The document analyzes the relationship between the North Atlantic Oscillation (NAO) index and temperature extremes in Europe based on observed data and climate model simulations. It finds that the NAO index explains a significant amount of variability in winter temperature extremes, particularly cold extremes, across large parts of Europe. Differences between high and low NAO index days exceed 5°C for some cold extremes in northern Europe. While climate models generally reproduce this relationship, they perform better for mean temperatures than extremes. The study also finds similar NAO impacts on temperatures in future climate simulations.

1. Links between the NAO Index and
temperatures in Europe in climate models
Romana Beranová, Jan Kyselý
Technical University of Liberec
Institute of Atmospheric Physics, Academy of Sciences of the Czech Republic

2. Introduction
 Motivation: Is NAO index a useful covariate for a
nonstationary statistical model?
Outline:
 Data
 The North Atlantic Oscillation index (NAOI)
 Relationship between NAOI and seasonal temperature and
temperature extreme indices
 Conclusions

3. Observed / Model data
• NCEP/NCAR reanalysis, which represents the observed
data
• Ensemble of 10 Global Climate Models (GCM) with 14 runs
• Simulations for 1961-2000, performed with forcing agents as
specified by the IPCC protocol for the 20C3M experiment
• 2071-2100 - sresA1B scenario
centre/model resolution [°] runs source
MPI / ECHAM5-MPI 1.875 x 1.875 3 http://www.mad.zmaw.de
DMI / ECHAM5-MPI 1.875 x 1.875 1 http://www.mad.zmaw.de
GFDL / CM2.1 2.5 x 2.0 1 http://nomads.gfdl.noaa.gov
CCCMA / CGCM3 (T63) 2.8 x 2.8 1 https://esg.llnl.gov:8443
BCCR / BCM2.0 2.8 x 2.8 1 http://www.mad.zmaw.de
CNRM / CM3 2.8 x 2.8 1 https://esg.llnl.gov:8443
IPSL / CM4 V1 2.5 x 3.75 1 https://esg.llnl.gov:8443
UKMO / HadCM3 2.75 x 3.75 1 http://cera-www.dkrz.de
UKMO / HadGEM2AO 1.25 x 1.875 1 http://cera-www.dkrz.de
FUB / EGMAM 3.75 x 3.75 3 http://cera-www.dkrz.de

4. Data
• Temperature data: daily temperature maxima (Tx) and
minima (Tn)
• Temperature extreme indices:
• cold extremes – Q10 and Q1 of Tn
• warm extremes – Q90 and Q99 of Tx
• Data cover European Area (15°W-40°E, 30-70°N)
• Winters (December- January- February) for recent climate
1961-2000 and future time slice 2071-2100
• Pressure data: daily mean sea level pressure (SLP)

5. The North Atlantic Oscillation
The NAO is the dominant mode of winter climate variability in the
North Atlantic region. The NAO is a large scale seesaw in
atmospheric mass between the subtropical high and the polar low
Positive NAO Index:
• A stronger than usual subtropical high pressure centre and a deeper
than normal Icelandic low
• The increased pressure difference results in more and stronger
winter storms crossing the Atlantic Ocean on a more northerly track
• This results in relatively warm and wet winters in northern and
western Europe
Negative NAO Index:
• A weak subtropical high and a weak Icelandic low
• The reduced pressure gradient results in fewer and weaker winter
storms
• They bring moist air into the Mediterranean and cold air to northern
Europe

6. The North Atlantic Oscillation in GCMs
Observed
NAO as the 1st leading empirical orthogonal function for the
winter SLP field

7. Definition of NAO index
• Daily NAO Index is based on the difference of the
normalised SLP between Ponta Delgada (37.7°N a 25.7°W,
Azores) and Stykkisholmur (65.1°N, 22.7°W, Iceland)
• Daily mean SLP values
for the two stations were
interpolated from the four
nearest grid points for
every GCM

8. Definition of NAO index
• NAO index was classified as negative (-) when below the 25%
quantile and positive (+) when above the 75% quantile of its
distribution over 1961-2000

9. Methods
We focus on identification of areas in which the NAO index is
linked to temperature extremes (on the daily time scale)
• We identified days with positive NAOI (NAO+ days) and
negative NAOI (NAO- days)
• We calculated mean, Q90, Q99 of Tx for NAO+ and
NAO- days; and mean, Q10, Q1 of Tn for NAO+ days and
NAO- days
• We displayed differences in these variables between the
NAO+ and NAO- days

10. Results – mean temperature
Mean of Tn
Observed
Fig.: Differences between mean of Tn for days with NAO+ index and NAO-
index

11. Results – mean temperature
Mean of Tx
Observed
Fig.: Differences between mean of Tx for days with NAO+ index and NAO-
index

12. Results – mean temperature
• NAO index influences Tx and Tn in large parts Europe
• Positive difference in the area of 45°N-75°N
• Negative difference in north Africa, in south of the Iberian
peninsula and in Turkey
• The majority of the GCM simulations depict almost the same
pattern as reanalysis
• In nearly all GCMs the area of positive differences is reduced
• Some GCMs (CM21, BCM2, CM3, CM4 and EGMAM) show
an extended area with negative differences
• Intra-model variability for EGMAM runs is larger than for
ECHAM runs

13. Results – cold extremes
Q10 of Tn
Observed
Fig.: Differences between Q10 of Tn for days with NAO+ index and NAO-
index

14. Results – cold extremes
Q1 of Tn
Observed
Fig.: Differences between Q1 of Tn for days with NAO+ index and NAO-
index

15. Results – cold extremes
Q10:
• For all GCMs the area with positive differences is
extended to the south and the differences of Q10 between
NAO+ and NAO- days are larger in comparison with mean
Tn
• In some models and some areas (e.g. HadGEM2AO in
Denmark) the difference for Q10 is >8°C
• The inter-model variability is larger for Q1 than Q10
• The ability of models to simulate relationships between
NAOI and temperature extremes is worse for Q1 than
Q10
• Some grids with negative differences (ECHAM5 run1,
BCM2, HadCM2) appear inside the north area with
positive differences of Q1

16. Results – warm extremes
Q90 of Tx
Observed
Fig.: Differences between Q90 of Tx for days with NAO+ index and NAO-
index

17. Results – warm extremes
Q99 of Tx
Observed
Fig.: Differences between Q99 of Tx for days with NAO+ index and NAO-
index

18. Results – warm extremes
• NAO affects smaller area of Europe for warm extremes than
for cold extremes, and this is reproduced by most GCMs
• For all GCMs the area with positive differences is smaller and
the area with negative differences tends to be larger than for
mean Tx
• Spatial patterns of projected differences in Q90 are much
more coherent compared to those of Q99

19. Results – future climate (2071-2100)
Mean of Tn
• Majority of GCMs presents similar relationships as for recent
climate (exception – model CM3)

20. Results – future climate (2071-2100)
Q1 of Tn
• For majority of GCMs the area with positive differences is
extended in comparison with recent climate

21. Results – future climate
Q99 of Tx
• Similar results as for recent climate
• Some models simulate larger area with negative differences
(CM2.1 and BCM2)

22. Conclusions
1. The NAO index represents a useful covariate that explains
an important fraction of variability of temperature extremes
in winter in large parts of Europe. It is more important for
cold than warm temperature extremes
2. Differences between NAO+ and NAO- days in Q1 exceed
5°C in large parts of northern, western and central Europe
3. The GCMs reproduce the observed spatial patterns
reasonably well, although better for means than extremes
4. The inter-model and intra-model variability is larger towards
tails of temperature distributions
5. Although the NAO index based on two fixed points is not
able to detect spatial structure of NAO, we show that the
index can be used in extreme value models as variable
which affects temperature extremes

23. Future work
Central European Zonal Index (CEZI) as a covariate for summer
temperature extremes
CEZI is defined as the difference Fig.: Differences between
of the normalised SLP between Q90 of Tx for days with
Northern Europe (60-65°N a
CEZI+ and CEZI-,
0-20°E) and Southern Europe
(35-40°N, 0-20°E)
1961-1990, JJA