22 -0.18 -0.27 Large cyprinids Barbus bocagei 0.16 -0.09 -0.11 -0.10 -0.03 -0.12 -0.12 Barbus sclateri -0.05 0.12 -0.35 -0.14 0.08 0.04 0.08 Pseudochondrostoma duriense 0.04 -0.13 -0.09 -0.42 -0.06 -0.18 -0.22 Pseudochondrostoma polylepis -0.09 -0.13 -0.14 0.05 0.08 -0.09 -0.06 Cobitids Cobitis paludica 0.04 -0.17 -0.06 -0.15 -0.26 -0.09 -0.23

Table 15 – Pearson correlation between Pearson residuals of linear regression calibration models and each pressure-type synthetic variables (significant correlations are shown in bold; values in blue are negative correlations and values in red are positive correlations).

Global Water Species Connect. Hydrol. Morphol. Biotic Global + quality Biotic Small cyprinids Achondrostoma arcasii 0.00 0.01 0.00 -0.01 -0.02 0.00 -0.01 Squalius alburnoides -0.01 0.02 0.00 0.00 -0.02 0.00 -0.01 Squalius aradensis 0.02 0.04 0.02 0.10 0.06 0.04 -0.02 Squalius carolitertii 0.00 0.00 -0.03 -0.01 -0.02 -0.02 -0.03 Squalius pyrenaicus 0.09 0.01 -0.13 -0.08 -0.08 -0.04 -0.07 Large cyprinids Barbus bocagei -0.01 0.25 0.07 0.22 0.18 0.21 0.24 Barbus sclateri 0.18 0.01 -0.08 -0.17 -0.10 -0.02 -0.10 Pseudochondrostoma duriense 0.00 -0.18 0.01 -0.13 -0.67 -0.10 -0.41 Pseudochondrostoma polylepis 0.00 -0.11 -0.06 -0.11 -0.03 -0.09 -0.10 Cobitids Cobitis paludica -0.03 -0.06 0.01 -0.03 -0.05 0.02 -0.04

40

91 Table 16 – Measures of calibration model adjustment (AUC – Area under the ROC) for calibration and non-calibration sites.

Logistic regression Linear regression

AUC AUC R2 R2 Espécies calibration non-calibration calibration non-calibration

Achondrostoma.arcasii 0.76 0.51 0.06 0.00 Barbus.bocagei 0.95 0.74 0.24 0.00 Barbus.sclateri 0.91 0.66 0.42 0.03 Cobitis.paludica 0.92 0.78 0.30 0.00 Pseudochondrostoma.duriense 0.84 0.67 0.24 0.01 Pseudochondrostoma.polylepis 0.99 0.69 0.29 0.03 Squalius.alburnoides 0.99 0.62 0.50 0.00 Squalius.aradensis 1.00 0.47 0.67 0.07 Squalius.carolitertii 0.88 0.78 0.25 0.08 Squalius.pyrenaicus 0.82 0.73 0.31 0.01

3.2.3.3 Taxa-based metrics – widespread invasive species

The assessment of the two main invasive species as responsive taxa-based metrics to pressure yielded less surprising results than those obtained for cypriniformes. Significant relationships with all pressure-type combinations and global pressures yielded positive responses (Table 17). Metrics based on abundance data resulted in less significant relationships for Lepomis gibbosus, but stronger responses for Gambusia holbrokii, than metrics based on presence-absence data (Tables 9 and 10). The stronger positive response to pressure-type combination by Lepomis gibbosus was attained for hydrological alterations (16% of explained variation), while Gambusia holbrooki showed the strongest positive response to the overall water quality deterioration (52% of model explained variation) and to the global pressure combination (34% of model explained variation) (Table 18).

Both L. gibbosus and G. holbrooki can be used as metrics.

41

92 Table 17 - Coefficient estimates of each pressure type for each regression model with L. gibbosus and G. holbrooki (negative effects of pressure types are indicated in bold; darker blue indicates stronger contributions of pressure variables to the model according to the values of table 10; empty cells correspond to non-selected variables).

Water Data Species Connectivity Hydrology Morphology Global quality L. gibbosus 0.43 0.99 - 0.72 1.25 Presence G. holbrooki - 0.41 0.30 0.48 0.85 L. gibbosus - - - - - Abundance G. holbrooki - 74.80 40.31 188.03 193.39

Table 18 - Contribution of pressure variables to the explained variation of L. gibbosus and G. holbrooki, according to the deviance-based pseudo-R2 (logistic regression) and the R2 (linear regression).

Water Data Species Connectivity Hydrology Morphology Global quality

L. gibbosus 2.93 16.21 0.00 8.38 10.10 Presence G. holbrooki 0.00 3.47 1.90 4.36 4.72 L. gibbosus 0.00 0.00 0.00 0.00 0.00 Abundance G. holbrooki 0.00 16.97 7.63 52.09 33.98

3.2.3.4 Ecotaxa guild-based metrics

As was found for taxa-based metrics the response of ecotaxa to the several pressure- type and global pressure combinations varied among different approaches (Tables 19 and 21). Nevertheless, ecotaxa-based metrics were globally more responsive to disturbances than taxa-based metrics.

Non-calibration methods

The approach that assessed the contribution of each combined pressure-type and global pressure variables to each taxon model resulted in quite similar trends, either according to analysis based on presence/absence data or on abundance data (Table 11). The main difference was found for small cyprinids, which did not respond significantly to any combined pressure variable when abundance data was used.

42

93 According to presence/absence data, except for the overall morphological river alterations, small cyprinids respond positively to all combined pressure variables. Large cyprinids showed even stronger positive responses to the same pressure-type and global combinations, both according to presence and abundance data (Table 11 and 12). As an exception, presence data yielded a negative response of large cyprinids to overall morphological river alterations.

According both to presence and abundance data, in opposition to cyprinids, salmonids showed significant negative responses to all combined pressure variables, with the exception to overall river connectivity disturbances where a positive and non-significant response resulted, respectively, for presence and abundance data. (Table 11). The strongest relationship with pressure variables was found for the response of large cyprinids to biotic (44% of model explained variation) and global+biotic combined pressure variables (47% of total variation) (Table 12).

Table 19 - Coefficient estimates of each pressure type for each regression model with the three eco-guild metric (negative effects of pressure types are indicated in bold; darker blue indicates stronger contributions of pressure variables to the model according to the values of table 12; empty cells correspond to non-selected variables).

Global + Data type Ecotaxon Connect. Hydrol. Morph. Wqual. Biot. Global. Biot. Small 0.54 0.54 - 0.34 0.15 0.90 0.88 cyprinids Large Presence 0.68 0.99 -0.21 0.37 0.56 1.21 1.70 cyprinids

Salmonids 0.70 -0.23 -0.65 -1.04 -0.78 -0.95 -1.54

Small ------cyprinids Large Abundance 283.68 497.07 - 153.31 313.49 620.17 848.32 cyprinids

Salmonids - -510.66 -342.46 -431.08 -68.54 -888.52 -733.56

43

94 Table 20 - Contribution of pressure variables to the explained variation of each one of the guild metric, according to the deviance-based pseudo-R2 (logistic regression) and the R2 (linear regression) (Numbers in bold represent negative responses).

Global + Data type Ecotaxon Connect. Hydrol. Morph. Wqual. Biot. Global Biot. Small 5.98 5.63 0.00 1.93 1.32 5.66 5.83 cyprinids Large Presence 23.51 35.67 3.36 6.73 32.34 23.89 39.50 cyprinids

Salmonids 3.64 0.47 4.94 6.86 5.87 2.65 6.00

Small 0.00 0.00 0.00 0.00 0.00 0.00 0.00 cyprinids Large Abundance 20.08 39.47 0.00 4.80 44.01 28.84 46.62 cyprinids

Salmonids 0.00 12.43 7.42 8.34 0.74 14.61 11.38

Calibration methods

The approach that was based on calibration models produced results that contrasted to those of the non-calibration method (Table 21). The results of the calibration approach using presence/absence data estimated overall responses of metrics to pressures that contradicted those of non-calibration methods. Models using presence/absence data showed moderately good overall agreements, except for the extrapolation of small and large cyprinids’ models to non-calibration sites that showed poor overall agreements (Table 22). According to this approach models’ residuals for both small and large cyprinids showed a negative relationship with all combined pressure variables, except for connectivity-related disturbances. This suggest that both small and large cyprinids tend to be absent from environmentally suitable sites due to human disturbances. Salmonids showed an opposite response to disturbances to that found for cyprinids and therefore their absence from environmentally suitable sites cannot be attributed to any pressure-type.

The calibration-based method using proportional data produced similar results to those based on presence/absence data, except for salmonids. This ecotaxa guild also resulted negatively related to all pressure-type and global pressure combination, except for connectivity-related disturbances that now resulted positively correlated to all ecotaxa-based guild proportions. Models using proportional data also showed moderately good overall agreements, except for large cyprinids and for and the

44

95 extrapolation of small cyprinids’ model to non-calibration sites, that showed poor overall agreements (Table 22).

Calibration methods using abundances produced very poor relationships between calibration model residuals and disturbances.

Table 21 – Pearson correlation between residuals of logistic (presence data) and linear (abundance data) regression calibration models and each pressure-type synthetic variables (significant negative correlations are shown in bold).

Global Data type Ecotaxon Connect. Hydrol. Morph. Wqual. Biot. Global + Biot.

Small cyprinids 0.09 -0.32 -0.33 -0.29 -0.22 -0.41 -0.44

Presence Large cyprinids 0.04 -0.30 -0.40 -0.18 -0.12 -0.41 -0.39

Salmonids -0.01 0.20 0.05 0.35 0.16 0.25 0.32

Small cyprinids 0.07 -0.32 -0.32 -0.32 -0.31 -0.44 -0.49

Proportions Large cyprinids 0.16 -0.13 -0.30 -0.13 -0.20 -0.33 -0.33

Salmonids 0.11 -0.27 -0.27 -0.37 -0.16 -0.42 -0.42

Small cyprinids -0.02 -0.05 0.01 -0.03 -0.05 -0.01 -0.03

Abundance Large cyprinids -0.02 -0.05 0.01 -0.03 -0.05 -0.01 -0.03

Salmonids 0.01 0.01 -0.01 0.01 -0.11 0.00 -0.03

Table 22 – Measures of calibration model adjustment (AUC – Area under the ROC) for calibration and non-calibration sites.

Data type Adjustment measure Small cyprinids Large cyprinids Salmonids

AUC calibration 0.88 0.83 0.85 Presence AUC non-calibration 0.69 0.67 0.91

R2 calibration 0.29 0.11 0.34 Proportions R2 non-calibration 0.04 0.04 0.54

R2 calibration 0.32 0.19 0.38 Abundance R2 non-calibration 0.00 0.00 0.03

45

96 3.2.3.5 Size-class guild-based metrics

The assessment of size-class based metric response to pressure-type and global pressure synthetic variables were based exclusively on calibration methods using logistic regression models with proportional data. This approach produced contradictory results amongst the large cyprinid species considered for analysis (Table 23). For B. bocagei, residuals of calibration models for all size-class proportions showed significant positive correlations with most combined pressure variables. The exception was found for connectivity-related overall disturbance that showed a significant negative correlation with models’ residuals for both the proportion of juveniles and small adults. On the other hand, for B. sclateri, negative correlations were found between models’ residuals and four pressure combinations (Hydrological, Morphological and biotic-related disturbances, and global + biotic synthetic pressure) for both the proportions of juveniles and small adults. Furthermore the negative relationships found for B. sclateri were much stronger than the positive relationships found for B. bocagei. Therefore, besides B. sclateri´s abundance as metric, size- classes of this species can also be used.

Surprisingly, for the proportion of juveniles of both Pseudochondrostoma species and also for the total proportion of juveniles, only positive correlations between residuals and combined pressure variables were found, although only very weak relationships were found.

Table 23 – Pearson correlation between residuals of logistic regression calibration models using size-class proportions and each pressure-type synthetic variables (significant correlations are shown in bold; values in blue are negative correlations and values in values in red are positive correlations).

Global Water Species/length class Connect. Hydrol. Morphol. Biotic Global + quality Biotic B. bocagei juv -0.11 0.24 0.09 0.07 0.08 0.19 0.20 B.bocagei small ad -0.14 0.36 0.13 0.19 0.11 0.30 0.32 B.bocagei large ad 0.12 0.31 0.11 0.40 0.13 0.26 0.35 B. sclateri juv -0.03 -0.32 0.28 -0.35 -0.44 -0.02 -0.24 B. sclateri small ad -0.05 -0.41 0.17 -0.48 -0.65 -0.19 -0.47 B. sclateri large ad 0.10 0.42 0.22 0.47 0.26 0.45 0.48 P. duriense juv 0.20 0.18 -0.02 0.20 0.05 0.03 0.13 P. polylepis juv 0.13 0.17 -0.09 -0.07 0.07 -0.09 -0.01 Total % of juveniles 0.03 0.25 0.04 0.16 0.04 0.14 0.18

46

97 4. Discussion and recommendations

In the present periodical report of sub-task 3.7 new metrics for a specific Mediterranean biogeographical unit (the iberian Peninsula) were tested using several alternative approaches, with the aim of selecting those metrics that most consistently responded to disturbances. Metrics may respond differently to each kind of disturbance and therefore separate analysis were carried out for different pressure-type combined variables. A first overall impression of the results presented in this report is that different methods basically produced variable results concerning metric responses to pressures, and that it is difficult to draw a conclusion about which metrics should be further selected for the biotic index development. Besides the variability displayed by different methods, various ways to quantify the fish populations, or parts of the population, also yielded different outputs. Nonetheless, some of the most consistent and general negative responses to disturbance came from B. sclateri, B. comizo and S. pyrenaicus.

However, the different approaches used in this report tried to deal with distinct questions that can be asked in the context of biological responses to human disturbances (see subsection 3.1.1). Therefore, a metric response detected with a given approach is not necessarily contradicting the result of another approach; the two approaches may simply being addressing different questions. For example, in theory it is possible that the abundance of a species showing a higher upper tolerance value will be more negatively responsive to disturbance than the abundance of a species showing a lower upper tolerance.

An important feature of river fish communities in the Mediterranean region is the high level of endemic species with restricted geographical ranges. The ecology of many of such species is largely unknown, which poses a problem for their guild classification. An example is the species classification into tolerance guilds, which have been typically carried out through expert judgment. In this report we propose two possible approaches to objectively quantify species upper tolerance to each pressure-type and to global pressure variability. The procedure can be used to help on guild classification, and also for creating new testable metrics, by considering for example the mean species tolerance per site or the percentage of species with upper tolerances below or above certain thresholds. Metrics using upper tolerance information separately to each kind of pressure may also be considered.

47

98 A potential problem of both approaches used to quantify species tolerance is that environmental variability may interfere with estimates. In fact, there is a geographical coincidence between pressure and environmental variability: for example most disturbed sites tend to be concentrated in the lower river segments. As a consequence, species showing low estimated upper tolerance values may in fact be tolerant species that are responding to environmental variability and not to pressure. This is why tolerance estimates cannot be directly used as surrogates to species response to pressures. In this report metric response to pressure was based on methods that tried to account for environmental variability. These included non-calibration and calibration methods. The calibration method is probably more reliable, since theoretically more realistic metric responses to environmental variability are obtained using a calibration dataset that includes only the most undisturbed sites. However, the non-calibration method have the advantage of allowing the assessment of a larger number of taxa- based metric responses. The use of a calibration dataset can also be limited by its representativeness of environmental variability. If the full range of environmental conditions are not represented in the calibration dataset then there is the problem of predicting metrics beyond the environmental conditions at calibration sites. In fact, this was the case with the calibration dataset used in this report: calibration sites were found to be more frequently located in the North-western region of Iberia and in Central Spain and, therefore, clearly did not represent the full environmental spectrum (see Fig. 6). Hence, caution is needed when interpreting metrics responses to pressure using the calibration method, since there is the danger of spurious relationships to be found.

Another challenge of calibration methods is to discriminate whether metrics are higher or lower than expected at less disturbed sites or higher or lower than expected at more disturbed sites, or both. This is a relevant question because in the context of biotic river quality assessments, the main concern is to find metrics that helps to identify the most disturbed sites. Therefore, analyses should focus on whether the metrics has higher or lower values than expected at the most disturbed sites.

Another potential problem that should be accounted for when interpreting results of metric responses to pressure variables is the spatial autocorrelation among sites, i.e., the tendency of closer sites to show similar biotic and abiotic conditions. The effects of spatial autocorrelation in regression and correlation are well known and include biased coefficient estimates and inflation of statistical significance of relationships (see e.g. Lennon 2000). Since unimpaired locations probably tend to be aggregated in space, the calibration subset will be more susceptible of showing an even stronger spatial

48

99 autocorrelation than the original dataset, which may represent an additional problem to calibration approaches.

Despite all the above mentioned possible problems, the results suggest that there are potentially good ecotaxa and taxa-based indicators among the several endemic cypriniformes and the most widespread invasive fish species, although surprisingly some of these metrics responded positively to pressures. The ecotaxa guild-based metrics were globally more responsive to metrics than simple taxa-based metrics. The results also showed that the two invasive species analysed (L. gibbosus and G. holbrookii) are also good candidate metrics, namely as positive indicators of hydrological, water quality and global disturbances.

The most consistent result across the several approaches used was that connectivity- related pressures showed the lowest relationship with both the taxa and ecotaxa-based metrics considered in this study. In fact, a stronger response of large potamodromous cyprinids with connectivity-related disturbances was expected. Nonetheless, the stronger negative relationship with connectivity disturbances was indeed found for the proportion of juveniles of a potamodromous species (B. bocagei), suggesting that it may be a good metric for connectivity problems. It should be emphasized that connectivity pressure may be underestimated due to the lack of a thorough field inventory on the number of small and/or old barriers and their capacity to be transposed by fish undergoing migratory movements.

Species-based metrics are limited by the geographical range of the species, sometimes quite small. This is not the case for widespread L. gibbosus and G. holbrooki. Also, metrics based on eco-taxa guilds are more promising because of its wider distribution.

The methods and procedures used in this report represent just a few among many other approximations to the problem of isolating and quantifying metric responses to pressures in Mediterranean areas. Based on the results of this report and the considerations made above, important challenges for further development of this issue would be to i) using estimated tolerances to the several combined pressure variable in metric calculations, ii) assess different methods for combining single pressure-type variables, iii) consider additive or interaction effects among different types of pressure, iv) test further responses of length-age-based metrics, both for other species and for eco-taxa guilds, v) use different sets of geographical and environmental variables, vi) develop a straightforward procedure to account for the effect of spatial autocorrelation

49

100 when testing metric responses, vii) taking into account fish home-range by using a segment-base and not a site-based approach.

5. References

Blondel J. & Aronson J. (1999) Biology and Wildlife of the Mediterranean Region. Oxford University Press, Oxford.

Bonada N, Zamora Munoz C, Rieradevall M, Prat N. (2005) Ecological and historical filters constraining spatial caddisfly distribution in Mediterranean rivers. Freshwater Biology 50, 781–797.

Davies B.R., Thoms M.C., Walker K., O_Keefe F. & Gore J.A. (1994) Dryland rivers: their ecology, conservation and management. In: P. Calow & G.E. Petts (eds) The Rivers Handbook: Hydrological and Ecological Principles. Oxford: Blackwell Science, pp. 484–511.

Doadrio, I (2001) Atlas y Libro Rojo de los peces continentals de Espana. Direccion Generale de la Naturaleza. Museo Nacional de Ciencias Naturales. Madrid.

Fausch, K.D., Lyons, J., Karr, J.R. & Angermeier, P.L., (1990): Fish communities as indicators of environmental degradation. American Fisheries Society Symposium 8, 123-144.

Ferreira M.T., Cortes R.M., Godinho F.N. & Oliveira J.M. (1996) Biological indicators of water quality applied to the Guadiana basin. Recursos Hidricos 17, 9–20.

Ferreira M.T., Oliveira J., Caiola N., de Sostoa A., Casals F., Cortes R., Economou A., Zogaris S., Garcia-Jalon D., Ilhéu M., Pont D., Rogers C., Prenda J. (2007) Ecological traits of fish assemblages from Mediterranean Europe and their responses to human disturbance. Fisheries Management and Ecology 14, 473– 481.

Gasith A. & Resh V.H. (1999) Streams in Mediterranean climate regions: abiotic influences and biotic responses to predictable seasonal events. Annual Review of Ecology and Systematics 30, 51–81.

Granado-Lorencio C. (1996) Ecologia de Peces. Secretariado de Publicaciones de la Universidad de Sevilla, Sevilla.

50

101 Harding A.E. (2006) Changes in Mediterranean Climate Extremes: Patterns, Causes, and Impacts of Change. PhD Thesis, University of East Anglia.

Hooke J.M. (2006) Human impacts on fluvial systems in the Mediterranean region. Geomorphology 79, 311 – 335.

Jongman, R.H.G.; Ter Braak, C.J.F. & Van Tongeren, O.F.R. (1995). Data analysis in community and landscape ecology. 2nd ed., Cambridge Univ. Press, Cambridge.

Kennard M.J., Arthington A.H., Pusey B.J. & Harch B.D. (2005). Are alien fish a reliable indicator of river health? Freshwater Biology 50, 174–193.

Lennon, J.J. (2000). Red-shifts and red herrings in geographical ecology. Ecography 23, 101–113.

Magalhães, M.F., Beja, P., Canas, C. & Collares-Pereira, M.J. (2002) Functional heterogeneity of dry-season fish refugia across a Mediterranean catchment: the role of habitat and predation. Freshwater Biology 47, 1919-1934.

Matthews W.J. & Marsh-Matthews E. (2003) Effects of drought on fish across axes of space, time and ecological complexity. Freshwater Biology 48, 1232–1253.

Miller D.L., Leonard P.M., Hughes R.M., Karr J.R., Moyle P.B., Shrader L.H., Thompson B.A., Daniels A.R, Fausch K.D., Fitzhugh G.A., Gammon J.R., Halliwell D.B., Angermeier P.L. & Orth D.J. (1988) Regional applications of an index of biotic integrity for use in water resource management. Fisheries 13, 12–20.

Moyle P.B. & Marchetti M.P. (1999) Applications of indices of biotic integrity to California streams and watersheds. In: T.P. Simon (ed.) Assessing the Sustainability and Biological Integrity of Water Resources Using Fish Communities. Boca Raton, FL: CRC Press, pp. 367–380.

Oliveira J.M. & Ferreira M.T. (2002) Development of an index of biotic integrity to assess environmental quality in warmwater streams. Revista de Ciencias Agrarias 25, 198–210.

Pires A.M., Cowx I.G. & Coelho M.M. (2001) Diet and growth of two sympatric Iberian barbel, Barbus steindachneri and Barbus microcephalus, in the middle reaches of the Guadiana Basin (Portugal). Folia Zoologica 50, 291–304.

Pont D., Hugueny B., Beier U., Goffaux D., Melcher A., Noble R., Rogers C., Roset N. & Schmutz S. (2006) Assessing river biotic condition at a continental scale: a

51

102 European approach using functional metrics and fish assemblages. Journal of Applied Ecology 43, 70–80.

Rivas-Martínez R., Sanchez-Mata D., Costa M. (1999) North American Boreal and Western Temperate Forest Vegetation. Itinera Geobotanica 12, 5-16.

Rivas-Martínez S. (2004). Global Bioclimatics (Clasificación Bioclimática de la Tierra). http://www.ucm.es/info/cif/book/bioc/bioc2.pdf.

Romero R, Guijarro J.A., Alonso S. (1998) A 30-year (1964-1993) daily rainfall data base for the Spanish Mediterranean regions: first exploratory study. International. Journal of Climatology 18, 541-560.

Ross S.T. (1991) Mechanisms structuring fish assemblages: are there lessons from introduced species? Environmental Biology of Fishes 30, 359–368.

Schmutz, S., U. Beier, J. Böhmer, J. Breine, N. Caiola, M.T. Ferreira, C. Frangez, D. Goffaux, G. Grenouillet, G. Haidvogl, J. de Leeuw, A. Melcher, R.A.A. Noble, J. Oliveira, N. Roset, I. Simoens, A. Sostoa & T. Virbickas. (2007) Spatially-based assessment of the ecological status in European ecoregions. Fisheries Management and Ecology 4, 441-452.

Statistical Sciences (1999) S-Plus, version 2000 for Windows. Mathsoft Inc., Seattle, WA.

Ter Braak, C.J.F. & Looman, C.W.N. (1986) Weighted averaging, logistic regression and the Gaussian response model. Vegetatio 65, 3–11.

Velasco J, Millán A, Vidal-Abarca M.R., Suárez M.L., Guerrero C. & Ortega M. (2003) Macrophytic, epipelic and epilithic primary production in a semiarid Mediterranean stream. Freshwater Biology 48, 1408-1420.

Vernaples W.N., & Ripley, B.D. (1997). Modern Applied Statistics with S-PLUS. Springer-Verlag, New York.

Vila-Gispert A., Moreno-Amich R. & Garcia-Berthou E. (2002) Gradients of life-history variation: an inter-continental comparison of fishes. Reviews in Fish Biology and Fisheries 12, 417–427.

Welsh, H.H., Hodgson G.R. In press. Amphibians as metrics of critical biological thresholds in forested headwater streams of the Pacific Northwest, USA. Journal of Freshwater Biology.

52

103 Whittier, T.R., Hughes, R.M., Lomnicky, G.A. & Peck, D.V. (2007) Fish and Amphibian Tolerance Values and an Assemblage Tolerance Index for Streams and Rivers in the Western USA. Transactions of the American Fisheries Society 136, 254-271.

53

104

http://efi-plus.boku.ac.at/

Project no.: 0044096 Project acronym: EFI+

Improvement and spatial extension of the European Fish Index

Instrument: STREP Thematic Priority: Scientific Support to Policies (SSP) - POLICIES-1.5

D 3.4, Task 3.7 – Large Floodplain rivers assessment

Due date of deliverable: 31.12.2007 Actual submission date: 18.04.2008

Start date of project: 01.01.2007 Duration: 24 Month

Organisation name of lead contractor for this deliverable: Leibniz-Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany Responsible author: Christian Wolter

Revision [draft 1]

Project co-funded by the European Commission within the Sixth Framework Programme (2002-2006)

Dissemination Level PU Public X PP Restricted to other programme participants (including the Commission Services) RE Restricted to a group specified by the consortium (including the Commission Services) CO Confidential, only for members of the consortium (including the Commission Services)

105 WP 3.7 Large floodplain rivers A separate large rivers data base has been compiled at IGB containing fishing data obtained by electric fishing as well as complementary methods like trawling, seining, fyke- and gillnetting. Additional sampling methods should be essentially considered for large rivers to representatively survey the mid channel section and type-specific potamal fish species. The data base was completed in December 2007 and contains 2730 data sets covering 18 river systems.

Pre-test A subset of 250 river stretches was evaluated using the existing German fish-based assessment scheme. This was used as a pre-test to elucidate the improvement of the assessment results by incorporating data from various sources as well as from complementary sampling gears. Both, the assessment result and the Ecological Quality Ratio (EQR) were highly significantly related to the number of species and individuals caught.

Table 1 Significant correlation results between the Ecological Quality Ratio (EQR) and species respectively individuals in a sample (** Correlation is significant at the 0.01 level (2- tailed)).

Species Individuals EQR Spearman's rho .712(**) .601(**) Sig. (2-tailed) .000 .000 Pearson Correlation .699(**) .430(**) Sig. (2-tailed) .000 .000 Sum of Squares and Cross-products 163.503 51994.758 Covariance .667 212.224 N 246 246

0.60

0.50

0.40

EQR 0.30

0.20

0.10

0.00 1.00 1.50 2.00 2.50 3.00 3.50 Assessment value (2.51 = GES)

Fig. 1 Correlation between assessment result and Ecological Quality Ratio (EQR) according to the German national fish-based assessment system for 250 sites (GES= good ecological status).

In addition, there was a slight indication, that the assessment results were biased towards the good ecological status, if the sample increased a threshold of about 10,000 individuals.

106 Therefore, the recursive segmentation procedure C&RT (Classification and Regression Trees) was used to separate the influence of species and individuals caught, respectively on the assessment results. At each classification level, this binary procedure divided the target variable (assessment value) in two homogenous subgroups with increasing homogeneity in the new subgroups, and indicated this variable which most contributed to the separation. The procedure stopped when complete homogeneity was reached. However, the procedure was terminated after the second segmentation level, because it becomes inefficient when the same predictor is selected twice.

There was a strong indication for species number as main predictor for the good ecological status, with 14 species per sample as significant threshold, whilst at a total of >20 species per sample the average site reached the quality aim of the WFD (≥2.51 represents the GES).

Node 0 Mean 2.157 Std. Dev. 0.528 n246 %100.0 Predicted 2.157 Species Improvement=0.114

<= 14 > 14 Node 1 Node 2 Mean 1.828 Mean 2.503 Std. Dev. 0.442 Std. Dev. 0.367 n126 n120 %51.2 % 48.8 Predicted 1.828 Predicted 2.503

SpeciesRatio Species Improvement=0.032 Improvement=0.015

<= 0.215 > 0.215 <= 20 > 20

Node 3 Node 4 Node 5 Node 6 Mean 1.371 Mean 1.965 Mean 2.306 Mean 2.658 Std. Dev. 0.216 Std. Dev. 0.398 Std. Dev. 0.330 Std. Dev. 0.319 n29n97n53 n67 %11.8%39.4%21.5 %27.2 Predicted 1.371 Predicted 1.965 Predicted 2.306 Predicted 2.658

Fig. 2 Result of the C&RT segmentation of the assessment result predicted by various sample characteristics. Significant predictors are shown. The procedure was terminated at the second level, as species had been selected the second time.

As indicated by the C&RT segmentation results, fish species number in the sample seemed to be a paramount impact on the result of the fish-based assessment. Even if representative sampling is almost a prerequisite for reliable assessments, these findings might imply that it becomes essentially to sample all present fish species even in large rivers. This would essentially require the implementation of additional sampling gears for mid channel potamal species in the survey design to properly evaluate large rivers.

107 Frequency of occurrence As a first step, the frequency of occurrence (% of samples containing a certain species) has been compared between standard electric fishing during the day and complementary fishing gears, like seines, trawls, gill- or trap nets. This analysis aimed to characterize the contribution of complementary gears to the species inventory obtained. It has been performed for all rivers in the data base with sufficient numbers of samples collected with both kinds of sampling gears.

For most of the species recorded, their frequencies in the catch significantly differed between standard electric fishing and alternative gears (Appendix Table 1). However, the number of species substantially more frequently recorded by additional gears ranged between 0 (Narew River, Poland) and 8 (Ijssel River, The Netherlands) with an average of 4.43 (± 3.29 standard deviation). In contrast, much more species have been more frequently recorded by standard electric fishing (Appendix Table 1). Further analyses focused on rivers with more than 100 samples in each strategy to account for rare species by keeping the contribution of a single record below 1%. The number of species exclusively caught by additional gears ranged between 1 (Meuse River) and 6 (Ijssel River), whilst the number of species exclusively caught by electric fishing during the day was in minimum 5 (Oder River) and in maximum 15 (Elbe River) (Fig. 3).

# species 16 electro other 14

12

10

8

6

4

2

0 Elbe (811) Ijssel (239) Meuse (319) Oder (339) Rhine (624) Coregonus peled Cobitis taenia Abramis sapa Coregonus maraena Osmerus eperlanus Hypophthalmichthys molitrix Carassius gibelio Vimba vimba Abramis sapa Coregonus lavaretus Osmerus eperlanus Petromyzon marinus Salmo trutta, (Sea-) Silurus glanis Abramis sapa Coregonus oxyrhynchus Coregonus oxyrhynchus

Fig. 3 Number of fish species exclusively caught by either electric fishing during the day or an alternative gear. Species caught by alternative gears only are indicated in blue.

By introducing a 1% criterion of rarity, i. e. omitting all species with less than one percent frequency, because they only accidentally occur in a sample, the number of exclusively recorded species substantially dropped to 0-4 in the alternative gears and to 4-10 in the electric fishing (Fig. 4). At a 5% criterion of rarity, in two rivers one species each remained exclusively for additional gears and 1-2 for electric fishing samples (Fig. 5)

108 # species 16 electro other 14

12

10

8

6

4

2

0 Elbe (811) Ijssel (239) Meuse (319) Oder (339) Rhine (624) Cobitis taenia Coregonus maraena Osmerus eperlanus Carassius gibelio Vimba vimba Abramis sapa Coregonus lavaretus Osmerus eperlanus Salmo trutta, (Sea-)

Fig. 4 Number of fish species with frequency >1% exclusively caught by either electric fishing during the day or an alternative gear. Species caught by alternative gears only are indicated in blue.

# species 16 electro other 14

12

10

8

6

4

2

0 Elbe (811) Ijssel (239) Meuse (319) Oder (339) Rhine (624) Coregonus maraena Osmerus eperlanus

Fig. 5 Number of fish species with frequency >5% exclusively caught by either electro fishing during the day or an alternative gear. Species caught by alternative gears only are indicated in blue.

109 In total, the contribution of additional sampling gears to the species inventory was surprisingly low compared to standard electric fishing during the day.

day electric fishing alternative fishing methods Elbe (316/495) 80

60

40

20

0 ns Ijssel (117/122) 80

60

40

)

s 20

le

p

m 0 a Maas (150/169) ns

s

%

( 80

e

c

n

e 60

r

r

u

c 40

c

o

f

o 20

y

c

n

e 0 u Oder (122/217)

q

e r 80 F ns

60

40

20

ns 0 Rhine (289/335) ns 80

60

40

20 ns

ns 0 i s a s la s s a a s g s is s s is s s a is ru m u il iu u n n iu n u il u u il u lu rc n a n u p b k e c li u t id n t s ti e a le r r g s r r a e n ia a ia le l l b u a e r lu r v s rl v f ru p g a lb n a b jo a b e u e io b is a s m x c lu c lu s s c s s a a u s b o R f s p f y u u i m s ll i u a s s / s a i e a h il lu r a u i p b c u E o lu r c s c t t r lu m r n u s r c n i a t u u r h u e i ra b r g A a li o b h e e r e ic R d S b A u n B B g o p p L e P t n b A e g e m la A l r c a m P a A o G o L s S C n O m y G

Fig. 6 Observed differences in species records obtained by electro fishing and other gears, respectively for the 19 most frequent species (ns= not significant, all other comparisons significantly different at the p<0.05 level of confidence, Fisher’s exact test). Rheophilic species are highlighted in blue. Number of samples in parentheses electric fishing / others.

110 day electric fishing alternative fishing methods Elbe (316/495) 80

60

40

20

0 ns Ijssel (117/122) 80

60

40

)

s 20

le

p

m 0 a Maas (150/169) ns

s

%

( 80

e

c

n

e 60

r

r

u

c 40

c

o

f

o 20

y

c

n

e 0 u Oder (122/217)

q

e r 80 F ns

60

40

20

ns 0 Rhine (289/335) ns 80

60

40

20 ns

ns 0 i s a s la s s a a s g s is s s is s s a is ru m u il iu u n n iu n u il u u il u lu rc n a n u p b k e c li u t id n t s ti e a le r r g s r r a e n ia a ia le l l b u a e r lu r v s rl v f ru p g a lb n a b jo a b e u e io b is a s m x c lu c lu s s c s s a a u s b o R f s p f y u u i m s ll i u a s s / s a i e a h il lu r a u i p b c u E o lu r c s c t t r lu m r n u s r c n i a t u u r h u e i ra b r g A a li o b h e e r e ic R d S b A u n B B g o p p L e P t n b A e g e m la A l r c a m P a A o G o L s S C n O m y G

Fig. 7 Observed differences in species records obtained by electro fishing and other gears, respectively for the 19 most frequent species (ns= not significant, all other comparisons significantly different at the p<0.05 level of confidence, Fisher’s exact test). Lithophilic species are highlighted in blue, psammophilic in brown. Number of samples in parentheses electric fishing / others.

However, among the 19 most frequently recorded species in large rivers, only six have been classified as rheophilic, which was considered of certain indicator value for habitat structures

111 and characteristics typical for more natural river systems. Three of them occurred – if anything – more frequently in the samples obtained by additional fishing gears (Fig. 6). Lithophilic fish are obligate gravel spawner with benthic larvae, and thus, of highest indicator value for hydrodynamic integrity of riverine habitats. Three lithophilic species belonged to the most frequently caught, of them only one – the river lamprey in the Rhine River – was predominantly caught by alternative gears. The two other lithophilic fish were more frequently caught by electric fishing, except for barbel in the Ijssel River (Fig. 7). The psammophilic gudgeons require sandy substrates for spawning. The have been most frequently caught by additional gears, which was probably because of the typically high abundance of the river gudgeon Romanogobio belingii in the mid channel sections of large lowland rivers. However, in the data set of Netherland’s large rivers, this species was not distinguished.

To summarize, eurytopic species were most frequent in samples from large rivers obtained by both standard electric fishing during the day and alternative fishing gears. Most species have been more frequently or exclusively obtained by electric fishing, while the exclusive contribution of additional gears to the species inventory was rather limited. However, three out of six indicator species for rheophilic conditions and one out of three for lithophilic were more frequently contributed by additional sampling gears.

Relative abundance The relative abundance of species has been compared between standard electric fishing during the day and complementary fishing gears to identify species which essentially require sampling by additional gears for their sufficient representation in the dominance structure. Means of relative abundance in the samples have been compared for all rivers in the data base with sufficient numbers of samples collected with both kinds of sampling gears. In most cases, the relative abundance of those species exclusively contributed by one method was very low (<0.1%), except the electric fishing results from the rivers Vistula and Narew (Appendix Table 2). Additional sampling gears performed highly significantly better for the abundance estimation of common bream and silver bream. They also performed significantly better for catching blue bream, whitefish, smelt or pikeperch, however, these pattern were more heterogeneous, the species much lesser abundant in general, respectively rare or limited to very few rivers. In two cases the comparably high abundance of tench and rude respectively resulted from the exposition of additional gillnets in backwaters. Corresponding to the results of the frequency analysis, most species have been recorded in higher abundances by standard electric fishing along the banks (Appendix Table 2).

Looking at the rheophilic indicator species, gudgeons (probably only river gudgeon, which has not been distinguished in the data from The Netherlands) and asp in the rivers Elbe and Vistula were caught in higher abundance by alternative gears, whilst burbot and ide were more abundant in the electric fishing samples (Fig. 8). Correspondingly, asp was the only lithophilic indicator species with higher abundance recorded by using additional gears (Fig. 9). In general only two lithophilic and psammophilic species belonged to the most abundant in the large river data set analysed.

112 )

%

(

e

c

n

a

d

n

u

b

a

e

iv

t

la

e

R

a s s a a o i a a na lus us ili ius c ll c us k i ius n uu c ui ingi in lot am r ut p r n idus iat u er t L an br r v l gobi el T l oe as bu cer L lu E iop ang G b er A bj R A al f R B A G P luc A ep S O

Fig. 8 Observed differences in relative abundance of the most numerous species in samples obtained by electric fishing and other gears. Rheophilic species are highlighted in blue.

113 )

%

(

e

c

n

a

d

n

u

b

a

e

iv

t

la

e

R

a s s a a o i a a na lus us ili ius c ll c us k i ius n uu c ui ingi in lot am r ut p r n idus iat u er t L an br r v l gobi el T l oe as bu cer L lu E iop ang G b er A bj R A al f R B A G P luc A ep S O

Fig. 9 Observed differences in relative abundance of the most numerous species in samples obtained by electric fishing and other gears. Lithophilic species are highlighted in blue, psammophilic in brown.

In total, large rivers were dominated by eurytopic, phyto-lithophilic species. Indicator species with higher respectively more structurally specialized habitat requirements were comparably rare amongst the most frequent and abundant fish. Accordingly, in large rivers there was only a limited number of indicator species for a fish-based assessment available. Unfortunately, alternative sampling methods did not significantly improve the amount of indicator species.

114 Further attempts have to be made to analyse the assessment power of the results obtained by electric fishing only for the evaluation of structural deficits in large rivers. Metrics based on presence absence or abundance thresholds of mid channel potamal species would require a sampling effort similar to electric fishing except for common bream, silver bream, river gudgeon, pike perch, and diadromous fish. However, the latter were so rare in general, that it seemed most suitable to assess these species at their upstream spawning stretches instead of the main channel migration corridor.

Floodplain assessment The fish-based assessment of the integrity of floodplains requires the additional survey of floodplain water bodies, permanently and/or temporarily with the main channel connected still waters. In the data base they are typically represented as backwaters surveyed together with the main channel sites. This type of connected backwaters represents only a part of the typical floodplain, the frequently inundated ecotones. Older stages of succession, water bodies naturally connected to the main channel less than 20 years on average were not represented in the data base. A detailed comparison of floodplain water bodies in different stages of succession has been performed at the lower Oder River. In the National Park “Lower Oder Valley” the water bodies in the non-floodable polder have been flooded for the last time in 1947. Therefore, these water bodies represent typical floodplain waters in a stage of 60 years of succession.

In general, highly significant differences between floodplain water bodies have been detected at the level of functional guilds. A significant dispersion of rheophils and limnophils (both species numbers and abundance) has been detected along the gradient of lateral connectivity. Main channel samples contained on average 5.1 (± 1.0 standard deviation) rheophilic species, those of frequently flooded water bodies 2.7 (± 1.7). Also the abundance of rheophilic fish dropped significantly with the flooding frequency of the water body (Fig. 10).

10

80 8

60 6

40 4

# rheophilic species 20 2

Rheophilis relative abundance (%) 0 0 main channel floodable non-floodable main channel floodable non-fl oodable pol der pol der polder pol der

Fig. 10 Number and abundance of rheophilic fish in the main channel and both floodplain water body types. All differences were significant at p<0.05 (ANOVA, post hoc Dunnett T3).

Functionally intact floodplains are characterized by well developed ecotone connectivity as well as the availability of stagnant water bodies or temporarily inundated terrestrial areas. Samples in water bodies representative for floodplain integrity will yield significantly lower

115 numbers and amounts of rheophilic fish. However, it is highly difficult to distinguish a drop in rheophils abundance due to well developed floodplain connectivity from those due to human impacts like flow regulation, siltation, damming etc. from the data base. While number and abundance of eurytopic fish did not significantly differ between the habitats, those of limnophils were inversely related to the inundation frequency, and patterned opposite to the rheophils (Fig. 11).

10

80 8

60 6

40 4

# limnophilic species

2 20

Limnophilis relative abundance (%) 0 0 main channel floodable non-fl oodable m ain channel floodabl e non-floodable polder pol der polder polder

Fig. 11 Number and abundance of limnophilic fish in the main channel and both floodplain water body types. All differences were significant at p<0.05 (ANOVA, post hoc Dunnett T3).

The amount of limnophilic and correspondingly of phytophilic fish were found to be highly indicative for ecotone connectivity and seemed to be a suitable metric for functionally intact floodplains. However, similar to the typical decline of rheophils in relation to lower inundation frequencies, the correlated raise of limnophils is potentially indicative for both floodplain integrity and human impacts. Lowered flow dynamics, e.g. by damming, might increase limnophils in the same way like improved ecotone connectivity.

Backwaters and floodplain water bodies provide environmental conditions similar to lakes with the ability of species exchange via permanent and/or temporary connections. Degraded rivers, especially flow regulated and dammed, increasingly provide habitats corresponding to stagnant water bodies, which improves their suitability for eurytopic and limnophilic fish. The resulting fish assemblage structure may be similar to floodplain waters or backwaters. This makes it nearly impossible to distinguish intact floodplain sites from degraded river sites just from the data set, without knowing further site details. Similarly, the differentiation potential of metrics for certain pressures is much lower when calibrated for both situations in the same way. However, normally a site will not be evaluated as a black box, so that the site characteristics and pressures are well known. Therefore, the floodplain could be evaluated separately, to not mix the contrary indications of the same metric in the main channel. It is suggested to evaluate backwaters and floodplain water bodies distinct from the main river, preferably by using metrics based on limnophilic and phytophilic fish.

116 Table 1 Comparison of fish species frequencies (%) obtained by electro fishing during the day and other gears, respectively. Significant differences (Fisher’s exact test, p<0.05) are marked in bold coloured; total= number of samples.

Species Elbe (811) Ijssel (239) Meuse (319) Narew (44) Oder (339) Rhine (624) Vistula (30) electro other electro other electro other electro other electro other electro other electro other Total 316 495 117 122 150 169 31 13 122 217 289 335 21 9 Abramis ballerus 4.43 16.97 0.82 41.47 Abramis brama 66.14 77.78 43.59 92.62 49.33 97.04 29.03 53.85 56.56 74.65 49.48 97.01 61.90 88.89 Abramis sapa 0.00 0.82 0.00 0.59 0.00 1.49 4.76 0.00 Alburnoides bipunctatus 14.29 0.00 Alburnus alburnus 84.18 23.84 47.01 18.85 34.00 11.24 61.29 0.00 72.95 23.50 45.33 12.84 95.24 0.00 Ameiurus nebulosus 1.90 1.01 Anguilla anguilla 81.33 5.05 70.94 12.30 86.00 76.33 21.31 4.15 61.94 20.00 4.76 0.00 Aspius aspius 65.82 49.29 19.66 8.20 46.67 10.65 0.00 7.69 27.05 4.15 32.53 18.21 57.14 33.33 Barbatula barbatula 1.27 0.00 0.85 0.00 2.00 0.00 0.82 0.00 0.35 0.00 66.67 0.00 Barbus barbus 10.13 0.81 6.84 18.03 19.33 0.00 0.82 0.00 13.15 16.42 47.62 11.11 Blicca bjoerkna 69.62 77.58 32.48 84.43 11.33 70.41 77.42 23.08 68.85 71.43 20.07 83.88 71.43 22.22 Carassius carassius 3.80 0.00 2.56 0.00 35.48 7.69 2.46 0.46 1.04 0.00 Carassius gibelio 4.43 0.81 0.00 1.64 1.33 0.00 9.68 7.69 0.82 0.46 0.69 0.00 14.29 33.33 Chelon labrosus 4.67 0.59 0.35 0.00 Chondrostoma nasus 3.16 0.40 6.84 8.20 10.67 2.37 9.34 3.88 9.52 11.11 Cobitis taenia 9.49 0.00 0.00 2.46 2.67 0.00 38.71 0.00 66.39 0.00 1.04 0.30 66.67 0.00 Coregonus lavaretus 0.00 1.64 Coregonus maraena 0.00 18.43 Coregonus oxyrhynchus 0.00 0.82 0.00 0.30 Coregonus peled 0.00 0.20 Cottus gobio 2.22 0.00 5.13 22.95 15.33 14.20 7.61 6.87 Ctenopharyngodon idella 2.00 0.00 Cyprinus carpio 3.16 0.40 6.84 0.82 17.33 3.55 4.10 0.92 6.23 0.90 4.76 0.00 Dicentrarchus labrax 1.33 0.00 1.38 0.00 Esox lucius 59.18 17.17 29.06 3.28 37.33 6.51 93.55 38.46 77.05 11.98 28.37 4.78 90.48 0.00 Gasterosteus aculeatus 1.58 0.00 18.80 4.92 20.67 4.73 6.45 0.00 7.38 0.46 13.49 1.19 23.81 0.00 Gobio gobio 76.58 3.03 7.69 50.00 9.33 44.97 45.16 0.00 42.62 0.46 12.11 60.60 90.48 0.00 Gymnocephalus cernuus 39.24 31.52 49.57 81.15 37.33 88.76 16.13 0.00 71.31 16.59 50.17 71.94 28.57 0.00 Hypophthalmichthys molitrix 0.00 0.20

117 Hypophthalmichthys nobilis 0.32 0.00 Lampetra fluviatilis 1.27 0.00 3.42 5.74 2.00 4.73 5.19 16.42 Leucaspius delineatus 1.58 0.00 0.67 0.00 16.13 0.00 3.28 0.00 0.69 0.00 Leuciscus cephalus 85.44 8.89 14.53 0.00 20.00 0.59 86.07 7.37 29.41 0.30 76.19 0.00 Leuciscus idus 94.94 31.31 72.65 51.64 91.33 29.59 80.65 38.46 74.59 32.26 69.20 40.60 90.48 66.67 Leuciscus leuciscus 58.86 1.01 11.11 0.82 6.67 1.18 6.45 0.00 29.51 1.84 14.88 2.39 80.95 0.00 Lota lota 54.43 0.20 80.65 0.00 94.26 15.21 52.38 0.00 Misgurnus fossilis 32.26 0.00 7.38 0.00 Neogobius gymnotrachelus 19.05 0.00 Neogobius melanostomus 0.35 0.00 Oncorhynhus mykiss 1.27 0.00 Osmerus eperlanus 12.82 45.08 2.00 36.69 0.00 2.30 0.00 5.07 Perca fluviatilis 93.35 28.69 68.38 46.72 89.33 75.74 80.65 30.77 94.26 16.59 67.82 58.51 100.00 22.22 Perccottus glenii 23.81 0.00 Petromyzon marinus 0.00 0.60 Phoxinus phoxinus 0.63 0.00 Platichthys flesus 3.42 40.98 52.67 77.51 21.45 35.52 Proterorhinus marmoratus 3.33 0.59 4.15 4.48 Pseudorasbora parva 14.29 0.00 Pungitius pungitius 0.63 0.00 3.42 0.82 1.33 1.18 2.08 0.00 Rhodeus amarus 0.32 0.00 2.56 0.00 4.00 0.00 48.39 0.00 4.15 0.30 38.10 0.00 Romanogobio belingii 25.32 1.21 36.89 64.06 4.76 0.00 Rutilus rutilus 92.09 70.30 94.87 93.44 92.67 92.31 100.00 53.85 83.61 44.24 87.89 87.76 100.00 22.22 Salmo salar 2.22 0.81 Salmo trutta, Meer- 0.00 1.64 1.33 0.00 0.35 0.30 Salmo trutta, Bach- 0.95 0.00 Salvelinus fontinalis 0.32 0.00 Sander lucioperca 25.95 15.15 21.37 64.75 46.00 95.86 19.67 59.45 37.02 85.07 42.86 0.00 Scardinius erythrophthalmus 18.67 0.61 9.40 0.00 16.67 0.00 61.29 7.69 18.85 34.56 9.00 1.19 28.57 0.00 Silurus glanis 2.53 0.20 0.67 1.18 6.56 19.82 0.00 0.60 14.29 0.00 Thymallus thymallus 1.58 0.00 Tinca tinca 5.70 1.21 3.42 0.00 2.00 0.00 70.97 61.54 9.84 0.46 7.96 0.00 14.29 11.11 Vimba vimba 0.95 0.20 0.00 4.61 4.76 11.11

118 Table 2 Comparison of mean relative abundance (%) obtained by electric fishing during the day and other gears, respectively. Significant differences (Students t-statistics, at least p<0.05) are marked in bold coloured.

Elbe Ijssel Meuse Narew Oder Rhine Vistula electro other electro other electro other electro other electro other electro other electro other Abramis ballerus 0.05 2.10 0.01 6.55 Abramis brama 5.17 32.37 2.96 25.91 1.84 40.66 0.43 14.89 1.23 20.74 4.29 40.85 1.13 79.60 Abramis sapa 0.00 0.01 0.00 0.00 0.00 0.02 0.00 0.00 Alburnoides bipunctatus 0.03 0.00 Alburnus alburnus 9.29 5.27 5.28 0.19 2.05 0.32 2.80 0.00 5.99 1.38 7.37 0.15 29.57 0.00 Ameiurus nebulosus 0.14 0.39 Anguilla anguilla 3.79 0.30 12.42 0.12 13.40 2.98 0.15 0.09 8.94 1.35 0.03 0.00 Aspius aspius 1.38 6.14 0.48 0.05 1.62 0.05 0.00 0.59 0.16 0.05 1.55 0.17 0.40 1.09 Barbatula barbatula 0.01 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.01 0.00 2.06 0.00 Barbus barbus 0.18 0.03 0.26 1.81 0.60 0.00 0.00 0.00 0.71 1.10 0.22 0.38 Blicca bjoerkna 6.69 24.96 1.70 19.36 0.39 1.18 9.21 8.01 7.74 33.09 1.64 13.90 1.43 1.19 Carassius carassius 0.03 0.00 0.01 0.00 0.37 1.33 0.00 0.00 0.01 0.00 Carassius gibelio 0.02 0.00 0.00 0.00 0.02 0.00 0.15 1.92 0.00 0.01 0.00 0.00 0.05 1.92 Chelon labrosus 0.07 0.00 0.00 0.00 Chondrostoma nasus 0.03 0.04 0.51 0.10 0.21 0.01 0.42 0.03 0.02 0.29 Cobitis taenia 0.10 0.00 0.00 0.00 0.04 0.00 0.92 0.00 3.54 0.00 0.01 0.00 2.13 0.00 Coregonus lavaretus 0.00 0.02 Coregonus maraena 0.00 0.57 Coregonus oxyrhynchus 0.00 0.02 0.00 0.00 Coregonus peled 0.00 0.00 Cottus gobio 0.01 0.00 0.12 0.66 0.18 0.55 0.11 0.15 Ctenopharyngodon idella 0.02 0.00 Cyprinus carpio 0.01 0.00 0.12 0.02 0.93 0.01 0.04 0.01 0.25 0.00 0.00 0.00 Dicentrarchus labrax 0.00 0.00 0.02 0.00 Esox lucius 1.28 1.61 1.77 0.00 0.63 0.06 9.02 15.06 3.97 0.82 0.90 0.05 1.38 0.00 Gasterosteus aculeatus 0.02 0.00 0.86 0.06 0.43 0.01 0.20 0.00 0.09 0.01 0.45 0.02 1.15 0.00 Gobio gobio 6.96 0.27 0.08 3.03 0.07 1.98 2.52 0.00 1.00 0.00 0.47 6.00 6.89 0.00 Gymnocephalus cernuus 1.02 4.05 3.86 14.07 1.60 6.74 0.18 0.00 5.20 2.88 3.90 6.79 0.08 0.00 Hypophthalmichthys molitrix 0.00 0.00 Hypophthalmichthys nobilis 0.00 0.00

119 Lampetra fluviatilis 0.00 0.00 0.09 0.16 0.00 0.05 0.29 1.13 Leucaspius delineatus 0.01 0.00 0.02 0.00 2.68 0.00 0.03 0.00 0.01 0.00 Leuciscus cephalus 6.32 0.75 0.57 0.00 0.20 0.00 3.19 0.16 1.55 0.00 5.44 0.00 Leuciscus idus 16.57 2.95 7.61 2.35 16.31 0.58 3.49 5.53 4.24 2.42 13.00 1.52 4.53 3.41 Leuciscus leuciscus 2.20 0.04 0.28 0.05 0.08 0.00 0.08 0.00 0.40 0.02 0.79 0.01 2.53 0.00 Lota lota 1.23 0.01 7.24 0.00 26.38 1.08 0.29 0.00 Misgurnus fossilis 0.66 0.00 0.04 0.00 Neogobius gymnotrachelus 1.13 0.00 Neogobius melanostomus 0.01 0.00 Oncorhynhus mykiss 0.01 0.00 Osmerus eperlanus 4.81 10.92 0.01 0.29 0.00 0.03 0.00 0.04 Perca fluviatilis 16.90 2.38 8.61 1.36 18.61 7.07 6.62 5.28 11.83 1.08 13.60 2.51 11.71 5.59 Perccottus glenii 0.38 0.00 Petromyzon marinus 0.00 0.01 Phoxinus phoxinus 0.00 0.00 Platichthys flesus 0.07 0.89 5.10 6.41 1.52 4.71 Proterorhinus marmoratus 0.01 0.00 0.50 0.12 Pseudorasbora parva 0.03 0.00 Pungitius pungitius 0.01 0.00 0.10 0.02 0.01 0.00 0.06 0.00 Rhodeus amarus 0.00 0.00 0.02 0.00 0.01 0.00 1.89 0.00 0.18 0.01 0.66 0.00 Romanogobio belingii 0.99 0.12 5.77 16.40 0.13 0.00 Rutilus rutilus 18.99 15.15 45.84 17.31 32.74 12.24 47.02 16.06 18.31 3.41 35.10 13.86 26.00 6.13 Salmo salar 0.01 0.02 Salmo trutta, Meer- 0.00 0.10 0.03 0.00 0.04 0.00 Salmo trutta, Bach- 0.01 0.00 Salvelinus fontinalis 0.00 0.00 Sander lucioperca 0.33 0.98 0.64 1.38 2.41 18.79 0.30 4.22 1.81 5.45 0.11 0.00 Scardinius erythrophthalmus 0.17 0.02 0.93 0.00 0.32 0.00 2.87 0.43 0.28 2.99 0.42 0.01 0.24 0.00 Silurus glanis 0.01 0.00 0.00 0.00 0.04 0.46 0.00 0.01 0.02 0.00 Thymallus thymallus 0.01 0.00 Tinca tinca 0.03 0.03 0.02 0.00 0.02 0.00 1.66 29.62 0.06 0.00 0.07 0.00 0.23 0.20 Vimba vimba 0.00 0.00 0.00 0.06 0.01 0.20

120 1

Low Species Rivers

BADY P., LOGEZ M., D. PONT & VESLOT J.

CEMAGREF. HYAX UNIT. Aix en Provence

Report on February 2008

Contents

1. Dataset definition 1.1 Calibration and reference sites datasets 1.2 Fish length data 1.3 Concllusion

2. Environmental characteristics of low species rivers 2.1 Typology of sites according to richness 2.2 Environment 2.3 Conclusion

3. Modelling river size 3.1 Motivation 3.2 Description of the variables 3.3 Modelling the width 3.4 Validation on the reference dataset 3.5 Conclusion

4. Development of new petrics based on age classes 4.1 Methodology 4.2 First results 4.3 Conclusion

References Appendix A to L

1. Datasets definition

121 2

1.1 Calibration and Reference Datasets

Among the total number of fishing occasions, we first only consider one fishing occasion per site (aleatory procedure) and only sites where the following environmental variables are fulfilled. Environmental variables: "Geomorph.river.type" "Size.of.catchment" "Flow.regime" "Altitude" "Geological.typology" "temp.ann" "temp.jan" "temp.jul" "Actual.river.slope" "Water.source.type" "Floodplain.site" "Natural.sediment" "Lakes.upstream" "Sampling.strategy" "Method" "Fished.area" "STRAHLER" "PREC.AN.du” "forest25.du" "urban.du" "ECOREG"

We then retain a total of 10,158 sites, distributued among the following countries.

AT CH DE ES FI FR HU IT LT NL PL PT RO SE UK 840 601 782 1735 220 971 165 498 109 121 866 922 239 522 1567

For these sites, the following pressure variables are in general fulfilled. "Barriers.catchment.down", "Barriers.river.segment.up", "Barriers.river.segment.down", "Impoundment", "Hydropeaking", "Water.abstraction", "Hydro.mod", "Temperature.impact", "Velocity.increase", "Reservoir.flushing", "Sedimentation", "Channelisation", "Cross.sec", "Instream.habitat", "Riparian.vegetation", "Embankment", "Floodprotection", "Floodplain", "Toxic.substances", "Acidification", "Water.quality.index", "Eutrophication", "Organic.pollution", "Organic.siltation", "Navigation", "Colinear.connected.reservoir"

Sites with some gaps in the pressure variables are accepted at that step, in order to not exclude a country from our selection: - missing values in Switzerland for "Colinear.connected.reservoir" and "Toxic.substances" - missing values in United Kingdom for "Riparian vegetation"Temperature.impact", "Velocity.increase", "Organic.siltation", "Floodplain" and "Navigation". - in most of case, missing values

122 3

- missing values for "Water.quality.index" and "Toxic.substances" in a large part of French sites

Reference sites

Sites considered as reference sites are sites without any local pressure, except alteration of the flood plain (presence of flood protection works). But sites for which a barrier is present downstream are not excluding.

We then retain 459 reference sites. Most of these reference sites are situated in Spain and Romania AT ES FI FR IT LT PL RO SE 1 193 14 26 53 13 13 133 13

Calibration sites

In order to calibrate our models on a more representative dataset, we selected a calibration dataset. For these sites, the following pressures are accepted: - no or presence of partial barrier downstream and/or upstream from the segment - no or Weak water abstraction - no or intermediate level for channelization - no “Impoundment” - no “Hydropeaking” - no “Reservoir flushing” - no alteration of “Cross section” - no or weak “Sedimentation” - no alteration of “Instream habitat” - no or intermediate level for “Toxic substances” - no “Acidification” - Water quality classes 1 or 2 - no or missing values for “Velocity increase” - no or missing values for “Colinear.connected.reservoir” - no or only local “Embankment” - no or slight alteration of “Riparian vegetation” (or missing values) - no or missing values for “Temperature Impact” - no “Hydrological alteration” - no or low level of “Eutrophication” - no or Weak level of “Organic pollution” - no or missing values for “Organic siltation”

The calibration dataset considers 959 sites, without including reference sites

AT DE ES FI FR HU IT LT PL PT RO SE UK 49 21 372 44 89 6 58 20 124 2 34 112

123 4

1.2. Fish length Data

The aim of this part is to give a detailled description of the available fish length data. These data are included in the table “Length” from our common European database. Fish length data are described for both all sites, reference sites and calibration sites.

1.2.1 Fish length for all sites

General presentation of the table “Length”

This table contains 7 columns: Length_id (id of the line), Site_code (the id of the sampling site), Date (the date of sampling), Species (the names of the species), Run (the number of the run where fishes were caught), Total_length (the total length of fishes) and Number_of_individuals (the number of fishes sharing the same length). This table contains 2,195,914 rows; each one represents the number of individuals of species Y having a length of X millimetres, sampled at the sampling site Z at the date D. The data covers thirteen countries: Austria (AT), Switzerland (CH), Germany (DE), Spain (ES), France (FR), Italy (IT), Lithuania (LT), Netherlands (NL), Poland (PL), Portugal (PT), Romania (RO), Sweden (SE) and United Kingdom (UK). No data are available for Finland. The data represent: - 9,422 sites on the 14,221 sites occurring in the data base; - 23,204 fishing occasions on the 29,509 total ones; - 146 species on the 162 species sampled (omitting hybrids); - 6,291,717 individuals on the 7,706,588 fishes caught (Table 1). Those fishes were measured across the period 1955-2007. However, seventy percent of individuals were measured during the last decade. The sixteen species for which we don’t have lengths are very rare; their catches are ranging from 1 to 1,850 (it is to say less than 0.03% of the total catches).

Number of Fishing Number of Number of Countries Number of sites occasions species lengths Austria 938 1,172 57 326,032 Switzerland 539 667 35 171,583 Germany 803 1,817 57 648,243 Spain 1,902 2,210 47 233,337 France 1,145 6,542 64 3,867,694 Italy 471 785 57 62,847 Lithuania 115 129 37 17,801 Netherlands 159 748 47 135,866 Poland 636 687 52 73,140 Portugal 421 421 37 60,431 Romania 198 219 37 27,722 Sweden 598 5,345 41 425,910 United Kingdom 1,497 2,462 22 241,111 Table 1: Distribution of sites, fishing occasions, species and lengths per country.

Lengths were calculated with four methods: - Total lengths, fishes were sized on the field, for some of them the real total length (from the head to the end of the caudal fin) was measured and for

124 5

others it was the fork length (from the head to the beginning of the caudal fin) which was translated in total length with an equation; - Classes, the sizes of fishes were divided into different classes with a constant interval between classes, and each fish was assign to a class; - Min-Max, only the minimal and the maximal sizes of a species was reported as a size (concern very few individuals); - Subsample, not all lengths were measured but only a sample of the fishes caught. With this classification we can see that fish lengths arise from are very heterogeneous sources. Inside the category “Classes”, there is also different type of classes. The interval between classes can be very different: some fishing occasions exhibit only three classes with 125 mm between them (100, 225 and 350 mm); others can have interval of 5mm between classes (70, 75, etc.). Moreover for some fishing occasions you can have a mix of both total individual lengths and classes. For example, all fishes less than 100 mm could have been measured precisely and all fish over 100 mm could have been assigned to a size class: 105, 110 mm…

Lengths This table contains lengths for 146 species and 6,291,717 fishes for all runs and 5,020,994 fishes for the first run (after removing all hybrids and all individuals with undetermined species). Considering all the runs, the number of species length available for each species is ranging from 1 to 926,908 whereas is ranging from 1 to 721,930 considering only the first run. For all the runs, the ten most abundant species are: minnow, brown trout, roach, gudgeon, stone loach, bullhead, chub, bleak, perch and Atlantic salmon (Table 2). For only the first run, the ten most abundant species are: minnow, roach, brown trout, gudgeon, chub, stone loache, bullhead, bleak, perch and eel (Table 3).

Species names Number of lengths Phoxinus phoxinus 926,908 Salmo trutta fario 876,033 Rutilus rutilus 713,614 Gobio gobio 575,677 Barbatula barbatula 441,269 Cottus gobio 392,365 Leuciscus cephalus 381,197 Alburnus alburnus 262,827 Perca fluviatilis 165,642 Salmo salar 143,633 Table 2: The number of lengths available for the ten most abundant species, for all runs.

Species names Number of lengths Phoxinus phoxinus 721,930 Rutilus rutilus 662,634 Salmo trutta fario 626,009 Gobio gobio 479,851 Leuciscus cephalus 339,192 Barbatula barbatula 337,856 Cottus gobio 262,175 Alburnus alburnus 252,676 Perca fluviatilis 157,214 Anguilla anguilla 102,670

125 6

Table 3: The number of lengths available for the ten most abundant species, first run.

The number of lengths for each species for all runs is detailed in appendix A and in appendix B for the first run. The five most abundant species (in number of lengths) per country are detailed in appendix C for all runs and in appendix D for the first run.

Lengths of fishes were collected during different runs. The number of runs is ranging from 1 to 5. The majority of lengths (79.8%) were measured during the first run. There are also 6% of lengths for which we don’t have information concerning the run. Spain and Sweden are the only two countries concerned buy this lake of information (Table 4). It represented 15 and 80% respectively of the total of lengths for those countries. If we take into account only lengths measured during the first run the percent of remaining lengths varies greatly between countries (Table 5). In Sweden due to the lack of information 16% of the lengths remained if we take into account only the first run. No Data 1 2 3 4 5 Austria 0 270,511 42,173 13,348 0 0 Switzerland 0 121,236 37,401 12,906 40 0 Germany 0 648,243 0 0 0 0 Spain 35,210 130,173 48,856 18,896 193 9 France 0 3,233,453 630,539 3,686 16 0 Italy 0 59,073 3,569 205 0 0 Lithuania 0 17,801 0 0 0 0 Netherlands 0 135,866 0 0 0 0 Poland 0 73,140 0 0 0 0 Portugal 0 60,431 0 0 0 0 Romania 0 27,722 0 0 0 0 Sweden 342,930 69,345 9,426 4,209 0 0 United Kingdom 0 174,000 47,369 19,265 477 0 Table 4: Distribution of length between runs and country.

AT CH DE ES FR IT LT NL PL PT RO SE UK 82.97 70.66 100 55.79 83.6 93.99 100 100 100 100 100 16.28 72.17 Table 5: percent of lengths measured during the first run

On other important issue for lengths is the date of fishing, especially for young of the year and juveniles. The date reflects indirectly the amount of energy collected by fishes during their annual growing period. We expect young fish to be bigger in a fishing occasion occurring in the end of year than fishes caught earlier in the year. Two-thirds of the fishing occasions were sampled between August and October, representing also two thirds of the lengths available (both for all runs and first run, Table 6).

Number of Number of Number of Months fishing occasions lengths all runs lengths run 1 January 101 17,035 15,609 February 97 12,341 11,094 March 514 61,618 58,214 April 929 156,751 141,777 May 1,356 368,064 321,795 June 2,076 757,917 633,925 July 2,598 561,556 420,377 August 4,861 1,124,631 805,940

126 7

September 6,398 2,129,885 1,670,756 October 3,179 878,877 745,602 November 927 191,129 167,364 December 168 31,913 28,541 Table 6: Distribution of fishing occasions and lengths per months.

The distribution of fishing occasions (by the way the number of lengths) across the year is very different between countries (Table 7 and Table 8). If we consider only August, September and October, the proportion of available lengths varies greatly between countries. For Italy and Portugal the number of lengths for this period only represents less than 10% of the total lengths provided. On the contrary for Lithuania, Romania and Sweden it represents more of 80% of the lengths provided by those countries (Table 9).

January February March April May June July August September October November December AT 3 16 49 60 47 84 124 129 211 268 136 45 CH 0 0 0 0 9 10 124 233 139 126 25 1 DE 8 4 39 162 275 114 72 311 370 275 163 24 ES 2 12 55 49 47 87 189 396 725 512 110 26 FR 4 3 14 174 520 1,230 665 851 2,158 829 83 11 IT 66 47 92 65 53 85 47 9 71 29 176 45 LT 0 0 0 0 0 1 27 45 52 4 0 0 NL 0 2 235 171 43 0 0 0 32 195 69 1 PL 0 0 0 2 23 55 61 72 186 249 39 0 PT 8 0 8 49 106 95 120 31 4 0 0 0 RO 0 7 0 0 2 0 6 95 56 24 29 0 SE 0 1 3 5 12 41 709 2,183 1,911 445 26 9 UK 10 5 19 192 219 274 454 506 483 223 71 6 Table 7: Distribution of fishing occasions per country across the year.

January February March April May June July August September October November December AT 634 5,524 10,652 19,396 7,722 18,612 39,654 36,377 58,757 87,805 31,609 9,290 CH 0 0 0 0 2,932 2,346 21,820 63,030 34,089 34,324 12,959 83 DE 8,157 1,076 3,007 26,554 64,695 35,036 20,879 156,108 161,296 114,231 49,119 8,085 ES 10 751 4,464 3,441 5,046 6,181 26,419 47,308 88,688 41,770 7,782 1,477 FR 1,536 686 5,942 54,773 224,780 640,627 319,887 555,437 1,531,822 481,112 46,144 4,948 IT 5,322 3,949 4,639 3,767 5,837 10,338 3,356 868 5,280 1,979 11,191 6,321 LT 0 0 0 0 0 82 2,343 6,515 8,119 742 0 0 NL 0 45 32,310 22,922 12,774 0 0 0 6,287 43,884 16,961 683 PL 0 0 0 42 4,729 9,009 9,396 12,927 14,847 20,217 1,973 0 PT 710 0 148 4,956 18,366 10,468 19,947 5,405 431 0 0 0 RO 0 201 0 0 130 0 411 10,141 11,338 2,023 3,478 0 SE 0 7 42 164 499 2,001 50,983 184,284 155,641 30,030 1,742 517 UK 666 102 414 20,736 20,554 23,217 46,461 46,231 53,290 20,760 8,171 509 Table 8: Distribution of lengths per country across the year for all runs.

AT CH DE ES FR IT LT NL PL PT RO SE UK 56.1 76.6 66.6 76.2 66.4 12.9 86.4 36.9 65.6 9.7 84.8 86.9 49.9 Table 9: Proportion of lengths for the period August-October.

127 8

1.2.2 Fish length for reference sites Introduction

On the 459 reference sites, fish lengths are available for 364 of them. The 364 sites are located in eight countries: Austria, Spain, France, Italy, Lithuania, Poland, Romania and Sweden. No data is available for Finland. The 364 sites include 460 fishing occasions, 59 species and 44,894 lengths during the period 1986-2007. Sixty percent of the lengths are located in Spain and Romania (Table 10). Number of Number of sites with Number of fishing Number of Number of sites lengths occasions with lengths species lengths

Austria 1 1 1 5 49 Spain 193 135 145 17 12,945 Finland 14 0 0 0 0 France 26 26 48 7 6,313 Italy 53 51 68 15 4,212 Lithuania 13 13 23 29 5,170 Poland 13 12 12 6 499 Romania 133 115 126 21 14,238 Sweden 13 11 37 11 1,468 Table 10: Distribution of reference sites per country.

Lengths Regarding reference sites, this table contains lengths for 59 species and 44,894 fishes taking into account all runs, where as it contains lengths for 58 species and 37,860 individuals for the first run. Without distinction of runs, number of length per species is ranging from 1 to 21,895, whereas for the first run it is ranging from 1 to 16,002. The ten most abundant species are the same if we consider all runs or only the first run: brown trout, minnow, stone loach, barbell, chub, stream bleak, bullhead, gudgeon, golden spined loach and roach (Table 11 and 12). Species Lengths Salmo trutta fario 21,895 Phoxinus phoxinus 5,681 Barbatula barbatula 2,902 Barbus petenyi 2,347 Leuciscus cephalus 1,853 Alburnoides bipunctatus 1,535 Cottus gobio 1,522 Gobio gobio 1,508 Sabanejewia balcanica 929 Rutilus rutilus 681 Table 11: Number of lengths for the ten most abundant species for all runs.

Species Lengths Salmo trutta fario 16,002 Phoxinus phoxinus 5,419 Barbatula barbatula 2,897 Barbus petenyi 2,347 Leuciscus cephalus 1,850 Alburnoides bipunctatus 1,535 Gobio gobio 1,413

128 9

Cottus gobio 1,296 Sabanejewia balcanica 929 Rutilus rutilus 681 Table 12 : Number of lengths for the ten most abundant species taking the first run.

The number of lengths for each species for all runs is detailed in appendix E and in appendix F for the first run. The five most abundant of species (in number of lengths) per country are detailed in appendix G for all runs and in appendix H for the first run.

Lengths where measured during run 1 to run 5, but the majority (84%) were measured during the first run. As previously seen for all sites, there is some individuals for which the run was not filled in. They represent 2.5% of the total lengths for reference sites and are all located in Sweden (Table 13). Except for this country, the majority of individual were measured during the first run (Table 14).

No data 1 2 3 4 Austria 0 49 0 0 0 Spain 0 8,644 3,063 1,233 5 France 0 5,030 1,283 0 0 Italy 0 4,002 210 0 0 Lithuania 0 5,170 0 0 0 Poland 0 499 0 0 0 Romania 0 14,238 0 0 0 Sweden 1,132 228 94 14 0 Table 13: Distribution of lengths between countries and runs.

AT ES FR IT LT PL RO SE 100 66.77 79.68 95.01 100 100 100 15.53 Table 14: Ratio of length measured during the first run.

The majority of fishing occasions (72%) takes place between August and October. The number of lengths measured during this period follows the same trend (Table 15). However, this pattern is not followed by each country: France and Italy only have 46 and 20% of their lengths measured during this period (Table 16, 17 and 18). If we considered only the first run those proportions don’t change except for Sweden, where it moves from 82% to 40% (Table 18). Number of fishing Number of lengths Number of lengths Months occasions all runs run 1 January 9 451 431 February 5 150 146 March 10 395 361 April 8 370 272 May 2 228 228 June 22 3,122 2,404 July 34 3,440 2,682 August 124 14,282 12,850 September 159 16,328 13,213 October 49 3,201 2,624 November 37 2,897 2,638 December 1 30 11 Table 15: Distribution of fishing occasions and number of lengths across the year.

129 10

January February March April May June July August September October November December Austria 0 0 0 0 0 0 0 1 0 0 0 0 Spain 0 1 1 4 0 4 5 47 57 23 2 1 France 0 0 0 0 0 10 9 6 19 2 2 0 Italy 9 0 9 4 1 8 10 0 9 4 14 0 Lithuania 0 0 0 0 0 0 1 11 11 0 0 0 Poland 0 0 0 0 0 0 0 0 5 7 0 0 Romania 0 4 0 0 1 0 5 56 29 12 19 0 Sweden 0 0 0 0 0 0 4 3 29 1 0 0 Table 16: Number of fishing occasions per country and months.

Table 17: Number of lengths per country and months for all runs. January February March April May June July August September October November December Austria 0 0 0 0 0 0 0 49 0 0 0 0 Spain 0 10 31 207 0 1,000 652 3,967 5,376 1,622 50 30 France 0 0 0 0 0 1,423 1,503 874 1,775 260 478 0 Italy 451 0 364 163 205 699 811 0 689 186 644 0 Lithuania 0 0 0 0 0 0 80 2,566 2,524 0 0 0 Poland 0 0 0 0 0 0 0 0 157 342 0 0 Romania 0 140 0 0 23 0 136 6,787 4,651 776 1,725 0 Sweden 0 0 0 0 0 0 258 39 1,156 15 0 0

Table 18: Proportion of lengths for the period August-October. AT ES FR IT LT PL RO SE All run 100 84.71 46.08 20.78 98.45 100 85.79 82.43 1st run 100 86.14 49.02 20.79 98.45 100 85.79 39.91

1.2.2 Fish length for calibration sites Introduction On the 959 calibrations sites recorded, fish lengths are available for 721 of them. The 721 sites are distributed between 10 countries: Austria, Germany, Spain, France, Italy, Lithuania, Poland, Romania, Sweden and United Kingdom. Three countries (Finland, Hungary and Portugal) with calibration sites don’t have lengths associated with the fishing occasions. The 721 sites were sampled during the period 1978-2007 and correspond to 2,016 fishing occasions, 78 species and 277,470 lengths (Table 19). The number of fishing occasions varies a lot between countries with more than one thousand one for Sweden and no data for Finland, Hungary and Portugal. The number of species per country is ranging from 17 (Italy and Romania) to 36 (France) and the number of length is ranging from 2,164 to more than 100,000 (Table 19). Consequently we observe a lot of variability in the number of fishing occasions and in the number of lengths per country. Number of sites Number of fishing Number of Number of Number of sites with length occasion with lengths species lengths Austria 49 49 62 19 9,858 Germany 21 21 27 18 6,503 Spain 372 255 295 25 33,136 Finland 44 0 0 0 0 France 89 89 280 36 118,515 Hungary 6 0 0 0 0 Italy 58 26 37 17 3,274

130 11

Lithuania 20 20 23 23 2,164 Poland 124 108 109 30 6,832 Portugal 2 0 0 0 0 Romania 34 26 30 17 3,753 Sweden 112 111 1,124 23 91,330 United Kingdom 28 16 29 15 2,105 Table 19: Number of calibrations sites per country.

Lengths Focusing our interest only on calibration sites we have data for 721 sites, 78 species, 277,470 individuals all runs confounded and 160,231 individuals for the first run. Without differentiating runs, the number of length per species is ranging from 1 to 99,778 and the tens most abundant species are: trout, minnow, Atlantic salmon, bullhead, sea trout, lake trout, stone loach, gudgeon, eel and chub (Table 20). Considering only the first run the number of lengths per species is ranging from 1 to 61,902 and the ten most abundant species are: brown trout, minnow, bullhead, Atlantic salmon, stone loach, gudgeon, eel, chub, roach and an Iberian endemic fish Pseudochondrostoma duriense (Table 21).

Species Lengths Salmo trutta fario 99,778 Phoxinus phoxinus 37,773 Salmo salar 23,868 Cottus gobio 23,638 Salmo trutta trutta 20,643 Salmo trutta lacustris 14,149 Barbatula barbatula 10,249 Gobio gobio 6,978 Anguilla anguilla 6,110 Leuciscus cephalus 4,265 Table 20: The ten most abundant species considering all runs.

Species Lengths Salmo trutta fario 61,902 Phoxinus phoxinus 29,137 Cottus gobio 11,923 Salmo salar 11,542 Barbatula barbatula 7,933 Gobio gobio 5,245 Anguilla anguilla 4,135 Leuciscus cephalus 3,888 Rutilus rutilus 3,357 Pseudochondrostoma duriense 2,188 Table 21: The ten most abundant species considering only the first run.

The number of lengths for each species for all runs is detailed in appendix I and in appendix J for the first run. The five most abundant species (in number of lengths) per country are detailed in appendix K for all runs and in appendix L for the first run.

Lengths were measured between the run 1 and 5 but mostly during the first run. Like previously observed for reference and for all sites, a lot of information concerning the run is

131 12

missing in Sweden (Table 22). Except for this country, the majority of lengths were collected during the first run (Table 23).

No data 1 2 3 4 5 Austria 0 8,249 1,245 364 0 0 Germany 0 6,503 0 0 0 0 Spain 0 21,041 8,410 3,558 118 9 France 0 94,241 24,274 0 0 0 Italy 0 2,694 388 192 0 0 Lithuania 0 2,164 0 0 0 0 Poland 0 6,832 0 0 0 0 Romania 0 3,753 0 0 0 0 Sweden 74,900 13,349 2,082 999 0 0 United Kingdom 0 1,405 418 282 0 0 Table 22: distribution of lengths per run and country.

AT DE ES FR IT LT PL RO SE UK 83.68 100 63.5 79.52 82.28 100 100 100 14.62 66.75 Table 23 : ratio of length measured during the first run.

As previously observed for both all sites and reference sites, fishing occasions and lengths are mostly distributed between August and October. Those three months concentrate more than three-quarters of fishing occasions and lengths of calibration sites (Table 24). Nevertheless, this general pattern is not true for all countries (Table 25 and 26). Germany and especially Italy have lees than fifty percent of their available lengths measured during this period (Table 27). Number of fishing Number of lengths Number of Months occasions all runs lengths run 1 January 2 115 76 February 5 587 296 March 5 426 355 April 16 3,108 2,304 May 25 3,468 2,635 June 65 16,449 12,766 July 175 26,572 13,766 August 744 93,514 40,945 September 695 106,960 67,076 October 235 20,256 15,119 November 35 4,922 3,947 December 14 1,093 946 Table 24: Distribution of fishing occasions and lengths per months.

January February March April May June July August September October November December Austria 0 0 0 0 4 6 8 9 14 17 3 1 Germany 0 0 1 4 0 0 0 2 5 11 2 2 Spain 0 2 2 7 11 1 29 72 92 56 12 11 France 0 1 1 0 6 36 27 64 132 11 2 0 Italy 1 1 1 4 1 8 6 3 5 0 7 0 Lithuania 0 0 0 0 0 1 1 8 12 1 0 0 Poland 0 0 0 1 1 8 2 10 39 43 5 0

132 13

Romania 0 1 0 0 1 0 0 15 1 9 3 0 Sweden 0 0 0 0 1 4 98 550 387 83 1 0 United Kingdom 1 0 0 0 0 1 4 11 8 4 0 0 Table 25: Distribution of fishing occasion per months and per country.

January February March April May June July August September October November December Austria 0 0 0 0 716 599 1,118 766 2,641 3,672 230 116 Germany 0 0 240 1,310 0 0 0 1,308 1,153 898 945 649 Spain 0 448 84 1,494 695 262 6,389 6,537 11,479 4,419 1,001 328 France 0 108 57 0 1,674 12,537 9,783 31,114 58,884 2,975 1,383 0 Italy 78 20 45 274 78 795 690 421 381 0 492 0 Lithuania 0 0 0 0 0 82 47 580 1,120 335 0 0 Poland 0 0 0 30 63 1,893 20 699 1,844 2,202 81 0 Romania 0 11 0 0 107 0 0 1,379 277 1,213 766 0 Sweden 0 0 0 0 135 278 8,058 50,059 28,368 4,408 24 0 United 37 0 0 0 0 3 467 651 813 134 0 0 Kingdom Table 26: Distribution of lengths per months and per country considering all runs.

AT DE ES FR IT LT PL RO SE UK All runs 71.81 51.65 67.71 78.44 24.5 94.04 69.45 76.44 90.7 75.92 First run 72.82 51.65 71.12 79.45 20.01 94.04 69.45 76.44 95.87 67.4 Table 27: Proportion of lengths for the period August-October.

1.2.3 Conclusion

The table tbllength contains a large amount of fish lengths distributed in thirteen countries, that is to say all countries involved in this project except Finland. The repartition of lengths per country is very unbalanced, which one country providing half of the data (France). The period concerned spreads over fifty years but is mostly concentrated during the last decade. Also fish lengths are very heterogeneous, with some fishes precisely measured (total length) and with other fishes for which we have only a coarse estimation of their lengths (large interval between classes). This data set contains lengths for the majority of species caught and that for both reference sites, calibrations sites and all sites. Six species are always among the ten most abundant species: stone loach (Barbatula barbatula), bullhead (Cottus gobio), gudgeon (Gobio gobio), chub (Leuciscus cephalus), minnow (Phoxinus phoxinus) and brow trout (Salmo trutta fario). However, the most abundant species differ between countries and this for all categories of sites. The sampling dates are spread all over the year, but they are not distributed similarly between countries. This pattern is observed for all kind of sites. The dare will be a very important factor to take into account in further analysis, especially if we focus on young fishes. The majority of individual lengths were measured during the first run and that for all countries except Sweden. For Sweden the low proportion of lengths coming from the first run, is not a problem of efficiency of sampling during the first run, but it’s mainly a lack of information. Most of the time the information related to the run is missing. It concerns 3,829

133 14 fishing occasions on the 5,345 Swedish fishing occasions. 482 Spanish fishing occasions are concerned by the same problem. In order to keep those fishing occasions and especially the Swedish ones, we propose to determine randomly if the length of each fish was determined during the first run or not. This will be possible because in the table tblcatch we have the details of how many fish of a certain species was sampled during the first run. Consequently if one hundred brown trout would have been sampled during the first run, we will assign to one hundred brown trout lengths the run number one. With this methodology we can keep 3,760 Swedish fishing occasions but no Spanish fishing occasions.

2. Environmental characteristics of low species rivers

2.1Typology of sites according to richness

By defining arbitrary levels - say only 1; 2 to 3; and 4 or more species- reference and calibration sites can be sorted according to their absolute richness as measured during fishing events. This results in the following distribution:

Richness 1 2-3 4+

Figure 1. Distribution of sites by richness level.

134 15

Species richness 1 2-3 4+ Reference and calibration sites 438 542 437 Reference and calibration sites with at least 50 fishes 179 294 351

Mapping sites by richness levels gives first information on locations of low and high richness level sites (See figure 1). Notice for instance the great number of low richness sites in Northern Spain and the Asturies, as well as in the Alps; and the relative lack of such sites in Central and Eastern Europe.

2.2 Environment

To go further in the exploratory analysis of site environment according to richness, factorial analyses were performed on a set of available variables. The following factors were considered: • Geomorphological factors: Geomorph.river.type, Valley.form, Floodplain.site and Actual.valley.slope; • Geological factors: Geological.typology, Natural.sediment, ERODI.du; • Geographical factors: Altitude, ALT.GRADIENT, ELEV.MN.du; • Hydrological factors: Size.of.catchment, STRAHLER, Lakes.upstream, Water.source.type; • Meteorological factors: precmean.ann, temp.ann, temp.jan, temp.jul, PREC.AN.du, TEMP.AN.du; • Percentages for soil occupation in upstream catchment: clastseddu, calcsed.du, igneous.du, morphic.du, fluvdep.du, glacdep.du, eolian.du, organic.du; • Percentages for deposit composition in upstream catchment: urban.du, agri.du, past.du, fores.du, scrub.du, noveg.du, wetl.du, mwetl.du, est.du. A Multiple Correspondence Analysis of this set of factors - grouped into interval classes when necessary, that is for continuous factors –helps find out relevant pattern in our calibration dataset. Sites and factor levels can be then displayed into main resulting factorial plans. Note that too asymmetrically distributed variables were previously transformed using usual transormations; Size of catchment was log-transformed whereas actual river slope and percentages were squared-root-arcsine transformed.

Beyond expected gradients, notably that of temperature and that of altitude, which correspond more or less to respectively the first and second axes, other factors prove to be particularly discriminating (See figure 2). Among them: sand sediment, glacial water source, nival water source and braided river type, which are all closely related to geographically consistent site groups.

135 16

1.0 Water.source.type.Glacial Natural.sediment.Sand 1.5 Valley.form.Gorges Water.source.type.Glacial TEMP.AN.du.2precmean.ann.2temp.jan.1Actual.river.slope.3 Geological.typology.Calcareoustemp.jan.2PREC.AN.du.2 temp.ann.2TEMP.AN.du.1ALT.GRADIENT.3scrub.du.0 TEMP.AN.du.0 ELEV.MN.du.3 0.5 Geomorph.river.type.Meandigneous.du.0 glacdep.du.3 1.0 Water.source.type.Nivalagri.du.0 Altitude.3 calcsed.du.3 ERODI.du.1Size.of.catchment.0ELEV.MN.du.0 temp.jul.0 Natural.sediment.Boulder.Rock temp.ann.1 Altitude.0 Floodplain.site.Yestemp.ann.0 Valley.form.PlainsAltitude.3past.du.2temp.jul.1noveg.du.3scrub.du.3Floodplain.site.Noclastseddu.0 temp.jan.0 noveg.du.3calcsed.du.3 Actual.river.slope.3Valley.form.Gorges fores.du.0urban.du.0PREC.AN.du.0STRAHLER.1 Water.source.type.Nival clastseddu.3ERODI.du.2fores.du.3 ALT.GRADIENT.3 Geomorph.river.type.Naturally.constraint.no.mobActual.river.slope.0Valley.form.V.shape agri.du.0 igneous.du.1PREC.AN.du.1 fluvdep.du.3ELEV.MN.du.3ELEV.MN.du.2precmean.ann.1Size.of.catchment.1fores.du.1agri.du.3 0.5 Geomorph.river.type.Naturally.constraint.no.mobagri.du.1 ALT.GRADIENT.0STRAHLER.2ERODI.du.2past.du.0 STRAHLER.5.6 0.0 Water.source.type.PluvialSTRAHLER.3fores.du.2wetemp.jul.2tl.du.0Natural.sediment.Boulder.Rockfluvdep.du.0 wetl.du.3temp.jul.0 Geological.typology.Calcareousscrub.du.2 STRAHLER.4agri.du.2Altitude.2noveg.du.0 Actual.river.slope.2ELEV.MN.du.2precmean.ann.2temp.jan.2 temp.jul.3agri.du.1Natural.sediment.Gravel.Pebble.CobbleERODI.du.3 calcsed.du.0 precmean.ann.1scrub.du.1 ALT.GRADIENT.2TEMP.AN.du.2 past.du.3Geomorph.river.type.Sinuousglacdep.du.0Geological.typology.Siliceousurban.du.2scrub.du.2 Geomorph.river.type.BraidedValley.form.U.shapepast.du.2 STRAHLER.3 urban.du.0Size.of.catchment.0PREC.AN.du.2 fores.du.3 temp.ann.0 wetl.du.3 STRAHLER.4 fluvdep.du.0glacdep.du.0Altitude.2 temp.ann.2ERODI.du.1 Size.of.catchment.3igneous.du.2Actual.river.slope.2ALT.GRADIENT.2 precmean.ann.0 igneous.du.0 scrub.du.3 urban.du.3 Size.of.catchment.2precmean.ann.0 Comp2 0.0 Size.of.catchment.3 urban.du.2 wValley.form.V.shapeetltemp.jul.3.du.0 past.du.0 Comp4 ALT.GRADIENT.1ERODI.du.0PREC.AN.du.3 temp.jan.1past.du.3 Geological.typology.SiliceousNatural.sediment.Gravel.Pebble.CobbleSize.of.catchment.2Actual.river.slope.1 igneous.du.2 PREC.AN.du.1 precmean.ann.3 fores.du.2calcsed.du.0ALT.GRADIENT.1Floodplain.site.Noclastseddu.0STRAHLER.2Size.of.catchment.1 Altitude.1 TEMP.AN.du.0 temp.ann.1 Water.source.type.PluvialAltitude.1noveg.du.0fores.du.1STRAHLER.1 -0.5 Actual.river.slope.1 temp.jul.1Geomorph.river.type.Sinuousfores.du.0 Valley.form.U.shape temp.ann.3 temp.jan.0 PREC.AN.du.0ERODI.du.3glacdep.du.3 scrub.du.1 ELEV.MN.du.1 fluvdep.du.3 urban.du.3 agri.du.2 PREC.AN.du.3 STRAHLER.5.6TEMP.AN.du.3temp.jan.3 TEMP.AN.du.1 ELEV.MN.du.1 -0.5 temp.jul.2 ERODI.du.0precmean.ann.3 ALT.GRADIENT.0scrub.du.0 temp.ann.3 Valley.form.Plains Actual.river.slope.0 temp.jan.3 -1.0 clastseddu.3 ELEV.MN.du.0 TEMP.AN.du.3 igneous.du.1 Geomorph.river.type.Meand Altitude.0 -1.0 agri.du.3 Floodplain.site.Yes Geomorph.river.type.Braided Natural.sediment.Sand -1.5

-1.5 -1.0 -0.5 0.0 0.5 1.0 -1.5 -1.0 -0.5 0.0 0.5 1.0

Comp1 Comp3

d = 0.5 d = 0.5

DE IT PL AT RO SE IT FR PT UK FI FR LT AT PT SE ES FI HU UK HU ES

LT RO

PL

DE

Figure 2. Factor levels and sites grouped by country in the first two factorial plans of a Multiple Correspondence Analysis of environment data.

Using hierarchical clustering, the distribution of sites by richness levels and by the resulting typology site groups can be analysed further using a Factorial Correspondence Analysis of the contingency table or by mapping sites (See figures 3 and 4). It is obvious that the distribution of richness levels differs according to environment types. Notably, low levels are relatively more abundant within groups 1, 6 and 7, which are mainly located in the Asturies, whereas high levels are relatively abundant in groups 4 and 5, which are mainly located in Central and Eastern Europe.

Eigenvalues d = 0.5 Cluster Dendrogram 0

8 04

3

03 2-3

1 2

4 4+

1 02 7 5 0 1 32 6596339 3406336 819 6896118 344 322347 655 764332 865334 876188 189 396 936297 706 261402 101648 191 419 208 134210 135 345346 348 252254 398 470471 164 162166 684167 682683 742744 299 67116243 3337 3687 8687 65509782 25697037 6117 6246 64746378 6383 6387298 61957244 23692451 6247 40403830 33116507 3756 64726111 6236 44664022 40734104 46954827 4624 4747 51975198 80847466 67107963 79708392 97539736 9709 9738 7236 65887182 65036338 7189 5746 6504 2575 6311 9750 70816864 6301 6310 6359 6134 63236499 70726617 24726362 7321 70347341 4685 73967410 4622 7111 71247145 63126377 6353 6376 26222625 6509 57226034 2646 4554 65426536 6309 47316496 7203 4159 42934290 9789 7253 6580 68687245 38453762 6473 39953996 33203890 40214086 4828 7478 8507 72397237 6337707 72427197 7201 57505751 47494764 61336135 73567380 7377 47034826 45744620 7144 27393255 45814730 6537 65416535 65386544 572811299 999010225 36673808 38783771 3663 3807 3752 37494256 35854379 36614427 33753497 34233500 34583469 34554358 34823533 3568 35413565 35593577 3384 33904754 41293383 3967 4139 42183351 39003904 34984366 44734477 42994081 4023 41154312 40274680 47674783 59896245 40544061 40644069 41974210 40803342 40754326 40984314 41004283 39794134 40934094 38873636 36983926 36064217 39213608 39024121 35983619 3531 37473658 41543720 41554175 44373818 3794 38013860 43833790 3792 44564458 38494172 42944168 3859 41663591 34843613 3652 35363623 36893509 3348 33544675 40964097 46744052 41494089 38194647 38624224 46444223 4044 41834050 41034195 40354151 34364144 40253395 4157 3532 41463760 42604147 4452 3425 40203350 43496530 40194087 40634768 4684 47074673 4751 61216131 4137 41433750 34993869 34384145 34243413 4156 41484158 41694167 42304196 4277 3784 36693783 42583846 42484329 36624251 4246 3454 41813597 44184431 34803593 34723596 36153656 36593569 43284140 35344422 35013473 3470 44303385 3554 38683837 3851 44464180 38164174 38333832 3843 44534160 42734244 34833485 39233924 3683 39053961 3622 37393507 3730 36283972 42193732 36923522 36453948 37093723 39473695 3651 40673918 39653895 3909 61156259 58616258 5611 6403 4053 42874036 33434099 40624769 47744469 33473349 4304 42124305 33574114 33564296 40744213 40914211 47394829 48134793 48074792 47946469 46686470 48014710 46764712 33774755 33553434 34214113 33826494 65004759 47434789 4790 4791 51845189 5437 74627536 75147546 75587564 75497559 74737474 7467 74687496 74657557 75677510 75097513 23182326 99119870 8549 81298548 998110125 78907896 8508 82348647 79758195 839810009 85677905 83948547 79607964 86398648 79618389 83888657 79017902 84558370 8478 84428480 83598131 85687893 85507891 8168 8362 8384 2300 79777978 8175 83748510 83638380 81348135 84058235 84498509 81698448 84797903 83938197 84628196 8485 7889 83838361 79527955 81808382 79507967 84578458 78768570 7875 7895 97019698 9732 97299722 9746 97129716 9684 96559714 96459651 96439621 9680 98179708 96889731 96289820 96929828 97269727 9711 9737 9819 97079810 97909812 98239693 97029694 97139815 97449816 97799818 96749689 9788 97839786 9787 96359636 6758 6759 9774 97759776 96279748 97929749 97949795 96249780 9659 96549662 96429650 96869687 96539660 9613 96149615 96499670 96589647 9806 96959798 98059791 97579821 96739678 96719677 97409773 97639765 97679830 98249766 97699799 96919741 96239793 96859822 96269656 97399747 97179733 72277103 71057101 7232 7223 71987226 6827 72357200 7225 7074 7092 6882 68886889 65797215 71027231 7183 5690 56845683 5685 2586 25892587 29647055 63887251 26266464 6828 70366830 68667254 68697247 47654692 6543 72987451 6940 74367437 48066126 47374803 47224734 63586455 63516386 7438 73097454 74336911 74236912 7424 47954756 4804 47574705 4758 45764785 6112 61036113 72057206 26172618 26152616 6589 65916592 65906598 5682 58855789 59055908 59095883 58766024 58806011 2744 32653264 2632 3274 71927193 71807190 72347104 7184 96399640 46984825 4291 3763 3997 7238 51925193 5194 97429743 24602461 6865 6364 64906491 36683872 4403 37653766 35233537 35024426 34574354 4423 35583588 3578 4083 41054106 41884198 3637 39083911 41774184 3828 38234441 4170 35183519 4408 40264671 4746 42764744 34264360 4348 37874400 35383539 3474 3838 41734176 4451 3976 4651 3935 35084407 4416 36843936 37644389 3696 36423643 6262 40424068 4677 33863387 75637560 75617562 9871 7897 86588659 8371 7898 81767972 8559 23492350 79748391 78847877 7880 78737878 9700 9745 9632 96999720 9728 98139814 9768 97849785 9781 9661 9652 96969697 96729676 2588 7452 47984796 4797 4786 27412742 2743 3266 42524253 3540 3664 35263527 3556 3627 4138 4182 44804481 47724773 3630 36313632 37053706 3707 4435 38154436 38134445 38144434 35174413 35064405 4406 35204414 3681 39623617 36203618 3621 3530 4421 3529 4425 4355 4356 3840 38414457 37893791 3599 3600 3941 39383939 39593963 5211 5212 9881 9882 5190 51995191 51965200 53595241 52435242 5244 7532 75337534 75657566 75687569 23212314 23192315 23232325 23082304 23072311 23052306 8452 84538454 2301 2302 7871 7872 78747879 78927899 7900 8372 83668373 83678368 8369 96319633 96299630 9760 97619751 97589752 97779759 9778 9796 9797 97349735 97189719 97239724 97039704 67036704 73687369 73717370 7372 47874788 6548 6552 3463 34623464 32493250 32533251 3252 4614 4615 9641 11302 1176912407 11474 117536313 1121811188 10465 10452 10574 1224012442 11731 1175412417 11595 12358 1043110456 1044310434 1176810544 12447 10295 1024110387 1024410240 10243 10376 1027510316 10200 1017010224 1025110101 1010310004 10247 1039710344 10319 1027710357 10354 1028610421 1021410287 1036510273 1027810249 1006010211 100709995 10085 10427 10458 1037010430 1033310368 1038110334 1044710445 100059883 10075 10350 10360 1092311765 1228912292 10453 10457 1043210433 10545 1029310294 1032110324 10141 10346 1020910210 1031410329 10279 1021510216 10336 10446 10269 1030710308 1030910311 1036210363 1030110302 1025310254 1031810300 10303 10315 1027410326 1038410276 10385 10015 1001610017 10454 1046010455 10461 10419 1041610417 10449 1045010451 10437 1043510438 1025610436 10257 1072812268

6 Figure 3. Typology of sites in terms of environment and factorial analysis of the contingency table of resulting groups vs. richness levels

136 17

Environment Richness 1 1 2 2-3 3 4+ 4 5 6 7 8

Figure 4. Geographical distribution of sites according to richness and environment

Finally, looking at factors separately to investigate differences in environmental features according to richness level, marked trends can be stressed.

First, notice the increase in the proportions of meanders and braided river types, of plains and U-shaped forms, and of floodplain sites with increasing richness; and conversely, look at decreasing river slope (See figure 5).

137 18

Geomorph river type Actual river slope

1.0 0.6 Braided 0.8 0.5 Meand 0.4 0.6 Constraint 0.3 Sinuous 0.4 0.2

0.2 0.1

0.0 0.0 12-34+ 12-34+

Floodplain site Valley form

1.0 1.0

No Gorges 0.8 0.8 Yes Plains

0.6 0.6 U-shape

V-shape 0.4 0.4

0.2 0.2

0.0 0.0 12-34+ 12-34+

Figure 5. Barplot and boxplot for geomorphological factors within each richness level

Geological typology Natural sediment

1.0 1.0

Calcareous Boulder/Rock 0.8 0.8 Siliceous Gravel/Pebble/Co

0.6 0.6 Sand

0.4 0.4

0.2 0.2

0.0 0.0 12-34+ 12-34+

Altitude ERODI du 2000 5

1500 4

1000 3

500 2

0 1 12-34+ 12-34+

Figure 6. Barplot and boxplot for geological factors and altitude within each richness level

138 19

As regards geological factors, there is an increase in the proportion of calcareous types and a marked decrease in natural sediment size, with notably a sharp increase in sand proportions. Altitude is lower for higher richness level in average and, conversely, catchment size and Strahler order are lower (See figures 6 and 7). Note also apparent relationships between glacial water source and low richness, and between upstream lakes and higher richness level, which were quite expectable.

Size of catchment STRAHLER

10 1.0

1 8 0.8 2 6 0.6 3

4 4 0.4 5-6 2 0.2

0 0.0 12-34+ 12-34+

Water source type Lakes upstream

1.0 1.0

Glacial No 0.8 0.8 Nival Yes

0.6 Pluvial 0.6

0.4 0.4

0.2 0.2

0.0 0.0 12-34+ 12-34+

Figure 7. Barplot and boxplot for hydrological factors within each richness level There is a decreasing trend for richness with mean annual temperature and to a greater extent with annual precipitations (See figure 8). As regards, soil occupation factors, low richness levels appear to be related with high percentages of scrub or/and herbaceous vegetation as well as no vegetation areas (See figure 10).

139 20

precmean ann temp ann 1600

15 1400

1200 10

1000

5 800

600 0

400

12-34+ 12-34+

temp jan temp jul

10 24

22 5

20 0 18

-5 16

14 -10 12 -15 10 12-34+ 12-34+

Figure 8. Boxplot for meteorological factors within each richness level

2.3 Conclusion

In short, some more or less expectable relationships between environment patterns and richness levels can be found out. As expected, higher richness levels are distributed in sites with relatively higher catchment size, higher strahler order, lower altitude, lower slope. In proportions, higher richness levels also mean more calcareous geology, more sand, more floodplains and more meanders and, as expected too, more braided rivers and more sites with upstream lakes.

Maybe more interesting is the geographical distribution of low level sites. They are mainly located in two areas: first continental Europe mountains, and second in non- continental Europe, that is United Kingdom and Scandinavian Europe. Conversely, there are almost no such sites in eastern and central Europe plains.

First location in southern Europe Mountains can be quite well explained by previously mentioned factors. Other interesting associated factors, notably glacial and nival water source, as well as scrub or/and herbaceous vegetation might be helpful in explaining other locations.

140 21

3. Modelling river size

This document contains a procedure to estimate the wetted width in function of the drainage area, annual precipitation, temperature and elevation and water source type. We propose two equations to compute the theoretical width.

3.1 Motivation

The aim of this study is the modelling of theoretical width based on the drainage area and some other variables. In the river, the width is important parameter associated with several hydraulic mechanisms. For example, the width is strongly linked to the flow discharge (more details in Leopold et al. 1992, Knighton 1998, figure 9). The relationship between these two variables is defined as follow: a QW = a QW c

log()log ()+= log ()QcaW where a and c are empirical coefficients. In the same way, the discharge is linked to the drainage area (pages 6-8, Knighton 1998): b AQ = b AQ d

()loglog ()+= log ()AdbQ where b and d are empirical coefficients. As results, we can postulate that the relationship between width and drainage area is defined as follow: A aW ≈ aW ' A c'

log() ()+≈ log''log ()AcaW where a’ and c’ are empirical coefficients. In the literature, the authors (ex. Rational Method, http://www.lmnoeng.com/Hydrology/rational.htm, see the section References) suggest that discharge values (Q) is proportional to the product of the Drainage Area, intensity of rainfall and Runoff coefficient to account for surface characteristics (C). The relationship is defined as follow: = kCiAQ where k is a conversion factor (empirical coefficient). This model is relatively similar to the Mulvaney equation (Beven 2003):

p = CARQ

where Qp, A, R and C corresponded to the hydrograph peak, the catchment area, a maximum catchment average rainfall intensity and empirical coefficient. Under the above hypotheses, we propose to model the theoretical width values in function of the potential flow defined by average precipitation (mm/year) in the drainage area (km2) and the water source type (ex. Nival, Glacial, Pluvial, …).

141 22

Figure 9. Examples of relation between width, discharge and drainage area (Knighton 1998).

In our study, we complete the list of explicative variables by average temperature and average elevation of the drainage area. These variables appear to be important to integrate the climatic condition associated with drainage area and potentially associated with the Runoff.

3.2 Description of the variables

We use the calibration dataset to model the potential width, and reference dataset to evaluate the quality of the model. Precipitation: Mean annual precipitation upstream catchment (mean of primary catchments, mm/year) Catchment size: Area drained by segment (upstream area + primary catchment, km2) Water source type: glacial, groundwater, pluvial, nival. Wetted width (meter).

142 23

Average temperature of the drainage area (degree Celsius) Average elevation of the drainage area (meter)

variables N description log-width 941 1.386294/1.791759/2.302585 Log(P*A) 959 9.942976/10.802038/12.015321 mean of Temperature of drainage area 959 6.097255/ 8.414133/11.731278 Log-elevation 959 5.462796/6.192375/6.732177 Water.source.type : 959 Glacial 1% ( 7) Groundwater 5% ( 44) Nival 24% (228) Pluvial 71% (680) STRAHLER : 959 1 18% (174) 2 34% (323) 3 24% (231) 4 14% (131) 5 7% ( 71) 6 3% ( 29) Country.abbreviation : 959 AT 5% ( 49) CH 0% ( 0) DE 2% ( 21) ES 39% (372) FI 5% ( 44) FR 9% ( 89) HU 1% ( 6) IT 6% ( 58) LT 2% ( 20) NL 0% ( 0) PL 13% (124) PT 0% ( 2) RO 4% ( 34) SE 12% (112) UK 3% ( 28) Table 28. Statistical description for the calibration dataset (N=959, coding for continuous variables: 1st Qu. / Median / 3RD Qu.).

variables N description log-width 442 1.386294/1.791759/2.302585 Log(P*A) 459 9.756642/10.560738/11.644547 mean of Temperature of drainage area 459 5.323545/ 7.852267/11.107016 Log-elevation 459 6.306282/6.769081/7.011304 Water.source.type : 459 Glacial 1% ( 4) Groundwater 2% ( 10) Nival 11% ( 51) Pluvial 86% (394) STRAHLER : 459 1 18% ( 83)

143 24

2 33% (153) 3 31% (143) 4 12% ( 57) 5 5% ( 23) Country.abbreviation : 459 AT 0% ( 1) CH 0% ( 0) DE 0% ( 0) ES 42% (193) FI 3% ( 14) FR 6% ( 26) HU 0% ( 0) IT 12% ( 53) LT 3% ( 13) NL 0% ( 0) PL 3% ( 13) PT 0% ( 0) RO 29% (133) SE 3% ( 13) UK 0% ( 0) Table 29. Statistical description for the reference dataset (N=959, coding for continuous variables: 1st Qu. / Median / 3RD Qu.).

The modalities Groundwater and Glacial are relatively rare. This could be induced some numerical problem in model fitting. We propose to agglomerate these modalities with the modality Nival. We obtain a variable which discriminated Non-Pluvial or Pluvial status.

Precipitations in function of the water source type:

Figure 10. Precipitation in function of the water source type for calibration Dataset.

144 25

Figure 11. Precipitation in function of the water source type for reference dataset.

Discordance between the two measures of the catchment size There are some discordance between the two measures of the catchment (AREA.ctch from CCM2 and Size.of.catchment from the database). We conserve the variable catchment of size (completed by EFI+ partners) to limit the potential systematic error associated with the GIS estimation. We propose to skip the observations characterized by a difference between the two measures superior to 5%. We conserve only 191 reference sites and 318 calibration sites. Small, medium and large Rivers The distributions of the sites among the Strahler order show that the number of small rivers in the calibration and reference datasets are overrepresented (Figure 12). To regularise the sample, we propose to weight the observations by classification based on Strahler order. We obtain three groups: 0-2, 3-4 and 5-6 (see table 30).

Weight 1 2 3 4 5 6 1/170 0 0 100 70 0 0 1/98 21 77 0 0 0 0 1/46 0 0 0 0 33 13 Table 280. Relationship between weight and Strahler order for the calibration dataset. The weight correspond to 1/(site number in the category).

145 26

Figure 12. Distributions of sites in function of the Strahler order for reference and calibration datasets. The term “Limited” refers to the sites characterized by difference between the two measures of catchment size superior to 5% respectively.

Selection bias of GIS estimation of drainage is clearly shown graphically in the figure 12. After selection of sites characterized by difference between the two measures of catchment size superior to 5% respectively, we observe that the proportion of the small river decreases (figure 12). Distribution by country The distribution of calibration and reference sites is unbalanced (Figures 13 and 14). The Spanish and Romania sites are largely dominant in reference dataset. In calibration dataset, Spanish sites are dominant, but the sample is more representative of others countries.

146 27

Figure 13. Distribution of reference sites by country.

Figure 14. Distribution of calibration sites by country.

3.3 Modelling the width

We use linear model to estimate the coefficient of the regression of width (W) on potential flow (PA, product of the potential area and averaged annual precipitation), water source type (S=0 for Non-pluvial and S=1 for Pluvial), average temperature (T) and average elevation (E) of the drainage area. The model equation is given as follow:

log()+= ⋅log ( ) ⋅+ log( )+ ⋅ + ⋅ + ⋅log( )⋅ SPAfSeTdEbPAadW + ε

Where a, b, c, d, e and f correspond to the empirical coefficient and ε corresponds to model error.

147 28

Table 31. Summary statistics of the fitted linear model. The table gives coefficients, standard errors, etc. and additionally gives "significance stars" (Signif. codes: 0 '***' ; 0.001 '**' ; 0.01 '*'; 0.05 '.'; 0.1 ' ' ). coefficients Estimate Std t- Pr(>|t|) Error value (intercept) -1.78105 0.50099 -3.555 0.0004 log(P*A) 0.25277 0.02585 9.777 < 0.0001 S.Pluvial -3.40902 0.41630 -8.189 < 0.0001 Log(E) 0.11926 0.03806 3.133 0.0019 T 0.04131 0.01006 4.107 < 0.0001 Log(P*A):S.pluvial 0.28295 0.03226 8.770 < 0.0001

All parameters are significantly different to zero (Table 31) and the explained variance is equal to 0.6981% (F-statistic: 139.2 on 5 and 301 DF, p-value: < 2.2e-16 ).

Figure 15. Graphical representation associated with the regression to EFI values on global pressure index. The first graphic corresponds to residuals in function of the fitted values. The second shows the QQ-plot representation of standardised residuals against normal theoretical quantiles. The third correspond to the representation of the square root of standardized residuals against the fitted values. The fourth graphic plots the leverage against the standardised residuals to detect the potential influent points (the red dotted line correspond to leverage limit 3*p/n). The last graphic corresponds to the histogram of the Pearson residuals.

148 29

Figure 16. Observed and expected values from linear model and robust linear model.

In the figure 15, we observe that the proportion of outliers is relatively high and the graphic entitled “leverage vs standardized residuals” shows that there were some potential influent points (see Fox, 2002, chapter 6: Diagnosing problems in linear and generalized linear models, pages 191-234). For this reason, we complete the analysis by robust regression to evaluate deviation of the parameter estimations compared with the classical linear model. For more detail on the robust regression, you can consult these references: Venables and Ripley 1999, pages 167-174; MacKimmon & White 1985, Greene 2002, Zeileis 2004). In this study, the version of the robust model is based on the random resampling procedure (Marazzi 1993). In the software R (R Development Core Team 2007), the function lmRob (package robust) performs a robust linear regression with high breakdown point and high efficiency regression.

149 30

Figure 17. Observed log-width of reference dataset in function of the predicted values by country. The equation of a red dotted line is y=x.

To examine performances of both models, we use the following values (see Greene 2002, Potts & Elith 2006): RMSE (Root Mean Square Error) provides indication on the divergence between the observed and predicted values. Pearson correlation (CORP) and Spearman correlation (CORS) provide indication of similarities between the ranks of the observed and predicted values. AVER (Residuals distribution) provides indication on the goodness of fit.

Table 32. Summary statistics of the fitted robust linear model. The table gives coefficients, standard errors, etc. and additionally gives "significance stars" (Signif. codes: 0 '***' ; 0.001 '**' ; 0.01 '*'; 0.05 '.'; 0.1 ' ' ). coefficients Estimate Std t-value Pr(>|t|) Error (intercept) - < -3.04708 0.69660 4.37423 0.0001 log(P*A) < 0.34922 0.03588 9.73425 0.0001 S.Pluvial - -2.05556 0.58418 3.51873 0.0005 Log(E) 0.15967 0.05285 3.02091 0.0027 T 0.02167 0.01369 1.58300 0.1145 Log(P*A):S.pluvial 0.17199 0.04643 3.70392 0.0003

The parameters associated with the robust model are given in table 32 (R2=0.487936; test for Bias: M-estimate= -24.07663 with p-value=1 and LS-estimate =1899.16898 with p- value= 0). The robust regression shows that the effect of the temperature of the drainage area is not significant (p < 0.11). This result moderates the importance of the temperature in the

150 31 linear model. The significant test associated with temperature coefficient in the linear model is probably induced by some site characterized by particular environment. However, the expected values obtained by the both methods were relatively similar (figure 18, table 33). We observe deviations for some sites with theoretical width superior to 15 meters. The linear model appeared to be slightly more efficient than the robust model but the variability is more stabilized in the robust model.

Table 33. Performance analyses of the models. The terms RMSE, AVER, CORP and CORS correspond to Root Mean Square Error, averaged residuals, Pearson correlation and Spearman correlation. ESTIMATION RMSE AVER CORP CORS by Linear Model 0.2283 0.0130 0.7760 0.7714 Robust Linear 0.7258 0.0414 0.7718 0.7788 Model

Figure 18. Comparison between the predictive values based on classical linear model and robust linear model. The equation of the red dotted line is y=x.

Analyses of variance on the residuals show that the effect of Strahler order is no significant for the both models (P= 0.7142 for the linear model and P=0.1987 for the robust linear model, table 34).

Table 34. Summary of analyses of variance. We test the effect of Strahler (river size) on the residuals of the both models (LM=linear model; Robust LM=Robust linear model). The table contains coefficients, degree of freedom (df) and F-value and probability (Pr(F) Sq Sq value LM Strahler 5 0.703 0.141 0.5815

151 32

residuals 301 72.772 0.242 Robust LM Strahler 5 1.838 0.368 0.1987 residuals 301 75.163 0.250

We don’t test the potential effect of the Flow regime on residuals, because there are only 4 sites characterized by the modality “Summer dry” (Figure 19).

Figure 19. Regression residuals in function of the Strahler order and Flow regime. The equation of the red dotted line is y=0.

Figure 20. Residuals of linear model in function of the Strahler order and the variable Flood plain. The equation of the red dotted line is y=0.

152 33

Figure 21. Residuals of robust linear model in function of the Strahler order and the variable Flood plain. The equation of the red dotted line is y=0.

The both models trend to underestimate the width of the sites with Floodplain and to overstimate the width of the sites without floodplain (figures 20 and 21). In addition, we observe that the robust linear model is more affected by the Floodplain variable than the linear model. This result is more in favour of the linear model estimation.

Figure 22. Residuals of robust linear model in function of the Strahler order and water source type. The equation of the red dotted line is y=0.

153 34

Figure 23. Residuals of robust linear model in function of the Strahler order and countries. The equation of the red dotted line is y=0.

3.4 Validation on the reference dataset

To complete our study, we add a validation step based on the reference dataset. The values of correlations and AVER are relatively similar. The third picture of the figure 24 confirms that both models converge on similar expected values. The Root Mean Square Error (table 35) is lower for the robust linear model. The two histograms illustrate that the shape of error distribution of the linear model tends to be slightly asymmetric.

Table 35. Performance analyses of the models. The terms RMSE, AVER, CORP and CORS correspond to Root Mean Square Error, averaged residuals, Pearson correlation and Spearman correlation. ESTIMATION RMSE AVER CORP CORS by Linear Model 2.5890 -0.1924 0.7341 0.7476 Robust Linear 1.8625 -0.1384 0.7521 0.7569 Model

154 35

Figure 24. Predictive performance of models. Histograms correspond to the distribution of the difference between the observed and expected values of log-width. The last graphics illustrates the comparison between the predictive values based on classical linear model and robust linear model. The equation of the red dotted line is y=x.

Figure 25. Observed log-width of reference dataset in function of the predicted values based on linear model by country. The equation of the red dotted line is y=x.

155 36

3.5 Conclusion

To conclude, we propose two procedures to compute a theoretical width as follow: Classical estimation ()W ⋅+−= (PA)+ ⋅ (E)+ ⋅T − S +⋅ ⋅log2829.04090.30413.0log1193.0log2528.07811.1log ( )⋅ SPA

Robust estimation : ()W ⋅+−= (PA)+ ⋅ (E)+ ⋅T − S +⋅ ⋅log1720.00556.20217.0log1597.0log3492.00471.3log ( )⋅ SPA

Where W, PA, S, T and E correspond to theoretical width, potential flow (PA, product of the potential area and averaged annual precipitation), water source type (S=0 for Non- pluvial and T=S for Pluvial), Temperature of the drainage area and Elevation of the drainage area. In practice, we suggest that the linear model can be more appropriate to estimate the width. The performances of the both models are very similar, but the use of the linear model is easier.

156 37

4. Development of new metrics based on age classes

4.1 Methodology

Different ideas could be explore in order to take into account the age classes of fishes in the new European fish index. We will expose in this part, only one of the possibilities.

In this work we will focus only one species: the brown trout, Salmo trutta fario. By working on this species the main idea is to extend the detection capacities of the future index to headwater systems, and also to low species rivers. Headwater systems very often present particular environmental conditions due to their geographical position and to their proximity to the source. They very often exhibit a high slope, high velocities, cold water and a high oxygenation (suitable conditions for the occurrence of brown trout). Moreover the brown trout is a widespread species which occur all over Europe. For all those reasons we think that focusing on brown trout could be very interesting for the development of a metric based on age class. Nonetheless the brown trout is also known to be very plastic and to adapt itself to a high range of environmental conditions (i.e. this species also occur in Mediterranean streams). In the data base we don’t have the information about age of fishes. We will have to determine it from the lengths of fishes.

For the development of a new metrics, one possibility is to focus more specifically on one age class: the young of the year (YOY or 0+). This fraction of the population is normally easier to distinguish from the rest of the population than older age classes (i.e. Fig. 26).

0+ Olde

0

0

Frequencies 0

0

Lengths (mm) Figure 26: An example of distribution of brown trout lengths for one fishing occasion.

157 38

The main idea is to use this portion of the populations to create new metrics. When the metrics will be defined we will have to test if those metrics could be related to environment and if they can detect human pressure. It’s only at the end of those three steps that the metrics could be integrated in an index.

This part is strongly influenced by a previous report realised by Reyjol et al. (2005). To create a metric based on brown trout YOY, we must be able to identify those fishes within their population. The solution proposed for this first step is to use a mixture of normal laws to estimate the two parameters of the distribution (mean and standard deviation) of the 0+. By using this methodology we suppose that the distributions of brown trout lengths are composed of at least two parts: - a normal law for the YOY; - one or more normal law(s) for the older fishes. Consequently we have to estimate, the mean and the standard error of the distribution of the YOY, the mean(s) and the standard error(s) of the older fish distribution(s) and the proportions of each laws in the total distribution. Hence, for two normal laws we have to estimate 6 parameters. Here clearly appear the two first limitations. First it’s not sure that the distribution of lengths for the YOY follows (or always follows) a normal law and second estimating automatically those parameters could be sometimes delicate. For this part we will only used data from undisturbed sites. With the mean and the standard error of the normal law of the YOY we would be able to compute a “theoretical” maximal length than a YOY can reach (i.e. Fig. 26). YOY Older fishes

0.05 Theoretical maximal length = 91 mm

0.04

0.03

Densities 0.02

0.01

0

0 0 00 20 40 60 80 00 Lengths (mm) Figure 16: A theoretical distribution of lengths, with µ = 80, σ = 5 for YOY and µ = 150, σ = 15 for the older fishes. The proportions of each class are respectively 62.5 and 37.5 % of the total number of lengths. In this example the “theoretical” maximal length for a 0+ would be 91 mm.

158 39

Estimating the first normal law involves to have enough data of lengths for each fishing occasion and especially data on young fishes. As a consequence we will have to select the fishing occasions with enough data. The next step will be to determine the number of YOY in each fishing occasion. At this point we will only have maximal lengths for some undisturbed sites. We will need a tool to determine also the maximal length for all sites: reference and calibration sites not used in the first step and the disturbed sites. We propose to develop a model which links the environment of the sites with the maximal lengths. This model will be calibrated with the maximal lengths computed during the first step on the undisturbed sites (Fig. 27).

Fitted values

One maximal length Theoretical maximal lengths from normal maximal laws lengths Theoretical Confidence interval

Environment Figure 27: Hypothetic model relating maximal lengths and environment.

Several types of models are possible to compute, the multiple linear regression and the partial least square regression are two examples. The purpose here is not to have an explicative model to determine which environmental factor explain the variability of maximal length, but to have as far as possible the best predictive model. This step is very important because without a model relating maximal length with environment we would not be able to determinate the maximal length of the YOY for disturbed sites. The next step will be to estimate for each fishing occasion the number of YOY in the population. For that we will compute the maximal length of a YOY with the previous model and count the number of fishes smaller than this size (Fig. 26). With this information it will be possible to create three new metrics: - occurrence of brown trout 0+; - proportion of brown trout YOY in the population; - density of brown trout YOY.

159 40

It is clear that at each step of this methodology there are many limitations which can lead to difficulties in computing the three metrics. Once the three new metrics would be available, they will enter in a process of selection to know if they can be relevant for the new index. The methodology is derived from the “reference condition approach” (Bailey et al. 1998) and is divided into two parts: - modelling the metric with environmental variable; - test of the discriminant power of the metric. The first part consists in estimating a model to explain the variability of the metrics by environmental variable. It is totally possible that the metric could not be related with environment, or that environment only explains a short amount of variance of the metric. Without a model explaining a significant part of variance, the metric could not be integrated in the further analysis. The second part consists in determining if the metric is relevant to detect an anthropogenic disturbance or not. We will compare the deviation to the model (difference between the observed values and the fitted values, residuals), of calibrations sites and disturbed sites. If there is no significant difference between the two distributions the metric is considered as non-informative. We can test if the metric is relevant to detect human impact without distinguishing the pressure, or we can test if the metric reponse to a specific pressure. The metric can reflect an environmental gradient but not a human pressure.

4.2 First results

Estimation of the maximal lengths

In this part we decided to use both reference sites and calibration sites because we did not expect a difference of maximal lengths between YOY living in reference or calibration sites. It also increased the amount of data. As previously discussed we had to select fishing occasions for which we could estimate a mixture of normal laws. The brown trout occured in 1,121 of the 1418 fishing occasions (959 calibration fishing occasions and 459 reference fishing occasions). We decided to retain only the fishing occasions for which at least 70 brown trouts were caught. This threshold (which can be disccussed) was decided to maximize our chance to find enough data for YOY in the fishing occasions. After this selection it remained only 286 fishing occasions on the 1,211 ones. On the 286 fishing occasions we had fish lengths in the table tbllength for 247 of them. The next step was to select the fishing occasions for which we had the information on the run. In a process of standardisation we decided to keep only lengths of fishes caught during the first run. After this selection we only lost one fishing occasion. After having a look on length data for those 246 fishing occasions we kept 181 of them. During this selection we did not conserve the fishing occasions with too few lengths for small fishes (some fishing occasions only exhibit lengths for what seemed older fish) or when the data was not appropriated to fit normal mixture law (i.e. data in classes with high interval between each class). After those selections we finally had a data set composed of 181 fishing occasions divided between 141 calibration fishing occasions and 40 reference fishing occasions. Those fishing occasions are distributed between nine countries but France and Spain represents 85% of the data (Table 31). This data set is strongly unbalanced. Table 31: Distribution of fishing occasions per country. AT DE ES FR IT PL RO SE UK 12 5 104 51 3 1 1 3 1

160 41

After this selection we tried to estimate the mixture of normal laws with an Expectation-maximisation algorithm (EM). It is an iterative algorithm which criteria of maximisatio is to maximise the likelihood or it is to say to minimise the log-likelihood. Those estimations was computed under the software R (version 2.5.1, R Development Core Team 2007) with the fonction “normalmixEM” of the library “mixtools” (Young et al. 2007). 0.035

0.030

0.025

.020 Densiti 0.015

es 0.010

0.005

0.000

100 150 200 250 0 Lengths (mm)

Figure 28: A good estimation of the normal law for the YOY.

0.020

0.015

0.010

Densities Densities 0.005

0.000

100 150 200 250 0 Lengths (mm)

Figure 29: Problem in the estimation of the normal law for the YOY.

First we computed for each fishing occasion a mixture of two normal laws without entering initial values for the parameters. With omitting initial values, we wanted to let the algorithm free of any judgment on the distributions. For some fishing occasions the

161 42 distributions estimated for the YOY seemed to estimate well the distribution of YOY whereas for others the fit was not good (Figure 28 and 29).

For fishing occasions for which we encounter some problems we decided to use a mixture of three normal laws instead of two. This enhanced in general the goodness of fit of the normal law for the YOY. The maximal length was computed as a 95%quantile of the first normal law. We checked the distribution of fish lengths for each of the 181 fishing occasions and we decided to left all fishing occasions for which we had a doubt. After this selection we only kept 85 fishing occasions. We made this selection because sometimes the distribution of the YOY was not clearly separated from the rest of the population (Fig. 30), or because the date of sampling was too early in the year or because we were not enough self-confident in the results. One possible solution to keep more fishing occasion and to be more self-confident could be to ask people who have provided the data if they agree or not with the results found by the mixture of normal laws. We can also ask them to provide us the maximal length that they thought it occurs in the fishing occasion, without giving any information about our results. Afetr we can compare maximal lengths estimate by an expert judgment and with the EM algorithm. 35

30

25

20

Frequenci 15

10

5

0

50 100 150 200 250 300 350 Lengths Figure 30: Strong overlap between distributions.

Currenlty we are trying to link the maximal lengths we computed, with environment. We used multiple linear regressions. The best model that we have computed until now, explain 53% of the variance. It is composed of annual temperature, squared annual temperature, geological typology, natural logarithm of distance from source and the mean annual precipitations. If we look at the deviation from this theoretical model we can see that for 80% of the fishing occasions we have an absolute residuals lower than 10 millimetres. However, the absolute deviation can reach 30 millimetres (Fig. 31). To improve our model, we have to identify the reason why for some fishing occasions we have an “acceptable” deviation and why for others the deviation is so important.

162 43

0.04

0.03 D 0.02

0.01

0.00

-20 -10 0 10 20 30 Re Figure 31: Distribution of the residuals. We can try to use different models than multiple linear regressions, like for exemple partial least square regression.

All age classes of the population

The idea here is not to focus only on YOY but to try to take into account the whole population. We will expose there the results from a French experiment which was done for by Reyjol et al. (2005).

The authors try to estimate the proportion of each age class in the population. The methodology they have used for the whole population is very close than the one used for the YOYs, except for the determination of the proportions. Due to a strong overlap between lengths distribution of older age classes, they could not used a mixture of normal laws to estimate the boundaries between classes. To face this problem they used Von Bertalanffy growth functions (VBGF) to estimate the maximum lengths that a fish of a certain age can reach (Fig. 32). Leng

th

Ti me Figure 32: an example of Von Bertalanffy growth curve. The VBGF related the length at one age with the growth rate, the maximum possible size and the theoretical age when length = 0:

163 44

LL=−1 e()−−Kt() t0 t ∞ () With: Lt: Fish length at age t; L∞: maximum possible size; K: rate of growth; t0: theoretical age when length = 0. They computed the theoretical maximal length for all age classes, which was used to delimiter the age classes. They related those limit of classes to the annual mean temperature (Fig. 33).

Total lengths (mm) Total lengths

Temperature (°C)

Figure 33: Relationship between mean air temperature and length of different age classes. In red the 1+ (fish in their second year), in blue the 2+ (fish in their third year) and in green the fish older than three years (Reyjol et al. Date).

With those relationships they were able to compute the proportion of each age classes for each fishing occasions. They defined their metrics for each age class. They modelled the proportion of each age class with environmental variables and they compared the distribution of the residuals between calibration sites and disturbed sites.

4.3 Conclusion

On the large amount of data present in the table tbllength only a very short part seems available for the construction of the new metrics based on brow trout young of the year. Moreover, this small data set is very unbalanced between countries with one country providing half of the data. This can be a strong limitation for the analysis.

164 45

Concerning the determination of the maximal lengths of YOY by normal mixture laws, currently we lose also half of the remaining data due to doubts on the distribution of the YOY. Sometimes it appears very difficult to estimate which part of the distribution of fish lengths is only composed of YOY. The process we used during this part of the methodology did not appear to be so automatic, we had to check the results for each fishing occasion, and sometimes we have to modify the initial parameters (i.e. three normal laws instead of two). For some fishing occasions, as seen on the figure 5, we were sure that the estimate parameters where not good, because the normal laws fitted, were very far from the supposed distributions. But for others we were not sure that the fitted distribution really mismatched with the real distribution. We decided to add a third normal law, because we observed an other maximum in the distribution but we were not always sure that this maximum was related to an older age class. One solution to keep more data and to validate our primary results would be to compare our results with maximal lengths estimated by expert judgment.

This first part is crucial because without a good estimation of the threshold which distinguishes the YOY from the older fishes we could not compute metrics precisely.

Concerning the model linking the maximal lengths and the environment, we have to improve it but we also have some satisfactions. For 80% of the 85 fishing occasions the error we made with a multiple linear regression was less than ten millimetres. Although it remains 20% of the data for which we have a too important errors.

Currently we only try one type of model and we just estimated a model on a priori important factors for fish populations. We will have to try other combinations of environmental variables, maybe with procedure of selection like stepwise, and we will have to try others kind of statistical model.

One of the possible biases explaining the biais can be the date of sampling. After visualisation of the distribution of lengths, it seems obvious that the date of samplinbg influence the maximal length. Currently we don’t have any solution to correct this bias.

There are also different alternatives to the proposed methodology. We can for example use class of size instead of class of age, and cut the distribution of lengths with fixed threshold. We can also try to do the same work as done by Reyjol et al. with the age classes of the whole population.

References

Beven, J. K. (2003) Rainfall – Runoff Modelling: The Primer, John Wiley & Sons, New-york, 372 pages Knighton D. (1998) Fluvial forms & processes, Arnold Publishers, New-York, 218 pages. Leopold, L.B., Wolman, M.G. & Miller, J.P. (1992) Fluvial processes in geomorphology Freeman, W.H. and Co, San Fransisco, California. Fox, J. (2002) Applied Regression, Linear Models, and Related Methods. Sage. Greene (2002) Econometric Analysis, Prentice Hall,US, 5e International Ed, 1026 pp. MacKinnon JG, White H (1985). Some Heteroskedasticity-Consistent Covariance Matrix Estimators with Improved Finite Sample Properties. Journal of Econometrics, 29, 305–325.

165 46

Marazzi, A. (1993) Algorithms, routines, and S functions for robust statistics. Wadsworth & Brooks/Cole, Pacific Grove, CA. Potts, J. & Elith, J. (2006) Comparing species' abundance models. Ecological Modelling 199, 153-163. Reyjol Y., J. De Bortoli & D. Pont 2005. Indice Poisson Rivière. Modélisation et sensibilité aux perturbations de métriques basées sur la structure en âge. Etude réalisée pour le compte du Conseil Supérieur de la Pêche. 73 pp. R Development Core Team (2007). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org. Venables W. N. & Ripley B. D. (1999) Modern Apllied Statistics with S-plus, Third edition, Statistics and computing, Springer-Verlag, New York, pp. 501. Young D., Elmore R., Hettmansperger T., Hunter D., Thomas H. and Xuan F. 2007. mixtools: Tools for mixture models. R package version 0.2.0. Zeileis A (2004). Econometric Computing with HC and HAC Covariance Matrix Estimators. Journal of Statistical Software, 11(10), 1–17. URL http://www.jstatsoft.org/v11/i10/.

166 47

Appendix A: Fish lengths for all species and all runs

Species names Lengths Species names Lengths Phoxinus phoxinus 926 908 Achondrostoma arcasii 5 025 Salmo trutta fario 876 033 Gobio lozanoi 4 784 Rutilus rutilus 713 614 Sander lucioperca 4 624 Gobio gobio 575 677 Cyprinus carpio 4 563 Barbatula barbatula 441 269 Pseudorasbora parva 4 523 Cottus gobio 392 365 Chondrostoma miegii 4 470 Leuciscus cephalus 381 197 Salmo marmoratus 4 055 Alburnus alburnus 262 827 Rutilus aula 3 988 Perca fluviatilis 165 642 Achondrostoma oligolepis 3 587 Salmo salar 143 633 Barbus petenyi 3 389 Anguilla anguilla 123 443 Platichthys flesus 3 283 Salmo trutta trutta 113 635 Carassius carassius 2 971 Leuciscus leuciscus 107 005 Romanogobio belingi 2 795 Barbus barbus 102 953 Barbus graellsii 2 500 Alburnoides bipunctatus 80 763 Barbus plebejus 2 365 Rhodeus amarus 60 833 Sabanejewia balcanica 1 990 Leuciscus souffia 60 395 Gambusia affinis 1 925 Thymallus thymallus 52 029 Padogobius martensii 1 504 Salmo trutta lacustris 49 267 Cobitis paludica 1 486 Oncorhynchus mykiss 45 834 Salvelinus fontinalis 1 472 Abramis brama 45 054 Carassius gibelio 1 392 Blicca bjoerkna 44 833 Silurus glanis 1 324 Lampetra planeri 42 411 Iberochondrostoma lemmingii 1 302 Leuciscus idus 36 381 Micropterus salmoides 1 296 Gasterosteus aculeatus 34 907 Barbus peloponnesius 1 286 Lepomis gibbosus 33 729 Salaria fluviatilis 1 141 Esox lucius 29 403 Barbus comizo 1 042 Chondrostoma nasus 24 027 Petromyzon marinus 1 001 Pseudochondrostoma duriense 22 925 Carassius auratus 878 Barbus bocagei 19 755 Leuciscus muticellus 868 Lota lota 19 021 Gasterosteus gymnurus 828 Scardinius erythrophthalmus 17 813 Lampetra fluviatilis 817 Pungitius pungitius 16 149 Pachychilon pictum 695 Ameiurus melas 15 142 Gambusia holbrooki 636 Gymnocephalus cernuus 14 757 Barbus tyberinus 613 Squalius carolitertii 12 886 Rutilus rubilio 563 Squalius pyrenaicus 12 069 Gobio kesslerii 502 Chondrostoma toxostoma 9 874 Ameiurus nebulosus 467 Squalius alburnoides 9 855 Liza ramada 437 Tinca tinca 9 806 Pseudochondrostoma willkommii 432 Cobitis taenia 8 061 Hucho hucho 422 Barbus sclateri 7 996 Proterorhinus marmoratus 406 Leucaspius delineatus 7 889 Iberochondrostoma lusitanicum 383 Barbus meridionalis 7 639 Knipowitschia punctatissima 373 Pseudochondrostoma polylepis 6 775 Chelon labrosus 365 Osmerus eperlanus 6 488 Mugil cephalus 357 Aspius aspius 6 016 Neogobius gymnotrachelus 336 Cottus poecilopus 5 833 Cobitis calderoni 334

Species names Lengths

167 48

Eudontomyzon mariae 317 Barbus microcephalus 285 Vimba vimba 279 Salmo trutta macrostigma 232 Abramis ballerus 229 Chondrostoma genei 214 Misgurnus fossilis 195 Zingel streber 175 Salvelinus alpinus 168 Australoheros facetus 100 Barbus caninus 93 Zingel zingel 77 Gobio albipinnatus 62 Hypophthalmichthys molitrix 58 Barbus guiraonis 55 Sabanejewia aurata 48 Gobio uranoscopus 46 Sabanejewia larvata 40 Alosa alosa 36 Padogobius nigricans 34 Barbus haasi 31 Leuciscus lucumonis 31 Dicentrarchus labrax 30 Neogobius fluviatilis 30 Perccottus glenii 30 Atherina boyeri 25 Atherina presbyter 23 Salvelinus umbla 20 Pomatoschistus microps 19 Alburnus albidus 18 Pomatoschistus minutus 15 Abramis sapa 12 Ctenopharyngodon idella 12 Zingel asper 12 Coregonus albula 8 Salvelinus namaycush 8 Gymnocephalus schraetser 5 Knipowitschia panizzae 5 Pleuronectes platessa 5 Squalius malacitanus 5 Chondrostoma soetta 4 Umbra pygmaea 4 Alosa fallax 3 Aristichthys nobilis 2 Acipenser naccarii 1 Anaecypris hispanica 1 Coregonus lavaretus 1 Coregonus peled 1 Liza aurata 1 Neogobius melanostomus 1

168 49

Appendix B: Fish lengths for all species for the first run

Species names Lengths Species names Lengths Phoxinus phoxinus 721 930 Gobio lozanoi 4 361 Rutilus rutilus 662 634 Pseudorasbora parva 4 181 Salmo trutta fario 626 009 Cyprinus carpio 4 151 Gobio gobio 479 851 Achondrostoma arcasii 4 101 Leuciscus cephalus 339 192 Rutilus aula 3 944 Barbatula barbatula 337 856 Salmo marmoratus 3 939 Cottus gobio 262 175 Achondrostoma oligolepis 3 587 Alburnus alburnus 252 676 Barbus petenyi 3 389 Perca fluviatilis 157 214 Chondrostoma miegii 3 199 Anguilla anguilla 102 670 Romanogobio belingi 2 795 Leuciscus leuciscus 96 043 Carassius carassius 2 775 Barbus barbus 88 497 Platichthys flesus 2 479 Alburnoides bipunctatus 71 434 Barbus plebejus 2 193 Salmo salar 68 095 Cottus poecilopus 2 055 Rhodeus amarus 58 154 Sabanejewia balcanica 1 986 Leuciscus souffia 49 076 Barbus graellsii 1 911 Thymallus thymallus 45 168 Gambusia affinis 1 778 Blicca bjoerkna 44 213 Padogobius martensii 1 446 Abramis brama 43 847 Carassius gibelio 1 387 Oncorhynchus mykiss 37 408 Cobitis paludica 1 386 Leuciscus idus 36 320 Silurus glanis 1 318 Gasterosteus aculeatus 33 144 Barbus peloponnesius 1 285 Lepomis gibbosus 29 912 Micropterus salmoides 1 279 Lampetra planeri 27 561 Barbus comizo 1 039 Esox lucius 25 452 Iberochondrostoma lemmingii 916 Chondrostoma nasus 22 895 Salaria fluviatilis 881 Barbus bocagei 18 570 Carassius auratus 811 Scardinius erythrophthalmus 17 250 Gambusia holbrooki 630 Lota lota 16 352 Salvelinus fontinalis 615 Salmo trutta trutta 14 889 Barbus tyberinus 612 Pseudochondrostoma duriense 14 858 Pachychilon pictum 603 Ameiurus melas 14 520 Rutilus rubilio 563 Gymnocephalus cernuus 14 270 Petromyzon marinus 535 Pungitius pungitius 12 346 Leuciscus muticellus 521 Squalius carolitertii 10 159 Lampetra fluviatilis 505 Tinca tinca 9 208 Gobio kesslerii 499 Squalius pyrenaicus 9 111 Gasterosteus gymnurus 482 Chondrostoma toxostoma 8 799 Ameiurus nebulosus 467 Squalius alburnoides 8 752 Liza ramada 437 Cobitis taenia 7 941 Pseudochondrostoma willkommii 419 Leucaspius delineatus 7 713 Proterorhinus marmoratus 405 Osmerus eperlanus 6 486 Iberochondrostoma lusitanicum 383 Aspius aspius 6 013 Knipowitschia punctatissima 369 Barbus meridionalis 5 981 Hucho hucho 356 Pseudochondrostoma polylepis 5 952 Mugil cephalus 353 Salmo trutta lacustris 5 707 Neogobius gymnotrachelus 336 Barbus sclateri 5 075 Cobitis calderoni 326 Sander lucioperca 4 536 Eudontomyzon mariae 286

169 50

Species names Lengths Barbus microcephalus 285 Chelon labrosus 283 Vimba vimba 266 Salmo trutta macrostigma 230 Abramis ballerus 229 Chondrostoma genei 201 Misgurnus fossilis 195 Zingel streber 132 Australoheros facetus 100 Barbus caninus 63 Zingel zingel 59 Gobio albipinnatus 56 Hypophthalmichthys molitrix 55 Sabanejewia aurata 48 Gobio uranoscopus 46 Salvelinus alpinus 46 Barbus guiraonis 44 Sabanejewia larvata 40 Alosa alosa 36 Padogobius nigricans 34 Leuciscus lucumonis 31 Dicentrarchus labrax 30 Neogobius fluviatilis 30 Perccottus glenii 30 Barbus haasi 28 Atherina boyeri 25 Atherina presbyter 23 Pomatoschistus microps 19 Alburnus albidus 18 Salvelinus umbla 18 Pomatoschistus minutus 15 Abramis sapa 12 Ctenopharyngodon idella 12 Zingel asper 12 Coregonus albula 8 Salvelinus namaycush 8 Gymnocephalus schraetser 5 Knipowitschia panizzae 5 Pleuronectes platessa 5 Squalius malacitanus 5 Chondrostoma soetta 4 Umbra pygmaea 4 Alosa fallax 3 Aristichthys nobilis 2 Acipenser naccarii 1 Anaecypris hispanica 1 Coregonus lavaretus 1 Coregonus peled 1 Liza aurata 1 Neogobius melanostomus 1

170 51

Appendix C: The five most abundant species per country all runs confounded

Country Species names Lengths Country Species names Lengths Salmo trutta fario 109 978 Rutilus rutilus 62 603 Oncorhynchus mykiss 42 113 Perca fluviatilis 22 468 Austria Thymallus thymallus 38 086 Netherlands Leuciscus idus 8 355 Leuciscus cephalus 24 745 Anguilla anguilla 6 587 Cottus gobio 22 766 Osmerus eperlanus 6 068 Salmo trutta fario 98 587 Rutilus rutilus 18 763 Cottus gobio 21 423 Salmo trutta fario 10 307 Switzerland Phoxinus phoxinus 14 134 Poland Phoxinus phoxinus 5 318 Barbatula barbatula 11 149 Perca fluviatilis 4 755 Leuciscus cephalus 7 001 Alburnus alburnus 4 425 Rutilus rutilus 139 408 Barbus bocagei 15 314 Perca fluviatilis 71 512 Squalius alburnoides 6 688 Germany Gobio gobio 55 824 Portugal Squalius carolitertii 5 354 Alburnus alburnus 51 075 Squalius pyrenaicus 4 780 Cottus gobio 33 059 Pseudochondrostoma duriense 4 359 Salmo trutta fario 133 263 Phoxinus phoxinus 5 165 Pseudochondrostoma duriense 18 566 Leuciscus cephalus 5 137 Spain Phoxinus phoxinus 10 032 Romania Barbatula barbatula 3 403 Anguilla anguilla 8 783 Barbus petenyi 3 389 Barbus sclateri 7 835 Gobio gobio 2 204 Phoxinus phoxinus 822 673 Salmo trutta trutta 110 196 Gobio gobio 457 973 Salmo salar 103 194 France Rutilus rutilus 407 838 Sweden Salmo trutta fario 75 401 Barbatula barbatula 373 992 Salmo trutta lacustris 48 727 Salmo trutta fario 362 063 Cottus gobio 34 876 Salmo trutta fario 21 256 Rutilus rutilus 65 986 Alburnus alburnus 5 154 Salmo trutta fario 36 577 United Italy Leuciscus souffia 4 411 Kingdom Gobio gobio 24 512 Leuciscus cephalus 4 374 Leuciscus cephalus 20 504 Salmo marmoratus 4 055 Leuciscus leuciscus 18 370 Rutilus rutilus 4 211 Phoxinus phoxinus 1 894 Lithuania Gobio gobio 1 845 Alburnoides bipunctatus 1 450 Cottus gobio 1 161

171 52

Appendix D: The five most abundant species per country first run

Country Species names Lengths Country Species names Lengths Salmo trutta fario 91 610 Rutilus rutilus 62 603 Oncorhynchus mykiss 34 061 Perca fluviatilis 22 468 Austria Thymallus thymallus 32 857 Netherlands Leuciscus idus 8 355 Leuciscus cephalus 21 427 Anguilla anguilla 6 587 Cottus gobio 15 355 Osmerus eperlanus 6 068 Salmo trutta fario 77 607 Rutilus rutilus 18 763 Cottus gobio 10 428 Salmo trutta fario 10 307 Switzerland Phoxinus phoxinus 8 576 Poland Phoxinus phoxinus 5 318 Barbatula barbatula 5 903 Perca fluviatilis 4 755 Leuciscus cephalus 4 988 Alburnus alburnus 4 425 Rutilus rutilus 139 408 Barbus bocagei 15 314 Perca fluviatilis 71 512 Squalius alburnoides 6 688 Germany Gobio gobio 55 824 Portugal Squalius carolitertii 5 354 Alburnus alburnus 51 075 Squalius pyrenaicus 4 780 Cottus gobio 33 059 Pseudochondrostoma duriense 4 359 Salmo trutta fario 66 461 Phoxinus phoxinus 5 165 Pseudochondrostoma duriense 10 499 Leuciscus cephalus 5 137 Spain Phoxinus phoxinus 7 484 Romania Barbatula barbatula 3 403 Barbus sclateri 4 914 Barbus petenyi 3 389 Squalius carolitertii 4 805 Gobio gobio 2 204 Phoxinus phoxinus 652 403 Salmo salar 34 815 Rutilus rutilus 377 519 Salmo trutta trutta 11 528 France Gobio gobio 375 742 Sweden Salmo trutta fario 6 168 Salmo trutta fario 297 669 Salmo trutta lacustris 5 181 Barbatula barbatula 288 622 Phoxinus phoxinus 2 599 Salmo trutta fario 19 216 Rutilus rutilus 48 650 Alburnus alburnus 5 148 Salmo trutta fario 28 370 United Italy Leuciscus souffia 4 226 Kingdom Gobio gobio 17 019 Leuciscus cephalus 4 218 Leuciscus cephalus 14 960 Rutilus aula 3 944 Leuciscus leuciscus 13 888 Rutilus rutilus 4 211 Phoxinus phoxinus 1 894 Lithuania Gobio gobio 1 845 Alburnoides bipunctatus 1 450 Cottus gobio 1 161

172 53

Appendix E: Fish lengths of reference sites, for all species and all runs

Species Lengths Species Lengths Salmo trutta fario 21 895 Cobitis paludica 4 Phoxinus phoxinus 5 681 Rhodeus amarus 4 Barbatula barbatula 2 902 Chondrostoma genei 3 Barbus petenyi 2 347 Gymnocephalus cernuus 2 Leuciscus cephalus 1 853 Leuciscus idus 2 Alburnoides bipunctatus 1 535 Tinca tinca 2 Cottus gobio 1 522 Abramis brama 1 Gobio gobio 1 508 Carassius carassius 1 Sabanejewia balcanica 929 Platichthys flesus 1 Rutilus rutilus 681 Pungitius pungitius 1 Leuciscus souffia 446 Salvelinus alpinus 1 Salmo marmoratus 439 Vimba vimba 1 Anguilla anguilla 393 Salmo trutta trutta 354 Leuciscus leuciscus 320 Barbus plebejus 212 Salmo salar 208 Pseudochondrostoma polylepis 190 Alburnus alburnus 146 Salmo trutta macrostigma 127 Gobio kesslerii 124 Cottus poecilopus 114 Barbus barbus 110 Achondrostoma arcasii 100 Pseudochondrostoma duriense 100 Thymallus thymallus 98 Blicca bjoerkna 75 Squalius pyrenaicus 64 Padogobius martensii 53 Barbus bocagei 46 Perca fluviatilis 40 Salmo trutta lacustris 34 Chondrostoma nasus 33 Pseudorasbora parva 29 Barbus meridionalis 23 Salvelinus fontinalis 23 Esox lucius 20 Scardinius erythrophthalmus 13 Petromyzon marinus 12 Lota lota 11 Carassius gibelio 10 Cobitis taenia 10 Lampetra planeri 10 Oncorhynchus mykiss 10 Salvelinus namaycush 8 Gasterosteus gymnurus 7 Squalius carolitertii 6

173 54

Appendix F: Fish lengths of reference sites, for all species caught during the first run

Species Lengths Species Lengths Salmo trutta fario 16 002 Cobitis paludica 2 Phoxinus phoxinus 5 419 Gymnocephalus cernuus 2 Barbatula barbatula 2 897 Leuciscus idus 2 Barbus petenyi 2 347 Squalius carolitertii 2 Leuciscus cephalus 1 850 Tinca tinca 2 Alburnoides bipunctatus 1 535 Abramis brama 1 Gobio gobio 1 413 Carassius carassius 1 Cottus gobio 1 296 Lampetra planeri 1 Sabanejewia balcanica 929 Platichthys flesus 1 Rutilus rutilus 681 Salvelinus alpinus 1 Salmo marmoratus 439 Vimba vimba 1 Leuciscus souffia 436 Pungitius pungitius 0 Salmo trutta trutta 351 Leuciscus leuciscus 320 Anguilla anguilla 215 Barbus plebejus 206 Salmo salar 153 Alburnus alburnus 146 Pseudochondrostoma polylepis 143 Salmo trutta macrostigma 127 Gobio kesslerii 124 Barbus barbus 110 Thymallus thymallus 95 Blicca bjoerkna 75 Pseudochondrostoma duriense 62 Padogobius martensii 53 Achondrostoma arcasii 48 Cottus poecilopus 43 Perca fluviatilis 40 Salmo trutta lacustris 34 Chondrostoma nasus 33 Barbus bocagei 32 Pseudorasbora parva 29 Squalius pyrenaicus 24 Barbus meridionalis 21 Esox lucius 20 Salvelinus fontinalis 20 Scardinius erythrophthalmus 13 Lota lota 11 Carassius gibelio 10 Cobitis taenia 10 Oncorhynchus mykiss 10 Salvelinus namaycush 8 Petromyzon marinus 4 Rhodeus amarus 4 Chondrostoma genei 3 Gasterosteus gymnurus 3

174 55

Appendix G: The five most abundant species per country for reference sites, all runs

Country Species Lengths Salmo trutta fario 21 Cottus gobio 20 Austria Thymallus thymallus 5 Oncorhynchus mykiss 2 Salvelinus fontinalis 1 Salmo trutta fario 11 515 Anguilla anguilla 328 Spain Gobio gobio 235 Phoxinus phoxinus 221 Pseudochondrostoma polylepis 190 Salmo trutta fario 5 127 Cottus gobio 626 France Phoxinus phoxinus 458 Anguilla anguilla 53 Barbatula barbatula 25 Salmo trutta fario 2 567 Leuciscus souffia 446 Italy Salmo marmoratus 439 Barbus plebejus 212 Leuciscus cephalus 147 Phoxinus phoxinus 1 237 Alburnoides bipunctatus 963 Lithuania Rutilus rutilus 681 Cottus gobio 517 Gobio gobio 460 Salmo trutta trutta 325 Salmo trutta fario 124 Poland Salmo trutta lacustris 33 Perca fluviatilis 9 Esox lucius 4 Phoxinus phoxinus 3 531 Barbatula barbatula 2 558 Romania Barbus petenyi 2 347 Leuciscus cephalus 1 626 Salmo trutta fario 1 303 Salmo trutta fario 1 156 Phoxinus phoxinus 204 Sweden Cottus poecilopus 73 Lampetra planeri 9 Cottus gobio 8

175 56

Appendix H: The five most abundant species per country for reference sites, first run

Country Species Lengths Salmo trutta fario 21 Cottus gobio 20 Austria Thymallus thymallus 5 Oncorhynchus mykiss 2 Salvelinus fontinalis 1 Salmo trutta fario 7 813 Anguilla anguilla 166 Spain Phoxinus phoxinus 145 Pseudochondrostoma polylepis 143 Gobio gobio 140 Salmo trutta fario 4 192 Cottus gobio 400 France Phoxinus phoxinus 360 Anguilla anguilla 38 Barbatula barbatula 20 Salmo trutta fario 2 381 Salmo marmoratus 439 Italy Leuciscus souffia 436 Barbus plebejus 206 Leuciscus cephalus 144 Phoxinus phoxinus 1 237 Alburnoides bipunctatus 963 Lithuania Rutilus rutilus 681 Cottus gobio 517 Gobio gobio 460 Salmo trutta trutta 325 Salmo trutta fario 124 Poland Salmo trutta lacustris 33 Perca fluviatilis 9 Esox lucius 4 Phoxinus phoxinus 3 531 Barbatula barbatula 2 558 Romania Barbus petenyi 2 347 Leuciscus cephalus 1 626 Salmo trutta fario 1 303 Phoxinus phoxinus 116 Salmo trutta fario 86 Sweden Cottus gobio 8 Lota lota 8 Thymallus thymallus 6

176 57

Appendix I: Fish lengths of calibration sites, for all species and all runs

Species Lengths Species Lengths Salmo trutta fario 99 778 Barbus tyberinus 87 Phoxinus phoxinus 37 773 Blicca bjoerkna 73 Salmo salar 23 868 Carassius gibelio 67 Cottus gobio 23 638 Lampetra fluviatilis 47 Salmo trutta trutta 20 643 Cobitis taenia 45 Salmo trutta lacustris 14 149 Gasterosteus gymnurus 43 Barbatula barbatula 10 249 Pseudorasbora parva 41 Gobio gobio 6 978 Carassius carassius 40 Anguilla anguilla 6 110 Cobitis paludica 36 Leuciscus cephalus 4 265 Cyprinus carpio 33 Pseudochondrostoma duriense 4 032 Padogobius martensii 30 Rutilus rutilus 3 656 Chondrostoma nasus 27 Lampetra planeri 2 467 Lepomis gibbosus 23 Thymallus thymallus 2 173 Tinca tinca 18 Leuciscus souffia 1 718 Platichthys flesus 15 Lota lota 1 243 Petromyzon marinus 11 Cottus poecilopus 1 175 Vimba vimba 9 Barbus sclateri 1 166 Chelon labrosus 7 Oncorhynchus mykiss 1 076 Hucho hucho 7 Leuciscus leuciscus 1 011 Leuciscus idus 7 Esox lucius 917 Sabanejewia aurata 6 Barbus meridionalis 915 Carassius auratus 5 Squalius carolitertii 768 Abramis brama 4 Sabanejewia balcanica 705 Chondrostoma genei 4 Scardinius erythrophthalmus 629 Chondrostoma toxostoma 4 Pseudochondrostoma polylepis 619 Gambusia affinis 4 Perca fluviatilis 594 Gymnocephalus cernuus 4 Barbus barbus 577 Pseudochondrostoma willkommii 4 Achondrostoma arcasii 555 Ameiurus melas 1 Squalius pyrenaicus 490 Barbus graellsii 1 Barbus petenyi 445 Eudontomyzon mariae 1 Salvelinus fontinalis 428 Mugil cephalus 1 Gobio lozanoi 238 Sander lucioperca 1 Alburnoides bipunctatus 212 Silurus glanis 1 Barbus bocagei 212 Pungitius pungitius 198 Alburnus alburnus 196 Barbus plebejus 176 Gasterosteus aculeatus 138 Rutilus rubilio 137 Salaria fluviatilis 128 Gobio kesslerii 121 Salmo marmoratus 104 Rhodeus amarus 93

177 58

Appendix J: Fish lengths of calibration sites, for all species caught during the first run

Species Lengths Species Lengths Salmo trutta fario 61 902 Carassius gibelio 67 Phoxinus phoxinus 29 137 Cobitis taenia 45 Cottus gobio 11 923 Pseudorasbora parva 41 Salmo salar 11 542 Salvelinus fontinalis 36 Barbatula barbatula 7 933 Padogobius martensii 30 Gobio gobio 5 245 Cottus poecilopus 28 Anguilla anguilla 4 135 Chondrostoma nasus 27 Leuciscus cephalus 3 888 Cyprinus carpio 27 Rutilus rutilus 3 357 Cobitis paludica 23 Pseudochondrostoma duriense 2 188 Carassius carassius 20 Salmo trutta trutta 2 019 Gasterosteus gymnurus 17 Thymallus thymallus 1 951 Lepomis gibbosus 17 Leuciscus souffia 1 551 Tinca tinca 17 Salmo trutta lacustris 1 289 Platichthys flesus 14 Lampetra planeri 1 080 Vimba vimba 9 Leuciscus leuciscus 916 Chelon labrosus 7 Oncorhynchus mykiss 864 Hucho hucho 7 Barbus meridionalis 782 Leuciscus idus 7 Barbus sclateri 754 Sabanejewia aurata 6 Esox lucius 716 Chondrostoma genei 4 Sabanejewia balcanica 705 Chondrostoma toxostoma 4 Scardinius erythrophthalmus 629 Petromyzon marinus 4 Lota lota 611 Carassius auratus 3 Barbus barbus 560 Abramis brama 2 Barbus petenyi 445 Gymnocephalus cernuus 2 Perca fluviatilis 418 Ameiurus melas 1 Pseudochondrostoma polylepis 405 Barbus graellsii 1 Squalius carolitertii 383 Eudontomyzon mariae 1 Achondrostoma arcasii 321 Gambusia affinis 1 Alburnoides bipunctatus 212 Lampetra fluviatilis 1 Squalius pyrenaicus 207 Mugil cephalus 1 Pungitius pungitius 197 Pseudochondrostoma willkommii 1 Alburnus alburnus 181 Sander lucioperca 1 Barbus plebejus 176 Silurus glanis 1 Barbus bocagei 167 Rutilus rubilio 137 Gasterosteus aculeatus 134 Gobio kesslerii 121 Gobio lozanoi 118 Salmo marmoratus 104 Salaria fluviatilis 102 Rhodeus amarus 93 Barbus tyberinus 87 Blicca bjoerkna 73

178 59

Appendix K: The five most abundant species per country for calibration sites, all runs

Country Species Lengths Country Species Lengths Salmo trutta fario 6 242 Gobio gobio 381 Cottus gobio 1 258 Cottus gobio 358 Austria Thymallus thymallus 1 028 Lithuania Salmo trutta fario 255 Oncorhynchus mykiss 974 Phoxinus phoxinus 253 Lota lota 159 Leuciscus leuciscus 223 Salmo trutta fario 2 948 Salmo trutta fario 2 622 Phoxinus phoxinus 826 Rutilus rutilus 1 222 Germany Gobio gobio 554 Poland Scardinius erythrophthalmus 621 Thymallus thymallus 548 Esox lucius 579 Cottus gobio 463 Salmo trutta trutta 443 Salmo trutta fario 21 311 Leuciscus cephalus 1 467 Pseudochondrostoma duriense 4 032 Sabanejewia balcanica 705 Spain Gobio gobio 1 721 Romania Barbus petenyi 445 Anguilla anguilla 1 228 Gobio gobio 297 Barbus sclateri 1 166 Phoxinus phoxinus 259 Salmo trutta fario 40 343 Salmo trutta fario 24 116 Phoxinus phoxinus 31 859 Salmo trutta trutta 20 098 France Cottus gobio 13 849 Sweden Salmo salar 18 759 Barbatula barbatula 9 556 Salmo trutta lacustris 13 937 Anguilla anguilla 4 829 Cottus gobio 6 635 Salmo trutta fario 1 193 Salmo trutta fario 671 Cottus gobio 1 005 Salmo salar 317 United Italy Leuciscus souffia 318 Kingdom Leuciscus leuciscus 209 Barbus plebejus 176 Phoxinus phoxinus 195 Rutilus rubilio 137 Rutilus rutilus 174

179 60

Appendix L: The five most abundant species per country for calibration sites, first run

Country Species Lengths Country Species Lengths Salmo trutta fario 5 200 Gobio gobio 381 Cottus gobio 989 Cottus gobio 358 Austria Thymallus thymallus 982 Lithuania Salmo trutta fario 255 Oncorhynchus mykiss 781 Phoxinus phoxinus 253 Lota lota 101 Leuciscus leuciscus 223 Salmo trutta fario 2 948 Salmo trutta fario 2 622 Phoxinus phoxinus 826 Rutilus rutilus 1 222 Germany Gobio gobio 554 Poland Scardinius erythrophthalmus 621 Thymallus thymallus 548 Esox lucius 579 Cottus gobio 463 Salmo trutta trutta 443 Salmo trutta fario 14 460 Leuciscus cephalus 1 467 Pseudochondrostoma duriense 2 188 Sabanejewia balcanica 705 Spain Gobio gobio 825 Romania Barbus petenyi 445 Barbus sclateri 754 Gobio gobio 297 Anguilla anguilla 635 Phoxinus phoxinus 259 Salmo trutta fario 33 148 Salmo salar 7 375 Phoxinus phoxinus 26 590 Salmo trutta fario 1 652 France Cottus gobio 9 139 Sweden Salmo trutta trutta 1 478 Barbatula barbatula 7 256 Salmo trutta lacustris 1 077 Salmo salar 3 675 Phoxinus phoxinus 566 Salmo trutta fario 1 019 Salmo trutta fario 521 Cottus gobio 599 Salmo salar 261 United Italy Leuciscus souffia 318 Kingdom Leuciscus leuciscus 135 Barbus plebejus 176 Leuciscus cephalus 105 Rutilus rubilio 137 Phoxinus phoxinus 96

180