Future Projections of Extreme Rainfall and Temperature
Conditions over the Mediterranean Region: Scenarios from Three
Updated Regional Climate Models Dr. Konstantia Tolika ESF-MedCLIVAR
Workshop: Climate Change Modeling for the Mediterranean region
Trieste, Italy, 13-15 October 2008 Co authors: Efthimia
Kostopoulou, Ioannis Tegoulias, Christina Anagnostopoulou and
Panagiotis Maheras Department of Meteorology and Climatology
Aristotle University of Thessaloniki, Greece Slide 2 Main scope:
study the occurrence of specific climate extremes in the
Mediterranean region Outline: Description of Regional Climate
Models used Evaluation of Regional Climate Models vs. Observational
Gridded data Spatial analysis of Climate extreme Indices for
reference and future periods Maximum and minimum values of Indices
Trend Analysis of Climate Indices Conclusions Slide 3 Description
of Regional Climate Models used C4IRCA3 Community Climate Change
Consortium for Ireland (C4I). Based on IPCC SRES A2 Driven by
ECHAM5 (the 5th generation of the ECHAM General Circulation Model
(GCM)) developed at the Max Planck Institute for Meteorology in
Hamburg. RCA3 the third version of the Rossby Centre Atmospheric
model (Kjellstrm et al., 2005). Spatial resolution 25x25km 206x206
horizontal grid-points 31 vertical levels. Time period 1960-2050.
Slide 4 Description of Regional Climate Models used CNRM-RM4
Mto-France/CNRM (Centre National de Recherches Mtorologiques) (Dqu
and Somot (2007) and Radu et al. (2008)). SRES A1B scenario.
120x128 horizontal grid points 31 vertical levels. Spatial analysis
25x25km Time period 1960-2050. Parent model ALADIN (Aire Limite
Adaptation Dynamique INitialisation)
(http://www.cnrm.meteo.fr/aladin,).http://www.cnrm.meteo.fr/aladin
Slide 5 Description of Regional Climate Models used KNMI-RACMO2:
Royal Netherlands Meteorological Institute (KNMI- Koninklijk
Nederlands Meteorologisch Instituut) (Lenderink et al., 2003; van
den Hurk et al., 2006) Parent ECHAM5 Time period 1960-2100 SRES
A1B. Physical parameterizations of CMWF (European Centre for Medium
Range Weather Forecasts) used also for ERA-40
(http://www.ecmwf.int/research/ifsdocs).http://www.ecmwf.int/research/ifsdocs
Spatial Resolution 25x25km. 114 grid points in longitudinal
direction and 100 in latitudinal direction. 40 vertical levels
Slide 6 Description of Regional Climate Models used
InstituteScenario Driving GCM RCMResolutionAcronymTime
C4IA2ECHAM5RCA325kmC4IRCA3 1951-2050 CNRMA1BARPEGEALADIN25km CNRM-
RM4.5 1951-2050 KNMIA1BECHAM5RACMO25km KNMI- RACMO2 1951-2100 Slide
7 Climate extreme indices studied for the Mediterranean region
IndexDefinition TXQ90 Days with Tmax > 90th percentile of daily
Tmax of the base period TNQ10 Days with Tmin < 10th percentile
of daily Tmin of the base period FDTotal number of frost days (days
with absolute Tmin < 0 o C) HWD Heatwave duration index
(intervals (>5 days) with Tmax>5 o C above the daily Tmax
normal of the base period PQ9595th percentile of wet day amounts
PX5DMaximum 5-day precipitation total CDDMaximum number of
consecutive dry days (rainfall < 1mm) Slide 8 Evaluation of
Regional Climate Models ENSEMBLES-RT5 daily gridded observational
datasets Daily precipitation - temperature have been developed on
the basis of a European network of high quality station series
(http://eca.knmi.nl).http://eca.knmi.nl Period 1950-2006 (Haylock
et al., 2008). 1961-1990 period used Domain of study (12.5oW -
37.5oE and 30oN - 45oN). Only grids with 80% temporal coverage of
data for the reference period were utilised. Slide 9 Evaluation of
Regional Climate Models Grids with 80% temporal coverage of
temperature (a) and precipitation (b) data for the 1961-1990
reference period. A. B. Slide 10 Evaluation of Regional Climate
Models Differences of TXQ90 between CNRM (left), KNMI (right) and
gridded observations, for summer. Summer: CNRM, extreme
temperatures underestimated in E. Mediterranean (-6o C western
Balkans), whereas positive differences are observed
(overestimation) in the west of the Iberian Peninsula. In contrast,
KNMI model underestimates TXQ90 in the Iberian Peninsula while in
Italy and Greece simulates better the index (small differences with
the observational gridded data). Similar behaviour of the three
models in autumn (underestimation) Slide 11 Evaluation of Regional
Climate Models Seasonal differences of TNQ10 between C4I and
gridded observations TNQ10 is generally underestimated C4I: the
winter spring and autumn similar patterns, more pronounced in
winter. In contrast, during summer, the index is underestimated
(KNMI best simulations, while CNRM shows large negative biases.)
Slide 12 Evaluation of Regional Climate Models Winter differences
of FD between models and gridded observations. C4I and KNMI:
Similar patterns in all seasons underestimating the index
especially in winter and in high altitude regions Slide 13
Evaluation of Regional Climate Models Winter differences of PQ95
between models and gridded observations CNRM underestimates PQ95
more than the other two models All three models reveal wetter
conditions along the western coast of the Balkan Peninsula in
winter. Such wet conditions are also found in spring and autumn at
the mountainous western part of the Balkans. Slide 14 Evaluation of
Regional Climate Models Winter differences of PX5D between models
and gridded observations All models overestimate PX5D in large
parts of the study region C4I and KNMI present similar patterns in
winter spring and autumn with increased values for the index, in
several scattered sub-regions over the Iberian Peninsula and the
western coast of the Balkan Peninsula. CNRM shows drier conditions
compared to the other two models. Slide 15 Evaluation of Regional
Climate Models Summer differences of CDD between models and gridded
observations CDD is underestimated in the majority of the study
region (negative differences). Summer the KNMI model displays a
reverse behaviour with large positive differences (overestimation)
especially in the entire Italian and Balkan Peninsula, as well as
in the northern part of Turkey. Slide 16 Spatial analysis of
Climate extreme Indices Summer TXQ90 as estimated by KNMI for the
future periods 2021-2050 (left) and 2071-2100 (right) Mediterranean
will get warmer in the future, especially during 2071-2100. KNMI
expects the heat to get worse by the end of the 21st century for
summer, and estimates that the majority of the study region will
frequently experience temperatures of up or greater to 40 o C Slide
17 Spatial analysis of Climate extreme Indices Number of winter
frost days (FD) estimated by all models for the reference and
future periods. Decrease of the index most evident in central and
western Mediterranean. CNRM is the coldest model Slide 18 Spatial
analysis of Climate extreme Indices Winter PQ95 as estimated by all
models for the reference and future periods Future projection do
not show changes in the patterns: more intense in the last 30 years
of the century Slide 19 Spatial analysis of Climate extreme Indices
Consecutive dry days (CDD) in summer as estimated by KNMI for the
future periods 2021-2050 (left) and 2071-2100 (right). General
swift to drier conditions is predicted by all models under
consideration. KNMI presents the longest maximum dry spells.
According to this model, it seems that in the future, the southern
Iberian Peninsula and Greece will be characterised by a persisting
absence of rainfall, since the length of the dry spells approaches
90 days Slide 20 Maximum and minimum values of the Indices: TXQ90
all seasons The extremes occur in the lower part of each zone of
latitude The models reveal some sort of persistence regarding
regions where the extreme values of the indices are found! Slide 21
Maximum and minimum values of the Indices: TNQ10 (extreme minimum
values) winter and summer Agreement for all seasons for the present
and future periods and for all models. Three distinguished regions:
Morocco, southern France and eastern Turkey Slide 22 Maximum and
minimum values of the Indices: PQ95 and Px5d winter The extreme
values display relatively similar spatial distribution The extreme
index values are observed in the same areas for both reference and
future period. Slide 23 Maximum and minimum values of the Indices:
CDD Summer. In winter and spring spatial inconsistency among models
results Extreme summer values of the CDD are observed in the
easternmost parts of the study area, and in low latitudes of each
zone Slide 24 Trend analysis of Climate Indices Trends of TQX90 for
summer as estimated by the three examined models. Trends are
statistically significant at 0.05 level of significance Generally
positive trends of TXQ90 are seen in the Mediterranean The highest
positive trends are found inland, especially for summer,
particularly by CNRM and KNMI. In many cases the positive trends
exceed the 0.5 o C per decade. Slide 25 Trend analysis of Climate
Indices Trends of TQN10 for winter and spring for the three models.
Trends are statistically significant at 0.05 level of significance
Generally all models show positive trends, with the highest
increasing trends during summer (0.4-0.5 o C / decade). Maximum of
the positive trends in the eastern Mediterranean for all seasons
Slide 26 Trend analysis of Climate Indices Trends of PQ95 for
summer and autumn as estimated by KNMI for the 1951-2100 period.
Trends are statistically significant at 0.05 level of significance.
Less spatial coherence describes the PQ95 trend results, as both
positive and negative trends are observed. Negative summer trends
cover large part of the study area, indicating that extreme
precipitation tends to decrease during the warm part of the year by
the end of the 21st century. In contrast this model shows that
intense precipitation episodes should be more often expected in
autumn Slide 27 Trend analysis of Climate Indices Trends of PX5D as
estimated by KNMI for all seasons for the 1951-2100 period. Trends
are statistically significant at 0.05 level of significance. No
clear signal. Positive and negative trends are observed for all
models and all seasons. Winter KNMI Positive trends at the northern
parts of the Iberian, Italian and Balkan Peninsulas Slide 28
Conclusions All models underestimate extreme warm temperatures
(TXQ90). The extreme cold temperatures (TNQ10) seemed to be better
reproduced especially by KNMI. The coldest model was found to be
CNRM, particularly for low temperatures of the transitional
seasons. Also more frost days especially in the eastern part of the
Mediterranean. C4I simulates better the low temperatures for spring
and autumn than in the other two seasons and high skill for HWD
index. All models lower skill in simulating precipitation indices.
PQ95 was reproduced better than PX5D Models show drier conditions,
than those defined by the observational data. Despite the dry
characteristics of models, they underestimate the CDD index. Slide
29 Conclusions All models marked a shift towards warmer climates
High temperatures (TXQ90) getting warmer in the future. HWD
increase in future summers. Increase in summer low temperatures
(TNQ10). KNMI and C4I present reduced number of frost days (spring
and autumn) Precipitation indices the models present similar
present and future spatial patterns as regards extremes
precipitation amounts (PQ95) in winter the most extreme
precipitation observed along the western boarders of all the
peninsulas of northern Mediterranean. CNRM is found to be the drier
among models, KNMI predicts larger dry periods (CDD) in the future,
which seem to be more pronounced in the eastern part of the basin.
Slide 30 Conclusions Large positive trends for both extreme high
and low (TXQ90, TNQ10) temperatures Trends seem to be getting
larger in summer Precipitation trends no clear picture for the
future behaviour of precipitation extremes. All models showed some
increasing trends in winter extreme precipitation amounts (PQ95)
(northern areas of the domain) Positive trends winter consecutive
dry days (CDD). HWD large positive trends in the eastern
Mediterranean (CNRM) Frost days negative trends in all seasons.
Slide 31 Conclusions The models appeared sensitive to define
regions vulnerable to experience extremes in both present and
future periods. In most cases the three models marked the same
grids having the maximum (or minimum) values of indices. Models
generally agreed regions showing extremes in present are the most
vulnerable to experience extreme climate events in the future too.
Slide 32 Thank you! ENSEMBLES ACKNOWLEDGEMENTS: Work was funded by
the European Commission, as part of the ENSEMBLES Project (Contract
number GOCE-CT-2003-505539)