Introduction
Climate is one of the most important limiting factors
in agricultural production. In recent years, a change in
climate has been documented in many locations
throughout the world. Global climate models coupling the
atmosphere and the oceans replicate certain climate
changes well, in particular, changes in temperature
observed over the last century, as well as the last few
decades, if the following observed variants are
incorporated (Santer et al., 1996): greenhouse gas
concentrations, stratospheric ozone, aerosols in the
troposphere, aerosols in the stratosphere, and solar
activity (sunspots). The first 3 of these are largely due to
human activities; the last 2 are natural. The application of
statistical tests to temperature measurements during this
century leads to the conclusion that the warming is manmade, with a confidence of 99% (Toll, 1994). The global
mean temperature trend between 1975 and 1997, and
even regional variations in this trend, can be explained
very well by observed changes in greenhouse-gas
concentrations (Hunt, 1998). In other words,
greenhouse gas emissions are by far the leading cause of
climate change during the last quarter of the 20th
century. The last decade of the 20th century (especially
the years 1990, 1993, and 1997) is almost certainly the
warmest since 1400 (Hunt, 1998). The minimum
temperature increased almost everywhere, and the
maximum and mean temperature increased in northern
and central Europe, across the Russian Federation and
Canada (Bootsma, 1994), and in Australia and New
Zealand (Plummer et al., 1999).
Atmospheric models have predicted that our planet
will be influenced by climate change (Toll, 1994). It is
widely accepted today that every change in the climatic
system is important for water and natural resources
management. As climate changes and temperature
increases, the hydrologic cycle changes accordingly. As a
result, there would be an increase in rainfall intensities
Turk J Agric For
30 (2006) 439-448
© TÜB‹TAK
439
Time Trend in the Mean Annual Temperature of Iran
Bijan GHAHRAMAN
Irrigation Department, College of Agriculture, Ferdowsi University of Mashhad, Mashhad 91775, IRAN
Received: 03.01.2006
Abstract: Many researchers around the world have reported a gradual increase in mean annual temperature. Yet, there are some
reports of a reduction in this parameter. In this study, we investigated the long-term trend of mean annual temperature at 34
synoptic stations in Iran (2 stations in cool humid climates, 14 stations in temperate humid climates, 11 stations in steppe climates,
and 7 stations in desert climates, based on the Koppen climatic division) with a minimum record of 30 years by Student’s t-test.
Results showed that there was a positive trend in 50% of the stations, while 41% of stations had a negative trend. Considering the
significance level (α = 5%), there were 3 trend zones for mean annual temperature in Iran, i.e. positive trend, negative trend, and
zero trend; however, it was difficult to define a specific spatial scheme for such a division. The results were supported by the MannKendall method, while low harmony was found with the Wald-Wolfowitz test. As far as record length is concerned, during a common
time period (1968-1998), 65% of the stations showed a positive trend, while 32% of them followed a negative trend. There were
some shifts from one trend to another for some of the stations in the study, yet with no well-defined spatial structure. In this case,
and at the 5% level of significance, 44%, 15%, and 41% of the stations had a positive trend, a negative trend, and zero trend for
the parameter of the study, respectively. In general, the behavior of trend direction was different for different climates and no
specific pattern was found. Based on the results, one may hypothesize that in the future more regions will experience higher
temperatures. Some stations did not show any significant trend, yet their positive trends may be indicative of future warming.
Including the years 1999-2002 in the data verified the results of our trend analysis. All of the stations showed higher average mean
annual temperature compared to the average of the period 1968-1998.
Key Words: Temperature, climate, time trend, Iran
* Correspondence to: bijangh@ferdowsi.um.ac.ir
and runoff, in addition to increasing soil moisture
demands. Meanwhile, the universal hydrologic cycle
would be influenced by changes that are man-made
(Lettenmaier et al., 1994; Karl et al., 1996; Vorosmarty
et al., 2000). It was shown in Slovakia that during a 94-
year period (1901-1994), air temperature rose by 0.8 °C
(Lapin, 1995). Hess (1998) reported that monthly mean
daily maximum, and in particular, minimum temperatures
of 3 stations in the northeast arid zone of Nigeria
experienced an increase of up to 1.5 °C during the period
1961-1991. It is also reported that from 1982 to 1998,
Maryland experienced not only an increase of 0.43
°C/decade for January and July temperature, but also its
mean annual maximum and minimum temperature
decreased by 0.6 °C/decade (Menglin and Dickinson,
2002).
In contrast to the following reports on the positive
trend of temperature, the literature reports the
possibility of temperature decrease. Perez et al. (2000)
utilized temperature data for 41 years (1958-1998)
from the National Centers for Environmental
Prediction–National Center for Atmospheric Research
(NCEP–NCAR) for reanalysis. They showed that in the
northern hemisphere a positive trend in temperature
encompassed most of Europe, North America, and the
Atlantic Ocean in the 30-50°N band, while there was a
negative trend over Iceland, Greenland, and the
Northeastern coast of Canada, including Hudson Bay. Box
(2002) analyzed temporal and spatial variability from
Greenland instrumental temperature records for 24
coastal and 3 ice sheet locations. Trends over the longest
period available, 1873-2001, at Ilulissat/Jakobshavn
indicate statistically significant spring and summer
cooling. Such a cooling is also reported by Nicholls et al.
(1996) for the northern and northwestern Atlantic
Ocean, and middle latitudes of the north Pacific Ocean
between 1951 and 1999. With the Illinois State Water
Survey, Angel (2004) showed that several Illinois stations
presented no warming during the last few decades. Yet,
some Illinois stations, particularly in the southern part of
the state, followed a cooling trend. Srivastava et al.
(1992) observed the increasing trends of annual mean,
maximum, and minimum temperatures south of 23°N,
and cooling trends north of 23°N. There was no sign of
mean annual temperature increase for 122 years (1878-
1999) as recorded by the Pisa meteorological station in
Italy (Moonen et al., 2002).
As the mean annual temperature trend varies by
location and time period worldwide, and since there is no
report on temperature trend analysis in Iran, this study
aimed to fill the gap.
Materials and Methods
Trend definition and its calculation
Any time series of a stochastic variable consists of 2
stochastic and deterministic components. Deterministic
components are in the form of trend, cyclicity, or jump
(Salas, 1992). Generally, there isn’t any cyclicity of the
annual scale. In addition, there is no report for annual
temperature series. Therefore, this paper is focused on
trend analysis of the deterministic components of mean
annual temperature of synoptic stations in Iran.
There are different statistical methods considering
trend analysis (Haan, 1977; Bobee and Ashkar, 1991;
Salas, 1992). Student’s t-test is a common method for
trend analysis of climatic parameters (e.g.,
Chattopadhyay and Hulme, 1997). In this method, a
regression line establishes differences between
independent (time; x) and dependent (temperature; y)
variables. Intercept and line slope can be calculated
through error minimization. Afterwards t statistics
(=b/sb, where b is the intercept of the regression line and
s
b is the standard deviation of the data, Eq. (1)) is
computed.
S
2b
= S2 / Σ (Xi – X – )2 (1)
where S2 can be determined through Eq. (2).
S2 = Σ (Yi – Y ^i)2 / (n – 2) (2)
The null and alternate hypothesis are defined as:
H
0: b = 0, H1: b ≠ 0 (3)
The H
o hypothesis at α level of significance is rejected
whenever | t | > t
1 – α
2
, n–2 from the t-student table (n =
sample size). As a result, the slope of the trend line is
significantly different from zero; there is a trend in time
series.
Time Trend in the Mean Annual Temperature of Iran
440
Station used
Record length is important in statistical investigations.
It is well known that as record length increases the
validity of the results increases accordingly. There are
160 synoptic stations in Iran. Record lengths from these
stations vary between 10 and 48 years (up to 1998, for
which we could obtain the data). It is not wise to proceed
with low record lengths. Therefore, a minimum record
length of 30 years was considered. This length is
appreciated in climatological studies. With this limiting
criterion, 34 stations were available. These stations were
located in different climatic conditions. Based on Koppen
climatic divisions, there were 2 stations (6%) in cool
humid (D), 14 stations (41%) in temperate humid (C), 11
stations (32%) in steppe (Bs), and 7 stations (21%) in
desert (Bw) climates. See Table 1 for geographical
specifications and Figure 1 for spatial distribution of the
selected stations.
Trend computations
Trend analysis, or significance test for line slope, was
performed in 3 stages: (a) for the entire time period for
all stations (up to 1998), (b) for a common period for all
stations (1968-1998), and (c) consecutive time periods,
in which in every step, the oldest data was dropped and
trend analysis was performed accordingly, the analysis
was stopped for a 5-year record length.
Results and Discussion
1. Entire time period
Table 2 shows the slopes of the long-term trend lines
for the 34 synoptic stations studied. Of the 34 stations,
17 (50%) showed a positive trend. On the other hand, a
negative trend was demonstrated by 14 stations (41%).
The remaining 3 stations (9%) showed a zero trend. Yet,
this ratio was not the same for the 4 different climates.
B. GHAHRAMAN
441
Table 1. Geographical specifications (degrees for longitude and latitude, and meters for altitude) for the selected stations in Iran.
(1)* (2) (3) (4) (5) (1) (2) (3) (4) (5)
Desert climate (Bw) Temperate humid (C)
Bw01 Abadan 4825 303611 C01 Arak 4970 3410 1720
Bw02 Ahwaz 4866 3133 22 C02 Babulsar 5265 3671 21
Bw03 Bandar Abbas 5636 2721 10 C03 Gorgan 5446 3681 155
Bw04 Bushehr 5085 2895 8 C04 Hamedan 4853 3485 1749
Bw05 Kerman 5696 3025 1754 C05 Kermanshah 4711 3426 1322
Bw06Yazd 5440 3190 1230 C06 Khoram Abad 4830 3350 1125
Bw07 Zabol 6148 3133 489 C07 Mashhad 5963 3626 980
Steppe (Bs) C08 Orumieh 4508 3753 1312
Bs01 Bam 5840 2910 1067 C09 Ghazvin 5000 3625 1278
Bs02 Birjand 5920 3286 1491 C10 Ramsar 5066 3690 20
Bs03 Fasa 5368 2896 1383 C11 Rasht 4960 3725 7
Bs04 Isfahan 5166 3261 1590 C12 Shiraz 5258 2953 1491
Bs05 Kashan 5145 3398 982 C13 Tabriz 4628 3808 1361
Bs06 Sabzevar 5766 3621 941 C14 Zanjan 4848 3668 1663
Bs07 Semnan 5338 3555 1171 Cool humid (D)
Bs08 Shahrud 5503 3641 1345 D01 Sanandaj 4700 3533 1373
Bs09 Tehran 5135 3568 1191 D02 Shahre-Kord 5085 3231 2078
Bs10 TorbatHeydarieh 5921 35261333
Bs11 Zahedan 6088 2946 1370
* (1) Code, (2) Name, (3) Longitude, (4) Latitude, (5) Altitude
Time Trend in the Mean Annual Temperature of Iran
442
Steppe
Desert
Cool Humid
Temperate
Humid
N S
30 0 30 60 90 120 150 Kilometers
Figure 1. Location of selected stations in different climatic regions of Iran.
Table 2. Slope of mean annual temperature line for synoptic stations of Iran. x denotes the level
of significance at α = 5%.
Code Whole Common Code Whole Common
period period period period
Desert climate (Bw) Temperate humid (C)
Bw01 0.01 0.02x C01 –0.03x –0.02
Bw02 0.03x 0.06x C02 0.02x 0.03x
Bw03 –0.03x –0.02 C03 –0.01 –0.01
Bw04 0.02x 0.04x C04 –0.02 0.02
Bw05 –0.02 0.03x C05 0.02x 0.05x
Bw060.02x 0.04x C06–0.05x –0.07x
Bw07 0.00 0.00 C07 0.03x 0.08x
Steppe (Bs) C08 –0.04x 0.01
Bs01 0.3x 0.3x C09 –0.03x –0.01
Bs02 –0.03x –0.03x C10 0.01 0.02
Bs03 –0.04x –0.05x C11 0.02 0.05x
Bs04 0.02x 0.02 C12 0.03x 0.06x
Bs05 –0.03 –0.03x C13 0.02x 0.04x
Bs060.06x 0.07x C14 –0.03x –0.02
Bs07 0.01 0.02 Cool humid (D)
Bs08 0.01 0.05x D01 0.00 0.01
Bs09 0.03x 0.05x D02 –0.03 –0.04x
Bs10 –0.01 –0.02
Bs11 0.00 0.01
We summarized the results in Table 3. Based on Table 3,
desert climate had the highest positive trend (57%),
while the highest negative trend was in humid climates
(temperate and cool).
One may obtain different results if the significance of
the slopes is considered. Based on all stations, 14 stations
(41.2%) showed no significant slope (α = 5%), while
significant slopes were found for the other 20 stations
(58.8%), of which 12 (35.3%) and 8 (23.5%) stations
showed positive and negative trends, respectively (Table 4).
Previous studies have principally confirmed the positive
trend of mean annual temperature in most parts of the
world (e.g., Hess, 1998, for Nigeria). In addition, there
are some forecasts for temperature increase in the future
(Zhang et al., 2000). Yet some reports indicate local
temperature decrease. Microclimatic conditions may be the
cause. The results of this study on the possibility that there
may exist a region with a non-uni-trend in mean annual
temperature are supported by other studies. Angel (2004)
not only showed that some parts of Illinois did not
demonstrate any signs of warming, but also that
temperature cooling was recorded in some parts of the
state. The writer did not report the reason, but stated,
“possible candidates include, but are not limited to,
changes in the distribution and amount of sulfates (small
airborne particles that result from the burning of fossil
fuels, especially coal), the shift in land use from prairie and
forest to agriculture, increases in the amount of cloud
cover, and changes in the sea surface temperatures of the
Pacific and Atlantic”. Chattopadhyay and Hulme (1997)
indicated some signs of cooling in India. The writers
analyzed temperature trends of 27 stations for 4 seasons,
winter, monsoon, pre-monsoon, and post-monsoon, from
1940-1990. Three zones of cooling may be distinguished
from their Figure 1 in all seasons. The writers did not
explain this cooling, either on a seasonal scale or an annual
scale. Elagib and Mansell (2000) reported that 2 out of 13
stations under study in Sudan (1941-1996) had a negative
temperature trend. Greenland and Kittel (2002) showed a
negative trend for 4 stations (out of 18) around the USA
from 1957 to 1990. However, as statistical methods are
not crucial, the results may change as a consequence of
changes in the methods used.
B. GHAHRAMAN
443
Table 3. Numbers and percentages of Iranian synoptic stations with
definite trends of mean annual temperature in different
climactic regions.
Trend Entire period Common period
Climate direction
No. % No. %
Desert (7*) positive 4 57 5 72
negative 2 29 1 14
zero 1 14 1 14
Steppe (11) positive 655 7 64
negative 4 364 36
zero 1 9
Temperate positive 7 50 9 64
humid (14) negative 7 50 5 36
zero 0 0
Cool positive 0 0 1 50
humid (2) negative 1 50 1 50
zero 1 50
Total number of positive 17 50 22 65
stations (34) negative 14 41 11 32
zero 3 9 1 3
* number of stations
Table 4. Numbers and percentages of Iranian synoptic stations with
definite trends of mean annual temperature under different
climates (α = 5%).
Trend Entire period Common period
Climate direction
No. % No. %
Desert (7*) positive 3 42.9 5 14.3
negative 1 14.3 0 0.0
zero 3 42.9 2 28.6
Steppe (11) positive 4 36.4 4 36.4
negative 2 18.2 3 27.3
zero 5 45.4 4 36.4
Temperate positive 5 35.7 642.5
humid (14) negative 5 35.7 1 7.1
zero 4 28.67 50.0
Cool positive 0 0.0 0 0.0
humid (2) negative 0 0.0 1 50.0
zero 2 100 1 50.0
Total number of positive 12 35.3 15 44.1
stations (34) negative 8 23.5 5 14.7
zero 14 41.2 14 41.2
* number of stations
The detailed results of significant temperature trend
for different Iranian climates are presented in Table 4.
Desert climate showed the maximum significant positive
trend (42.9%), while the minimum trend (0%) was in
cool humid climate. On the other hand, the maximum
(35.7%) significant negative trend was observed in
temperate humid climate and 0% of the stations had a
minimum significant negative trend in cool humid climate.
Both humid climates (cool and temperate) possessed the
maximum (100%) and minimum (28.6%) of stations
with zero trends. Overall, 35%, 24%, and 41% of the
stations showed positive, negative, and zero trends,
respectively. Karim-Zadeh and Ghahraman (2002)
support this study with the Mashhad temperature trend
(1970-2000).
To check whether the trend was influenced by one or
more outliers, we counted the number of events higher
and lower than the mean annual temperature. The
difference between these 2 numbers may be considered
as criteria denoting the outlier situation. This difference
was divided by the record length of its station, as record
length differed for each station. Our results showed that
this criterion was only 6.5%; the maximum deviation
(9.5%) was due to cool humid climate. Therefore, one
may conclude that the trend was not affected by some
outlier data; however, the results had more fluctuations
during different decades (data not shown). Low record
length (10 years) may have been a cause for such
fluctuations
Figure 2 shows the spatial significant trends for mean
annual temperature of Iran for the entire record length.
It seems that there exists no geographical pattern of
different trend directions throughout the country.
2. Common period
We chose a common period of 31 years (1968-1998)
for all stations studied. Comparing the results with
previous cases showed that there are some changes in
trend direction (Table 2). In this case, 22 stations (65%)
had a positive trend, 11 stations (32%) had a negative
trend, while a zero trend was recorded from only 1
station (Bw07) (3%). Trend directions for 3 stations
(Bw05 in the southeast, and C04 and C08 in the
northwest) shifted from negative to positive, as
compared with the previous case. One possible reason for
the shift of these 3 stations is that all of them experienced
a cool period followed by a warm one (Table 5).
Therefore, excluding the warm period, there would be a
greater chance for a positive trend direction.
The literature supports such a change in trend
direction. Box (2002), in Greenland, showed that general
periods of warming occurred from 1885 to 1947 and
1984 to 2001, and cooling occurred from 1955 to
1984. Greenland and Kittel (2002) showed that the
Time Trend in the Mean Annual Temperature of Iran
444
Positive Trend
Negative Trend
Zero Trend
N S
30 0 30 60 90 120 150 Kilometers
Figure 2. Spatial pattern of mean annual temperature in Iran for total record length.
mean annual temperature trend for the Jornada station in
New Mexico (subtropical desert climate) and for the
Luquillo Exp. Forest station located in Puerto Rico
(tropical rainforest climate) from 1957-1990 was
positive and negative, respectively. However, increasing
the record length of these stations to 76 and 62 years,
respectively, reversed their trend directions.
Changing the record length to the common period of
1968-1998 provided some significant changes in trend in
the present study. A comparison of this common period
and the entire time period is presented in Table 4. Based
on the results, a significant trend was present in 20
stations (58.8%), of which 15 (44.1%) and 5 (14.7%)
had positive and negative significant trends, respectively.
There were no trends in the other stations. Moreover, the
number of stations with significant trends diminished.
There were also some shifts in trend directions, e.g., 5
stations without a trend changed to a significant trend (4
positive, and 1 negative).
In this case, humid climates (temperate and cool) had
the maximum (42.5%) and minimum (0%) percentages
of positive trend, respectively. The results for significant
negative trend were 50% for maximum and 0% for
minimum, respectively, in cool humid and desert climates.
On the other hand, the 2 humid climates (cool and
temperate) had the maximum (50%) and desert climate
had the minimum (28.6%) of zero trends (Table 4).
The geographical pattern for trend direction in Iran
during the common period is presented in Figure 3. One
may compare Figure 1 with Figure 2 to verify the effect
of record length on trend analysis. Despite different views
of these Figures, the results shows that trend direction
for only 5 stations (15%) changed. These stations are
staggered throughout the country. Yet, a positive trend is
common among these stations.
3. Trend for different record lengths
We analyzed trend after increasingly shortening the
record length of the data (variable starting year, and fixed
end year, i.e. 1998). There was a fluctuation in trend
B. GHAHRAMAN
445
Table 5. Mean temperature for 3 stations corresponding to 2 different
time series.
Mean temperature for:
Station up to 1968 1968-1998
Bw05 16.04 15.42
C04 11.5610.61
C08 12.41 10.86
Positive Trend
Negative Trend
Zero Trend
N S
30 0 30 60 90 120 150 Kilometers
Figure 3. Spatial pattern of mean annual temperature in Iran for the common period of 1968-
1998.
direction for consecutively shortened record length (Table
6). Not all stations offered unique responses to record
length shortening. Some stations maintained a unidirection trend. In such stations (22) there was only a
positive significant trend, yet their start year was not the
same; Bandar Abbas (Bw03), in the south, was an
exception. Hamedan (C04), in the west, did not show any
significant trend throughout the analysis, yet it had a
positive and negative trend for the entire and common
record lengths, respectively (Table 2). The other 11
stations fluctuated, of which 10 stations with an initial
negative trend changed to positive; Isfahan (Bs04), in the
center of the country, was an exception. This leads to a
major conclusion that 31 out of 34 stations experienced
a positive temperature trend, either initially or some
years later.
The reason for trend direction fluctuation, though not
significance, for different record lengths, may have been
annual cyclic variations of temperature. One may compare
the sinusoidal cycle through a 5-year moving run average.
This moving run average makes cycles, which are above
or below the long-term mean; based on this, the trend
line is augmented by cycle and a negative or positive trend
will be the result.
4. Sensitivity to the method used
In addition to Student’s t-test, which is the main trust
of this paper, we used 2 other methods; Wald-Wolfowitz
(Bobee and Ashkar, 1991) and Mann-Kendall (Salas,
1992). There was virtually complete harmony between
Student’s t-test and Mann-Kendall; for the entire record
length, only 3 stations were different. However,
Student’s t-test showed a significant negative trend, while
the Mann-Kendall method showed only a negative trend.
On the other hand, during the common period, 2 stations
were different. As neither station showed a nonsignificant trend under either method, the difference may
not be regarded. Skewness coefficients of all stations for
both overall and common time periods were computed.
Based on the skewness test of normality for sample sizes
less than 150 (Salas et al., 1980), all the stations, except
Time Trend in the Mean Annual Temperature of Iran
446
Table 6. First year for significant (α = 5%) mean annual temperature trend slope for synoptic stations
of Iran.
first year with first year with
start of significant trend: start of significant trend:
record record
code length positive negative code length positive negative
Desert climate (Bw) Temperate humid (C)
Bw01 1951 1967 ---- C01 1958 1989 1958
Bw02 1957 1957 ---- C02 1951 1951 ----
Bw03 1957 ---- 1957 C03 19561991 ----
Bw04 1951 1951 ---- C04 1955 ---- ----
Bw05 1951 1978 1955 C05 1951 1951 ----
Bw061953 1953 ---- C061951 1981 1951
Bw07 1963 1991 ---- C07 1951 1951 ----
Steppe (Bs) C08 1953 1971 1953
Bs01 1957 1957 ---- C09 1959 1973 1959
Bs02 19561981 1956 C10 1965 1972 ----
Bs03 1967 1992 1967 C11 1956 1956 ----
Bs04 1951 1951 1987 C12 1951 1951 ----
Bs05 1967 1990 ---- C13 1951 1951 ----
Bs061955 1955 ---- C14 1955 1989 1955
Bs07 1966 1972 ---- Cool humid (D)
Bs08 1954 1961 ---- D01 1960 1971 ----
Bs09 1951 1951 ---- D02 1960 1992 1960
Bs10 1959 1991 1973
Bs11 1951 1955 ----
for Bw05, passed the test. Therefore, for normal series,
there is some support in the literature (Yue et al., 2002;
Önöz and Bayazit, 2003) that both parametric t-test and
non-parametric Mann-Kendall tests perform identically.
There was not much similarity between Wald-Wolfowitz
and either Student’s t-test or Mann-Kendall. There were
59% and 68% similar results for the entire record length
and common period, respectively. There were a few
published documents dealing with the comparison of
different trend-detection tests. Rakhecha (Rakhecha, P.R.
Secular changes in the extreme rainfalls in India. 6 p.
balwois.mpl.ird.fr/balwois/administration/ full_paper/ffp-
527.pdf) investigated secular changes in the annual extreme
rainfall series of 1 to 3 days duration at 316 stations in
India during the period of 1901 to 1980. He used 5 tests,
including Mann-Kendall and Wald-Wolfowitz. The number
of stations showing a significant trend at the 95%
confidence level was completely different (42 for MannKendall and only 15 for Wald-Wolfowitz, for 1-day rainfall
time series). Though Rakhecha did not explain the probable
reason, this finding is in agreement with our trend analysis
of temperature in Iran. More research is needed to delineate
the cause of such differences.
5. Validity of the results with more recent data
For the period of 1999-2002, temperature data are
now available from the Iranian Meteorological Society.
Averages of annual temperature for these years were
greater than the average of the 31-year common period for
all the stations we studied. While this result confirms our
previous positive trend, some stations did possess negative
trends. This is not a controversy, however. Slope of the line
trend only determines the deterministic component of the
variable under study. Yet the stochastic component may
cause such undulations over the trend line, which is not
studied here. Rate change of the 31-year average to the
recent 4-year period was a minimum of 0.4% for Bushehr
(Bw04), in desert climate, to a maximum of 18.57% for
Hamedan (C04), in temperate humid climate, with an
overall average of around 7.16% for all stations.
Discussion and Conclusion
Mean annual temperatures in recent years were
greater than those in previous years for many stations.
This finding is consistent with the literature (e.g., KarimZadeh and Ghahraman, 2002, for Mashhad in Iran; Lapin,
1995, for the Danube in Slovakia). The majority of the
stations we studied revealed increasing temperature;
however, since the majority of the synoptic stations in
Iran are located in airports and suburbs, it is hard to
relate these findings to climate change and global
warming. Land use change in suburbs, from agriculture
and pasture to residential areas, is another cause for
gradual increase in temperature. Yet, it is not possible to
separate these 2 factors, while the synoptic stations in
Iran are used extensively for irrigation water requirement
computations. Land use change occurs slowly and,
therefore, cannot lead to a jump in temperature time
series. Meanwhile, constructing any instrument in a
region may not necessarily lead to a jump in temperature
(e.g., hydroelectric construction on the Danube River in
Slovakia, as reported by Lapin, 1995).
Two of the 3 methods used (Student’s t-test and MannKendall) verified each other. Based on the results of longrun mean annual temperature, it can be hypothesized that
the majority of the region will experience an increase in
temperature in the future. Although the mean annual
temperature trend lines in some stations were not
significant, their positive signs may be a clue to
temperature increase. This is especially true for the final
years of the record. The results of the Wald-Wolfowitz
method are relatively inconsistent with those of the other 2
methods. However, we could not find any preference of
one over the other in the literature. As 2 of the 3 methods
completely verified each other, it is possible to rely on the
results. Excluding the Wald-Wolfowitz method, for nonskewed series, Student’s t-test may be preferred over the
Mann-Kendal method. The latter method requires a
continuous data set, while the former performs equally well
for series with missing data. Such series are common in
undeveloped countries.
This study clearly confirmed that the results may be
sensitive to the initial and final years of the recorded data;
therefore, in any comparison, one must pay attention to
this point.
The announcement of temperature increase in the
future is merely important as a pre-information. Other
meteorological parameters, e.g., rainfall, must also be
analyzed for a comprehensive understanding of the global
B. GHAHRAMAN
447
situation. Furthermore, as far as agriculture is concerned,
study of mean seasonal temperature may bring about more
insights for future planning. The literature clearly supports
that trend slopes differ for different seasons; however,
temperature increase may act negatively and unidirectionally in the sense of increasing requirements for soil
moisture, decreasing water content in the upper layers of
soil, decreasing ground water levels, and discharge in rivers.
Time Trend in the Mean Annual Temperature of Iran
448
References
Angel, J., 2004. Climate Change and Variability in Illinois. Illinois State
Water Survey. URL: http://www.sws.uiuc.edu/atmos/statecli/
Climate_change/cc.htm.
Bootsma, A. 1994. Long-term (100 years) climate trends for
agriculture at selected locations in Canada. Clim. Change 26: 65-
88.
Bobee, B. and F. Ashkar. 1991. The Gamma family and derived
distribution applied in hydrology. Water Resources Publications,
USA.
Box, J.E. 2002. Survey of Greenland and instrumental temperature
records: 1873-2001. Int. J. Climatol. 22: 1829-1847.
Chattopadhyay, N. and M. Hulme. 1997. Evaporation and potential
evapotranspiration in India under conditions of recent and future
climate change. Agric. Forest Meteorol. 87: 55-73.
Elagib, N.A. and M.G. Mansell. 2000. Climate impacts of environmental
degradation in Sudan. GeoJournal 50: 311-327.
Greenland, D. and T.G.F. Kitte. 2002. Temporal variability of climate at
the US long-term ecological research (LTER) sites. Clim. Res. 19:
213-231.
Haan, C.T. 1977. Statistical Methods in Hydrology. The Iowa State
Univ. Press, Ames.
Hess, T.M. 1998. Trends in reference evapo-transpiration in North East
arid zone of Nigeria 1981-91. J. Arid. Envir. 38: 99-115.
Hunt, B.G. 1998. Natural climatic variability as an explanation for
historical climatic fluctuations. Clim. Change 38: 133-57.
Karim-Zadeh, M and B. Ghahraman. 2002. A statistical outlook to
gradual increase in Mashhad crop reference evapotranspiration
and its consequences. In: Proceedings of the First National
Conference on Water Crisis, Vol. 2, Zabol, Iran, pp. 95-108 (in
Persian).
Karl, T.R., R.W. Knight, D.R. Easterling and R.G. Guayle. 1996. Indices
of climate change for the United States. Bull. Am. Meteorol. Soc.
77: 279-292.
Lapin M., 1995. Climatological Monitoring of Territory Affected by
Construction of the Danube Hydroelectric Power Project and
Evaluation of Initial Impact. Gabcyvo Part of the Hydroelectric
Power Project - Environmental Impact Review. Faculty of Natural
Sciences, Comenius University, Bratislava, 15-22. URL:
http://www.mpsr.sk/slovak/dok/gn/book/03kap/03kap.htm.
Lettenmaier, D.P., E.F. Wood and J.R. Wallis 1994. Hydroclimatological trends in the continental United States. J. Climatol.
7: 586-607.
Menglin, J.R. and E. Dickinson. 2002. New observational evidence for
global warming from satellite. Geophys. Res. Let. 29: 39-42.
Moonen, A.C., L. Ercoli, M. Mariotti and A. Masoni. 2002. Climate
change in Italy indicated by agrometeorological indices over 122
years. Agric. Forest Meteorol. 111: 13-27.
Nicholls, N., G.V. Gruza, J. Jouzel, T.R. Karl, L.A. Ogallo and D.E.
Parker. 1996. Observed Climate Variability and Change. In:
Climate Change 1995: The Science of Climate Change. (Eds.: J.T.
Houghton, L.G. Meiro Filho, B.A. Callander, N. Harris, A.
Kattenberg and K. Maskell). Cambridge University Press,
Cambridge, UK, pp. 133-192. URL: http://cybele.bu.edu/
courses/gg312fall00/chap03/chap03.html#twod.
Önöz, B. and M. Bayazit. 2003. The power of statistical tests for trend
detection. Turkish J. Eng. Env. Sci. 27: 247-251.
Perez, J.F., L. Gimeno, P. Ribera, D. Gallego, R. Garia and E.
Hernandez. 2000. Influence of the North Atlantic oscillation on
winter equivalent temperature. In: Proceedings of the AGU
Chapman Conference of "The North Atlantic Oscillation",
University of Vigo (Orense campus), Orense, Galicia, Spain. URL:
http://xtide.ldeo.columbia.edu/~visbeck/nao/poster/Perez.pdf.
Plummer, N., M.J. Salinger, N. Nicholls, R. Suppiah, K.J. Hennessy,
R.M. Leighton, B. Trewin, C.M. Page and J.M. Lough. 1999.
Changes in climate extremes over the Australian region and New
Zealand during the twentieth century. Clim. Change 42: 183-
202.
Salas, J.D. 1992. Analysis and modeling of hydrologic time series, In:
Handbook of Hydrology, (Ed. D.R. Maidment), McGraw Hill Book
Company, USA, chapter 19.
Salas, J.D., J.W. Delleur, V. Yevjevich and W.L. Lane. 1980. Applied
modeling of hydrologic time series. Water Resources Publications,
Colorado.
Santer, B.D., K.E. Taylor and T.M. Wigley. 1996. A search for human
influences on the thermal structure of the atmosphere. Nature
382: 39-46.
Srivastava, H.N., B.N. Dewan, S.K. Dikshit, G.S. Prakash Rao, S.S. Singh
and K.R. Rao. 1992. Decadal trends in climate over India.
Mausam 43: 7-20.
Tol, R.S.J. 1994. Greenhouse statistics - time series analysis. Theor.
and Appl. Climatol. 49: 91-102.
Vorosmarty, C.J., P. Green, J. Salisbury and R.B. Lammers. 2000.
Global water resource: Vulnerability from climate change and
population grow. Bull. Am. Met. Soc. 289: 284-288.
Yue, S., P. Pilon and G. Caradias. 2002. Power of the Mann-Kendall and
Spearman's rho tests for detecting monotonic trends in
hydrological series. J. Hydrol. 259:254-271.
Zhang, X., L.A. Vincent, W.D. Hogg and A. Niitsoo. 2000. Temperature
and precipitation trends in Canada during the last century. Atm.
Ocean 38: 305-429.