Réf. Dankers & Feyen 2008

Référence bibliographique complète

DANKERS, R., FEYEN, L. 2008. Climate change impact on flood hazard in Europe: An assessment based on high resolution climate simulations,Journal of Geophysical Research, 113, D19105, doi :10.1029/2007JD009719.

Abstract: Global warming is generally expected to increase the magnitude and frequency of extreme precipitation events, which may lead to more intense and frequent river flooding. This work assesses the implications of climate change for future flood hazard in Europe. Regional climate simulations from the HIRHAM model with 12-km horizontal resolution were used to drive the hydrological model LISFLOOD, and extreme value techniques were applied to the results to estimate the probability of extreme discharges. It was found that by the end of this century under the Special Report on Emission Scenarios (SRES) A2 emissions scenario in many European rivers extreme discharge levels may increase in magnitude and frequency. In several rivers, most notably in the west and parts of eastern Europe, the return period of what is currently a 100-year flood may in the future decrease to 50 years or less. A considerable decrease in flood hazard was found in the northeast, where warmer winters and a shorter snow season reduce the magnitude of the spring snowmelt peak. Also in other rivers in central and southern Europe a decrease in extreme river flows was simulated. The results were compared with those obtained with two HIRHAM experiments at 50-km resolution for the SRES A2 and B2 scenarios. Disagreements between the various model experiments indicate that the effect of the horizontal resolution of the regional climate model is comparable in magnitude to the greenhouse gas scenario. Also, the choice of extreme value distribution to estimate discharge extremes influences the results, especially for events with higher return periods.


Organismes / Contact
Institute for Environment and Sustainability, Joint Research Centre, Ispra, Italy.

(1) - Paramètre(s) atmosphérique(s) modifié(s)
(2) - Elément(s) du milieu impacté(s)
(3) - Type(s) d'aléa impacté(s)
(3) - Sous-type(s) d'aléa
Precipitation   Floods  

Pays / Zone
Massif / Secteur
Site(s) d'étude
Période(s) d'observation

(1) - Modifications des paramètres atmosphériques

Changes in Meteorological Forcing

Changes in precipitation between scenario and control run in the three HIRHAM experiments:
Averaged over Europe, the changes in average annual precipitation are fairly small; however, the regional differences are large. In line with other studies with different RCMs and GCMs [see Christensen and Christensen, 2007], the three scenario runs show a strong increase in annual precipitation over northern Europe, and a strong decrease in the south, particularly over the Iberian Peninsula. [...]

Change in average seasonal precipitation in the HIRHAM H12A2 scenario run compared to the corresponding control run:
Over most of Europe the precipitation is increasing in winter and, to a lesser extent, in spring, while in summer and autumn the rainfall is generally decreasing, except in the north.

Change in the annual maximum 5-day accumulated rainfall (average over 30 years), as an indicator of extreme precipitation, in the H12A2, H50A2, and H50B2 scenario runs relative to the respective control runs:
The patterns of change in extreme rainfall are much more heterogeneous than in the average precipitation. [...] Strong but localized increases in heavy rainfall are simulated over much of Europe, except for the very south. The patterns of change are roughly similar in the two A2 scenario runs, but locally there are differences, for example over Finland. The H50B2 scenario results in a stronger increase over much of eastern Europe and southern France, but much smaller changes in the west. Overall, the magnitude of changes in precipitation in the B2 scenario is comparable to that in the two A2 experiments, while the temperature increase is much smaller: averaged over Europe, the temperature rise is 3.7 and 4.0 K in the H12A2 and H50A2 runs, respectively, compared to 2.7 K in the H50B2 run. This suggests that the regional precipitation changes are, at least to some extent, determined by interdecadal variability in the driving global model [...], rather than by the different greenhouse gas pathways.


Informations complémentaires (données utilisées, méthode, scénarios, etc.)

Model simulations:
[The authors] used results from a recent experiment with the RCM HIRHAM [Christensen et al., 1996] in which a very high horizontal resolution of ~12 km was adopted. This experiment has been conducted within the framework of the PRUDENCE project [Christensen et al., 2007]. The simulations consisted of two 30-year time slices: a control run with a GHG forcing corresponding to 1961–1990, and a scenario run corresponding to 2071–2100 according to the A2 scenario of the Intergovernmental Panel on Climate Change [IPCC, 2000]. These runs are referred to as ‘‘H12CL’’ and ‘‘H12A2,’’ respectively. In the control run, the lateral boundaries were derived from the HadAM3H high-resolution global atmosphere model, which itself had been forced by low-resolution observed sea surface temperature (SST) and sea-ice extent [see references in the study]. Furthermore [the authors] applied three HIRHAM simulations at a lower resolution of ~50 km that used essentially the same boundary conditions: a 30-year control run (‘‘H50CL’’) and two scenario runs with GHG forcing according to the IPCC scenarios A2 and B2 (‘‘H50A2’’ and ‘‘H50B2,’’ respectively).

Overview of the HIRHAM Regional Climate Model Runs Used in This Study:

Resolution, km
Gas Forcing
control run
scenario A2
control run
scenario A2
scenario B2

For more details on the climate simulations used in this study, see Christensen and Christensen [2007] and Frei et al. [2006].

(2) - Effets du changement climatique sur le milieu naturel

Sensibilité du milieu à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)

(3) - Effets du changement climatique sur l'aléa

Flood Magnitude

Change in the estimated 100-year discharge level in the H12A2, H50A2, and H50B2 scenario runs relative to the respective control runs:
The general patterns of change are similar between the three scenarios. In all cases the 100-year return level decreases strongly in the northeastern part of the continent (i.e., Finland, northern Russia and part of the Baltic States). Decreases can also be seen along the Norwegian coast and, to a varying degree, in central Europe and the southern half of the Iberian Peninsula. Strong increases in the 100-year flood level are simulated across much of western and central Europe, including parts of the Balkan, northern Italy and locally also in Sweden and southern Norway. The most notable difference in the H50B2 run compared to the two A2 scenarios is a strong increase over eastern Europe, including the downstream parts of the Danube River. [...]
Especially rivers originating in the Scandinavian mountains show an increase in flood hazard in spring and a decrease in summer, reflecting an earlier snowmelt. A similar pattern can be seen for rivers originating in the other mountain ranges in Europe: an increase in spring and a decrease in summer in the Alpine region, and an increase in winter and a decrease in spring, for example in the Pyrenees.

Change in the seasonal 100-year return level in H12A2 scenario run (Gumbel fit):
In other regions of Europe the changes in flood hazard reflect changes in extreme rainfall amounts rather than in snow accumulation. In the H12A2 run the flood hazard is increasing particularly in winter and, to a lesser extent, in spring. In parts of eastern Europe it is also rising strongly in summer. In northern Italy the flood hazard is rising in all seasons except summer. Increases in the 100-year discharge are projected even in areas where the climate is getting much drier on average, like in Spain and southern France.
On average the H12A2 run shows somewhat larger increases (and slightly smaller decreases) in the 100-year return level than the H50A2 run, especially in smaller rivers and upstream tributaries. However, at about 15% of the river grid cells the two experiments do not agree on the sign of change (that is, one experiment shows an increase of more than 5%, while the other shows a decrease of more than 5%). This dissimilarity is about the same for the two 50-km scenarios (H50A2 and H50B2).

Recurrence of Floods

Change in recurrence of a 100-year flood in the control run in H12A2 scenario run (Gumbel distribution):
[...] In several major European rivers, such as the Loire, Po, Elbe, Oder and parts of the Danube, the return period decreases to 50 years or less, sometimes even to 20 years, meaning that an event associated with the 100-year return level in the control period is more than twice as likely in the scenario period. At the same time the probability of occurrence decreases significantly in the northeast of Europe, as well as in several rivers in central and eastern Europe, and the Iberian Peninsula.

[see also Trend Analysis in the Scenario Period + Comparison of GEV and Gumbel + Discussion...]


In this paper [the authors] analyzed changes in flood hazard in Europe due to climate change, on the basis of simulations with a spatially distributed hydrological model driven by high-resolution regional climate model output. The results suggest that, by the end of this century under both the SRES A2 and B2 scenarios, in many rivers in Europe extreme discharge levels may become more frequent and more intense, while in other areas the flood hazard decreases. Contrary to previous studies a considerable decrease in extreme flows was found in the northeast, where warmer winters and a shorter snow season reduce the magnitude of the spring snowmelt peak. A decrease in flood hazard was also found in several rivers in central and southern Europe. In many parts of western and eastern Europe, on the other hand, the simulations suggest that what is currently a 100-year flood may become more than twice as likely in the scenario period, meaning that the return period decreases to 50 years or less. The disagreements between the various model experiments that were performed in the current study are in part related to the horizontal resolution of the regional climate model, which were found to be comparable in magnitude to differences due to the scenario of greenhouse gas development. Also the choice for a particular extreme value distribution to estimate discharge extremes may influence the results, especially for events with a longer return period of more than 20–50 years. More robust estimates may be obtained by applying resampling techniques to reduce the uncertainties in the parameter estimation of the extreme value distribution and by adopting a multimodel ensemble of different RCMs and driving GCMs to better take account of the uncertainty in climate simulations.


Paramètre de l'aléa
Sensibilité des paramètres de l'aléa à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)

Model simulations:
The HIRHAM simulations of temperature, precipitation, solar and thermal radiation, humidity and wind speed were used to drive the hydrological model LISFLOOD [De Roo et al., 2000]. [see details in the study]

Model simulations:
[The study] is based on the very high resolution climate simulations that were evaluated by Dankers et al. [2007]. These simulations were derived from a recent experiment whereby the RCM HIRHAM was run with a horizontal resolution of 12 km, much higher than the 25 or 50 km typically used in RCM simulations. The HIRHAM output was used to drive the hydrological model LISFLOOD that has been developed for operational flood forecasting at European scale. The simulated river discharges were analyzed with respect to changes in the maximum flows. In order to investigate to what extent the projected changes are determined by the resolution of the climate data, two further HIRHAM simulations were included in the analysis: one driven by the same global atmosphere model and GHG emission scenario as the 12-km experiment but with a lower horizontal resolution of ~50 km, and a second one, also at 50-km resolution but driven by a different scenario of GHG emissions. In this way [the authors] could not only test whether the (tentatively) better representation of precipitation events in the 12-km simulation leads to different patterns of change in flood hazards, but also how these differences compare to uncertainties in the future development of GHG emissions.

Statistical Analysis:
Since LISFLOOD is a spatially distributed model, it produces runoff for every grid cell, which is then routed through the river network using a kinematic wave approach. To estimate the probability of extreme discharge levels, a generalized extreme value (GEV) distribution [Coles, 2001; Katz et al., 2002] was fitted to the annual maximum values in every grid cell. The GEV is a three-parameter distribution defined by a location, scale and shape parameter.
[...] To test the sensitivity of the results to uncertainties related to the estimation of the shape parameter, a second fit of the extreme value distribution was done with the shape parameter set to 0, which is commonly known as the Gumbel distribution. [see details in the study]

Validation Results:
The simulated extreme discharges from the H12CL and H50CL LISFLOOD model runs were compared with observations at 209 gauging stations across Europe for which long enough daily data (30 years covering 1960–1990 or 1970–2000) were available.
[see details in the study]

(4) - Remarques générales

(5) - Syntèses et préconisations

Références citées :

Christensen, J. H., and O. B. Christensen (2007), A summary of the PRUDENCE model projections of changes in European climate by the end of this century, Clim. Change, 81, 7– 30, doi:10.1007/s10584-006- 9210-7.

Christensen, J. H., T. R. Carter, M. Rummukainen, and G. Amanatidis (2007), Evaluating the performance and utility of regional climate models: The PRUDENCE project, Clim. Change, 81, 1 – 6, doi:10.1007/s10584-006-9211-6.

Coles, S. (2001), An Introduction to Statistical Modeling of Extreme Values, Springer, London.

Dankers, R., O. B. Christensen, L. Feyen, M. Kalas, and A. de Roo (2007), Evaluation of very high resolution climate model data for simulating flood hazards in the Upper Danube Basin, J. Hydrol., 347, 319–331, doi:10.1016/j.jhydrol.2007.09.055.

De Roo, A. P. J., C. G. Wesseling, and W. P. A. Van Deurzen (2000), Physically-based river basin modelling within a GIS: The LISFLOOD model, Hydrol. Processes, 14, 1981 – 1992, doi:10.1002/1099-1085(20000815/30)14:11/12<1981::AID-HYP49>3.0.CO;2-F.

Frei, C., R. Schöll, S. Fukutome, J. Schmidli, and P. L. Vidale (2006), Future change of precipitation extremes in Europe: Intercomparison of scenarios from regional climate models, J. Geophys. Res., 111, D06105, doi:10.1029/2005JD005965.

Katz, R. W., M. B. Parlange, and P. Naveau (2002), Statistics of extremes in hydrology, Adv. Water Resour., 25, 1287 – 1304, doi:10.1016/S0309-1708(02)00056-8.