Spat. Inf. Res.

https://doi.org/10.1007/s41324-019-00250-8

Analysis of long-term seasonal and annual temperature trends

in North Bengal, India
Pritha Datta1 • Soumik Das2

Received: 3 November 2018 / Revised: 17 January 2019 / Accepted: 22 January 2019

 Korean Spatial Information Society 2019

Abstract In the recent past, scientific modeling of climatic minimum temperature is rising rapidly than maximum
elements, in particular, temperature data has attained con- temperature across the region. Additionally, Sequential
siderable importance as it affects many aspects of the Mann–Kendall test exhibits the periodic fluctuation of
environment and also indicates a clear sign of climate trends, which are more prominent in pre-monsoon and
change. The temperature warming is mainly associated monsoon season.
with the increasing concentration of greenhouse gases
triggered by the land use and land cover changes. In the last Keywords Climate change  Temperature trend  North
century, exploitation of forest resource, population influx, Bengal  Mann–Kendall test  Theil-Sen’s slope estimator 
and expansion of agricultural land has changed the natural Sequential Mann–Kendall test
landscape of North Bengal to a great extent. Thus the
present study intends to find out long-term changes in
maximum and minimum temperature for six northern dis- 1 Introduction
tricts of the state of West Bengal, popularly known as
North Bengal. The non-parametric Mann–Kendall test and Surface air temperature is a crucial element of Earth’s
Theil-Sen’s slope estimator reveals the presence of climate owing to its direct influence on almost all hydro-
warming trends in both maximum and minimum temper- meteorological variables. It varies across different regions
ature. Annual temperature is rising 0.006 C per year and both seasonally and annually on the basis of latitude, alti-
0.007 C per year at most of the districts for the maximum tude, surface albedo, evapotranspiration and different
and minimum temperature respectively. However, the geographical features [1, 2]. Several studies have provided
seasonal analysis of trend exposes that post-monsoon and evidence of global atmospheric warming since the early
winter temperature rise are predominantly contributing to twentieth century. The air temperature has increased
the upward annual trend. The highest increasing trend in 0.6 ± 0.2 C during the last century [3]. However, a
maximum and minimum temperature is observed at Malda substantial amount of spatial and temporal variation can be
(0.013 C per year) in the post-monsoon and winter season present in climatically different regions [4]. Increasing
respectively. Except the post-monsoon season, the temperature leads to the intensification of hydrological
cycles which further affects the spatio-temporal distribu-
tion of runoff, soil moisture, groundwater reserves as well
& Pritha Datta as the frequency of extreme events viz. droughts and floods
[email protected] [5]. Thus the warming trend of temperature under the effect
1
of climate change plays a pivotal role in determining the
Department of Geography, Aryabhatta School of Earth
mitigation and adaptation strategies in the fields of
Sciences, Assam University, Diphu Campus, Diphu, Assam,
India hydrology, environmental engineering, agriculture, and
2 forestry [6]. The word diversity significantly defines India
Department of Humanities and Social Sciences, Indian
Institute of Technology Kharagpur, Kharagpur, West Bengal, whether it is culture, religion, ethnicity, or geography. This
India diverse geographical aspect is a triggering factor for

123
 P. Datta, S. Das

climatic variability across the country. Hingane et al. [7] composition of this region is completely different from the
found pronounced warming trends (1901–1982) over north rest of the state. In the last century, the natural landscape of
central, north-eastern and interior peninsula regions of North Bengal has changed to a great extent while the
Indian landmass. In contrast, Srivastava et al. [8] found emission of Greenhouse Gases (GHGs) and Land Use and
decreasing trends in northern parts (north to 23N) and Land Cover (LULC) changes were marked as major con-
increasing trends in southern parts (south to 23N) of the tributors to the warming during the twentieth century [16].
country. Additionally, Kumar et al. [9] reported that The changes in LULC took place through three parallel
increase in average temperature is solely due to the processes viz. exploitation of forest resource, expansion of
increase in maximum temperature, particularly in winter agricultural land, and population influx. Establishment of
and post-monsoon seasons. Whereas minimum temperature tea gardens, army cantonments, settlements, railways,
records remained trendless during 1901–1987. Similarly, highways as well as extraction of forest wood for timber
Arrora et al. [5] also found strong warming trends in winter and lumber, caused the deterioration of natural landscape.
and post monsoon, but falling trend in the pre-monsoon In consequence, deforestation, forest fragmentation fol-
season. However, considerable acceleration in both maxi- lowed by man-animal conflicts has become the major
mum and minimum temperature trend was noticed by environmental threat in North Bengal [17–20]. The
Kothawale and Kumar [10] precisely from the year of expansion of agricultural lands and settlements completely
1971–2003. So, India has already witnessed the evidence of rely on the growing population pressure (both due to the
changing climate with extensive regional and seasonal natural growth as well as immigration especially from
diversities. The review of past studies regarding tempera- Bangladesh during the independence of India and 1971
ture trends in India, reveals three significant drawbacks. liberation war of Bangladesh). With the documentation of
Firstly, several studies [7, 9, 11] analyzed trends consid- such remarkable LULC changes, how the climate changed
ering seven major temperature homogeneous regions as in the twentieth century is still not investigated precisely.
provided by the Indian Institute of Tropical Meteorology Moreover, the rate and duration of warming in the twen-
(IITM). However, incorporating the entire West Bengal tieth century has reported been larger than any other time
including Bihar, Jharkhand, and parts of Orissa state into during the last millennium by IPCC [21]. Thus, the present
the North-East (NE) homogeneous zone seems a debat- study intends to analyze the long-term district-level sea-
able issue as diversities concerning physiographic and sonal and annual trends as well as their approximate
climatic aspects usually observed between the eight NE potential trend turning points in surface air temperature
states1 and the rest of the region. Consequently, in terms of (ST) data of a longer period of 102 years (1901–2002).
climatic parameters, it can manifest both inter and intra- Analysis of district-wise trend is essential in Indian sce-
regional variations at different spatio-temporal scale. Sec- nario as it is an important administrative unit for practical
ondly, studies [5, 9, 10, 12] either used simple linear implementation of various adaptation and mitigation
regression or the classic Mann–Kendall (MK) test without strategies. The findings of this study will add a solid
considering the effect of autocorrelation. Both can produce understanding of last century’s temperature warming at the
erroneous results, as the parametric test assumes a normal regional scale in North Bengal.
distribution and the MK test assumes a random distribu-
tion. Such conditions are rare in hydro-climatic observa-
tions. Last but not least, many researchers [13–15] have 2 Description of the study area
already explored the long-term precipitation trend, whereas
the single most crucial climatic parameter, i.e., the tem- North Bengal consists of six districts, namely: Darjeeling,
perature is not studied extensively. These shortcomings Jalpaiguri, Cooch Behar, Uttar Dinajpur, Dakshin Dinajpur
motivated us to revise the last century’s temperature trends and Malda (Fig. 1). The region extends from 24390 2700 N
especially when the understanding of twentieth century’s to 27130 14‘‘N and 8745’54’’E to 89520 5800 E, covering
warming is not well documented in many ecologically approximately an area of 21,859 km2. Out of total geo-
sensitive areas of the country including the North Bengal graphical area, 3089 km2 area is covered with forests [22].
region. The Northern part of the region comprises lofty hills
The northern part of the Indian state of West Bengal is (Darjeeling Himalaya) while the immediate foothills
popularly known as North Bengal. This division is not (southern parts of Darjeeling, entire Jalpaiguri district,
based on any natural or administrative purpose, but the northern parts of Cooch Behar and Uttar Dinajpur) consists
physical structure, economic and socio-cultural of a wavy rugged topography which is known as Terai and
Dooars. Rest of the areas are characterised by monotonous
1
NE states: Assam, Meghalaya, Arunachal Pradesh, Mizoram, plain [23]. As per the recent climatic regionalisation based
Nagaland, Tripura, Manipur, and Sikkim. on the Köppen-Geiger world climatic scheme and updated

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Fig. 1 Geographical location of the study area

by Peel et al. [24], the entire region comes under three (http://www.indiawaterportal.org/metdata). Numerous
climatic divisions such as Cwb2 (Darjeeling and upper part studies [14, 26–31] used these datasets to detect tempera-
of Jalpaiguri district), Cwa3 (Cooch Behar, lower part of ture and precipitation trend. However, they misguided the
Jalpaiguri, and Uttar Dinajpur district), and Aw4 (Dakshin actual data source of IWP by mentioning it as a sister web
Dinajpur and Malda district). The hilly parts, Terai and portal of India Meteorological Department (IMD). Further,
Dooars region experience heavier rainfall owing to its rich they interpolated the district-level test results as the station
vegetation cover, orographic effects and the northward or point specific which are completely ambiguous. These
shift of monsoon trough during the break phase of monsoon district-wise climatic data of IWP were actually obtained
[23, 25]. The maximum and minimum temperature records by simple linear averaging from the gridded CRU TS 2.1
are observed in the month of April and January respec- datasets. This gridded data were constructed from the
tively. The region in comparison with its southern kin is observatory networks spread across the globe. Prior grid-
still backward. With the increasing population, agriculture ding, data were homogenized using a modified version of
and allied activities continues to be the prime source of the Global Historical Climatology Network (GHCN)
economy. Apart from this, poor knowledge about manu- method to remove errors due to changes in observation
facturing activities, the absence of mineral resources, lack practice, station environment, or other non-climatic factors.
of trained man-power, fragile marketing and infrastructural More details about these homogenization and gridding
facilities restricted the industrial growth [23]. approaches are well documented in the work of Mitchell
and Jones [32]. Furthermore, a significant correlation
between the IMD station observations CRU datasets for
3 Materials and methods temperature has already been reported in the studies of
Robertson et al. [33] and Piyoosh and Ghosh [34].
3.1 Database The data series are continuous, without any missing
values. From the monthly data, using Ms. Excel Spread-
Monthly maximum and minimum temperature data of 6 sheet 2016, the mean maximum temperature (Tmax) and
districts covering a period of 102 years (1901–2002) were mean minimum temperature (Tmin) were computed annu-
downloaded from the website of India Water Portal (IWP) ally and seasonally. The annual series was considered as
the occurrence of climatic events may take a longer period,
2
C represents temperate, w for dry winter, and b for warm summer. such as multi-year and multi-decade. In addition, seasonal
3
a represents hot summer. data were also incorporated as several studies [35, 36] have
4
A represents tropical, w for savannah. already observed that changes in temperature trends do not

123
 P. Datta, S. Das

solely arise in annual scale but also within the different 3.3 Theil-Sen’s slope estimator
seasons of a specific year. It is also evident that a signifi-
cant positive trend of an annual series may show non-sig- The slope or magnitude of the trend is identified using a
nificant negative trends in some seasons [36, 37]. The non-parametric procedure. It was first developed by Theil
seasonal scales were adopted as defined by IMD [38], [42] and later modified by Sen [43], thereby Theil-Sen’s
namely: pre-monsoon (March–May), monsoon (June– slope estimator. This method is not greatly affected by
September), post-monsoon (October–November) and win- gross data errors or outliers and missing values [41]. The
ter (December–February). slope estimates (Q) of N data pairs is computed as:
xj  xk
Qi ¼ for i ¼ 1; 2; . . .N ð5Þ
3.2 Mann–Kendall test jk

A non-parametric Mann–Kendall (MK) test [39, 40] was where (j [ k), Qi= slope between data points xj and xk ,
used to analyze trends in long-term surface temperature xj = data measurement at time j, xk = data measurement at
data. The test is suitable as the data need not follow any time k; and N is the number of data pairs.
particular distribution. It uses the relative magnitudes of The median of N values of Q is Sen’s estimator of the
the data rather than their measured values [41]. MK test slope. If there is only one datum in each timespan, then
statistic is calculated from the sum of the signs of the N ¼ nðn  1Þ=2; where n is the number of time scale. If
slopes. The MK statistic, S is: there are multiple observations in one or more timespan,
then N\nðn  1Þ=2; where n is the total number of
X
n1 X
n
observations not the number of timespans. The N values of
S¼ sgnðxj  xk Þ ð1Þ
k¼1 j¼kþ1
Qi are ranked from smallest to largest, and the median of
slope or Sen’s slope estimator is computed as follows:
where (j [ k), n is the number of data points; xj is the jth 8
< Q½ðNþ1Þ=2 if N is odd
observations; xk is the kth observations.
  Qmed ¼ QðN=2Þþ½ðNþ1Þ=2 ð6Þ
The sgn xj  xk is an indicator function that takes on : if N is even
  2
the values 1, 0, or - 1 according to the sign of xj  xk :
8   The Qmed sign indicates data trend reflection, while its
  < 1 if xj  xk [ 0 value indicates the steepness of the trend.
sgn xj  xk ¼ 0 if xj  xk  ¼ 0 ð2Þ
:
1 if xj  xk \0
3.4 Test of autocorrelation
When, n C 10, the statistic S is approximately normally
distributed with the mean E(S) = 0 and the variance is: The presence of positive autocorrelation in the time series
P enhances the chance of detecting trends when there is no
nðn  1Þð2n þ 5Þ  pj¼1 ðtj  1Þð2tj þ 5Þ
varðsÞ ¼ ð3Þ trend, and vice versa [44, 45]. Thus, all the surface tem-
18 perature data series are first tested for autocorrelation using
where, p is the number of the tied groups in the dataset; tj is the autocorrelation coefficient (Pk) at 95% confidence
the number of data points in the jth tied group. interval. The following formula was used [46] to compute
Then, the S and VarðSÞ are used to compute the test the autocorrelation coefficient (Pk) of the discrete time
statistic Z, which is stated as follows: series for lagk.
8
> S1 nP
k
>
> if S [ 0 ðxt  xÞðxtþk  xtþk Þ
>
< ½VarðsÞ1=2 t¼1
Z¼ 0 if S ¼ 0 ð4Þ pk ¼  1=2 ð7Þ
> nP
k nP
k
>
> Sþ1 2
ðxt  xÞ  ðxtþk  xtþk Þ 2
>
: if S\0
½VarðsÞ1=2 t¼1 t¼1

The trend is considered as decreasing if Z is negative where xt and Var (xt ) is considered as the sample mean and
and increasing if Z is positive. In this analysis, 95% and sample variance of the first (n - k) terms respectively.
99% confidence intervals were used. The null hypothesis Whereas, xtþk and Var (xt þ k) are the sample mean and
(H0) of no trend, is rejected if |Z| [ 1.960 and |Z| [ 2.576, sample variance of the last (n - k) terms respectively.
respectively. Further, the hypothesis of serial independence is tested by
the lag-1 autocorrelation coefficient, as H0:P1 = 0 against
H1: |P1| [ 0 using:

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

sffiffiffiffiffiffiffiffiffiffiffiffiffi
3.6 Sequential Mann–Kendall test
n2
t ¼ j p1 j ð8Þ
1  p21
A time series may contain numerous fluctuations in the
where the test statistic t has a student’s t-distribution with trend (non-significant positive or negative) over the
(n - 2) degrees of freedom. If |t| C ta/2, the null hypothesis investigation period. These can be detected by using the
about serial independence is rejected at the significance sequential version of MK test [50]. In the present study,
level a (here 0.05). Sequential Mann–Kendall (SQMK) test proposed by
Sneyers [51] was applied to detect the beginning of trend
3.5 Modified Mann–Kendall test change in the surface temperature time series. It calculates
a progressive u(t) and retrograde u’(t) series which are the
To eliminate the autocorrelation from time series, von standardized variables with zero mean and unit standard
Storch and Navarra [47] recommended pre-whitening deviation. So its sequential behavior fluctuates around the
before applying the MK test. However, pre-whitening can zero levels [11]. The points where these two lines intersect
underestimate the presence of trends in the time series [4]. each other are considered as the approximate potential
Also, it is not exceptional to find significant lag-1 auto- trend turning points. If they cross each other and diverge
correlation even after the pre-whitening [48]. In this study, beyond the 95% confidence limit, then there is a statisti-
Modified Mann–Kendall (MMK) test [44, 49] was used for cally significant change point. The following steps were
trend detection of an autocorrelated series. MMK test is applied to calculate u(t) and u0 ðtÞ:
based on the modified variance of (V*(S)). 1. The values of original series Xi are replaced by their
n ranks Yi and arranged in ascending order. The magni-
V  ðSÞ ¼ VarðSÞ  ð9Þ
nS tudes of Yi (i = 1,…n) were compared to Yj,
where Var(S) is the same used in the MK test. (j = 1…,i - 1). For each comparison, the number of
The autocorrelation between the ranks of the observa- cases Yi [ Yj was counted and denoted by ni.
tions, Pk, is first evaluated. The values of Pk were calcu- 2. The test statistics t was then calculated by the equation:
lated after subtracting a suitable non-parametric trend X
i

estimator, i.e., Sen’s slope estimator from the data. ti ¼ ni ð11Þ

1
Insignificant values of Pk will adversely affect the accuracy
of the variance of S. Therefore, only significant Pk values 3. The mean and variance of the test statistics are:
are used to calculate nnS . tðt  1Þ
Eðti Þ ¼ and Varðti Þ ¼ ½iði  1Þð2i þ 5Þ=72
n 2 4
¼1þ
nS nðn  1Þðn  2Þ ð12Þ
X
n1
 ðn  kÞðn  k  1Þðn  k  2Þpk ð10Þ 4. The sequential values of the statistics u(t) were then
k¼1 calculated as:
where n is the actual number of observations, n*S is con- ti  Eðti Þ
uðti Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð13Þ
sidered as an ‘effective’ number of observations to account Varðti Þ
for autocorrelation in the data and Pk is the autocorrelation
function of the ranks of the observations. According to Similarly, the values of u0 ðtÞ were computed, starting
Hamed and Rao [44], without any loss of power, the from the end of the series and then plotted using OriginPro
accuracy of the MMK test, in terms of the empirical sig- 9.4 to figure out approximate trend turning points. On the
nificance level is more robust to that of the original MK other hand, ArcGIS 10.2.1 was used to generate and
trend test. However, Rao et al. [49] found only first three visualize the spatial concentration of trend magnitudes and
significant autocorrelation coefficients are enough for the their variations in each season over the entire North
variance correction approach as the power of MK test is no Bengal.
longer affected by autocorrelation after a lag of three.
Therefore, in the present study, the first three significant
lags were considered for the variance correction approach.

123
 P. Datta, S. Das

4 Results and discussion season, and lowest Tmin was 7.85 C at Darjeeling district
in the winter season during the study period. The standard
4.1 Preliminary analysis of annual and seasonal deviation of Tmax and Tmin varied between 0.98 C (in pre-
temperature data monsoon at Malda) to 0.32 C (in the monsoon at Cooch
Behar) and 0.81 C (in post monsoon at Malda) to 0.34 C
The statistical analysis (mean, standard deviation, coeffi- (in the monsoon at Jalpaiguri) respectively. The skewness
cient of variation, coefficient of skewness and coefficient of which is a measure of the asymmetry of the probability
kurtosis) of the maximum (Tmax) and minimum (Tmin) distribution from its mean ranges from 0.62 to - 0.34 and
annual and seasonal temperature data series have been from 0.60 to - 1.62 for Tmax and Tmin data series respec-
computed for each district (Tables 1, 2). Highest Tmax was tively. Predominantly positive skewness was found in Tmax,
38.05 C at Dakshin Dinajpur district in pre-monsoon with an average of around 0.18 and in case of Tmin, data

Table 1 General statistical characteristics of Tmax data at six districts of North Bengal from 1901 to 2002
District Variable Mean Standard Coefficient of Coefficient of Coefficient of Seasonal and Seasonal and
(C) deviation (C) variation (%) skewness kurtosis annual Tmax annual Tmin
(oC) (oC)
Tmax Year Tmin Year

Cooch Pre-monsoon 32.36 0.86 2.66 - 0.01 - 0.47 34.30 1999 30.38 1990
Behar Monsoon 31.93 0.32 1.00 0.17 0.04 32.82 1983 31.13 1954
Post 29.79 0.68 2.27 0.27 - 0.30 31.43 1915 28.31 1967
monsoon
Winter 25.32 0.64 2.52 0.36 0.15 26.86 1960 23.75 1905
Annual 30.03 0.38 1.26 0.33 0.29 31.2 1999 29.1 1905
Dakshin Pre-monsoon 35.36 0.98 2.77 - 0.08 0.16 38.05 1972 32.98 1981
Dinajpur Monsoon 32.90 0.36 1.10 0.12 1.04 33.98 1958 31.75 1971
Postmonsoon 30.46 0.59 1.95 0.37 - 0.43 32.05 1915 29.18 1934
Winter 26.13 0.69 2.65 0.62 0.43 28.38 1963 24.51 1905
Annual 31.42 0.41 1.30 0.13 0.32 32.60 1958 30.45 1971
Darjeeling Pre-monsoon 30.48 0.85 2.78 - 0.04 - 0.36 32.84 1999 28.56 1905
Monsoon 30.50 0.33 1.10 0.48 1.26 31.69 1983 29.72 1954
Post 28.41 0.77 2.70 0.20 0.03 30.38 1958 26.23 1968
monsoon
Winter 23.02 0.63 2.74 0.19 0.22 24.59 1993 21.37 1905
Annual 28.28 0.37 1.30 0.07 - 0.12 29.36 1999 27.36 1905
Jalpaiguri Pre-monsoon 31.72 0.84 2.65 0.02 - 0.39 34.03 1999 29.83 1905
Monsoon 31.55 0.32 1.01 0.34 0.50 32.60 1983 30.76 1954
Postmonsoon 29.52 0.75 2.56 0.09 - 0.02 31.51 1958 27.54 1968
Winter 24.80 0.64 2.57 0.26 0.21 26.33 1960 23.23 1905
Annual 29.57 0.37 1.25 0.20 0.20 30.71 1999 28.66 1905
Malda Pre-monsoon 35.50 0.98 2.77 - 0.33 0.29 37.78 1958 32.73 1981
Monsoon 32.58 0.43 1.31 0.33 1.18 33.78 1974 31.32 1971
Postmonsoon 30.12 0.62 2.07 0.26 0.53 31.60 1915 28.74 1971
Winter 25.75 0.72 2.78 0.45 0.17 27.74 1963 24.11 1905
Annual 31.19 0.44 1.40 - 0.19 0.84 32.46 1958 29.81 1971
Uttar Pre-monsoon 34.24 0.92 2.67 - 0.34 - 0.19 35.96 1972 31.88 1981
Dinajpur Monsoon 32.36 0.38 1.16 0.18 1.06 33.51 1983 31.22 1971
Post 30.00 0.63 2.09 0.47 - 0.46 31.57 1915 28.87 1971
monsoon
Winter 25.18 0.68 2.70 0.49 0.24 27.16 1963 23.50 1905
Annual 30.64 0.40 1.32 - 0.13 0.47 31.74 1958 29.48 1971

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Table 2 General statistical characteristics of Tmin data at six districts of North Bengal from 1901 to 2002
District Variable Mean Standard Coefficient of Coefficient of Coefficient of Seasonal & Seasonal &
(C) deviation (C) variation (%) skewness kurtosis annual Tmax annual Tmin
(C) (C)
Tmax Year Tmin Year

Cooch Pre- 20.13 0.74 3.66 - 0.33 - 0.25 21.54 1999 18.22 1905
Behar monsoon
Monsoon 25.57 0.35 1.38 0.20 0.05 26.56 2001 24.60 1972
Post 19.47 0.74 3.79 0.35 0.15 21.50 1998 17.85 1954
monsoon
Winter 11.57 0.70 6.04 0.60 0.78 13.75 1999 9.89 1905
Annual 19.69 0.43 2.18 0.49 0.62 20.97 1998 18.72 1905
Dakshin Pre- 21.49 0.76 3.52 - 0.45 - 0.21 22.86 1922 19.39 1905
Dinajpur monsoon
Monsoon 25.73 0.43 1.66 - 0.76 1.46 26.55 1998 24.25 1972
Post 19.96 0.78 3.90 0.26 0.77 22.25 1998 17.89 1972
monsoon
Winter 12.32 0.74 5.99 0.41 0.09 14.50 1999 10.61 1905
Annual 20.36 0.46 2.72 0.01 0.26 21.42 1999 19.09 1972
Darjeeling Pre- 18.08 0.74 4.11 - 0.19 - 0.09 20.00 1999 16.14 1905
monsoon
Monsoon 23.30 0.36 1.54 - 0.32 1.64 24.23 2001 21.92 1972
Post 17.18 0.75 4.34 0.27 0.20 19.14 1998 15.16 1967
monsoon
Winter 9.57 0.68 7.09 0.30 0.39 11.35 1990 7.85 1905
Annual 17.54 0.41 2.33 0.32 0.44 18.69 1999 16.56 1905
Jalpaiguri Pre- 19.64 0.74 3.77 - 0.22 - 0.16 21.47 1999 17.74 1952
monsoon
Monsoon 24.84 0.34 1.39 0.15 0.33 25.81 2001 23.75 1972
Post 18.89 0.76 4.02 0.26 - 0.06 20.87 1998 17.21 1967
monsoon
Winter 11.29 0.69 6.15 0.51 0.66 13.39 1988 9.63 1905
Annual 19.16 0.42 2.19 0.44 0.54 20.38 1999 18.20 1905
Malda Pre- 20.80 0.80 3.82 - 0.56 - 0.05 22.13 1966 18.58 1905
monsoon
Monsoon 24.73 0.49 1.97 - 1.62 4.74 25.59 1998 22.71 1972
Post 18.49 0.81 4.38 0.36 0.95 20.80 1998 16.17 1972
monsoon
Winter 10.91 0.75 6.87 0.21 0.08 12.97 1999 9.01 1974
Annual 19.25 0.49 2.56 - 0.52 1.35 20.29 1958 17.64 1972
Uttar Pre- 19.89 0.76 3.84 - 0.50 - 0.18 21.22 1922 17.76 1905
Dinajpur monsoon
Monsoon 24.85 0.45 1.82 - 1.27 3.71 25.64 1937 22.95 1972
Post 18.33 0.78 4.23 0.33 0.76 20.53 1998 16.25 1972
monsoon
Winter 10.27 0.70 6.86 0.31 0.24 12.26 1999 8.53 1905
Annual 18.88 0.46 2.42 - 0.14 0.46 19.93 1999 17.55 1972

series was negatively skewed with an average around and Tmin records. The coefficient of variation (CV) was
- 0.04. The coefficient of kurtosis, a measure for the performed to compare the degree of variation from one
degree of tailedness in the variable distribution varies from administrative unit to another. The CV in Tmax ranges from
1.26 to - 0.47 and from 4.74 to - 0.25 both for the Tmax 2.78% (Darjeeling) to 2.65% (Jalpaiguri) for pre-monsoon,

123
 P. Datta, S. Das

1.31% (Malda) to 1.00% (Cooch Behar) for monsoon, 4.2 Analysis of annual and seasonal temperature
2.70% (Darjeeling) to 1.95% (Dakshin Dinajpur) for post trend
monsoon, 2.78% (Malda) to 2.52% (Cooch Behar) for
winter and 1.40% (Malda) to 1.25% (Jalpaiguri) for annual The magnitude of the trends present in the time series,
scale. On the other hand, CV for Tmin varied from 4.11% determined using the Theil-Sen’s slope estimator, is given
(Darjeeling) to 3.52% (Dakshin Dinajpur) for pre-mon- in Table 3. The annual long-term temperature (Tmax and
soon, 1.97% (Malda) to 1.38% (Cooch Behar) for mon- Tmin) trends during the analysis period (1901–2002) show
soon, 4.38% (Malda) to 3.79% (Cooch Behar) for post significant warming trends. The increase of annual Tmin is
monsoon, 7.09% (Darjeeling) to 5.99% (Dakshin Dinajpur) faster than Tmax across the entire region. This result is
for winter and 2.56% (Malda) to 1.25% (Jalpaiguri) for the slightly in disagreement with the findings of Kothawale
annual scale. The results of CV confirmed that Darjeeling et al. [10] who reported non-significant warming trend of
and Malda are receiving more variabilities in terms of both Tmin since 1901–2003 in the north-eastern region. How-
Tmax and Tmin than the rests. ever, from 1971 they found that Tmin significantly
increased faster than Tmax. Interestingly, the annual tem-
perature trend is almost similar in all the districts. Hence

Table 3 Details of Theil-Sen’s

District Variable TS slope for Tmax TS slope for Tmin
slope estimation at annual and
seasonal scale for Tmax and Tmin Cooch Behar Pre-monsoon 0.007 0.009**
records
Monsoon 0.003* 0.003
Post monsoon 0.011** 0.009**
Winter 0.008** 0.012**
Annual 0.006** 0.007**
Dakshin Dinajpur Pre-monsoon 0.002** 0.004
Monsoon 0.003 0.003
Post monsoon 0.012** 0.011**
Winter 0.010** 0.013**
Annual 0.006** 0.007**
Darjeeling Pre-monsoon 0.007* 0.008**
Monsoon 0.002 0.002
Post monsoon 0.009** 0.008**
Winter 0.008** 0.010**
Annual 0.006** 0.006**
Jalpaiguri Pre-monsoon 0.008* 0.009**
Monsoon 0.002 0.002
Post monsoon 0.011** 0.008**
Winter 0.008** 0.011**
Annual 0.006** 0.007**
Malda Pre-monsoon 0.002** 0.004
Monsoon 0.002 0.003
Post monsoon 0.013** 0.011**
Winter 0.010** 0.013**
Annual 0.006** 0.007**
Uttar Dinajpur Pre-monsoon 0.004 0.005*
Monsoon 0.002 0.003
Post monsoon 0.011** 0.010**
Winter 0.009** 0.012**
Annual 0.006** 0.007**
TS Theil-Sen
*Statistically significant trends at 95% significance level
**Statistically significant trends at 99% significance level

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

the spatial plotting is not shown for the annual series. may have negative impacts on its renowned tea and tourism
These annual trend magnitudes are consistent with the industry. The seasonal trends of Tmax and Tmin in different
IPCC’s global scale warming rate of 0.6 ± 0.2 C. It is districts of North Bengal reflect variabilities in trend
evident from the results that the usual place of soothing magnitude. In pre-monsoon season, all the districts except
climate, i.e., Darjeeling and parts of Jalpaiguri are also Cooch Behar and Uttar Dinajpur experience significant
warming at the same magnitude like other districts which increasing trends in Tmax, and it varies between 0.002 to

Fig. 2 Spatial variations of Tmax trend magnitude (C/year) in a pre-monsoon, b monsoon, c post-monsoon and d winter seasons

123
 P. Datta, S. Das

0.008 C per year (Fig. 2a). Jalpaiguri district shows the who reported an upward trend in both the minimum and
highest increase followed by Darjeeling. Whereas, a sig- maximum temperature in the north-eastern part of India
nificant upward trend is present in Tmin throughout the during 1901–2003. However, in the same reported period,
region except for Dakshin Dinajpur and Malda. It varies Kothawale and Kumar [10] observed an increasing trend
between 0.004 to 0.009 C per year (Fig. 3a). Results are only in maximum temperature in the north-east region. The
comparable with the findings of Sonali and Kumar [11], results can be illustrated by the findings of Rai et al. [52],

Fig. 3 Spatial variations of Tmin trend magnitude (C/year) in a pre-monsoon, b monsoon, c post-monsoon and d winter seasons

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Fig. 4 Graphical representation of progressive and retrograde series of pre-monsoon Tmax (a) and Tmin (b) at six districts of North Bengal from
1901 to 2002

123
 P. Datta, S. Das

Fig. 4 continued

who examined the possible relationship between cloud During post monsoon, all the districts show significant
cover and the rise of temperature during all seasons. It was upward trends both for Tmax and Tmin (Figs. 2c, 3c). The
apparent from their study that decreasing cloud cover has highest trend in Tmax is observed at Malda (0.013 C per
strong anti-correlation with Tmax, while a weak correlation year) followed by Dakshin Dinajpur (0.012 C per year),
with Tmin during the pre-monsoon season. and the trend is least in case of Darjeeling district
The trends during monsoon season in Tmax and Tmin (0.009 C per year). Similarly, the highest increase of Tmin
indicate a general pattern of rising temperatures at maximum (0.011 C per year) is observed at Malda and Dakshin
districts, albeit not significant (Figs. 2b, 3b). However, Dinajpur and lowest (0.008 C per year) for Darjeeling and
monsoon Tmax has significantly increased only in Cooch Jalpaiguri district. Tmax is increasing faster than Tmin dur-
Behar at 0.003 C per year. Kothawale and Kumar [10] ing the post-monsoon period. This finding agreed with the
reported significant increasing trend present only in the study of Rai et al. [52]. A substantial increase in post
monsoon Tmax over the entire north-eastern region during the monsoon temperature was also observed by previous
same period. Generally, the monsoon season is associated studies in India [53, 54]. It can be noted that the annual rise
with cloudiness and rainfall, which tends to reduce the in Tmax and Tmin is mainly attributed due to upward trends
temperature. Mondal et al. [53], linked net deficit of rainfall present in the post-monsoon followed by the winter season.
with increasing temperature in India. It was apparent from A similar type of result was observed in the study of Dash
their study that the annual rainfall trend magnitude decreased et al. [55]. They also reported that such trends in temper-
and there was an increase in the mean temperature for India ature rise might change the spatial position of heating over
as a whole. The previous study of Chatterjee et al. [14] also the Indian region and hence capable of altering atmo-
indicated the overall presence of the decreasing trend of spheric circulation patterns. Interestingly, the highest
monsoon rainfall in the districts of North Bengal, although increase in Tmax is also observed in the post-monsoon
most of them were not significant. Thus, the monsoonal period in the study area.
decrease in rainfall and rising temperature are indicating the A significant increase in both Tmax and Tmin is observed
sign of global warming in this region. in all the districts for the winter season (Figs. 2d, 3d). It is

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Fig. 5 Graphical representation of progressive and retrograde series of monsoon Tmax (a) and Tmin (b) at six districts of North Bengal from 1901
to 2002

123
 P. Datta, S. Das

Fig. 5 continued

also evident in the results of Srivastava et al. [54]. How- monsoon may be attributed to the decreasing cloud cov-
ever, in North Bengal, Tmin is increasing at a faster rate erage [52] followed by rainfall [57]. As the maximum
than Tmax in all the districts throughout the analysis period precipitation occurs in monsoon followed by pre-monsoon
which is opposite to the post-monsoon situation. It proves season, it can dampen out the effect of GHGs during these
that the difference between the Tmax and Tmin is decreasing seasons. It is further supported by the presence of maxi-
and winter months are becoming significantly warmer. mum non-significant trends in monsoon and diverse results
Kothawale et al. [56] mentioned that the ever-increasing of pre-monsoon season. Whereas, the pronounced warming
concentration of GHGs after 1951 (post-industrial era) was in post-monsoon and winter may be due to the effects of
responsible for such warming of Tmin in India. Malda and GHGs. As the study area practically remains cloud free
Dakshin Dinajpur experienced the highest (0.010 C per during post-monsoon and winter, the warming effect of
year) increase in Tmax followed by Uttar Dinajpur GHGs is even more accelerated due to the absence of
(0.009 C per year). Simultaneously the rest of the districts dampening effects of clouds [58]. The rice centric cropping
show a similar increasing pattern with a magnitude of system and increasing population are the two important
0.008 C per year. On the other hand, the rise of Tmin contributors to the GHG emission in North Bengal. The
trends varies from 0.010 to 0.013 C per year across the annual warming trends show that Tmin has increased more
region. than Tmax. It can be associated with urbanization and
It is observed that the southern part of North Bengal agricultural expansion. In praxis, the effect of urbanization
experienced a higher increasing trend of Tmax and Tmin than occurs during the night, when buildings and streets release
the northern parts while the slope of the trend is gradually the absorbed solar heating. Besides, the effect of agricul-
increasing southward in all seasons except pre-monsoon. It tural expansion, i.e., irrigation would increase the heat
perhaps, because the northern parts are closer to the Dar- capacity of the soil thereby increasing Tmin [59].
jeeling Himalayas and typically have a different climatic
condition than the extreme southern districts, i.e., Malda
and Dakshin Dinajpur. The exceptional case of pre-

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Fig. 6 Graphical representation of progressive and retrograde series of post monsoon Tmax (a) and Tmin (b) at six districts of North Bengal from
1901 to 2002

123
 P. Datta, S. Das

Fig. 6 continued

4.3 Analysis of trend turning points beginning of change point on or around 1938. In almost all
the districts a decreasing trend is observed up to 1920 in
The Sequential Mann–Kendall test graphically illustrated monsoon Tmax. Approximate change points are detected at
the progressive and retrograde trends of Tmax and Tmin Dakshin Dinajpur (1937), Darjeeling (1917), Jalpaiguri
series both at seasonal and annual scale in six districts of (1942), Malda (1910), Uttar Dinajpur (1940). Periodic
North Bengal during the period of 1901–2002 (Figs. 4, 5, increasing and decreasing trends with two possible change
6, 7, 8). If the progressive MK value crosses the upper or points in monsoon Tmin temperature are found at Cooch
lower confidence limit, it indicates a significant positive Behar, Dakshin Dinajpur, Jalpaiguri, Malda, and Uttar
and negative trend at 95% significance level respectively Dinajpur. The monsoon Tmin trend started to decrease after
[36]. Considering u(t) statistics, periodic fluctuations in the early 1950s and continued approximately 1987. At this
pre-monsoon Tmax and Tmin are found in all the six districts time, the progressive and retrograde lines came too close to
and Uttar Dinajpur, Dakshin Dinajpur and Malda respec- each other, but they did not cross. Cooch Behar district also
tively. Approximately after 1950, pre-monsoon Tmax star- followed the similar pattern. Although more observations
ted to show possible change points. However, after 1960 all are required for the confirmation as the upward trend
the districts are showing a slight downward trend in pre- started very recently. The u(t) and u0 (t) statistics are almost
monsoon Tmax. Significant pre-monsoon Tmin change point parallel at Darjeeling during the entire period. However,
is observed around 1921 at Uttar and Dakshin Dinajpur u(t) statistics depict an increasing warming trend. It can
districts whereas at Malda such change occurred in 1928. It also be noted that MK Z-value at the end of the study
indicates that the warming trends of Tmin started a long ago period was not significant at all over the region. Perhaps,
and it can be prolonged in future also. this may be due to the fact that significant positive and
An overall warming trend is observed in monsoon Tmax negative trends canceled each other out [36].
in the entire region. In Cooch Behar, a decreasing trend in Considering the progressive line, the post-monsoon Tmax
Tmax can be observed approximately up to the year of 1918. and Tmin show an overall increasing trend. The u(t) and
Later on, it started to increase gradually, and there is a u0 (t) statistics do not intersect each other in any of the

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Fig. 7 Graphical representation of progressive and retrograde series of winter Tmax (a) and Tmin (b) at six districts of North Bengal from 1901 to
2002

123
 P. Datta, S. Das

Fig. 7 continued

districts throughout the time period. However, in the late estimator that, there are variations in the trends at spatial
1940s, u(t) and u0 (t) statistics of the post-monsoon Tmin at scales, but the increase of Tmin and Tmax is predominant
Cooch Behar clearly begin to diverge without crossing over the entire region. Excluding monsoon, the significant
each other indicating upward trend [60]. The progressive upward trend is observed in most of the districts. On the
and retrograde lines do not intersect each other for winter other hand, maximum locations showed much warming
Tmax and Tmin also. Interestingly near 1950, u(t) and trend during the post-monsoon and winter season con-
u0 (t) statistics in Tmax for all the districts except Darjeeling cerning Tmax and Tmin than the monsoon and pre-monsoon.
come too close to each other and diverge. Whereas the Findings of the present study are quite consistent with the
u(t) and u0 (t) statistics remained parallel at Darjeeling and previous macro-level studies. Wherever the dissimilarities
Jalpaiguri districts for the winter Tmin. Considering occurred may be attributed to the presence of regional
u(t) statistics an overall increasing pattern of Tmin trend is variations within a faster climatic sub-system or due to the
observed in all the districts. Significant trend turning points application of different trend detection methods. Addi-
are also absent in case of annual Tmax and Tmin. However, tionally, the SQMK graphs show increasing trends in the
the nonexistence of trend turning points in the time series annual, post-monsoon and winter seasons whereas varied
does not indicate the absence of a trend in the entire series results found in pre-monsoon and monsoon seasons. It
[61]. The reason for such failure to show exact trend makes the finding more intense that the annual temperature
turning points by the Sequential Mann–Kendall method is increase is mostly due to post-monsoon and winter tem-
explained by Nalley et al. [36]. perature rise. It should also be noted that the warming of
seasonal temperature in North Bengal is not attributed to
any single mechanism because of the variation in trends
5 Conclusion and magnitude differs from one district to another. It is also
known that from the twentieth century a significant amount
The study investigated the seasonal and annual trends in of deforestation activities and LULC changes have taken
Tmax and Tmin at North Bengal from 1901 to 2002. It is place in this region. Presence of diverse physiography and
evident from the results of MK test and Theil-Sen’s slope agricultural practices also played some role in determining

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

Fig. 8 Graphical representation of progressive and retrograde series of annual Tmax (a) and Tmin (b) at six districts of North Bengal from 1901 to
2002

123
 P. Datta, S. Das

Fig. 8 continued

the differential warming trends. Results of the study pro- 2. Akinsanola, A. A., & Ogunjobi, K. O. (2014). Analysis of rainfall
vide useful documentation of warming trends in the and temperature variability over Nigeria. Global Journal of
Human-Social Science Research, 14(3), 10–28.
twentieth century which will be valuable for considering 3. IPCC. (2001). Climate change 2001: Synthesis report. In R.
different adaptation and mitigation policies. Future studies T. Watson & The Core Writing Team (Eds.), A contribution of
are required to address the impact of such changes on working groups I, II, and III to the third assessment report of the
natural and human resources as the overall warming may intergovernmental panel on climate change (p. 398). Cambridge:
Cambridge University Press.
have a significant bearing on agriculture, water availability 4. Yue, S., & Hashino, M. (2003). Long term trends of annual and
as well as on human health. monthly precipitation in Japan 1. Journal of the American Water
Resources Association, 39(3), 587–596.
Acknowledgements The authors would like to express their sincere 5. Arora, M., Goel, N. K., & Singh, P. (2005). Evaluation of tem-
gratitude to India Water Portal managed by Arghyam and encouraged perature trends over India/Evaluation de tendances de tempéra-
by National Knowledge Commission, Bangalore, India for providing ture en Inde. Hydrological Sciences Journal, 50(1), 81–93.
the district-level meteorological data. The authors are also grateful to 6. Anandhi, A., Srinivas, V. V., Kumar, D. N., & Nanjundiah, R. S.
the Editor-in-Chief, Dr. Jung-Sup Um and the three anonymous (2009). Role of predictors in downscaling surface temperature to
reviewers for their critical review and constructive suggestions. river basin in India for IPCC SRES scenarios using support vector
machine. International Journal of Climatology: A Journal of the
Compliance with ethical standards Royal Meteorological Society, 29(4), 583–603.
7. Hingane, L. S., Kumar, K. R., & Murty, B. V. R. (1985). Long-
Conflict of interest The authors declared no potential conflicts of term trends of surface air temperature in India. Journal of Cli-
interest with respect to the research, authorship, and/or publication of matology, 5(5), 521–528.
this article. 8. Srivastava, H. N., Dewan, B. N., Dikshit, S. K., Rao, G. P., Singh,
S. S., & Rao, K. R. (1992). Decadal trends in climate over India.
Mausam, 43(1), 7–20.
9. Kumar, K. R., Kumar, K. K., & Pant, G. B. (1994). Diurnal
References asymmetry of surface temperature trends over India. Geophysical
Research Letters, 21(8), 677–680.
1. Jain, S. K., & Kumar, V. (2012). Trend analysis of rainfall and
temperature data for India. Current Science, 102(1), 37–49.

123
Analysis of long-term seasonal and annual temperature trends in North Bengal, India

10. Kothawale, D. R., & Rupa Kumar, K. (2005). On the recent 30. Ghosh, K. G. (2018). Analysis of rainfall trends and its spatial
changes in surface temperature trends over India. Geophysical patterns during the last century over the Gangetic West Bengal,
Research Letters, 32(18), L18714. Eastern India. Journal of Geovisualization and Spatial Analysis,
11. Sonali, P., & Kumar, D. N. (2013). Review of trend detection 2(2), 15.
methods and their application to detect temperature changes in 31. Kundu, S., Khare, D., Mondal, A., & Mishra, P. K. (2015).
India. Journal of Hydrology, 476, 212–227. Analysis of spatial and temporal variation in rainfall trend of
12. Kothawale, D. R., & Kumar, K. R. (2002). Tropospheric tem- Madhya Pradesh, India (1901–2011). Environmental Earth Sci-
perature variation over India and links with the Indian summer ences, 73(12), 8197–8216.
monsoon: 1971–2000. Mausam, 53(3), 289–308. 32. Mitchell, T. D., & Jone, P. D. (2005). An improved method of
13. Subash, N., & Sikka, A. K. (2014). Trend analysis of rainfall and constructing a database of monthly climate observations and
temperature and its relationship over India. Theoretical and associated high resolution grids. International Journal of Cli-
Applied Climatology, 117(3–4), 449–462. matology, 25(6), 693–712.
14. Chatterjee, S., Khan, A., Akbari, H., & Wang, Y. (2016). 33. Robertson, A. W., Bell, M., Cousin, R., Curtis, A., & Li, S.
Monotonic trends in spatio-temporal distribution and concentra- (2013). Online tools for assessing the climatology and pre-
tion of monsoon precipitation (1901–2002), West Bengal, India. dictability of rainfall and temperature in the Indo-Gangetic Plains
Atmospheric Research, 182, 54–75. based on observed datasets and seasonal forecast models. CGIAR
15. Guhathakurta, P., & Rajeevan, M. (2008). Trends in the rainfall Research Program on Climate Change, Agriculture and Food
pattern over India. International Journal of Climatology, 28(11), Security (CCAFS). Working paper no. 27. Cali, Colombia: https://
1453–1469. cgspace.cgiar.org/handle/10568/28992. Accessed 12 September
16. Basha, G., Kishore, P., Ratnam, M. V., Jayaraman, A., Kouchak, 2018.
A. A., Ouarda, T. B., et al. (2017). Historical and projected 34. Piyoosh, A. K., & Ghosh, S. K. (2016). A comparative assess-
surface temperature over India during the 20th and 21st century. ment of temperature data from different sources for Dehradun,
Scientific Reports, 7(1), 2987. Uttarakhand, India. Journal of Meteorological Research, 30(6),
17. Lahiri Choudhury, D. K. (1975). Report on elephant movement 1019–1032.
and depredation in Jalpaiguri Division and part of Madarihat 35. Zhang, X., Vincent, L. A., Hogg, W. D., & Niitsoo, A. (2000).
Range of Cooch Behar Division in June-July, 1975. Submitted to Temperature and precipitation trends in Canada during the 20th
West Bengal Government, p. 60. century. Atmosphere-Ocean, 38(3), 395–429.
18. Roy, M. (2010). Habitat use and foraging ecology of the Asian 36. Nalley, D., Adamowski, J., Khalil, B., & Ozga-Zielinski, B.
elephant (Elephas maximus) in Buxa Tiger Reserve and adjoining (2013). Trend detection in surface air temperature in Ontario and
areas of northern West Bengal. Doctoral dissertation, Vidyasagar Quebec, Canada during 1967–2006 using the discrete wavelet
University, West Bengal, India. transform. Atmospheric Research, 132, 375–398.
19. Das, K. (2015). Man elephant conflicts in North Bengal. Teri 37. Karaburun, A., Demirci, A., & Kara, F. (2011). Analysis of
University. http://www.teriuniversity.ac.in. Accessed 14 spatially distributed annual, seasonal and monthly temperatures
September 2018. in Istanbul from 1975 to 2006. World Applied Science Journal,
20. Vyas, P., & Sengupta, K. (2014). Human-Leopard conflict in 12(10), 1662–1675.
North Bengal, India. Tigerpaper: Regional Quarterly Bulletin on 38. Rathore, L. S., Attri, S. D., & Jaswal, A. K. (2013). State level
Wildlife and National Parks Management, 41(1), 1–6. climate change trends in India. Government of India Ministry of
21. IPCC Third Assessment Report—Climate Change. (2001). Earth Sciences, Earth System Science Organisation, India
Working group I: Technical summary. Geneva: WMO and Meteorological Department.
UNEP. 39. Mann, H. B. (1945). Nonparametric tests against trend. Econo-
22. Annual Report of West Bengal Forest Department, 2014–15. metrica, 13(3), 245–259.
http://www.westbengalforest.gov.in. Accessed 12 September 40. Kendall, M. G. (1975). Rank correlation methods. London:
2018. Charles Griffen.
23. Biswas, A. (2011). Prospect of eco tourism in North Bengal a 41. Gilbert, R. O. (1987). Statistical methods for environmental
case study of Dooars region Doctoral dissertation, University of pollution monitoring. New York: Wiley.
Calcutta, West Bengal, India. 42. Thiel, H. (1950). A rank-invariant method of linear and polyno-
24. Peel, M. C., Finlayson, B. L., & McMahon, T. A. (2007). mial regression analysis, part 3. Proceedings of Koninalijke
Updated world map of the Köppen-Geiger climate classification. Nederlandse Akademie van Weinenschatpen A, 53, 1397–1412.
Hydrology and Earth System Sciences Discussions, 4(2), 43. Sen, P. K. (1968). Estimates of the regression coefficient based on
439–473. Kendall’s tau. Journal of the American Statistical Association,
25. Dastidar, A. G., Ghosh, S., De, U. K., & Ghosh, S. K. (2010). 63(324), 1379–1389.
Statistical analysis of monsoon rainfall distribution over West 44. Hamed, K. H., & Rao, A. R. (1998). A modified Mann-Kendall
Bengal, India. Mausam, 61(4), 487–498. trend test for autocorrelated data. Journal of Hydrology,
26. Duhan, D., & Pandey, A. (2013). Statistical analysis of long term 204(1–4), 182–196.
spatial and temporal trends of precipitation during 1901–2002 at 45. Akinsanola, A. A., & Ogunjobi, K. O. (2017). Recent homo-
Madhya Pradesh, India. Atmospheric Research, 122, 136–149. geneity analysis and long-term spatio-temporal rainfall trends in
27. Duhan, D., Pandey, A., Gahalaut, K. P. S., & Pandey, R. P. Nigeria. Theoretical and Applied Climatology, 128(1–2),
(2013). Spatial and temporal variability in maximum, minimum 275–289.
and mean air temperatures at Madhya Pradesh in central India. 46. Basistha, A., Arya, D. S., & Goel, N. K. (2009). Analysis of
Comptes Rendus Geoscience, 345(1), 3–21. historical changes in rainfall in the Indian Himalayas. Interna-
28. Goyal, M. K. (2014). Statistical analysis of long term trends of tional Journal of Climatology, 29(4), 555–572.
rainfall during 1901–2002 at Assam, India. Water Resources 47. von Storch, H., & Navarra, A. (1995). Analysis of climate vari-
Management, 28(6), 1501–1515. ability (p. 334). New York: Springer.
29. Meshram, S. G., Singh, V. P., & Meshram, C. (2017). Long-term 48. Zhang, Q., Sun, P., Singh, V. P., & Chen, X. (2012). Spatial-
trend and variability of precipitation in Chhattisgarh State, India. temporal precipitation changes (1956–2000) and their
Theoretical and Applied Climatology, 129(3–4), 729–744.

123
 P. Datta, S. Das

implications for agriculture in China. Global and Planetary 56. Kothawale, D. R., Kumar, K. K., & Srinivasan, G. (2012). Spatial
Change, 82, 86–95. asymmetry of temperature trends over India and possible role of
49. Rao, A. R., Hamed, K. H., & Chen, H. L. (2003). Nonstation- aerosols. Theoretical and Applied Climatology, 110(1–2),
arities in hydrologic and environmental time series (p. 362). The 263–280.
Netherlands: Kluwer Academic Publishers. 57. Nair, P. J., Chakraborty, A., Varikoden, H., Francis, P. A., &
50. Makokha, G. L., & Shisanya, C. A. (2010). Trends in mean Kuttippurath, J. (2018). The local and global climate forcings
annual minimum and maximum near surface temperature in induced inhomogeneity of Indian rainfall. Scientific Reports, 8(1),
Nairobi City. Kenya. Advances in Meteorology, 2010, 6. 6026.
51. Sneyers, S. (1990). On the statistical analysis of series of 58. Rao, P. G. (1993). Climatic changes and trends over a major river
observations. Technical note no. 143, WMO No. 725 415, Sec- basin in India. Climate Research, 2, 215–223.
retariat of the World Meteorological Organization, Geneva. 59. Kalnay, E., & Cai, M. (2003). Impact of urbanization and land-
52. Rai, A., Joshi, M. K., & Pandey, A. C. (2012). Variations in use change on climate. Nature, 423(6939), 528.
diurnal temperature range over India: under global warming 60. Nasri, M., & Modarres, R. (2009). Dry spell trend analysis of
scenario. Journal of Geophysical Research: Atmospheres, Isfahan Province, Iran. International Journal of Climatology,
117(D2), 2114. 29(10), 1430–1438.
53. Mondal, A., Kundu, S., & Mukhopadhyay, A. (2012). Rainfall 61. Rahman, M. A., Yunsheng, L., Sultana, N., & Ongoma, V.
trend analysis by Mann–Kendall test: A case study of north- (2018). Analysis of reference evapotranspiration (ET 0) trends
eastern part of Cuttack district, Orissa. International Journal of under climate change in Bangladesh using observed and CMIP5
Geology, Earth and Environmental Sciences, 2(1), 70–78. data sets. Meteorology and Atmospheric Physics, 1–17. https://
54. Srivastava, A. K., Kothawale, D. R., & Rajeevan, M. N. (2017). doi.org/10.1007/s00703-018-0596-3.
Variability and long-term changes in surface air temperatures
over the Indian subcontinent. In Observed climate variability and Publisher’s Note Springer Nature remains neutral with regard to
change over the Indian Region (pp. 17–35). Singapore: Springer. jurisdictional claims in published maps and institutional affiliations.
55. Dash, S. K., Jenamani, R. K., Kalsi, S. R., & Panda, S. K. (2007).
Some evidence of climate change in twentieth-century India.
Climatic Change, 85(3–4), 299–321.

123