A weather type is a representative of the weathers that are similar to each
other in terms of discriminant atmospheric variables. If analyzing weather types
is done in daily extent, a weather type will include days that have similar
weathers. Occurrence of a weather type in a definite region depends on air masses
that entered the region on one hand and the result of influence of geographical
conditions (topography, adjacency with water masses and etc.,) of that region
on the other hand. Because geographical conditions of region are usually constant,
different weather types that come in and go away from a region are function
of air masses that enter the region. Thus, time series of weather types of a
region is related to the air masses that cover a large area including the subject
region. It is on the basis of this logic that in climatology literature, after
stationary analysis of weather types in temporal dimension, spatial analysis
of weather types will become important itself (Kalkstein
et al., 1996). Clearly, the concept of weather type is not a simple
topic to be considered along with other synoptic climatology categories; but
it is an important concept which is valuable as a theory. By this theory, new
concepts will be introduced and available climatology concepts and categories
will find new definitions and novel concepts.
Since, one of the main aims of synoptic climatology is identifying weather
types, climatology literature is rich in this aspect. Bissolli
et al. (2006), after studying weather types in Germany and comparing
it with tornado occurrence, concluded that there is a significant correlation
between frequency of stormy days and weather type. Morabitoa
et al. (2006) studied the relation between winter weather types of Florence,
Italy and myocardial infarction through synoptic method and they showed that
although because of methodological limitations they cannot generalize their
findings to other geographical regions; it seems that there is a statistical
relationship between weather types and myocardial infarction in Florence. Fowler
and Kilsby (2002) have studied relation of hydrologic droughts of Yorkshire
of Britain with Lamb weather types and they found that sometimes droughts are
results of atmospheric conditions and thus, they have relation with weather
types and sometimes they are results of mal-management. Littmann
(2000) has classified the data of pressure and geopotential height of 500
hpa using cluster analysis and has studied the relationship between resultant
weather types and Mediterranean Basin rainfall. He has identified 20 different
weather types and believes that these types explain Mediterranean rainfall patterns
to a high degree. Krichak et al. (2000) have
studied the relation of synoptic patterns and wet and dry conditions in the
Eastern Mediterranean. They have showed that anomalies of sea level pressure
and geopotential height of 500 hpa explain occurrence of wet and dry conditions
in the Eastern Mediterranean. Kalkstein et al. (1998)
have used analysis of air masses as a tool for studying climate change. They
believe that analysis of air mass frequency is a more useful method for explaining
of climate change than the trend analysis of climatic variables. Sheridan
(2002) has reviewed spatial synoptic classification from different aspects
and he believes that this method is a useful method for environmental studies.
He also found a strong relationship between weather types of North America and
teleconnection indices i.e. North Atlantic Oscillation (NAO) and Pacific Ocean-North
America (PNA); Sheridan (2003). Rainham
et al. (2005) paid attention to mortality in Toronto in Canada using
spatial synoptic classification. Their study showed that weather quality (weather
pollution) is in general a function of synoptic type but in order to understand
the relationship of atmospheric conditions and health, more studies are needed.
MATERIALS AND METHODS
Study area: Tehran, the capital city of Iran, is located in Tehran province. The city has more than 7 million people and it is the most populated city in the country. Figure 1 shows the location of Tehran. Geographical conditions of Tehran are stable approximately. Therefore, entrance of air masses causes occurrence of different weather types in Tehran. Present objective is to study these weather types. Data sources: the data of 22 variables from January 1, 1978 to December 31,
2004 are provided in a matrix in p mode (atmospheric variables in columns and
days in rows) (Fig. 2).
The atmospheric variables should be used to identify weather types that are
representative of temperature and humidity conditions. Among different variables
that are measured in synoptic station of Tehran, 22 variables were selected
that they are listed in Table 1. The data of these 22 variables
(1978 to 2004) are provided in a matrix in p mode (atmospheric variables on
columns and days on rows) (Fig. 2). So, data matrix has the
size of 9862x22. Since, the data have different units (Celsius degree, millimeter,
percent, degree, Knot), it is necessary to standardize it before doing any analysis,
so that weight of all variables in separating synoptic types could be the same.
Of course before standardization, the speed and direction of wind convert to
u-v wind components to prevent problem of wind speed and direction standardization.
The following equation has been used to standardize the variables (Johnson
and Wichern, 2001).
where, STNDij is standardized amount of the jth variable in the
ith day, Dataij is the jth variable in the ith day, Minjis the minimum value of the jth variable and Maxj is the maximum value of jth variable.
|| Location of Tehran in Iran that highlighted by black color
|| Sample of first and last rows of data matrix
|| List of basic data for identifying synoptic types of Tehran
After standardization, rows with a gap (even for one variable), were omitted
from matrix and so, the final matrix of size 9823×22 was obtained (stnd9823×22).
This matrix was the basis of calculation of dissimilarities using Euclidean
Before classification, we do not have any idea about the number of weather types, so cluster analysis for identifying weather types seems practical. In this case, for example k variables belongs to one day (t1), is compared with k variables belongs to another day (t2), in order to find degree of dissimilarity. Then, all days were clustered according to degree of dissimilarity. Therefore, there are two important steps in cluster analysis: first step is calculating the degree of dissimilarity of variables and the second step is how to link variables according to their degree of dissimilarity.
Depending on the methods that we choose for calculating the degree of dissimilarity
of variables and linkage of them, a cluster analysis can be implemented in different
ways. In this study, a hierarchical cluster analysis with ward linkage method
was used for identifying weather types. In order to calculate degree of dissimilarity,
Euclidean distance have been used and for linking items that show the highest
dissimilarity, the ward linkage method was used. It is clear that for n observations
n(n-1)/2 distance are considerable. Suppose xr is observation vector
on r and xs is observation vector on s; so, Euclidean distance is
calculated using the following equation (Johnson and Wichern,
In ward method, s and r are linked if the variance increase due to agglomeration
is minimal in comparison with combining them with any other group (Johnson
and Wichern, 2001), that is:
the distance between group r and group s that is obtained from centroid linkage
method. In climatological studies, ward linkage method primarily is used; because
in this case, inter group variance will be minimized and homogeneity of resulted groups will be maximized. After clustering analysis based on ward linkage method, the weather classification carried out according to characteristics of temperature, wind direction and speed, humidity, precipitation in each cluster.
RESULTS AND DISCUSSION
The study of air masses is important for forecasting air and information about
climate circumstances. The air masses determine daily humidity and thermal circumstances.
Repetition of these circumstances in long time results in climate formation.
The air masses that enter to a region are depending on general circulation.
Therefore different regions according to geographic situation have different
air masses types. The 27 individual weather types for Scotland, classified at
a daily resolution by Mayes (1991), were categorized
into five groups using cluster analysis. Schwartz (1991)
classifies weather types over the North-central USA and six weather types are
Tehran weather types: Applying a cluster analysis on standardized matrix and linking days using ward method, showed that Tehran has 4 distinctive weather types (Fig. 3, Table 2). The nominate of each weather types carried out according to characteristics of temperature, wind (direction and speed), humidity and precipitation in each cluster.
Cold, frosty, rainy type: This type is active for 17.3% of year, from
15th October to 15th April (Fig. 4, Table 3).
Temperature of day is almost 8.1°C and at night is about 1.4°C. In 39.6%
of time that this type occurs, there is frost. This type is very important because
in 44.6% accompany with precipitation and daily precipitation mean approximately
6/1 mm. During recent decades decrease frequency of this type (Fig.
8). In 29% of cases that this type is seen; there is morning fog (Table
3). But after sunrise, relative humidity decreases considerably and the
weather is dry.
Moderate type: Tehran weather at the end of summer and early autumn on one hand and at the end of winter to nearly late spring on the other hand can be moderate (Fig. 5). This type is seen from 15th September to the beginning of December and from 15th February to 15th June (Fig. 5). In the time of this type dominance that is observed in 25.4% of years, Tehran has moderate days. During this period temperature varies from 14.6° to 24.6°C. In the dominance time of this type almost all atmospheric variables are closer to the total mean. During the past decades, frequency of this type has had a very good stability (Fig. 9).
Warm, dry type: This type is the most dominant observable weather type
Tehran. This type is seen from 20th April to 20th of November (Fig.
|| Tehran weather types
|| Dendrogram of 4 weather types of Tehran
|| Monthly frequency percent of weather type No. 1
|| Characteristics of 4 weather types of Tehran
In the dominance time of this type, it has never been seen a weather except
hot and dry weather in Tehran (Fig. 6). Temperature varies
between 22.6° to 34.3°C during day and night. Relative humidity minimizes
(Table 3). Frequency of this weather type relatively has increased
during the last decades in Tehran (Fig. 10).
Cold, windy weather type: In autumn and spring, from 15th October 15th May there is cold weather and wind in Tehran (Fig. 7). In dominance time of this type that its frequency is about 20.4%, temperature is varying between 4.3° and 13.6°C during day and night and by passing the day wind speed increases (Table 3). It seems that during past decades frequency of this type has been reduced (Fig. 11).
Daily variety of weather types: As, we saw in explaining monthly frequency
occurrence of weather types, each weather type tends to be active in certain
months. In other words, weather types have seasonal behavior. Because of this,
some of the types of weather disagree with some others and have much harmony
with the rest. But, there are some types that can appear after each type of
weather and may connect disagreeing patterns. These two different kinds of behavior
can mean that in some days of year, we should expect a certain weather type
on one hand and it means that in some days there can be many weather types on
the other hand. Measuring degree of weather type variety on each day of year,
beside theoretical value, can also help forecasting from practical view. So,
we attempt to introduce a scale for measuring degree of daily variety of weather
||Monthly frequency percent of weather type No. 2
||Monthly frequency percent of weather type No. 3
||Monthly frequency percent of weather type No. 4
||Annual frequency percent of weather type No. 1
||Annual frequency percent of weather type No. 2
||Annual frequency percent of weather type No. 3
||Annual frequency percent of weather type No. 4
||Occurrence variety percent of weather types in each day of
In this study, because the base of classification of weather types has
been daily data, we determine degree of weather type variety for daily comparison. Assume that we can attribute a certain weather type to each of days of a year
and assume that there are m different weather types. In addition, assume that
we are studying N different years. If nij is frequency of ith weather
type in jth day, then it can be written (Massodian, 2007):
where, nij is frequency percentage of the ith weather type in the
jth day. In this case, we can determine the degree of each of the types of weather
variety in the following way (Massodian, 2007):
In our problem, number of weather Types is four (m = 4) and the number of considered years is twenty seven (N = 27). So, Di can obtain degree of weather type variety for ith day, i for ordinary years varies between 1 to 365 and for leap years varies between 1 to 366. For a day that in all considered years had same weather type, D is equal to zero. In other words, this certain day just accepts one weather type and it does not have weather type variety. In opposite, for a day that has accepted all weather types and frequency percent of occurrence for all weather types has been the same on it; D will be one: that is, such a day has maximum of weather type variety. Therefore, D varies between zero and one. Zero introduces complete monotony and one introduces complete variety of occurrence weather types in a certain day. In warm period of the year, Tehran is under dominance Azores subtropical dynamic high pressure. In the dominance time of this system, warm, dry weather type and hot, dry weather type occurs in Tehran; because of this, weather variety in warm period of the year is little. In cold period of the year that by progressing westerly winds to South latitude, Tehran rests in direction of westerly winds waves; dependent on station location relative to waves axis, more spread spectrum of weather types can be seen in Tehran (Fig. 12).
Sequentiality: One of the important characteristics of weather types is their sequentiality condition. Sequentiality means the number of times that a weather type can be seen after itself or after another weather type. By studying sequentiality we can identify disagreeable weather types and subsequent weather types. We call weather types i and j disagreeable; when after observing weather type i, weather type j never be observed. Studying sequentiality of weather type number 1 makes it clear that this type is disagreeable with weather types 2 and 3. On the other hand, the most probable weather type after observing weather type number 1 is the weather type number 4. The other important characteristic that can be determined by sequentiality counting of weather types is stability of each weather type. Obviously, the probability of observing a weather type after occurrence the same type is more because similar weather types tend to appear after each other. Even some of weather types that are middle limit completely opposite weather types have transition duty.
If occurrence frequency of each type after another weather type (sequentiality)
is presented on percentage basis, we can obtain a criterion for integration
of weather types. For example, assume that weather type number 4 has been seen
during 3642 days and in all cases except one time (in the last day) has been
repeated after itself; in this case, this pattern has been occurred during the
27 year period one time and for 3642 days and it has complete integration; while
it could for example, occur 244 times and each time almost for 15 days that
in this case. It has had less integration. This pattern has been repeated in
95% cases after itself (Table 4). This characteristic can
be expressed in a precise way. If N is the number of days that a weather type
has been observed and n is times of observing that weather type, then we can
express occurrence index as follows (Massodian, 2007):
In the case of completely separated weather types, this index will be zero
and for completely continual pattern it will be 1-1/N. Calculation of the occurrence
index of the types of weather of Tehran (OI) shows that the occurrence index
of the 4 types of weather that have occurred 75, 80, 93 and 71%, respectively.
|| Sequentiality percent of Tehran weather types
|| Occurrence characteristics of Tehran weather type
These types have had almost 4, 5, 15 and 3/5 day durability each time and their
maximum permanence has been 43, 44, 135 and 31 days, respectively (Table
In this study cluster analysis and linking days on the basis of ward method on the standardized matrix of data are employed in order to recognition Tehran weather types. The results showed that Tehran has 4 weather types (WT) including: 1) Cold, frosty, rainy WT, 2) Moderate WT, 3) Warm, dry WT and 4) Cold, windy WT. Tehran has a cold, frosty and rainy winters (weather type 1) and it has warm and dry summers (weather type 3). Only in 25.4% days of a year, moderate climate can be seen (weather type 2). In warm and dry period that involves almost 40% of days a year, because of dominance of Azores subtropical high pressure, Tehran climate has much stability and just weather type 3 occur. The dominant of warm and dry weather types in most days of year rather than other weather types and increase of relative frequency of this weather types in recent years shows that Tehran has a dry climate and become warmer The characteristics and annual frequency percent of weather type 1 is testimony to this matter. By increasing temperature and dislocation snowline to higher heights, in fact necessary water resources for continuing lives may be at risk. For further studies the recognition circulation patterns that result in these weather types may be very useful for better planning and management resources that effected by circumstances atmospheric.