A statistic model Eel Density Analysis (EDA 2.0) to predict the European eel (Anguilla Anguilla) density in the French river network.

Since the early 1980s, the European eel ( Anguilla anguilla ) stock has been declining and continues to decline at an alarming rate. It is presently considered to be outside safe biological limits.

EDA 2.0 (Eel Density Analysis) is a modelling tool based on a GIS river network database to predict yellow eel densities and silver escapement. The principle of this approach is (1) to relate observed yellow eel densities to different parameters: fishing method, environmental conditions (distance to the sea, relative distance, temperature, Strahler stream order, elevation and slope…), anthropogenic conditions (obstacles, fisheries…) and time (year trends), (2) to calculate the yellow eel density in each reach by applying the statistical model calibrated in step 1, (3) to calculate the yellow eel stock abundance by multiplying these densities by the surfaces of the reaches (4) to calculate a potential escapement from the catchment (mortality during downstream mortality are neglected) from the yellow eel stock estimate in step 3. It is also possible to give an estimate of the pristine escapement by running the EDA model with anthropogenic conditions artificially set to zero.

It presently runs with BD_Carthage®, a spatial referencing system for surface water in France and will be tested on two European hydrographical databases, CCM (Catchment Characterisation and Modelling) (Vogt et al. 2007) and Ecrins (European Catchment and RIvers Network System). The presence/absence and densities of yellow eel in France are obtained from the Aquatic environment and fish database (BDMAP - more than 16,000 fishing samples collected on more than 6,000/8,968 (?) sampling stations (cf.coordination juillet 2008)) from the National Office of Water and the Aquatic Environments (ONEMA) and other databases from the Brittany watershed.

Values of the explanatory variables are calculated for each segment of the river network. The distance to the sea, the relative distance (between sea limit and the more upstream source) are directly calculated from the river network topology. The temperatures are extracted from the CRU (Mitchell et al., 2004), and Worlclim (www.worldclim.org/). Elevation and slope are extracted from the National Height Elevation Database (BD ALTI® - spatial resolution of 50m) from the National Geographic Institute. The obstacle pressure (characteristics, Steinbach rank…) comes from the National list of obstacles to river flows (ROE) from the ONEMA. Glass eel fisheries data set comes from Castelnaud (1994), non-professional/leisure fisheries data set from the ONEMA and professional fisheries data set from Thomas Changeux and Wenes (2004). The data sets used to extract the water quality parameters (which ones?) are obtained from the ROM or/and ? the RHP from the ONEMA.

The statistical model is calibrated with a Generalized Additive Model (Hastie and Tibshirani, 1990), using the ‘gam’ library in the R software. This semi-parametric extension of generalised linear models is flexible and allows combination of both linear and complex additive responses within the same model. The best model is selected by the Akaike’s Information Criterion (AIC), and Kappa when presence-absence models were used.

For technical reasons, silver eel densities are unachievable and exhaustive samplings are rare, so an indirect method is used to estimate the silver eel stock from the knowledge of the yellow eel stock. The silver eel stock is obtained with a conversion rate which will be calibrated according to known silver eel productions.

Within the POSE project, this EDA 2.0 model will be implemented in at least four geographical areas: the Brittany region (Leprévost, 2007), the Loire-Brittany basin (Hoffmann, 2008) , the Rhône and Vaccares basins, and the Basque country river basins. For management purpose, it will also be implemented across all of France (Beaulaton in French EMP). The implementation of the model in Loire Brittany allowed to predict the impact of river obstacles on densities, and test the obstacles grid from Pierre Steinbach.

REFERENCES

Castelnaud G., Guérault D., Désaunay Y. and Elie P., 1994. Production et abondance de la civelle en France au début des années 90. Bulletin Français de la Pêche et de la Pisciculture, 335, 263-288.

Crouzet, C. and W. Simonazzi (2008). Building the EEA European Catchment and Rivers Network System (ECRINS) from CCM v2.1, European ENvironment Agency: 15.

Hastie, T.J. and Tibshirani, R.J., 1990. Generalized Additive Models, New York: Chapman and Hall.

Hoffmann, 2008. Modélisation de l’impact des ouvrages sur les densités d’anguilles, dans le bassin Loire-Bretagne. Rapport de stage.

Leprévost, 2007. Développement d’un indicateur pour caractériser l’impact migratoire sur le stock d’anguille européenne à l’échelle des basins. Mémoire technique.

Mitchell, T. D., T. R. Carter, et al. (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), Tyndall Centre Working Paper.

Vogt, J., P. Soille, et al. (2007). A pan-European river and catchment database. Luxembourg, Joint Research Centre-Institute for Environment and Sustainability: 120.