Content uploaded by Demetris Koutsoyiannis
Author content
All content in this area was uploaded by Demetris Koutsoyiannis on Jul 08, 2014
Content may be subject to copyright.
Influence of atmospheric circulation types in space - time
distribution of intense rainfall
Nikos Mamassis and Demetris Koutsoyiannis
Department of Water
Resources,
Faculty of Civil Engineering, National Technical University, Athens, Greece
Abstract. The influence of the prevailing weather situation on the temporal evolution and
geographical distribution of
intense
rainfall is studied, as a potential tool to improve rainfall
prediction. A classification scheme of the atmospheric circulation over the east Mediterranean
territory is used for the analysis. The study area is the Sterea Hellas region (central Greece)
with an area of about 25,000 km
2
. Daily data from
71
rain gages and hourly data from three
rain recorders over a 20 year period are used. From these data sets, the intense rainfall events
were extracted and analyzed. Several empirical and statistical methods (also including the
available tools of
a
Geographical Information System) are used for the analysis and comparison
of rainfall distribution both in time and in space. The analysis shows that the contribution of
the
concept of weather types to the quantitative point rainfall prediction in short timescale is small,
and only the estimation of
the
probability of occurrence of an intense event is feasible. On the
contrary, the relation between the spatial distribution of rainfall and the atmospheric circulation
patterns is significant and may be used for improving the forecasting of the geographical
distribution of rainfall.
1.
Introduction
The space-time evolution of
the
rainfall process is related to
the characteristics of the prevailing weather situation that
caused the rainfall. To study more effectively this relationship,
several researchers have classified similar weather situations
of particular regions into specific types. The classification
schemes generally fall into three main categories, each corre-
sponding to a different approach to the compilation of the
meteorological data and weather maps.
The first category includes the schemes that are based on
the combination of the different value ranges of several local
meteorological variables. McCabe [1989] separated the range
of wind direction into three classes and the range of cloudiness
values into two classes. Combining these classes, he defined
six weather types, which were used by Hay et al. [1991] for
the classification, modeling, and simulation of daily rainfall in
a watershed in the United States. Wilson et al. [1991] used
surface air pressure and 850-hPa level temperature to classify
and simulate daily rainfall in the northern Pacific. McCabe
[1994] used height anomalies (in meters) for the 700-hPa
level, to represent atmospheric circulation of the western
United States, and relate that to the variability of the snow-
pack accumulation. Hughes and Guttorp [1994] used sea
air pressure over a wide area larger than Europe. This scheme
was based on a timescale of several days (at least three) con-
sidering that the main features of weather across Europe
remain constant during one time step of that length. Bardossy
and Plate
[1991,
1992] used this scheme for daily rainfall
modeling and simulation in Germany. Following the scheme
ofBauretal.
[1944],
Duckstein et
al.
[1993] developed a cir-
culation pattern classification scheme for the continental
United States and used it to relate flood occurrence in central
Arizona to circulation patterns. Schuepp [1968] developed a
weather classification scheme for the European Alpine region,
based on the airmass advection (related to source areas), tem-
perature characteristics, and cyclonicity. This scheme was
used for the study of daily runoff in northern Italy by Van de
Gried and Seyhan [1984]. Lamb [1950, 1972] defined a clas-
sification scheme of atmospheric circulation for the British
Isles,
based on the specific patterns in weather maps of the
surface and the 500-hPa level. Wilby [1994] used Lamp's
scheme to simulate circulation patterns and hence daily rain-
fall in England and
Wilby
et
al.
[1994] to simulate daily flows.
Wilby [1995] used the same scheme to study and model daily
rainfall in England by incorporating also in the classification
the presence or absence of weather fronts. Maheras
[
1982],
based on the Lamb [1972] method, introduced a classification
scheme of atmospheric circulation over the East Mediterra-
nean territory. This scheme was used by Mamassis and Kout-
soyiannis [1993] and Mamassis et al: [1994] for the analysis
of intense rainfall and flood events in Greece and was also
adopted in this study. Details of this classification scheme are
given in section 3.
Weather classifications may be viewed as a systematization
of meteorological experience in a particular region. Obviously,
there is a relationship between the geographical characteristics
of a particular area (such as the geographical location, the
relative position with regard to the sea, the orography), and the
climatological regime in this area. This relation may be cap-
tured empirically by using weather types and studying their
influence on hydrometeorological processes. This approach
can potentially be combined in a complementary way with the
outputs of general circulation models (GCM) in order to
downscale these ouputs to a finer spatial scale, localized to the
specific area of interest. In particular, the empirical study of
the relationship between weather types and the rainfall process
in a specific area may be useful to translate the synoptic con-
ditions used to define the weather types, into quantitative
information about rainfall. If
such
a relationship can indeed be
established, it will improve rainfall prediction on a local scale.
With regard to these aspects, research was recently carried out
daily rainfall fields and process them statistically. Similar
studies have typically used multivariate models for that pur-
pose.
This paper is outlined as follows: In section 2 the study
area and the available data sets are described and in section 3
the weather type classification scheme is presented. In section
4 the point rainfall process on an hourly basis as well as the
total storm characteristics of intense rainfall events are
studied. In section 5 the methodology of analyzing the spatial
distribution of rainfall using a grid-based technique is
described. Finally, the conclusions are presented in section 6.
2.
Study Area and Data Used
The study area is the Sterea Hellas region (central Greece)
with an area of approximately 25,000 km
2
(about one fifth of
the total area of Greece, Figure 1). This region includes five
important and many smaller rivers. One of
them
(the Acheloos
River) is the largest river (in discharge) in Greece providing
water for irrigation and hydropower. Three others (Evinos,
Momos, B. Kifissos) provide water supply to the area of
Athens. The Pindus mountain chain on the west side of this
region causes heavy orographic rainfall and therefore a wetter
rainfall regime, as compared to that of the east side. Thus the
annual rainfall varies from about 2000 mm in the north-
western part of the region to about 400 mm in the southeastern
part (Athens).
Daily rainfall data from
71
rain gages over the entire region
and hourly rainfall data from three of
them
equipped with rain
recorders (Krikello, Aniada, Drymonas, located at the Evinos
River basin) were available for a 20 year period (1970-1990).
Figure 1 shows the general location of the study area, its
morphology, and the available rain gages.
From the continuous records we have extracted and studied
only the intense rainfall events. Intense rainfall has the most
practical interest as it is responsible for extreme floods in the
study area. From the hourly point rainfall data sets the intense
rainfall events were extracted using a criterion based on a
threshold of hourly rainfall (above 7 mm) or daily rainfall
(above 25 mm). Analysis was performed on the data of all the
three rain-recording stations mentioned above, but here we
present the results from one of
them
(Krikello), as those of the
other stations were very similar. The intense rainfall events
were separated into rainy (October to April) and dry (May to
September) season events. In total, 200 events belong to the
rainy and 93 events to the dry season.
For the daily data sets of the 71 gages the intense rainfall
days were extracted using a criterion based on a threshold of
centers of the anticyclones, (2) the main trajectories of the
cyclones, and (3) some special synoptic situations at the sur-
face and at the 500-hPa level.
According to this scheme the circulation patterns of
the
ter-
ritory were classified into five anticyclonic, six cyclonic, two
mixed, and three characteristic weather types [Maheras,
1982].
Table 1 shows a summary description of the above
weather types. Figure 2 shows the main trajectories of the six
cyclonic weather types which are responsible for bad weather
and produce the main amount of rainfall in the study area. In
four of them (Wl, SW1, NW1, W2) the disturbance passes
near the study area and provokes intense rainfalls, especially
on the west side of the area where the Pindus mountain chain
lies.
The SW1 and NW1 weather types are very common in
Greece especially in the rainy season and have a significant
influence on the annual rainfall regime of the study area. In
the other two cyclonic types the disturbance passes a long
distance away from the study area and the generation of
intense rainfall is rare. Especially for the SW2 weather type,
the trajectory of the cyclone passes through the Aegean Sea
and is often responsible for intense rainfall on the eastern side
of
the
study area.
Figure 3 shows a weather map of the 500-hPa level corre-
sponding to the weather type DOR. The presence of a cold air
mass at this level above Greece combined with a field of low
pressures with a weak gradient on the surface is typical for
this weather type. This weather situation has a high frequency
of occurrence in the dry season and causes atmospheric insta-
bility and intense convective rainfall at several sites in Greece
without a typical areal distribution.
Table 1
Figure 2
Figure 3
4.
Influence of Weather Types to Point Rainfall
In this section we use statistical methods to detect whether
the prevailing weather type affects the point rainfall charac-
teristics or not. More specifically, we examine the probability
of occurrence of intense rainfall, the total duration and depth,
and the hourly point rainfall structure and compare statisti-
cally these characteristics among different weather types.
The conditional probability of occurrence of an intense
rainfall event, given the prevailing weather type, was calcu-
lated using the Krikello data set and the daily calendar of
weather types. Table 2 shows this probability for each weather
type and Figure 4 for grouped weather types. To examine if
there are statistically significant differences in probabilities
(proportions of rainy days to the total number of days) among
weather types, we have applied the proportion statistical test
[Freund and Simon 1991 pp 386-388] The analysis shows
Table 2
Figure 4
the dry season the differences in the duration and total depth
among the various weather types are statistically significant
(at a 1% significance level) almost in all cases. However, in a
more quantitative point of view these differences are not so
important, as they explain a small percentage of variance of
these characteristics. Specifically, as shown in Table 4, the
percentage of variance explained by weather types for the
various characteristics varies from 3% to 7% for the rainy sea-
son and from 8% to 18% for the dry season.
Furthermore, we have calculated the marginal statistics
(mean and standard deviation), as well as the autocorrelation
function of hourly depths, which are shown in Table 3 for the
different weather types. No statistically significant differences
in the hourly depth are detected among different weather
types.
Hence only
1 %
of the variance of the hourly depth is
explained by the weather type concept for the rainy season and
2%
for the dry season (Table 4). Finally, there are no statisti-
cally significant differences on the autocorrelation function of
hourly depths among weather types, but there are differences
between the two seasons (Table 3). The atmospheric instabil-
ity which characterizes the rainy days of the dry season
explains the strong variability of rainfall which leads to a
weaker autocorrelation function of hourly depth on this sea-
son.
Table 3
Table 4
5. Influence of Weather Types to the Geographic
Distribution of Rainfall
In this section we examine the geographical distribution of
rainfall on a daily basis using a grid-based approach. The
daily data values were stored and analyzed using a GIS. For
each intense rainfall day the measured values of point rainfall
were used to determine a representative precipitation depth at
all grid cells of the study area. The algorithm used consists of
the following steps [Dingman, 1994]: (1) A grid covering the
study area is established. The grid spacing depends on the
analyst but is usually about one tenth of the average distance
between rain gages. In this study, the average distance
between the rain gages was about 19 km and thus the grid
spacing was chosen as 2 km. (2) Values of precipitation at
each grid cell are estimated as linear combinations of the
measured values, that is,
(1)
where
/>,
is the estimated precipitation at the ith grid cell,
p
%
is
the measured precipitation at station g, and w
Ig
is the weight
tional Thiessen method. In this study the IDW method, which
combines simplicity, availability in the GIS, and preservation
of the measured point values, was selected for estimating the
weights w
lg
. Another advantage of the IDW method is that it
does not require the rainfall field to be spatially homogeneous,
as it simply performs a local linear estimation. This feature is
very important in our case, where there exist apparent spatial
inhomogeneities in the rainfall characteristics over the study
area. On the contrary, the kriging method typically makes the
assumption of a constant semivariogram over the entire area,
which is not strictly valid in our case.
We remind that in the IDW method the weights w
ig
are a
function of the distance between each of the grid cells and
each of the rain gage locations, given by
where d(i, g) is the distance between the grid cell /' and the
gage g, G is the total number of gages, and b is a chosen
parameter, usually assigned to
1
or 2. In the case where a grid
cell is also a gage location, the rain depth assigned to this cell
is the value measured by the gage and the weights for the
other gages become zero.
The grid-based approach described above, which is based
on the fitting of a surface to point data, has some advantages
when compared to other methods, such as the multivariate
analysis. For example, it does not require filling of missing
data in order to fit a surface and offers a better understanding
and visualization of the rainfall field (e.g., localization of
regions with specific characteristics). Also, the approach
allows for the calculation of statistical surfaces instead of
point statistics, as described later in this section. The rainfall
surfaces can be combined with geographical, geological, and
land use surfaces for rainfall-runoff modeling. An apparent
weakness of
the
approach is that in most grid cells the rainfall
values are estimates rather than measurements, which obvi-
ously introduces errors in the field representation.
The analysis was performed only for the rainy season due to
the small number of intense rainfall days in the dry season.
Several methods were used for the analysis and comparison of
the rainfall spatial distribution.
As a first step, the rainfall fields in each weather type were
plotted using an appropriate color or gray scale, thus visual-
izing the spatial rainfall distribution. This assists the localiza-
tion of areas attracting intense rainfall. Then, the statistical
f ( dd dii ffii f rii) f
provokes intense rainfall in the eastern side of the study area.
In mixed types the combination of an anticyclone over Europe
and a cyclone over the Aegean Sea (Table 1) provokes bad
weather and rainfall in the eastern part of our study area, and
especially in the northern Evia. In the remaining types,
including all the anticyclonic and dry types, the few intense
rainfall events are related to the local atmospheric instability,
and hence no typical spatial distribution appears.
Furthermore, the correlation coefficients among all the
rainfall surfaces of
the
same weather type were calculated, and
their empirical distribution function was studied. Figure 7
shows the sample characteristics (median, maximum and
minimum value, and upper and lower quartile) of correlation
coefficients for each weather type. The majority of
the
correla-
tion coefficients are positive for all weather types. That means
that different intense rainfall events of the same weather type
are positively correlated in space. This indicates a similarity in
the spatial distribution of different rainfall events belonging to
the same weather type. Notably, as shown in Figure 7, all cor-
relation coefficients of the events of the Wl type are positive,
and most of
the
others have their lower quartile positive. There
are two exceptions, related to the types NW1 and SW2, whose
lower quartiles are negative and ranges of the correlation
coefficients are wider, thus indicating larger spatial variability
of rainfall. Yet, the median remains significantly higher than
zero for both types. In addition, as we observe in Figure 6, the
rainfall produced by these two weather types is generally
attracted at certain locations of the study area, as NW1 causes
heavy rainfall in the western part and SW2 in the eastern part
of
the
study area. For comparison we have plotted in Figure 7
(last box) a box plot of the sample correlation coefficients of
all events, regardless of the prevailing weather type. We
observe that in this case the range of
the
correlation coefficient
covers almost all the feasible interval [-1, 1]. The positive
median (0.15) reflects the inhomogeneity of the general rain-
fall regime in the study area, with higher rainfall in the Pindus
mountain chain.
As another technique to quantify the influence of weather
types on the rainfall distribution, apart from the grid-based
analysis described above, we also performed an analysis based
on the separation of the study area into subareas. Specifically,
we have separated the study area into 10 subareas (Figure 1)
climatically homogeneous, considering also the borders of
hydrologic basins for the separation. The statistics of the areal
rainfall of each subarea and each weather type were calcu-
lated. The mean daily rainfall per subarea and weather type is
presented in Table 5. We observe in Table 5 that each weather
type affects a nubr of neighboring subareas and each
Figure 7
Table 5
rainfall depth (more than 20% in 7 out of 10 subareas) is
explained by the concept of weather type. The values of the
explained variance given in Table 6 are not negligible like
those of hourly depth (Table 4). This may be interpreted as an
indication that the influence of the weather type on rainfall
depth increases with the increase of
the
timescale (from hourly
in Table 4 to daily in Table 6). To get a more solid under-
standing of the influence of timescale to the percentage of
variance explained by weather types, we aggregated the
Krikello hourly point rainfall depths of the rainy season to 6-,
12-,
and 24-hour timescales and we performed analysis of
variance for each scale. The analysis showed that the percent-
age of explained variance has a very small increase with the
increase of the timescale. Specifically, the percentages are
1.0%, 2.3%, 2.9%, and 3.9% for 1, 6, 12, and 24 hours,
respectively.
Thus the increase of timescale does not explain much of the
difference between the results of Table 4 and Table 6. Conse-
quently, the major part of this difference is explained by the
different criteria used for extracting the intense rainfall events
in the two cases. In the first case (Table 4), the selection of
events was based on the value of rainfall intensity or depth at
a point location, whereas in the second case (Table 6), the
selection was based on the magnitude of the rainfall depth at
any point of the study area. The data set of the second case
includes events that did not give rainfall at all at the point
location of the first case (Krikello). It is anticipated that, if
every event was considered, regardless of the rainfall magni-
tude (i.e., without a selection criterion), the percentage of vari-
ance explained by weather types would be even higher, as the
anticyclonic types would have a larger participation in the
sample. The apparent dissimilarities of anticyclonic and
cyclonic weather types would affect further the results of
analysis of variance. This is the case, for example, in the study
of Bardossy and Plate
[1991,
1992], where all events are
considered and the connection between rainfall and circulation
patterns is quite stronger.
6. Conclusions
In this study we have explored the connection between the
structure of intense rainfall in space and time and the atmo-
spheric circulation patterns in a large area in Greece. In the
analyses performed, we have used records (in hourly through
daily timescale) of intense rainfall events, in order to investi-
gate whether weather types can contribute to the prediction of
flood-producing storms.
The analysis of point rainfall characteristics shows that
weather types and the physiographical characteristics of the
study area.
Overall, the analysis shows that the contribution of the con-
cept of weather types to the quantitative rainfall prediction in
short timescales is small, but the estimation of the probability
of occurrence of an intense event is feasible. On the contrary,
the strong relation between the spatial distribution of rainfall
and the atmospheric circulation patterns may be used for
improving the forecast of the geographical distribution of
rainfall.
Acknowledgments. A part of this study was performed under the
framework of
the
research project AFOR1SM funded by the European
Union, DG XII (EPOC-CT90-0023). We are grateful to P. Maheras
for providing us with the daily calendar of weather types. We thank
A. Kazakos for the English review. We appreciate the constructive
comments by the Guest Editor E. Foufoula-Georgiou and two anony-
mous reviewers. Computer resources and precipitation data were pro-
vided by the Hydroscope Hydrometeorological Database System of the
National Technical University of Athens.
References
Bardossy, A, and E. Plate, Modelling daily rainfall using a semi-
Markov representation of circulation pattern occurrence, J,
Hydrol., 122,33-47,1991.
Bardossy, A, and E. Plate, Space-time model for daily rainfall using
atmospheric circulation patterns. Water Resour. Res., 28(5), 1247-
1259,
1992.
Baur, F., P. Hess, and H. Nagel, Kalender der Grosswetterlagen
Europas
1881-1939,
DWD, Bad Homburg, Germany, 1944.
Dingman, L., Physical Hydrology, Prentice-Hall, Englewood Cliffs,
N.
J., 1994.
Duckstein, L., A. Bardossy, and I. Bogardi, Linkage between the
occurrence of daily atmospheric circulation patterns and floods: An
Arizona case study, J. Hydrol., 143, 413-428, 1993.
Freund, J., and G. Simon, Statistics, A First Course, Prentice-Hall,
Englewood Cliffs, N. J., 1991.
Hay, L. E., G. J. McCabe Jr., D. M. Wolock, and M. A. Ayers, Simu-
lation of precipitation by weather type analysis. Water Resour.
Res.,
27(4),
493-501,
1991.
Huff,
F., Climatological assessment of natural precipitation character-
istics for use in weather modification, J. Appl. Meteorol., 8, 401-
410,
1969.
Hughes, J., and P. Guttorp, A class of stochastic models for relating
synoptic atmospheric patterns to regional hydrologic phenomena,
Water
Resour.
Res., 30(5), 1535-1546, 1994.
Joulis, V., Appraisal of space distribution of rainfall using
Geographical Information Systems, Diploma thesis, Natl. Tech.
Univ., Athens, 1996.
Lamb,
H. H., Types and spells of weather around the year in the Brit-
ish Isles, Annual trends, seasonal structure of the year. Singulari-
ties,
Q.
J.
R.
Meteorol. Soc.,
76,
393-438, 1950.
Lamb,
H. H., British Isles weather types and a register of the daily
sequences of circulation patterns 1861-1971, 86 pp., Geophys.
Mem 116 1972
10
Schuepp, M., Kalender der Wetter und Witterungslagen von 1955 bis
1967,
Veroffntl.
Schweiz.
Meteorol.
Zentralanst.,
II, 1-43, 1968.
Shaw, E., An analysis of the origins of precipitation in northern Eng-
land, 1956-1960,
Q.
J. R. Soc., 88, 539-547, 1962.
Van de Gried, A., and E. Seyhan, Statistical analysis of weather-type
runoff phenomena in an alpine environment, J. Hydroi,
70,
51-69,
1984.
Wilby, R., Stochastic weather type simulation for regional climate
change impact assessment, Water Resour. Res., 30{\2), 3395-
3403,
1994.
Wilby, R., Simulation of precipitation by weather pattern and frontal
analysis,
J.
Hydroi.,
173, 91-109, 1995.
Wilby, R., B. Greenfield, and C. Glenny, A coupled synoptic-hydro-
logical model for climate change impact, J. Hydroi,
153,
265-290,
1994.
Wilson, L., D. Lettenmaier, D., and E. Wood, Simulation of daily pre-
cipitation in the Pacific Northwest using a weather type classifica-
tion scheme, Surv. Geophys., /
2(1-3),
127-142, 1991.
D.
Koutsoyiannis and N. Mamassis, Department of Water
Resources, Faculty of Civil Engineering, National Technical Univer-
sity, Heroon Polytechniou 5, Zogragou, GR-157 80, Greece, (e-mail:
dk@hydro.civil.ntua.gr, nikos@hydro.civil.ntua.gr)
(Received October
27,
1995; revised April 15, 1996;
accepted April 25, 1996.)
Copyright 1996 by the American Geophysical Union
Paper number 96JD01377.
0148-0227/96/96JD-O1377S09.0O
Table 1. Summary Description of Weather Types
Main Category Abbreviation
Description of circulation
location of center in western Europe or northern Atlantic
location of center in Russian or Siberian region
location of center in Balkan region
location of center in eastern Mediterranean
location of center in western Mediterranean and northern Africa
cyclone passes from the Balkans over
45°
latitude
cyclone passes through Greece below
45°
latitude
cyclone originates from western Mediterranean or northwestern Europe and passes through Greece
cyclone circulates from Scandinavia to Black Sea
cyclone passes to the west of the line Malta - western Macedonia - Ukraine
cyclone passes to the east of the line Malta - western Macedonia - Ukraine
presence of an anticyclone over the central-northern Europe and a cyclone over Black sea;
isobars at surface have meridional arrangement
presence of an anticyclone over the central - northern Europe and a cyclone over eastern
Mediterranean or Aegean sea; isobars in surface have no specific arrangement
special combination between low pressure in southeastern Asia and very weak gradient in Balkans
(dry type)
very weak pressure gradient over Greece
presence of a cold air mass at the 500-hPa level above Greece
Continental
anticyclones
Maritime
anticyclones
Cyclones with
zonal orbit
Cyclones with
meridional orbit
Mixed types
Al
A2
A3
A4
A5
Wl
W2
NWl
NW2
SWl
SW2
MT1
MT2
Characteristic
types (special,
mostly dry
period types)
DES
MB
DOR
Table 2. Probability of Occurrence of Intense Point Rainfall
Events, Conditional on the Prevailing Weather Type
Rainy Season
Dry Season
Weather Total Number Prob- Total Number Prob-
Type Number of ability Number of ability
of
Days
Intense of of Days Intense of
Rainfall Occur- Rainfall Occur-
Days rence, % Days rence, %
Al
A2
A3
A4
A5
Wl
W2
NWl
NW2
SWl
SW2
MT1
MT2
346
372
199
146
101
155
123
571
266
615
215
149
278
1
1
0
0
0
15
17
63
14
71
14
0
0
0.3
0.3
0.0
0.0
0.0
9.7
13.8
11.0
5.3
11.5
6.5
0.0
0.0
237
268
161
67
13
155
4
36
268
238
48
91
4
0
0
0
0
0
4
1
4
9
14
2
4
0
0.0
0.0
0.0
0.0
0.0
2.6
25.0
11.1
3.4
5.9
4.2
4.4
0.0
Table 3. Statistics of Hourly Depth
12
Weather
Type
SW1
SW2
NW1
NW2
Wl
W2
Rainy Period
Mean Value,
mm
1.8
1.2
1.6
2.0
1.5
2.0
Standard
Deviation, mm
2.6
1.9
2.2
2.6
2.0
3.5
Lag
1
Auto-
correlation
0.55
0.54
0.55
0.58
0.38
0.56
Weather
Type
SW1.SW2,
W1.W2
NW1.NW2
DOR
Rest
Mean Va
mm
2.0
3.6
3.2
3.4
Dry Period
ue,
Standard
Deviation, mm
3.7
6.1
5.5
4.9
Lag
1
Auto-
correlation
0.44
0.20
-0.03
0.01
Table 4. Percentage of Variance in Point Intense Rainfall
Characteristics Explained by Weather Types
Variable
Rainy Season Dry Season
Duration
Total depth
Mean intensity
Hourly depth
7%
3%
7%
1%
18%
18%
8%
2%
Table 5. Mean Daily Depth of Intense Rainfall Days (mm) for Each Subarea and Each Weather Type
Subarea
Lefkada
Upper Acheloos
Lower Acheloos
Mornos-Evinos
Sperchios
Biotilcos Kifissos
Assopos
Athens
Evia
Slcyros
Total area
Number of events
Wl
13.9
16.4
32.8
25.1
12.2
4.2
0.9
0.6
0.7
0.6
12.4
18
W2
21.2
23.1
29.8
27.6
15.6
8.4
4.3
3.6
4.1
9.1
15.3
23
SW1
22.6
24.0
33.7
25.5
15.2
7.4
3.6
3.9
3.7
2.7
15.5
88
Weather Type
SW2
8.7
7.5
13.4
13.7
19.5
19.7
15.9
15.5
25.0
9.1
16.4
28
NW1
16.9
16.9
25.9
22.8
19.7
14.9
9.4
9.4
12.4
7.2
17.0
96
NW2
12.3
14.0
13.8
17.6
6.7
4.0
2.1
1.5
3.3
4.3
8.1
10
MT1,2
2.1
2.2
5.3
5.4
14.6
18.2
13.9
12.0
36.5
7.0
13.8
41
Rest
2.6
3.2
3.2
5.9
4.6
2.9
0.5
0.3
0.9
0.8
2.6
12
Table 6. Percentage of Variance of Areal Daily Intense
Rainfall Explained by Weather Types
Subarea Percentage, % Subarea Percentage, %
Lefkada
24.8
B.
Kifissos 21.8
Figure Captions
Figure 1. Study area: (a) morphology, (b) separated subareas
and rain gages (with the Krikello rainrecorder being the trian-
gle),
(c) general location.
Figure 2. Main trajectories of cyclonic weather types.
Figure 3. Map of the 500 hPa level of the event of July 7,
1970 06 00 UT.
Figure 4. Probability of occurrence of intense rainfall events
per group of weather types: (a) Groups of rainy period (A:
Wl,
W2,
NWl, SWl;
B:
SW2, NW2; C: MT2, DOR; D: Al-
A5,
MTl, DES, MB); (b) Groups of dry season (A: W2,
NWl, DOR; B: SWl, SW2; C: Wl, NW2, MTl, DES, MB;
D:
A1-A5.MT2).
Figure 5. Box plots of rainfall event characteristics at
Krikello: (a) duration, (b) total depth, and (c) mean intensity.
The middle line of
each
box represents the median, the bottom
and top lines represent the lower and the upper quartile, and
the whiskers represent the minimum and maximum observed
values.
Figure 6. Statistical surfaces of weather types. First column,
mean (mm); second column, standard deviation (mm); third
column, coefficient of variation.
Figure 7. Box plots of computed correlation coefficients
between rainfall fields per weather type. The middle line of
each box represents the median, the bottom and top lines
represent the lower and the upper quartile, and the whiskers
represent the minimum and maximum observed values.
Figure 1. Study area: (a) morphology, (b) separated subareas and rain gages (with the Krikello rainrecorder
being the triangle), (c) general location.
Figure 2. Main trajectories of cyclonic weather types.
Figure 3. Map of the 500 hPa level of
the
event of July 7, 1970 06 00 UT.
Figure 4. Probability of occurrence of intense rainfall events per group of weather types: (a) Groups of rainy
period (A: Wl, W2, NWl, SWl; B: SW2, NW2; C: MT2, DOR; D: A1-A5, MTl, DES, MB); (b) Groups of
dry season (A:
W2,
NWl, DOR; B: SWl, SW2; C: Wl, NW2, MTl, DES, MB; D: A1-A5, MT2).
Figure 5. Box plots of rainfall event characteristics at Krikello: (a) duration, (b) total depth, and (c) mean
intensity The middle line of
each
box represents the median the bottom and top lines represent the lower and
(a)
(b)
0
30 60 90 km
^
LL
PROBABILITY (%) PROBABILITY (%)
f
CD
15
100
80
Iso
1.
20--
(a)
y
---
J.
1
1
—-n
--u-
n
-
^
^
"____!
L____
T
--
Wl
Figure
5
Please reduce it
v\e
NWI
N\AC
SW1 SW2
REST
300
250
200
g-150
•a
3 100
50
(b)
B
T
T
1
"X"
W1 W2 NWI
NV\C SWI SW2 REST
20
I"
I
5
(c)
±:
^ ^
Wl V\C NWI NV\C SWI SV\C REST
W1
18 days
W2
23 days
NW1
96 days
NW2
10 days
SW1
88 days
SW2
28 days
MT1,2
41 days
16
1 -
0.75 -
S 0.5!
E 0.25
^
8
° A
o
3 -0.25
£
O "0-5
-0.75
-
—
T
r-l
L-
pr
1
1
hr.
£__(-•-
r
f-
L
1
Wl W2 NW1 NW2 SW1 SW2 MT2 REST
Weather types
•-
—
ALLTYPES
Figure 7
Please reduce it
