Synopsis on Proposed Thesis
Submitted to the
Dean, Post Graduate Studies
By
Shekhar Singh
M.Tech (Agricultural Engineering)
Id. No. 52783
Under the guidance of
Dr. Deepak Kumar
Assistant Professor
Department of Soil and Water Conservation Engineering
College of Technology
Govind Ballabh Pant University of Agriculture and Technology
Pantnagar-263145 (U.S.Nagar)
Uttarakhand, INDIA
SYNOPSIS
Name : Shekhar Singh
Id. No. : 52783
Degree Programme : M.Tech (Agricultural Engineering)
Major : Soil and Water Conservation Engineering
Minor : Nil
Advisor : Dr. Deepak Kumar, Assistant Professor
Department of Soil and Water Conservation Engineering
Thesis Title : Study of rainfall variability using wavelet theory (Tentative)
Advisory Committee
Dr. Deepak Kumar
Assistant Professor
Dept. of Soil & Water Cons. Engg.
(Chairman)
Dr. Anil Kumar Dr. Pankaj Kumar
Professor & Head Assistant Professor
Dept. of Soil & Water Cons. Engg. Dept. of Soil & Water Cons. Engg.
(Member) (Member)
Department of Soil and Water Conservation Engineering
College of Technology
Govind Ballabh Pant University of Agriculture and Technology
Pantnagar-263145 (U.S.Nagar)
Uttarakhand, INDIA
1. INTRODUCTION
India is well known for its unique geography as well as for its geology which influences its climate.
It experiences a heterogeneous climatic because of its vast size. According to Indian Metrological
department (IMD), four type of climates are found and they are winter (January- February),
summer or pre-monsoon (March-May), monsoon or rainy season (July to September) and post-
monsoon or autumn season (October- November). Every season plays its own role. They all are
interlinked with each other. Out of all of these, rainy season is one of them. As per meteorological
data, average annual rainfall over the whole India is about 120 cm and average annual rainfall is
300-650 mm. The rainfall distribution in any region is primarily influenced by its relief features
on Earth. In the north-eastern state of India, they receives rainfall more than 500 cm of annual
rainfall while in the places like that desert is receives rainfall less than 10 cm of annual rainfall.
The plants that we grow the animals that we raise and the need of a human to stay healthy and
hydrated, all these factors depends on rain. Rain is like blood for the existence of all creatures on
Earth. In normal cases, rainwater is used for crop production. Therefore, the first thing which
comes in mind is that whether the amount of rainfall received is adequate or not and whether the
rainfall droplets are well distributed among the crop or not.
Climatic change is a serious problem for today’s scientific community. According to the
study made by The World Meteorological Organization, they found that the period from 2011-
2015 was the hottest period, and out of that 2015 was the hottest. Despite of increase or decrease
in temperature, other factor was also considered viz. Rainfall. Out of all the meteorological
parameter, rainfall is considered to be a major parameter which directly influences the socio-
economic wealth of the country and the people who depend on agriculture (rain-fed agriculture).
India receives, it is broadly divided into heavy rainfall area, moderate rainfall areas, low rainfall
areas and inadequate rainfall areas.
Rainfall type Rainfall(cm)
Heavy rainfall >200
Moderate rainfall 100-200
Low rainfall 60-100
Inadequate rainfall <60
A large degree of variation in the rainfall distribution pattern is found from region to region
and also with respect to time. This change in the pattern of rainfall is termed as “rainfall
variability”. Rainfall have direct impact on agriculture, 60% of Indian agriculture depends on rain.
With a regular pattern in rainfall result to a healthy production of the crop whereas variation in the
pattern of rainfall leads to devastating scenario in the production of crop. Time series analysis is a
statistical approach which deals with the time series data and analysis of trend. Time series data is
a series of data for a given time period or for a particular interval. It is used to forecasting related
to weather, sales and studies related to invention etc. Trend analysis in time series contains the
magnitude of the trend and its significance (statistically). Generally the magnitude of time series
for the time series is determined by using regression analysis or by Sen’s estimator method.
Regression analysis is a parametric method and Sen’s estimator method is a non-parametric
method. The wavelet transform is a recent advance in signal processing that has attracted much
attention since its theoretical development in 1984 by Grossman & Morlet (1984). Its use has
increased rapidly as an alternative to the Fourier transform in preserving local, non-periodic,
multiscaled phenomena, and it has advantages over classical spectral analysis, because it allows
analysis of different scales of temporal variability and it does not need a stationary series (Smith
et al., 1998).
Uttarakhand, is known for its natural beauty. About 83% of the landmass comes under the
hills and rest lives in the belt of Tarai plains. Uneven topography leads to a variation in rainfall in
temporally as well as spatially. This state greatly experience disasters related to the climatological
factor in the recent year. Temperature as well as rainfall varies from place to place. The average
annual of Uttarakhand is found to be 1896 mm and average annual tempature is found to be around
21.80C. A variation of about 200m is found in relief in the southern part of state and for northern
part it is more than 7500 m. it is found that at every rise of 1000 m there is decrease in temperature
by 60C. Slope factor also plays role in determining the climate like the northern part of the slope
much cooler than the southern part. Studies based on temperature and rainfall for data of 100 years
shows a decline in the rainfall rate which has put the area under stress (water resource). The trend
is not uniform in all region. Shortage in rainfall is found in Pithoragarh, Almora, Champawat,
Bageshwar and Nainital district. While the region is experiencing the rise in temperature and the
places are Uttarkashi, Chamoli, Rudraprayag and Pithoragarh. On the other hand, places like
Hardwar, Dehradun and Garhwal district which are found in the low relief areas are coolest places
in the state. There is a large variation in the pattern because of the melting of glaciers and also is
topography. Analysis based on the rainfall series of the state is highly recommended for the
management of resources which is one of the challenging task. Therefore, it is required to first
understand the rainfall distribution pattern over the areas which may cause flood or which may
also cause drought in the areas. Understanding the pattern of rainfall distribution is important for
rainfall frequency analysis, rainfall trend analysis and stochastic modelling.
The proposed study will be done with the following objectives…
1) Statistical analysis of rainfall pattern of selected area of Uttarakhand.
2) Rainfall variability analysis of study area using entropy theory.
3) Study of rainfall variability for rainy days using entropy theory.
4) Study of rainfall using Wavelet Transform technique.
2. REVIEW OF LITERATURE
Mahsin et al. (2011) studied on rainfall modelling time series of rainfall data using Box-
Jenkins (ARIMA) methodology. In this study they have used data from 1981 to 2010 (June) with
the 354 readings and they obtained very good result between actual values and the predicted values
using ARIMA (0 0 1), (0 1 1) model. The results also suggested that the selected model was used
to forecasting the monthly rainfall for upcoming two years.
Longobardi and Villani (2010) studied on annual and seasonal rainfall data from 1918 to
1999 with 211 gauged stations over 25000 km2 area. The findings showed that 76% of data (out
of all dataset) were used to test for trend detection both at annual and seasonal dataset. The annual
data shows negative trend whereas seasonal data represented positive trend and also detected
trends looks very severe during last 30 years.
Wu et al. (2010) developed a model for rainfall forecasting on monthly basis as well as
daily basis. In this study three techniques have been used i.e. Moving average (MA), Principal
Component Analysis (PCA) and Singular Spectral Analysis (SSA) and some forecasting models
also have been used like Linear Regression (LR), K-nearest-Neighbors (K-NN), Artificial Neural
Network (ANN) and Modular Artificial Neural Network (MANN). The results showed that SSA
technique (singular spectral analysis) was better than moving average (MA) or principal
component analysis (PCA) techniques and Modular artificial neural network (MANN) has showed
good results on daily rainfall forecasting if MANN associated with SSA technique.
Chattopadhay and Chattopadhay (2010) developed a univariate model to forecasting the
rainfall of data from 1871 to 1999 using Auto Regressive Integrated Moving Average (ARIMA)
and Auto Regressive Neural Network (ARNN). The results suggested that ARIMA (0,1,1) model
has been indicated very good results between actual and predicted values but ARNN model has
been shown better results as compared to the ARIMA (0,1,1) model.
Cheung et al. (2008) have studied on rainfall data from 1960 to 2002 consisting 134
stations of 13 watershed. In this study interstation & watershed correlation analysis, watershed &
national level analysis, and seasonal & annual analysis have been done. It was found that most of
watersheds were homogeneous and it has been also investigated that no significant change for
annual rainfall data in the analysis of watershed and National level. In contrast annual rainfall
analysis a significant declination was observed but in seasonal rainfall analysis 4 watershed was
declined out of 13 watersheds for the same time interval.
Ellouze et al. (2009) studied on spatial and temporal variability of rainfall data (monthly
and annually basis) from 1930 to 2000 consisting of 20 stations. In this study, performances were
determined by statistical and descriptive analysis. The results suggested that less variability in
trend of rainfall data. It was also found that slight correlation between monthly data but no
relationship between annual data.
Penarrocha et al. (2002) conducted the spatial analysis of rainfall data from 1950 to 2000
in Valencia area of Spain using cluster and principal component analysis (CA and PCA). The study
was performed in two phase. In the first phase, the study has been done for the geographical
distribution of rainfall data from 1971 to 1995 and in the second phase, the study was conducted
for statistical analysis of rainfall data from 1981 to 1990. The study was investigated that three
kind of distribution has been detected by both the methods i.e. PCA and CA and study which was
carried in 1971 to 1995, 800 mm daily rainfall appeared and it was found very severe in nature.
Maruyama et al. (2005) developed a map of the potential water resource availability
(PWRA) of rainfall data from 1697 to 2002, consisting of 20614 stations on worldwide basis using
entropy techniques i.e. Intensity entropy and Apportionment entropy after that they have used
simple clustering and K-means clustering to classify PWRA and also explain the importance of
PWRA. The results has given on the basis of Binary, Tetramerous and Octonary classification. In
the binary classification, both methods have showed same results in the dividing of PWRA and in
the Tetramerous classification, 1st and 3rd points were found more or less PWRA respectively
whereas 2nd and 4th points were not found exactly. In Octonary classification, all eight clusters
were identified in a different frame.
Joshi and Pandey (2011) studied on gridded precipitation data of 1384 rain gauge stations
from 1901-2004 in the all over India i.e. southwest, southeast, central, and northwest regions by
using spectral analysis, Mann-Kendal and linear square fitting techniques. The results suggested
that, in southwest and northwest region, precipitation was found heavy and low respectively
similarly in the southwest and central region, precipitation was average and below average
respectively. It results also suggested that no significant trend was found in the last 10 decadal
analysis which have been done by Mann-Kendal and linear square tests and not much significant
change has examined by spectral analysis method.
Fragoso and Gomes (2008) examined an abundant rainfall pattern and atmospheric pattern
of rainfall data from 1983 to 1998(15 years) consisting of 98 sites in the region of southern Portugal
by using PCA (principal component analysis) and K-Means clustering techniques. All rainfall
patterns have divided into four daily abundant rainfall patterns (DARP) i.e. DARP1, DARP2,
DARP3, DARP4) similarly atmospheric patterns (AP) have also divided into five patterns i.e. AP1,
AP2, AP3, AP4, AP5. The results showed that out of 4 DARP, only two patterns have showed
heavy rainfall. In all patterns of APs, AP4 has indicated depression at very least extent and AP5
and AP1 both have some affinities. The results also suggested that no significant relation was
observed between rainfall and atmospheric pattern.
Cheng et al. (2017) examined the variability of rainfall data form 1961 to 2015 consisting
of 29 metrological stations in the region of Hexi Corridor of the northwest China using entropy
approach. The results demonstrated that annual rainfall was found to be increasing from north to
southwest area and the value of marginal disorder index (MDI) looked higher for the winter
seasons in the month of December and seasons of spring as well that means the variability was
high in the both seasons. The analysis has also be done on decadal basis, in this study the variability
were appeared more from 2001 to 2010 in the month of January , May , June and July.
Santos and Freire (2008) analyzed rainfall data from 1901 to 2010 in the northeast region
of Brazil consisting of nine states using wavelet transform technique. In this study, wavelet spectra
approach had been taken to study the variability in monthly rainfall. The results suggested that a
high concentration of rainfall was observed using wavelet spectra technique. It was also found that
low and high fluctuations were recognized on average fluctution scale which was assumed as a
high wet span. Furthermore, the authors also described the future study which indicated that a
model could be developed for runoff-erosion using stream flow analysis.
Adamowaski et al. (2009) analyzed trend of rainfall from 1911 to 1997 for the Canada
region consisting of five main stations using new approach of continuous wavelet transform. The
study was undertaken for detecting the trend in two ways i.e. short term and long-term change.
The results suggested that, in short term change, trend was observed to be decreasing whereas in
long-term change, trend was increasing in nature.
He et al. (2014) developed an LR model using wavelet to forecast monthly rainfall in the
southern region of Australia. The study was conducted using rainfall data from 1959 to 2008 by
Multi Regression Analysis (MRA) and Multi Linear regression i.e. WMLR and TMLR models.
The study has been done in two phase i.e. 1959 to 1995 and 1996 to 2008. The developed models
were compared with each-other. The results indicated that WMLR model predicted significantly
realistic as compared to the TMLR model and also the value of correlation coefficient in WMLR
model was high. The detection of error varied from 0.11 to 0.46. It was also investigated that, the
obtained results (using MRA-model) were up to the mark.
Adarsh and Reddy (2015) studied trend analysis of rainfall data from 1871 to 2011 using
Linear Regression (LR), Discrete Wavelet Transform (DWT) technique, sequential Mann-Kendall
(MK) test and Sen’s slope estimator methods. The study has been done for four regions of India
i.e. Kerala, Tamil Nadu, Karnataka and Telangana. The results suggested that out of four
subdivisions, there was three subdivisions have increasing trend whereas one subdivisions i.e.
Kerala experienced decreasing trend in annual analysis. The sequential change analysis has also
been done for the given regions. It was investigated that a significant increasing trend was observed
for post-monsoon season of Kerala subdivision in seasonal analysis.
Nalley et al. (2012) detected the trend of rainfall data and stream flow data from 1954 to
2008 using Discrete Wavelet Transform (DWT) technique and non-parametric Mann-Kendall
(MK) test. The study has been done for the region of Quebec and Ontario in Canada and analysis
was based on monthly, annually and seasonal type. The results suggested that intra and inter-
annual analysis play very extreme role for the detecting of trend by wavelet technique and non-
parametric Mann-Kendall (MK) test at 5% level of significance. It was investigated that from 1965
to before 1970, a positive trend has observed.
3. MATREIAL AND METHODS
3.1 Study Area
Uttarakhand is the state of India consisting of 13th districts, having an area of 53483 km2 and it is
divided into two division i.e. Kumaun and Garhwal. The latitudes and longitudes of Uttarakhand
are 30.066o N and 79.019o E respectively and it is bordered by Tibet to north and Nepal to the east.
Fig.1: Study Area
3.2 Data Acquisition
Rainfall monthly data of six rain gauge stations for the period of 116 years from 1901 to 2016 was
obtained from Indian Metrological department (IMD) but during this period i.e. 1901 to 2016,
there was missing values in the dataset.
3.3 Analysis of data
The annual and monsoonal rainfall will be evaluated through monthly rainfall data for the time
period of 116 years from 1901 to 2016. The analysis of data will be analyzed on Entropy theory,
wavelet transform and basic statistical techniques.
3.3.1 Calculation of Entropy
First of all entropy theory was given by Shannon in 1948. Entropy gives the brief knowledge about
the variability or uncertainty of system domain. Entropy can be defined as, it is measures of
variability or randomness in the dataset. For selecting a numerical model and quantify the
disorderness, entropy theory play very extreme role. In last few years, entropy theory is used to
analyze the uncertainty of an event of a random variables. The formula of entropy was as follows…
Entropy H(x)
H(x) =− ∑𝑲
𝒌=𝟏 𝒑(𝒙𝒌 )𝒍𝒐𝒈𝟐 [𝒑(𝒙𝒌 )] …1)
Where k = time interval of K events
xk = Event corresponding to the interval k
𝑝(𝑥𝑘 ) = Probability of 𝑥𝑘
H(x) = Value of Entropy
If the probability of the random variables are same then the value of H would be maximum.
H=0; it means probability of that event is 1 and for all other events, it will be zero. Hence value of
H lies between zero to 𝑙𝑜𝑔2 𝐾 and if the no. of constraints are increasing it means values of entropy
would be decreases.
3.3.2 Wavelet Analysis
Wavelet transform analysis is an advanced technique in signal processing which has gained
attention because of theoretical development by Grossman & Morlet (1984). It is use as an
alternative to Fourier transformation and it is also superior to classical spectral analysis as it allows
to use different scale for analysing the temporal variations and the main advantage is that the use
of stationary series is not required (Smith et al., 1998). Thus, it is appropriate to analyse irregular
distributed events and time series that contain non stationary power at many different frequencies.
The formula of wavelet transform is given by as follows…
𝟏 𝒕−𝒃
WT (b, a)= ∫𝒈( ) 𝒔(𝒕)𝒅𝒕 …2)
√𝒂 𝒂
Where
a = Scaling Factor
b = Time domain
s(t) = Signals at time b
g(t) = Mother wavelet at a and b having values are 0 and 1 respectively
Although, the use of wavelet is not common in the field of hydrology, it is used characterization
of basin (Smith et al, 2002).It maintains the time and frequencies localisation which used in signal
analysing by transforming one-dimensional time series to a diffused two dimensional time-
frequency image at the same time, which helps the researcher to get information regarding
amplitude of any signal that is periodic within the series and the time varies with respect to time.
3.3.3 Calculation of Statistics Indices
i) Mean
It is measures of central tendency which is average of dataset and it is given as follows.
∑𝐱
𝐱̅ = …3)
𝐍
Where 𝑥̅ = Mean of observation and 𝑁 = Number of data
ii) Maximum and Minimum
It is maximum and minimum value calculated in the dataset.
iii) Standard Deviation (SD)
It gives the knowledge about data i.e. data are in disperse from. Low values of SD indicates, data
are closer to mean, it means less variation in the dataset and high values of SD indicates that data
are wider to mean it means more variation or dispersion in the data and it is given by
𝟏
𝐒𝐝 = √𝐍 ∑𝐍𝐢=𝟏(𝐱 𝐢 − 𝐱̅) …4)
Where N= sample size, 𝑥̅ = Mean of sample size, 𝑆𝑑 = Standard Deviation and 𝑥𝑖 is the values
in the observation
iv) Coefficient of variation
It is known as good index of reliability and if the coefficient of variation is less than 20% then it
is consider that data is reliable.
𝐒𝐝
COV (%) = × 𝟏𝟎𝟎 …5)
𝐱̅
v) Skewness
Skewness means lack of symmetry. If the value of coefficient of skewness is negative it means
negative skewness whereas value of coefficient of skewness is positive then it is known as positive
skewness and the mathematical expression is given as follows…
𝐌𝐞𝐚𝐧−𝐌𝐨𝐝𝐞
Coefficient of skewness = …6)
𝐒𝐃
vi) Kurtosis
Kurtosis means flatness or peakness of frequency distribution curve. It is used for finding the
differences between dataset. It is given by…
µ𝟒
𝛃𝟐 = …7)
µ𝟐𝟐
Where µ4 is the 4th central moment and µ2 is the 2nd central moment or it is also known as variance.
(𝐱 𝐢 −𝐱̅)𝟐
µ𝟒 = ∑𝐍𝐢=𝟏 …8)
𝐍
N= sample size, 𝑥̅ = Mean of sample size, 𝑥𝑖 is the values in the observation
4. REFERENCES
• Mahsin, M. (2011). Modeling rainfall in Dhaka division of Bangladesh using time series
analysis. Journal of mathematical modelling and application, 1(5), 67-73.
• Longobardi, A., and Villani, P. (2010). Trend analysis of annual and seasonal rainfall
time series in the Mediterranean area. International journal of Climatology, 30(10), 1538-
1546.
• Wu, C. L., Chau, K. W., and Fan, C. (2010). Prediction of rainfall time series using
modular artificial neural networks coupled with data-preprocessing techniques. Journal of
hydrology, 389(1-2), 146-167.
• Chattopadhyay, S. and Chattopadhyay, G. (2010). Univariate modelling of summer-
monsoon rainfall time series: comparison between ARIMA and ARNN. Comptes Rendus
Geoscience, 342(2), 100-107.
• Cheung, W. H., Senay, G. B., and Singh, A. (2008). Trends and spatial distribution of
annual and seasonal rainfall in Ethiopia. International Journal of Climatology, 28(13),
1723-1734..
• Ellouze, M., Azri, C., and Abida, H. (2009). Spatial variability of monthly and annual
rainfall data over Southern Tunisia. Atmospheric Research, 93(4), 832-839.
• Penarrocha, D., Estrela, M. J., & Millán, M. (2002). Classification of daily rainfall
patterns in a Mediterranean area with extreme intensity levels: the Valencia region.
International Journal of Climatology, 22(6), 677-695.
• Maruyama, T., Kawachi, T., & Singh, V. P. (2005). Entropy-based assessment and
clustering of potential water resources availability. Journal of Hydrology, 309(1-4), 104-
113.
• Joshi, M. K., and Pandey, A. C. (2011). Trend and spectral analysis of rainfall over India
during 1901–2000. Journal of Geophysical Research: Atmospheres, 116(D6).
• Fragoso, M., and Tildes Gomes, P. (2008). Classification of daily abundant rainfall
patterns and associated large‐scale atmospheric circulation types in Southern Portugal.
International Journal of Climatology: A Journal of the Royal Meteorological Society,
28(4), 537-544.
• Santos, C., and Freire, P. (2012). Analysis of precipitation time series of urban centers of
northeastern Brazil using wavelet. Int J Environ Ecol Geol Geophys Eng, 6(7), 64-69.
• Adamowski, K., Prokoph, A., & Adamowski, J. (2009). Development of a new method
of wavelet aided trend detection and estimation. Hydrological Processes: An International
Journal, 23(18), 2686-2696.
• He, X., Guan, H., Zhang, X., and Simmons, C. T. (2014). A wavelet‐based multiple
linear regression model for forecasting monthly rainfall. International Journal of
Climatology, 34(6), 1898-1912.
• Adarsh, S., and Janga Reddy, M. (2015). Trend analysis of rainfall in four meteorological
subdivisions of southern India using nonparametric methods and discrete wavelet
transforms. International Journal of Climatology, 35(6), 1107-1124.
• Nalley, D., Adamowski, J., and Khalil, B. (2012). Using discrete wavelet transforms to
analyze trends in streamflow and precipitation in Quebec and Ontario (1954–2008).
Journal of Hydrology, 475, 204-228.
• Grossann, A., & Morlet, J. (1984). Decomposition of hardy function into square integral
wavelets of constant sharp. SIAM J. Math. Anal., 15, 723-736.
• Smith, L. C., Turcotte, D. L., & Isacks, B. L. (1998). Stream flow characterization and
feature detection using a discrete wavelet transform. Hydrological processes, 12(2), 233-
249.
