PDF FILE

Quarterly Journal of the Royal Meteorological Society
Q. J. R. Meteorol. Soc. 136: 305–318, January 2010 Part B
Factors governing the interannual variation and the long-term
trend of the 850 hPa temperature over Israel
H. Saaroni,a * B. Ziv,b I. Osetinskyc and P. Alpertd
a
Dept. of Geography and the Human Environment, Tel Aviv University, Israel
b Dept. of Natural Sciences, The Open University of Israel
c Israel Meteorological Service, Bet Dagan, Israel
d Dept. of Geophysics and Planetary Sciences, Tel Aviv University, Israel
*Correspondence to: H. Saaroni, Department of Geography and the Human Environment, Tel Aviv University, Tel Aviv,
Israel. E-mail: [email protected]
This study examines the ability of the interannual variability in the occurrence of
synoptic types, intensity of large-scale circulations and global temperature to explain
that of the 850 hPa temperature in Israel for the summer and the winter. The synoptic
factor was represented by 19 types defined by Alpert et al. (2004b). For the summer,
the deep and the weak Persian Trough explained 35% of the interannual temperature
variance. For the winter, the lows to the east and to the north explained 44% of
the interannual temperature variance. Two additional factors were incorporated:
large-scale circulations, the North Atlantic Oscillation for the summer and the Arctic
Oscillation for the winter; and global radiative forcing, represented by the global
temperature. Both of them were found to be significant, and the variance explained
by all of them is 56% for the summer and 64% for the winter. In the summer
the variation is dominated by warm and cool types whereas in the winter the cold
systems dominate.
The individual contribution of each factor to the long-term temperature trend
was estimated. While the global radiative forcing contribution was positive and
large in both seasons, the synoptic contribution was positive, four times larger in
the summer. The large-scale contribution was negative, three times larger in the
winter. The considerable warming in the summer results from a rapid increase in the
occurrence of the weak Persian Trough, which is a warm type. The study approach
may be useful for predicting future temperature regimes, based on predicted synoptic
c 2010 Royal Meteorological Society
features in climatic models. Copyright Key Words: temperature interannual variability; synoptic classification; eastern Mediterranean; Persian
Trough; Red Sea Trough; Cyprus Low; Subtropical High
Received 6 March 2008; Revised 3 November 2009; Accepted 8 December 2009; Published online in Wiley
InterScience 9 February 2010
Citation: Saaroni H, Ziv B, Osetinsky I, Alpert P. 2010. Factors governing the interannual variation
and the long-term trend of the 850 hPa temperature over Israel. Q. J. R. Meteorol. Soc. 136: 305–318.
DOI:10.1002/qj.580
1.
Introduction
those in the large-scale circulations (e.g. Xoplaki et al.,
2003a). A common approach for examining the relationships
Interannual temperature variations are closely related to between local conditions and larger-scale systems is to
variations in the regional synoptic-scale systems beyond correlate interannual variations of monthly or seasonal
c 2010 Royal Meteorological Society
Copyright 306
H. Saaroni et al.
anomalies of atmospheric fields, such as sea-level pressure
(SLP), with the desired local variable. A more sophisticated
approach, though similar in essence, uses interannual
variations of the monthly or seasonally-averaged indices
that represent large-scale circulations.
Statistical downscaling is one of the approaches developed
and used to regionalize weather variables, especially
temperature and precipitation, as reviewed by Giorgi
et al. (2001). Most studies use upper-level parameters and
circulation modes. For example, Rauthe and Paeth (2004)
quantified the relative importance of Northern Hemisphere
circulation modes with respect to twentieth-century nearsurface temperature and precipitation, using stepwise multiregression. Spak et al. (2007) pointed out the importance
of downscaling projections from general-circulation model
(GCM) simulations for assessing local and regional impacts
of climate change. They employed a multilinear regression
model to downscale June, July and August monthly
mean surface temperatures over eastern North America
under greenhouse-gas-driven climate change simulation.
The weakness in the use of monthly or seasonal averages
is their inability to capture individual synoptic events on
the daily time-scale. Huth (2002) intercompared statistical
downscaling methods and potential large-scale predictors
for winter daily mean temperatures in central and western
Europe using various methods, including multilinear
regression. He used four potential predictors, two circulation
variables: SLP and 500 hPa geopotential heights (GPH)
and two temperature variables: 850 hPa temperature and
1000–500 hPa thickness. The best choice of predictor
appeared to be a pair composed of one circulation and
one temperature predictor.
The linkage between atmospheric circulations and the
temperature regime over the Mediterranean Basin (MB) has
been addressed mainly by analysing monthly (or seasonal)
averages of pressure and wind fields. For instance, Ben-Gai
et al. (1999) analysed the correlation between the North
Atlantic Oscillation (NAO) index and the variations in the
winter surface temperatures in Israel. Kutiel and Benaroch
(2002) defined the North Sea–Caspian Pattern (NCP),
which is based on the 500 hPa GPH difference between these
two regions. They found the NCP capable of differentiating
between below- and above-normal temperatures over the
Eastern Mediterranean (EM), (Kutiel et al., 2002).
Otterman et al. (2002) found that the strengthening of
warm air transport, via south-westerlies, from the eastern
North Atlantic to Europe between January and March for
the years 1948–1995 accounts for a large part of the observed
warming there. They found also that for 1996–2001
this temperature trend was broken, and attributed their
findings to the long-term variations in the NAO. Ziv
et al. (2005) attempted to identify Long-Term Trends
(LTTs) in dynamic variables, including lower-level advection
and vertical velocity, which may explain the considerable
warming trend in the summer over the MB. However, they
did not find any significant signal, and therefore suggested
that the MB responds directly to the global radiative forcing
as the source of the regional warming in the summer season.
While Ziv et al. (2005) checked trends in the immediate
proximity of the eastern Mediterranean and could not find
the dynamic cause for the warming, Xoplaki et al. (2003a,
2003b) showed that large-scale circulations can explain it.
Xoplaki et al. (2003a) found that three large-scale predictor
fields (300 hPa geopotential height, 700–1000 hPa thickness
c 2010 Royal Meteorological Society
Copyright and SSTs) account for more than 50% of the total summer
temperature variability over the Mediterranean Basin. They
found that the warming over the MB and Europe can be
explained, at least in part, by an increasing trend in the
occurrence of blocking highs over western Europe.
The synoptic systems prevailing over the EM, and their
associated weather conditions, are extensively reviewed in
various publications and are summarized in HMSO (1962),
Ziv and Yair (1994) and Goldreich (2003). Ziv and Yair
(1994) also describe the causal relationships between the
structure, location and orientation of the synoptic systems
and the weather conditions in Israel.
The present study aims to analyse the factors that determine the variations of the seasonal average temperatures
(SAT) over Israel. The analysis differentiates among the
synoptic-scale systems, the large-scale circulations and the
global radiative forcing for both the interannual variations and the long-term trend of winter and summer
temperatures. The main product is prediction equations
that calculate the SAT from the occurrence of the regional
synoptic types, the amplitude of the dominant large-scale
circulation and the global temperature, as a proxy for the
global radiative forcing. These equations serve as a quantitative measure for evaluating the climatic contribution of
each factor, and are proposed as a tool for deriving future
temperatures from the output of climate models.
2.
2.1.
Materials and methods
Materials
The analysis is performed for the midsummer, July and
August (JA), and winter, December–March (DJFM), as
defined by Alpert et al. (2004a), for the period 1948–2008.
The transitional seasons, which are affected by both the
winter and the summer synoptic types and are characterized
by large day-to-day and year-to-year variations, are not
included.
The temperatures used are the 850 hPa daily temperatures
at the grid point 32.5◦ N, 35◦ E (Figure 1) located at the
centre of the study area, taken from the National Centers for
Environmental Prediction/National Center for Atmospheric
Research (NCEP/NCAR) CDAS-1 archive (Kalnay et al.,
1996; Kistler et al., 2001). The values related to a grid point
are averaged over a 2.5◦ × 2.5◦ area, which is comparable
to the area of Israel. Moreover, it was found to be well
correlated (by >0.9) with the averaged temperature over the
larger square of 7.5◦ × 7.5◦ (3 × 3 grid box).
Following Saaroni et al. (2003) and Ziv et al. (2005),
we chose the 850 hPa level as a reference level for the
temperature. The surface temperatures are noisy since they
reflect the sea–land discontinuity, topographic influences
and urban effects. These are considerably smoother at the
850 hPa level. The 1000 hPa level is located below the surface
in the majority of the study region and the 925 hPa level
intersects the marine inversion, which prevails in the study
region for most of the year (Dayan et al., 2002) and is
highly variable both in space (between the coastline and
the inner regions) and time (from day to day). The above
considerations imply that the use of surface temperature may
reduce the representativeness of the results. However, recent
analysis has shown good correspondence (the correlation
coefficient, R, is in the order of +0.7) between the 850 hPa
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
Factors Governing Temperature Variation and Trend Over Israel
307
types that differ only in temperature, except for the ‘low
to the south’ (‘deep’ and ‘shallow’) which is identical to
the ‘Sharav low to the west’, that is similar in structure
and location but differs in temperature. Since the total
occurrence of these types is only 1.4%, their practical effect
can be ignored.
In order to verify the extent to which the surface systems
represent the 850 hPa level, we derived a correlation map
between the daily SLP and the 850 hPa GPH (not shown).
The correlations found within the EM exceeded 0.8, so that
the synoptic types seen in the SLP fields can be considered
as representing the flow patterns in the 850 hPa level. It is
worth noting that the synoptic types as defined by Alpert
et al. (2004b) and used here have several flaws. First, the
categorization procedure was designed for the synopticscale systems according to the meteorological re-analysis
data with no respect for their local mesoscale environmental
implications. The second drawback is the limited size of
a manually classified set of synoptic situations, such as the
coexistence of two different systems (e.g. two cyclones). This
is a common weakness for all classifications.
Figure 1. The study region and the grid point (32.5◦ N 35◦ E) selected to
represent Israel. The shaded rectangle denotes the respective grid box.
temperatures and surface temperatures and heat-stress in
Israel.
The synoptic types are based on the semi-objective
synoptic classification (Alpert et al., 2004b), containing
five systems that are divided into 19 types. This classification
is the only one used operationally for climate research in
the EM. For instance, it has been used for season definition
for the study region (Alpert et al., 2004a) and for analysing
the calendaricity of cyclonic activity (Osetinsky and Alpert,
2006). The rationale for the development of this classification
was formulated by Alpert et al. (2004b), and states that
the large-scale European patterns classification cannot be
used for the EM weather, because the latter is governed
by much smaller synoptic systems, originating from both
midlatitudinal and tropical areas and modified over the
complex EM terrain. The classification is based on the
1200 UTC 1000 hPa temperature and horizontal wind and
GPH, reflecting the surface pressure systems. According
to Alpert et al. (2004b), the regional weather phenomena
are well defined by these synoptic systems and the upperlevel systems are clearly reflected when one considers the
dynamics of the EM surface synoptic systems. The daily types
by themselves are not expected to explain the SAT. However,
since each type is characterized by typical temperature (see
section 3), their occurrence, which varies considerably from
one season to another, can explain it. A brief description
of the synoptic classification, together with a representative
map for each of the five systems is given at the appendix.
It should be noted that the weather-typing scheme
includes thermal variables. Therefore, it could happen that
types which are similar in structure or dynamics but differ in
temperature may tend to be identified more as warm types
due to anthropogenic warming, although their dynamical
properties remain unchanged. In this way the greenhousegas forcing may be inadvertently included in the changes in
the frequency of occurrence, and so mask physical predictors
necessary to explain the temperature trend. Fortunately, the
synoptic typing used in this study does not contain identical
c 2010 Royal Meteorological Society
Copyright 2.2.
Method
The analysis is divided into three parts. The first deals with
the impact of the occurrence of the individual synoptic
types on the SAT. For each season, the types with average
occurrences of <0.4% are excluded. Accordingly, four types
for the summer and 13 for the winter were included. The
average temperature of the days belonging to each type is
calculated. Then, the interannual variation in the occurrence
of each individual type is correlated with that of the SAT.
For the upper and lower tenths of the studied seasons (i.e.
six hottest/coolest summers and winters), the degree of
synoptic extremity is analysed. This is done by calculating
the deviation in the occurrence of the synoptic types from
the average. The sensitivity of the results concerning extreme
years was addressed by repeating the analysis for the eight
coolest and warmest seasons.
In the second part, prediction equations are derived by
a multi-regression scheme based on the occurrence of the
synoptic types, the amplitude of the regional large-scale
circulations and the global average temperature:
• Synoptic types – The size of the sample (60 seasons)
necessitates a limited number of potential predictors.
This was done by grouping several individual synoptic
types that reflect similar properties, such as being both
‘cold’ and negatively correlated with the SAT and vice
versa. Another grouping, applied for the summer, was
the difference between the occurrence of the cool and
hot types.
• Large-scale circulations – Two regional circulations
were found to be correlated with the SAT over the
study region. These are the NAO and the Arctic Oscillation (AO). For the summer, the most correlated is
the NAO (averaged over the pertinent months) and
for the winter, the AO, which is highly correlated
with the NAO (Thompson and Wallace, 1998). The
NAO data (Jones et al., 1997; Vinther et al., 2003)
is taken from the Climatic Research Unit (CRU),
University of East Anglia, Norwich, United Kingdom (http://www.cru.uea.ac.uk/cru/data/nao.htm)
and the AO data is taken from the
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
308
H. Saaroni et al.
Figure 2. Average 850 hPa temperature (light grey columns), and the correlation coefficient (R) with the occurrence (dark grey columns), of the six
summer synoptic types. The long-term seasonal average temperature is denoted by the solid horizontal line. The standard deviation of the temperature
is presented in Table I.
National Oceanic and Atmospheric Administration’s
NOAA–CIRES,
Climate
Indices:
Monthly Atmospheric and Ocean Time Series
(http://www.cdc.noaa.gov/data/correlation/ao.data)
• Global temperature anomaly (with respect to the
1951–1980 period) – This variable is regarded
here as an estimate for the global radiative forcing,
and its average for the pertinent months is added
to the prediction scheme for each of the studied
seasons. The data are taken from NOAA–CIRES,
Climate Indices: Global Average Surface Air
Temperature Anomalies: GHCN 1880–12/2008
(http://data.giss.nasa.gov/gistemp/tabledata/GLB.
Ts.txt). These data are based on surface air temperature measurements (stations only) which undergo
quality control including elimination of the urban
heat island (Hansen et al., 1999, 2001).
The multi-regression procedure was the ‘stepwise’ option
of the Statistical Package for the Social Sciences (SPSS)
software, which retains only the potential predictors
significant at the 0.05 level, regarded hereafter as ‘predictors’.
In order to assess the validity of the predictors found in
the multi-regression scheme, we performed cross-validation
using the ‘holdout validation’ technique. Observations
are chosen randomly from the initial sample to form
the validation data, while the remaining observations are
retained as the training data. In each experiment, 40 of the
61 seasons (for the winter, 60) were chosen randomly and
were used as a basis for a multi-regression scheme. In each
experiment, we obtained a different prediction equation
according to the sub-sample used. We then applied it to
the remaining seasons (independent sample) and correlated
the calculated and the observed temperatures. The two parts
are carried out separately for the summer and the winter
(sections 3.1 and 3.2, respectively).
The third part (section 3.3) deals with the contribution
of each of the above factors, i.e. the synoptic, the largescale and the global radiative forcing, to the observed LTT.
This was achieved by multiplying the trend in each of the
predictors, representing them by the respective coefficient
in the prediction equations. This was done separately for the
summer and the winter.
c 2010 Royal Meteorological Society
Copyright 3.
3.1.
Results
Summer
The summer is represented here by the ‘high season’,
July–August (JA), the period in which the Persian Trough
(PT, Figure A2) is most prominent (Alpert et al., 1990; Bitan
and Saaroni, 1992; Alpert et al., 2004a). Ziv et al. (2004)
found that the factor dominating the interdiurnal variations
of the summer temperature is the magnitude of the lowerlevel temperature advection by the Etesian winds, i.e. permanent westerly winds associated with the PT. Eighty-five
percent of the midsummer days belong to one of the three
types defined as PT. (For the distinctions between the types
belonging to the PT system, see appendix). Figure 2 presents
the average temperature for each of the six summer synoptic
types, which cover 99.8% of the days, together with the correlation between the seasonal occurrence of each one and
the SAT. The six types differ in their average temperatures by
up to 5.6 ◦ C. The warmest type is the Red Sea Trough (RST,
Figure A1)–eastern axis and the coolest is the strong (deep)
PT. The most frequent type is the medium PT, whose average
temperature is very close to the seasonal average, 21.7 ◦ C.
Figure 2 and Table I (4th column) show the correlation
between the seasonal occurrences of each synoptic type
and the SAT. The most negatively correlated types are the
strong PT (R = −0.53) and the shallow low north of Cyprus
(R = −0.41), in agreement with their being the coolest types.
These two types are infrequent (2.5% for the strong PT and
0.4% for the shallow low north of Cyprus) and occur only
in part of the summer seasons. But, when they occur, they
denote an exceptionally cool season (e.g. summer 1966).
The role of these two types as indicators for a cool season
is also expressed in the high deviation in their occurrence
from the average in the six coolest seasons, at +1.9 standard
deviations (STDs, Table I, right panel).
The most positively correlated types are the weak PT and
the RST–eastern axis; both imply weakening of the Etesian
winds and higher temperatures. The RST–eastern axis does
not occur in the majority of summers, but seasons in which
this type is observed are extremely hot (deviation of +0.81
STD in the six hottest summers, Table I).
One would expect that the temperature deviation from
normal for each type will be proportional to the correlation
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
c 2010 Royal Meteorological Society
Copyright 2.4
0.4
14.5
44.5
37.6
0.5
Average monthly
occurrence (%)
Correlation (R)
with 850 hPa SAT
−0.53
−0.41
−0.30
−0.20
+0.54
+0.41
850 hPa temp (◦ C)
deviation from normal and STD
−3.3 ± 2.4
−1.7 ± 2.3
+0.2 ± 2.5
−0.3 ± 2.5
+0.5 ± 2.5
+2.3 ± 2.9
−0.57
−0.29
−0.08
−0.61
+0.69
+0.81
6 hottest summers
(+1.9◦ C above normal)
+1.90
+1.87
+1.22
−0.35
−1.25
−0.41
6 coolest summers
(−1.8◦ C below normal)
Deviation in the occurrence
from the average (STD units)
Average monthly occurrences (%), 850 hPa temperature deviation from the normal (seasonal long-term average temperature) and STD, correlation with the SAT (those significant at the 0.05 level are in bold)
and the deviation in the occurrence from the average (in STD units) for the upper and lower percentile, i.e. the six coolest and the six hottest seasons.
Strong PT
Shallow low north of Cyprus
High west of Israel
Medium PT
Weak PT
RST–eastern axis
Synoptic types for July–August
Table I. The midsummer (JA) synoptic types.
Factors Governing Temperature Variation and Trend Over Israel
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
309
310
H. Saaroni et al.
between its occurrence and the SAT. This holds for the
RST–eastern axis, the strong PT and the shallow lows north
of Cyprus. An exception is the high west of Israel, which
has an average temperature anomaly of only +0.2◦ C but is
negatively correlated with the SAT (R = −0.30, significant
at the 0.05 level). This can be explained by a tendency of this
type to be correlated with types that are associated with low
temperatures. Indeed, this type is significantly correlated
with the two coolest types, the shallow low north of Cyprus
and the strong PT (+0.52 and +0.51, respectively, Figure 3).
Table I also shows the standardized anomalies (the
anomaly of each type divided by its respective STD) in
the occurrences of the summer synoptic types for the
upper and lower tenths. The occurrence of the medium
PT deviates negatively from the normal in both the coolest
and hottest summers. It indicates that this type represents
normal conditions, so when it is less frequent it ‘gives
way’ to other types that produce anomalous conditions. As
for the remaining types, for four of them the deviations
in the coolest summers are significantly larger than those
for the hottest seasons, except for the RST with eastern
axis, whose deviation for the hottest summers is twice as
large as that for the coolest ones. This can be attributed
to the infrequent occurrence of this type (0.3 days/season),
which, when appear causes severe warm events. For similar
reasons (but with a reversed sign) the deviation in the
occurrence of the shallow low north of Cyprus in the coolest
seasons is over six times larger than that in the hottest
seasons.
In order to examine the sensitivity of the results to
the number of extreme years, we repeated the analysis
for the eight warmest and coolest summers. The ratio
between the negative and positive deviations in the
occurrence of the relevant synoptic types remained similar.
As expected, the amplitude of the deviations was reduced by
10–20%.
The multi-regression model for the SAT was applied
to five potential predictors, three compound predictors
based on the six synoptic types, the NAO anomaly and
the global temperature anomaly. Each of the synoptic
combined predictors is the difference between two synoptic
types that were found negatively correlated with each other
(Figure 3). These are the weak and the strong PT, the weak
and the medium PT and the weak PT and the high to
the west. The multi-regression was first operated using
only the three synoptic compound potential predictors
and yielded only one significant predictor, the difference
between the occurrence of the weak and the strong PT with
a correlation of +0.59 between the calculated and observed
SATs. The multi-regression, using the entire set of potential
predictors described above, yielded the following prediction
equation:
TJA = 20.19 + 0.0567 × fwPT−sPT − 0.662 × NAO
+ 1.199 × TglobJA ,
(1)
where fwPT−sPT denotes the difference between the
occurrence (in days per season) of the weak and the
strong PT and TglobJA is the global average temperature
anomaly for JA. The correlation between the modelled and
the observed SAT for the full sample was 0.76, when all
potential predictors were included, and 0.75 when only
the three significant predictors included in Eq. (1) were
used. The correlation implies that the seasonal distribution
c 2010 Royal Meteorological Society
Copyright of the surface pressure systems, together with the NAO
anomaly and the global temperature anomaly, explain over
50% of the interseasonal variance of the high summer
temperature.
The next step was the holdout validation for the multiregression (see section 2, above). The results of the three
experiments, each carried out for 40 randomly selected
years, are shown in Table II. In two of the experiments, the
significant predictors were the same as for the full sample,
and in one, only two of the three predictors appear, but in
the same order. The validation confirms the validity of the
predictors found in the multi-regression done for the full
sample.
The ability of the prediction equation (1) to reconstruct
extreme seasons can be inferred from Figure 4, which
presents the observed and the modelled SAT for the study
period. The standard error of the prediction was 1.1 ◦ C.
The deviation of the average temperature for the upper
and lower tenths, i.e. the six hottest and the six coolest
seasons, is +1.9◦ C and −1.8◦ C, respectively (over 1.5 larger
than the interannual STD, which is 1.1 ◦ C). The bias of
the modelled temperature was −0.93◦ C for the upper and
+0.57◦ C for the lower tenth, reflecting underestimation of
the prediction equation for both hottest and coolest seasons.
Underestimation of extremes is a well-known drawback of
regression-based methods (von Storch, 1999).
3.2.
Winter
The temperature in the winter season is determined by the
advection related to a large number of different synoptic
systems (Goldreich, 2003). Several studies, based on casestudies, point to synoptic systems with which daily abnormal
temperatures are associated. Such are continental polar
outbreaks (Saaroni et al., 1996), rain events accompanied by
low temperatures associated with lows located east of Cyprus
(Kahana et al., 2002) and warm conditions associated with
the Red Sea Trough (Dayan et al., 2001) and with highs east
of Israel (Saaroni et al., 1998).
Figure 5 shows the average temperature for 13 types with
frequencies of ≥0.4% (hereafter ‘winter types’) together
with the correlation between their occurrence and the SAT
(see Table III for more details). The two coldest types are
the deep low to the east (average temperature of +1.6◦ C,
compared to the seasonal average of 5.8 ◦ C) and the shallow
low to the east (+3.4◦ C). The two warmest systems are
the RST–western axis (+11.3◦ C) and the Sharav Low west
of Israel (+10.0◦ C). The characteristic temperatures of the
extreme winter types deviate from the normal by ∼5◦ C,
over twice the deviation of the summer types. This is in line
with the interdiurnal temperature STD, which in the winter
is over twice that of the summer (Ziv et al., 2004).
The correlation between the occurrence of the winter types
and the SAT agrees with the temperature deviation, i.e. types
having positive correlation are warmer than average and
vice versa, except for two types (Figure 5). The prominent
example is the Sharav Low west of Israel, which is the second
warmest type, but is not correlated with the SAT. This can
be explained by its being positively correlated (+0.41) with
the shallow low north of Cyprus, which is characterized by
normal temperatures, and by its being negatively correlated
(−0.26) with the high east of Israel, the second warmest
type.
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
Factors Governing Temperature Variation and Trend Over Israel
311
Figure 3. Average seasonal occurrence of each synoptic type and the correlation between each pair of types for the summer season (when the absolute
value of R exceeds 0.4).
Table II. Summary of the multi-regression model for July–August (bold), together with the three holdout validation experiments.
Experiment
Multi-regression Full sample
Predictor no.1
Weak-strong PT
Predictor no.2
Predictor no.3
Multi-correlation (R, stepwise)
Correlation R, between calculated and observed temperature for the independent subsample
NAO
TglobeJA
0.75
Holdout validation
Exp. 1
Exp. 2
Exp. 3
weak–strong
PT
NAO
TglobeJA
0.75
weak–strong
PT
NAO
TglobeJA
0.80
weak–strong PT
0.80
0.64
0.71
NAO
–
0.76
Figure 4. The observed and calculated (Eq. (1)) SAT for the summer season.
The multi-regression model for the SAT was applied to
nine potential predictors including four individual synoptic
types, three groups of types, the AO and the global
temperature anomaly (TglobeDJFM ). The two ‘cold’ groups
are defined as lows to the east (comprising the shallow and
deep lows east of Cyprus) and lows to the north (comprising
the shallow and deep lows north of Cyprus). The ‘warm’
group includes the five types which are mostly positively
correlated with the SAT (high east of Israel, RST–western
and central axes, high west of Israel and high over Israel).
The correlation between the collective occurrence of the five
warm types and the SAT was found to be +0.48.
c 2010 Royal Meteorological Society
Copyright The multi-regression yielded the following prediction
equation:
TDJFM = 7.768 − 0.0916 × fLOWSe − 0.595 × AO − 0.0492
× fLOWSn + 0.875 × TglobDJFM ,
(2)
The correlation between the modelled and the observed
SAT for the full sample was +0.81 when all the potential
predictors were included, and +0.80 when only the four
predictors found significant (Eq. (2)) were used. Following
the absence of warm types in the prediction equation
we tried several alternative combinations of warm groups,
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
Average monthly
occurrence (%)
7.1
1.3
13.8
10.6
6.2
5.7
16.1
5.8
1.0
10.4
6.6
5.4
8.9
Synoptic types for
December-March (DJFM)
High east of Israel
RST – western axis
High west of Israel
RST – central axis
High over Israel
Cold low west of Cyprus
RST – eastern axis
Shallow low north of Cyprus
Sharav low west of Israel
High north of Israel
Deep low north of Cyprus
Deep low east of Cyprus
Shallow low east of Cyprus
Correlation (R)
with 850 hPa SAT
+0.27
+0.17
+0.17
+0.17
+0.15
+0.09
+0.07
−0.02
−0.04
−0.05
−0.21
−0.41
−0.52
850 hPa temp (◦ C)
deviation from normal and STD
+2.4 ± 3.5
+5.6 ± 3.3
−0.1 ± 3.8
+1.6 ± 4.0
+0.3 ± 3.1
+0.4 ± 3.5
+0.6 ± 4.1
+0.4 ± 4.2
+4.2 ± 4.9
−0.8 ± 3.9
−1.6 ± 3.3
−4.2 ± 3.2
−2.4 ± 3.5
c 2010 Royal Meteorological Society
Copyright +0.00
+0.39
+0.28
+0.39
+0.65
+0.11
+0.21
+0.08
−0.01
−0.25
−0.84
−0.68
−0.40
−0.73
−0.30
−0.55
−0.26
−0.24
−0.07
+0.27
−0.03
+0.12
+0.08
+0.25
+0.59
+1.17
6 coldest winters
(2.1 ◦ C below normal)
Deviation in the occurrence
from the average (STD units)
6 warmest winters
(1.8 ◦ C above normal)
Table III. As in Table I, but for the 13 most frequent types of the winter season (DJFM).
312
H. Saaroni et al.
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
Factors Governing Temperature Variation and Trend Over Israel
but none of them was found significant for the SAT. It is
worth noting that when the synoptic predictors only were
used, the correlation for those found significant was +0.66.
The results of the holdout validation experiments are
shown in Table IV. In each of the three experiments,
three significant predictors were found. The lows to the
east and the AO appeared in all of them, the lows
to the north in two of them, and the TglobDJFM in
one of them. The holdout validation implies that the
regional cold advection and the large-scale circulation
are most prominent for the regional winter temperature
regime.
Figure 6 compares the observed and the calculated
(Eq. (2)) SAT for the winter season during the study period.
The standard error of the prediction was 0.7 ◦ C. As for the
summer, we examined the anomaly of the occurrence of
the synoptic types for the 10% coldest and warmest winters
(Table III). The deviation of the average temperature for
the warmest and the coldest seasons from the normal is
+1.8◦ C and −2.1◦ C, respectively (over 1.5 times larger than
the interannual STD, at 1.1◦ C). Here also the prediction
equation underestimates both extreme seasons, with a bias
of −0.75 and +0.71 for the warmest and the coldest seasons,
respectively. Unlike the summer, no consistent difference
was found here between the coolest and warmest seasons in
the occurrence of the types that are significantly correlated
with the SAT (Table III). The same results were found when
applied for the eight coolest and warmest winters.
3.3.
The contributors to the long-term trend
The trend of the observed temperature for the study period
is +0.02 K y−1 (significant at the 0.05 level) for the
summer season and +0.002 K y−1 (non-significant) for
the winter season. The contributions of the three factors,
the synoptic, the large-scale and the global radiative forcing,
are summarized in Table V. In spite of the warming trend
observed in both seasons, not all of the factors make a
positive contribution. The global radiative forcing is the
largest in both seasons. The synoptic factor, also being
positive, is considerably more pronounced in the summer.
The large-scale factor is negative in both seasons. This can be
explained by the combination of the positive trend in both
the AO and the NAO during most of the study period till
the mid-1990s (Holland, 2003; Overland and Wang, 2005;
IPCC, 2007) and the significant negative correlation between
the AO (in the winter) and the NAO (in the summer) and
the 850 hPa SAT over the EM (Figure 7).
The contribution of the synoptic factor can be inferred
from the trend observed in the occurrence of the types found
to be significant predictors. Figure 8(a) shows the trend of
the two types included in the prediction equation of the
summer season, i.e. the weak and the strong PT. The weak
PT occurrence increased at a rate of +83%/100 years, while
that of the strong PT decreased at a rate of −42%/100 years.
The extreme reversed trends explain the high positive
contribution (+1.28 K/100 y) of the synoptic factor in
the summer. Figure 8(b) shows the trend of the occurrence
for the two groups included in the prediction equation of
the winter season, i.e. the lows to the north and the lows
to the east. The occurrence of lows to the north decreased
at a rate of −77%/100 years, while that of the lows to the
east increased at a rate of +9%/100 years. The contradictory
trends of the two groups of these cool types explain the small
c 2010 Royal Meteorological Society
Copyright 313
(positive) contribution (+0.30 K/100 y) of the synoptic
factor in the winter.
4.
Discussion and conclusions
This study examines the ability of the interannual variability
in the occurrence of the EM synoptic types, the large-scale
circulations and the global temperature to explain that of
the 850 hPa temperature in Israel. The analysis was done
separately for the midsummer months (JA) and the winter
months (DJFM), each characterized by a specific collection
of prevailing synoptic types. The differences between the
average temperatures for the days belonging to the various
synoptic types deviated by up to ±3◦ C for the summer and
±5◦ C for the winter, similar to the interdiurnal STD of the
temperature for the respective seasons (e.g. Ziv et al., 2004).
For the summer, the prevailing six synoptic types were
transformed into three compound predictors. One of them,
the difference between the occurrences of the strong PT
(the coolest type) and the weak PT (the second warmest)
was found significant, explaining 35% of the interannual
variance of temperature. For the winter, the 13 types
were grouped into seven potential predictors. Two groups
(containing four types) were found significant. These are the
lows to the east and the lows to the north, explaining 44%
of the interannual temperature variance.
The limited variance explained by the synoptic factor
alone suggests that additional factors should be examined.
We incorporated two additional factors: large-scale circulations, which were found to be correlated with the regional
temperature (the NAO for the summer and the AO for
the winter), and the global temperature, as a representation
of the global radiative forcing. Both were found significant
predictors for the summer and the winter. The interannual
variation of the temperature explained by the full set of
significant predictors was 56% for the summer and 64% for
the winter.
A weakness of prediction equations is their tendency to
underestimate extreme seasons. The bias in the prediction
of the SAT for the upper and lower tenths for the summer
was 49% for the hot seasons and 32% for the cold seasons.
For the winter, the bias was 41% for the hot seasons and
34% for the cold ones.
The individual contribution of each of the three factors
to the long-term temperature trend was estimated. The
contribution of the global radiative forcing was large and
positive in both seasons. The contribution of the synoptic
factor, though positive in both seasons, was found to be
four times larger in the summer. The contribution of
the large-scale factor, though negative in both seasons,
was found to be three times larger in the winter. The
considerable warming in the summer imparted by the
synoptic factor was found to be the result of rapid increase in
the occurrence of the weak PT ‘on the account’ of the strong
PT. The cumulative warming contribution of both the global
radiative forcing and the synoptic factor has a threatening
implication, especially if they are correlated. Indeed, the
correlation between these two factors was significantly
high (R = +0.47). This implies that the temperature over
the EM may continue to rise more rapidly than the
global one.
To summarize, the results of this study demonstrate
how the combination of the synoptic-scale systems, the
large-scale circulations and the global radiative forcing
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
314
H. Saaroni et al.
Figure 5. Average 850 hPa temperature (light grey columns), and the correlation coefficient (R) with the occurrence (dark grey columns), of the most
frequent winter (DJFM) synoptic types. The long-term seasonal average temperature is denoted by the solid horizontal line. The standard deviation of
the temperature is presented in Table III.
Table IV. As in Table II, but for the months December–March.
Experiment
Multi-regression full sample
Predictor no.1
Lows to the east
Predictor no.2
Predictor no.3
AO
Lows to the north
Predictor no.4
Multi-correlation (R, stepwise)
Correlation R, between calculated and observed temperature – for the independent
sub-sample
T globDJFM
0.80
Holdout validation
Exp. 1
Exp. 2
Exp. 3
Lows to the
east
AO
Lows to the
north
–
0.79
AO
Lows to the east
TglobDJFM
Lows to the
east
–
0.83
Lows to the north
AO
0.70
0.63
0.77
–
0.76
Figure 6. The observed and calculated (Eq. (2)) SAT for the winter season.
controls the temperature over the EM. The approach Acknowledgements
elaborated here may be useful in predicting the future
temperature regime, based on synoptic features and large- This study was supported by the Israeli Science Foundation
scale circulations extracted from output of climatic models. (ISF, grants no. 764/06 and no. 828/02). Partial support
c 2010 Royal Meteorological Society
Copyright Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
Factors Governing Temperature Variation and Trend Over Israel
315
Figure 7. The correlation between (a) the AO in December–March and (b) the NAO in July–August with the 850-hPa SAT over the MB, based on the
NCEP/NCAR CDAS-1 archive (Kalnay et al., 1996; Kistler et al., 2001).
Table V. The individual contributions of the synoptic, large-scale
and the global radiative forcing factor (K/100 y) for the period
1950–2008.
Factor
Summer
Winter
Synoptic
Large-scale
Global radiative forcing
Total
Observed
+1.28
−0.34
+1.57
+2.50
+2.13
+0.30
−1.03
+1.28
+0.55
+0.23
Appendix
The Eastern Mediterranean Synoptic Systems and Types
Alpert et al. (2004b) defined five synoptic systems,
subdivided into 19 types, according to their intensity and
location with respect to Israel. The types’ definition is based
on the GPH, wind and temperature fields at the 1000-hPa
level, taken from the NCEP-NCAR initialized data (Kalnay
et al., 1996; Kistler et al., 2001). The systems and types
are three types of the Red Sea Trough (RST), three types
of the Persian Trough (PT), four types of highs, seven
types of Mediterranean Cyclones, or Cyprus Lows (referred
hereafter as ‘lows’) and two types of North African cyclone,
or Sharav Lows (SL). Each map shows one of the types
(Figures A1–A5), belonging to each of the synoptic systems,
and was taken from the date on which the conditions were
the most representative for the pertinent synoptic type,
i.e. the specific day was located right in the middle of the
respective cluster in the states’ space. The average annual
frequency and STD of each type and the months in which it is
most frequent are specified in Table AI and their description
follows.
A1.
Figure 8. Time series and LTT of the occurrence (% days per season
from normal) (a) for the summer types included in the prediction
equation, i.e. the weak (grey) and the strong (black) PT, and (b) for
the two groups included in the prediction equation of the winter
season, i.e. the lows to the north (black) and the lows to the east
(grey).
is due to the GLOWA-JR funded by the BMBF and
Ministry of Science (MOS), Israel. NCEP Reanalysis data
was provided by the NOAA-CIRES Climate Diagnostics
Center, Boulder, Colorado, USA, from their website at
http://www.cdc.noaa.gov/.
c 2010 Royal Meteorological Society
Copyright Red Sea Trough (RST)
The RST is a non-migratory pressure trough extending
from the south. It is most frequent during the autumn and
significant also during the winter and fades by mid-spring
(Alpert et al., 2004a). The wind direction and consequently
the weather conditions in Israel are highly dependent on
the position of the RST axis (Saaroni et al., 1998; Tzvieli
and Zangvil, 2005). Therefore, its location with respect
to Israel was used for categorizing the three types of the
RST. The first is the RST–eastern axis, which induces a
northwesterly maritime flow over Israel. The second, the
RST–western axis, induces southeasterly warm and dry
flow and the third is the RST–central axis, located over
Israel (Figure A1). When combined with an upper-level
trough, the RST may become active and produce local
showers, especially over south and east Israel (Kahana et al.,
2002). The annual frequency of the RST (all three types)
is 19%.
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
316
H. Saaroni et al.
Figure A1. SLP (hPa) for a selected date representing the synoptic system of
the Red Sea Trough, based on the NCEP/NCAR CDAS-1 archive (Kalnay
et al., 1996; Kistler et al., 2001).
Figure A3. As in Figure A1 but for a synoptic type of Highs.
Figure A2. As in Figure A1 but for a synoptic type of the Persian Trough.
Figure A4. As in Figure A1 but for a synoptic type of the Mediterranean
Cyclone.
A2.
Persian Trough (PT)
pressure and weak maritime advection (Figure A2), and the
The PT is also a non-migratory lower-level trough ‘deep’ (strong) PT has the lowest pressure and the strongest
extending from the Asian monsoon over the Persian Gulf maritime cool advection.
toward southern Turkey and the Aegean Sea. This system
persists during the summer months, mostly from June A3. Highs
to September, and its annual frequency is 31%. The PT
induces northwesterly flow over the EM, known as the Four types of highs were defined according to their location
Etesian wind, which transports relatively cool and humid with respect to Israel: highs located to the east, to the west, to
air toward the EM (Alpert et al., 1990; Saaroni and Ziv, the north and over Israel. Their annual frequency is 32%, and
2000; Ziv et al., 2004). Dayan et al. (2002) divided the PT they are observed all year round. The highs to the west are
into three types, ‘deep’, ‘moderate’ and ‘shallow’, according the most frequent (Tables III and A1), typically connected
to the pressure difference between Cairo and Nicosia. They to the Azores High (the Subtropical High) through a ridge
found that the three types differ in the wind speed they extending along North Africa. They induce northwesterly
produce, being maximal under a ‘deep’ PT and minimal humid flow from the Mediterranean Sea onshore. The rest
under a ‘shallow’ one. The semi-objective classification of the highs, including a small proportion of the highs
also divides the PT into three types, in line with Dayan to the west, are connected to the Euro-Asian polar highs,
et al. (2002), according to the depth of the trough and which dominate in the cool season (Saaroni et al., 1996).
the pressure gradient. The ‘weak’ PT has relatively high The highs east of Israel also induce southeasterly warm and
c 2010 Royal Meteorological Society
Copyright Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
Factors Governing Temperature Variation and Trend Over Israel
317
are the major source of the rainfall and floods in Israel
(Goldreich, 2003). Shay-El and Alpert (1991) showed that
they transport cold air originating from eastern Europe into
the southeastern Mediterranean, which becomes moist and
unstable there, and so produces convective and widespread
rains over Israel. Seven types of low are defined according to
their momentary location with respect to Cyprus (along their
path, normally from east to west) and their depth (illustrated
in Figure A4). These criteria determine the pressure gradient
and the wind direction and speed. An individual low can
change its type definition while moving across the study
region.
A5.
Sharav Lows (SL)
These warm and shallow cyclones originate mainly in
North Africa. They are most frequent during March–May,
peaking in April, and move toward the EM along the
Figure A5. As in Figure A1 but for the synoptic type of the North African North African shoreline, much faster than do the Cyprus
(Sharav) Cyclone.
Lows (Alpert and Ziv, 1989). Two SL types are defined
according to their location relative to Israel. When the
dry flow, sometimes strengthening to gale force (Saaroni SL is located to the west, Israel is affected by its warm
et al., 1998). When located over Israel, the highs cause calm sector, typified by southeasterly winds, resulting in warm
and dry conditions. While it crosses Israel (Figure A5),
weather.
southerly winds induce extremely warm and dry conditions
(Sharav). Both SL types are often accompanied by dust
A4. Mediterranean Cyclones (Cyprus Lows)
storms.
It is worth noting that since the synoptic types are defined
These cyclones, which are midlatitude disturbances, have
an annual frequency of 17% and are most frequent in according to the 1000-hPa fields alone, they reflect mainly
the cool season, i.e. DJFM (Alpert et al., 2004a). They advective features.
Table A1. Average annual frequencies and standardized interannual STD (%) for all the synoptic types.
Synoptic type
RST - eastern axis
RST - western axis
RST - central axis
Total for the RST
Weak PT
Medium PT
Strong PT
Total for the PT
High east of Israel
High west of Israel
High north of Israel
High over Israel
Total for the highs
Deep low north of Cyprus
Deep low east of Cyprus
Deep low south of Cyprus
Low west of Cyprus
Shallow low north of Cyprus
Shallow low east of Cyprus
Shallow low south of Cyprus
Total for the lows
Sharav Low west of Israel
Sharav Low over Israel
Total for the Sharav Lows
Average annual
frequency (%)
12.0
0.7
6.4
19.1
14.3
14.7
2.1
31.1
3.3
19.9
4.4
4.3
31.9
2.7
2.3
0.1
2.2
4.7
4.6
0.1
16.7
0.6
0.6
1.2
c 2010 Royal Meteorological Society
Copyright Interannual
STD (%)
Main period
of occurrence
Main peak
21
61
39
October-April
Nov-Mar
October-May
Oct–Nov
Dec
Nov
24
22
87
May-Oct
May-Sep
May-Sep
Aug
Jul
Jun, Sep
39
19
36
37
Nov-Mar
All year round
Nov-Apr
October-May
Nov-Dec
May, Oct
Dec
April
43
46
192
44
29
36
221
Dec-Mar
Nov-Apr
Dec-Mar
Nov-Mar
late Sep-May
Oct-Apr
Dec- Feb
Feb
Feb
Feb
Dec, Feb
Mar-Apr
Jan
Dec
79
76
Jan-May
Mar-May, Oct
Mar
April
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)
318
H. Saaroni et al.
References
Alpert P, Ziv B. 1989. The Sharav cyclone: Observations and some
theoretical considerations. J. Geophys. Res. 94: 18495–18514.
Alpert P, Abramsky R, Neeman BU. 1990. The prevailing summer
synoptic system in Israel: Subtropical high, not Persian trough. Isr.
J. Earth Sci. 39: 93–102.
Alpert P, Osetinsky I, Ziv B, Shafir H. 2004a. A new seasons’ definition
based on the classified daily synoptic systems: An example for the
eastern Mediterranean. Int. J. Climatol. 24: 1013–1021.
Alpert P, Osetinsky I, Ziv B, Shafir H. 2004b. Semi-objective classification
for daily synoptic systems: Application to the eastern Mediterranean
climate change. Int. J. Climatol. 24: 1001–1011.
Ben-Gai T, Bitan A, Manes A, Alpert P, Rubin S. 1999. Temporal and
spatial trends of temperature patterns in Israel. Theor. Appl. Climatol.
64: 163–177.
Bitan A, Saaroni H. 1992. The horizontal and vertical extension of the
Persian Gulf trough. Int. J. Climatol. 12: 733–747.
Dayan U, Ziv B, Margalit A, Morin E, Sharon D. 2001. A severe autumn
storm over the Middle-East: Synoptic and mesoscale convection
analysis. Theor. Appl. Climatol. 69: 103–122.
Dayan U, Lifshitz-Goldreich B, Pick K. 2002. Spatial and structural
variation of the atmospheric boundary layer during summer in
Israel: Profiler and rawinsonde measurements. J. Appl. Meteorol.
41: 447–457.
Giorgi F, Hewitson B, Christensen J, Hulme M, von Storch H,
Whetton P, Jones R, Mearns L, Fu C. 2001. Regional climate
information: Evaluation and projections. Pp 583–638 in: Climate
Change 2001: The Scientific Basis. Contribution of Working Group I
to the Third Assessment Report of the Intergovernmental Panel on
Climate Change, Houghton JT, Ding Y, Griggs DJ, Noguer M, van der
Linden PJ, Dai X, Maskell K, Johnson CA (eds), Cambridge University
Press: Cambridge, UK and New York. http://www.ipcc.ch (Chapter
10 of the IPCC WG1 Assessment).
Goldreich Y. 2003. The climate of Israel: Observation, research, and
application. Kluwer Academic/Plenum: New York.
Hansen J, Ruedy R, Glascoe J, Sato M. 1999. GISS analysis of surface
temperature change. J. Geophys. Res. 104: 30997–31022.
Hansen J, Ruedy R, Sato M, Imhoff M, Lawrence W, Easterling D,
Peterson T, Karl T. 2001. A closer look at United States and global
surface temperature change. J. Geophys. Res. 106: 23947–23963.
HMSO. 1962. Weather in the Mediterranean. I: General meteorology. 2nd
edition. Her Majesty’s Stationery Office: London.
Holland MM. 2003. The North Atlantic Oscillation–Arctic Oscillation
in the CCSM2 and its influence on Arctic climate variability. J. Climate
16: 2767–2781.
Huth R. 2002. Statistical downscaling of daily temperature in central
Europe. J. Climate 15: 1731–1742.
IPCC. 2007. Intergovernmental Panel on Climate Change – The Scientific
Basis (contribution of WG I to the 4th Assessment Report of the
IPCC). Cambridge University Press: Cambridge.
Jones PD, Jónsson T, Wheeler D. 1997. Extension to the North Atlantic
Oscillation using early instrumental pressure observations from
Gibraltar and south-west Iceland. Int. J. Climatol. 17: 1433–1450.
Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L,
Iredell M, Saha S, White G, Woollen J, Zhu Y, Leetmaa A,
Reynolds R, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC,
Ropelewski C, Wang J, Jenne R, Joseph D. 1996. The NCEP/NCAR
40-year reanalysis project. Bull. Am. Meteorol. Soc. 77: 437–471.
Kahana R, Ziv B, Enzel Y, Dayan U. 2002. Synoptic climatology of major
floods in the Negev Desert, Israel. Int. J. Climatol. 22: 867–882.
Kistler R, Kalnay E, Collins W, Saha S, White G, Woollen J, Chelliah M,
Ebisuzaki W, Kanamitsu M, Kousky V, van den Dool H, Jenne R,
Fiorino M. 2001. The NCEP-NCAR 50-year reanalysis: Monthly
means CD-ROM and documentation. Bull. Am. Meteorol. Soc. 82:
247–267.
c 2010 Royal Meteorological Society
Copyright Kutiel H, Benaroch Y. 2002. North Sea–Caspian Pattern (NCP) – an
upper level atmospheric teleconnection affecting the eastern
Mediterranean: Identification and definition. Theor. Appl. Climatol.
71: 17–28.
Kutiel H, Maheras P, T&urkeş M, Paz S. 2002. North Sea–Caspian
Pattern (NCP) – an upper level atmospheric teleconnection affecting
the eastern Mediterranean: Implications on the regional climate.
Theor. Appl. Climatol. 72: 173–192.
Osetinsky I, Alpert P. 2006. Calendaricities and multimodality in the
eastern Mediterranean cyclonic activity. Nat. Hazards Earth Syst. Sci.
6: 587–596.
Otterman J, Angell JK, Ardizzone J, Atlas R, Schubert S, Starr D,
Wu M-L. 2002. North-Atlantic surface winds examined as the
source of winter warming in Europe. Geophys. Res. Lett. 29: 1912,
DOI:10.1029/2002GL015256.
Overland JE, Wang M. 2005. The Arctic climate paradox: The recent
decrease of the Arctic Oscillation. Geophys. Res. Lett. 32: L06701,
DOI:10.1029/2004GL021752.
Rauthe M, Paeth H. 2004. Relative importance of Northern Hemisphere
circulation modes in predicting regional climate change. J. Climate
17: 4180–4189.
Saaroni H, Ziv B. 2000. Summer rain episodes in a Mediterranean
climate, the case of Israel: Climatological–dynamical analysis. Int.
J. Climatol. 20: 191–209.
Saaroni H, Bitan A, Alpert P, Ziv B. 1996. Continental polar outbreaks
into the Levant and eastern Mediterranean. Int. J. Climatol. 16:
1175–1191.
Saaroni H, Ziv B, Bitan A, Alpert P. 1998. Easterly wind storms over
Israel. Theor. Appl. Climatol. 59: 61–77.
Saaroni H, Ziv B, Edelson J, Alpert P. 2003. Long-term variations in
summer temperatures over the eastern Mediterranean. Geophys. Res.
Lett. 30: 1946, DOI: 10.1029/2003GL017742.
Shay-El Y, Alpert P. 1991. A diagnostic study of winter diabatic heating
in the Mediterranean in relation to cyclones. Q. J. R. Meteorol. Soc.
117: 715–747.
Spak S, Holloway T, Lynn B, Goldberg R. 2007. A comparison of statistical
and dynamical downscaling for surface temperature in North America.
J. Geophys. Res. 112: D08101, DOI:10.1029/2005JD006712.
Thompson DWJ, Wallace JM. 1998. The Arctic oscillation signature in
wintertime geopotential height and temperature fields. Geophys. Res.
Lett. 25: 1297–1300.
Tsvieli Y, Zangvil A. 2005. Synoptic climatological analysis of ‘wet’ and
‘dry’ Red Sea Troughs over Israel. Int. J. Climatol. 25: 1997–2015.
Vinther BM, Andersen KK, Hansen AW, Schmith T, Jones PD. 2003.
Improving the Gibraltar/Reyjavik NAO index. Geophys. Res. Lett. 30:
2222, DOI:10.1029/2003GL018220.
von Storch H. 1999. On the use of ‘inflation’ in statistical downscaling.
J. Climate 12: 3505–3506.
Xoplaki E, González–Rouco JF, Luterbacher J, Wanner H. 2003a.
Mediterranean summer air temperature variability and its connection
to the large-scale atmospheric circulation and SSTs. Clim. Dyn. 20:
723–739.
Xoplaki E, González–Rouco JF, Gyalistras D, Luterbacher J, Rickli R,
Wanner H. 2003b. Interannual summer air temperature variability
over Greece and its connection to the large-scale atmospheric
circulation and Mediterranean SSTs 1950–1999. Clim. Dyn. 20:
537–554.
Ziv B, Yair Y. 1994. Introduction to meteorology, Vol. II. The Open
University of Israel: Tel Aviv (in Hebrew).
Ziv B, Saaroni H, Alpert P. 2004. The factors governing the summer
regime of the eastern Mediterranean. Int. J. Climatol. 24: 1859–1871.
Ziv B, Saaroni H, Baharad A, Yekutieli D, Alpert P. 2005. Indications
for aggravation in summer heat conditions over the Mediterranean
Basin. Geophys. Res. Lett. 32: L12706, DOI:10.1029/2005GL022796.
Q. J. R. Meteorol. Soc. 136: 305–318 (2010)