Landscape and Urban Planning 190 (2019) 103586

Contents lists available at ScienceDirect

Landscape and Urban Planning

journal homepage: www.elsevier.com/locate/landurbplan

Research Paper

Assessing urban drivers of canopy layer urban heat island: A numerical T

modeling approach
⁎
Rajashree Kotharkar , Anurag Bagade, Aparna Ramesh
Department of Architecture and Planning, Visvesvaraya National Institute of Technology Nagpur, Maharashtra 440010, India

A R T I C LE I N FO A B S T R A C T

Keywords: The physical and geometrical properties of urban regions are responsible for influencing canopy layer air
Urban heat island temperature leading to urban heat island effect (UHI). Local climate zone (LCZ) land cover classification scheme
Local climate zone has been widely adopted to examine UHI and inter urban heat island (IUHI). In tropical cities with hetero-
Multiple regression geneous mix of urban surface cover, morphology, thermal and radiative properties pose the question–which built
Numerical modeling
properties have a higher influence on UHI? This study evaluates UHI response in tropical city Nagpur (India)
Mitigation
using LCZ and city specific built parameters. It uses mobile survey to collect nocturnal air-temperature data
during two consecutive summers (2016 and 2017). The study identifies eleven built environment parameters
from literature survey. It adopts a unique approach of all possible regression technique with series of checks and
model validation (coefficient signage, variance inflation factor, p-value & t-statistics, R square, F value and k-fold
cross validation). The result shows distance from central business district (DI), surface albedo (AL), aspect ratio
(AR) and vegetation density ratio (VDR) as major predictors explaining UHI response. Every 500 m increase in DI
reduces inter urban heat island (IUHI) by 0.13 °C. Increasing AL by 0.01 decreases UHI by 0.18 °C whereas
increasing VDR by 0.10 yields 0.17 °C reduction in IUHI. 10% increase in AR suggests IUHI increase by 0.17 °C.
This study contributes in exploring mitigation strategies for complex built environment. It enables urban
planners, designers and policy makers to approach urban intervention with scientific and sustainable approach.

1. Introduction heatwaves and its effects are more profound in the sub-surface layer. Its
impacts are also well associated with human health, thermal comfort,
Some of the major challenges of today’s economies include rapid environmental quality and energy efficiency (Phelan et al., 2015;
pace of urbanization and dealing with the potential impacts of climate Santamouris, 2014). Thus, making the compound effect of climate
change. Today, more than half of the world population (i.e., 4.2 Billion change and urbanization even more concerning. Effective urban plan-
people) lives in urban areas (United Nations, 2018). Cities of devel- ning and designing strategies provide solution to some of these chal-
oping countries, where 70% of urban population lives, are likely to lenges. However, it requires quantitative data to understand the role of
borne the severe impacts for changing climate (Field et al., 2012). Asia urban parameters affecting UHI magnitude to formulate required mi-
already observes an alarming rise in number of warm days and heat- tigation policy.
wave frequency since the past century (Hijioka et al., 2014). Urban
areas affect the overlying climate and induce variation in cloud cover, 1.1. UHI reporting
precipitation, temperature, wind speed, and wind direction (Collier,
2006; Han, Baik, & Lee, 2014). Considering the need of hour, urban UHI studies are advancing towards developing universal, standar-
climate studies are now focusing over the underlying principles that dized and robust framework of evaluation. Local climate zone (LCZ) is a
affect and alter the radiative, aerodynamic, thermal and moisture universal land cover classification system defined as regions spanning
properties of atmosphere. Urban heat island (UHI), an elevated tem- hundreds of meters in horizontal scale with uniform surface cover,
perature rise phenomenon observed in urban areas as against its rural geometry, material, human activity and having characteristic near-
surroundings, is one of the most documented urban climate study (Oke, surface air temperature patterns (Stewart & Oke, 2012). LCZs are
1981). UHI exacerbate the detrimental impact of warm days and grouped into 'built' (classes 1–10) and 'land-cover' (classes A–G)

⁎
Corresponding author at: Department of Architecture and Planning, Visvesvaraya National Institute of Technology, VNIT Campus, South Ambazari Road, Nagpur,
Maharashtra 440010, India.
E-mail address: rskotharkar@arc.vnit.ac.in (R. Kotharkar).

https://doi.org/10.1016/j.landurbplan.2019.05.017
Received 27 January 2019; Received in revised form 16 May 2019; Accepted 24 May 2019
Available online 26 June 2019
0169-2046/ © 2019 Elsevier B.V. All rights reserved.
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

categories, each with distinct yet measurable physical properties 1.2.2. Building envelope
(Stewart & Oke, 2012). It provided an invaluable opportunity for re- Building envelope indices affect the wind speed, wind direction,
searchers across the globe to explicitly define the UHI intensity among shortwave and long wave radiation balances. These indices constitute
different neighborhoods having discrete built environments. The ex- the urban structure and covers urban parameters such as building
tensive observational and numerical simulations from diverse cities height (Bottyan & Unger, 2003), sky view factor (Unger, 2004), aspect
across latitudes, have given credence to the LCZ classification system as ratio, total height-to-floor area ratio (Giridharan, Lau, & Ganesan,
a valid structural framework to evaluate UHI magnitude. Analysis of 2005), and building volume density (Petralli et al., 2014). Sky view
historically observed air temperatures in the ten different LCZs of factor (SVF) is a geometric factor that describes the ratio of the amount
Glasgow, UK revealed that the urban features added to the increase in of sky hemisphere visible from ground level to that of an unobstructed
regional temperatures despite the reduction in city population hemisphere. It comprehensively characterizes the radiation geometry of
(Emmanuel & Krüger, 2012). After classifying the warm-humid city of the site of near surface air temperature measurement (Stewart & Oke,
Colombo into multiple LCZs on the basis of observational data, it was 2012; Unger, 2004). Regions with low SVF traps the long wave radia-
found that significant urban warming was seen at sites with high-rise tion leading to the loss of surface cooling rate and increased tempera-
developments (Perera, Emmanuel, & Mahanama, 2012). In another ture of urban areas during night (Yuan & Chen, 2011). Aspect ratio (AR)
study, the Greater Nancy area in France was divided into 13 LCZs to is calculated as mean height of building-to-width ratio of street can-
investigate the spatial distribution of screen-height air temperature via yons. It examines how densely buildings are spaced with respect to their
mobile measurements and it was observed that the differences in heights (Stewart & Oke, 2012; Unger, 2004). With rise in AR and re-
average nocturnal air temperature varied by up to 4 °C for dissimilar duction in SVF, long wave radiant loss is constrained, leading to a
LCZs (Leconte, Bouyer, Claverie, & Petrissans, 2015). To address the marked increase in canopy-layer UHI intensity (Erell, Pearlmutter, &
thermo-methodological concerns in quantifying the UHI intensity in Williamson, 2012; Stewart & Oke, 2012; Unger, 2004). Fabric density
Hong Kong SAR, LCZs were used to classify 17 weather stations from ration (FDR), ratio of vertical surface area to total plan area, increases
the Hong Kong observatory’s fixed-point meteorological observation the amount of solar radiation absorbed, retained and converted to
network (Siu & Hart, 2013). The study indicated that ‘urban’ re- sensible heat (Kolokotroni & Giridharan, 2008).
presentative stations were used in previous studies instead of ‘rural’.
Therefore, UHI intensities previously calculated for Hong Kong SAR 1.2.3. Thermal and radiative indices
may have been underestimated (Siu & Hart, 2013). Twelve mobile Albedo, thermal mass and thermal conductivity constitute some of
surveys were conducted in the city of Kochi in southern India to the thermal and radiative indices of urban surfaces. Albedo is defined as
quantify UHI intensity, thermal gradient and cooling rates in different the ratio of solar radiation reflected to the amount of solar radiation
LCZs which further corroborated the importance of classification absorbed by a surface. It plays an important role in governing surface
system in studying UHI (Thomas, Sherin, Ansar, & Zachariah, 2014). energy balance. Temperature variation within built environment also
depends on the heat absorption and storage capacity represented by
thermal mass. Neighborhoods with high thermal mass follow a slow
1.2. UHI drivers rate of heat dissipation which further contributes in reducing its diurnal
temperature.
There are various uncontrolled and controlled variables that con-
tribute to the development of UHI. Uncontrolled variables include cli- 1.2.4. Anthropogenic factors
mate based parameters and controlled variables include built and an- Anthropogenic heat is measured as the heat flux density with
thropogenic factors. LCZ classification system considers built human activity as a main source. Heat produced due to transportation,
parameters that influence screen-height temperature such as spacing of space heating, industrial manufacturing and human metabolism fuels
trees and buildings, height of roughness and surface cover (Stewart & the temperature variations in urban areas. As quantifying the exact heat
Oke, 2012). For detailed investigation, these urban parameters could be flux is difficult, many researcher use proxy indicators to represent an-
further grouped into four categories, namely, surface indices, building thropogenic factors. This include distance to main streets (Alcoforado &
envelope, thermal and radiative indices, and anthropogenic factors. Andrade, 2006), population density (Kotharkar & Surawar, 2016) and
distance to urban area (Ivajnšič, Kaligarič, & Žiberna, 2014).

1.2.1. Surface indices 1.3. Modeling approach

Land cover fractions such as built surface fraction (BSF), vegetation
density ratio (VDR), pervious surface fraction (PSF) and impervious With the advancement in computational capacity, numerical mod-
surface fraction (ISF) and are some of the widely used surface indices. eling method has been widely adopted by various researchers to assess
Many remote sensing studies explored the land surface composition and UHI relationship with climatic, built cover and anthropogenic factors.
its implication over surface heat island (Deng & Wu, 2013; Xiao et al., Numerical modeling using regression technique such as linear, lasso,
2007; Zhang, Odeh, & Han, 2009). The presence of tree canopy, par- ridge, and geographic ridge regression have shown merits in UHI and
ticularly at heights above 1 m, has been proved to reduce canopy-layer built parameters evaluation (Ivajnšič et al., 2014; Szymanowski &
UHI intensities by dissipating the received solar radiation as latent heat Kryza, 2009; Unger et al., 2010). Mesoscale meteorological models such
energy (Bottyan & Unger, 2003; Giridharan, Lau, Ganesan, & Givoni, as Weather Research and Forecasting (WRF) combined with surface and
2007; Oke, 2006). Vegetation density ratio (VDR) was examined to energy schemes such as Noah land-surface-hydrology model (LSM),
represent the variation in deciduous vegetation cover in summer urban canopy model (single layer, multi layer and advance) and total
(Kolokotroni & Giridharan, 2008). Study shows that during stable at- energy balance schemes, have gained momentum in evaluating urban
mospheric conditions, accompanied by the absence of a considerable climate (Ek et al., 2003; Kusaka, Kondo, Kikegawa, & Kimura, 2001;
anthropogenic heat source, maximum UHI intensity can be explained Martilli, Clappier, & Rotach, 2002; Masson, 2000; Oleson, Bonan,
by the level of 'openness' of the urban structure. Other surface para- Feddema, & Vertenstein, 2008; Wang, Bou‐Zeid, & Smith, 2013; Yang
meters exploring canopy heat island includes built and water surface et al., 2015). These surface models present a multiscale, multiphysics
ratio (Bottyan & Unger, 2003), tree canopy surface fraction (Coseo & approach to evaluate heat islands (Powers et al., 2017). They consider
Larsen, 2014), lawn cover ratio, tree cover ratio, green cover ratio, various parameters including urban geometry, building shadow, re-
street cover ratio, and building cover ratio (Petralli, Massetti, Brandani, flection of radiation, hydrological and thermal properties of all land
& Orlandini, 2014). cover as well as heat fluxes from roof, wall and road (Kusaka, 2001,

2
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Table 1
Previous published studies of urban heat islands in Nagpur.
Authors Method Classification of zones Results/UHI intensity

Katpatal et al. (2008) Calculation of land surface temperature (LST) Land use land cover (LU/LC) classification into Most densely populated urban areas, paved roads,
Landsat 5 TM image, 11:00 am, December 21, water-body, thick vegetation, sparse vegetation, barren land/wasteland or stony waste showed
2001. Validated by field observation of air open spaces, dense urban area, sparse urban area, higher temperatures compared to dense vegetation
temperature slum, pavement, agriculture land, waste land and water bodies (Katpatal et al., 2008)
Khandar and Garg Calculation of LST from Landsat LU/LC classification into open space, spare urban Open space and waste lands showed highest LST.
(2014) ETM + image, 14 December 2006. Validated area, water body, dense urban area, thick Urban areas showed higher temperature compared
by comparing LST to mean near surface vegetation, sparse vegetation, waste land to semi-urban and the rural areas (Khandar & Garg,
temperature from 2006 2014)
Agarwal, Sharma, Calculation of LST from Landsat TM and LU/LC classification into built-up, vegetation, Established strong linear relationship model
and Taxak ETM + images water body, bare soil, shrubs and paved surfaces between the LST and normalized difference
(2014) vegetation index. (Agarwal et al., 2014)
Kotharkar and Nocturnal mobile traverse survey measuring LU/LC categories based on type and percentage of Observed canopy layer UHI intensity 2.14 °C in the
Surawar (2016) temperature and relative humidity at 96 open land, built-up, vegetation and water summer and 2.36 °C in the winter (Kotharkar &
points during summer and winter 2012–14 Surawar, 2016)
Kotharkar and Station and nocturnal mobile traverse survey LCZ Classification and evaluation of heat island UHI of 4.09 °C was observed between LCZ 73 and
Bagade (2018a, measuring temperature using LCZ magnitude during winter season LCZ 9. LCZ 3 stands out as a critical LCZ for Nagpur
2018b) classification (Kotharkar & Bagade, 2018a, 2018b)

Martilli et al., 2002). These models are able to simulate UHI considering 1.5. Study area
the spatial and temporal dynamics (Masson, 2000). Depending on the
database, extensive data on Land-Use Land Cover (LULC) determines 1.5.1. Understanding city
the model’s prediction accuracy. Several assumptions are required to be The city of Nagpur, Maharashtra covers 225.08 square km of geo-
considered to setup planetary boundary layer condition (Mirzaei & graphical area on the Deccan Plateau in India (CRISIL, 2015). The city
Haghighat, 2010). Studying UHI having heterogeneous built environ- experiences extreme summers from March to June, tropical monsoon
ment, especially in South Asian context, the WRF simulations still rains from June to August and dry winters from November to January
present many challenges at local scale. To gain insight on urban para- (Nagpur Municipal Corporation, 2014; Nandankar, Dewangan, &
meters at local scale, empirical studies are found to be sufficient Surpam, 2011). Classified as tropical-savannah climate under the
(Masson, 2000). Koppën climate classification system, Nagpur records highest tem-
perature upto 48 °C in summer and lowest temperature upto 12 °C in
winter respectively (Nagpur Municipal Corporation, 2014; Nandankar
1.4. Research objective et al., 2011; Peel, Finlayson, & McMahon, 2007).
Nagpur city could be interpreted on the basis of three zones – city’s
LCZ–UHI study lays foundation for identifying UHI critical zones in commercial business district (CBD) namely ‘core’ region, the peripheral
the urban areas. However, absolute UHI magnitude could not be at- sparsely built ‘fringe’ zone and the in-between sandwiched region
tributed to any particular LCZ class as its properties may differ within ‘inner-city’. Old and congested neighborhoods are found closer to the
the standard range, thus, varying the UHI. Additionally, studies suggest CBD. These regions feature high population density and anthropogenic
the possibilities of finding different intra-urban UHI within same LCZs activity. Almost half the developed land falls under residential land-use
situated at different location in a city (Siu & Hart, 2013; Stewart, 2011; with the remaining land occupied mainly by public institutions, com-
Thomas et al., 2014). During summer night-time, these variations are mercial enterprises and industries (CRISIL, 2015). Slums occupy more
often found stronger (Stewart, 2011; Thomas et al., 2014) due to micro- than 17 square kilometer of horizontal surface area accounting for al-
scale variation in built surface properties, building roughness, material most 10% of the urban area (CRISIL, 2015). While small parcels of the
properties and anthropogenic factors. Urban regions have diverse city have extensive tree cover, < 2% of the urban land is officially
physical, structural, fabric and metabolic properties which further classified as parks and gardens (CRISIL, 2015). The city’s population is
varies with cultural, regional and economic disparities. While the in- 2.4 million and is projected to be double by the year 2041 (Census,
tangible components are difficult to examine, evaluation of physical 2011; CRISIL, 2015). Encumbered by government policies of low floor
and geometrical drivers of UHI would assist in proposing effective mi- space index, the city is compelled to resort to urban sprawl and expand
tigation solutions. rapidly along its southern fringe (CRISIL, 2015).
Development of South Asian cites is impeded with the adverse im-
pact of climate change. Most of the UHI studies conducted in South 1.5.2. Case specific UHI literature
Asian region merely quantify the urban heat island. There are very few Previous research literature indicate variation in UHI with changing
studies that examine the effects of urban morphology and overlying urban factors (refer Table 1) such as population density and built to
climate. There is wide research gap on the statistical and empirical open land cover ratios. An increasing trend of UHI could be observed on
model studies (Kotharkar & Bagade, 2018a, 2018b) in these regions. assessing timeline of these studies. However, as urban-rural classes vary
This study contributes to the research literature of this geographic re- in these studies, collective interpretation of these studies are difficult to
gion. collate.
The primary objective of this paper is to examine and prioritize the A meticulous attempt to classify the city on the basis of LCZs and
built controllers of urban heat island. It evaluates canopy-layer urban evaluate UHI was carried out by Kotharkar and Bagade (2018a, 2018b)
heat island intensities on a typical summer night using all possible (refer Fig. 2). At the resolution of 500 m, the city's heterogeneous mix of
multiple regression technique. It also attempts to explore hierarchy of built form was reflected in the juxtaposition of dissimilar classes and
built parameters affecting UHI magnitude. The model would benefit subclasses across the LCZ map prepared as a result of their study. The
urban local bodies in effective planning and resource allocation for concentration of 'built' subclasses at the city's center was due to the
mitigation. recurrence of dissimilar built form (Kotharkar & Bagade, 2018a,
2018b). The surrounding inner-city areas were typically compact or
open low-rise construction while the peri-urban fringe exhibited

3
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 1. Research design for UHI modeling.

characteristics of the 'sparsely built' class. Compact low-rise (LCZ 3) Measurement of near-surface air temperature in the street canyons
regions are prevalent in the city, followed by sparsely built (LCZ 9) and through automobile traverse surveys has been validated as a funda-
open low-rise (LCZ 6) forms. LCZs 3, 6 and 9 along with their subclasses mental method to study the spatio-temporal behavior of canopy-layer
account for approximately 80% of the built form in the city of Nagpur UHI in cities with diverse urban forms (Roth, 2012; Runnalls & Oke,
(Kotharkar & Bagade, 2018a, 2018b). Slums in the city are non-uni- 2000).
formly distributed and show considerable variation in the types of Meteorological data collection follows the world meteorological
construction materials used. As a result, only small regions with pre- organizations (WMO) guideline and methodology described by
dominantly lightweight construction satisfies the LCZ 7 criteria. LCZ Kotharkar and Bagade (2018a, 2018b). Traverse and station survey
subclassification (nesting of one LCZ into other) was also observed as regimes were conducted on calm and clear nights (between 22:00 and
the inherent feature of the city which further adds to the complexity of 00:00 h Indian standard time) on 21, 22 and 25th April in 2016. They
urban heat island evaluation. were repeated for the subsequent year on 30, 31st March and 1st of April
2017 (refer Fig. 2). Five air temperature recording stations were placed
2. Research design in LCZs having major coverage area in the city (LCZ 3, LCZ 32, LCZ 9,
LCZ 3F and LCZ 65). The experimental setup includes HOBO U-23 Pro
The study explores the urban form and fabric of Nagpur city and V2 temperature and humidity (RH) sensor, placed inside a radiation
collects air temperature data from major LCZ. Based on city char- shield, positioned in the representative LCZ at a height of 1.5 m–2 m
acteristics and UHI literature, urban parameters are deduced for mod- from the ground. The temperature sensor has an accuracy of ± 0.21 °C
eling nocturnal UHI. The study adopts all possible regression modeling from 0° to 50 °C and resolution of 0.02 °C at 25 °C whereas humidity
approach evaluates urban parameters using a three stage methodology sensor has an accuracy of ± 2.5% from 10% to 90% RH (typical) to a
(Fig. 1). First stage includes identification and filtering of urban para- maximum of ± 3.5%. Wind speed and cloud cover data retrieved from
meters through literature study and establish most influencing UHI regional meteorological department, a fixed weather station near the
predictors through quantitative assessment of significant regression airport in the south-west of the city. Atmospheric stability on survey
models. Second stage deals with the selection of best model predicting days was examined using modified Pasquill–Gifford–Turner (PGT)
nocturnal UHI. The final stage explore implication of urban parameters classification system (Krüger & Emmanuel, 2013; Mohan & Siddiqui,
in different existing LCZ scenarios. The study concludes with the dis- 1998). Each survey regime was observed to lie within E-G range of PGT
cussion on most influencing urban parameters and viable mitigation classes indicating calm and clear atmospheric conditions.
strategies applicable for the city. Traverse surveys routes were planned to collect air temperature
data from core, inner city and fringe area as shown in Fig. 2 (route
2.1. Monitoring network and data collection codes: MHL, VNIT and DBA respectively). Each route was designed in a
loop for obtaining continuous air temperature data. HOBO U-23 V2
Diurnal variations in the formation of canopy-layer UHI have been sensors and Holux GPS logger instruments were placed inside PVC tube
attributed to difference in nocturnal cooling rates (NCR) between and mounted on the top of each car. With controlled speed and three
densely built city-centers and non-built up peripheries (Roth, 2012). continuous traverse in a loop, air temperature, humidity data and

4
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 2. LCZ map of Nagpur showing traverse routes and station points modified from Kotharkar and Bagade (2018a, 2018b).

location details were continuously logged at an interval of 1 second. Air UHI = TurbanLCZ − TruralLCZ (1)
temperature at 11 pm was calculated for of eleven sites representative
IUHI = TurbanLCZ − TurbanLCZ (2)
of three standard built classes (LCZs 3, 7 and 9), six built subclasses
(LCZs 32, 37, 3F, 56, 65, and 73) and one land cover class (LCZ F). These where UHI is urban heat island intensity measured between LCZ at
LCZs were selected based on spatial coverage and location in the city. It urban and rural reference which is LCZ F for this case. LCZ F was found
provides an opportunity to examine the thermal performance of major to have lowest nocturnal air temperature throughout the survey. The
LCZ in city and associated anthropogenic activities based on LCZs lo- maximum IUHI and UHI was found to be 3.53 °C and 4.05 °C respec-
cation. tively. After LCZ 73, LCZ 3 (compact low-rise) and their subtypes
Average temperature range of 31.1–34.9 °C was observed during the showed higher UHI and IUHI intensity.
traverse survey. Air temperature was found to be highest in LCZ 73
(Lightweight low-rise nesting compact low-rise) while minimum air 3. Methodology for empirical model
temperature (TLCZ min) was recorded in LCZ F, located in the fringe
areas of the city, during each survey regime. 3.1. Selecting urban parameters
ANOVA test result supported the selection of LCZ sites as it ex-
plained the variations in air temperature is a function of LCZ char- Major LCZ defining built parameters such as pervious surface frac-
acteristics. Since variations were found in the UHI intensity of same tion (PSF), impervious surface fraction (ISF), built surface fraction
classes at different locations in the city, inter-urban heat island was (BSF), sky view factor (SVF), aspect ratio and albedo (AL) were selected
investigated. which explained the surface cover properties and building envelope
As canopy-layer air temperature trend has been found to be uniform properties. Additional parameters were considered based on previous
after approximately four hours past sunset, air temperature at 11:00 pm UHI study and existing urban fabric. This includes vegetation density
(Indian standard time) was interpolated from traverse survey mea- ratio (VDR), fabric density ratio (FDR), thermal mass (TM), population
surements (Chow & Roth, 2006; Roth, 2012). Urban heat island is density (PD) and distance from the central commercial district (DI). The
computed using the formula: coefficient signage (polarity) was estimated from the empirical models
described in previous UHI research literature.

5
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

3.2. Regression model were studied for each LCZs (Fig. 4). Densely compact urban areas were
observed to have small percentage of green cover. Within built LCZs,
Present study uses all possible regression technique using ‘lmtest’ VDR variation was observed between 0.10 and 0.29. Open type LCZs
package in ‘R’-statistical software (Hothorn et al., 2015) to develop UHI featured higher VDR ranges and sparsely built LCZ and LCZ F observed
& IUHI model. This method overcomes the drawbacks of stepwise re- the lowest VDR values. Morphological indicators, represented in
gression technique as the latter may eliminate some useful parameters Fig. 4b, represents FDR, SR and SVF wherein maximum range of SVF is
(Harrell, 2015). The standard framework of multiple linear regressions 1. FDR was observed highest in compact and lightweight LCZs and their
is as follows: subclasses. Although diverse compositions of thermally distinct natural
n and synthetic materials are observed across LCZs, the range of Albedo
Y = β0 + ∑ βn Xn (0.20–0.24) observed for different LCZs was small (Fig. 4c). The small
i (3) range of values received for the parameters is expected in developing
cities in the tropics (Chow & Roth, 2006). A wide range of TM was
where ‘Y’ is the response variable UHI and IUHI, ‘X’ represents urban
observed in old heritage quarters of Mahal (located LCZ 3) and light-
predictors; ‘β’ is coefficient values, ‘β0’ is the intercept and ‘n’ is number
weight structures along the fringe (LCZ 9). PD, represented in Fig. 3d,
of variables.
was found highest for standard 'built' LCZ classes 3, 7 and subclass 73
Each predictor was tested for multicolinearity using Pearson cor-
located close to the city's CBD. Highest DI was observed for built LCZ 9
relation technique. To validate the model, k-fold (5:1) test was applied.
and land cover LCZ F as they lies in the fringes of the city. LCZ F soil had
Each k fold is a set of response and predictors data obtained from in-
no anthropogenic, building envelope and thermal properties (except
dividual nocturnal traverse. Out of six folds of data, five folds were used
SVF and albedo).
to train the model while the remaining fold was used for validation.
Correlation matrix between built parameters and UHI response was
Only significant models (p < 0.05) passing statistical checks were
examined for multicolinearity issue (refer Fig. 5). It was anticipated
shortlisted for further examination. These checks verifies whether
that surface fractions would present co-linearity issue as they are pro-
model’s variation inflation factor (VIF) being less than 5 (indicating no
portional to each other. Built surface fraction established high corre-
multicolinearity), adjusted R-square value is greater than 0.70 and F
lation (R square more than 0.90) with pervious surface fraction, aspect
values above critical. Urban parameter’s priority was set by ranking the
ratio and fabric density ratio. Among building envelope indicators, AR
predictors based on the number of times it appeared in significant
and FDR showed high correlation. Pervious surface ratios are found
models. Predictor’s polarity, corresponding to previous UHI studies, and
more in LCZs with more playgrounds and open spaces. Average sky
significance was studied (as mentioned in Table 2).
view factor computed in these LCZs are bound to be higher than
Final model was selected based on performance in statistical checks
compact type LCZs. Consequently, a high correlation value is observed
and minimum standard residual error occurred in k-fold validation. The
between PSF and SVF. In Nagpur, LCZs situated near old parts of the
urban parameters in final UHI and IUHI models were evaluated in
city often bears high thermal mass due to old construction practices.
different LCZs using partial plot method wherein effect of single pre-
These areas are located near the commercial business district of the
dictor is studied while controlling all other parameters in the model.
city. Built form in rest of the city have similar thermal mass due to
Hierarchy of urban parameters is proposed and viable options of urban
common construction practice and material composition. As a result, a
interventions are examined based on the overall study.
high correlation is also observed in distance from the CBD parameter
and thermal mass.
4. Results

4.1. Urban parameters 4.2. Assessing UHI models

For understanding built variation amongst LCZs, urban parameters Each set of regression run generated 2047 models amounting a total

Table 2
Urban parameter measurement and signage based on reference station.
Sr No. Parameter Signage Method of Measurement

1 Building surface fraction (BSF) + Footprint of buildings were mapped from Google Earth Pro satellite imageries to calculate area; this area was divided by the
total sample area of the LCZ
2 Impervious surface fraction + Footprint of impervious surfaces (paved, rock) were mapped from relevant Google Earth satellite data to calculate area; this
(ISF) area was divided by the total sample area of the LCZ
3 Pervious surface fraction (PSF) − Footprint of pervious surfaces (bare soil, vegetation, water) were mapped from relevant Google Earth satellite data to
calculate area; this area was divided by the total sample area of the LCZ
4 Vegetation density ratio (VDR) − Footprint of vegetation cover were mapped from relevant Google Earth satellite data to calculate area; this area was divided
by the total sample area of the LCZ
5 Fabric density ratio (FDR) + Perimeter of the building footprint multiplied by the average height of buildings in a homogenous LCZ resulted in the vertical
surface area of the built mass; this value was divided by the total plan area. Footprints of buildings were mapped from the
relevant Google Earth satellite images
6 Aspect ratio (AR) + Weighted average road width was calculated for each LCZ. Average height of buildings in homogenous LCZ were verified by
field survey and their mean was calculated. Footprints of buildings and roads were mapped from the relevant Google Earth
satellite images
7 Sky view factor (SVF) − Average SVF was calculated using building and tree data input in RayMan and Skyhelios software (Matzarakis & Rutz, 2007)
8 Surface albedo (AL) − Weighted average of surface albedo of built, pervious and impervious material components in the horizontal surface plane of
a homogenous LCZ area. Footprints of various components were mapped from the relevant Google Earth satellite images
9 Thermal mass (TM) in KJ/kgK + Weighted average specific heat capacity was calculated for all material components of a typical building section in a
homogenous LCZ; this value was multiplied by the perimeter of the building footprint, average thickness of wall section and
average building height in the chosen sample area of the LCZ. Footprints of buildings were mapped from the relevant Google
Earth satellite images
10 Population density (PD) + Municipal ward-wise population density data acquired from Nagpur Municipal Corporation and noted as per the ward-wise
location of the chosen sample area of each LCZ
11 Distance from CBD (DI) − Distance from the sample LCZ area to the primary central business district as measured from Google Earth satellite images

6
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 3. Nocturnal air-temperature observed through traverse.

Fig. 4. LCZ parameters including (a) Surface cover; (b) Morphological indicators; (c) Thermal and radiative; (d) Anthropogenic factors.

of 24,564 number of equations explaining UHI and IUHI magnitude. statistics of coefficients and root mean square error (RSME) found after
From R analysis platform, summary of each regression set was imported k-fold validation.
to Excel spreadsheet where the data was structured for filtering. Initial Eight models explaining response variables (seven IUHI and one
filters (p value, coefficient signage, F factor, Adjusted R square and VIF UHI model) were observed to pass all statistical checks. Most significant
values) reduced the total equation to 1548 and 1647 model for IUHI and relevant models with top five predictors are shown in Table 3.
and UHI respectively. These equations were considered significant for Predictors with more than mean recurrence value (310) were identified
identifying the most influencing urban predictors of UHIs. Predictors and ranked accordingly. Based on ranking, five IUHI explanatory
with more than mean recurrence value (357 and 252 for IUHI and UHI variables were observed. These consists of DI (p = 0, t value ranging
respectively) were identified and ranked accordingly (refer Fig. 6). Top between −2.59 and −6.96), AL (p ranging between 0.05 and 0, t value
five predictors for IUHI and UHI, based on recurrence ranking, were ranging between −2.29 and −2.97), AR (p = 0, t value ranging be-
found to be DI, AL, AR, VDR and PD. Models with combination of these tween 3.67 and 4.26) and VDR (p ranging between 0.05 and 0, t value
variables were scrutinized further using statistical tests based on t- ranging between −2.62 and −3.576). UHI model was explained by

7
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 5. Correlation matrix for response and predictors.

Table 3
Selected IUHI and UHI models.
Model IUHI IUHI IUHI UHI UHI UHI
Sr 186 513 996 186 513 990
Intercept 1.862 5.795 5.874 2.256 3.558 4.127
(***) (**) (**) (***) (***) (**)
DI (in 10−4) −2.59 −2.62 −2.54 −2.86 −2.69 −2.60
(***) (***) (***) (***) (***) (***)
AR 1.202 1.577 1.531 1.088 1.501 1.449
(**) (***) (**) (***) (***) (***)
VDR −1.398 −3.447 −3.350 −1.693 −7.356 −8.787
(**) (*) (*) (*) (*) (*)
AL −17.950 −18.570 −2.617 −2.689
(*) (*) () ()
PD (+) 2.21E-04
()
SVF (−) −0.422
Fig. 6. Quantitative assessment of significant models. ()

F 78.36 53.01 41.50 130.50 82.72 79.88

VDR (t value −2.1), AR (t value 4.55) and DI (t value −7.68). Final R2 0.84, 0.82 0.82 0.885 0.894 0.891
IUHI model was selected based on root mean square error and k fold Adjusted R2 0.83 0.80 0.80 0.878 0.883 0.880
p 0.00 0.00 0.00 0.00 0.00 0.00
validation R square (0.96) (refer Table 3) having 95% confidence in-
terval ( ± 1.4 °C) as shown in Fig. 7. K-fold 0.92 0.96 0.96 0.90 0.92 0.91
Based on the geographic coverage area and urban setting, four im- Validation
R2
portant LCZs (LCZ 3, 6, 9 and F) were selected to evaluate the perfor- RSME 0.41 0.13 0.36 1.02 1.01 1.02
mance of individual indicators using partial effect plots in Fig. 8a–d.
These plots explains the UHI trend for the case by varying one of the (***) p = 0.00; (**) p < 0.01; (*) 0.01 < p < 0.05; (.) 0.05 < p < 0.1; ()
predictor’s value wherein the other predictors are kept constant. The x- p > 0.1. (Bold) Best significant model.
axis of the graphs represents the predictors from the model whereas y-
axis represents UHI magnitude. (Fig. 7d). Although the magnitude of parametric change is different,
For both the models, partial plot shows that every 500 m increase in similar IUHI-UHI trend is seen over different LCZs.
DI would reduce the nocturnal IUHI and IUHI by 0.13 °C and 0.14 °C A comparative assessment of the two models (IUHI and UHI) reveals
respectively (Fig. 7a). Increasing AL by 0.05 would decrease IUHI by the importance of selecting reference station. An urban reference sta-
0.9 °C (Fig. 7b) whereas increasing VDR by 10% would result in gradual tion having minor temperature variation compared to rural station (LCZ
reduction of IUHI by 0.17 °C (Fig. 7c). Similarly, increasing AR by 10% 9 and LCZ F) will exhibit smaller UHI–IUHI difference. When both the
would lead to an increase in IUHI and UHI by 0.17 °C and 0.08 °C models are tested for changes in DI, the difference between UHI and

8
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 7. Model fit plot for statistical models with confidence intervals.

IUHI magnitude is constant. This indicates that even when the distance negatively proportional to the distance from the major anthropogenic
is altered, the LCZ properties remains intact and its response to rural heat source (Coseo & Larsen, 2014; Kolokotroni & Giridharan, 2008;
reference station continue to follow the uniform trend. However, other Unger, Sümeghy, & Zoboki, 2001).
parameter’s partial plots shows that UHI & IUHI trend lines converges AR enable swift release of radiant heat by buildings and paved
to or diverges from each other depending on the parameter and LCZ surfaces in open type LCZs and their subclasses, as a result nocturnal
selected. Increasing AR for LCZ 6 shows gradual decrease in the dif- temperatures are observed to lie closer to the city’s average tempera-
ferences between UHI and IUHI. On the Contrary, LCZ 9’s partial plot ture. In addition to their high values of DI, the city's peripheries are
for the same parameter show an opposite trend. UHI model shows interspersed with large open grounds (lower AR), often only sparsely
higher temperature difference between the stations as compared to vegetated, resulting in the rapid loss of radiant heat after sunset. In
IUHI model. On the contrary, IUHI model considers and explains the contrast, the compact 'built' LCZs closest to the city's CBD (along the
significance of AL which further assists in providing mitigation solu- MHL route) unfailingly show a positive deviation from the city’s mean
tion. Although partial plots help in understanding UHI manifestation, it temperature (Refer Fig. 4). The beneficial effects of 'openness' in built
is practically not possible to alter parameters such as DI or AR in an form was observed in the subclasses of compact LCZs (LCZ 3F).
existing LCZ as a solution. However, AL and VDR are controllable Despite the insignificant bivariate correlation observed between
parameters and provide an opportunity to reduce IUHI with minimum UHI intensity and VDR before the regression modeling (as shown in
intervention. Fig. 5), post regression analysis indicates the presence of vegetation
strongly contributes to significant nocturnal cooling. This result is in
4.3. Study of relevant parameters line with those reported in other low-latitude cities of Hong Kong,
Singapore, Colombo, and Kochi (Chow & Roth, 2006; Kolokotroni &
A change in the combination of AL and VDR was explored to un- Giridharan, 2008; Senanayake, Welivitiya, & Nadeeka, 2013; Thomas
derstand its contribution in varying IUHI (as shown in Fig. 9). The x- et al., 2014). In case of Hong Kong, it was proposed that careful balance
axis represents the VDR, y-axis represents the AL and resultant z-axis of vegetation cover with other urban structure parameters must be
shows the temperature difference with respect to LCZ 9 or IUHI for maintained in order to achieve appreciable reductions in UHI magni-
typical summer night. The result shows that every 5% increase in VDR tudes (Giridharan et al., 2005, 2007; Giridharan, Lau, Ganesan, &
and AL altogether results in 1.07 °C drop in canopy layer IUHI. It was Givoni, 2008). Senanayake et al. (2013) emphasized the cooling effect
observed that if vegetation cover drops to zero and AL reduces to 0.20, of vegetation cover in urban areas in Colombo. Previous studies in
UHI of LCZ 3 would increase up to 3.78 °C. Although UHI model doesn’t Nagpur reported similar results, indicating that areas with the max-
provide significant relationship with AL for the case, it’s partial plot of imum built-up area and least vegetation cover were the hottest in the
VDR shows similar trend to that observed in IUHI model (Fig. 8b). summer season (Katpatal, Kute, & Satapathy, 2008; Kotharkar &
Surawar, 2016).
5. Discussion AR is encountered most often in the previous UHI literature. Recent
studies in Hong Kong have hypothesized that tall, concrete residential
5.1. LCZ’s thermal behavior buildings with sharp edges and low aspect ratio could trap both short
and long wave radiation. Subsequently, multiple regression analysis
Night time summer canopy-layer UHI in Nagpur city shows that identified Sky View factor as a significant predictor of nocturnal UHI in
most compact 'built' classes exhibit highest UHI & IUHI magnitude residential estates (Giridharan et al., 2005, 2007). Bottyan and Unger's
whereas most open 'land cover' classes show lowest UHI intensities. The study of Szeged notes that since the deep street canyons in dense urban
maximum difference in nocturnal near surface air temperature between areas are not fully accounted in urban cover parameters, it is essential
two LCZs (ΔTLCZ) is 3.53 °C in summer which corresponds to the same to factor in the vertical geometry of cities with the help of AR or SVF
order of magnitude as those reported in other tropical cities (Chow & (Bottyan & Unger, 2003). Lan and Zhan (2017) also presented sig-
Roth, 2006; Perera et al., 2012; Thomas et al., 2014). nificance of building density and building height in addressing night-
time temperature variation.
5.2. Urban planning and UHI parameters Unger (2004) reviewed several landmark studies that used AR and
SVF to predict UHI magnitude, following which he statistically calcu-
The distance from the major anthropogenic heat source in the city – lated significant negative coefficients of correlation between UHI in-
the primary central business district (DI), showed a strong negative tensity (ΔT) and SVF for all seasons in Szeged (Unger, 2004). In the
linear relationship with UHI intensity. In tropical city of Singapore, it course of the present study, the results of simple linear regression noted
was observed that increased UHI intensity lie close to the CBD which that the controllable predictors AL and VDR explained a larger portion
could be attributed to the presence of enhanced anthropogenic heat of variation in UHI intensities during summer nights in comparison to
sources well past sunset (Runnalls & Oke, 2000). Studies from Szeged, other built parameters in the model.
London and Chicago have shown similar results, with UHI intensity

9
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 8. IUHI and UHI partial effect plots with (a) DI; (b) VDR; (c) AL and; (d) AR.

5.3. Modeling technique and application minimum when variable input lied within the data range used for
training the model. Even though the model statistically shows marginal
A rural reference station used for predicting heat island response values of multicolinearity, careful considerations are required for pre-
was found preferable. Stepwise linear regression using similar data dictor’s value input to generate different urban scenario. For example,
produced unfavourable results which failed in the coefficient signage’s increasing vegetation density ratio will certainly affect the surface al-
checks. It’s significant model predicted an inverse relationship between bedo. Therefore, it was found that model inputs are required to be dealt
impervious surface fraction and heat island magnitude. All possible with a holistic perspective, with consideration to any physical limita-
regression allows to examine the recurring variables in significant tion, before generating UHI scenarios.
models. Meticulous care should be taken while selecting the signage or
polarity checks for the coefficient. Since response variable is a tem-
6. Conclusion
perature difference, variation in polarity of some of the coefficients may
depend on selection of reference station. Prediction errors were found
DI, AL, AR and VDR (in the order of hierarchy) contributes in

10
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

Fig. 9. Possible variation in IUHI on changing albedo and vegetation density.

explaining the UHI and built relation in more than 50% and 40% of the Journal of Information and Computation Technology, 4, 1069–1074. https://www.
significant IUHI and UHI models respectively. DI, AL, AR and VDR ripublication.com/irph/ijict_spl/ijictv4n11spl_08.pdf.
Alcoforado, M. J., & Andrade, H. (2006). Nocturnal urban heat island in Lisbon
explains 88–96% of IUHI variation during calm summer nights. Only (Portugal): Main features and modelling attempts. Theoretical and Applied Climatology,
three significant variables (DI, AR and VDR) were able to explain UHI 84, 151–159. https://doi.org/10.1007/s00704-005-0152-1.
model. Appropriate mitigation technique could be based on this study. Bottyan, Z., & Unger, J. (2003). A multiple linear statistical model for estimating the
mean maximum urban heat island. Theoretical and Applied Climatology, 75, 233–243.
Mitigation techniques may involve modification or changes in aforesaid https://doi.org/10.1007/s00704-003-0735-7.
parameters considering the existing urban scenario wherein modifica- Census (2011). District census handbook Nagpur – Village and town wise primary census
tion in some parameters (Like AR, DI and PD) might be very difficult to abstract. Retrieved 12 1, 2018, from District Census Hand Book:http://censusindia.
gov.in/2011census/dchb/2709_PART_B_DCHB_%20NAGPUR.pdf.
implement or practice. On the contrary, AL and VDR provides com-
Chow, W. T., & Roth, M. (2006). Temporal dynamics of the urban heat island of
paratively better mitigation opportunities in such areas. Local bodies Singapore. International Journal of Climatology, 26, 2243–2260. https://doi.org/10.
could therefore consider the findings from this study and evolve urban 1002/joc.1364.
Collier, C. G. (2006). The impact of urban areas on weather. Quarterly Journal of the Royal
planning policies, framework and recommendation to impede UHI de-
Meteorological Society, 132(614), 1–25.
velopment. Coseo, P., & Larsen, L. (2014). How factors of land use/land cover, building configuration,
Nagpur city features flat topography, a smaller number of water and adjacent heat sources and sinks explain Urban Heat Islands in Chicago. Landscape
bodies, mono-centric development with radial road network. On one and Urban Planning, 125, 117–129. https://doi.org/10.1016/j.landurbplan.2014.02.
019.
hand, the model provides best result for this city whereas its limitation CRISIL (2015). City development plan. Retrieved 12 1, 2018, from Nagpur Municipal
lies in explaining UHI variation due to presence of water bodies or Corporation:http://203.129.203.8/assets/250/2018/10/mediafiles/Final_CDP_
changing topography. The application of this study extends to cities Nagpur_-Mar_15.pdf.
Deng, C., & Wu, C. (2013). Examining the impacts of urban biophysical compositions on
with similar climate, geography and urban climate setting. The research surface urban heat island: A spectral unmixing and thermal mixing approach. Remote
approach mentioned could be used to further explore advance case Sensing of Environment, 131, 262–274.
specific models using different time of interest, additional variables and Ek, M. B., Mitchell, K. E., Lin, Y., Rogers, E., Grunmann, P., Koren, V., ... Tarpley, J. D.
(2003). Implementation of NOAH land surface model advances in the National
consideration of climatic factors. Future research opportunities lies in Centers for Environmental Prediction operational mesoscale Eta model. Journal of
examining more high density heterogeneous cities and developing a Geophysical Research: Atmospheres, 108(D22).
generic model for UHI mitigation. Emmanuel, R., & Krüger, E. (2012). Urban heat island and its impact on climate change
resilience in a shrinking city: The case of Glasgow, UK. Building and Environment, 53,
137–149. https://doi.org/10.1016/j.buildenv.2012.01.020.
Acknowledgements Erell, E., Pearlmutter, D., & Williamson, T. (2012). Urban microclimate: Designing the spaces
between buildings. Retrieved fromRoutledgehttps://www.taylorfrancis.com/books/
9781136539435.
The study is supported and funded by Science and Engineering
Field, C. B., Barros, V., Stocker, T., Qin, D., Dokken, D., Ebi, K., ... Tignor, M. (2012).
Research Board, Department of Science and Technology, New Delhi, IPCC, 2012: Managing the risks of extreme events and disasters to advance climate
India [DST-SERB project SB/S4/AS-110/2013]. Special thanks to Dr. change adaptation. A special report of Working Groups I and II of the Intergovernmental
Ravindra B Keskar and Varun Kotharkar for their guidance on statistical Panel on Climate Change, Cambridge University Press, Cambridge, UK, and New York,
NY, USA, 30(11), 7575–7613. https://www.ipcc.ch/site/assets/uploads/2018/03/
analysis. SREX_Full_Report-1.pdf.
Giridharan, R., Lau, S. S., & Ganesan, S. (2005). Nocturnal heat island effect in urban
References residential developments of Hong Kong. Energy and Buildings, 37, 964–971. https://
doi.org/10.1016/j.enbuild.2004.12.005.
Giridharan, R., Lau, S. S., Ganesan, S., & Givoni, B. (2007). Urban design factors influ-
Agarwal, R., Sharma, U., & Taxak, A. (2014). Remote sensing based assessment of Urban encing heat island intensity in high-rise high-density environments of Hong Kong.
Heat Island phenomenon in Nagpur metropolitan area. Retrieved from International Building and Environment, 42, 3669–3684. https://doi.org/10.1016/j.buildenv.2006.

11
R. Kotharkar, et al. Landscape and Urban Planning 190 (2019) 103586

09.011. Oke, T. R. (2006). Initial guidance to obtain representative meteorological observations at

Giridharan, R., Lau, S. S., Ganesan, S., & Givoni, B. (2008). Lowering the outdoor tem- urban sites. Retrieved fromhttps://weather.gladstonefamily.net/UrbanMetOps.pdf.
perature in high-rise high-density residential developments of coastal Hong Kong: Oleson, K. W., Bonan, G. B., Feddema, J., & Vertenstein, M. (2008). An urban para-
The vegetation influence. Building and Environment, 43, 1583–1595. https://doi.org/ meterization for a global climate model. Part II: Sensitivity to input parameters and
10.1016/j.buildenv.2007.10.003. the simulated urban heat island in offline simulations. Journal of Applied Meteorology
Han, J. Y., Baik, J. J., & Lee, H. (2014). Urban impacts on precipitation. Asia-Pacific and Climatology, 47(4), 1061–1076.
Journal of Atmospheric Sciences, 50(1), 17–30. Peel, M. C., Finlayson, B. L., & McMahon, T. A. (2007). Updated world map of the
Harrell, F. E. (2015). Multivariable modeling strategies. Regression modeling strategies (pp. Koppen-Geiger climate classification. Hydrology and Earth System Sciences, 11,
63–102). Springer. https://doi.org/10.1007/978-3-319-19425-7_4. 1633–1644. https://doi.org/10.5194/hess-11-1633-2007.
Hijioka, Y., Lin, E., Pereira, J. J., Corlett, R. T., Cui, X., Insarov, G. E., ... Surjan, A. (2014). Perera, N. G., Emmanuel, M. P., & Mahanama, P. K. (2012). 576: Mapping “Local Climate
Asia. In V. R. Barros, C. B. Field, D. J. Dokken, M. D. Mastrandrea, K. J. Mach, T. E. Zones” and relative warming effects in Colombo, Sri Lanka. ICUC8–8th international
Bilir, M. Chatterjee, K. L. Ebi, Y. O. Estrada, R. C. Genova, B. Girma, E. S. Kissel, A. N. conference on urban climates.
Levy, S. MacCracken, P. R. Mastrandrea, & L. L. White (Eds.). Climate change 2014: Petralli, M., Massetti, L., Brandani, G., & Orlandini, S. (2014). Urban planning indicators:
Impacts, adaptation, and vulnerability. Part B: Regional aspects. contribution of working Useful tools to measure the effect of urbanization and vegetation on summer air
group II to the fifth assessment report of the intergovernmental panel on climate change (pp. temperatures. International Journal of Climatology, 34, 1236–1244. https://doi.org/
1327–1370). Cambridge, United Kingdom and New York, NY, USA: Cambridge 10.1002/joc.3760.
University Press. Phelan, P. E., Kaloush, K., Miner, M., Golden, J., Phelan, B., Silva, H., III, & Taylor, R. A.
Hothorn, T., Zeileis, A., Farebrother, R. W., Cummins, C., Millo, G., & Mitchell, D. (2015). (2015). Urban heat island: Mechanisms, implications, and possible remedies. Annual
lmtest: Testing linear regression models. R package version 0.9-34. URLhttps://CRAN.R- Review of Environment and Resources, 40, 285–307.
project.org/package=lmtest. Retrieved from https://cran.r-project.org/web/ Powers, J. G., Klemp, J. B., Skamarock, W. C., Davis, C. A., Dudhia, J., Gill, D. O., ... Grell,
packages/lmtest/lmtest.pdf. G. A. (2017). The weather research and forecasting model: Overview, system efforts,
Ivajnšič, D., Kaligarič, M., & Žiberna, I. (2014). Geographically weighted regression of the and future directions. Bulletin of the American Meteorological Society, 98(8),
urban heat island of a small city. Applied Geography, 53, 341–353. https://doi.org/10. 1717–1737.
1016/j.apgeog.2014.07.001. Roth, M. (2012). Urban heat islands. Handbook of environmental fluid dynamics (pp. 160–
Katpatal, Y. B., Kute, A., & Satapathy, D. R. (2008). Surface- and air-temperature studies 177). CRC Press.
in relation to land use/land cover of Nagpur urban area using landsat 5 TM data. Runnalls, K. E., & Oke, T. R. (2000). Dynamics and controls of the near-surface heat island
Journal of Urban Planning and Development-ASCE, 134, 110–118. https://doi.org/10. of Vancouver, British Columbia. Physical Geography, 21, 283–304. https://doi.org/10.
1061/(ASCE)0733-9488(2008) 134:3(110). 1080/02723646.2000.10642711.
Khandar, V., & Garg, R. D. (2014). Heat island analysis of Nagpur and surrounding area Santamouris, M. (2014). On the energy impact of urban heat island and global warming
using geomatics techniques. Retrieved from International Journal of Engineering on buildings. Energy and Buildings, 82, 100–113.
Research & Technology (IJERT), 3, 1929–1933. https://www.ijert.org/heat-island- Senanayake, I. P., Welivitiya, W., & Nadeeka, P. M. (2013). Remote sensing based analysis
analysis-of-nagpur-and-surrounding-area-using-geomatics-techniques. of urban heat islands with vegetation cover in Colombo city, Sri Lanka using Landsat-
Kolokotroni, M., & Giridharan, R. (2008). Urban heat island intensity in London: An in- 7 ETM+ data. Urban Climate, 19–35, 5. https://doi.org/10.1016/j.uclim.2013.07.
vestigation of the impact of physical characteristics on changes in outdoor air tem- 004.
perature during summer. Solar Energy, 82, 986–998. https://doi.org/10.1016/j. Siu, L. W., & Hart, M. A. (2013). Quantifying urban heat island intensity in Hong Kong
solener.2008.05.004. SAR, China. Environmental Monitoring and Assessment, 185, 4383–4398. https://doi.
Kotharkar, R., & Bagade, A. (2018a). Evaluating urban heat island in the critical local org/10.1007/s10661-012-2876-6.
climate zones of an Indian city. Landscape and Urban Planning, 169, 92–104. https:// Stewart, I. D. (2011). A systematic review and scientific critique of methodology in
doi.org/10.1016/j.landurbplan.2017.08.009. modern urban heat island literature. International Journal of Climatology, 31, 200–217.
Kotharkar, R., & Bagade, A. (2018b). Local climate zone classification for Indian cities: A https://doi.org/10.1002/joc.2141.
case study of Nagpur. Urban Climate, 24, 369–392. https://doi.org/10.1016/j.uclim. Stewart, I. D., & Oke, T. R. (2012). Local climate zones for urban temperature studies.
2017.03.003. Bulletin of the American Meteorological Society, 93, 1879–1900. https://doi.org/10.
Kotharkar, R., & Surawar, M. (2016). Land use, land cover, and population density impact 1175/BAMS-D-11-00019.1.
on the formation of canopy urban heat islands through traverse survey in the Nagpur Szymanowski, M., & Kryza, M. (2009). GIS-based techniques for urban heat island spa-
Urban Area, India. Journal of Urban Planning and Development, 142. https://doi.org/ tialization. Climate Research, 38(2), 171–187.
10.1061/(ASCE)UP.1943-5444.0000277. Thomas, G., Sherin, A. P., Ansar, S., & Zachariah, E. J. (2014). Analysis of urban heat
Krüger, E., & Emmanuel, R. (2013). Accounting for atmospheric stability conditions in island in Kochi, India, using a modified local climate zone classification. In L. Beesley
urban heat island studies: The case of Glasgow, UK. Landscape and Urban Planning, (Ed.). Urban environmental pollution 2013 – Creating healthy, livable cities (pp. 3–13). .
117, 112–121. https://doi.org/10.1016/j.landurbplan.2013.04.019. https://doi.org/10.1016/j.proenv.2014.09.002.
Kusaka, H., Kondo, H., Kikegawa, Y., & Kimura, F. (2001). A simple single-layer urban Unger, J. (2004). Intra-urban relationship between surface geometry and urban heat is-
canopy model for atmospheric models: Comparison with multi-layer and slab models. land: Review and new approach. Climate Research, 27, 253–264. https://doi.org/10.
Boundary-layer meteorology, 101(3), 329–358. 3354/cr027253.
Lan, Y., & Zhan, Q. (2017). How do urban buildings impact summer air temperature? The Unger, J., Gál, T. M., Rakonczai, J., Mucsi, L., Szatmári, J., Tobak, Z., ... Fiala, K. (2010).
effects of building configurations in space and time. Building and Environment, 125, Modeling of the urban heat island pattern based on the relationship between surface
88–98. https://doi.org/10.1016/j.buildenv.2017.08.046. and air temperatures. Időjárás/Quarterly Journal of the Hungarian Meteorological
Leconte, F., Bouyer, J., Claverie, R., & Petrissans, M. (2015). Using local climate zone Service, 114, 287–302.
scheme for UHI assessment: Evaluation of the method using mobile measurements. Unger, J., Sümeghy, Z., & Zoboki, J. (2001). Temperature cross-section features in an
Building and Environment, 83, 39–49. https://doi.org/10.1016/j.buildenv.2014.05. urban area. Atmospheric Research, 58, 117–127. https://doi.org/10.1016/S0169-
005. 8095(01)00087-4.
Martilli, A., Clappier, A., & Rotach, M. W. (2002). An urban surface exchange para- United Nations. (2018). Revision of world urbanization prospects, 2018.
meterisation for mesoscale models. Boundary-Layer Meteorology, 104(2), 261–304. Wang, Z. H., Bou-Zeid, E., & Smith, J. A. (2013). A coupled energy transport and hy-
Masson, V. (2000). A physically-based scheme for the urban energy budget in atmo- drological model for urban canopies evaluated using a wireless sensor network.
spheric models. Boundary-layer meteorology, 94(3), 357–397. Quarterly Journal of the Royal Meteorological Society, 139(675), 1643–1657.
Matzarakis, A., & Rutz, F. (2007). RayMan: A tool for tourism and applied climatology. Xiao, R. B., Ouyang, Z. Y., Zheng, H., Li, W. F., Schienke, E. W., & Wang, X. K. (2007).
Development in Tourism Climatology, 129–138. Spatial pattern of impervious surfaces and their impacts on land surface temperature
Mirzaei, P. A., & Haghighat, F. (2010). Approaches to study urban heat island – Abilities in Beijing, China. Journal of Environmental Sciences, 19(2), 250–256.
and limitations. Building and Environment, 45(10), 2192–2201. Yang, J., Wang, Z. H., Chen, F., Miao, S., Tewari, M., Voogt, J. A., & Myint, S. (2015).
Mohan, M., & Siddiqui, T. A. (1998). Analysis of various schemes for the estimation of Enhancing hydrologic modelling in the coupled weather research and forecasting-
atmospheric stability classification. Atmospheric Environment, 32(21), 3775–3781. urban modelling system. Boundary-Layer Meteorology, 155(1), 87–109.
Nagpur Municipal Corporation (2014). Environmental status report 2013-14. Nagpur. Yuan, C., & Chen, L. (2011). Mitigating urban heat island effects in high-density cities
Nandankar, P. K., Dewangan, P. L., & Surpam, R. V. (2011). Climate of Nagpur. Nagpur: based on sky view factor and urban morphological understanding: A study of Hong
India Meteorological Department. (accessed 29 March 2016) http://imdnagpur.gov. Kong. Architectural Science Review, 54(4), 305–315.
in/ClimateNGP.pdf. Zhang, Y., Odeh, I. O., & Han, C. (2009). Bi-temporal characterization of land surface
Oke, T. R. (1981). Canyon geometry and the nocturnal urban heat island: Comparison of temperature in relation to impervious surface area, NDVI and NDBI, using a sub-pixel
scale model and field observations. Journal of Climatology, 1, 237–254. https://doi. image analysis. International Journal of Applied Earth Observation and Geoinformation,
org/10.1002/joc.3370010304. 11(4), 256–264.

12