linked in logo | IS-ENES | Contact

Exploring climate model data

Impact of global changes on freshwater fish biodiversity at large scale


Developing a hydro-climatic-ecological (HCE) model-chain to link large scale climate data to local scale occurrence of aquatic biodiversity.


The Water Framework Directive (WFD) aims to improve the chemical and ecological status of European freshwaters (European Parliament, 2005). Such improvement is likely to be achieved by reductions in pollutant. However, natural freshwater ecosystems are structured by regional (altitude, geology, slope, land use) and local (point or diffuse inputs of elements, stream depth and width, stream velocity); environmental factors which are highly variable in space and time and likely to be sensitive to future climate variations, particularly those in temperature and precipitation. To understand how projected climate change will impact the freshwater ecology, it is important to determine the inter-relationships between climate and hydrology and the response of the aquatic ecology to changes in habitat and food-web structure. This understanding is required to develop informed management plans regarding the use of water resources whilst protecting the ecological services of surface waters. As part of this research effort, a hydro-climatic-ecological (HCE) model-chain was developed for south-west France to test hypotheses regarding how the climate controls fish communities through invoked changes in the regional hydrology and temperature.

Author (s)

Mathieu Vrac (LSCE/IPSL, France).


Climate scientists from the IPSL (Paris); ecological scientists from the EDB laboratory (Evolution et Diversité Biologique, Paul Sabatier University, Toulouse); and people implied in the Euro-Limpacts European project.

Related Use Cases


Data Needs

Observation data:
Hydrological data: Adour river monthly flows for 1970-2000 period from MEDDAD database.
Meteorological data: monthly temperature and precipitation of SouthWest France observation stations for 1970-2000 period from Météofrance database.
Biodiversity data: occurrence (presence or absence) for the 13 most prevalent freshwater fish species in the region. The ONEMA (“Office National des Eaux et des Milieux Aquatiques”) database includes values from 50 sites for the 1992-2000 period.

Modelling data:
Monthly data from 13 IPCC models for 1970-2000. 4 variable groups around 1) temperature, 2) pressure, 3) precipitation, 4) short wave radiation are used. Monthly data from 5 IPCC models for 2005-2100 period. Same variables.

NCEP data for the same variable than modelling data. Period:1970-2000.

Typical Course of Events

Gathering observed data sets needed for the study from several agencies, laboratories…

Development and validation of the HCE model
The HCE model is composed of 2 parts: a statistical downscaling step (SD) converts global gridded data (temperature, pressure, precipitation, short wave radiation groups) into local fields measured at stations (temperature, precipitation, flow); an “ecological niche” model (EN) turns the local fields into freshwater fish occurrence. These two parts of the HCE model are based on aggregated boosted trees (ABT) statistical method. HCE model has been developed from NCEP reanalysis data, local meteorological data from Météofrance, local flows from MEDDAD and occurrence data for the 1970-1985 period (apprentissage phase). Validation phase of the model has been performed from the 1985-2000 period. The lack of biodiversity data (available only for 1992-2000) is tackled in mixing data from the 50 sites in the same data set.

Selection of the “best” model data sets
Analysing the main strengths and weaknesses of 13 IPCC model data in comparison to the NCEP reanalysis fields for both periods 1970-1985 and 1985-2000. The analysis is based on the 3 following indicators: mean; standard deviation; pdf deviation (Kolmogorov-Smirnov test). Selection of the most appropriate simulated datasets for the study (5 models have been chosen).

Correction of modelling data
Selection of a correction method regarding the data biases, the geographical zones of interest and the data availability (cdft method).

Selection of a data set to perform the correction
observations, re-analyses. NCEP reanalysis has been selected for this step. Development or adaptation of correction code source. Validation of corrected data.

Testing the HCE model performances for model corrected data
Period: 1985-2000. Comparison with observed flow, temperature, precipitation, and fish occurrence data. Performing fish occurrence projections from climate model data set (2005-2100).

Statistical analysis of the multi-models results
Trend and variability Evaluation of the signal trend compared to natural variability;
Results significativity and robustness Interpretation and graphical display;
Writing papers for international scientific review and thesis report.


  • C. Tisseuil, M. Vrac G. Grenouillet, M. Gevrey, T. Oberdorff, A.J. Wade, J.-B. Grodwohl, S. Lek (2012): 'Strengthening the link between hydro-climatic downscaling and species distribution modelling: Climate change impacts on freshwater biodiversity', accepted by Science of the Total Environment (STOTEN).
  • Tisseuil C., 2009. Impact of global changes on freshwater fixh biodiversity at large scale. PhD Thesis.
  • Tisseuil C., Vrac M., Lek S., Wade A.J.,2009 (in revision). Statistical downscalling of river flows.
  • Tisseuil C., Vrac M., Wade A.J., Grenouillet G., Gevrey M., Lek S., 2009 (in preparation). Validating a hydro-ecological model to project fish community structure from general circulation models using downscaling techniques.


CDFT correction method.

Aggregated boosted trees (ABT) statistical method for the HCE model.

Model evaluation based on mean, standard deviation and pdf comparisons of observed and simulated data. For the pdf one the Kolmogorov-Smirnov test has been used.

Software Use


Files Format(s)

NetCDF Ascii



The ENES3 project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 824084.