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

net/publication/330824269

RAINFALL TREND ANALYSIS BY USING THE MANN-KENDALL TEST & SEN'S

SLOPE ESTIMATES: A CASE STUDY OF DISTRICT CHAKWAL RAIN GAUGE,
BARANI AREA, NORTHERN PUNJAB PROVINCE, PAKISTAN

Article · January 2015

CITATIONS READS
32 10,480

3 authors:

Fiaz Hussain Ghulam Nabi

PMAS - Arid Agriculture University Universiteti Europian i Tiranës
49 PUBLICATIONS   348 CITATIONS    58 PUBLICATIONS   1,453 CITATIONS   

SEE PROFILE SEE PROFILE

Muhammad Waseem Boota

University of Lahore
19 PUBLICATIONS   79 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Student View project

Development of conjunctive smart irrigation management model View project

All content following this page was uploaded by Fiaz Hussain on 27 June 2020.

The user has requested enhancement of the downloaded file.

Sci.Int.(Lahore),27(4),3159-3165,2015 ISSN 1013-5316; CODEN: SINTE 8 3159

RAINFALL TREND ANALYSIS BY USING THE MANN-KENDALL TEST &

SEN’S SLOPE ESTIMATES: A CASE STUDY OF DISTRICT CHAKWAL RAIN
GAUGE, BARANI AREA, NORTHERN PUNJAB PROVINCE, PAKISTAN
i
Fiaz Hussain1, Ghulam Nabi2, Muhammad Waseem Boota3
1
Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore, Pakistan. E-mail:
fiaz_hussain@yahoo.com
2
Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore, Pakistan. E-mail:
gnabi60@yahoo.com
3
Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore, Pakistan. E-mail:
engr.waseemboota@gmail.com
ABSTRACT: This study focused on a key climatic variable i.e. the precipitation for analyzing changing trend of rainfall of a
Barani Area Chakwal District located in Northern Punjab Province Pakistan. In this area agriculture is dependent on rainfall
but occurrence of rainfall is unpredicted and erratic which produces an adverse effects on the cropping system and also causes
negative effects on natural water resources. Rainfall data of 37 years (1977 to 2013) on daily basis was processed to find out
the monthly, seasonally and annual rainfall variability (trend and slope magnitude) by using nonparametric analysis. The
Mann-Kendall (MK) Test for monotonic analysis of trend together with nonparametric Sen’s Slope Estimator was used to
estimate the magnitude of trend for time series data.
Monthly, seasonally and annually precipitation trends has been drawn using 37 years daily data. The statistical analysis of
whole reference time series data highlight that (1) the trend looks predominantly positive (increasing), both at the annual and
seasonal scales (2) the spring, pre-monsoon and post-monsoon seasons showed decreasing trend (3) Individually seven months
(Jan, Feb, May, Jun, July, Aug and Sep) represents rising trend while other five months (Mar, Apr, Oct, Nov and Dec)
represent falling trend (4) The total annual rainfall and monsoon seasons showed positive significant increasing trend with 99
% and 95 % confidence level respectively.
Key Words: Rainfall trend; Mann-Kendall test; Sen’s Slope Estimates; Trend Analysis

1. INTRODUCTION ostentatious floods, and severe drought conditions etc. in

Globally water resources of any region has become a major different parts of country which swaying the social and
concern for planning and development of projects such as economic situations and produced affects particularly on
sustainable development of agricultural and food production those peoples which mainly depends upon agriculture and
projects, effective water resource management techniques food. The Task Force on Climate Change (TFCC) [3] give a
and erosion and flood estimation and control methodologies report on climate change in Pakistan and according to TFCC
etc. Rainfall is a source of fresh water and important element the country is exposed to a number of natural disasters,
of all types of life and agriculture mainly depends upon time including intense rainfall storms, cyclones, tornedoes, severe
of occurrence and amount of rainfall on any area to meets drought, unadorned floods, and earthquakes.
others numerous needs and demands such as industrial, The recent studies of Pakistan on climate change predicted
domestic water supply. The changes in rainfall trends/patters that in most regions the rainfall patterns have become very
are directly associated with climatic changes and according to unreliable, uneven and unpredicted. These types of climatic
IPCC 2007 reports “global surface temperature is increasing changes produces difficulties for peoples to live in this
at the rate of 0.74 ± 0.18°C over the period of 1906–2005 and environment, to make appropriate solutions and arrangements
due to this climate change the availability of fresh water will for their life safety, agricultural crops and livestock. For
reduced in future and it has been also demonstrated that in the instance Pakistan has faced a super flood on 29th July 2010,
middle of 21st century the available water and average annual due to heavy monsoon rainfall in provincial areas of Khyber
runoff will reduced up to 10 – 30%” [1]. The changing in Pakhtunkhwa, Sindh, Punjab and some parts of Balochistan
precipitation trend based on time series is the temporal shift and it was estimated that about two thousand people were
in rainfall patterns which may be either increasing or dead and seven hundred thousand homes were destroyed and
decreasing and this trend change is the dominant component damaged. A record-breaking 274 mm rainfall was measured
of the climatic variations. Human interventions are also the in Peshawar during 24 hours; while the previous record was
primary and leading factors of climate change with reference 187 mm in April 2009. On the other hand the southern and
to changing landuse from the impact of agricultural and central parts of the country has faced severe droughts from
irrigation practices [2]. So the subject of climate change with 1998 to 2001, So keeping in view these problems related to
changes in rainfall patterns directly affects water resources, climate change and its effects in Pakistan, this study was
agricultural divisions and disaster management sectors. conducted for precipitation trends analysis of district
Pakistan is agronomic and agrarian country, having diverse, Chakwal.
uneven and arid to semi-arid type climate where agriculture Trend analysis can be done using parametric and non-
depends upon rainfall occurrence and amount. Due to non- parametric approached such as t-test is parametric test while
friendly environmental activities the temperature variation of nonparametric The Mann-Kendall (MK) Test [4, 5] give
country is increasing beyond its normal limits which creates positive or negative trend for a given confidence level both at
an optimistic impact on the production of crop and this seasonal and annual scale. Mann-Kendall Test mostly used
climatic change produces the problems like glacier melting, by different researchers to find out the trends of temperature,

July August
3160 ISSN 1013-5316; CODEN: SINTE 8 Sci.Int.(Lahore),27(4),3159-3165,2015
evapotranspiration (ET) and rainfall on monthly, seasonally rainfall and cropping system. Depending upon temperature
and annually scale, the results of which had showed some variation of a whole year four seasons named as winter
significance in case of rainfall and temperature. In Zambia (October to March), summer (April to September), spring
MK Test and CUSUM (Cumulative Summation) test was (March to May) and autumn (September to October) were
used to analyze the long term rainfall trend by [6]. The trend used. According to amount of rainfall variability three
of precipitation and run-off was studied by Xu et al. [7] in seasons were used named as pre-monsoon (March to May),
major Chinese rivers by Mann-Kendall statistics for the monsoon (June to September) and post-monsoon (October to
detection of trends. The objective of this study is to analyze December) similarly depending upon the cropping system
rainfall trends/patterns using long time series data and to two seasons Rabi (November to April) and Kharif (May to
assess the significance of rainfall trends over selected study October) were used for seasonal trend analysis. Generally the
area by using The Mann-Kendall (MK) Test to gather with statistical distribution was applied to assess the behavior of
Sen’s Slope Estimator. data and the trend/pattern analysis was done as discussed
2. Study Area below.
This research was conducted as a case study for precipitation 3.1 Trend Analysis
trend analysis of district Chakwal raingage installed by Statistically trend is a significant change over time which is
Pakistan Meteorological Department having Latitude 32º 55’ detectable by parametric and non-parametric procedures
and Longitude 72º 51’ at 519 m elevation. while trend analysis of a time series consists of the magnitude
District Chakwal is a Barani district having total drainage of trend and its statistical significance. In this study statistical
area 6609 Km2, located in the southern zone of Pothwar significance trend analysis was done by using Man- Kendall
plateau, bordered by district Mianwali in West, district test while for the magnitude of trend was determined by non-
Jhelum in East, district Khushab in South and districts parametric Sen’s estimator method.
Rawalpindi and Attock in North as shown in Fig.1. The 3.1.1 Mann–Kendall Test
terrain is mostly hilly, due to laying at the beginning of This is a statistical method which is mostly used to check the
Pothwar plateau and salt range. The soil is silt loam to loam null hypothesis of no trend versus the alternative hypothesis
and its considerable area is covered by forests and plains of of the existence of monotonic increasing or decreasing trend
district are cultivated by even those placed in hilly range and of hydro-climatic time series data. The non-parametric Mann-
cultivation is mainly depend upon occurrence of rainfall Kendall test is fit for those data series where the trend may be
because there is no proper canal irrigation system so assumed to be monotonic (i.e. mathematically the trend
agriculture in this area is very precarious. consistently increasing and never decreasing or consistently
decreasing and never increasing) and no seasonal or other
cycle is present.
MAKESENS [8] performs two types of statistics depending
upon the number of data values i.e. S – statistics is used if
number of data values are less than 10 while Z – statistics
(normal approximation/distribution) for data values greater
than or equal to 10.
The statistic S is calculated as shown in equation (i)
∑ ∑ (i)
Where,
and are annual values in years j and i, j > i respectively,
n is the number of data points and ( ) is
calculated using equation (ii)

( ) { (ii)

Figure 1: Location map of study area A positive or negative value of S indicates an upward
(increasing) or downward (decreasing) trends respectively. If
3. MATERIAL AND METHODS number of data values are 10 or more, the S – statistics
Daily rainfall data of thirty seven years (1977 – 2013) for approximately behave as normally distributed and test is
district Chakwal was collected from Pakistan Metrological performed with normal distribution with the mean and
Department (PMD) Lahore and Soil and Water Conservation
variation as given below equations (iii) & (iv).
Research Institute (SAWCRI) Chakwal. This data included E(S) = 0 (iii)
daily precipitation, maximum and minimum precipitation ∑
during a month and peak rainfall event with intensity and (iv)
duration. The time series analysis of rainfall data was done on Where, n is number of tied (zero difference between
monthly, seasonally and yearly basis. Monthly data were compared values) groups and is the number of data points
arranged in such a way that nine number of seasons were in the ith tied group. The standard normal distribution (Z –
used and then these seasons were separated into three sub statistics) is computed using equation (v).
categories depending upon temperature variation, amount of

July August
Sci.Int.(Lahore),27(4),3159-3165,2015 ISSN 1013-5316; CODEN: SINTE 8 3161
The estimated Sen’s Slope (Q) was also calculated for each
√ month separately and the month Jan, Feb, May, Jun, July,
(v) Aug and Sept give increasing slope magnitude and the month
of March and April showed non-significant decreasing trend.
{ √ But the month of Oct, Sept and Dec showed no change in
Statistically the significance of trend is assessed using Z- Sen’s Slope magnitude as shown in Fig.4.
value. A positive value of Z shows upwards (increasing) The analysis mentioned in Fig.3 and Fig.4 to gather indicated
trend while the negative value indicates downward that the months of Feb May, Jun, July and Aug give
(decreasing) trend. significant increasing trend due to positive value of both Z
In MAKESENS the two-tailed test is used for four different and Q statistics while the months of Jan and Sept showed
significance levels α: 0.1, 0.05, 0.01 and 0.001. The non-significant increasing trend. The month of March give
significance level 0.001 means that there is a 0.1% significant decreasing trend due negative value of Z and Q
probability that the values xi are from a random distribution statistics and the month of April give non-significant
and with that probability we make a mistake when rejecting decreasing trend.
H0 (null hypostasis) of no trend. Thus the significance level
0.001 means that the existence of a monotonic trend is very
probable. Respectively the significance level 0.1 means that
there is a 10% probability that we make a mistake when
rejecting H0.
3.1.2 Sen’s Estimator method
Sen’s non-parametric estimator method has been used for
predicting the magnitude (true slope) of hydro-metrological
time series data. The Sen’s slope estimator method uses a
Figure 3: The Mann-Kendall Z -Statistics for Monthly Trend
linear model for the trend analysis. The slope (Ti) of all data
Analysis
pairs is calculated using equation (vi) by [9].
(vi)
Where, and are data values at time j and k (j > k)
separately.
The median of these n values of Ti is represented by Sen’s
slope of estimation (true slope) which is calculated using
equation (vii)
Figure 4: The magnitude of trend using Q –Statistics
{ (vii)
( )
4.2 Seasonally Rainfall Trend Analysis
Sen’s estimator is calculated using above equation Rainfall trend was analyzed on three sub-categories of
depending upon value of n is either odd or even and then seasons as discussed below.
is computed using 100 (1 – α) % confidence interval 4.2.1 Seasonal variation of rainfall trend depends upon
using non-parametric test depending upon normal temperature variation
distribution. A positive value of indicate increasing Four seasons were used depending upon temperature i.e.
(upward) trend while negative value of represent winter, summer, spring and autumn. The value of Z and Q –
downward or decreasing trend of time series data. Statistics for three seasons, winter, summer and autumn is
positive which showed significant increasing trend while the
4. RESULTS AND DISCUSSION spring season indicate decreasing trend as shown in Fig.5
In the present study trend analysis of rainfall of district
Chakwal rain gauge was done using Mann-Kendall to gather
with Sen’s slope estimator for 37 years of time series data
(1977- 2013) on monthly, seasonally and yearly basis.
4.1 Monthly Rainfall Trend Analysis
The variation in rainfall data (trend) on monthly basis is
calculated individually for each month using Mann-Kendall
statistically method and magnitude of slope is calculated with
Sen’s slope estimator as represented in Fig. 2. It was analyzed
that there is significant changes in monthly rainfall data mean
some of the months showed increasing (upward) trend and Figure 5: Seasonal Variation of Rainfall Trend Depending upon
some showed decreasing trends. Seven months (Jan, Feb, Temperature
May, Jun, July, Aug and Sept) give positive values of Z– 4.2.2 Seasonal variation of rainfall trend depend upon
Statistics which represent rising trend while other months amount of Rainfall variation
(Mar, Apr, Oct, Nov and Dec) represent falling trend as Depending upon the amount of rainfall, three type of seasons
shown in Fig.3 were used, i.e. pre-monsoon (March to May), monsoon (June

July August
3162 ISSN 1013-5316; CODEN: SINTE 8 Sci.Int.(Lahore),27(4),3159-3165,2015
to September) and post-monsoon (October to December). The monsoon and post-monsoon seasons showed downward trend
monsoon seasons showed significant upward trend having due to negative value of Z and Q – Statistics as represented in
positive values of both Z and Q – Statistics while pre- Fig.6

Figure 2: Monthly Rainfall Trend Analysis

July August
Sci.Int.(Lahore),27(4),3159-3165,2015 ISSN 1013-5316; CODEN: SINTE 8 3163
(443 mm) of year 1997. The calculated mean annual rainfall
for 37 year data is 625 mm and maximum total rainfall (1241
mm) occurred in 1997.

Figure 2: Seasonal Variation of Rainfall Trends Depending

upon Amount of Rainfall

4.2.3 Seasonal variation of rainfall trend depend upon Figure 4: Annual Rainfall Trend Analysis
cropping system
Rabi season consist of Nov to April months and this seasons The Mann-Kendall trend, its statistical significance along
showed increasing trend due to positive values of Z and Q with magnitude of Sen’s slope for 1977 to 2013 year rainfall
statistics similarly Kharif season (May to Oct) also showed data is shown in Table.1. The months of February and June
increasing trend as shown in Fig.7 showed increasing monotonic trend with significance level
0.01, that mean there is a 1% probability that we make a
mistake when rejecting Ho (null hypothesis) of no trend.
Similarly those months and seasons which have single star
(*) sign as given in significance (α) column of this table
showed significance level 0.05 having 95 % significance
interval of Q and the month of august and annual time series
data having plus (+) sign showed 0.1 significance level
Figure 3: Seasonal Variation of Rainfall Trend Depending upon
Cropping System having 99 % significance interval of Q and there is 10 %
probability that we make a mistake when rejecting Ho. While
4.3 Annual Rainfall Trend Analysis the other months and seasons having column with (-) sign
The trend of total annual rainfall of 37 years data is showed significance level greater than 0.1.
represented in Fig.8 which depicted that the total maximum Generally those time series data give quite significant trends
annual rainfall (1241 mm) has occurred in the year 1997 whose both Mann-Kendall trend (Z statistics) and Sen’s
while minimum total annual rainfall (225 mm) has occurred Slope magnitude (Q statistics) having either increasing
in 1979 of whole reference period. Average annual rainfall of (positive)or decreasing (negative) values.
37 years data is 625 mm. The positive vale of Z and Q –
Statistics showed that there is rising trend with upward slope 5. CONCLUSION and RECOMMENDATION
of rainfall on annual basis. The results of study depicted that there is substantial year to
4.4 Mann-Kendall Trend and Sen’s Slope year and season to season variability in rainfall pattern and
The statistical analysis of rainfall data was done using central rainfall pattern is generally erratic in nature. The statistical
tendency parameters i.e. (mean and median) and the result of tests indicated that in some months there is an
dispersion of data from mean was done using standard increasing trend while in some other months decreasing trend
deviation and coefficient of variance as shown in Table.1. of precipitation. (1) Individually seven months (Jan, Feb,
Table.1 illustrated that minimum mean monthly rainfall May, Jun, July, Aug and Sept) represents increasing trend
occurred in the month of November (6 mm) for all these 37 while other five months (Mar, Apr, Oct, Nov and Dec)
year data followed by December (14 mm) and January (25 represent decreasing trend (2) the trend looks predominantly
mm) while maximum mean monthly rainfall occurred in positive (increasing), both at the annual and seasonal scale (3)
month of august (144 mm). It is also depicted from the table the spring, pre-monsoon and post-monsoon seasons showed
that total maximum rainfall was occurred in month of august decreasing trend while other showed increasing trend.

July August
3164 ISSN 1013-5316; CODEN: SINTE 8 Sci.Int.(Lahore),27(4),3159-3165,2015

Table 1: Statistical Analysis of Rainfall data along with Mann-Kendall Trend and Sen's Slope

Mean Median Maxi Mini St. Coef. Mann-Kendall Signific. Sen's

Time Series
(mm) (mm) (mm) (mm) Dev Var trend (α) Slope
JAN 25 14 130 4 32 130 1.003 - 0.121
FEB 48 33 213 7 51 106 3.011 ** 1.332
MAR 52 40 201 3 45 87 -1.452 - -0.823
APR 42 23 223 6 45 109 -0.052 - -0.011
MAY 27 25 71 2 20 75 1.191 - 0.392
JUN 61 48 235 6 50 81 2.970 ** 1.817
JUL 124 109 322 19 69 56 0.955 - 1.045
AUG 144 138 443 20 83 57 1.727 + 2.036
SEP 64 52 291 7 55 85 0.288 - 0.123
OCT 19 15 105 4 22 117 -0.052 - 0.000
NOV 6 1 42 2 10 153 -0.217 - 0.000
DEC 14 4 99 3 24 173 -1.092 - 0.000
Annual (J-D) 625 640 1241 225 197 32 1.674 + 5.515
Winter (O-M) 164 153 344 57 76 47 0.981 - 1.356
Summer (A-S) 461 454 1057 120 172 37 1.530 - 4.215
Cold Winter(DJF) 86 69 213 3 60 69 2.368 * 2.117
Spring (MAM) 121 103 363 30 75 62 -0.785 - -0.597
Hot Summer (JJA) 329 326 759 84 134 41 2.538 * 5.182
Autumn (SON) 89 83 305 19 60 67 0.105 - 0.076
Pre-Monsoon
121 103 363 30 75 62 -0.785 - -0.597
(MAM)
Monsoon (JJAS) 393 373 856 119 149 38 2.106 * 4.863
Post-Monsoon
39 30 128 3 31 79 -0.864 - -0.151
(OND)
Rabi (ND, J-A) 186 172 418 65 90 49 0.876 - 1.265
Kharif (M-O) 438 414 1032 144 169 38 2.236 * 5.216

The monthly analysis indicated that the months of February, productivity in rainfed areas of Pakistan and this increasing
May, June, July and August give significant increasing trend trend of rainfall on annual and seasonal basis can be used for
due to positive value of both Z and Q statistics while the better planning of water resources development and
January and September showed non-significant increasing management schemes as well as conservation of soil moisture
trend. The month of March give significant decreasing trend in study area of Pothwar region Pakistan.
due negative value of Z and Q statistics and the month of
April give non-significant decreasing trend. The trend of 6. REFERENCES
whole data on annual basis showed positive increasing trend. [1] IPCC, Climate change 2007, “Climate change impacts,
The statistical analysis of whole series data indicated that the adaptation and vulnerability”. Working group II
average annual rainfall of study area raingage is 625 mm with contribution to the Intergovernmental Panel on Climate
maximum average annual is 1241 mm and minimum average Change Fourth Assessment Report. Summery for
annual amount of rainfall is 225 mm also the value of policymakers, 23 (2007).
standard deviation of rainfall data depicted that there is a [2] Kalnay, E. and Cai, M., “Impact of urbanization and
great fluctuation in rainfall, about 70 % amount of rainfall landuse changes on climate”. Nature 423, 528 – 531
occurs in the months of monsoon season also the months of (2003).
July and August give maximum amount of rainfall while the [3] T.F.C.C., “Climate change in Pakistan: TFCC (Task
months of November and December give minimum amount Force for Climate Change) recommends various
of rainfall. adaptations, mitigation measures” (2010).
Rainfall is the most important agro-climatic variable that [4] Mann, H.B., “Non-parametric test against trend”,
determines the cropping system and overall agricultural Econometrica 13, 245 – 259 (1945).

July August
 Sci.Int.(Lahore),27(4),3159-3165,2015 ISSN 1013-5316; CODEN: SINTE 8 3165
[5] Kendall, M.G., “Rank Correlation Methods”, 4th edition. [7] Xu, K. Milliman, J.D. Xu, H., “Temporal trend of
Charles Griffin, London, UK (1975). precipitation and runoff in major Chinese Rivers since
[6] Kampata, J.M. Parida, B.P. Moalafhi, D.B, “Trend 1951”. Global and Planetary Change 73, 219 – 232
analysis of rainfall in the headstreams of the Zambezi (2010).
River Basin in Zambia”, Physics and Chemistry of [8] MAKESENS 1.0, “Mann-Kendall Test and Sen’s Slope
Earth 33, 621 – 625 (2008). Estimates for Trend of Annual Data” Finnish
Meteorological Institute, (2002).
[9] Sen, P.K., “Estimates of regression coefficient based on
Kendall’s tau”, Journal of American Statistical
Association 39, 1379 – 1389 (1968).
i
Corresponding Author:
Email: fiaz_hussain08@yahoo.com

July August

View publication stats