Version classiqueVersion mobile
OpenEdition Books

The Mediterranean region under climate change

Jean-Paul Moatti
Stéphane Thiébault

Part 1. Mechanisms, observed trends, projections

Sub-chapter 1.3.1. Heavy precipitation in the Mediterranean basin

Observed trends, future projections

Gilles Molinié, Michel Déqué, Erika Coppola, Juliette Blanchet et Luc Neppel

Texte intégral

1This sub-chapter focuses on extreme rainfall around the Mediterranean basin. It provides an overview of the state-of-the-art knowledge of past and expected future changes.

2The high vulnerability of the Mediterranean area to future anthropogenic climate change has been pointed out since the early 2000s (Giorgi 2006) based on the consistency of the climate projections made by many atmosphere-ocean general circulation models (AOGCM) in the successive phases of the Coupled Model Intercomparison Projects (CMIP).

3Whatever the model types, climate projections are resolved at scales of a few kilometers to hundreds of kilometers. The question of how humans will perceive these projections at the scale of an urban area or a small catchment basin is still open. Behind this is the question of the scaling of rainfall with horizontal resolution and of the processes involved when a physical risk becomes a human hazard. We tackle these questions in part by examining how changes in extreme rainfall in the recent past scale between regional and local diagnostics in the following section. In the second section, we review the main results of regional climate simulations of extreme rainfall.

Extreme precipitation trends in the recent past

Methods for trend analysis of extreme rainfall

4Trends in extreme rainfall reported in the literature are examined using a variety of methodologies to which the results are sensitive.

5Two main approaches exist. The first is based on climate indices. Most indices describe the occurrence and the intensity of moderate extremes that typically occur several times a year. The Expert Team on Climate Change Detection, Monitoring and Indices (ETCCDI) recommends the use of 11 indices related to precipitation. This broad spectrum of indices is used to cope implicitly with the variety of rainfall. The relevance of the indices can be discussed when comparing their use from a region to another and their partial assessment of the extreme rainfall phenomenon.

6The second set of approaches focuses on the heaviest rainfall fully characterizing them by modeling their probability density function in the framework of the “Extreme value theory” (Coles, 2001). These approaches have the advantage of being scale independent. Thus the same theoretical framework can be used to analyze precipitation at different scales (rain gauges, climate model fields, etc.). Moreover fitting a probability density function onto the data implicitly mixes the intrinsic information of the dataset and the theoretical knowledge behind the statistical model. This leads to implicit filtering of sampling issues in rainfall data. However such an approach is seldom used in the literature.

7Most studies assess changes in extreme precipitation by regressing climate indices with time or in the case of the statistical model approach, by capturing the temporal evolution of one or more parameters of the statistical model.

8Another approach is to consider, rather than time, state parameters of the atmosphere (temperature, humidity) or a large scale diagnosis of atmospheric circulation (North Atlantic oscillation, NAO) as covariates either in the climate index regressions or in the statistical model parameters. This framework is useful to compare past and future climates, however it does not really provide a quantitative assessment of the rate of change.

From regional to local trends in the recent past

9Casanueva et al. (2014) provided a regional picture of the trend in extreme rainfall on the western side of the Mediterranean basin by analyzing changes in the relative contribution of daily rainfall above the 95th percentile to the total amount of rain (R95pTOT) for the period 1950-2010. Their study was based on the E-OBS data (high-resolution gridded data set of daily climate over Europe) at a resolution of 0.5°. At this spatial resolution, the trend is either significantly positive at specific points or null for most of the areas, but seldom negative. Significant positive trends are in the range of 1.5% to 4% per decade, which corresponds to changes in the range of 8% to 25% from the 1950s to the present.

10Over North Africa, R95pTOT shows almost no trend except at the extreme east of the Atlas Massif, the north of the Algerian-Tunisian border where it tends to increase by about 4% per decade. None of the local studies in this region (Tramblay et al. (2012a); Donat et al. (2014) and Taibi et al. (2015)) corroborates the increase in heavy rainfall at the large scale of 0.5° reported in Casanueva et al. (2014) nor reveals any consensus in this area.

11In the north-western part of the Mediterranean basin, the positive trends of R95pTOT at 0.5° resolution are reported over the main mountain massifs from the eastern Pyrenees in Spain to the Alps and Balkans around the Adriatic and Black Sea. In these regions, at some local spots, the extreme rainfall shows a positive trend of around 40% for the 90th percentile of daily rainfall, but the results depend on the area studied and on how extreme rainfall is defined. This is particularly apparent over the Iberian Peninsula, where contrasting rainfall regimes are observed from the ocean shore to inland Spain.

12Kostopoulou and Jones (2005) collected rain gauge data from 10 countries around the Mediterranean from Italy to Israel and computed the ETCCDI indices characterizing extreme rainfall. The dataset spans the period from 1958 to 2000. In the western part of the basin, they obtained similar results to those described above, but their results in the eastern part differed, although they were in agreement with those of other local studies. The quasi-homogeneously positive trend in extreme rainfall in Italy, can become significantly negative moving eastwards toward the Balkans, Peloponnese, Turkey and Cyprus. In Israel, these trends vary from significantly negative to significantly positive passing through zero among sites located only a few tens of kilometers apart.

13Again it appears that designing a robust method to study extreme rainfall is still a scientific challenge (see Box 1.)

Regional change in extreme rainfall in future climates

Modeling background

14Most models predict a reduction in annual precipitation, associated with a larger increase in temperature than in the surrounding regions, except in snow covered areas that are subject to strong winter warming due to snow-albedo feedback.

Box 1
Trends in extreme rainfall in southern France
The Mediterranean region of southern France is prone to heavy rainfall that can have major impacts on humans and society in general. In the framework of the HyMeX program, a sensitive study was conducted on how the definition of extreme rainfall influences the resulting extreme rainfall trends. Different methods are being applied independently by three French labs (HSM-Montpellier, CNRM and LTHE Grenoble) on the same dataset. The first result is that the statistical extreme value theory is the most efficient framework in the context of extreme rainfall and its possible evolution.
Figure 1 shows the relative trend in the annual maxima of daily rainfall computed from both rain gauges and spatialized rainfall data from 1958 to the present. Extreme rainfall in the study region was most likely stationary from 1958 to 1985. From 1985 on, we hypothesize a linear evolution in the statistical distribution of extreme rainfall. The figure below shows that the trend is spatially highly variable over the region and sometimes even at neighboring rain gauges such as over the mountain range delineated by the location of the main mountain peaks (triangles). There is a wide region in the center of the map extending from the Cévennes mountain range to the Alps passing though the Rhône river valley, in which both local and spatially extreme rainfall have increased by 20% to 60% in 30 years.

Figure 1
Increase (in%) in the annual maxima of daily rainfall during the period 1985-2014 relative to the average of the series of annual maxima of daily rainfall for each rain gauge (colored dots) and of the spatial data (colored 8x8 km2 squares). Adapted from Blanchet et al. (2016).

15However, the complexity of both the topography and the land/sea distribution of the Mediterranean area makes it impossible for either the atmospheric or the ocean models used in the Coupled Model Intercomparison Project (CMIP) to faithfully represent the Mediterranean features. For example, atmospheric models with a horizontal resolution of more than 100 km cannot perfectly represent the Massif Central and the Rhône valley and are consequently unable to reproduce the northerly Mistral winds over the Rhône valley.

16For these reasons, a concerted strategy of climate simulations at higher resolution has been implemented in the Mediterranean basin. In the PRUDENCE project (Christensen et al. 2007) a first approach was attempted with 50 km resolution regional climate models over Europe (RCMs), driven by observed sea surface temperature (SST) in the present climate or with SST plus the local SST response by Hadley Centre Coupled Model. The results confirmed and even amplified the risk of drought over southern Europe. Unfortunately several RCMs did not cover the southern part of the Mediterranean basin. Another limitation of the project was the use of a single greenhouse gas (GHG) scenario (SRES-A2) and a single SST scenario (HadCM). In the following European project, ENSEMBLES (van der Linden, 2009), the horizontal resolution was increased to 25 km, the whole Mediterranean basin was taken into account in the RCM domains, and several SST scenarios were used (most including the bias of the coupled GCM). The GHG scenario was SRES-A1B, considered to be more credible than SRES-A2. ENSEMBLES results confirmed the consistent response over the Mediterranean area, and added a probability layer, with the use of weights for the different GCMs (Christensen et al. 2010)

17The CORDEX project (Giorgi et al. 2009) was an extension of the former PRUDENCE, ENSEMBLES, and other extra-European projects to all continents of the world. The common denominator is a minimum horizontal resolution of 50 km and the use of two GHG scenarios (RCP45 and RCP85). Two sub-projects concern the Mediterranean area: EURO-CORDEX (Jacob et al. 2010) covers a larger domain (from Iceland to the Red Sea) with a higher resolution (12 km). MED-CORDEX (Ruti et al. 2016) kept the original 50 km resolution, but included a regional model for the Mediterranean Sea, rather than prescribing SST from CMIP5. CORDEX is not only a scenario experiment, but includes validation with the RCM driven by ERA-interim reanalyses. CORDEX is still an on-going project. Despite the abundant literature on the validation phase, still no synthesis papers have been published specifically concerning extreme precipitation scenarios over the Mediterranean basin.

Regional climate change: response to +2 °C

18The most recent attempt to synthesize the climate response of a large number of RCMs over the Mediterranean basin (and more generally over Europe) was the IMPACT2C European project. This project relied on high resolution (12 km) contributions to EURO-CORDEX. The originality of the approach, compared with FP5-PRUDENCE or FP6-ENSEMBLES, is the choice of the time period to define the future climate. Rather than a fixed 30-year period like 2021-2050 or 20712100, in which all models use the same GHG concentration (for a given scenario), each RCM uses a 30-year period for which the mean global temperature in the driving GCM is 2 °C above the pre-industrial level. As the reference period to evaluate the climate change at regional scale is 1971-2000, this means a global temperature increase of 1.54 °C with respect to this reference.

19With this new approach, the model spread is reduced, at least as far as temperature is concerned: by construction the spread of the global temperature responses is zero. The GCMs with high climate sensitivity to GHG impose an early period for the “future climate” in the RCMs they control. For example, in the RCP8.5 scenario, the center of such a period among the CMIP5 GCMs spans 2020 to 2060. Mean precipitation (not shown) tends to decrease in any season, but the decrease is generally modest (less than 5%). Heavy precipitation can be measured by the 95th percentile of the amount of precipitation on rainy days (defined as days with precipitation of more than 1 mm). This parameter is maximum in the fall (up to 35 mm/day in some areas). More than 80% of the models agree on the sign of the response over the major part of the domain, which is a good consensus. Figure 2 shows the mean model response for the change in the 95th percentile of daily precipitation in summer and fall. Some areas of France show a negative response (less heavy precipitation) in summer. The same goes for Morocco in winter/spring (not shown). But most of the Mediterranean basin undergoes an increase, which is maximum in summer in the southern part and in fall/winter in the northern part. In summer, the relative increase in the 95th percentile of the daily rainfall reaches 20%. However, despite this spectacular increase, the 95th percentile in summer remains well below that in the fall and the risk involved in this increase might be low. Yet in the fall, although lower, the relative increase reaches more than 10% at some spots in the northern part of the basin where the 95th percentile is already among the highest and where flooding events already take place.

Figure 2
95th percentile of precipitation (mm/day) for the Eurocordex multimodel (12 km resolution) for 1971-2000 (first row) for summer (left) and autumn (right); changes (mm/day) in this percentile during the +2 °C period with respect to 1971-2000 (second row).

Regional climate change: variability of the response to socio-economic scenarios

20Jacob et al., 2013 reported climate projections at the end of the century (2071-2100) from the new Euro-CORDEX ensemble. They showed a reduced northwards shift of Mediterranean drying evolution and slightly stronger increase in mean precipitation over most of Europe compared to the previous projections of the ENSEMBLE models.

21Improvements due to the high-resolution in the EURO-CORDEX simulations are clearly visible in the pattern on change for heavy precipitation events. The projected seasonal mean changes in heavy precipitation for the three emission scenarios (RCP8.5, A1B, RCP4.5) are similar, but some regional differences are visible.

22The results of RCP8.5 include a possible decrease of around 25% in heavy summer precipitation over some parts of the Iberian Peninsula and southern France, accompanied by regional increases in parts of Spain and Portugal. For winter, RCP8.5 projects up to 35% increases in heavy precipitation in Central and Eastern Europe, whereas A1B projects changes only up to 25% in the same region.

23In this study, both GCM and RCM ensembles show a reduction in weak intensity precipitation and an increase in strong intensity precipitation, and these are more pronounced in the RCP8.5 scenario.


24The first part of this chapter shows how heavy precipitation scales from regional to local scales in the recent past. At local scale, the extreme rainfall trend can increase by a factor of 2 compared to the regional assessment. In the future climate characterized by an increase of about 2 °C in the global temperature, extreme daily rainfall (95th percentile) is expected to increase by about 10% relative to the current level. This is the same order of magnitude as the increase observed at regional scale in the recent past. At the local scale, it corresponds to an increase of 30% to 50% in southern France, for example. Research is underway on the attribution of extreme rainfall trends and of their spatial variability. Efforts will also made to evaluate the new ‘convection enabled’ simulations at kilometric scale that will be performed in Euro-CORDEX/Med-CORDEX in the future and the differences in the way the simulated precipitations respond to climate change from those simulated by the current lower resolution climate models.



Blanchet J., G. Molinié and J. Touati, 2016
Spatial analysis of trend in extreme daily rainfall in southern France, Climate Dynamics, doi: 10.1007/s00382-016-3122-7, accepted.

Casanueva A., Rodriguez-Puebla C., Frias M. D., Gonzalez-Reviriego, N., 2014
Variability of extreme precipitation over Europe and its relationships with teleconnection patterns. Hydrology and Earth System Sciences, 18 (2): 709-725, doi: 10.5194/hess-18-709-2014.

Christensen J. H., Carter T. R., Rummukainen M., Amanatidis G., 2007
Evaluating the performance and utility of regional climate models: the PRUDENCE project, Clim. Change, 81:1-6.

Coles S., 2001
An Introduction to Statistical Modeling of Extreme Values. Springer-Verlag London Limited 2001.

Donat M., et al., 2014
Changes in extreme temperature and precipitation in the arab region: long-term trends and variability related to ENSO and NAO. Int. J. Climatol., 34:581-592.

Gallant A. J. E., Karoly D. J., Gleason K. L., 2013
Consistent trends in a modified climate extremes index in the United States, Europe, and Australia. Journal of Climate, 27 (4): 1379-1394, doi: 10.1175/JCLI-D-12-00783.1.

Giorgi, F., 2006
Climate change hot-spots. Geophys. Res. Lett., 33: L08707, doi: 10.1029/2006GL025734.

Giorgi F., Jones C., Asrar G. R., 2009
Addressing climate information needs at the regional level: the CORDEX framework, WMO Bulletin, 58: 175-183.

Giorgi F., IM E.-S., Coppola E., Diffenbaugh N. S., Gao X. J., Mariotti L., Shi Y., 2011
Higher hydrocli- matic intensity with global warming. Journal of Climate, 24 (20): 5309-5324, doi: 10.1175/2011JCLI3979.1.

Jacob et al., 2013
EURO-CORDEX: new high-resolution climate change projections for European impact research. Regional Environmental Change, Springer Berlin Heidelberg Ed., 1-16.

Kostopoulou, E., D. P. Jones, 2005
Assessment of climate extremes in the eastern Mediterranean. Meteorology and Atmospheric Physics, 89 (1): 69-85, doi: 10.1007/s00703-005-0122-2.

Ruti P. M., Somot S., Giorgi F., et al., 2016
MED-CORDEX initiative for Mediterranean Climate studies Bulletin of the American Meteorological Society. 97: 1187-1208

Taibi, S., M. Meddi, G. Mahe, A. Assani, 2015
Relationships between atmospheric circulation indices and rainfall in northern Algeria and comparison of observed and RCM-generated rainfall. Theor Appl. Climatol., 121 (3-4), doi: 10.1007/s00704-015-1626-4.

Tramblay, Y., W. Badi, F. Driouech, S. E. Adlouni, L. Neppel, and E. Servat, 2012
Climate change impacts on extreme precipitation in Morocco. Global and Planetary Change, 82-83, 104-114, doi: 10.1016/j. gloplacha. 2011.12.002.

Van der Linden P. Mitchell J. F. B., 2009
Climate change and its Impacts: Summary of research and rresults from the Ensembles projectresults from the ENSEMBLES project, Met Office Hadley Centre, Exeter, UK.

Table des illustrations

Légende Figure 1Increase (in%) in the annual maxima of daily rainfall during the period 1985-2014 relative to the average of the series of annual maxima of daily rainfall for each rain gauge (colored dots) and of the spatial data (colored 8x8 km2 squares). Adapted from Blanchet et al. (2016).
Fichier image/jpeg, 280k
Légende Figure 295th percentile of precipitation (mm/day) for the Eurocordex multimodel (12 km resolution) for 1971-2000 (first row) for summer (left) and autumn (right); changes (mm/day) in this percentile during the +2 °C period with respect to 1971-2000 (second row).
Fichier image/jpeg, 171k


Climatologist, Laboratoire d’étude des Transferts en Hydrologie et Environnement, Université Joseph Fourier, Grenoble, France

Forest ecologist, UR RECOVER, IRSTEA, France

Climatologist, Earth System Physics Section, The Abdus Salam International Centre for Theoretical Physics, Italy

Mathematician, Laboratoire d’étude des Transferts en Hydrologie et Environnement, Université Joseph Fourier, Grenoble, France

Hydrologist, HydroSciences Montpellier, Université de Montpellier, France

© IRD Éditions, 2016

Conditions d’utilisation :