See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/284168230

A new methodology for trend analysis: A case study in Burdur and Isparta,
Turkey

Article  in  Fresenius Environmental Bulletin · October 2015

CITATIONS READS

12 6,461

1 author:

Kemal Saplıoğlu
T.C. Süleyman Demirel Üniversitesi
61 PUBLICATIONS   178 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Kemal Saplıoğlu on 19 November 2015.

The user has requested enhancement of the downloaded file.

© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

A NEW METHODOLOGY FOR TREND ANALYSIS:

A CASE STUDY IN BURDUR AND ISPARTA, TURKEY

Kemal Saplıoğlu

Süleyman Demirel University, Civil Engineering Department, Isparta, Turkey

ABSTRACT Gregory [3] has researched the changes of rainfall

trends in England and Suzuki [4] has done a similar study
Nowadays where global climate change effects have for Japan. The trend changes using rainfall data in India
been remarkably perceived, the determination of trends of were analyzed by Jagannathan and Parthasarathy [5] and
rainfall, stream flow, and temperatures has become crucial Parthasarathy and Dhar [6]. Besides, Goswami et al. [7]
for the planning of water sources and engineering phases of have studied the trend changes of the monsoon rains which
future projects. A bunch of methods are suggested by many were accounted for 80% of total rainfall amount. Singh et
researchers to determine these trends and some estimations al. [8], have studied the trend changes of central and north-
are tried to be made. A new method based on a graphical west regions of India annually and monthly using the rain-
method developed by Şen is proposed in this study. fall data.
This new method was compared with current methods Nasrallah and Balling [9] have researched the temper-
of Mann-Kendall test statistic, regression model and Sen’s ature trend analysis for Kuwait between the years of 1950
graphical method. For comparison, the rainfall data of Bur- and 1990 and found a temperature increase of 0.070C in
dur and Isparta have been chosen, and the trend analyses place. El Nesr et al. [10] have studied the trend analysis
were performed monthly and annually. As one of the con- using maximum, minimum and average temperature values
clusions of this study, it iwas observed that Sen’s graphical of 29 years using 29 stations in Saudi Arabia. It was con-
test and this new method have similar results and it was cluded that there was a trend for winter months (November
seen that the results of the regression test support both of to January) in that study. Chan et al. [11] have studied the
these two tests. On the other hand, Mann-Kendall test re- Hanjiang basin in China and have not found any tempera-
sults were found as partly similar to the results of this ture trend with 95% accuracy in accordance with the study.
study. In conclusion, it is considered that this new test Also in this study an increasing temperature trend was de-
method can be used for trend analysis. termined for some parts of the basin. In Danjiangkou res-
ervoir, it was found that this increase had a variety from
10C to 20C.
KEYWORDS: Mann-Kendall, Sen’s graphical test, regression, new Mann-Kendall test is frequently used in analysis of hy-
methodology, Isparta, Burdur, trend, rainfall drometeorology data [12]. In Mann-Kendall test, when the
results of tests exceed some certain limits, positive or neg-
ative trends are observed [13]. Wang et al. [14] have mod-
elled the rainfall data of West China for the second half of
1. INTRODUCTION the 20th century and they have determined an increase in
precipitation amount. Nevertheless, there not any increase
In recent years, a lot of studies have been made on trend in East China was observed in the same study.
global and regional climate changes. Most of these studies Partal and Kahya [15] analyzed annually and monthly
have presented some remarkable findings. However they trend changes in the amount of rainfall. They completed
seem to require further test in order to be regarded as being the analysis using nonparametric methods of Sen’s T test
accurate or not. and Mann-Kendall using 96 rainfall measurement stations
in various places of Turkey. As a result of this study, espe-
In order to mention about climate change in a region, cially in 14 stations, the decrease of rainfall in January was
there should be substantial differences between average remarkable. Tabari et al. [16] made a study about climate
values of years for a long period of decades or even more changes in 13 stations covering the east, south and south-
[1]. However, in the case of global climate change, the re- east part of Iran between 1966-2005. They used Mann-
gions should affect each other [2]. It is seen that the climate Kendall, Mann-Whitney, and Mann-Kendall Rank Corre-
change studies focus on rainfall, temperature and flowrates lation tests. According to the results, from the beginning of
of rivers in general. 1970’s, an increase of 0,412°C in average values, 0,452°C

3344
© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

in maximum values and 0,493°C in minimum values were coordinates of 37° 72’ North latitude and 30° 29’ East lon-
obtained. gitude. This station has an elevation of 957 m over the sea
Karmesha [17] has studied the rainfall and temperature level and the average annual rainfall height is 429,16 mm.
trends of 9 states in the northwest part of the USA for the Isparta central station is located on the coordinates of 37°78’
years in between 1900-2011. In this study, statistically sig- N and 30°56’ E. This station has an elevation of 997 m and
nificant increases in trends were observed for both rainfall the average annual rainfall height is 514.7mm [23].
and temperature values. Only for New Hampshire and
2.2. Methods
Maine states, the increases determined were not be the
2.2.1. Mann-Kendall Test
ones statistically supported. Gocic and Trajkovic [1] have
determined increase and decrease trends with 95% signifi- The Mann Kendall test [24, 25] which is nonparamet-
cance level using Mann-Kendall and Spearman’s tests in ric, is independent from the data distribution. Whether
their studies on rainfall amounts in Serbia. In these tests, there is a trend in time series is checked by zero hypotheses
there seemed a remarkable decrease for February and Sep- “Ho: no trend”. The time series to which the test will be
tember for all the stations studied. For the other months, applied was divided into two groups as x1, x2, … xn
there were decrease and increase trends varying at different (xi,,xj) pairs. If the number of pairs where xi<xj for i<j is
stations. In addition, during the intervals of 1987-1994 and shown as P and the number of pairs where xi>xj as M, the
2000-2003, a remarkable drought was brought up as a mat- test statistics is defined as S=P-M.
ter. The year 2000 is found to be as the droughtiest year for
all the stations in that study. Mohamed and Saavaenije [18] For n>10 the Kendall correlation coefficient is:
have used Mann-Kendall test to understand local, regional
and general trend changes in Sudd basin around the Nile μ 0 ve σ n n 1 2n 5 /18 (1)
River, Sudan. In the study, the data of 100 years in between
1900 and 2000 were used. One of the most striking results s 0
of that study of the basin was the increase trend in Victoria
Lake for 1960 and later as well as a temperature increase
of 0,6°C and 1,5°C degrees for maximum and minimum z 0 s 0 (2)
values successively.
Şen [19] has represented an innovative way of graphics s 0
splitting the data into 2 groups in his published study. The
study was applied for the annual flow of the Danube River
in Turkey, Cizre River, Aslantaş Dam and Menzelet Dam. It is concluded that if the absolute value of Z, calcu-
Şen [20] benefited from his innovative graphical method- lated according to the above method, is less than the Z/2
ology to visualize trend analysis of temperature values in value of a normal distribution at the α significance level,
Marmara region. Saplıoğlu et al. [21] have compared the zero hypothesis is accepted and time serial does not include
Mann Kendall Test results of annual and monthly trend a trend. If it is greater than Z/2, time serial includes a trend,
analysis of three different observation station in Southwest and if S is positive, this trend is an increasing one. If S is
Turkey with the results calculated by Şen’s graphical negative, on the other hand, trend is a decreasing trend. As
method. Besides, the degree of slope in the graphical the data is not required to be compatible with a distribution,
method was compared to see whether it is a trend indicator this test is particularly useful [26].
or not. Kişi and Ay [22] have used both Mann-Kendall
trend test and Şen’s graphical method for their study on the 2.2.2. Regression Analysis
Kızılırmak River. In linear Regression analysis, the solution is based on
In the present study, rainfall data of Burdur (17238) entering two different variables for different axis. A line
and Isparta (17240) obtained from measurement stations, which fits best on data is determined and trend of this line
were evaluated by Mann-Kendall, Regression, Şen’s graph- is calculated as an equation [27].
ical method and also by Proposed method. The results ob- . (3)
tained were compared with each other and the usability of In the equation above, is a constant value while
the recommended model was discussed further. corresponds to the slope. If this equation is used for calcu-
lating the trendline, shows the amount of increase or de-
crease in the trend.
2. MATERIALS and METHOD
2.2.3. Şen’s Graphical Test
2.1. Field of Study
Şen [19] demonstrated a trend analysis method depend-
As field study, Burdur (17238) and Isparta (17240) sta- ing 1:1 line on the Cartesian coordinate system. Where the
tions located in Central Mediterranean Region were used. 1: 1 line and regions close to this line refers to the portion
The data of the years between 1975 and 2006 were taken which is not the trend. There are triangular areas on both
into consideration. Burdur central station is located on the sides of the 1:1 line. These triangular areas contain infor-

3345
© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

mation about the formation and direction of the trend Here the “i” refers to index, “ ” refers to the first data
(Figure 1). First, time serial data is divided into two groups set, “ ” refers to the second data set, “n” refers to the
with equal numbers. The first serial is the data from the be- number of data in each data set, and “MX” refers to the
ginning of the time series to the midpoint of the time interval, mean of the difference of data between data sets.
the second serial is the data from the midpoint of the time
interval to the last data. Both of these data is sorted from big H0 hypothesis is formed. This hypothesis is accepted
to small in itself. Scatter diagrams are created where the first as there is no change. The hypothesis is evaluated by t-test
group data is shown in the x-axis and the second group data where the results are n≤30 and evaluated by z test where
is shown in the y-axis. Finally, the 1: 1 line is passed through the results are n>30 evaluated. In this evaluation if the hy-
right in the middle of the scatter diagram. If the results ob- pothesis is rejected, it is decided that that there is an in-
tained in the scatter diagram fall below the 1:1 line the trend crease or decrease trend in place. If the results are positive,
line may be considered towards reduction, and vice versa. it is estimated that there is a positive trend and vice versa,
Besides, in case of observed values exactly on or close to the since the first data set covers the half of data.
1:1 line, it can be considered that there is no trend. Apart
from those, if there are values on both side of the 1:1 line in
the diagram, it can be estimated that there are only the low 3. RESULTS
values or high values have increase or decrease trends ac-
cording to the form of the graphic [19, 20]. The annual and monthly data obtained from three rain-
fall measurement stations are separately modelled and the
2.2.4. Recommended Method results are compared with each other using Mann-Kendall
test, Regression test, Şen’s graphical test and the recom-
This is a method inspired by the Şen graphical test. In
mended method. There are 32 data sets which each include
this method as in Şen’s test, the data is divided evenly into
Mann-Kendall and Regression tests, however there are
two portions and ranked from lower to higher. In the next
16 data sets which each for Şen’s test and the recommended
step the first value of first data set is subtracted from the
method each. Therefore the recommended method is tested by
first value of the second data set. Then the second value of
t test. A program is coded using Matlab in order to transact
the first data set is subtracted from the second value of the
the recommended method and Mann-Kendall test automat-
second data set. This process is maintained until the end of
ically. Mann-Kendall test is applied in accordance with
data. In the next step, the average of all values obtained is
both 95% and 90% confidence intervals and then the trends
calculated (Equation 4).
are determined. In regression analysis, the differentiation
of the annual and the monthly rainfall heights are found in
∑ / (4) mm and percentage equivalents of the results are calcula-

FIGURE 1 - Şen’s Graphic test representation of increase and decrease trends

3346
© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

lated. The results were evaluated and trends were found in changes in January and August were seen as supported by
Şen’s graphical method. The results obtained from the rec- any other method. Trends of Sen’s graphical method with
ommended method were tested by t test with 95% confi- the recommended method were overlapping with each
dence interval and the trends were found again. All the re- other. Also, according to regression test in April, June and
sults obtained by methods were compared and the recom- July, 1% and over precipitation changes were supported
mended method was discussed in terms of usability. with the graphic method of Sen and the recommended
In Table 1, the test results belonging to Burdur prov- method.
ince are summarized. According to these results, there was Considering annual and monthly trend analysis of Bur-
not any trend found with a 95% confidence interval in dur province, while there was a trend of decrease in Janu-
Mann-Kendall trend analysis. However, a decrease trend in ary, February and June, a trend of increase was observed in
January and an increase trend in August was observed with April, May, July, and August. Besides a slight increase on
90% confidence interval of Mann-Kendall trend test. Trend an annual basis a trend was observed (Figure 2).

TABLE 1 - Monthly and annual rainfall data trend analysis of Burdur (17238) province

Month Mean Mann- Mann- Mann- Annual Percentage Şen New New
Flow Kendall Kendall Kendall Change change Trend Method Method
Rate Z Trend Trend for Regres- Test Z Trend
(mm) (0,95) (0,9) sion (mm) (0,95) (2,12)
(1,96) (1,64)
January 51,76 -1,84   -0,69 -1,33  -2,67 
February 38,97 -0,69   -0,31 -0,80  -4,54 
March 47,86 0,73   0,42 0,88  2,01 
April 50,81 1,51   0,99 1,95 ↑ 8,33 ↑
May 41,22 0,19   0,085 0,21 ↑ 3,19 ↑
June 26,08 -0,6   -0,31 -1,19  -3,14 
July 15,38 0,66   0,43 2,80 ↑ 3,15 ↑
August 7,84 1,75  ↑ 0,37 4,72 ↑ 2,41 ↑
September 16,84 0,67   0,14 0,83  -0,01 
October 34,30 -0,52   0,15 0,44  -0,02 
November 40,27 0,32   0,30 0,74  -0,64 
December 57,82 -0,31   0,47 0,81  1,84 
Annual 429,16 1,05   1,86 0,43 ↑ 3,22 ↑

180 90 180
Rainfall (mm) 1991-2006
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Jan. Feb. March

120 60 120

60 30 60

0 0
0
0 30 60 90 0 60 120 180
0 60 120 180
Rainfall (mm)  Rainfall (mm)
Rainfall (mm)  1975‐1990
1975‐1990 1975-1990

165 165 90
Rainfall (mm) 1991‐2006

June
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Apr. May
110 110 60

55 55 30
0 0
0
0 55 110 165 0 55 110 165
0 30 60 90
Rainfall (mm)
Rainfall (mm) 
1975‐1990 Rainfall (mm) 
1975‐1990
1975‐1990

3347
© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

120 90 120
Aug. Sept.
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006
July
80 60 80

40 30 40

0 0 0
0 40 80 120 0 30 60 90 0 40 80 120
Rainfall (mm)  Rainfall (mm)  Rainfall (mm) 
1975‐1990 1975‐1990 1975‐1990

165 210 120

Oct. Nov.
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006
Dec.
110 140 80

55 70 40

0 0 0
0 55 110 165 0 70 140 210 0 40 80 120
Rainfall (mm)  Rainfall (mm) Rainfall (mm) 
1975‐1990 1975‐1990 1975‐1990

750
Rainfall (mm) 1991‐2006

Annual
500

250
250 500 750
Rainfall (mm)
1975‐1990

FIGURE 2 - Burdur (17238) province Sen’s graphical test monthly and annual trends

TABLE 2 - Isparta (17240) province’s monthly and annual trend analysis of rainfall data

Month Mean Mann- Mann- Mann- Annual Percentage Şen New New
Flow Kendall Kendall Kendall Change change Trend Method Method
Rate Z Trend Trend for Regres- Test Z Trend
(mm) (0,95) (0,9) sion (mm) (0,95)
(1,96) (1,64) (2,12)
January 65,11 -1,13   -0,86 -1,32  -2,79 
February 56,54 -1,02   -0,75 -1,33  -6,79 
March 54,26 0,37   0,45 0,83 ↑ 2,55 ↑
April 60,07 0,79   0,33 0,55 ↑ 1,76 
May 47,87 0,60   0,1 0,21 ↑ 4,59 ↑
June 28,59 -1,64   -0,79 -2,76  -4,65 
July 13,15 0,61   -0,28 -2,13 ↑ 2,72 ↑
August 12,30 1,37   0,15 1,22 ↑ 1,94 
September 15,65 0,52   0,55 3,51 ↑ 3,27 ↑
October 38,50 -0,99   -0,43 -1,12  -1,59 
November 50,00 0,36   0,11 0,22  -2,29 
December 72,53 -1,08   -0,16 -0,22 ↑ 0,55 
Annual 514,58 -0,84   1,23 0,24  -0,64 

3348
© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

210 150 180

Rainfall (mm) 1991‐2006
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006
Jan. Feb. March
140 100 120

70 50 60

0 0 0
0 70 140 210 0 50 100 150 0 60 120 180
Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990
Rainfall (mm) 1975‐1990

165 120 90

Rainfall (mm) 1991‐2006
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006
Apr. May June
110 80 60

55 40 30

0 0 0
0 55 110 165 0 40 80 120 0 30 60 90
Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990

90 60 75
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006
July Aug. Sept.
60 40 50

30 20 25

0 0 0
0 30 60 90 0 20 40 60 0 25 50 75
Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990

150 210 225

Oct.
Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Rainfall (mm) 1991‐2006

Nov. Dec.
100 140 150

50 70 75

0 0 0
0 50 100 150 0 70 140 210 0 75 150 225
Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990 Rainfall (mm) 1975‐1990

810
Rainfall (mm) 1991‐2006

Annual
640

470

300
300 470 640 810
Rainfall (mm) 1975‐1990

FIGURE 3 - Burdur (17238) province’s Sen graphical test monthly and annual trends

3349
© by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

In Table 2, the test results belonging to Isparta prov- REFERENCES

ince are summarized. According to these results, there was
not any trend with 95% confidence interval in Mann-Ken- [1] Gocic, M. and Trajkovic, S. (2013) Analysis of precipitation
dall trend analysis. However, a decrease trend in June was and drought data in Serbia over the period 1980-2010. J. Hy-
drol, 494, 32-42.
observed with 90% confidence interval in Mann-Kendall
trend test. Trend changes in June seemed to be supported by [2] Trajkovic and, S. and Kolakovic, S. (2009) Wind-adjusted
Turc equation for estimating reference. Hydrol. Res, 40:1, 45-
any other method. Even if trends of Sen’s graphical method 52
and the recommended method appear not to overlap overlap
[3] Gregory, G. (1956) Regional variations in the trends of annual
in April, August, October and December (according to 95% rainfall over the British Isles. George. J.,122, 346-353.
confidence interval); the recommended method had a 90%
[4] Suzuki, E. (1968) Secular variations of the rainfall in Japan.
of confidence interval in the first three months.
Pap. Met. Geophys, 19, 363-399.
In December, the growth trend appeared to be decreas- [5] Jagannathan, R. and Parthasarathy, B. (1973) Trends and peri-
ing considering the small observation data, and increasing odick citiles of rainfall over India. Mon. Weath. Rev., 101,
considering the large rainfall data. Therefore there was a 691-700.
separation between the two methods. It is also important [6] Parthasarathy, B. and Dhar, O. (1974) Secular variations of the
that in the study a 3.51% increase in September was supported rainfallover India. Quart. J. Roy. Met.Soc, 100, 245-257.
by the other methods except for the Mann -Kendall test. [7] Goswami, B., Venugopal, V., Sengupta, D., Madhusoodanan
M. and Xavier, P. (2006) Increasing trend of extreme rain
Considering annual and monthly trend analysis of Bur- events over India in a warming environment, Science, 314,
dur province, while there is a trend of decrease in January, 1442-1444.
February, October and November; a trend of increase was [8] Singh, P., Kumar, V., Thomas, T. and Arora, M. (2008)
observed in March, April, May, July, and August. In De- Changes in rainfall and relative humidity in river basins in
cember both increase and decrease exist in trend. Also it northwest and central India. Hydrolog. Process., 22:16, 2982–
was observed that there was not any change in trend annu- 2992.
ally (Figure 3). [9] Nasrallah, H. and Balling, R. (1995) Impact of desertification
on temperature trends in the Middle East. Environ. Monitor.
Assess, 37,265-271.
4. CONCLUSIONS [10] ElNesr, M., Abu-Zreig, M. and Alazba, A. (2010) Tempera-
ture trends and distribution the Arabian Peninsula. Am. J. En-
viron. Sci. 6:2, 191–203.
In this study, a new method has been recommended for
trend analysis. This method was compared with Mann- [11] Chen, H., Guo, S., Xu, C.-Y. and Singh V. P. (2007) Historical
Temporal Trends of Hydro-climatic Variables and their rele-
Kendall, Regression and Sen’s graphical method in order vance in water resource management in the Hanjiang basin. J.
to investigate its usability. Comparison of Mann-Kendall Hydrol., 344, 175-185.
and the recommended test indicated that trend results of [12] Yue, S., Pilon, P. and Cavadias, G. (2002) Power of the Mann
13 monthly data sets were supported each other and they Kendall and series. J. Hydrol, 259 , 254-271.
are same for the 22 monthly data sets with 95% confidence
[13] Bayazıt, M. (2007) To prewhiten or not to prewhiten in trend
interval. This number increased to 25 with 90% confidence analysis," Hydrol. Sci. J., 52:4, 611-624.
interval. In the light of these results it is observed that the
[14] Wang, S., Zhu, J. and Cai, J. Interdecadal variability of tem-
method is compatible with Sen’s graphic method and re- perature and precipitation in China since 1880. Adv. Atmos.
gression analysis and it is also practicable. The proposed Sci., 21:3, 307–313.
method’s results were consistent with the Sen’s graphical [15] Partal, T. and Kahya, E. (2006) Trend analysis in Turkish pre-
method and regression analysis and partly compatible with cipitation data. Hydrolog. Process., 20: 2011-2026.
the results of Mann-Kendall trend test. Additionally, trends [16] Tabari, H., Somoe, B. and Zadeh, M. R. (2011) Testing for
were detected by analyzing the data sets of Burdur (17248) long-term trends in climatic variables in Iran, Atmos. Res.,
and Isparta (17240) rain gauge stations. Considering all the 100, 132-140.
results obtained in Burdur, a decrease in January and an [17] Karmesha, N. (2012) Trend Detection in Annual Temperature
increase in August were observed. In Isparta, on the other & Precipitation using the Mann-Kendall Test. Master of envi-
hand, June had a decreasing trend. Another common point ronmentail studies Capstone Project, Northeastern United
among all the trend tests is the time periods in which there States, 2012.
are no trends. In all tests, with 95% confidence interval no [18] Mohamed, Y. and Saavaenije, H. G. (2014) Impact of climate
trend in March, September, October, November and De- variability on the hydrology of the Suddwettland. Wetlands
Ecol. Manage, 22, 191–198.
cember was observed in Burdur while in Isparta there is no
trend annually. [19] Şen, Z. (2012) Innovative Trend Analysis Methodology. J.
Hydrol. Eng., 17, 1042-1046.

The author has declared no conflict of interest. [20] Şen, Z. (2014) Trend Identification Simulation and Applica-
tion. J. Hydrol. Eng, 19:3, 635-642.
[21] Saplıoğlu, K., Kilit, M. and Yavuz, B. K. (2014) Trend analy-
sis of streams in the western Mediterranean basin of Turkey,
Fresenius Environmental Bulletin, 23:1a, 313-324.

3350
 © by PSP Volume 24 – No 10a. 2015 Fresenius Environmental Bulletin

[22] Kişi Ö. and Ay M., (2014) Comparison of Mann–Kendall and

innovative trend method for water quality parameters of the
Kizilirmak River, Turkey, J. Hydrol., 513, 362-375.
[23] MGM, http://www.mgm.gov.tr/kurumsal/istasyonlarimiz.
aspx? sSirala= AL&m=ISPARTA (2014)
[24] Mann, H. (1945) Non-parametric test against trend, Econo-
metrika, 13, 245-259.
[25] Kendall, M. (1975) Rank Correlation Methods, London:
Charles Griffin.
[26] Yue, S. and Hashino, M. (2003) Long term trends of annual
and monthly precipitation in Japan. J. Am. Water Resour. As-
soc., 39, 587-596.
[27] Saplıoğlu, M. (2010) An accident prediction model for urban
uncontrolled intersections, Natural and Applied Sciences,
Süleyman Demirel Uni. Isparta, Turkey. (in Turkish).

Received: February 09, 2015

Revised: April 23, 2015
Accepted: May 13, 2015

CORRESPONDING AUTHOR

Kemal Saplıoğlu
Süleyman Demirel University
Civil Engineering Department
Isparta
TURKEY

E-mail: kemalsaplioglu@sdu.edu.tr

FEB/ Vol 24/ No 10a/ 2015 – pages 3344 - 3351

3351

View publication stats