Understanding and modelling surface urban heat island phenomenon across major Indian cities
Student name: Ms Tisha Sehdev
Guide: Dr P K Joshi
Year of completion: 2012
Host Organisation: TERI University
Abstract: In this study, we have followed a multi-objective approach to understand and
characterize Surface urban heat island phenomenon across the major Indian
cities- Delhi, Mumbai, Kolkata, Chennai and Hyderabad. In the past, Surface
Gaussian fitting has been greatly employed to quantify and characterize Surface
Urban Heat Islands using the low resolution LST of urban cities worldwide.
However in context to Indian cities with multiple as well as newly emerging
growth centres, it was important to construct a methodology which enables to
characterize these local scale Surface urban heat islands corresponding to the
clusters of newly urbanised areas in and around a city. In order to achieve so, an
attempt was made to utilize the algorithm of Expectation Maximization for
parameter estimation of the Gaussian Mixture Models in terms of the
characteristics of the local scale Surface urban heat islands-Gaussian height
(average rise in temperature), variance in longitudinal (x-direction) and
latitudinal (y-direction) extents.
In the First stage, MODIS LST product was utilized to investigate the seasonal
variation in LST for all the cities. This helped in identification of the months or
seasons where there is maximum variation on LST and thus appropriate for
further analysis and computation of SUHI Intensity. Whereas in the Second
stage, difference in Land surface temperature of the cities from Landsat Thermal
Imagery in the past decade (2001-2010) was used to compute SUHI
characteristics of newly emerging urban growth centres using the algorithms for
Gaussian mixture models and Expectation-Maximization.
Keywords: Surface Urban Heat Islands (SUHI), Gaussian Mixture Model,
Expectation Maximization, LST, MODIS
