SVNIT Study
SVNIT Study
SVNIT Study
SUBMITTED TO
SARDAR VALLABHBHAI NATIONAL INSTITUTE OF TECHNOLOGY, SURAT IN
PARTIAL FULFILMENT OF THE REQUIREMENT FOR THE AWARD OF THE
DEGREE OF
MASTER OF TECHNOLOGY
IN
CIVIL ENGINEERING
WITH SPECIALIZATION IN
TRANSPORTATION ENGINEERING & PLANNING
Submitted by
ABHIJEET B. BHOSALE
(P14TP015)
GUIDE
Dr. G. J. Joshi
Associate Professor,
Department of Civil Engineering,
SVNIT, Surat
2015 – 2016
P.G. CENTRE IN TRANSPORTATION ENGINEERING & PLANNING
CIVIL ENGINEERING DEPARTMENT
SARDAR VALLABHBHAI NATIONAL INSTITUTE OF TECHNOLOGY,
SURAT- 395 007, GUJARAT, INDIA
SARDAR VALLABHBHAI NATIONAL INSTITUTE OF TECHNOLOGY, SURAT
CERTIFICATE
This is to certify that the dissertation entitled “AIR TRAVEL DEMAND MODELLING FOR
SURAT AIRPORT” submitted by ABHIJEET B. BHOSALE (P14TP015) in partial fulfilment
of the requirement for the award of the degree of “Master of Technology in Civil Engineering
with specialization in Transportation Engineering and Planning” of the Sardar Vallabhbhai
National Institute of Technology, Surat, is a record of his own work carried out under research
supervisors‘ supervision and guidance. The thesis report has been approved as it satisfies the
academic requirements prescribed for M. Tech degree.
Dr. G. J. Joshi
Guide
Associate Professor &
Section Head (TEP)
Civil Engineering Department
SVNIT, Surat
ii
Sardar Vallabhbhai National Institute of Technology, Surat
This is to certify that the dissertation entitled “AIR TRAVEL DEMAND MODELLING FOR
SURAT AIRPORT” submitted by Mr. ABHIJEET B. BHOSALE (P14TP015) in
partial fulfilment of the requirement for the award of the degree of “Master of
Technology in Transportation Engineering and Planning” of the Sardar
Vallabhbhai National Institute of Technology, Surat, is examined by us and is hereby
approved for the award of the degree.
Date: Examiners
1. ________________
2._________________
_________________
CHAIRMAN
iii
ACKNOWLEDGEMENT
First and foremost, I must thank The Almighty god for blessing, protecting and guiding me
throughout my life and particularly during M. Tech programme. I would have never accomplished
this without faith I have in the Almighty.
I would like to express my sincere gratitude to my honourable guide and master Dr. G. J. Joshi,
Associate Professor and Section Head, M. Tech (TEP), SVNIT, Surat. I am very thankful to him
for his invaluable guidance, patience, motivation, enthusiasm, encouragement and immense
knowledge. He is the one of the motivator for my whole academic period. This thesis would not
have been possible without his guidance and support. His teachings are helpful to me not only in
M. Tech but throughout my life. I feel myself fortunate and lucky to have complete my dissertation
under him. I am thankful to Dr. Ashish Dhamaniya, Assistant Professor and PG In-charge, M.
Tech (TEP) SVNIT, Surat for his support and valuable time in the dissertation work. I sincerely
thank Dr. Rakesh Kumar, Associate Professor, SVNIT for his support. I am thankful to Mr. Amit
J Solanki, Assistant Professor, SVNIT for his valuable guidance throughout the M. Tech
programme. I am very thankful to Dr. S S Arkatkar, Assistant Professor, SVNIT for his
enthusiastic guidance throughout my dissertation work. I am very grateful to Mr. Pramod Kumar
Thakre, Airport Direcor,Surat and Mr. Sanjay Ezhava,(Founder of SAAC‐committee) for
helping me in getting the necessary data.I would like to thank Dr.J.E.M.Macwan, Professor
and Head of Civil Engg. Department, SVNIT, for providing us with excellent academic facilities.
I heartily thank Vasudevan N., Phd Scholar for his contributions and help in the analysis of my
dissertation. I am extremely thankful to my class mates Prashant Nisarata and Aathira K. Das
for helping me in analysing data. I am deeply thankful to my friends Jairam, Rajiv Mishra, Vinay
Gautam , for their help. I convey my special acknowledgement to my junior Omkar Bidkar and
summer interns for their help in Survey. Also, I am thankful to my other friends in SVNIT for their
untiring efforts in building my confidence throughout the programme. Words fail me to express
my appreciation towards my parents who are my inspiration. Their love and persistent confidence
in me has taken off the loads off my shoulder.
Abhijeet B. Bhosale
iv
DEDICATED WITH EXTREME AFFECTION AND
GRATITUDE TO
v
ABSTRACT
Air transportation is one of the fastest and most reliable mode of transfer which has its own
importance for long distance travel. Air connectivity is the key to unlock a city’s economic growth
potential which attracts business investment and human capital. One of the way of measuring air
connectivity is measuring total passenger movements. Surat city is the economic capital of Gujarat
state which is also textile and diamond hub of India. Being a metro city with over 45 lakhs
population, it lacks the air connectivity. Though Surat has very high economic and population
growth, it has very less number of domestic flights due to some reasons.
In this study, the potential domestic air passenger demand is estimated by comparative method.
For comparative study air travel demand models for other airports are developed for short term
and long term forecast. The purpose of short term forecast is to compare Surat airport’s domestic
air travel demand by current pattern of growth to another airport with similar growth pattern. For
short-term forecast, ARIMA modelling technique is used for other airports. The long-term forecast
is carried out using econometric model by using population and GDP at current rate as the
independent variables. Econometric models are developed for different airports and by using these
models, forecast scenarios for Surat airport is developed. The best econometric model is chosen
by comparing econometric parameters of Surat district with other districts/metropolitan regions.
While estimating air travel demand, consideration of other competing modes for long distance
travel is given. For understanding mode choice behaviour for long distance, mode choice analysis
is carried out. In this mode choice analysis, present long distance journey behaviour and mode
preference analysis is studied. The effect of purpose of travel, willingness to pay, sponsorship and
travelling distances on mode choice is studied. Travel attributes rating analysis is conducted to
find out whether according to the purpose of travels and by income category of travellers, is there
any change in importance given to the travel attribute by travellers.
From mode choice analysis, it is observed that travel distance and purpose of travel have impact
on mode preference given for long distance trips. Willingness to pay varies as income class of
travellers and distance to be travelled varies. Importance of travel attributes, travel time, waiting
time, travel cost, frequency and interchanges varies as the purpose of journey changes while
vi
importance given to all travel attributes doesn’t vary as income class changes except travel cost to
which higher income class travellers give less importance compared to lower income class
travellers. The latent domestic air travel demand for Surat is found to be 14 times higher than
current demand. This indicates Surat airport has very high potential for domestic air travel. For
mode choice analysis MS Excel and SPSS software are used while for forecasting E-Views and
Gretl software are used.
vii
LIST OF ABBREVIATIONS
viii
TABLE OF CONTENTS
CERTIFICATES............................................................................................................................iii
ACKNOWLEDGEMENT ............................................................................................................. iv
ABSTRACT ................................................................................................................................... vi
1.1 BACKGROUND................................................................................................................... 1
1.3.3 Surat Airport in Comparison with Other City Airports In The Vicinity: ....................... 4
1.4 METHODOLGY................................................................................................................... 6
ix
2.5.1 Discrete Choice Modelling ........................................................................................... 20
3.2.2 Rail................................................................................................................................ 26
3.2.2 Population and Domestic Annual Passenger Growth Pattern per Year ........................ 31
3.2.3 Air Traffic Scenario at Surat and other Airports in the Vicinity .................................. 33
x
5.1 GENERAL .......................................................................................................................... 39
xi
7.2 SURAT DISTRICT’S GDP AND POPULATION GROWTH RATE IN
COMPARISIONS TO OTHER METROPOLITAN REGION AND DISTRICTS ................. 90
7.3 LONG TERM FORECAST SCENARIOS FOR SURAT AIRPORT FOR LONG TERM
FORECAST .............................................................................................................................. 91
REFERENCES ............................................................................................................................. 99
ANNEXURE............................................................................................................................... 103
xii
LIST OF FIGURES
Figure 1.1: International comparison of air traffic penetration....................................................... 3
Figure 3.3: Domestic air passenger and population growth pattern for Surat District ................. 31
Figure 3.4: Domestic air passenger and population growth pattern for CMA.............................. 31
Figure 3.5: Domestic air passenger and population growth pattern for MMR ............................. 32
Figure 3.6: Domestic air passenger and population growth pattern for Pune district................... 32
Figure 3.7: Domestic air passenger and population growth pattern for Vadodara district ........... 32
Figure 3.8: Domestic air passenger and population growth pattern for Ahmedabad district ....... 32
Figure 5.6: Mode preferences for travelling distance less than 500 km for business/work purpose
trip ................................................................................................................................................. 48
Figure 5.7: Mode preferences for travelling distance in range of 500 to 1000 km for
business/work purpose trip ........................................................................................................... 48
Figure 5.8: Mode preferences for travelling distance more than 1000 km for business/work
purpose trip ................................................................................................................................... 48
Figure 5.9: Mode preferences for travelling distance less than 500 km for social/leisure purpose
trip ................................................................................................................................................. 49
xiii
Figure 5.10: Mode preferences for travelling distance in range of 500 to 1000 km for
social/leisure purpose trip ............................................................................................................. 49
Figure 5.11: Mode preferences for travelling distance more than 1000 km for social/leisure
purpose trip ................................................................................................................................... 49
Figure 5.12: Mode preference for non-sponsored trips up to 500 km travelling distance ............ 50
Figure 5.13: Mode preference for Sponsored trips up to 500 km travelling distance .................. 51
Figure 5.14: Mode preference for non-sponsored trips for 500 to 1000km travelling distance ... 51
Figure 5.15: Mode preference for Sponsored trips for 500 to 1000km travelling distance .......... 51
Figure 5.16: Mode preference for Non-sponsored trips for more than 1000km travelling distance
....................................................................................................................................................... 52
Figure 5.17: Mode preference for Sponsored trips for more than 1000km travelling distance .... 52
Figure 5.18: Graphical representation of travelling attributes ratings for business purpose trips 55
Figure 5.19: Graphical representation of travelling attributes ratings for social purpose trips .... 56
Figure 5.20: Spider diagram for income wise average ratings given to travelling attributes for
business purpose trips ................................................................................................................... 58
Figure 5.21: Spider diagram for income wise average ratings given to travelling attributes for
social purpose trips ....................................................................................................................... 63
Figure 7.1: GDP growth pattern of Surat, Pune and Vadodara Districts ...................................... 93
Figure 7.2: Population growth pattern for Surat and Ahmedabad District ................................... 94
xiv
LIST OF TABLES
Table 1.1: Year wise annual domestic passengers handled at Surat airport ................................... 4
Table 1.2: Domestic air passengers per capita for different districts .............................................. 5
Table 3.3: Yearly take-offs and landings at airports per million population of districts/
metropolitan regions ..................................................................................................................... 33
Table 3.4: Daily take-off and landing enplanements in Indian Cities .......................................... 33
Table 4.1: Generated GDP at current rate data for Surat, Ahmedabad and Vadodara district ..... 38
Table 5.2: GDP and Population data for Mumbai, Pune and Chennai airport.............................. 42
Table 5.3: GDP and Population data for Ahmedabad and Vadodara airports .............................. 42
Table 5.4: Long term elasticity values for Mumbai, Pune and Chennai airports ......................... 43
Table 5.5: Long term elasticity values for Ahmedabad and Vadodara airports ........................... 43
Table 5.9: Willingness to pay analysis for travellers with annual income less than 5 lakhs ........ 52
Table 5.10: Willingness to pay analysis for travellers with annual income between 5 to 10 lakhs
....................................................................................................................................................... 53
Table 5.11: Willingness to pay analysis for travellers with annual income between 10 to 20 lakhs
....................................................................................................................................................... 53
Table 5.12: Willingness to pay analysis for travellers with annual income more than 20 lakhs .. 53
Table 5.13: Travelling attributes ratings for business purpose trips ............................................. 54
Table 5.14: Travelling attributes ratings for social purpose trips ................................................. 55
Table 5.15: Group statistics for ratings given to travelling attribute for business purpose and
social purpose trip ......................................................................................................................... 57
xv
Table 5.16: Independent sample t-test .......................................................................................... 57
Table 5.17: Income wise average ratings given to travelling attributes for business purpose trips
....................................................................................................................................................... 58
Table 5.18: Details of ANOVA test for business purpose trips .................................................... 59
Table 5.19: Robust Tests of Equality of Means for business purpose trips .................................. 60
Table 5.20: Multiple Comparison by Post Hoc test for business purpose trips ............................ 60
Table 5.21: Income wise average ratings given to travelling attributes for social purpose trips .. 63
Table 5.22: Details of ANOVA test for social purpose trips ........................................................ 63
Table 5.23 Robust Tests of Equality of Means for social purpose trips ....................................... 64
Table 5.24: Multiple Comparison by Post Hoc test for social purpose trips ................................ 64
Table 6.7: Validation of ARIMA (1,1,1) model for Mumbai airport ........................................... 71
Table 6.8: Forecast evaluation statistics of ARIMA (1,1,1) model for Mumbai airport .............. 72
Table 6.9: Validation of ARIMA (0,1,1) model for Ahmedabad airport ..................................... 72
Table 6.10: Forecast evaluation statistics of ARIMA (0,1,1) model for Ahmedabad airport ...... 72
Table 6.11: Validation of ARIMA (0,2,2) model for Vadodara airport ....................................... 73
Table 6.12: Forecast evaluation statistics of ARIMA (0,2,2) model for Vadodara airport .......... 73
Table 6.13: Validation of ARIMA (0,1,1) model for Pune airport ............................................... 73
Table 6.14: Forecast evaluation statistics of ARIMA (0,1,1) model for Pune airport .................. 74
xvi
Table 6.16: Forecast evaluation statistics of ARIMA (1,1,0) model for Pune airport .................. 74
Table 6.17: Validation of ARIMA (0,1,1) model for Chennai airport ......................................... 74
Table 6.18: Forecast evaluation statistics of ARIMA (0,1,1) model for Chennai airport ............ 75
Table 6.25: Forecast evaluation statistics of econometric model for Mumbai Airport ................ 79
Table 6.27: Forecast evaluation statistics of econometric model for Mumbai Airport ................ 79
Table 6.29: Forecast evaluation statistics of econometric model for Vadodara Airport .............. 80
Table 6.31: Forecast evaluation statistics of econometric model for Pune Airport ...................... 80
Table 6.33: Forecast evaluation statistics of econometric model for Chennai Airport ................ 81
Table 6.36: Descriptive Statistics for travel time and travel cost ................................................. 84
xvii
Table 6.41: Parameter estimates for model 4................................................................................ 86
Table 6.42: Interpretation for estimated parameters for Air mode ............................................... 87
Table 6.43: Interpretation for estimated parameters for Rail mode .............................................. 87
Table 6.44: Interpretation for estimated parameters for Air-rail mode ........................................ 88
Table 6.45: Interpretation for estimated parameters for Air-road mode ....................................... 88
Table 7.2: Regression model for Surat airport’s domestic air travel demand .............................. 89
Table 7.3: Forecasted demand from 2014-15 to 2020-21 for Surat airport ................................. 90
Table 7.4: Forecasted demand for Airports using ARIMA models ............................................. 90
Table 7.5: Average annual GDP growth rate from year 2001-02 to 2015-16 .............................. 91
Table 7.6: Average annual population growth rate from year 2001 to 2011 ................................ 91
xviii
CHAPTER 1: INTRODUCTION
1.1 BACKGROUND
Air transportation is one of the fastest, safest and reliable mode of transportation. People normally
travel to fulfill business purpose, for leisure, for other personal reasons, or for some combination
thereof. Though air travel in not significantly different from other modes of intercity travel, it has
some unique characteristics like traveller’s perception of time involved in travel and restrictions
on the traveller’s desire to select a route, a carrier, a transport mode to reach final destination, in
addition to safety, cost, convenience, and accessibility to the traveller.
Air transportation requires large infrastructures in the form of airport and aircraft systems, and
depends on a technology which require large time for its development and implementation. Hence,
the forecasting the demand for air transportation services of airports become a relatively important
activity in economic and strategic planning.
Forecasting is the fundamental process for the planning and design process of airports. Airport
master plans are developed on the basis of forecasts. From forecasts, the relationships between
demand and the capacity of an airport’s various facilities can be established and airport
requirements can be determined. (Kumar & Chandra , 1999)The passenger handling capacity of
airport terminals, runways, freight-storage facilities, parking lots, and even the roadway network
to and from an airport are all based on the air-traffic demand forecasts. The forecasting process
can be the most critical factor in the development of an airport. Errors made in this phase of the
process can be very costly and damaging to local economies. For planning and development
purposes, the annual forecast is the first step from which a whole array of forecasts is derived. The
type of forecast and the level of effort depend on the purpose for which the forecast is being used.
Short, intermediate, and long-range (approximately 5, 10, and 20 years respectively) forecasts are
made to enable the planner to establish a schedule of development for improvements proposed in
the master plan. Short-term forecasts, up to 5 years, are used to justify near-term development and
support operational planning and incremental improvements or expansion of facilities. Medium-
Page | 1
term forecasts, a 6- to 10-year time frame, and long-term forecasts of 10 to 20 years are used to
plan major capital improvements, such as land acquisition, new runways and taxiways, extensions
of a runway, a new terminal, and ground access infrastructure. (Horonjeff, et al., 2010) Forecasts
beyond 20 years are used to assess the need for additional airports or other regional aviation
facilities.
Air travel demand relates primarily to certain basic economic, demographic, behavioral, and
market factors that simply provide people and business with the means to travel and connect. While
estimating demand, it is important to consider other competing modes which are rail and road.
Hence mode choice analysis for long distance travel is important to estimate the propensity to
travel by air.
The objective of this study is to carry out airport specific forecasting rather than corridor
forecasting. Corridor forecasting is concerned with the demand for air travel on in a given travel
route(corridor), say Surat-Delhi corridor. This type of forecast is of interest for airline companiess,
however, is not as useful to airport planners who are mainly concerned with the aggregate
passenger forecasts for a given airport.
Knowledge of the aggregate passenger forecasts will provide planners with the necessary
guidelines for the planning and design of the entire airport facilities as well as number of flights to
various potential destinations. Hence the objectives of this study are framed as under:
Page | 2
1. To study inter regional travel behaviour for South Gujarat Region.
3. To develop the forecasting model to estimate air travel demand for short term and long term
based on time series data of air travel from other airports.
4. To generate scenarios for air travel demand at Surat airport for planning horizon 2035.
The present study is focusing on the mode choice for inter-regional travel and aggregate air
travel demand for Surat airport. Estimate of city pair wise air travel demand is out of scope.
1.3.1 Indian Aviation at Global Level: Global comparison of air travel penetration shows that
India (at 0.04 air-trips per capita per annum) stands far behind the developed countries like US and
Australia (2 air-trips per capita per annum). China’s domestic traffic is five times the size of India’s
despite having a population just 10% larger. There is significant growth potential for the Indian
civil aviation industry as economy grows, disposable incomes rise and the value of time becomes
more precious. (12th Five Year Plan 2012-17 for Civil Aviation., 2011)
The following graph shows there is a direct correlation between development of an economy and
air travel penetration for India.
Table 1.1: Year wise annual domestic passengers handled at Surat airport
2009-10 33880
2010-11 35379
2011-12 101271
2012-13 174321
2013-14 188282
2014-15 80421
Source: AAI office, Surat (2016)
Due to lack of proper air-connectivity to various metro and tier II cities travellers from Surat and
South Gujarat region needs to travel around 250km either to Ahmedabad or Mumbai for
International flights. Unfortunately, even for domestic flights, travellers need to travel to other
nearby cities. Even city of Vadodara, with half of the population of Surat, has much better domestic
air connectivity compelling travellers to take up road or rail travel of 130 Km to Vadodara for
catching flights.
1.3.3 Surat Airport in Comparison with Other City Airports In The Vicinity: The ratio of
total number of domestic passengers handled at airport to total population of the corresponding
districts/region indicates provision of air travel services for a city of given economic condition.
Statistics domestic air passengers per capita for CMA (Chennai Metropolitan Area), MMR
(Mumbai Metropolitan Region), Pune, Vadodara and Surat districts is given in table 1.2
Page | 4
Table 1.2: Domestic air passengers per capita for different districts
From table 1.2, it can be seen that as compared to airports at Mumbai, Ahmedabad, Pune and
Vadodara, the per capita domestic air passengers are too low for Surat district. For year 2011, Surat
district had 42.24 percent growth in the population compared to population as per 2001 whereas
Ahmedabad district, Pune district, Vadodara district, MMR and CMA have 24.03, 30.37, 14.38,
28.41 and 35.20 percentage growth respectively(Census,2011).Though Surat district has highest
decadal growth compared to other districts/metropolitan regions it witnessed still very less growth
in domestic air travellers. Table 1.2 provides the comparative idea of ratio of domestic passengers
per capita(population) of city region of MMR, Pune district, CMA, Vadodara district and Surat
district.
A city with relatively weak economic base, this ratio can be quite low, however lower ratio can be
observed under constrained air travel services at an airport of city with strong economic base. If
this ratio is compared between cities with comparable economic strength, it reflects on the
deficiency of air services. It is found significantly lower than other cities in case of Surat.
To provide the better connectivity for Surat airport, there is need of forecasting air travel demand
considering air connectivity available at nearby airports and also by comparing socio economic
Page | 5
characteristics. This forecasting study will also help airport planners to plan various airport
components and facilities like terminal building, runway capacity, storage capacity etc.
1.4 METHODOLGY
The present study has been carried out as per the steps described below:
Problem Identification: To select the topic of dissertation, the current issues for regional
transport for Surat city has been studied. Being a metro city with high rate of economic and
population growth, Surat lacks in one key aspect of air transport that is air transport with
better connectivity.
Defining objectives: While defining the objective for the research, the study of present
scenario for inter-regional transport is included. Also, the effect of other competing modes
of transport like road and railway is taken into consideration.
Literature Review: To study methods and approaches used in other similar works several
research papers, manuals and literature from books has been referred. In what situation, the
methods are suitable is to be studied with possible alternatives and the various independent
variables they have considered for forecasting is also studied.
Defining the Study area: For forecasting the air travel demand at Surat airport, the entire
South Gujarat region is considered as its influence area. Since we are also going to analyze
demand at others airports in the vicinity their influence area will be decided after
conducting appropriate survey.
Data collection: For this study the following primary and secondary data is collected:
Page | 6
Primary data: As per the objectives for developing mode choice behavior, mode choice
survey of travellers through a questionnaire survey is conducted at Surat Airport. For this
a questionnaire survey is to be conducted at Surat Airport.
Secondary data: For making econometric model for forecasting a time series data of annual
passengers handled at given airports for last 20 years is used for short term forecasting.
The month wise number of passengers handled at given airports for last 7 years is used for
seasonality analysis. Also for econometric analysis, annual domestic passenger data and
socio-economic factors like are population and GDP are collected for calibrating
econometric model.
Data Analysis and development of model: After collection of primary and secondary data
the logit model is calibrated for mode choice. Using econometric analysis and time series
analysis, models for air travel demand estimation are developed. Then by comparative
method air travel demand at Surat airport is forecasted for future.
Results and Conclusions: After analyzing the data and developing air travel demand
model, the results are reported and its interpretation is presented. Based on this, suitable
conclusions are drawn
Page | 7
Problem Identification
Defining Objectives
Literature Review
Data Collection
This thesis in divided into eight chapters. Chapter 1 includes the introduction parts which consists
background about the topic, objectives and need of study and methodology to be adopted. Chapter
2 is literature review which has short descriptions about the methods to be used in analysis and the
review relevant studies done in this research area. Chapter 3 is study area profile which include
demographic and economic study of study area. Chapter 4 is data collection. Chapter 5 which is
data analysis consists of 3 types of analysis which are time series analysis, econometric analysis
and mode choice analysis. Chapter 6 in about development of models which formulation and
validation part of models. In Chapter 7, forecasting is done using models obtained from previous
chapter. Finally, in Chapter 8, the summary and conclusion of whole study is given with future
scopes in this study area.
Page | 9
CHAPTER 2: LITERATURE REVIEW
2.1 GENERAL
Reliable forecasts of air travel/traffic demand activity play a critical role in the planning process
of States, airports, airlines, engine and airframe manufacturers, suppliers, air navigation service
providers and other relevant organizations. Forecasting is not an independent discipline but is a
part of the overall aviation planning process.
The first consideration in terms of a forecast is its intended use. Forecasting has a short-term,
medium-term or long-term time horizon depending on the intended use, the length of which can
vary from industry to industry, as well as the particular application concerned. One of the useful
criteria for matching a specific forecasting situation with the appropriate methodology also
depends on the time horizon involved. For the aviation industry, the following time horizons are
generally used:
Short-term: up to 1 year
Medium-term: 1 to 5 years
Present study aims to bring out the medium term and long term forecasting for a metropolitan
city/airport. Medium-term forecasts are generally prepared for planning, scheduling, budgeting
and resource requirement purposes. The trend factor as well as the cyclical component plays a key
role in the medium term forecast because the year-to-year variations in traffic growth are an
important element in the planning process. As per ICAO manual for medium term forecasting time
series analysis/trend analysis should be carried out. Long-term forecasts are used mostly in
connection with strategic planning to determine the level and direction of capital expenditures and
to decide on ways in which goals can be accomplished. The trend element generally dominates
long-term situations and must be considered in the determination of any long run decisions. Since
the time span of the forecast horizon is long, it is also important that forecasts are calibrated and
Page | 10
revised at periodic intervals (every two or three years depending on the situation). The methods
generally found to be most appropriate in long-term situations is econometric analysis as per ICAO
manual for forecasting.
The theory and research work related to time series/trend analysis and econometric analysis for
medium-term and long term forecasting respectively is reviewed.
Time Series relates to values taken by a variable over time (such as daily sales revenue, weekly
orders, monthly overheads, yearly income) and tabulated or plotted as chronologically ordered
numbers or data points to yield valid statistical inferences.
When deriving a medium-term or long-term forecast by extrapolating from the traffic trend, the
forecaster assumes that the factors which determined the historical development of the traffic will
continue to operate in the future as in the past, except that their impact may change gradually, and
steady-state conditions will continue into the future. The appropriateness of using trend analysis
in forecasting depends heavily on stability in past developments and the confidence of the
forecaster that the assumption of continuing trends is appropriate to the particular operating
environment.
Secular trend: The increase or decrease in the movements of a time series is called Secular
trend. A time series data may show upward trend or downward trend for a period of years and this
may be due to factors like: increase in population, change in technological progress, large scale
shift in consumers demands etc.
Page | 11
Seasonal Variation: Seasonal variation are short-term fluctuation in a time series which occur
periodically in a year. This continues to repeat year after year. The major factors responsible for
seasonal variation are weather conditions and customs of people.
Cyclical variations: These are recurrent upward or downward movements in a time series but the
period of cycle is greater than a year. Also these variations are not regular as seasonal variation.
Irregular variation: Irregular variations are fluctuations in time series that are short in duration,
erratic in nature and follow no regularity in the occurrence pattern. These variations are also
referred to as residual variations since by definition they represent what is left out in a time series
after trend, cyclical and seasonal variations.
• Addition Model: O= T + S + C + I
• Multiplication Model: O = T x S x C x I
Ot Tt×St×It
= ≈ St×It ……………………………………2.1
Tt’ Tt’
Step-2: Capturing Seasonality (i.e. Seasonal indices): This involves Quantify the effect of
seasonality in each year by calculating overall average for each month/quarter for across all years.
De-seasonalize the dataset by dividing with seasonal indices Tt’
Ot Tt×St×It
= ≈ Tt×It ……………………………………2.2
St’ St’
Page | 12
Step-3: Irregular Component: Irregular means no trend which can not be modelled. After first
level of de trending & de seasonalising, if there is still any pattern left in the time series, repeat the
above steps once again.
Tt×It
≈ It……………………………...…………..………2.3
Tt’
Econometric demand forecasting involves various economic, social, and operational variables in
determining on the basis of historical data. A quantitative relationship between air travel and the
variables influencing the level of traffic is to be established.
The evaluation uses a variety of multivariate statistical techniques such as correlation analysis,
factor analysis, and linear and nonlinear regression analysis to define suitable predictive variables
selected as the independent variables of the model. The main steps in the development of a forecast
using an econometric model are explained in the figure 2.1.
1. Problem definition: Problem definition is related to the intended use of the forecast. It
includes the geographical coverage, whether it is global, regional or a particular traffic flow.
The time horizon of the forecast should also be specified.
2. Selection of explanatory variables: There are two approaches for selecting explanatory
variables. In first, the researcher identifies a long list of potentially useful causal variables
and in other way the selection of a small number of the most relevant variables is done.
3. Data availability: Certain variables can be rejected due to lack of available data. For some
variables, the data may not be directly available in an appropriate form but can be derived
from several sources or by careful estimation
4. Formulation of the model: In order to establish the relationship between the selected
independent (explanatory) variables and the dependent variable, i.e. traffic demand, it is
necessary to specify the mathematical form of the relationship (model). The validity of the
relationship can be established empirically through tests against actual historical data.
5. Estimation of model coefficient: For each traffic demand model specification, in terms of a
set of explanatory variables and a functional form, there is one set of values for the constant
coefficients (a, b, c, etc.) that provides a better fit of the equation to the data than all other
sets of coefficients
6. Final form of model: Once the model has satisfied all the statistical criteria, the econometric
model will be established in its final form. The model provides the first step in producing a
forecast but not necessarily the last.
7. Scenario for explanatory variables: Scenario analysis is to understand the range of various
possible outcomes. In constructing the scenarios, many important factors come into play,
which will have an influence on the final forecasts.
The performs of model can be measured by comparing predicted and actual values.For this
MAPE (mean absolute percentage error), MAE (mean absolute error) can be used.
Page | 14
2.4 RELEVANT WORKS ON AIR TRAVEL DEMAND FORECASTING
The first step in forecasting air traffic is usually to study the historical data (time series) and
determine the trend in traffic development. In the context of medium or long-term forecasting, air
traffic trend represents the regular or smooth development of air traffic over many years is isolated
from short-term fluctuations in traffic levels. When deriving a medium or long-term forecast
through extrapolating from the traffic trend, the historical growth of air traffic may continue to
operate in the future as in the past and their impact may change gradually.
The study of suitability of different models for reliable outputs is important. The comparison
between outputs of 3 different methods has been made by Ahmadzade(2010). The methods used
in this study are trend projection, econometric relationship and market and industry survey.
Stationarity of trend is checked by ADF unit root test. In market and industry survey, analysis of
the characteristics of the market for air transport in order to examine, empirically how the use of
air transport varies between different sectors of the population and different industries has been
made. Use of month dummy variables is done in mode making to incorporate monthly variation.
Decomposition technique is one of the time series analysis method. Decomposition methods
involve the dissection of the problem into various components. These methods are particularly
relevant when strong seasonality or cyclical patterns exist in the historical data. These methods
can be used to identify three aspects of the underlying pattern of the data: the trend factor, the
seasonal factor and any cyclical factor that may exist.
In the research work by Mallikarjuna & Raghu Kanth(2010), a new strategy for modelling and
forecasting the air traffic data series is presented using time series modelling with decomposition
technique to decompose the monthly air traffic time series into finite number of intrinsic modes.
This helps in identifying the last empirical mode as a trend and the summation of remaining modes
as the fluctuation in the data. The fluctuation part is handled by artificial neural network
techniques. In this study the monthly air passenger data of Guwahati, Imphal, Silchar, Agartala,
and Dibrugarh which are major cities in North East region of India are collected and the correlation
between air traffic at these airports are analysed. Air traffic data from Jan. 2002 to Feb. 2006 has
been used for modelling purpose and the remaining data has been used for validation purpose.
Page | 15
From the performance indices, it has been found that in both modelling and forecasting periods,
proposed methodology is able to predict the air traffic reasonably well. Proposed models could
explain 46–89% of the variability of five air traffic data series used in this study.
In similar study by Sen(1985), time series modelling is used for examining air travel demand for
short-term forecast and estimate impact of fare changes. In this, the analysis of total monthly
revenue passenger kilometres flown by Indian airlines from 1976 December to 1980 is done using
smoothening technique for easy computation and consequently easy modification. The median
smoothening technique is used by the author as decomposition technique.
The determination of a statistical time series model of demand for domestic air travel in Nigeria is
done by Eke, et al.(2014) using ARIMA models. After conducting test for stationarity of time
series and homogenity of variance,four ARIMA models are developed.The best model for
forecasting is selected on the basis of least Mean Square Error(MSE) value which is
ARIMA(1,0,1) with MSE value 0.6723.
In a study by Rengaraju & Arasan(1992), the air travel demand between 40 city pairs in India is
estimated using gravity model with special emphasis on rail/air travel time ratio and big city
proximity and validation of model is done by regression analysis by comparing estimated values
and actual values of households, percentage literates, percentage migrants, percentage employees
and percentage university degree holders. In this study, the cross sectional data for demand of
domestic air travel in India is used for calibration of model. The travel time ratio between rail and
air and the big-city proximity factor are found to be significantly influencing the demand for air
travel along with the other variables.
Karlaftis(2008) has attempted forecasting for highly seasonal demand in regional airports, where
peak flows approach airport capacity. He has useda modeling combination, dynamic Tobit models
with GARCH errors/disturbances, which is able to capture many of the shortcomings of most
traditional models.odels are calibrated using monthly passenger and flight data for a 20 year
period for the airport of Corfu in Greece, where traffic over the summer approaches airport
capacity and seasonal fluctuations in demand are very intense. Results show that: i. Not explicitly
accounting for seasonal variations in demand or for traffic approaching capacity may significantly
bias model parameter estimates and affect demand predictions; and, ii. Improved demand model
Page | 16
specifications are an invaluable tool in obtaining more accurate demand estimates.In another study,
byKarlaftis et al. (1996) examined the predictive ability and forecasting accuracy of airtravel
demand models. In particular, they developed an analytical framework for developing econometric
models, and used post-fact analysis to test the accuracy of the models. Statistical data describing
air-travel patterns for two major international airports, the Miami and Frankfurt International
Airports, were used to demonstrate the effectiveness of the proposed models. In addition, the effect
of external factors such as the deregulation of the air-transport industry was examined. The results
suggested that simple models with few independent variables perform as well as more complicated
and costly models, and that external factors have a pronounced effect on air-travel demand.
In another study carried out by Grosche, et al.,(2007), two gravity models are used for estimation
of passenger volume between cities in 28 European countries with 1228 city pairs. One model that
is basic gravity model is limited to city pairs with airports not subjected to competition from
airports in the vicinity, while the other model that is extended gravity model includes all city pairs.
The independent variables used in the basic model are population, catchment, buying power index,
GDP, geographical distance and average travel time. In extended model, additional variables of
number of competing airports, average distance to competing airports ,number of competing
airports weighted by their distance are taken. The airports within a distance of 200 km are
considered as competing airports. In this studies, with conventional geo-economic variables and
services related factors they have given emphasis on non-service related factors. The cross
sectional data is used for calibration.
By Fridstrom & Thune-Larsen(1989) ,the intercity gravity model for Norwegian network is made
which is calibrated econometrically using both combined cross-sectional and time series data on
traffic flows, fares, travel time, income and population.
Wadud(2011) developed a model to forecast of passenger demand at a new airport with limited
aggregate data. This model utilizes aggregate time series information on national level and
patronage at other airports to develop a gravity type model for the country. The model has been
developed using population, GDP or income, ratio of costs between the cities by air travel and
alternative road travel. In this study, the gravity model for is city pairs is used considering GDP
per capita, population, and travel time ratio as independent variables. A gravity-type econometric
Page | 17
model was estimated by using national aggregate data, and it was calibrated further for predicting
demand in a new airport by using a simple passenger questionnaire survey.
Alam & Karim,(1998), constructed air travel demand model for domestic air transportation in
Bangladesh using city pair models and for the analysis, data of 7 city pairs involving 8 cities in
Bangladesh is collected. The zone of influence is identified through in-flight passenger survey
which was found to be 40 km. The demand variables used in this study are population,
employment, GDP with use of dummy variables in some special cases where tourist attraction and
connecting flights from United Kingdom are more. The supply variables used are travel time ratio
as impedance parameter and frequency of service which is used indirectly as apart of demand
variable. For the purpose of analysis, stepwise multiple linear regression analysis has been adopted
in the model.
Drago, et al.,(2014) developed an econometric model to estimate demand of passenger air transport
in Croatia. Two models were made and tested. For one of the model, GDP is considered as an
independent variable, while in othe, the number of tourist arrivals arrivals is considered as
dependent variable since the study area, Crotia which a popular tourism place. Both models show
satisfactory theoretical, statistical and econometric values, and a high level of practical
applicability in projecting demand in air passenger transport.
Ba-Fail, et al.(2000) addressed the determinants of domestic air travel demand in the Kingdom of
Saudi Arabia. List of variables relevant to the demand for the domestic air travel which are
considered in this study are total non-oil gross domestic product,consumer price index, import of
goods and services, per capita income, population size,total expenditures and total consumption
expenditures.Four models are made using different combination of these variables.Finally the
model with number of passengers in millions, population and total expenditures found to be most
suitable based on t- values and R2 values.
Silva, et al.,(2012) studied the determination of the economic viability of new airports for which
demand forecasting is necessary. The Manual on Air Traffic Forecasting Part I (Doc 8991)
addresses three general approaches to model and forecast airport specific demand. These
approaches are: (i) Quantitative forecasting methods, (ii) Qualitative forecasting methods, and;
Page | 18
(iii) Decision analysis. All these approaches, however, require extensive information on individual
travel behaviour as well as the distribution of socioeconomic characteristics in the area influenced
by the new airport or judgment of several experts. This study explains an alternative approach
where passenger demand for Brazilian regional new airports is modelled using an econometric
approach. Per-capita income, population, yield and a constant is used to calculated economteric
model for airports that have similar socio-economic characteristics based on the Instituto Brasileiro
de Geografia e Estatística (IBGE) study that gives the areas of influence of cities. The alternative
model proposed which is based on the econometric method considers the following predictive
variables: (i) population of catchment area;(ii) average airfare per mile;(iii) per-capita income of
catchment area, and; (iv) constant calibrated to each regional airport to predict passenger
enplanements. The proposed model estimates demand with max percentage error up to 5.09%.
Alekseev & Seixas(2002), attempted artificial neural networks based models for the air transport
passenger demand forecasting. They used econometric modelling using GDP and mean ticket price
for transported kilometers with dummy variable.It is found that neural processing can outperform
the traditional econometric approach used in this field and can accurately generalise the learnt
time series behaviour, even in practical conditions, where a small number of data points is
available. Feeding the input nodes of the neural estimator with pre-processed data, the forecasting
error is evaluated to be smaller than 0.6%.
As per air travel demand studies carried out by Wang & Song,(2010) air travel demand research
shares some similar characteristics with tourism demand research in terms of the measurement of
demand, demand elasticity analysis, and forecasting modeling techniques. However, the research
of air travel demand shows its own features and trends such as driving forces of demand and
different forecasting techniques. The major variety of air travel demand studies is in connection
with two main market segments; that is, business travelers and leisure travelers, with no distinction
between these two markets, only few studies focused on leisure air travel in recent years.
Page | 19
2.5 MODE CHOICE MODELLING FOR LONG DISTANCE TRAVEL
Increasing congestion on intercity highways and at intercity air terminals has raised serious
concerns about the adverse impacts of such congestion on regional economic development,
national productivity and competitiveness, and environmental quality. To alleviate current and
projected congestion, attention has to be directed toward identifying and evaluating alternative
proposals to improve intercity transportation services. Among other things, a prior evaluation of
such large-scale projects requires the estimation of reliable intercity mode choice models to predict
ridership share on the proposed new or improved intercity service and identify the modes from
which existing intercity travellers will be diverted to the new (or improved) service. Discrete mode
choice modelling concept is used for this purpose.
2.5.1 Discrete Choice Modelling
Discrete choice models can be used to analyse and predict a decision maker’s choice of one
alternative from a finite set of mutually exclusive and collectively exhaustive alternatives. Such
models have numerous applications since many behavioural responses are discrete or qualitative
in nature; that is, they correspond to choices of one or another of a set of alternatives. The ultimate
interest in discrete choice modelling is to predict the decision making behaviour of a group of
individuals. The basis for estimating the probabilities of choice for each alternative lies in
evaluation of utilities of alternatives. For this, utility function is defined which measures the degree
of satisfaction or impedance that people derive from their choices. The utility function is typically
expressed as the linear weighted sum of independent variables or their transformation that is
U=a1+a2X2+a3X3+……..+arXr , where a1, a2……….ar are model parameters
Discrete choice models, such as the binary logit and multinomial logit, are used to predict the
probability a decision-maker will choose one alternative among a finite set of mutually exclusive
and collectively exhaustive alternatives.
2.5.2 Logit Models
In statistics, logistic regression; also, known as the logistic model or logit model; is used for
prediction of the probability of occurrence of an event by fitting data to logistic curve. Generalized
linear model is used for binomial regression. Like other forms of regression analysis, it makes use
of various predictor variables that may be either numerical or categorical. Logit analysis assumes
Page | 20
that is probability of occurrence of an event varies with respect to utility function F(x) as a sigmoid
curve called logistic curve.
Simple form of logit model; that is binary logit model is, P1=eu1 /eu2+ eu3
In binary logit, models are calibrated to estimate probabilities for only two discrete choice.
Where P1 =Probability of an individual choosing mode 1
u1, ,u2= utility functions of mode 1 and 2 =a1(c1 - c2)+a2(t1)+……
a1 ,a2= model parameter
c1,c2= cost of travel by modes 1 and 2
t1,, t2=time of travel by modes 1 and 2
Multinomial logit model is used when the dependent variable in question is nominal and
consists of more than two choices. Simple form of MNL is:
Nested Logit model is same as MNL except in this the unobserved component of utility is
correlated over alternatives rather than being independent over alternatives.
Investments for improving long-distance travel infrastructure are substantial and may have
significant impacts on travel demand, the environment, and the economy, and therefore, deserve
careful analysis before implementation. The vast majority of mode choice models developed over
the last few decades have been implemented for urban models with a focus on short-distance travel,
where modal availability is different from long-distance travel. The travel behaviour in long-
distance travel is quite different, too, as people tend to be more familiar with modal options for
short-distance travel than for long-distance travel. In addition, the trip purposes also differ. While
short-distance transit ridership is dominated by commute trips, long distance transit modes
(particularly rail and air) are heavily used for pleasure trips as well as by business travellers.
Because long-distance travellers tend to stay longer at their destination, travel time tends to be a
less dominating factor in long-distance mode choice than in short-distance travel.
Page | 21
Algers & Staffan(1993) presented an overview of the models involved in a new model system for
long-distance trips developed for Swedish national authorities. The model system consists of
nested logit models, partly estimated by the use of simultaneous estimation techniques. The trip
data source is a national travel study conducted in 1984-1985. The choice structure of the model
system spans from choice of access and egress mode over mode and destination choice to trip
generation. There are different models for business and private trip purposes. The models contain
cost parameters and mode-specific time parameters. The integrated structure implies that all
variables affect all choice levels.
Miskeen, et al.,(2013) carried out a study to recognize the behavior of intercity travel using
disaggregate models, for projecting the demand of nation-level intercity travel in Libya.
Multinomial Logit Model for all the intercity trips by modes car,intercity bus and airplane has been
formulated to examine the national-level intercity transportation in Libya.The Multinomial logit
model was calibrated using nationwide revealed preferences (RP) and stated preferences (SP)
survey. The model was developed for deference purpose of intercity trips (work, social and
recreational).The refrence mode taken in this study is car. The variables of the model have been
predicted based on maximum likelihood method. The data needed for model development were
obtained from all major intercity corridors in Libya. The final sample size consisted of 1300
interviews.The explanatory variables considered in making the logit models for intercity bus and
airplane are gender, nationality, age, monthly income,duration of stay at destination, purpose of
travel, car availability,egress distance from airport/bus terminal to final destination, total travel
cost, privacy and convenience.
Baik, et al.,(2008) developed a nationwide nested mode choice model for long-distance travel by
auto, air, and air taxi. The estimation was based on American Travel Survey (ATS) of the Bureau
of Transportation Statistics (BTS) from 1995. Unfortunately, the rail mode was left out because of
limitations in the survey data, which limits the usability of this model to areas that do not offer
significant rail service. Specifically to analyze highspeed rail ridership. The mode choice model
uses the trip distribution and a nested logit model to calculate the air taxi, commercial airline, and
automobile county-to-county demand. The probability of ridership for each mode is dependent on
its utility function.The utility for each mode is calculated by using its door-to-door travel time and
travel cost. The utility function is defined in this study have two variables:travel cost and travel
Page | 22
time. The travel time and cost coefficients vary depending on the trip purpose and the traveler’s
household income.
Cho,(2013) developed MNL models adopting mathematical function from the conditional logit
(CL) model. In estimating the model, the impact of travel time and cost on mode choice decisions
for long-distance travel is considered. This study assumes a new high speed rail system as a new
alternative mode over other ground transportation modes such as personal cars and bus. This study
also estimates travel times for personal cars, bus, airplane and train in the same travel distances.
These distances represent the service range where a high speed ground transportation mode,
including HSR system, is empirically expected to have competitiveness against air and personal
cars. In this study, travel time and travel cost data for both modes in reference to Amtrak and
Greyhound’s real service operations. The estimated coefficients of both travel time and travel costs
are negative in all models developed in this study. These negative signs indicate that travelers’
utility decrease as travel time and costs increase.
From above studies made by researchers it can be said that the factor of mode choice study or the
preferences given to other means of transport has not been considered while forecasting passenger
demand for air travel. Long distance travel by connecting flights and combination of modes such
as air-rail or air-road are also not considered.
Most of the studies are carried out with city pair models in India. City pair models are useful for
forecasting demand along specific corridor but for airport planners, total air travel demand at
airport that means total number of passengers handled at airport is more important. In India, study
at national and international level has been conducted but at regional level no study has yet been
carried out. Such studies may also help to introduce regional air connectivity for fast growing
metropolitan cities.
Page | 23
Also, the travelling attribute studies have not been reported in literature for long distance travel.
There are few studies in India considering mode choice behaviour for long distance travels with
comparison between different modes like air, rail and road.
Page | 24
CHAPTER 3: STUDY AREA PROFILE
Surat is one of the fastest growing city in India and second largest city in Gujarat. The city
registered an annualised GDP growth rate of 11.5 per cent over the seven fiscal years between
2001 and 2008.Surat city is also a part of smart city project of government of India. Surat is also
known as Diamond Hub of the World, Textile Hub of India, and having lots of corporate Industries
in its surrounding. Surat is among top exporter of Agricultural products as well.
South Gujarat region of Gujarat has population of more than 1 Crore has many NRIs,
Businessmen, Professionals, etc. who frequently needs to fly domestic or international destination.
Surat recorded 65% Youth Population and also 65% Internet Users out of around 45 lac city
population in year 2011.
Surat has regional connectivity by all three modes which are air, rail and road.
3.2.1 Air
Till May 2006, Surat was the only city in the world with a population of over 4.6 million people
without a functioning airport. Surat airport was closed down due to technical reasons in 2001. The
new airport finally became functional with Surat-Delhi IA flight on 6 May 2007. Currently there
is daily flight to Delhi in morning as well as evening and to Mumbai there is morning flight on
Wednesday, Friday and Sunday. Surat city also has good connectivity with rail and road modes
with other major cities with big airports like Mumbai and Ahmedabad which have better airport
connectivity than Surat airport for domestic and international travel. Proper air-connectivity is not
only beneficial to Surat but can also support in development of South Gujarat region, Khandesh
region (North-West Maharashtra), North Maharashtra, South West Madhya Pradesh, and UT of
Damman and Dadara & Nagar Haweli. At present, air travellers of these regions have to travel to
Page | 25
these big cities to reach their desired destination on domestic as well as international level. For the
convenience of local people of Surat and being a part of smart city project, Surat needs to have
better air connectivity with other airports.
3.2.2 Rail
Surat Railway Station is under the administrative control of the Western Railway Zone of the
Indian Railways. It connects Surat with Delhi, Ahmedabad, Vadodara, Jaipur, Jabalpur, Patna and
other cities of north India and southern and eastern Indian cities. There is also a connection to the
Central Railway Zone through Udhana Railway Junction and the Tapi line which connects Surat
with cities of Central India namely Jalgaon, Amravati and Nagpur. Daily around 250 passenger
trains range from local trains to express trains, superfast and premium halt at Surat station. In
addition, there are freight trains that move the goods produced by the city to the rest of the country
and beyond. Surat railway station was built by British is one of its kind in India, as the railway
platform is on the first floor of the railway station whereas ticketing and access facilities are at the
ground floor level.
3.2.3 Road
Surat is on junction of 3 important National Highways of India and many state highways of Gujarat
state. The major highways passing through Surat are the Udhana-Mumbai Highway also known as
Udhana-Navsari Highway, Surat-Ahmedabad Highway also known as Varachha Main Road and
Surat-Kolkata Highway also known as Dhulia Highway.
The Golden Quadrilateral highway system passes through Surat district. The city is connected to
the National Highway 8 through a 16 km connector highway. National Highway 6, also known as
Surat-Kolkata Highway, starts from Hazira and passes through Surat city and connects the city to
Dhule, Amravati, Nagpur, Raipur, Sambalpur, Kharagpur and Kolkata. The National Highway 228
also known as Dandi Heritage highway, starts from Sabarmati in Ahmedabad and passes through
Anand, Kheda , Ankleshwar and passes through Surat city to Navsari ending at Dandi. National
Highway 50, also known as Surat-Pune highway starts near a junction known as Kharvarnagar
Junction and passes through Sachin, Navsari, Dharampur, Nasik, Pune.
The city of Surat is commercial capital of the state and is of significant importance to the country.
The Arabian Sea is to its west at a distance of about 22 km along Tapi and about 16 km by Road.
Page | 26
The city is a pivotal centre on the Ahmedabad-Mumbai regional corridor as well as on the 225 km
long industrial belt, having direct linkages with the industrial centres of Vadodara, Ankleshwar
and Vapi.
Surat is known for its textile manufacturing, trade, diamond cutting and polishing industries. The
city economy is characterised by a large number of small and medium sized industries. In terms
of factory sector also, the share of Surat is significant. Much of the industrial development is
located within the limits of Surat city. In fact, Surat is truly an industrial city with over 50% of
workforce engaged in manufacturing activity.
Page | 27
Surat ranked 8th in India with a GDP of $40 billion in fiscal year 2011–2012 ($14 billion in 2010).
Surat GDP in 2020 will be around $57 billion estimated by The City Mayors Foundation, an
international think tank on urban affairs. The per capita GDP was $8,000 in 2010. Surat is known
for diamonds, textiles, and for diamond-studded gold jewellery manufacturing. Surat registered a
GDP growth of 11.5% for the seven fiscal years 2001–2008, the fastest growing GDP in India.
The city accounts for:
Total 136 number of airports are existing in India out of them 82 are operational. In the state of
Gujarat, total 10 airports are there out of which 8 are operational. Figure 3.2 shows all the airports
in India.
It would be appropriate similar to study the air travel demand in other cities with good air
connectivity compared to Surat. This eventually helps in estimating the potential air travel demand
at Surat. For this comparative study, Mumbai(MMR), Ahmedabad district, Pune district, Chennai
(Chennai Metropolitan Area) and Vadodara district are considered.
In MMR (Mumbai Metropolitan Area), Mumbai District, Mumbai Suburban District and some
part of Raigad and Thane District is included but in this study, whole Raigad and Thane District
is considered for easy study area delegation. Similarly, In CMA (Chennai Metropolitan Area)
Page | 28
Chennai district and some part of Tiruvallur and Kancheepuram district is included but for easy
study area delegation, whole Tiruvallur and Kancheepuram district is consider.
3.2.1 Population
Table 3.1 shows growth profile of population of different districts.
Chennai
YEAR Surat Vadodara Ahmedabad Pune Mumbai(MMR)
(CMA)
Since the population for year 2001 and 2011 only available, to estimate each year’s population,
annual growth rate is calculated using formula P2011=P2001 (1+i)10.
Surat district has high percentage of population living in urban area similar to Ahmedabad district,
MMR and CMA.
Page | 30
Table 3.2: Urbanisation percentage
Source: Census2011
3.2.2 Population and Domestic Annual Passenger Growth Pattern per Year
The following graphs shows the growth patterns of domestic air passengers and population for
given districts/metropolitan region. From graphs, it is observed that for Surat district, there is
significant drop in domestic air passengers in the year 2014-15 though there is increase in
population. For other airports increasing trend is observed in domestic air passengers for
increasing population.
1.80 68 140.00
100.00
Population in lakhs
Population in lakhs
1.60 66
120.00
1.40 64 80.00
1.20 62 100.00
1.00 60 60.00 80.00
0.80 58 60.00
0.60 56 40.00
40.00
0.40 54 20.00
0.20 52 20.00
0.00 50 0.00 0.00
2005-06
2001-02
2003-04
2007-08
2009-10
2011-12
2013-14
2015-16
Year Year
Figure 3.3: Domestic air passenger and Figure 3.4: Domestic air passenger and population
population growth pattern for Surat District growth pattern for CMA
Page | 31
350.00 300.00 60.00 120.00
Population in lakhs
250.00
Population in lakhs
250.00 40.00 80.00
200.00
200.00 30.00 60.00
150.00
150.00 20.00 40.00
100.00
100.00 10.00 20.00
50.00 50.00
0.00 0.00
0.00 0.00
year
year
domestic air passnegrs in lakhs population in laks domestic air passnegrs in lakhs
Figure 3.5: Domestic air passenger and population Figure 3.6: Domestic air passenger and population
growth pattern for MMR growth pattern for Pune district
Domestic air passengers in lakhs
Population in lakhs
8.00 40.00 70.00
Population in lakhs
Year
Year
domestic air passnegrs in lakhs
domestic air passnegrs in lakhs population in laks population in laks
Figure 3.7: Domestic air passenger and population Figure 3.8: Domestic air passenger and population
growth pattern for Vadodara district growth pattern for Ahmedabad district
Page | 32
3.2.3 Air Traffic Scenario at Surat and other Airports in the Vicinity
Status of a connectivity of a city can be judged through simple data ratio of number of flights
landing and taking off from the city. Table 3.3. gives yearly take-offs and landings at the airports
per million population of their respective districts/metropolitan regions. For 2015 year for Surat
district this value was approximately 230 which is very less in comparison to other
districts/metropolitan regions.
Table 3.3: Yearly take-offs and landings at airports per million population of districts/
metropolitan regions
Table 3.4 provides data of daily take offs and landings of 25 India cities including Surat and their
population.(Census 2011).
Page | 33
12 Nagpur 24.06 35
13 Indore 19.64 40
14 Bhopal 17.98 22
15 Visakhapatnam 17.28 45
16 Patna 16.84 40
17 Vadodara 16.71 20
18 Rajkot 12.87 14
19 Raipur 10.1 28
20 Thiruvananthapuram 7.44 71
21 Kochi 6.02 150
22 Nanded 5.5 3
23 Mangalore 4.89 37
24 Udaipur 4.51 17
25 Latur 3.83 3
For Surat city, daily take-off and landing enplanements is 5 which is very less compare to other
cities like Chennai, Kolkata which has almost same population as Surat. Even for cities with less
population than Surat like Pune, Jaipur, Lucknow, Nagpur, Kochi, Visakhapatnam etc. daily take-
off and landing enplanements are more than Surat.
Page | 34
CHAPTER 4: DATA COLLECTION
4.1 GENERAL
After deciding the methodology of the research work the relevant necessary data is collected. There
are two types of data viz., primary and secondary. The primary data are those which are collected
afresh and for the first time, and thus happen to be original in character. The secondary data, on
the other hand, are those which have already been collected by someone else and which have
already been passed through the statistical process.
As per the objectives for understanding mode choice behavior of travellers, a mode choice survey
is conducted at Surat airport. Currently there are daily two flights to Delhi and for 3 days in week
there is one flight to Mumbai. The survey is conducted in departure lounge of the airport with due
permission from Airport Director. The survey is conducted during flight timings for two weeks.
The passengers are selected randomly in the departure hall for interview for the survey.
4.2.1 Pilot Survey
Before main survey, a pilot survey is conducted at SVNIT campus. In this survey the faculty
members of various departments and students who make long distance trips from Surat were
interviewed. After pilot survey some changes in the questionnaire were made to make it more
effective.
4.2.2 Design of Questionnaire
The questionnaire of the survey is divided into four parts: 1. Passenger profile 2. Journey details
3. Mode preference 4. Travelling attributes rating
Part 1 collects the information about socio-economic characteristics of travellers like are age
group, profession and income class. Journey details about traveller’s recent long distance trips
from Surat by road, rail or air mode including destinations, mode used, journey duration and cost
of each trips are collected in Part 2. Journey expenditure includes the fare price or travel cost and
Page | 35
the incidental costs in journey. It has been also asked whether the journey is sponsored or not.
Travel time is divided into access time, waiting time, in rail/plane/vehicle journey time and egress
time. In Part 3, the travellers have been asked about their mode preference between air, rail and
road for making long distance journey for various travelling distance for each business/work as
well as social/leisure purpose. The willingness to pay for given travelling distance range is also
asked in this part. Part 4 consists of ratings of travelling attributes on a scale of 1 to 10 of travel
time, waiting time, travel cost, frequency, comfort, reliability and interchanges included. The
respondents have been asked to give ratings to these travelling attributes based on their importance
for business/work purpose travel and social/leisure purpose travel separately. Questionnaire is
placed in Annexure 1.
Page | 36
4.3 SECONDARY DATA
The GDP data for MMR, CMA and Pune district is obtained from data.gov.in website which is
open government data platform for India. Since district level GDP data in Gujarat was not
available, GDP data for Ahmedabad, Surat and Vadodara is generated using GDP data which was
Page | 37
available for Gujarat state. To generate the GDP data, the ratio of population of given district to
state population is taken. Then this ratio is multiplied with GDP of Gujarat state. The generated
data is shown in table 4.1.
Table 4.1: Generated GDP at current rate data for Surat, Ahmedabad and Vadodara district
The overall population data and GDP data used for analysis is given in Annexure 3 and 4.
Page | 38
CHAPTER 5: DATA ANALYSIS
5.1 GENERAL
After collection of necessary primary and secondary data, different types of analysis are done to
achieve the defined objectives. The data analysis chapter is divided in to three parts based on the
data collected through personal interviews of passengers at Surat airport as well as demographic
and economic indicators data obtained for other similar cities.
i)Time Series Analysis: In this the monthly domestic air passenger data at Mumbai, Pune,
Ahmedabad, Chennai and Vadodara is analysed to find out the monthly variation in terms of
seasonality indices.
ii)Econometric Analysis: In this analysis, the various elasticity values of domestic air travel
demand with respect to population and GDP of districts is calculated.
iii)Mode choice analysis: The data obtained from questionnaire is analysed to understand long
distance trip making behaviour of travellers of Surat region.
A time series is a collection of observations made sequentially in time. Most time series analysis
methods are concerned with the decomposition of the series into trend, cyclical, seasonal, and
irregular components.
The decomposition into the aforementioned components is achieved by using smoothing
models. We assume that the seasonality is multiplicative. That is, the seasonal variations are
directly proportional to the observations, as opposed to additive seasonality in which these
variations are assumed to be constant, as mentioned in chapter 2.
Page | 39
By using monthly domestic passenger data for last 9 years, the seasonality factors for districts
which are taken as case studies are calculated using moving average method. Calculation of indices
for other cities is placed at Appendix
1.First the 12-month moving average of domestic air travel demand for Ahmedabad is calculated.
In our case, since we have monthly data for each year and we want to estimate seasonality in each
month,12 month moving average is taken.
2. Then the centred moving average is calculated by taking average of first observation and the
next observation. This central moving average value gives us the baseline for the time series.
3. Compute the ratio of each observation in demand to the corresponding centred moving average
observation which gives the (St×It) component (seasonality and irregularity component).
4.Now to find out the seasonal component (St) for each month the average value of all (St×It)
component corresponding to the given month is calculated to remove irregularity component. For
example, in given table to find out seasonality index for August month the average of all (St×It)
values for August month in each year is taken.
Seasonality for August month =[(St×It) aug2007 + (St×It) aug2008 +………..+ (St×It)aug2007 ] / 8
Similarly, the monthly seasonality indices are calculated for Mumbai, Pune, Vadodara, And
Chennai.
i) if it is less than 1 it indicates the demand value for that month is less than average of
all months. (indicated by green colour in table)
ii) if it is more than 1 it indicates the demand value for that month is more than average
of all months. (indicated by red colour in table)
From table 5.1 considering overall average value of seasonality indices, it can be said that in
months January, May, June, November and December, the demand is above the monthly average
of all months. From this it can be said that there is more demand of domestic air passengers in
Page | 40
these months. And for rest of the months the demand is below the monthly average that is there is
less demand of domestic air passengers.
From above table, it can also be inferred that there is similarity in seasonal monthly patterns of
domestic passenger demand for different airports in India.
The long-term elasticity of air travel demand is calculated with respect to population and GDP for
airports in each selected districts and metropolitan regions. To do this the following data is used.
Page | 41
Table 5.2: GDP and Population data for Mumbai, Pune and Chennai airport
Table 5.3: GDP and Population data for Ahmedabad and Vadodara airports
Ahmedabad Vadodara
Annual GDP (in Annual GDP (in
YEAR
domestic Crore Population domestic Crore Population
passengers Rs.) passengers Rs.)
2001-02 596085 - 5816519 238804 - 3641802
2002-03 636064 - 5943139 361302 - 3690966
2003-04 724236 - 6072515 312447 - 3740794
2004-05 916548 23627 6204708 361894 14438 3791295
2005-06 1438969 28546 6339778 360489 17304 3842478
2006-07 2092026 33223 6477788 404260 19976 3894351
2007-08 2461909 38717 6618803 501744 23091 3946925
2008-09 2141609 43432 6762888 444573 25695 4000208
2009-10 2678756 51114 6910109 500852 29996 4054211
2010-11 3216542 62059 7060536 598335 36125 4108943
From above data, the long-term elasticity and long term arc elasticity is calculated for each airport.
The elasticity values are calculated with respect to population and GDP for long term period of 10
years. Since in case of Ahmedabad and Vadodara GDP data is insufficient elasticity value with
respect to GDP is calculated for 5-year period. (GDP value considered in analysis are current rate)
Page | 42
Table 5.4: Long term elasticity values for Mumbai, Pune and Chennai airports
Table 5.5: Long term elasticity values for Ahmedabad and Vadodara airports
From these long-term elasticity values, it can be said that demand of domestic air travel demand
is more sensitive to change is population rather than the change GDP value. Elasticity values with
respect to population for every airport is more than 1 which means the demand is elastic with
respect to population for all airports. Also, elasticity values with respect to GDP are more than 1
for Ahmedabad, Pune and Chennai but Mumbai and Vadodara airports elasticity values w.r.t GDP
are less than 1 which means for Mumbai and Vadodara the domestic air travel demand is inelastic
with respect to GDP in long term.
The data collected through questionnaire samples is entered into excel and SPSS to carry out travel
mode choice analysis. The profile of traveller’s socio-economic characteristics based on the age,
income and profession are analysed. The destinations where people are going from Surat is also
noted. The analysis of preferred modes out of air, rail and road by respondents for three distance
range (<500 km, 500-1000 km, >1000 km). The effect of sponsored trips on trip making behaviour
is also analysed. The willingness to pay by each category of income class is analysed for travelling
Page | 43
different trip lengths. The importance given by travellers to various travelling attributes based on
purpose of travel is also compared.
>60 up to 18
46-60 4% 5%
19-25
14%
15%
36-45
25%
26-35
37%
From above chart, it can be seen that major percentage of travellers belong to age category 26-35
and 36-45 that is the active age groups engaged in income earning activities.
Page | 44
other
Student 4% Service (govt)
10% 18%
Business
Service (Pvt)
36%
32%
From above chart, it can be seen that the maximum travellers travelling long distance do business
and private service.
10 to 20 lakh
28% 5 to 10 lakh
40%
Page | 45
5.3.1.4 Destination analysis
Since the survey is conducted at the Surat Airport during Delhi flight time, out of 813 journey
details given by travellers, more than 50% of passengers interviewed where making onward trips
or return trips to Delhi. To find out potential destinations other than Delhi, journey destinations by
passengers is analysed as follow:
Figures it can be observed that people are making more long trips to metro-cities of Mumbai,
Bangalore, Kolkata, Chennai, Ahmedabad, Chandigarh, Hyderabad and Jaipur. The other
destinations which are comparatively less travelled are Cochin, Trivandrum, Amritsar, Ajmer,
Vadodara, Beed, Warangal, Vishakhapatnam, Nagpur, Lucknow, Surathkal, Jalgoan, Jammu,
Calicut, Coimbatore, Dehradun, Nashik.
Page | 46
Table 5.8: Zone wise analysis of destinations
0.12%
6.66%
21.21%
15.78%
56.10%
0.37%
Travellers making long trips to Northern India are more than 50%. This is due to most of travellers
which were interviewed were making return journey to Delhi from Surat airport and the survey is
conducted during Delhi flight timing. After Northern India, percentage of travellers going to
Western India are more which might be because of in this region, business hubs like Mumbai,
Ahmedabad and Pune are present and trading between Surat and these regions is more.
Page | 47
100.00%
80.00%
Travellers
60.00% 49.53% 53.74%
44.63% 40.42%
40.00% 32.94%
23.83% 26.64% 22.43%
20.00%
5.84%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
Figure 5.6: Mode preferences for travelling distance less than 500 km for business/work purpose
trip
100.00%
82.48%
80.00% 67.29% 67.76%
Travellers
60.00%
40.00% 31.31%
16.12% 16.12% 16.59%
20.00%
1.40% 0.93%
0.00%
1 2 3
Mode preferences
Figure 5.7: Mode preferences for travelling distance in range of 500 to 1000 km for business/work
purpose trip
60.00%
40.00%
20.00% 4.91%0.00% 3.04% 3.04% 1.64%1.40%
0.00%
1 2 3
Mode preferences
Figure 5.8: Mode preferences for travelling distance more than 1000 km for business/work purpose
trip
Page | 48
100.00%
80.00% 71.46%
Travellers
60.00% 44.47% 49.56%
43.58%
40.00% 34.07%
21.90%
11.95% 16.37%
20.00% 6.64%
0.00%
1 2 3
Mode preferences
Figure 5.9: Mode preferences for travelling distance less than 500 km for social/leisure purpose trip
100.00%
74.56%
80.00%
56.19%
Travellers
60.00%
41.15% 42.26%
40.00% 35.18%
22.57% 23.23%
20.00%
2.65% 2.21%
0.00%
1 2 3
Mode preferences
Figure 5.10: Mode preferences for travelling distance in range of 500 to 1000 km for social/leisure
purpose trip
80.00%
Travellers
60.00%
40.00%
20.00% 9.07% 6.19% 3.76%
0.22% 2.65%1.77%
0.00%
1 2 3
Mode preferences
Figure 5.11: Mode preferences for travelling distance more than 1000 km for social/leisure purpose
trip
Page | 49
It is observed for business/work and social/leisure purpose trips to as the travelling distance
increases more number of travellers are preferring air as their first preference. For distance ranges
less than 500 km, 500-1000 km and more than 1000 km, the percentage of travellers preferring air
as first mode are 23.83%,67.29% and 95.09% respectively for business purpose trips. Similarly,
for distance ranges less than 500 km, 500-1000 km and more than 1000 km, the percentage of
travellers taking air as first mode are 11.95%,41.15% and 90.71% respectively for social/leisure
purpose trips, for travelling distance up to 500 km, most travellers are preferring rail as there first
mode.
Further, However, if we compare preferences based on purpose of trips, for given distance range,
more percentage of travellers are giving air as first preference for business/work purpose trips
compare to social/leisure purpose trips. For example, to travel for a distance up to 500 km 23.83%
travellers preferred air as their choice for business purpose trips compare 11.95% for social
purpose trips.
It can be observed that for distance up to 500 km, there is no significant change in percentage of
travellers choosing air as first preference for sponsored and non-sponsored trips, but in sponsored
trip, more percentage of travellers are preferring rail as first preference compare to a non-sponsored
trip.
100.00%
80.00%
Travellers
60.00% 54.13%
48.76%
43.39%
38.02%
40.00% 31.40% 30.58%
25.21% 20.66%
20.00% 7.85%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
Figure 5.12: Mode preference for non-sponsored trips up to 500 km travelling distance
Page | 50
100.00%
80.00%
57.23%
Travellers
60.00% 49.40%
39.16% 46.99%
40.00% 33.73%
23.49% 27.11%
19.28%
20.00%
3.61%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
Figure 5.13: Mode preference for Sponsored trips up to 500 km travelling distance
For distance range 500 to 1000 km, there is significant difference in percentage of travellers
choosing air as first preference which is 67.36% for non-sponsored trips and 71.08% for sponsored
trips. Also percentage of travellers choosing rail as first preference is more in case of sponsored
trips compare to non-sponsored trips.
100.00%
81.82%
80.00% 67.36% 68.18%
Travellers
60.00%
40.00% 30.99%
17.36%
15.29% 16.53%
20.00%
1.65% 0.83%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
Figure 5.14: Mode preference for non-sponsored trips for 500 to 1000km travelling distance
100.00% 86.75%
80.00% 71.08% 71.69%
Travellers
60.00%
40.00% 27.71%
16.27% 12.65%
20.00% 12.05%
1.20% 0.60%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
Figure 5.15: Mode preference for Sponsored trips for 500 to 1000km travelling distance
Page | 51
For distance more than 1000 km, there is no significant difference in percentage of travellers
choosing air as first preference.
120.00% 97.11%
95.87% 94.21%
100.00%
Travellers
80.00%
60.00%
40.00% 1.65%
20.00% 4.13%
0.00% 2.89% 2.89% 1.24%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
.
Figure 5.16: Mode preference for Non-sponsored trips for more than 1000km travelling distance
120.00%
94.58% 94.58% 98.19%
100.00%
Ttravellers
80.00%
60.00%
40.00%
1.20%
20.00% 5.42% 3.61% 1.81% 0.60%
0.00%
0.00%
1 2 3
Mode preferences
AIR RAIL ROAD
Figure 5.17: Mode preference for Sponsored trips for more than 1000km travelling distance
Table 5.9: Willingness to pay analysis for travellers with annual income less than 5 lakhs
Page | 52
Above 1000 Business/work 87% 13% 0%
5211 13000 800
km Social/leisure 87% 13% 0%
Table 5.10: Willingness to pay analysis for travellers with annual income between 5 to 10 lakhs
Table 5.11: Willingness to pay analysis for travellers with annual income between 10 to 20 lakhs
Table 5.12: Willingness to pay analysis for travellers with annual income more than 20 lakhs
From above tables, it can be observed that for income classes as travelling distance increases the
average willingness to pay is also increasing. As the income of travellers is increasing, they are
Page | 53
willing to pay more money for given travelling distance. For example, for distance above 1000
km, average willingness to pay amount for income class 1 is Rs. 5211 while for income class 4,
this amount is Rs. 8982 which is almost Rs. 3800 more compare to income class 1.
Also with increasing in travelling distance more percentage of travellers are choosing air as first
preference for long distance trips. For example, for income class 1, for travelling distance less than
500 km average willingness to pay amount is Rs.1584 while for travelling distance above 1000
km it is Rs. 5211. Compare to social purpose trips, travellers choosing air as first preference is
more in case of business purpose trips. As the income of travellers increases the willingness to pay
is also increasing
Travelling Rating
attributes 1 2 3 4 5 6 7 8 9 10
Travel time 0% 0% 0% 0% 1% 1% 1% 15% 18% 63%
Waiting time 1% 2% 1% 1% 3% 2% 10% 27% 26% 28%
Travel cost 1% 1% 2% 4% 18% 7% 22% 25% 11% 10%
Frequency 0% 1% 1% 2% 7% 5% 13% 29% 21% 20%
Comfort 0% 1% 0% 0% 3% 3% 9% 28% 31% 26%
Reliability 0% 0% 0% 1% 4% 3% 7% 19% 26% 39%
Interchanges 3% 1% 0% 4% 8% 6% 16% 27% 16% 19%
Page | 54
70%
60%
50%
PERCENTAGE
40%
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10
RATINGS
travel time waiting time travel cost frequency comfort reliablity interchanges
Figure 5.18: Graphical representation of travelling attributes ratings for business purpose trips
Form Figure 5.18, it can be seen that 63% of travellers have given 10 rating for travel time implying
that travel time is most important factor for business/social purpose trips for 63% of travellers.
Also 39% of travellers have given reliability the highest importance. After travel time and
reliability, travellers have given waiting time and comfort high importance. In general, it is found
that for business/work purpose trips travellers given more importance to travel time, reliability,
waiting time and comfort while choosing the mode to travel.
Travelling RATING
attributes 1 2 3 4 5 6 7 8 9 10
Travel time 0% 0% 0% 0% 2% 3% 20% 28% 15% 30%
Waiting time 1% 2% 1% 1% 5% 4% 23% 32% 18% 13%
Travel cost 1% 0% 0% 3% 10% 5% 10% 23% 24% 24%
Frequency 0% 0% 3% 3% 16% 11% 18% 28% 10% 11%
Comfort 0% 0% 0% 0% 3% 3% 9% 25% 34% 26%
Reliability 0% 0% 0% 2% 3% 5% 9% 20% 26% 36%
Interchanges 2% 1% 0% 4% 7% 4% 9% 23% 24% 25%
Page | 55
40%
35%
30%
PERCENTAGE
25%
20%
15%
10%
5%
0%
1 2 3 4 5 6 7 8 9 10
RATINGS
travel time waiting time travel cost frequency comfort reliablity interchanges
Figure 5.19: Graphical representation of travelling attributes ratings for social purpose trips
Figure 5.19 shows that 36% of travellers have given reliability the highest importance. Compare
to business/work purpose trips, more importance is given to travel cost, interchanges, and less
importance is given to waiting time, frequency, travel time in social/leisure purpose trips. The
importance given to travelling attributes which are reliability, travel time, interchanges and travel
cost is more compare to rest of travelling attributes.
In this test equality of means between two adjacent groups were found out with 95% confidence
interval. Here the ratings (which indicates importance) given to each travelling attribute for
business/work purpose is compared with ratings given for social/leisure purpose and it is found
out whether there is significant difference in importance between each purpose. Here the null
hypothesis is “there is no significant difference in importance given to travelling attribute in
business/work purpose trip and social/leisure purpose trip” and the alternate hypothesis is “there
is significant difference in importance given to travelling attribute in business/work purpose trip
and social/leisure purpose trip”. The following are the results for t-test for equality of means.
Page | 56
Table 5.15: Group statistics for ratings given to travelling attribute for business purpose and social
purpose trip
Page | 57
Since the significant value is less than 0.05 for travelling attributes; travel time, waiting time, travel
cost, frequency and interchanges, the null hypothesis is rejected which means there is significant
difference importance given to these travelling attributes in business/work purpose trip and
social/leisure purpose trip. For travelling attributes; comfort and reliability the importance given
does not change based purpose of trips.
Table 5.17: Income wise average ratings given to travelling attributes for business purpose trips
Travel Time
10.00
9.00
8.00
Interchanges 7.00 Waiting time
6.00
5.00
4.00
3.00
2.00
1.00
0.00
Reliability Travel cost
Comfort Frequency
Figure 5.20: Spider diagram for income wise average ratings given to travelling attributes for
business purpose trips
Page | 58
From above spider diagram, it is found that there is no significant difference in ratings given to all
travelling attributes except travel cost by different income group. In case of travel cost,
significantly less importance is given by travellers with income class ‘above 20 lakhs’ compare to
other income class for business purpose trip
Page | 59
Table 5.19: Robust Tests of Equality of Means for business purpose trips
From ANOVA test and Robust tests of equality of means, it is found that these tests are failed at
confidence level of 95% for travelling attribute travel cost.
Table 5.20: Multiple Comparison by Post Hoc test for business purpose trips
95% Confidence
Mean
Std. Interval
Dependent Variable Difference Sig.
Error Lower Upper
(I-J)
Bound Bound
5 to 10 lakhs -.091 .172 .996 -.554 .373
Less than 5
10 to 20 lakhs -.281 .172 .490 -.745 .182
lakhs
Above 20 lakh -.296 .216 .681 -.873 .281
Less than 5 lakhs .091 .172 .996 -.373 .554
5 to 10
10 to 20 lakhs -.190 .118 .495 -.503 .122
Travel Time lakhs
Above 20 lakhs -.205 .176 .815 -.675 .265
Less than 5 lakhs .281 .172 .490 -.182 .745
10 to 20
5 to 10 lakhs .190 .118 .495 -.122 .503
lakhs
Above 20 lakhs -.015 .176 1.000 -.485 .456
Less than 5 lakhs .296 .216 .681 -.281 .873
Page | 60
Above 20 5 to 10 lakhs .205 .176 .815 -.265 .675
lakhs 10 to 20 lakhs .015 .176 1.000 -.456 .485
5 to 10 lakhs -.449 .257 .408 -1.139 .241
Less than 5
10 to 20 lakhs -.505 .297 .438 -1.300 .289
lakhs
Above 20 lakhs -.788 .329 .104 -1.668 .092
Less than 5 lakhs .449 .257 .408 -.241 1.139
5 to 10
10 to 20 lakhs -.056 .234 1.000 -.678 .565
lakhs
Waiting Above 20 lakhs -.339 .273 .771 -1.071 .393
Time Less than 5 lakhs .505 .297 .438 -.289 1.300
10 to 20
5 to 10 lakhs .056 .234 1.000 -.565 .678
lakhs
Above 20 lakhs -.282 .312 .936 -1.114 .549
Less than 5 lakhs .788 .329 .104 -.092 1.668
Above 20
5 to 10 lakhs .339 .273 .771 -.393 1.071
lakhs
10 to 20 lakhs .282 .312 .936 -.549 1.114
5 to 10 lakhs .014 .273 1.000 -.718 .745
Less than 5
10 to 20 lakhs .250 .290 .948 -.526 1.027
lakhs
Above 20 lakhs 1.68849* .332 .000 .802 2.575
Less than 5 lakhs -.014 .273 1.000 -.745 .718
5 to 10
10 to 20 lakhs .237 .229 .884 -.370 .844
lakhs
Above 20 lakhs 1.67493* .280 .000 .927 2.423
Travel Cost
Less than 5 lakhs -.250 .290 .948 -1.027 .526
10 to 20
5 to 10 lakhs -.237 .229 .884 -.844 .370
lakhs
Above 20 lakhs 1.43818* .297 .000 .646 2.230
Less than 5 lakhs -1.68849* .332 .000 -2.575 -.802
Above 20
5 to 10 lakhs -1.67493* .280 .000 -2.423 -.927
lakhs
10 to 20 lakhs -1.43818* .297 .000 -2.230 -.646
5 to 10 lakhs -.044 .265 1.000 -.754 .666
Less than 5
10 to 20 lakhs -.134 .272 .997 -.862 .594
lakhs
Above 20 lakhs -.496 .289 .427 -1.269 .277
Less than 5 lakhs .044 .265 1.000 -.666 .754
5 to 10
10 to 20 lakhs -.090 .214 .999 -.656 .477
lakhs
Above 20 lakhs -.452 .235 .293 -1.078 .174
Frequency
Less than 5 lakhs .134 .272 .997 -.594 .862
10 to 20
5 to 10 lakhs .090 .214 .999 -.477 .656
lakhs
Above 20 lakhs -.362 .243 .589 -1.009 .285
Less than 5 lakhs .496 .289 .427 -.277 1.269
Above 20
5 to 10 lakhs .452 .235 .293 -.174 1.078
lakhs
10 to 20 lakhs .362 .243 .589 -.285 1.009
5 to 10 lakhs -.415 .210 .273 -.981 .152
Less than 5
10 to 20 lakhs -.281 .225 .766 -.885 .323
lakhs
Above 20 lakhs -.319 .271 .809 -1.045 .406
Comfort
Less than 5 lakhs .415 .210 .273 -.152 .981
5 to 10
10 to 20 lakhs .134 .161 .956 -.293 .560
lakhs
Above 20 lakhs .095 .221 .999 -.496 .687
Page | 61
Less than 5 lakhs .281 .225 .766 -.323 .885
10 to 20
5 to 10 lakhs -.134 .161 .956 -.560 .293
lakhs
Above 20 lakhs -.038 .235 1.000 -.666 .589
Less than 5 lakhs .319 .271 .809 -.406 1.045
Above 20
5 to 10 lakhs -.095 .221 .999 -.687 .496
lakhs
10 to 20 lakhs .038 .235 1.000 -.589 .666
5 to 10 lakhs -.004 .228 1.000 -.613 .606
Less than 5
10 to 20 lakhs -.091 .238 .999 -.729 .546
lakhs
Above 20 lakhs -.339 .256 .712 -1.024 .345
Less than 5 lakhs .004 .228 1.000 -.606 .613
5 to 10
10 to 20 lakhs -.088 .199 .998 -.614 .439
lakhs
Above 20 lakhs -.335 .219 .561 -.920 .249
Reliability
Less than 5 lakhs .091 .238 .999 -.546 .729
10 to 20
5 to 10 lakhs .088 .199 .998 -.439 .614
lakhs
Above 20 lakhs -.248 .230 .865 -.862 .366
Less than 5 lakhs .339 .256 .712 -.345 1.024
Above 20
5 to 10 lakhs .335 .219 .561 -.249 .920
lakhs
10 to 20 lakhs .248 .230 .865 -.366 .862
5 to 10 lakhs .430 .314 .681 -.411 1.271
Less than 5
10 to 20 lakhs .007 .321 1.000 -.851 .865
lakhs
Above 20 lakhs .137 .359 .999 -.823 1.097
Less than 5 lakhs -.430 .314 .681 -1.271 .411
5 to 10
10 to 20 lakhs -.423 .259 .480 -1.109 .263
lakhs
Above 20 lakhs -.293 .305 .916 -1.108 .521
Interchanges
Less than 5 lakhs -.007 .321 1.000 -.865 .851
10 to 20
5 to 10 lakhs .423 .259 .480 -.263 1.109
lakhs
Above 20 lakhs .130 .312 .999 -.703 .962
Less than 5 lakhs -.137 .359 .999 -1.097 .823
Above 20
5 to 10 lakhs .293 .305 .916 -.521 1.108
lakhs
10 to 20 lakhs -.130 .312 .999 -.962 .703
Post Hoc test is carried out considering that equal variances are not assumed. Tamhane’s T2 test
was performed in Post Hoc test and the results obtained are shown in above table. The above table
clearly shows that there is significant difference in importance given for travelling attribute ‘travel
cost’ only between income class ‘above 20 lakhs’ and rest of the income class.
Page | 62
Table 5.21: Income wise average ratings given to travelling attributes for social purpose trips
Travel Time
9.00
8.00
7.00
Interchanges 6.00 Waiting time
5.00
4.00
3.00
2.00
1.00
0.00
Reliability Travel cost
Comfort Frequency
Figure 5.21: Spider diagram for income wise average ratings given to travelling attributes for social
purpose trips
Similar to ratings analysis for business/work purpose trips, for social/leisure purpose trips also
there in case of travel cost, significantly less importance is given by travellers with income class
‘above 20 lakhs’
Table 5.23 Robust Tests of Equality of Means for social purpose trips
Travelling
attributes Statistic df1 df2 Sig.
Travel Time Welch .914 3 168.269 .435
Brown-Forsythe .830 3 286.914 .478
Waiting Time Welch .209 3 164.776 .890
Brown-Forsythe .225 3 288.397 .879
Travel Cost Welch 10.140 3 166.749 .000
Brown-Forsythe 12.200 3 287.510 .000
Frequency Welch .849 3 170.661 .469
Brown-Forsythe .882 3 309.953 .451
Comfort Welch 2.752 3 163.286 .044
Brown-Forsythe 2.552 3 283.557 .056
Reliability Welch 1.472 3 174.211 .224
Brown-Forsythe 1.508 3 330.139 .212
Interchanges Welch .518 3 175.342 .670
Brown-Forsythe .563 3 327.088 .640
Table 5.24: Multiple Comparison by Post Hoc test for social purpose trips
95% Confidence
Mean Interval
Difference Std. Lower Upper
Dependent Variable (I-J) Error Sig. Bound Bound
Travel time Less than 5 to 10 lakhs .189 .221 .951 -.403 .780
5 lakhs 10 to 20 lakhs -.075 .220 1.000 -.664 .514
Page | 64
Above 20 lakhs .164 .271 .991 -.559 .887
5 to 10 Less than 5 lakhs -.189 .221 .951 -.780 .403
lakhs 10 to 20 lakhs -.263 .171 .549 -.716 .190
Above 20 lakhs -.024 .233 1.000 -.646 .598
10 to 20 Less than 5 lakhs .075 .220 1.000 -.514 .664
lakhs 5 to 10 lakhs .263 .171 .549 -.190 .716
Above 20 lakhs .239 .232 .886 -.380 .859
Above 20 Less than 5 lakhs -.164 .271 .991 -.887 .559
lakhs 5 to 10 lakhs .024 .233 1.000 -.598 .646
10 to 20 lakhs -.239 .232 .886 -.859 .380
Waiting time Less than 5 to 10 lakhs -.128 .261 .997 -.829 .573
5 lakhs 10 to 20 lakhs -.059 .289 1.000 -.830 .713
Above 20 lakhs -.240 .330 .977 -1.121 .641
5 to 10 Less than 5 lakhs .128 .261 .997 -.573 .829
lakhs 10 to 20 lakhs .069 .219 1.000 -.512 .650
Above 20 lakhs -.112 .271 .999 -.837 .613
10 to 20 Less than 5 lakhs .059 .289 1.000 -.713 .830
lakhs 5 to 10 lakhs -.069 .219 1.000 -.650 .512
Above 20 lakhs -.181 .297 .991 -.975 .613
Above 20 Less than 5 lakhs .240 .330 .977 -.641 1.121
lakhs 5 to 10 lakhs .112 .271 .999 -.613 .837
10 to 20 lakhs .181 .297 .991 -.613 .975
Travel cost Less than 5 to 10 lakhs -.229 .267 .950 -.947 .489
5 lakhs 10 to 20 lakhs .052 .286 1.000 -.715 .818
Above 20 lakhs 1.34868* .339 .001 .441 2.256
5 to 10 Less than 5 lakhs .229 .267 .950 -.489 .947
lakhs 10 to 20 lakhs .281 .221 .749 -.307 .868
Above 20 lakhs 1.57776* .287 .000 .810 2.345
10 to 20 Less than 5 lakhs -.052 .286 1.000 -.818 .715
lakhs 5 to 10 lakhs -.281 .221 .749 -.868 .307
Above 20 lakhs 1.29683* .305 .000 .484 2.110
Above 20 Less than 5 lakhs -1.34868* .339 .001 -2.256 -.441
lakhs 5 to 10 lakhs -1.57776* .287 .000 -2.345 -.810
10 to 20 lakhs -1.29683* .305 .000 -2.110 -.484
Frequency Less than 5 to 10 lakhs -.156 .274 .994 -.892 .579
5 lakhs 10 to 20 lakhs -.265 .285 .928 -1.026 .497
Above 20 lakhs -.494 .326 .572 -1.365 .377
5 to 10 Less than 5 lakhs .156 .274 .994 -.579 .892
lakhs 10 to 20 lakhs -.108 .227 .998 -.710 .493
Above 20 lakhs -.338 .277 .782 -1.077 .402
10 to 20 Less than 5 lakhs .265 .285 .928 -.497 1.026
lakhs 5 to 10 lakhs .108 .227 .998 -.493 .710
Above 20 lakhs -.230 .287 .964 -.995 .536
Less than 5 lakhs .494 .326 .572 -.377 1.365
Page | 65
Above 20 5 to 10 lakhs .338 .277 .782 -.402 1.077
lakhs 10 to 20 lakhs .230 .287 .964 -.536 .995
Comfort Less than 5 to 10 lakhs -.472 .196 .103 -1.000 .055
5 lakhs 10 to 20 lakhs -.153 .216 .980 -.731 .425
Above 20 lakhs -.438 .243 .369 -1.087 .212
5 to 10 Less than 5 lakhs .472 .196 .103 -.055 1.000
lakhs 10 to 20 lakhs .319 .158 .239 -.100 .739
Above 20 lakhs .035 .193 1.000 -.483 .552
10 to 20 Less than 5 lakhs .153 .216 .980 -.425 .731
lakhs 5 to 10 lakhs -.319 .158 .239 -.739 .100
Above 20 lakhs -.285 .213 .705 -.854 .284
Above 20 Less than 5 lakhs .438 .243 .369 -.212 1.087
lakhs 5 to 10 lakhs -.035 .193 1.000 -.552 .483
10 to 20 lakhs .285 .213 .705 -.284 .854
Reliability Less than 5 to 10 lakhs -.100 .227 .998 -.706 .507
5 lakhs 10 to 20 lakhs -.110 .234 .998 -.735 .514
Above 20 lakhs -.499 .261 .302 -1.195 .198
5 to 10 Less than 5 lakhs .100 .227 .998 -.507 .706
lakhs 10 to 20 lakhs -.011 .195 1.000 -.529 .507
Above 20 lakhs -.399 .227 .397 -1.004 .206
10 to 20 Less than 5 lakhs .110 .234 .998 -.514 .735
lakhs 5 to 10 lakhs .011 .195 1.000 -.507 .529
Above 20 lakhs -.388 .234 .465 -1.012 .235
Above 20 Less than 5 lakhs .499 .261 .302 -.198 1.195
lakhs 5 to 10 lakhs .399 .227 .397 -.206 1.004
10 to 20 lakhs .388 .234 .465 -.235 1.012
Interchanges Less than 5 to 10 lakhs .210 .303 .982 -.598 1.018
5 lakhs 10 to 20 lakhs -.115 .300 .999 -.918 .687
Above 20 lakhs .028 .355 1.000 -.921 .977
5 to 10 Less than 5 lakhs -.210 .303 .982 -1.018 .598
lakhs 10 to 20 lakhs -.325 .262 .768 -1.020 .370
Above 20 lakhs -.182 .324 .994 -1.046 .681
10 to 20 Less than 5 lakhs .115 .300 .999 -.687 .918
lakhs 5 to 10 lakhs .325 .262 .768 -.370 1.020
Above 20 lakhs .143 .321 .998 -.715 1.001
Above 20 Less than 5 lakhs -.028 .355 1.000 -.977 .921
lakhs 5 to 10 lakhs .182 .324 .994 -.681 1.046
10 to 20 lakhs -.143 .321 .998 -1.001 .715
Similar to rating analysis results for business trips, there is significant difference in importance
given for travelling attribute ‘travel cost’ only between income class ‘above 20 lakhs’ and rest of
the income class for social purpose trip.
Page | 66
CHAPTER 6: DEVELOPMENT OF MODELS
6.1 GENERAL
Three models have been developed for different purposes. The models developed are as follows:
i) ARIMA model for short term forecast
For medium-term forecasting of domestic air travel demand for airports at Mumbai, Ahmedabad,
Vadodara, Pune and Chennai, ARIMA modelling technique is used. For ARIMA modelling,
domestic air passenger data from year 1995-1996 to 2015-16 i.e., 21 years’ data is used. Out 21
years’ data,16 data points (from year 1995-96 to 2010-11) have been used for formulation models
and using rest of the 5 years’ data (from year 2011-12 to 2015-16) the validation part for models
is done by trial and error method. To do ARIMA modelling analysis Gretl (Gnu Regression,
Econometrics and Time-series Library) software is used.
Test for stationarity: In order to model a time series for ARIMA method, the series has to be
stationary. Stationary time series data means a time series which do not have trend or seasonal
variation. Time series data is stationary when, mean, variance and coefficient correlation of them
is constant during the time. Stationarity tests is used to determine if differencing is necessary. One
of the method to identify whether data is stationary or not is ADF (Augmented Dickey- Fuller) test.
Hence ADF test is conducted for time series data of Mumbai, Pune, Vadodara, Chennai and
Ahmedabad airports. For data to be stationary the asymptotic p-value should be less 0.05. It can
be seen that after 1st order differencing, time series data is becoming stationary with asymptotic p-
value less than 0.05. Hence the differencing is necessary for making stationary.
Asymptotic p-value
For given After 1st order
time series differencing
Mumbai 0.81 0.05
Pune 0.99 0.00
Vadodara 1 0.00
Chennai 0.21 0.00
Ahmedabad 0.41 0.00
To determine ‘p’ and ‘q’, the autocorrelation function (ACF) and partial autocorrelation
(PACF) plots. Since the data points are insufficient to interpret ACF and PACF plots, the p and q
values are found using trial and error method.
3. Estimate suitable ARIMA model by trial and error method by checking whether all coefficients
related AR and MA components comes significant or not.
Page | 68
6.2.2 Formulation of ARIMA Models for Airports
ARIMA
Coefficient Std. Error z p-value significance
Sr.No. models
constant 868471.00 417067.00 2.08 0.04 **
1 ARIMA(1,1,1) phi_1 −0.690959 0.24 −2.9078 0.00 ***
theta_1 1.00 0.31 3.25 0.00 ***
constant 904507.00 433159.00 2.09 0.04 **
2 ARIMA(0,1,1)
theta_1 0.14 0.32 0.44 0.66
constant 903931.00 421445.00 2.14 0.03 **
3 ARIMA(1,1,0)
phi_1 0.09 0.27 0.35 0.72
constant 160404.00 81149.50 1.98 0.05 **
4 ARIMA(1,2,1) phi_1 −0.0126904 0.27 −0.0467 0.96
theta_1 −1 0.20 −5.0535 <0.0001 ***
From above table it can be seen that only ARIMA (1,1,1) model is suitable with constant and other
coefficients (coefficients related to AR and MA component of ARIMA model) being significant
based on p-value.
Page | 69
From above table, it can be seen that only ARIMA (0,1,1) model is suitable with constant and
coefficient related to MA component being significant based on p-value.
From above table it can be seen that only ARIMA (0,2,2) model is suitable with constant and
coefficients related to two MA components being significant based on p-value.
Page | 70
Table 6.6: ARIMA model for Chennai airport
ARIMA
Coefficient Std. Error z p-value significance
Sr.No. models
constant 438389 196854 2.23 0.03 **
1 ARIMA(1,1,1) phi_1 −0.371729 0.25028 −1.4853 0.14
theta_1 1 0.203823 4.91 <0.0001 ***
constant 426220 231679 1.84 0.07 *
2 ARIMA(0,1,1)
theta_1 0.56424 0.296578 1.90 0.06 *
constant 423021 213578 1.98 0.05 **
3 ARIMA(1,1,0)
phi_1 0.264624 0.237605 1.11 0.27
constant 439322 173089 2.54 0.01 **
phi_1 −0.0926775 0.822274 −0.1127 0.91
4 ARIMA(0,2,2)
theta_1 0.695631 0.82972 0.84 0.40
theta_2 −0.304368 0.802094 −0.3795 0.70
From above table it can be seen that only ARIMA (0,1,1) model is suitable with constant and
coefficient related to MA component being significant based on p-value.
Std.
Obs Demand Prediction 95% Interval
Error
2011 21043600 20298000 1370880 (1.76111e+007, 2.29848e+007)
2012 20277000 21557500 2258240 (1.71314e+007, 2.59836e+007)
2013 21880700 22155800 2712030 (1.68403e+007, 2.74712e+007)
2014 25205300 23210900 3202960 (1.69332e+007, 2.94886e+007)
2015 30045900 23950400 3564550 (1.69640e+007, 3.09368e+007)
Page | 71
Table 6.8: Forecast evaluation statistics of ARIMA (1,1,1) model for Mumbai airport
Std.
Obs DEMAND Prediction 95% Interval
Error
2011 3950170 3426700 260938 (2.91527e+006, 3.93813e+006)
2012 3344540 3584410 455985 (2.69069e+006, 4.47812e+006)
2013 3566450 3742110 589708 (2.58631e+006, 4.89792e+006)
2014 3834200 3899820 698276 (2.53123e+006, 5.26842e+006)
2015 4911580 4057530 792100 (2.50504e+006, 5.61002e+006)
Table 6.10: Forecast evaluation statistics of ARIMA (0,1,1) model for Ahmedabad airport
Page | 72
For Vadodara Airport:
Std.
Obs Demand Prediction 95% Interval
Error
2011 669931 618662 36838 (546461., 690862.)
2012 676072 665582 36869 (593320., 737843.)
2013 686235 715338 36994 (642832., 787844.)
2014 712441 767931 37272 (694879., 840983.)
2015 931092 823360 37762 (749348., 897373.)
Table 6.12: Forecast evaluation statistics of ARIMA (0,2,2) model for Vadodara airport
Std.
Obs Demand Prediction 95% Interval
Error
2011 3229350 3056280 173464 (2.71630e+006, 3.39626e+006)
2012 3234630 3225940 325009 (2.58894e+006, 3.86295e+006)
2013 3495540 3395610 425643 (2.56136e+006, 4.22985e+006)
2014 4067700 3565270 506668 (2.57222e+006, 4.55832e+006)
2015 5176210 3734930 576414 (2.60518e+006, 4.86468e+006)
Page | 73
Table 6.14: Forecast evaluation statistics of ARIMA (0,1,1) model for Pune airport
Std.
Obs Demand Prediction 95% Interval
Error
2011 3229350 3124500 179328 (2.77302e+006, 3.47598e+006)
2012 3234630 3405450 327887 (2.76281e+006, 4.04810e+006)
2013 3495540 3638250 461675 (2.73338e+006, 4.54311e+006)
2014 4067700 3845480 580451 (2.70782e+006, 4.98314e+006)
2015 5176210 4039150 686233 (2.69415e+006, 5.38414e+006)
Table 6.16: Forecast evaluation statistics of ARIMA (1,1,0) model for Pune airport
Std.
Obs Demand Prediction 95% Interval
Error
2011 8617180 8272360 617158 (7.06275e+006, 9.48197e+006)
2012 8314340 8668480 1179390 (6.35691e+006, 1.09800e+007)
Page | 74
2013 8358380 9064590 1549530 (6.02756e+006, 1.21016e+007)
2014 9592060 9460710 1846930 (5.84080e+006, 1.30806e+007)
2015 10340900 9856830 2102670 (5.73567e+006, 1.39780e+007)
Table 6.18: Forecast evaluation statistics of ARIMA (0,1,1) model for Chennai airport
MAPE values of best suited ARIMA models developed for Mumbai, Ahmedabad, Vadodara,
Pune and Chennai airports are 7.86%, 8.89%, 6.56%, 8.01% and 4.55% respectively.
To make long term forecast of domestic air travel demand for Mumbai, Ahmedabad, Vadodara,
Pune and Chennai airports, econometric model is used. In econometric model GDP and population
are considered as dependent variable. For Mumbai airport, the population and GDP of all the
districts in MMR is considered and for Chennai airport, the population and GDP of all districts in
CMA is considered. For Pune, Ahmedabad and Vadodara airports, the population and GDP of
corresponding districts is considered. For MMR, CMA and Pune district, the annual domestic air
travel demand, population and GDP values from year 2001-02 to 2011-12 is used for model
formulation while for Ahmedabad and Vadodara the annual domestic air travel demand,
population and GDP values from year 2004-05 to 2011-12 is used for model formulation due to
less data availability. For model validation part, corresponding data from year 2012-13 to 2015-
16 is used.
Page | 75
6.2.1 Formulation of Econometric Model
Population and GDP are used as variables for respective districts/region to estimate domestic air
travel demand model. For each district/region, three models are tested, i) by considering GDP and
population of district/region as independent variable ii) by considering only population of
district/region as independent variable and iii) by considering GDP as independent variable.
Depending on the significance of each variable and constant using t-statistic and R-square value
in equation obtained for econometric models, best model is selected. For making econometric
model, E-Views software is used and method used for finding out equation is least square method.
Page | 76
C -89432.65 285170.7 -0.31 0.76
2 0.932 0.921 0.000
GDP 55.67523 6.106983 9.12 0.00
C -15104498 2043049 -7.39 0.00
3 0.924 0.911 0.000
Population 2.607493 0.304626 8.56 0.00
Similar to Mumbai airport, for Ahmedabad airport, model with population as independent variable
and constant are significant in the equation. Even though in 1st and 2nd model shows better R-
square value, the variables and constant obtained in the equations are not significant. Hence 3rd
model is used for forecasting.
For Vadodara airport, though 1st model has high R-square value, the variable and constant are not
significant. Hence 2nd model with GDP as independent variable is used for forecast.
Page | 77
GDP 27.03264 1.716712 15.75 0.00
C -9577773 774786.2 -12.36 0.00
3 0.957 0.952 0.000
Population 1.329755 0.093 14.27 0.00
Similar to Vadodara airport, for Pune airport also though 1st model has high R-square value, the
variables and constant are not significant hence 2nd model is used for forecast.
For Chennai airport, also 2nd model gives high R-square value with GDP as independent variables
and a constant being significant.
Page | 78
Table 6.25: Forecast evaluation statistics of econometric model for Mumbai Airport
Observed Predicted
3.23 3.62
3.50 3.38
4.07 3.64
5.18 3.90
Table 6.27: Forecast evaluation statistics of econometric model for Mumbai Airport
Observed Predicted
0.68 0.71
0.69 0.81
0.71 0.81
0.93 0.86
Page | 79
Table 6.29: Forecast evaluation statistics of econometric model for Vadodara Airport
Observed Predicted
0.68 0.71
0.69 0.81
0.71 0.81
0.93 0.86
Table 6.31: Forecast evaluation statistics of econometric model for Pune Airport
Observed Predicted
8.31 9.17
8.36 9.83
9.59 10.48
10.34 11.14
Page | 80
Table 6.33: Forecast evaluation statistics of econometric model for Chennai Airport
From forecast evaluation statistics of econometric models developed for Mumbai, Ahmedabad,
Vadodara, Pune and Chennai airports, the MAPE values observed are 9.9%, 22.39%, 11.26%,
12.63% and 11.22% respectively.
For developing inter regional model, passenger profile data and travel details are collected from
questionnaire survey. For developing choice model, analysis of categorical dependent variable
with more than two responsive categories have to be done which in our case is mode choice for
long distance between air, rail, road, air-rail and air road. Hence to do analysis of this multinomial
data, multinomial logistic regression is used in which one of the reference category (in this case
one of the travel modes) has to be considered as aa reference category. From this multinomial
regression model is developed which will calculate log-odds for all other categories relative to
reference category.
Page | 81
multinomial logit model under discrete choice methods is an analytically convenient modeling
method.
Multinomial logistic regression compares multiple groups through a combination of binary logistic
regressions. It is sometimes considered an extension of binomial logistic regression to allow for a
dependent variable with more than two categories. Multinomial logistic regression has nominal
and/or continuous independent variables and can have interactions between independent variables
to predict the dependent variable.
In this study, mode choice analysis between long distance modes which are air, rail, road, air-rail
and air-road is done. In multinomial logistic regression, these modes are nominal dependent
variables and age, income, profession, travel cost and travel time are considered as independent
variables. Age, profession and income are the nominal or categorical independent variable while
travel time and travel cost are scale or continuous variables.
1.In model 1, travel time, travel cost, age, profession and income is taken independent variables
with an intercept constant.
Page | 82
From table 6.34, it can be seen that for intercept, Chi-square is 0. So, intercept cannot be tested
this model because removing the intercept simply causes one of the previously redundant factor
levels to become non-redundant. Because of this the parameters associated with the last category
of each categorical independent variable becomes redundant. Hence to avoid this in next trials
intercept is omitted.
2. In model 2, travel time, travel cost, age, profession and income are taken as predictors without
an intercept or a constant.
It can be interfered from table 6.34, that age and profession predictors does not significantly effect
the mode choice. Hence in next trial, profession and income variable is removed.
3. In model 3, travel time, travel cost and income are taken as predictors. In this models all the
predictors are significantly effecting the mode choice. Hence to estimate the parameter coefficients
this model is used.
Std.
MODES B Wald df Sig. Exp(B)
Error
Page | 83
TC .001 .000 17.153 1 .000 1.001
TT .004 .003 2.150 1 .143 1.004
[INCOME=1.00] -6.193 1.720 12.959 1 .000 .002
Air-Road
[INCOME=2.00] -4.392 1.223 12.897 1 .000 .012
[INCOME=3.00] -5.724 1.395 16.849 1 .000 .003
[INCOME=4.00] -4.975 1.355 13.480 1 .000 .007
Note: Here the reference category is road.
In above table, coefficients for travel time and travel cost are almost zero, but their significant.
Hence there is need to standardize the travel time and travel cost values.
Table 6.36: Descriptive Statistics for travel time and travel cost
Std.
N Minimum Maximum Mean Deviation
TRAVEL 745 200.0 17500.0 4934.3 2485.0
COST
TRAVEL 745 130.0 2340.0 439.5 311.9
TIME
Valid N 745
(listwise)
𝑋−𝑋̅
Hence the formula for standardized TC value is, ZTC=
𝜎
where X-actual travel cost value, 𝑋̅- mean of travel cost values , 𝜎- standard deviation of travel
cost values.
Similarly, ZTT is calculated. Then the standardized values of Travel cost and travel time i.e.,
ZTC and ZTT is used as predictors
Independent Model 1
variables Chi-Square df Sig.
ZTC 473.04 4 0.00
ZTT 380.73 4 0.00
INCOME 162.45 16 0.00
Page | 84
We can see that, standardizing the travel cost and travel time values does not change results of
likelihood tests results. In above table it can be seen that the number of road and air-road samples
are less.
Pseudo R-Square
Cox and Snell 0.87
Nagelkerke 0.91
The pseudo R-square tells us how much of the variance in the dependent variable is explained by
the model. Cox and Snell's R2 is based on the log likelihood for the model compared to the log
likelihood for a baseline model. However, with categorical outcomes, it has a theoretical maximum
value of less than 1, even for a "perfect" model. Nagelkerke's R2 is an adjusted version of the Cox
& Snell R-square that adjusts the scale of the statistic to cover the full range from 0 to 1. So the
selected model satisfies both criteria.
Marginal
N Percentage
MODES Air 429 57.6%
Rail 132 17.7%
Road 14 1.9%
Air-Rail 148 19.9%
Air-Road 22 3.0%
INCOME up to 5 lakh 95 12.8%
CLASS 5 to 10 lakh 286 38.4%
10 to 20 lakh 216 29.0%
above 20 lakh 148 19.9%
Valid 745 100.0%
Missing 0
Total 745
In above table it can be seen that the number of road and air-road samples are less.
Page | 85
Deviance and Pearson test for goodness of fit are the statistics test which determine how well the
model fits that data (expected and actual values) and p<0.05 means that there is a significant
difference between the two i.e. the model is not a good fit. Here according Pearson test criteria
model is not good fit while by Deviance test model is good fit. This could be due to low frequencies
in crosstabs.
Table 6.41: Parameter estimates for model 4
Page | 86
is more than 0.05 the predictor has no significant effect. If Exp(B) value for given mode is more
than one, predictors decrease the likelihood of a respondent choosing reference category which is
road. If Exp(B) value for given mode is less than one, predictors increase the likelihood of a
respondent choosing reference category which is road.
Independent
Description B significance Exp(B) Interpretation
Variable
Page | 87
Travellers with If income is above 20 lakhs it has no
[INCOME=4.00] annual income -0.353 0.657 0.703 significant effect on mode choice with road
more than 20 lakhs as reference category
Page | 88
CHAPTER 7: AIR TRAVEL DEMAND FORECAST
For Surat airport, domestic air passenger data from year 2009-10 is available which is not sufficient
to do forecast by ARIMA method. Hence for Surat airport, demand is calculated by simple linear
regression model as follows:
For simple linear regression, the data from year 2009-10 to 2013-14 is used. (since there is
abrupt drop in demand in year 2014-15 that reading is not used in regression)
Table 7.2: Regression model for Surat airport’s domestic air travel demand
Page | 89
Table 7.3: Forecasted demand from 2014-15 to 2020-21 for Surat airport
2014-15 240950
2015-16 285725
2016-17 330500
2018-19 375274
2019-20 420049
2020-21 464823
In above table, it can be seen that by current pattern of growth, Surat airport will have domestic
air passenger demand of around 4.6 lakhs by year 2020 while Mumbai, Pune, Ahmedabad,
Vadodara and Chennai airports will have demand of 2.8 Crore, 49 lakhs, 48 lakhs,11 lakhs and 1.2
Crore respectively.
Average annual growth rate of GDP and population is calculated so as to identify to which
metropolitan regions or districts, Surat district shows similar growth pattern.
Page | 90
Table 7.5: Average annual GDP growth rate from year 2001-02 to 2015-16
MMR 13.84
Pune 13.15
CMA 6.95
Ahmedabad 14.28
Vadodara 13.80
Surat 15.55
Surat district has the highest average annual GDP growth rate but it is close the average annual
growth rate of MMR, Pune district, Ahmedabad district and Vadodara district
Table 7.6: Average annual population growth rate from year 2001 to 2011
Surat 3.586
Vadodara 1.353
Ahmedabad 2.177
Pune 2.688
Mumbai(MMR) 1.803
Chennai(CMA) 2.130
Surat district has highest annual population growth rate but it is closest to average annual growth
rate of Pune district, Ahmedabad district and CMA.
Here Surat’s GDP and Population data is used to forecast domestic air travel demand using
econometric models obtained for other airports.
Page | 91
Table 7.7: Forecast scenarios for Surat airport
From table 7.7, we can see that the econometric model for Mumbai airport cannot be used for
estimating air travel demand for Surat airport since it is giving negative values. The reason for this
might be because MMR is a mega metropolitan and has population as well as GDP value much
greater than Surat.
The estimated potential domestic annual air travel demand values for Surat for current year 2015-
16 is 19.37 lakhs, 1.38 Crore, 11.86 lakhs and 31.54 lakhs by econometric models of Pune,
Page | 92
Chennai, Vadodara, and Ahmedabad airports while by the planning horizon it will be are 55.83
lakhs, 4.22 Crore, 27.41 lakhs and 1.36 Crore respectively.
The model for Chennai which is based on GDP is giving very high estimate of domestic air travel
demand for Surat. This might be because CMA has very low average annual GDP growth rate.
Hence it is not appropriate to use econometric.
The model for Vadodara and Pune which are also based on GDP gives comparatively low estimates
of domestic air travel demand for Surat. From figure 7.1, it can be observed that both Pune and
Vadodara districts have very different growth pattern compare to Surat district. Hence it would not
be appropriate to use these models to forecast Surat district model.
180000
160000
GDP at current rate in Crores
140000
120000
100000
80000 Vadodara
60000 Surat
40000 Pune
20000
0
2004-05
2005-06
2006-07
2007-08
2008-09
2009-10
2010-11
2011-12
2012-13
2013-14
2014-15
2015-16
Year
Figure 7.1: GDP growth pattern of Surat, Pune and Vadodara Districts
The model for Ahmedabad district is based on population. From figure 7.2, it can be observed
that both of these districts show almost similar population growth pattern. Hence it would
appropriate to use this model to forecast domestic air travel demand for Surat.
Page | 93
90.00
80.00
Population in lakhs
70.00
60.00
50.00
40.00
30.00
20.00
10.00
0.00
2009-10
2001-02
2002-03
2003-04
2004-05
2005-06
2006-07
2007-08
2008-09
2010-11
2011-12
2012-13
2013-14
2014-15
2015-16
Year
Ahemdabad Surat
Figure 7.2: Population growth pattern for Surat and Ahmedabad District
Using deduction method based on similarity of pattern in population and GDP growth of other
cities and Surat, the best model has been chosen. Among all cities, Ahmedabad shows nearly
similar pattern of population with some lag. Hence the prediction of annual domestic air passengers
using econometric model developed for Ahmedabad is most suitable for Surat airport.
Page | 94
CHAPTER 8: SUMMARY AND CONCLUSION
8.1 GENERAL
Developing the forecasting model to estimate potential domestic air travel demand is the main
objective of this study. Study on air travel demand pattern of other airports has been done and
comparative method forecast scenarios for Surat airport is developed. For forecasting, Gretl and
E-Views softwares are used. In addition to forecast, to understand the current travel behaviour of
travellers from Surat for long distance trip, mode choice analysis is done using Excel and SPSS.
For mode choice analysis total 453 samples have been collected by questionnaire survey. Out of
these samples, travellers with annual income less than 5 lakhs, 5 to 10 lakhs, 10 to 20 lakhs and
above 20 lakhs were 14%, 40% ,28% 18% respectively. The destination analysis revealed that,
travellers going to Mumbai, Bangalore, Kolkata, Chennai, Ahmedabad, Chandigad, Hyderabad
and Jaipur i.e. metro-cities are more in percentage.
The mode preference analysis revealed that purpose of trip for distance ranges less than 500 km,
500-1000 km and more than 1000 km, the percentage of travellers taking air as first mode are
23.83%,67.29% and 95.09% respectively for business purpose trips. Similarly, for distance ranges
less than 500 km, 500-1000 km and more than 1000 km, the percentage of travellers taking air as
first mode are 11.95%,41.15% and 90.71% respectively for social/leisure purpose trips. In
business/work purpose trips the effect sponsorship for travelling expenditure on mode preference
is analysed. For distance, up to 500 km percentage of travellers choosing rail as first preference is
more for sponsored trips compare to non-sponsored trips. For distance range 500 to 1000 km, more
percentage of travellers are choosing air as first preference for sponsored trips which is 71.08%
while for non-sponsored trips it is 67.36%. For distance above 1000 km, for both sponsored and
Page | 95
non-sponsored trips around 90% travellers giving air mode as first preference. Willingness to pay
analysis revealed that higher income groups are willing to spent more money for given distance
than lower oncome group. From T-test, it is found that the importance of travelling attributes
comfort and reliability remains same for both business and social purpose trips while there is
significant difference for other travel attributes which are travel time, waiting time, travel cost,
frequency and interchanges. For income group above 20 lakhs, travel cost is significantly not as
important as it for other groups.
The econometric models are developed by taking GDP at current rate and population as an
independent variable. These forecasting models are applied for estimating potential air travel
demand for Surat. The estimated domestic annual air travel demand values for Surat by the
planning horizon year 2035-36 by econometric models of Pune, Chennai, Vadodara, and
Ahmedabad airports are 55 lakhs, 4.22 Crore, 27 lakhs and 1.36 Crore respectively. But after
comparing population and GDP growth patterns of Surat with other districts/metropolitan regions
it is found that econometric model developed for Ahmedabad should be applied to Surat since
Surat and Ahmedabad districts show very similar growth patterns for population. For present year
that is 2015-16, by simple regression using current domestic air travel demand data estimated
demand for Surat is around 2.40 lakhs but by the potential demand estimated by using econometric
model for same year is 31.54 lakhs which is much higher than actual demand.
8.4 CONCLUSION
The main purpose of this study is to develop air travel demand model for Surat airport to estimates
its potential domestic air travel demand. For comparison of current growth pattern of Surat airport
and other airports ARIMA models are used. Long term demand is estimated by applying
Page | 96
econometric models developed for other airports to Surat airport on the basis on similarity in GDP
and population growth pattern. Mode choice analysis is done to understand travellers behaviour
for making long distance trip.
1. Form this study, it can be concluded that Surat has much higher potential domestic air travel
demand compare to present demand. For present year, the estimated potential domestic air travel
demand is 31.54 lakhs which almost 14 times more than present demand. It means the latent
domestic air travel demand of Surat airport is 28.69 lakhs.
2. In time series analysis from seasonality indices values found for different airports, it is revealed
that for months January, May, June, November and December there is high domestic air travel
demand compare to other months.
3. From mode choice analysis it is observed that as travelling distance increases more percentage
of travellers giving air as their first choice. It is also found that for business purpose long trips
more percentage of travellers prefer air compare social purpose trips. Willingness to pay varies
among different income classes. Higher income groups are willing to spend more money compare
to lower income class groups for given purpose of travel and travelling distance.
4. Travelling attribute study revealed that the importance of travelling attributes, travel time, travel
cost, waiting time, interchanges and frequency varies as the purpose of journey varies. While for
comfort and reliability attributes, equal importance is given for both social and business purpose
trip.
1. The current survey is carried at Surat airport by interviewing air passengers but since almost
40% of the passengers were making return journeys which were not residents of Surat. To have
better idea of possible destinations in future survey can be conducted at Surat railway stations by
interviewing AC passengers.
Page | 97
2. City pair wise forecast can be done in future so as to estimate potential demand between Surat
and other cities. For that air passenger data from airline agencies can be collected.
3. As per estimated demand the future plans for expansion of Surat Airport can be prepared.
4. The influence area of the airport can be estimated from flight booking data, from which
accordingly the data regarding influencing factors can be collected and better estimation model
can be made.
Page | 98
REFERENCES
Ahmadzade, F., 2010. Model for Forecasting Passenger of Airport. Dhaka,Bangladesh, s.n.
Alam, M. B. & Karim, D. M., 1998. Ait Travel Demand Model for Domestic Air Transportation
in Bangladesh. Journal of Civil Engineering, CE26(1).
Alekseev, K. & Seixas, J. M., 2002. Forecasting the Air Transport Demand for Passengers with
Neural Modelling. Rio de Janeiro, Brazil, IEEE Computer Society.
Algers & Staffan, 1993. Integrated Structure of Long-Distance Travel Behavior Models in
Sweden. Transportation Research Board, Issue 1413, pp. 141-149.
Anon., 2006. Manual on Air Traffic Forecasting. 3rd ed. s.l.:International Civil Aviation
Organization.
Anon., 2011. 12th Five Year Plan for Civil Aviation Sector, New Delhi: Planning
Commission,Government of India.
Anon., 2013. Discussion Paper 01: Aviation Demand Forecasting, s.l.: Airports Commission.
Ba-Fail, A. O., Abed, S. Y. & Jasimuddin, S. M., 2000. The Determinants of Domestic Air
Travel Demand in The Kingdom of Saudi Arabia. Journal of Air Transportation World Wide,
5(2).
Baik, H. et al., 2008. Forecasting Model for Air Taxi,Commercial Airline, and Automobile
Demand in the United States. Transportation Research Record: Journal of the Transportation
Research Board, pp. 9-20.
Page | 99
Cho, H. D., 2013. The Factors that Affect Long Distance Travel Mode Choice Decisions and
their implications for Transportation policy, s.l.: University of Florida.
Drago, P., Ratko, Z. & Zuko, R., 2014. Econometric Model for Forecasting Demand in
Passenger Air Transport. Transport Economics.
Eke, C. N., Okwuosha, K. C., Egwim, K. C. & Nkwazema, . O. A., 2014. Time Series Analysis
of Demand for Domestic Air Travel in Nigeria. The International Institute for Science,
Technology and Education, 4(24), pp. 123-132.
Frank S. Koppelman, V. S., Koppelman, F. S. & Sethi, V., 2005. Incorporating variance and
covariance heterogeneity in the Generalized Nested Logit model: an application to modeling
long distance travel choice behavior. Transportation Research Part B 39, p. 825–853.
Fridstrom, L. & Thune-Larsen, H., 1989. An Econometric Air Travel Demand Model for Entire
Conventional Domestic Network:The Case of Norway. Transport Research Part B, 23B(3), pp.
213-223.
GhobriaI, A. & Kanafani, A., 1995. Quality of Service Model of Intercity Air-Travel Demand.
Journal of Transportation Engineering, March/April, 121(2), pp. 135-140.
Ghobrial, A. & Kanafani, A., 1995. Qualtiy of Service Model of Intercity. Journal of
Transportation Engineering, March/April, 121(2), p. 1716.
Grosche, T., Rothlauf, F. & Armin , H., 2007. Gravity Models for Airline Passenger Volume
Estimation. Journal of Air Transport Management, Volume 13, pp. 175-183.
Horonjeff, R., McKelvey, F. X., Sproule, W. J. & Young, S. B., 2010. Planning and Design of
airports. 5th ed. s.l.:Mc Graw Hill.
Karlaftis, M. G., 2008. Demand Forecasting in Regional Airports:Dynamic Tobit Models with
Garch Errors. Rio De Janero, Sitraer 7, pp. 100-111.
Karlaftis, M. G., Zografos, K. G., Papastavrou, J. D. & Charnes, J. M., 1996. Methodological
Framework for Air-travel. Journal of Transportation Engineering, March/April.122(2).
Page | 100
Kumar, V. & Chandra , S., 1999. AIr transportation planning and design. 1st ed. New Delhi:
Galgotia.
Limtanakool, N., Dijst, M. & Schwanen, T., 2006. The Influence of Socioeconomic
Characteristics, Land Use and Travel Time Considerations on Mode Choice for Medium- and
Longer-Distance trips. Journal of Transport Geography , pp. 327-341.
Mallikarjuna, C. & Raghu Kanth, S. T. G., 2010. Forecasting Air Traffic for North East Indian
Cities. Advances in Adaptive Data Analysis, 2(1), pp. 81-96.
Mathisen, T. A., 2006. The relationship between travel distance and fares, time costs and
generalized costs in passenger transport, s.l.: Trafikdage på Aalborg Universitet.
Moeckel, R., Fussell, R. & Donnelly, R., 2015. Mode choice modeling for long-distance travel.
Transportation Letters: the International Journal of Transportation Research, 7(1), pp. 35-46.
Nepal, K. P., Hibino, N., Kanda, Y. & Morichi, S., 2006. Time Series Analysis of Domestic
Aviation Passenger Travel. Nagoya, Japan, Proceedings of 10th ATRS conference.
Reichert, A. & Holz-Rau, C., 2015. Mode use in long-distance travel. The Journal of Transport
and Landuse, 8(2), pp. 87-105.
Rengaraju, V. & Arasan, V. T., 1992. Modelling for Air Travel Demand. Journal of
Transportation Engineering, May/June, 118(3), pp. 371-380.
Sen, A., 1985. Examining Air Travel Data using Time Series Data. Journal of Transportation
Engineering, March, 111(2), pp. 155-161.
Silva, E., Cruz, R. & Vilelea, M., 2012. Forecasts of Scheduled Passenger Traffic for New
Airport with Limited Data. s.l., International Civil Aviation Organization(ICAO) Information
Paper.
Page | 101
Sivrikaya, O. & Tunç, E., 2013. Demand Forecasting for Domestic Air Transportation in Turkey.
The Open Transportation Journal, Volume 7, pp. 20-26.
Spitz, W. & Golaszewski, R., 2007. Airport Aviation Activity Forecasting. Transport Research
Board of National Academies.
Thomet, M. & Sultan, S., 1979. Traffic Forecasting for the New Riyadh International Airport.
ICAO Manual on Air Traffic Forecasting.
Wadud, Z., 2011. Modeling and Forecasting Passenger Demand for a New Domestic Airport
with. Transportation Research Record, pp. 11-0284 .
Wang, M. & Song, H., 2010. Air Travel Demand Studies: A Review. Journal of China Tourism
Research, 5(1), pp. 29-49.
Page | 102
ANNEXURE
1. QUESTIONNAIRE
Page | 103
Page | 104
2. SAMPLE CALCULATION FOR SEASONLITY INDICES OF
AHEMDABAD AIRPORT
Yt base line Yt/CMA
MA 12 CMA
Year month Ahmedabad St ,It St
MONTHS' average(12)
Page | 107
3. YEARWISE DOMESTIC AIR PASSENGER DATA
Page | 108
4. GDP AT CURRENT RATE DATA
Page | 109
5. AAI PERMISSION LETTER
Page | 110
6. MONTH WISE DOMESTIC AIR PASSENGER DATA
Year Month Ahmedabad Mumbai Pune Vadodara Chennai
2007 January 218568 1416053 147463 41824 587355
2007 February 203734 1310823 139715 38240 526479
2007 March 190312 1341873 142699 37270 535370
2007 April 209815 1448983 135234 38210 617519
2007 May 239425 1555907 148458 44502 669926
2007 June 207038 1420715 136153 41155 624126
2007 July 184897 1393828 150356 36884 588371
2007 August 188061 1436678 141467 37431 616826
2007 September 182695 1357237 123058 37049 568535
2007 October 187184 1429428 130960 40397 571402
2007 November 224141 1650769 151807 47892 595252
2007 December 211504 1626897 161684 47200 624076
2008 January 215977 1599372 152006 48085 593142
2008 February 211083 1498605 66890 42918 577177
2008 March 200138 1462505 143831 40021 583149
2008 April 226478 1558632 158388 40079 614787
2008 May 240344 1639655 165257 44245 636289
2008 June 194473 1328655 153489 41042 560055
2008 July 144953 1167677 135232 43907 494146
2008 August 137480 1182078 126766 33724 616826
2008 September 127097 1048530 114712 31187 453185
2008 October 156326 1175219 131522 34153 476384
2008 November 176164 1187585 141904 35032 446639
2008 December 185393 1239009 160737 37149 508990
2009 January 188691 1309182 154789 36428 527567
2009 February 194348 1288108 154380 35907 478466
2009 March 170582 1192483 146364 31735 472100
2009 April 198112 1282025 158605 35296 472100
2009 May 275753 1521138 185233 43374 555851
2009 June 232080 1379461 175310 36759 539705
2009 July 205370 1308586 174892 35101 540625
2009 August 211394 1389392 155183 37536 546642
2009 September 191613 1301132 155981 35603 514224
2009 October 227590 1564291 186628 45057 554280
2009 November 216817 1486401 192182 45368 542486
2009 December 259744 1720532 220241 52466 639828
2010 January 237385 1560037 205863 47990 616149
2010 February 223828 1445102 198076 43143 551984
Page | 111
2010 March 208071 1414134 201924 43159 569186
2010 April 220768 1560864 216825 43685 605418
2010 May 296382 1846260 235014 53361 680733
2010 June 250157 1630970 228571 46633 627923
2010 July 209051 1487168 219312 43760 598705
2010 August 200357 1500826 219351 42174 600629
2010 September 200877 1457573 210684 42035 592677
2010 October 225680 1623406 234121 48744 651702
2010 November 316680 1813749 239671 54409 659150
2010 December 336517 1885750 249569 60657 756113
2011 January 343045 1794579 236354 58256 706307
2011 February 317168 1701361 228174 54340 646455
2011 March 299862 1682937 235154 50248 678031
2011 April 281022 1653788 236990 45085 664591
2011 May 363014 1938684 266990 62416 764416
2011 June 314124 1735089 261325 57195 741277
2011 July 317563 1687202 262586 47892 729841
2011 August 309697 1676063 264284 46917 699745
2011 September 297712 1580176 242098 44639 672837
2011 October 355476 1748084 254980 54541 696896
2011 November 364509 1854209 296750 63845 710496
2011 December 375828 1901803 302590 67695 786760
2012 January 348585 1869616 295607 68971 774641
2012 February 336373 1748267 283349 63383 707531
2012 March 286249 1651227 261802 47362 668149
2012 April 299055 1751772 272850 59557 355744
2012 May 331905 1825668 293926 69021 785516
2012 June 293626 1647164 272437 61118 701915
2012 July 204685 1476792 258369 53053 660767
2012 August 236081 1507970 254685 47647 650614
2012 September 217259 1395597 231477 44024 342482
2012 October 239592 1538283 236121 47792 643711
2012 November 296974 1811259 274134 60394 676659
2012 December 302847 1928503 295475 63418 757314
2013 January 316404 1862373 285970 62039 734087
2013 February 295439 1751411 278913 55445 678672
2013 March 274671 1780159 280302 52564 723259
2013 April 285345 1780940 285369 52655 346809
2013 May 355236 1992150 320239 63902 770140
2013 June 277074 1689259 291729 52804 677283
2013 July 269969 1703153 291487 51947 677394
Page | 112
2013 August 298612 1915069 312548 52401 732065
2013 September 251520 1656203 272190 48766 357851
2013 October 260511 1724398 273260 51170 653468
2013 November 328030 1889104 292814 57825 669332
2013 December 353426 2007117 326941 62037 764200
2014 January 306521 1866477 284539 67463 714088
2014 February 300910 1802769 268432 65684 657684
2014 March 278832 1854047 275995 59581 679094
2014 April 276721 1891997 293163 59004 689365
2014 May 328798 2155202 347398 70019 809313
2014 June 283578 1939099 320998 62352 757499
2014 July 256291 1767718 305560 56925 764164
2014 August 300551 1995326 339492 57394 829938
2014 September 313978 2030720 336625 53150 841070
2014 October 326282 2124552 335229 60472 781607
2014 November 330730 2093230 341132 56300 795606
2014 December 349123 2343268 371889 63283 861464
2015 January 377504 2311493 365115 62706 857197
2015 February 351352 2254077 352528 58044 780480
2015 March 342802 2298593 358572 53428 824352
2015 April 380196 2418961 401146 65429 843560
2015 May 425985 2567032 427925 83366 913382
2015 June 362289 2268381 398740 70482 808419
2015 July 355171 2329524 422008 68352 834565
2015 August 360113 2392728 424137 71709 862049
2015 September 378121 2372492 415742 72581 840426
2015 October 389008 2419774 424958 70483 849918
2015 November 438515 2589297 433152 84671 851041
2015 December 455776 2770303 453101 91190 706003
Page | 113