WP5.4 Evaluation of extreme events in observational and RCM
data Institute of Environmental Research & Sustainable
Development National Observatory of Athens, Greece Effie
Kostopoulou Christos Giannakopoulos Progress and Plans Progress
meeting for ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 2
Aim: Evaluate extremes in observational and RCM data for the
eastern Mediterranean Deliverable: D5.33: Scientific paper on the
ability of different RCMs to represent extremes in the eastern
Mediterranean (Month 60) Progress meeting for ENSEMBLES-WP5.4
members De Bilt, 16 May 2008
Slide 3
Progress to Date Comparison of an ENSEMBLES Regional Climate
Model with observed data in the Balkan Peninsula Presentation of
results during the annual EGU 2008 meeting: E. Kostopoulou, K.
Tolika, G. Tegoulias, C. Anagnostopoulou, P. Maheras and C.
Giannakopoulos: Regional climate model temperature simulations
compared with observed station data over the Balkan Peninsula
Collaborative study between National Observatory of Athens and
Aristotle University of Thessaloniki (RT4/RT5) Submission of
journal paper: E. Kostopoulou, K. Tolika, I. Tegoulias, C.
Giannakopoulos, S. Somot, C. Anagnostopoulou and P. Maheras:
Evaluation of a Regional Climate Model using in-situ temperature
observations over the Balkan Peninsula Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 4
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula RCM data at 25-km
horizontal resolution driven by the ERA-40 (ALADIN-Climate
Mto-France/ CNRM & common with the GCM ARPEGE-Climate ) Station
data (53 stations, 8 countries) Domain of study - Data Variables:
Tmax, Tmin Period: 1961-1990 Progress meeting for ENSEMBLES-WP5.4
members De Bilt, 16 May 2008
Slide 5
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Temporal evaluation
of RCM (T X ) Average mean (middle curves), absolute max (upper)
and min (lower) T X for each calendar-day during the period of
study, for the observed (blue) and modelled (red) temperatures.
Mean, min T X : biases mainly during the cold season: Warm higher
elevation Cold low altitudes Max T X : cold biases in high
altitudes throughout the year (e.g. Zenica) Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 6
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Temporal evaluation
of RCM (T N ) observed (blue) modelled (red) More pronounced
biases. Mean T N : better results for low-altitude stations. High
altitudes: overestimated max/min T N. E.g. the overestimation is
such that the mean model T N is as high as the highest station T N
(local effects ?) Low altitudes: underestimated max/min T N.
Progress meeting for ENSEMBLES-WP5.4 members De Bilt, 16 May
2008
Slide 7
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Spatial evaluation
of RCM (T X ) seasonal Differences red: +ve blue: -ve
North-to-south gradient with positive values in the north and
negative in the south of the study region. Winter diff. of up to +3
o C Warm bias Progress meeting for ENSEMBLES-WP5.4 members De Bilt,
16 May 2008
Slide 8
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Spatial evaluation
of RCM (T N ) - seasonal Differences red: +ve blue: -ve Winter
& autumn tendency of the model to underestimate T N (-ve diff)
Summer ve diff mainly in Romanian stations Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 9
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Assessment of RCM
performance in determining climate extremes Analysis of warm and
cold spells The model generally overestimates the occurrence both
warm and cold spells. Progress meeting for ENSEMBLES-WP5.4 members
De Bilt, 16 May 2008
Slide 10
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Assessment of RCM
performance in determining climate extremes Analysis of warm and
cold spells (% of coincidence) The percentage of coincidence in
both cold and warm spells is relatively high for the transitional
seasons. Winter: better for warm spells Summer: better for cold
spells particularly for island stations located around the Aegean
Sea.
Slide 11
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Assessment of RCM
performance in determining climate extremes Analysis of
percentile-based indices (Tx90) TRENDS Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 12
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Assessment of RCM
performance in determining climate extremes Analysis of
percentile-based indices (Tn10) - TRENDS Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 13
Regional Climate Model Temperature Simulations Compared with
Observed Station Data over the Balkan Peninsula Conclusions The
model accurately described the seasonal cycle and simulated the
spatial distribution of T X and T N. Altogether, the model
performed better for T X than T N and better for the transitional
seasons of the year. The model performed better at low-altitude
stations along the coasts, highlighting the constraints of the
topographic forcing in the simulations. Assessing the performance
of the model to determine extremes, the model results did not seem
to be very sensitive in detecting particular events of warm/cold
spells. This could be attributed to the poor behaviour in terms of
time chronology, as it is expected for the RCM to lose their time
chronology in a region that lies so far from the models western
boundary. In contrast the model exhibited a remarkable ability to
reproduce the seasonal trends of Tx90 and Tn10. Progress meeting
for ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 14
Work in Progress NOA will perform an evaluation of ENSEMBLES
RCM data against observational datasets from Greek stations and
against gridded observational data available from WP5.1 for the
eastern Mediterranean. Comparison of these datasets will be made
using selected indices of extremes for temperature and
precipitation. The main focus will be to identify strengths and
weaknesses in RCM data in their ability to represent extremes. Some
examples of indices of extremes that can be used for the case study
region of the Mediterranean are shown in the following slide:
Progress meeting for ENSEMBLES-WP5.4 members De Bilt, 16 May
2008
Slide 15
Extremes Indices Temperature Tmax 90th percentile Tmin 10th
percentile Heat Wave Duration Number of summer days per year (T X
> 25 o C) Number of tropical nights per year (T N > 20 o C)
Number of frost days (T N < 0 o C) Number of very cold nights
per year (T N < -5 o C) Precipitation 95th percentile of rainday
amounts Greatest 5-day total rainfall Total No. of days per year
with precip >= 20mm Maximum length of dry spell in a year (days)
Maximum length of wet spell in a year (days)
Slide 16
Work in Progress Three ENSEMBLES Regional Climate Models are
compared to the ENSEMBLES gridded observational dataset Seven
climate extreme indices are obtained from a collaborative study
with RT4 (Aristotle University of Thessaloniki) TxQ90 TnQ10 hwd fd
pQ95 px5d cdd InstituteScenarioDriving GCMModel C4IA2ECHAM5RCA3
CNRMA1BARPEGEAladin KNMIA1BECHAM5RACMO Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008
Slide 17
Differences btwn model gridded observed TxQ90 autumn summer
spring winter C4ICNRMKNMI
Slide 18
Differences btwn model gridded observed TnQ10 autumn summer
spring winter C4ICNRMKNMI
Slide 19
Differences btwn model gridded observed pq95 autumn summer
spring winter C4ICNRMKNMI
Slide 20
Differences btwn model gridded observed px5d autumn summer
spring winter C4ICNRMKNMI
Slide 21
Work Planned NOA will perform a comprehensive analysis of
several indices that represent key aspects of climate extremes.
Using these indices, grid cells that present differences in RCM
data with observations will be identified. Physical explanations
for these discrepancies will be researched based on specific
topographical features of each location that is probably not
possible to be represented in RCMs. Several RCMs will be compared
with the gridded and station data and the ability of each RCM to
represent specific extremes will be rated. A reliability
(weighting) factor for each model is planned to be quantified
specifically for the eastern Mediterranean region. These weights
will be assigned to the different ENSEMBLES RCMs (instead of
treating separate simulations as "equally likely") and so their
results can be combined into a scenario to assess future climate
changes in the eastern Mediterranean. Progress meeting for
ENSEMBLES-WP5.4 members De Bilt, 16 May 2008