Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/262527017

EffectsofspatialpatternofgreenspaceonurbancoolinginalargemetropolitanareaofeasternChina

ARTICLEinLANDSCAPEANDURBANPLANNING·AUGUST2014

ImpactFactor:3.04·DOI:10.1016/j.landurbplan.2014.04.018

CITATIONS

5

READS

165

5AUTHORS,INCLUDING:

FanhuaKong

NanjingUniversity

22PUBLICATIONS372CITATIONS

SEEPROFILE

PhilipJames

UniversityofSalford

47PUBLICATIONS921CITATIONS

SEEPROFILE

HongSHe

UniversityofMissouri

197PUBLICATIONS3,749CITATIONS

SEEPROFILE

Availablefrom:HongSHe

Retrievedon:05February2016

R

Em

Fa

b

c

d

e

h

•••••

a

ARRAA

KCLLUU

1

p

(H

h0

Landscape and Urban Planning 128 (2014) 35–47

Contents lists available at ScienceDirect

Landscape and Urban Planning

j our na l ho me pa g e: www.elsev ier .com/ locate / landurbplan

esearch Paper

ffects of spatial pattern of greenspace on urban cooling in a largeetropolitan area of eastern China

anhua Konga,∗, Haiwei Yinb, Philip Jamesc, Lucy R. Hutyrad, Hong S. Hee

International Institute for Earth System Science (ESSI), Nanjing University, No. 163 Xianlin Ave., Nanjing 210023, ChinaDepartment of Urban Planning and Design, Nanjing University, No. 22 Hankou Road, Nanjing 210093, ChinaSchool of Environment and Life Sciences, University of Salford, Salford M5 4WT, UKDepartment of Earth & Environment, Boston University, 685 Commonwealth Avenue, Boston, MA 02215, USASchool of Natural Resources, University of Missouri, Columbia, MO 65203, USA

i g h l i g h t s

An optimal spatial scale for examining greenspace cooling effects was identified.The effect of greenspace configuration on the land surface temperature was quantified.Spatial configuration of a mainland-island greenspace enhances the cooling effect.Fragmented greenspace is also effective for cooling given a fixed amount of forest cover.Greenspace cooling intensity can indicate cool island characteristics well.

r t i c l e i n f o

rticle history:eceived 8 November 2013eceived in revised form 16 April 2014ccepted 24 April 2014vailable online 22 May 2014

eywords:ooling effectandscape metricsandscape patternrban greenspacerban heat island

a b s t r a c t

Urban areas will experience the greatest increases in temperature resulting from climate change due tothe urban heat island (UHI) effect. Urban greenspace mitigates the UHI and provides cooler microclimates.Field research has established that temperatures within parks or beneath trees can be cooler than innon-greenspaces, but little is known about the effects of the spatial pattern of greenspace on urbantemperatures or the optimal spatial patterns needed to cool an urban environment. Here, urban coolislands (UCIs) and greenspace in Nanjing, China were identified from satellite data and the relationshipbetween them analyzed using correlation analyses. The results indicate the following: (1) Areas with ahigher percentage of forest-vegetation experience a greater cooling effect and a 10% increase in forest-vegetation area resulted in a decrease of about 0.83 ◦C in surface temperature; (2) A correlation analysisbetween mean patch size, patch density, and an aggregation index of forest vegetation with temperaturereduction showed that for a fixed amount of forest vegetation, fragmented greenspaces also provide

effective cooling; (3) The spatial pattern of UCIs was strongly correlated with greenspace patterns; amainland-island greenspace spatial configuration provided an efficient means of enhancing the coolingeffects; and (4) the intensity of the cooling effect was reflected in cool island characteristics. These findingswill support better prediction of the effects of specific amounts and spatial arrangements of greenspace,helping city managers and planners mitigate increasing temperatures associated with climate change.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

Climate change predictions indicate that, in the near future, tem-eratures will rise and there will be an increased frequency and

∗ Corresponding author. Tel.: +86 25 89681033; fax: +86 25 83316892.E-mail addresses: fanhuakong@163.com (F. Kong), qzyinhaiwei@163.com

H. Yin), P.James@salford.ac.uk (P. James), lrhutyra@bu.edu (L.R. Hutyra),eH@missouri.edu (H.S. He).

ttp://dx.doi.org/10.1016/j.landurbplan.2014.04.018169-2046/© 2014 Elsevier B.V. All rights reserved.

intensity of heat waves; as a result there will be negative effects onboth human health and the environment (Luber & McGeehin, 2008;Oliveira, Andrade, & Vaz, 2011; Stott, Stone, & Allen, 2004; Tan et al.,2007; Yuan & Bauer, 2007). Large urban areas are warmer than thesurrounding countryside – a phenomenon known as the urban heatisland (UHI). The higher temperatures in an UHI result in reduced

thermal comfort with an associated increased in energy consump-tion as measures are taken to cool homes and offices. UHIs can alsoserve as a trap for atmospheric pollutants, contribute to increasedurban smog formation, and generate socioeconomic impacts on

3 Urban

cweT1tegtqiNGSgU1acM2bdK

eowd2&YCSof–dS2ts2Ksgbl2rm1etsermr

rbtSts

6 F. Kong et al. / Landscape and

ommunities all of which affect the quality of life of those living andorking in urban areas (Alghannam & Al-Qahtnai, 2012; Gobakis

t al., 2011; Mihalakakou, Santamouris, Papanikolaou, Cartalis, &sangrassoulis, 2004; Ng, Chen, Wang, & Yuan, 2012; Price, 1979,983). The predicted magnitude and speed of climate change overhe next 30–40 years (Gill, Handley, Ennos, & Pauleit, 2007; Hulmet al., 2002), suggests an urgent need for the development of strate-ies to adapt to and mitigate the expected continual increase inemperature (Bowler, Buyung-Ali, Knight, & Pullin, 2010). Conse-uently, the role of urban greenspace in moderating urban climates

s being studied extensively (Chen & Wong, 2006; Dimoudi &ikolopoulou, 2003; Eliasson, 2000; Georgi & Dimitriou, 2010;ivoni, 1991; Hamada & Ohta, 2010; Oliveira et al., 2011; Zoulia,antamouris, & Dimoudi, 2009) with researchers consistently sug-esting that an effective way to reduce or alleviate the effects ofHIs is to increase tree cover area and density (Rosenfeld et al.,995; Semrau, 1992). Vegetation, mainly through direct shadingnd evapotranspiration, can reduce temperatures and create a localool island within an urban area (Oke, Crowther, McNaughton,onteith, & Gardiner, 1989; Tyrväinen, Pauleit, Seeland, & de Vries,

005). Vegetation in a city can also achieve other environmentalenefits such as reduced storm water runoff, greater urban bio-iversity, and improved esthetics (Kong, Yin, & Nakagoshi, 2007;ong, Yin, Nakagoshi, & Zong, 2010).

Traditionally, on-site observation using fixed stations or mobilequipment has been widely used to investigate the cooling effectf greenspace. The type, size, and shape of greenspace patches, asell as tree shade area have been identified as important factors inetermining the cooling effect of greenspaces (Chang, Li, & Chang,007; Fahmy, Sharples, & Yahia, 2010; Giridharan, Lau, Ganesan,

Givoni, 2008; Jauregui, 1990; Jusuf, Wong, Hagen, Anggoro, &an, 2007; Katayama, Ishii, Hayashi, & Tsutsumi, 1993; Potchter,ohen, & Bitan, 2006; Shashua-Bar & Hoffman, 2000; Spronken-mith & Oke, 1998; Upmanis, Eliasson, & Lindqvist, 1998). In termsf the cooling effect of vegetation, researchers typically rank trees,ollowed by bushes and then grass, as having the greatest effect

even a single tree can affect the air temperature of the imme-iate area (Rosenfeld, Romm, Akbari, Romm, & Pomerantz, 1998;aito, Ishihara, & Katayama, 1990–1991; Shashua-Bar & Hoffman,000). However, most research provides only qualitative descrip-ions of cooling effects and fails to establish quantifiable effects andtatistically significant relationships (Cao, Onishi, Chen, & Imura,010; Hemiddi, 1991; Honjo & Takakura, 1990; Jonsson, 2004;awashima, 1994; Narita et al., 2004; Wong et al., 2007). On-ite studies often take measurements in only a small number ofreen sites; but confirm that vegetation lowers air temperaturesy shading, by absorbing heat, and by converting ambient heat to

atent heat through evapotranspiration at a local scale (Cao et al.,010). A consensus has also developed among researchers that theelationship between the cooling effect and the size of greenspaceay not be linear (Cao et al., 2010; Chang et al., 2007; Jauregui,

990). Yet conclusions drawn from an individual study cannot beasily verified or transferred (Bowler et al., 2010) and hence quan-ifiable cooling effects and statistical relationships at the urbancale cannot be established from these site-specific studies (Changt al., 2007; Spronken-Smith & Oke, 1998). Consequently, the cur-ent body of evidence base does not allow recommendations to beade on how to best incorporate greening in an urban area for

educing temperatures (Bowler et al., 2010).The effects of greenspace on cooling can be measured using

emote sensing at all scales from a greenspace patch to a city andeyond. Rao (1972) first demonstrated the possibility of detec-

ing the thermal footprint of urban areas from satellite images.ubsequently, a wide range of remote sensing images and GISechnology has been used to retrieve spatially explicit land-urface temperature (LST) datasets. Significantly, remote sensing

Planning 128 (2014) 35–47

images provide detailed spatial land-use and land-cover (LULC)information that can be combined with LST datasets (Cao et al.,2010; Schwarz, Lautenbach, & Seppelt, 2011; Small, 2006; Tran,Uchihama, Ochi, & Yasuoka, 2006). Relating LST data to sur-face cover characteristics and assessing thermal conditions hasenabled the development of city level climate strategies (Amiri,Weng, Alimohammadi, & Alavipanah, 2009; Carlson & Arthur, 2000;Keramitsoglou, Kiranoudis, Ceriola, Weng, & Rajasekard, 2011; Kim,1992; Nichol, 2005; Stathopoulou & Cartalis, 2007;Weng, Lu, &Liang, 2006; Weng & Lu, 2008; Xiao et al., 2008).

Two decades of urban surface temperature studies haveadvanced our understanding of spatial thermal patterns, andgreenspace is now recognized as one of the most importantland-use types that contributes to reducing urban thermal effects(Buyantuyev & Wu, 2010; Gallo et al., 1993; Lu & Weng, 2006;Mackey, Lee, & Smith, 2012; Nichol, 1998; Owen, Carlson, & Gillies,1998; 1999; Quattrochi & Ridd, 1998; Sobrino, Raissouni, & Li,2001; Sobrino, Jiménez–Munoz, & Paolini, 2004; Weng & Larson,2005; Weng, Lu, & Schubring, 2004; Wilson, Clay, Martin, Stuckey,& Vedder-Risch, 2003; Xian & Crane, 2006). However, most pre-vious remote sensing studies of urban areas focused on providingqualitative descriptions of thermal patterns and simple correlationsbetween LST and LULC types (Keramitsoglou et al., 2011; Voogt& Oke, 2003), and provide only a limited exploration of how thecomposition and spatial arrangement of greenspaces affect cooling(Chang et al., 2007; Li et al., 2011; Li, Zhou, Ouyang, Xu, & Zheng,2012; Weng, 2009).

Recent developments in landscape ecology have made it pos-sible to characterize composition and spatial arrangement and toquantitatively link greenspace spatial heterogeneity to its cool-ing effects (Cao et al., 2010; Li et al., 2011, 2012; Zhou, Huang,Mary, & Cadenasso, 2011). For example, numerous studies haveshown that increased greenspace cover has a positive relationshipwith cooling effects. Li et al. (2012) reported that a 10% increase ingreenspace cover produced a 0.86 ◦C decrease in LST. However, theurban greenspace cooling effect is scale dependent and the opti-mal spatial scale for its study is not yet known (Li et al., 2011; Liu& Weng, 2009). Using landscape metrics, Weng, Liu, and Lu (2007)assessed the effects of LULC patterns on thermal conditions, whileLi et al. (2011) did the same at pixel and landscape scales. Li et al.(2012) investigated the impact of greenspace spatial patterns onLST using a census tract method and recommended that multi-scaleresearch be conducted, a conclusion with which we concur.

As past research illustrates, the characteristics of the urbangreenspace cooling effect are not fully understood; this limits theplanning and design of such space to mitigate thermal effects atthe city level. Consequently, the main objectives of our study areto: (1) investigate the sensitivity of the cooling effect associatedwith greenspace to changes in scale; (2) identify urban cool islands(UCIs) associated with greenspace – greenspace cool islands (GCIs)– and identify any relationships between GCIs and greenspace spa-tial patterns; and (3) characterize the intensity of the effects ofGCIs.

2. Study area

Nanjing (31◦14′–32◦37′ N, 118◦22′–119◦14′ E), the capital ofJiangsu Province, China, is located in the western edge of theYangtze River Delta (Fig. 1). The Nanjing metropolitan region,comprising 11 districts, covers an area of 4733 km2, and in 2010had a population of about 6.3 million (Nanjing Municipal Bureau

Statistics, 2011). Nanjing has a subtropical monsoon climate withfour seasons and a mean annual temperature of 15 ◦C (Jim &Chen, 2003). Since 1951, the mean daily maximum temperaturebetween June and August has been 37.3 ◦C, and on three occasions

F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47 37

F is dis( r is re

tvla(u4

3

3t

2(r(LtItiobtcfLsnbbsLK

T

wtt

ig. 1. Location map of Nanjing and the study area. The false color IKONOS imagegreen). (For interpretation of the references to color in this figure legend, the reade

emperatures exceeded 40 ◦C (Miao, Pan, & Xu, 2008). The regionalegetation consists of deciduous broadleaf and evergreen needle-eaf forest and bush-grass communities. The dominant tree speciesre Platanus acerifolia, Juniperus chinensis and Ligustrum lucidumJim & Chen, 2003). The area examined in this study includes therbanized area of Nanjing and part of its suburbs, an area of about32 km2.

. Data and methodology

.1. Image pre-processing and retrieving land-surfaceemperature

The data used in this research include IKONOS images (June 18,009, four bands, 3.2-m spatial resolution), Landsat TM 5 imagesJune 13, 2009), and meteorological data. The IKONOS images wereectified and georeferenced to the Universal Transverse MercatorUTM) coordinate system. Based on these IKONOS images, an urbanULC map was created by manual interpretation supported byhe ArcMap platform (Environmental Systems Research Institute,nc., Redlands, CA, USA), combined with field surveys and ground-ruthing as necessary. Land use was classified into six types:mpervious surface, forest vegetation (trees with shrub and grass),ther vegetation (shrub and grass), water, agricultural land, andarren land (Fig. 2a). To conduct a moving-window analysis, vec-or land-use data were converted to a grid format with a 5 m × 5 mell size. The Landsat 5 Thematic Mapper image (Row/Path: 120/38)rom 10:29 a.m. (local time), June 13, 2009, was used to retrieve theST. This Landsat image was rectified to a common UTM coordinateystem based on the IKONOS image, and was resampled using theearest-neighbor algorithm with a pixel size of 30 m × 30 m for allands. The LST was retrieved from thermal infrared data suppliedy band 6 (120 m spatial resolution) of the Thematic Mapper sen-or on board the Landsat 5 satellite. The methodology for retrievingST is based on the Mono-Window Algorithm (MWA) from Qin,arnieli, and Berliner (2001), with the following equation:

s = 1C6

{a6(1 − C6 − D6) + [b6(1 − C6 − D6) + C6 + D6]T6 − D6Ta},

(1)

here a6 and b6 are the coefficients −63.1885 and 0.44411, respec-ively (according to Qin et al. (2001) when the air temperature is inhe range of 10–40 ◦C), T6 is the at-sensor brightness temperature,

played with RGB composition of band 4 (near infrared), band 3 (red), and band 2ferred to the web version of the article.)

and Ta represents the mean atmospheric temperature. C6 and D6are defined, respectively, by Eqs. (2) and (3):

C6 = ε6�6, (2)

D6 = (1 − �6)[1 + �6(1 − ε6)], (3)

where ε6 is the land-surface emissivity, and �6 is the total atmo-spheric transmissivity. Qin et al. (2001) provide additional detailsrelated to this algorithm. A program called Mono-window Algo-rithm was run in the Spatial Model in ERDAS software (9.2). TheLST map was then resampled to 5 m × 5 m to match the land-usemap (Fig. 2b).

3.2. Moving-window method and window size chosen

A moving window was applied over the entire landscape usingFRAGSTATS (3.3) to calculate landscape metrics, returning valuesto the center cell and creating a new continuous surface grid mapfor each selected metric (Kong & Nakagoshi, 2006; McGarigal &Cushman, 2002a). As previous studies found the greenspace cool-ing effect to be scale dependent (Li et al., 2011; Liu & Weng,2009), four window sizes were tested (60 m × 60 m, 120 m × 120 m,240 m × 240 m and 480 m × 480 m) to find the optimal window sizeto investigate the relationship between temperature change – thedifference between the local LST and the mean LST of the study area– and the pattern of greenspace within the city. As the resolution ofthe TM band 6 is 120 m, a window size was chosen that was an inte-ger multiple of 120 m and an analysis to determine if the smallerscale (60 m × 60 m) would influence the results was conducted.

Previous researchers demonstrated that the proportional areaof greenspace within a region is one of the key factors influenc-ing the cooling effect (Connors, Galletti, & Chow, 2012; Jeneretteet al., 2007; Li et al., 2011; Shashua-Bar & Hoffman, 2000). The per-centage of forest vegetation (PLAND) in the selected window sizewas captured through the moving-window method. A significantrelationship exists between the percentage of forest vegetationand temperature (Fig. 3a–d) in different window sizes. However,the mean patch size of forest vegetation is 1.4 × 104 m2 (about thesame size as a 120 m × 120 m window), and the statistical analysis

shows that 95% of the patches of forest vegetation are smaller than160 m × 160 m. If a smaller window size is chosen (which meansthat large greenspace patches will be divided into small areas bythe analysis window), then the cooling effect caused by the entire

38 F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47

F atial df

ptttitgsptct

F1

ig. 2. (a) Land-use map of the study area in 2009 based on an IKONOS image. (b) Sprom a Landsat 5 Thematic Mapper image.

atch will be estimated incorrectly, which will influence the rela-ionship of the variables being tested (Fig. 3a and b). However, ifhe window size is too large, the patches with little forest vege-ation will not show a significant cooling effect. This is especiallymportant because large patches of forest vegetation are rare inhe urban area studied here. The cooling effect associated withreenspace is weakened by the influence of the surrounding land-cape, which diminishes the apparent relationship between the

ercentage of forest vegetation and temperature change. In addi-ion, if a large window size is used, there are insufficient samples toonduct the analysis in the study area (shown as Fig. 3d). To ensurehat most of the patches of forest vegetation were not divided as a

ig. 3. (a–d) Relationship between percentage of landscape (PLAND) of forest vegetatio20 m × 120 m, (c) 240 m × 240 m, and (d) 480 m × 480 m.

istribution of land-surface temperature of the study area on June 13, 2009, derived

result of using a small window size, and that there were sufficientsamples to conduct an analysis of the relationship between temper-ature change and the spatial pattern of forest vegetation, a windowsize of 240 m × 240 m was chosen for the subsequent analysis. Theresults displayed in Fig. 3c show a good fit (R2 = 0.873) when the240 m × 240 m window size was used.

3.3. Quantifying the greenspace pattern and defining urban cool

island (UCI) and greenspace cool island (GCI)

Eleven spatial landscape metrics were chosen to quantify thespatial pattern of greenspace and to analyze the relationship

n and the temperature change (◦C) in different window sizes; (a) 60 m × 60 m, (b)

F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47 39

Table 1Definitions of landscape metrics (based on McGarigal et al., 2002b).

Landscape metrics Abbreviation Description Units Range

Percent of landscape PLAND The proportion of total area occupied by a particular patch type; ameasure of landscape composition and dominance of patch types

Percent 0 < PLAND < 100

Class area CA The total class area Hectares CA > 0, no limitPatch density PD The number of patches per unit area n/km2 PD > 0Mean patch size MPS The area occupied by a particular patch type divided by the

number of patches of that typeHectares MPS > 0, no limit

Number of patch NP The number of patches of the corresponding patch type (class). None NP ≥ 1, no limitMean patch shape

indexShape MN Mean value of shape index None Shape mn ≥ 1, no

limit,Area-weighted

perimeter area ratioPARA AM The sum, across all patches of the corresponding patch type, of the

corresponding perimeter area ratio multiplied by the value ofpatch area (m2) divided by the sum of patch areas.

None PARA AM > 0

Largest patch index LPI The area (m2) of the largest patch of the corresponding patch typedivided by total landscape area (m2), multiplied by 100 (to convertto a percentage).

Percent 0 < LPI < 100

Aggregation index AI The number of like adjacencies involving the corresponding class,divided by the maximum possible number of like adjacenciesinvolving the corresponding class, which is achieved when theclass is maximally clumped into a single, compact patch;multiplied by 100 (to convert to a percentage).

Percent 0 ≤ AI ≤ 100

Shannon’s diversityindex

SHDI SHDI equals minus the sum, across all patch types, of theproportional abundance of each patch type multiplied by that

None SHDI ≥ 1, no limit

ween

he sum

bc(ai((

tUstacai

d4tmts

4

4c

4c

oilo

src

proportion.Area-weighted mean

radius of gyrationGYRATE AM The mean distance (m) bet

patch centroid divided by t

etween these patterns and GCI (Table 1): proportional of landover (PLAND), class area (CA), patch density (PD), mean patch sizeMPS), patch number (NP), mean patch shape index (Shape MN),rea-weighted mean perimeter area ratio (PARA AM), largest patchndex (LPI), aggregation index (AI), Shannon’s diversity indexSHDI), and area-weighted mean radius of gyration (GYRATE AM)McGarigal et al., 2002b).

We calculated the mean LST (T) of the study area, the tempera-ure in each pixel (Ti) and the urban cool island (UCI) as the anomalyCI = �T = Ti − T(�T ≤ 0), where T =

∑n1Ti/n, that is to say the

ummary of each pixel land surface temperature (Ti) divided byhe total number of pixels in the study area. The spatial extent of

cool island was estimated as the area covered by the aggregatedluster of pixels with a LST lower than the T – which is treateds the reference land-surface temperature (RLST). The cool islandsdentified were then classified as objects and extracted.

UCIs that were primarily the result of forest vegetation wereesignated GCIs. The rule to define a GCI is shown in Section.2.2.Cooling intensity was estimated as the maximum tempera-ure reduction in the cool island, using the difference between the

inimum LST of a cool island and the T (in ◦C). It was capturedhrough the Zonal Analysis of the spatial analysis tool in ArcMapoftware.

. Results

.1. Relationship between the spatial pattern of greenspace andooling effect

.1.1. General characteristics of the land use and itsorresponding land-surface temperature

Forest vegetation, which covers about 9200 ha (shown by CA)r about 21.3% of the study area, has low AI and MPS and high NPndicating that it has a more highly fragmented pattern than otherand use types. However, the high LPI implies the existence, nearby,f some large greenspace patches (Table 2).

The land-surface temperature (LST) illustrates the thermal land-cape pattern of the study area (Fig. 2b). To better understand theelationship between LST and LULC, a zonal statistical analysis wasonducted to determine the meanLST of different land cover types

each cell in the patch and the of patch areas.

Meter GYRATE AM ≥ 0

(Table 3). The results indicate that the LST of urban surfaces cor-responds closely to the distribution of LULC characteristics. Thisis confirmed by most previous studies in which it was reportedthat water (−8.8 ◦C) and forest vegetation (−3.2 ◦C) had the largestcooling effects compared to the mean LST of impervious surfaces(Keramitsoglou et al., 2011; Weng et al., 2004).

4.1.2. The relationships of greenspace patterns with cooling effectat a fixed scale

Sample analysis was conducted to quantify the relationshipbetween the percentage of different LULC types in a 240 m × 240 mwindow and the associated temperature change. The relationshipbetween percentage of forest vegetation and temperature reduc-tion was assessed for sample points where the percentage of forestvegetation was >0, excluding pixels that are not forest vegeta-tion, impervious surface, and barren land so that effects of onlythese three classes were analyzed within each window. We foundthat a cool island will be created when forest vegetation occu-pies over 61.2% of the 240 m × 240 m window (Fig. 3c). The largestcooling effect (−4.9 ◦C) was found in a window completely occu-pied by forest vegetation. A sample analysis was also conductedto determine the impact of other vegetation (shrub and grass) onthe cooling effect (Fig. 4). The data indicate that if the entire win-dow (actually 99.69% according to the calculation) is covered byother vegetation (shrub and grass), then a cool island is formed.Linear regression analysis of the relationship between tempera-ture change and percentage of other vegetation (Fig. 4a) shows thatits relationship (R2 = 0.311, p < 0.01) is not as strong as with forestvegetation (R2 = 0.873, p < 0.01), although there is a cooling effectwith the increase of the percentage of other vegetation. Thus, thefollowing sections of this paper focus only on forest vegetation.

We found that water had a significant cooling effect when itcovered more than 20.7% of the chosen window size and there wasno vegetation or agricultural land (Fig. 4b). The effect of agricul-tural land on temperature change was also statistically significant(R2 = 0.702, p < 0.01), but the effect was not as strong as that of for-

est vegetation or water as indicated by their regression coefficients(Fig. 4c). The presence of impervious surfaces had a strong relation-ship with temperature when an impervious surface covered about20.5% of the 240 m × 240 m window size. There was a significant

40 F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47

Table 2General landscape characteristics in Nanjing, 2009.

LULC type CA (ha) PLAND (%) NP LPI MPS (ha) AI

Impervious surface 20,491.14 47.43 2786 44.95 7.36 91.64Forest vegetation 9200.91 21.30 8049 5.85 1.14 87.81Other vegetation 4270.95 9.89 5719 0.30 0.75 83.91Water 5590.05 12.94 1493 4.01 3.74 94.50Barren land 1030.49 2.39

Agricultural land 2618.83 6.06

Table 3Results of the mean LST of different LULC types and the cooling effect (Unit: ◦C).

LULC type MeanLST

Temperaturereductioncompared withthe mean LSTof study area

Temperaturereductioncompared with themean LST ofimpervious surface

Impervious surface 31.73 3.27 0Water 22.94 −5.52 −8.78Agricultural land 28.52 0.06 −3.21Forest vegetation 27.73 −0.73 −4.00Other vegetation 29.78 1.32 −1.94

iwoupe

otbtwcoM

F(

Barren land 30.05 1.59 −1.68

ncrease of the LST, created a warming effect (Fig. 4d), when thereas no other land use type except vegetation (i.e. forest vegetation

r other vegetation) in the window. The results of different landse types in decreasing or increasing the LST were consistent withrevious research (Buyantuyev & Wu, 2010; Li et al., 2012; Wengt al., 2006).

Other landscape metrics were explored to assess the influencef fragmentation and the shape of forest vegetation patches onhe cooling effect. Linear regression analysis of the relationshipetween the temperature change and MPS of the forest vegeta-ion shows that if the MPS is over 3.25 ha in the 240 m × 240 m

indow, a cooling effect will be present (Fig. 5a). However, the

ooling effect was also present when there was more than 61.2%,r 3.53 ha of forest vegetation in a window (Fig. 3c) even if thePS is less than 3.25 ha. In combination with the correlation

ig. 4. Relationships between the percentage of landscape (PLAND) and (a) other vegetat◦C) shown by linear regression and relativity analysis.

494 0.06 2.09 89.25268 0.61 9.77 94.00

analysis between PD, AI of vegetation forest, and temperaturechange (Fig. 5b and c), these data imply that cooling effects area function of the spatial extent of the area of forest vegetation.However, given a fixed amount of forest vegetation, fragmentedgreenspaces also effectively provide a cooling effect.

We found that temperature change has a weak, positive corre-lation with Shape MN (R2 = 0.187), but it has a strong relationshipwith the area-weighted mean perimeter area ration (PARA AM)(R2 = 0.803). The results indicate that if forest-vegetation area isheld constant, the temperature change has a strong relationshipwith the ratio of patch perimeter to area which implies that theconfiguration of the greenspace patches being studied has a stronginfluence on temperature change (Fig. 5d and e). At the land-scape level (for each window size), scatter plots of the relationshipbetween temperature change and the SHDI did not show constantnegative or positive trends (Fig. 5f). This does not appear to beconsistent with previous research (Connors et al., 2012; Li et al.,2012).

4.2. Relationship between spatial pattern characteristics of GCIand greenspace at the class level

4.2.1. General characteristics of cool islands in NanjingTwo extensive and dominant UCIs were found in the study area:

one along the Yangtze River (including Mufu Mountain) and the

second at the Purple Mountain (including Xuanwu Lake) (Fig. 6a).In addition, many smaller UCI corridors in the city follow the pathof the Qinhuai River (Fig. 6a). UCIs can be clearly identified in thesouthern portion of the study area but fewer UCIs are found near

ion, (b) water, (c) agricultural land, (d) impervious surface and temperature change

F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47 41

F (b) pa( dex (S

tw

at1Udto

TG

N

ig. 5. Regression analysis between forest vegetation and (a) mean patch size (MPS),e) area-weighted mean perimeter area ratio (PARA AM), (f) Shannon’s diversity in

he Xinjiekou (downtown) area (Fig. 6a). A total of 808 UCI patchesith a mean patch size of 17.18 ha were identified (Table 4).

Using the Tabulate Area model, a part of the spatial statisticalnalysis in ArcMap, and landscape spatial metrics to investigatehe general characteristics of each UCI revealed that UCIs covered38 km2 or about 32% of the entire study area. About 40% of the

CIs were composed of forest vegetation and 35% of water. Theseata indicate that the greatest cooling effect exceeds −9 ◦C fromhe mean LST of the study area and the average cooling effect of allf the cool islands is roughly −1.3 ◦C (Table 4).

able 4eneral characteristics of UCI landscapes in Nanjing.

Composition of UCIs LULC type

Impervious surfacWater

Agricultural land

Forest vegetation

Other vegetation

Barren land

Landscape characteristics of UCIs MPS (ha)

NP

PLAND (%)

Mean temperature reduction (◦C) of UCIs −1.3

ote: Differences in the statistical results were produced in the transformation between v

tch density (PD), (c) aggregation index (AI), (d) mean patch shape index (Shape MN),HDI) of land use and temperature change (◦C).

4.2.2. Spatial pattern relationship between the greenspace coolislands (GCIs) and greenspace landscapes

GCIs were classified as forest vegetation based (forest vegeta-tion made up > 60%of each UCI) and open water area < 20% (basedon the analysis of temperature change and the proportion of eachLULC type) (Figs. 3c and 4b). A case-by-case examination was

then completed to select UCIs caused by street canyons or shadefrom buildings. During the patch analysis, the Yangtze River coolisland was divided to separate the cool island created by the largegreenspace associated with the Mufu Mountain from the rest. The

km2 %

e 8.70 6.3049.11 35.3913.08 9.4855.20 40.0010.35 7.50

1.55 1.13

17.18808

31.86

ector data and grid data.

42 F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47

F tempe(

PaaGGo−

Gavlboofttlsapprs(

((GtpenFga

ig. 6. (a) Image used to identify urban cool islands (UCIs) where the land-surface

GCIs) in the study area.

urple Mountain cool island patch was also divided into two, sep-rating the Xuanwu Lake and the Purple Mountain. Finally, thenalysis identified 154 individual GCIs (Fig. 6b). When individualCIs induced by the same patch of greenspace were combined, 116CI patches of were identified for further analysis. The total areaf the GCIs was 43.94 km2, the mean temperature reduction was0.7 ◦C, and the maximum cooling was −6.9 ◦C.

To understand better the spatial pattern relationship betweenCIs and the corresponding greenspaces an analysis was performedt the GCI class scale (Fig. 7). After sorting the data by each forest-egetation patch area (CA forest vegetation), a corresponding trendine of each GCI class area (CA GCI) was plotted to reveal trendsetween CA forest vegetation and CA GCI (Fig. 7A and a). The CAf GCI and forest vegetation have a strong relationship; and mostf the CA of GCI is smaller than that of forest vegetation, exceptor two GCI patches identified that have a larger area (CA) thanhat of the corresponding greenspace patches (Fig. 7A). This implieshat other factors may also influence the effect of vegetation onowering the LST. The UCI was caused by both greenspace and theurrounding water (Fig. 8a). Similar results were reported by Rinnernd Hussain (2011) and Saaroni and Ziv (2003); specifically, thatonds in a park contribute to cooling. A large area of a greenspaceatch may extend its cooling effect beyond its boundary, with theesult that its UCI connects with neighboring UCIs associated withmaller greenspace patches, enhancing the overall cooling effectsFig. 8b).

The scatter plot of the positive relationship between the NPNumber of Patches) of forest vegetation and GCI (R2 = 0.396)Fig. 7B and b, Table 1) shows that generally there were fewerCI patches than patches of forest vegetation. These data indicate

hat most of the GCIs were created from more than one greenspaceatch; that is, fragmented greenspace patches may cause a coolingffect when their spatial distribution is aggregated. Further exami-

ation found that this occurs when the GCI area increases (shown asigs. 7B and 8b), suggesting that the GCI areas around fragmentedreenspaces can contribute to the cooling effect of a larger GCI cre-ted by a larger greenspace patch. Several locations occur where

rature (LST) is less than the LST for the study area. (b) The greenspace cool islands

one greenspace patch creates different GCIs (Fig. 7B), and the inva-sion of the urbanized area and associated boundary effects canresult in complex urban greenspace shapes and fragmented GCIs(Fig. 8c). The results indicate that for a fixed amount of greenspace,the complex greenspace patch shapes caused by urbanization maylead to the cool islands becoming fragmented and the area of a GCIas well as its cooling effect would be decreased.

The area-weighted perimeter area ratio (PARA AM) of aGCI is significantly but only weakly (R2 = 0.169) correlatedto the area-weighted perimeter area ratio of forest vegeta-tion (PARA AM forestvegetation) (Fig. 7C and c, Table 1). Thearea-weighted mean radius of gyration (GYRATE AM) of forestvegetation shows a high correlation (R2 = 0.836) with that of GCIand most of the GYRATE AMof forest vegetation values are largerthan the corresponding area-weighted mean radius of gyration(GYRATE AM GCI) of GCI. This indicates that the extent and com-pact shape of a cool island patch are strongly correlated with thegreenspace patches (Fig. 7D and d, Table 1).

4.3. Characteristics of greenspace cool island intensity

The greenspace cool island intensity (the maximum tem-perature reduction) was defined as the difference between theminimum LST (minLST) of the cool island and the meanLSTof the study area. When the GCI intensity characteristics wereinvestigated, it was found that GCI intensity was not correlateddirectly with the maximum or minimum LST. Rather, the analy-sis shows that it is strongly correlated with the spatial extent ofthe greenspace cool island (GCI extent) (Fig. 9a). The statisticalanalysis also indicates that the GCI intensity has a highly corre-lated linear relationship with the mean background temperatureof the GCI, that is, the mean temperature reduction of each GCI(Fig. 9b). These results imply that context plays an important role

in determining GCI intensity. A cool island is the combined resultof the surrounding cooler area, and the cumulative cooling effectsof the surrounding greenspaces lead to the minimum LST area inthe cool island. This finding contradicts the traditional method of

F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47 43

F eenspm tion (

iDrba

5

vmtpu

ig. 7. Spatial pattern relationship between the greenspace cool islands (GCIs) and grean perimeter area ratio (PARA AM) (C, c) and area weighted mean radius of gyra

nvestigating factors such as the relationship between Normalizedifference Vegetation Index (NDVI) with LST at the pixel scale. The

esults suggest that previous research using correlation analysisetween NDVI with LST may need to consider cumulative effects at

scale larger than pixels.

. Discussion and conclusion

Numerous previous studies have revealed that vegetation pro-ides a cooling effect in urbanized areas, and strategies relating to

aintaining and that increasing vegetation cover in urban areas has

he potential to significantly cool urban environments. However,ast studies have either been on a small scale, providing individ-al examples of the application of cooling methods, or they have

aces indicated by class area (CA). (A, a), number of patches (NP) (B, b), area-weightedGYRATE AM) (D, d) metrics ranked by the CA of forest vegetation.

been attempts to numerically model the results of urban greeningto determine what would occur if such methods were adopted ona city or larger scale. Understanding the impact of vegetation onurban cooling at a city scale is difficult; no two cities are identi-cal in design or climate. It is, therefore, challenging to compare theeffect that greenspaces, especially forest areas, have on the climateof urban areas (Armson, 2012; Bowler et al., 2010). By understand-ing the ways in which vegetation cools the environment in anurban area, it may be possible to maximize the cooling potentialof greenspaces and to incorporate these features in urban planning

for the express purpose of reducing the UHI effect.

However, there is a lack of published material on the appli-cation and sensitivity analysis of landscape metrics with respectto the relationship between greenspace patterns and their cooling

44 F. Kong et al. / Landscape and Urban Planning 128 (2014) 35–47

F is largI ded cou g isla

eotgshdgtl

eLsweaifoprttMsaeao

F(

ig. 8. Class-scale perspective on the locations where class area of GCI (CA GCI)

nfluenced by the surrounding water, (b) influenced by the nearby water and extenrban greenspace shape caused by urbanization, and resulting in fragmented coolin

ffects. In this study, using landscape metrics, a method was devel-ped to identify the scaling effect on the relationship betweenhe greenspace patterns and their cooling effect. We explored thereenspace cooling effect at multiple scales, rather than relying oningle case studies of individual green areas. This approach willelp land managers understand how urban greenspace should beesigned in terms of the type, size, and spatial arrangement ofreenspace across an urban area. We also confirmed that the meanemperature is significantly lower for greenspaces and highest forand use areas with impervious surfaces (Table 3).

Most previous studies have confirmed that there is a scalingffect on the relationship between urban land-use patterns andSTs (Li et al., 2012; Liu & Weng, 2009). For this study area, a fixedcale was selected with results suggesting that a 240 m × 240 mindow size is the optimal spatial scale to examine the cooling

ffects of greenspaces. An increase of 10% forest vegetation within 240 m × 240 m area resulted in a 0.83 ◦C decrease in temperature,ndicating that the creation of greenspace with approximately 60%orest vegetation coverage could create a cooling island capablef offsetting to the average UHI temperature increase currentlyresent in the study area. Sample analysis, including bivariateegression, indicates that at this fixed window size scale the MPS,he AI, and the PARA AM have a significant relationship withemperature reduction. In particular, the strong relationship with

PS, the binomial relationship with AI, and the weak relation-hip with the PD metric, all imply that with sufficient greenspace

reas, smaller green areas with a fragmented spatial pattern areffective for cooling. We also found that the relationship of temper-ture reduction with the SHDI index is not consistent with resultsf previous research that found a constant positive or negative

ig. 9. Correlation analysis between the greenspace cool island (GCI) intensity (◦C) (mLnGCI extent) (a, X axis) as well as mean temperature reduction of each GCI (◦C) (b, X ax

er or smaller than the class area of forest vegetation (CA forest vegetation). (a)oling effect caused by the larger greenspace patch (Purple Mountain), (c) complex

nds.

effect on the LST (Connors et al., 2012; Li et al., 2011), and theanalysis indicates the SHDI is a much more scale-sensitive metricand conclusions based on SHDI need to be treated with cau-tion.

In this study, the cool islands were extracted and representedas object in ArcMap. This allowed for the calculation of severalfeatures related to the cool islands taken from the original LSTmaps (for example, the landscape pattern, the areal extent, and themaximum, mean, and minimum cooling temperatures). A spatialpattern relationship between the GCIs and greenspace landscapeswas made at the class level. There is a strong relationship betweenthe area of GCI and forest vegetation, with the area of forest vege-tation usually being larger than the corresponding area of GCI. Thismay, in part, be caused by the impact of the designation of RLST,which will define the extent of qualifying GCIs. Alternatively, theboundary effect of urbanized land around greenspaces, by increas-ing the LST, may account for the relationship. Another interestingresult is that two GCI patches were identified with larger areasthan their corresponding greenspace patches, one the result ofnearby water bodies, and the other resulting from the contribu-tion to fragmented patches of a nearby larger greenspace. A largearea of a greenspace patch may extend its cooling effect beyondits boundary and enhance the cooling effects of the nearby smallergreenspaces. We defined this as the mainland-island spatial pat-tern and the use of multiple small greenspaces appears to be a goodurban greenspace planning strategy designed to improve the urban

thermal environment. The class-level analysis also shows that theinvasion of urbanized areas into rural landscapes and the associatedboundary effect can result in complex urban greenspace shapes andcause fragmented GCIs.

aximum temperature reduction) (a and b, Y axis) and the cooling island areais).

Urban

cc2srsecctiabc

tliUttetstm

aicaitete&elFoat

tptmatrmGpiigHo

A

e

F. Kong et al. / Landscape and

No universally accepted method exists to define exactly what aool island is or to define the intensity of the effects of greenspaceool islands in urban landscapes (Cao et al., 2010; Chang et al.,007). In this study, the effectiveness of a GCI (here called its “inten-ity”), the relationship of GCI intensity with the mean temperatureeduction, and the extent of GCIs were analyzed at a patch-levelcale. A strong relationship between GCI intensity with its spatialxtent and mean temperature reduction implies that it can indi-ate the cool island characteristics well. The analysis revealed thatooler pixels can be combined into a cool island. Such a cumula-ive cooling effect results in a sink area associated with the coolsland. The results also implies that previous studies focused onnalysis of the relationship between NDVI and surface temperatureased on the pixel scale may need to be revisited by considering theumulative cooling effect at the landscape scale.

Although apparently conclusive, the results of this research onhe cooling effects of urban forest vegetation at multiple scales haveimitations and further research is warranted. One important points that the choice of the RLST will influence the spatial extent of aCI. When increasing the threshold (for example, using the mean

emperature of an impervious surface as the RLST, rather than usinghe mean temperature of the study area to define the RLST), the UCIxtent will become larger and vice versa. If the mean temperature ofhe study area is used as the RLST, such as in areas of the greenspacesurrounded by a higher thermal environment that have a higherhan average temperature compared to that of the study area, this

ay prevent the development of a GCI.Future research on the cooling effect of forest vegetation as well

s cool island characteristics and impact factors should also exam-ne the influence of the shade of buildings and the effect of streetanyons on the urban thermal environment and microclimate. Inddition, the greenspace areas were calculated without consider-ng the impact of the topography of the mountainous area withinhis study area, which may have led to underestimating its coolingffect. In a greenspace with trees, the cooling effect is determined,o a large degree, by the amount of canopy shading. Different lev-ls of shading will produce different cooling effects (Shashua-Bar

Hoffman, 2000). Furthermore, on-site investigation on the influ-nce of the sky view factor, the tree crown structure, and the actualeaf area index needs to be analyzed through empirical studies.inally, this research was conducted during the one summer forne city; the results may differ for other cities or for the same cityt a different time. Thus, further research is needed to compare tohe study’s conclusions and recommendations.

This study offers insights into the scientific understanding ofhe cooling effects of greenspaces in urban landscapes. The resultsrovide urban planners and natural resource managers with impor-ant theoretical and practical information about the planning and

anagement of urban greenspaces, which they can use to adapt tond mitigate the impacts of UHIs that will continue to develop ashe climate changes. Furthermore, incorporating thermal infraredemote sensing into landscape studies allows better measure-ent of the linkages of patterns and processes on multiple scales.reenspace cooling effects exists, and decision makers and urbanlanners need to consider this when thinking about urban adapt-

on to global climate change. With rapid urban expansion, sprawlnvades greenspace and causes their fragmentation into hetero-eneous shapes with irregular boundaries and even complete loss.ence, understanding and applying the mechanisms and dynamicsf greenspace cooling is becoming increasingly important.

cknowledgements

The research was supported by the National Natural Sci-nce Foundation of China (No. 31170444). Special thanks to the

Planning 128 (2014) 35–47 45

anonymous reviewers and the editor for their valuable commentsto improve our manuscript.

References

Alghannam, A. R., & Al-Qahtnai, M. R. (2012). Impact of vegetation cover on urbanand rural areas of arid climates. Australian Journal of Agricultural Engineering,3(1), 1–5.

Amiri, R., Weng, Q. H., Alimohammadi, A., & Alavipanah, S. K. (2009).Spatial–temporal dynamics of land surface temperature in relation to fractionalvegetation cover and land use/cover in the Tabriz urban area, Iran. RemoteSensing of Environment, 113, 2606–2617.

Armson, D. (2012). The effect of trees and grass on the thermal and hydrological perfor-mance of an urban area Thesis. Manchester, UK: The University of Manchester.

Bowler, D. E., Buyung-Ali, L., Knight, T. M., & Pullin, A. S. (2010). Urban greening tocool towns and cities: A systematic review of the empirical evidence. Landscapeand Urban Planning, 97, 147–155.

Buyantuyev, A., & Wu, J. G. (2010). Urban heat islands and landscape heterogeneity:Linking spatio-temporal variations in surface temperatures to land-cover andsocioeconomic patterns. Landscape Ecology, 25, 17–33.

Cao, X., Onishi, A., Chen, J., & Imura, H. (2010). Quantifying the cool island intensityof urban parks using ASTER and IKONOS data. Landscape and Urban Planning, 96,224–231.

Carlson, T. N., & Arthur, S. T. (2000). The impact of land use—land cover changes dueto urbanization on surface microclimate and hydrology: A satellite perspective.Global and Planet Change, 25, 49–65.

Chang, C. R., Li, M. H., & Chang, S. D. (2007). A preliminary study on the local cool-island intensity of Taipei city parks. Landscape Urban Planning, 80, 386–395.

Chen, Y., & Wong, N. H. (2006). Thermal benefits of city parks. Energy and Buildings,38, 105–120.

Connors, J. P., Galletti, C. S., & Chow, W. T. L. (2012). Landscape configuration andurban heat island effects: Assessing the relationship between landscape char-acteristics and land surface temperature in Phoenix, Arizona. Landscape Ecology,28, 271–283.

Dimoudi, A., & Nikolopoulou, M. (2003). Vegetation in the urban environment:Microclimatic analysis and benefits. Energy and Buildings, 35, 69–76.

Eliasson, I. (2000). The use of climate knowledge in urban planning. Landscape andUrban Planning, 48, 31–44.

Fahmy, M., Sharples, S., & Yahia, M. (2010). LAI based trees selection for mid lati-tude urban developments: A microclimatic study in Cairo, Egypt. Building andEnvironment, 45, 34–57.

Gallo, K. P., McNab, A. L., Karl, T. R., Brown, J. F., Hood, J. J., & Tarpley, J. D. (1993).The use of a vegetation index for assessment of the urban heat island effect.International Journal of Remote Sensing, 14, 2223–2230.

Georgi, J. N., & Dimitriou, D. (2010). The contribution of urban green spaces to theimprovement of environment in cities: Case Study of Chania, Greece. Buildingand Environment, 45, 1401–1414.

Gill, S., Handley, J., Ennos, R., & Pauleit, S. (2007). Adapting cities for climate change:The role of the green infrastructure. Building and Environment, 30, 97–115.

Giridharan, R., Lau, S. S. Y., Ganesan, S., & Givoni, B. (2008). Lowering the outdoortemperature in high-rise high-density residential developments of coastal HongKong: The vegetation influence. Building and Environment, 43, 1583–2159.

Givoni, B. (1991). Impact of planted areas on urban environmental quality: A review.Atmospheric Environment, 25, 289–299.

Gobakis, K., Kolokotsa, D., Synnefa, A., Saliari, M., Giannopoulou, K., & Santamouris,M. (2011). Development of a model for urban heat island prediction using neuralnetwork techniques. Sustainable Cities and Society, 1(2), 104–115.

Hamada, S., & Ohta, T. (2010). Seasonal variations in the cooling effect of urban greenareas on surrounding urban areas. Urban Forestry & Urban Greening, 9, 15–24.

Hemiddi, N. A. (1991). Measurement of surface and air temperature over sites withdifferent land treatments. In Proceeding of passive and low energy conferenceSpain.

Honjo, T., & Takakura, T. (1990). Simulation of thermal effects of urban greenareas on their surrounding areas. Journal of Energy and Buildings, 15–16,443–446.

Hulme, M., Jenkins, G., Lu, X., Turnpenny, J., Mitchell, T., Jones, R., et al. (2002). Climatechange scenarios for the United Kingdom: The UKCIP02 scientific report. Norwich:Tyndall Centre for Climate Change Research, School of Environmental Sciences,University of East Anglia.

Jauregui, E. (1990). Influence of a large urban park on temperature and convec-tive precipitation in a tropical city. Journal of Energy and Buildings, 15–16,457–463.

Jenerette, G. D., Harlan, S. L., Brazel, A., Jones, N., Larsen, L., & Stefanov, W. L.(2007). Regional relationships between surface temperature, vegetation, andhuman settlement in a rapidly urbanizing ecosystem. Landscape Ecology, 22,353–365.

Jim, C. Y., & Chen, S. S. (2003). Comprehensive green space planning based on land-scape ecology principles in compact Nanjing city, China. Landscape and UrbanPlanning, 65, 95–116.

Jonsson, P. (2004). Vegetation as an urban climate control in the subtropical city ofGaborone, Botswana. International Journal of Climatology, 24, 1307–1322.

Jusuf, S. K., Wong, N. H., Hagen, E., Anggoro, R., & Yan, H. (2007). The influenceof land use on the urban heat island in Singapore. Habitat International, 31,232–242.

4 Urban

K

K

K

K

K

K

K

L

L

L

L

L

M

M

M

M

M

N

N

N

N

N

O

O

O

P

P

P

Q

Q

6 F. Kong et al. / Landscape and

atayama, T., Ishii, A., Hayashi, T., & Tsutsumi, J. I. (1993). Field surveys on cool-ing effects of vegetation in an urban area. Journal of Thermal Biology, 18,571–576.

awashima, S. (1994). Relation between vegetation, surface temperature, andsurface composition in the Tokyo region during winter. Remote Sensing of Envi-ronment, 50, 52–60.

eramitsoglou, I., Kiranoudis, C. T., Ceriola, G., Weng, Q., & Rajasekard, U.(2011). Identification and analysis of urban surface temperature patterns inGreater Athens, Greece, using MODIS imagery. Remote Sensing of Environment,115(3080–3090).

im, H. H. (1992). Urban heat island. International Journal of Remote Sensing, 13(12),2319–2336.

ong, F., & Nakagoshi, N. (2006). Spatial–temporal gradient analysis of urban greenspaces in Jinan, China. Landscape and Urban Planning, 78, 147–164.

ong, F., Yin, H., & Nakagoshi, N. (2007). Using GIS and landscape metrics in thehedonic price modeling of the amenity value of urban green space: A case studyin Jinan City, China. Landscape and Urban Planning, 79, 240–252.

ong, F., Yin, H., Nakagoshi, N., & Zong, Y. (2010). Urban green space network devel-opment for biodiversity conservation: Identification based on graph theory andgravity modeling. Landscape and Urban Planning, 95, 16–27.

i, J., Song, C., Cao, L., Zhu, F., Meng, X., & Wu, J. (2011). Impacts of landscape structureon surface urban heat islands: A case study of Shanghai, China. Remote Sensingof Environment, 115(12), 3249–3263.

i, X. M., Zhou, W. Q., Ouyang, Z. Y., Xu, W. H., & Zheng, H. (2012). Spatialpattern of green space affects land surface temperature: Evidence from theheavily urbanized Beijing metropolitan area, China. Landscape Ecology, 27,887–898.

iu, H., & Weng, Q. (2009). Scaling-up effect on the relationship between landscapepattern and land surface temperature. Photogrammetric Engineering and RemoteSensing, 75(3), 291–304.

u, D., & Weng, Q. (2006). Spectral mixture analysis of ASTER imagery for exam-ining the relationship between thermal features and biophysical descriptors inIndianapolis, Indiana. Remote Sensing of Environment, 104(2), 157–167.

uber, G., & McGeehin, M. (2008). Climate change and extreme heat events. AmericanJournal of Preventive Medicine, 35, 429–435.

ackey, Ch. W., Lee, X., & Smith, R. B. (2012). Remotely sensing the cooling effectsof city scale efforts to reduce urban heat island. Building and Environment, 49,348–358.

cGarigal, K., & Cushman, S. A. (2002). The gradient concept of landscape structure:Or, why are there so many patches. Available at: http://www.umass.edu/landeco/pubs/pubs.html

cGarigal, K., Ene, E., & Holmes, C. (2002). FRAGSTATS (Version 3): FRAGSTATS metrics.University of Massachusetts-Produced Program. Available at: http://www.umass.edu/landeco/research/fragstats/documents/fragstats documents.html

iao, Q. L., Pan, W. Z., & Xu, X. Z. (2008). Characteristic analysis of summer tem-perature in Nanjing during 56 years. Journal of Tropical Meteorology, 24(6),737–742.

ihalakakou, G., Santamouris, M., Papanikolaou, N., Cartalis, C., & Tsangrassoulis,A. (2004). Simulation of the urban heat island phenomenon in Mediterraneanclimates. Pure and Applied Geophysics, 161, 429–451.

anjing Municipal Bureau Statistics. (2011). Nanjing statistical yearbook 2010. Avail-able at: http://www.njtj.gov.cn/2004/2011/index.htm (in Chinese)

arita, K., Mikami, T., Sugawara, H., Honjo, T., Kimura, K., & Kuwata, N. (2004). Coolisland and cold air-seeping phenomena in an urban park, Shinjuku Gyoen, Tokyo.Geographical Review of Japan, 77(6), 403–420.

g, E., Chen, L., Wang, Y., & Yuan, C. (2012). A study on the cooling effects of greeningin a high-density city: An experience from Hong Kong. Building and Environment,47, 256–271.

ichol, J. E. (1998). Visualization of urban surface temperature derived from satelliteimages. International Journal of Remote Sensing, 19(9), 1639–1649.

ichol, J. E. (2005). Remote sensing of urban heat islands by day and night. Pho-togrammetric Engineering and Remote Sensing, 71, 613–623.

ke, T. R., Crowther, J. M., McNaughton, K. G., Monteith, J. L., & Gardiner, B. (1989).The micrometerology of the urban forest. Philosophical Transactions of the RoyalSociety B, 324, 335–349.

liveira, S., Andrade, H., & Vaz, T. (2011). The cooling effect of green spaces as acontribution to the mitigation of urban heat: A case study in Lisbon. Buildingand Environment, 46, 2186–2194.

wen, T. W., Carlson, T. N., & Gillies, R. R. (1998). An assessment of satellite remotely-sensed land cover parameters in quantitatively describing the climatic effect ofurbanization. International Journal of Remote Sensing, 19, 663–1681.

otchter, O., Cohen, P., & Bitan, A. (2006). Climatic behavior of various urban parksduring hot and humid summer in the Mediterranean city of Tel Aviv, Israel.International Journal of Climatology, 26, 1695–1711.

rice, J. C. (1979). Assessment of the urban heat island effect through the use ofsatellite data. Monthly Weather Review, 107, 1554–1557.

rice, J. C. (1983). Estimating surface temperature from satellite thermal infrareddata – A simple formulation for the atmospheric effect. Remote Sensing of Envi-ronment, 13, 353–361.

in, Z., Karnieli, A., & Berliner, P. (2001). A mono-window algorithm for retriev-ing land surface temperature from Landsat TM data and its application to

the Israel–Egypt border region. International Journal of Remote Sensing, 22(18),3719–3746.

uattrochi, D. A., & Luvall, J. C. (1999). Thermal infrared remote sensing for analysisof landscape ecological processes: Methods and applications. Landscape Ecology,14, 577–598.

Planning 128 (2014) 35–47

Quattrochi, D. A., & Ridd, M. K. (1998). Analysis of vegetation within a semi-aridurban environment using high spatial resolution airborne thermal infraredremote sensing data. Atmospheric Environment, 32(1), 19–33.

Rao, P. K. (1972). Remote sensing of urban “heat islands” from an environmentalsatellite. Bulletin of the American Meteorological Society, 53, 647–648.

Rinner, C., & Hussain, M. (2011). Toronto’s urban heat island-exploring the rela-tionship between land use and surface temperature. Remote Sensing, 3(6),1251–1265.

Rosenfeld, A. H., Akbari, H., Bretz, S., Fishman, B. L., Kurn, D. M., Sailor, D., et al. (1995).Mitigation of urban heat islands: Materials, utility programs, updates. Journal ofEnergy and Buildings, 22, 255–265.

Rosenfeld, A. H., Romm, G., Akbari, H., Romm, J. J., & Pomerantz, M. (1998). Coolcommunities: Strategies for heat island mitigation and smog reduction. Journalof Energy and Buildings, 28, 51–62.

Saaroni, H., & Ziv, B. (2003). The impact of a small lake on heat stress in aMediterranean urban park: The case of Tel Aviv, Israel. International Journal ofBiometeorology, 47, 156–165.

Saito, I., Ishihara, O., & Katayama, T. (1990–1991). Study of the effect of green areason the thermal environment in an urban area. Journal of Energy and Buildings,15(16), 443–446.

Schwarz, N., Lautenbach, S., & Seppelt, R. (2011). Exploring indicators for quanti-fying surface urban heat islands of European cities with MODIS land surfacetemperatures. Remote Sensing of Environment, 115, 3175–3186.

Semrau, A. (1992). Introducing cool communities. American Forests, July/August,49–52.

Shashua-Bar, L., & Hoffman, M. E. (2000). Vegetation as a climatic component in thedesign of an urban street: An empirical model for predicting the cooling effectof urban green areas with trees. Energy and Buildings, 31, 221–235.

Small, C. (2006). Comparative analysis of urban reflectance and surface temperature.Remote Sensing of Environment, 104, 168–189.

Sobrino, J. A., Raissouni, N., & Li, Z. L. (2001). A comparative study of land sur-face emissivity retrieval from NOAA data. Remote Sensing of Environment, 75,256–266.

Sobrino, J. A., Jiménez–Munoz, J. C., & Paolini, L. (2004). Land surface tem-perature retrieval from LANDSAT TM 5. Remote Sensing of Environment, 90,434–440.

Spronken-Smith, R. A., & Oke, T. R. (1998). The thermal regime of urban parks in twocities with different summer climates. International Journal of Remote Sensing,19, 2085–2104.

Stathopoulou, M., & Cartalis, C. (2007). Use of satellite remote sensing in sup-port of urban heat island studies. Advances in Building Energy Research, 1,203–212.

Stott, P. A., Stone, D. A., & Allen, M. R. (2004). Human contribution to the Europeanheat wave of 2003. Nature, 432, 610–614.

Tan, J., Zheng, Y., Song, G., Kalkstein, L. S., Kalkstein, A. J., & Tang, X. (2007). Heatwave impacts on mortality in Shanghai, 1998 and 2003. International Journal ofBiometeorology, 51, 193–200.

Tran, H., Uchihama, D., Ochi, S., & Yasuoka, Y. (2006). Assessment with satellite dataof the urban heat island effects in Asian mega cities. International Journal ofApplied Earth Observation and Geoinformation, 8, 34–48.

Tyrväinen, L., Pauleit, S., Seeland, K., & de Vries, S. (2005). Benefits and uses of urbanforests and trees. In C. C. Konijnendijk, K. Nilsson, T. B. Randrup, & J. Schipperijn(Eds.), Urban forests and trees (pp. 81–114). Berlin: Springer.

Upmanis, H., Eliasson, I., & Lindqvist, S. (1998). The influence of green areas on noc-turnal temperatures in a high latitude city (Göteborg, Sweden). InternationalJournal of Climatology, 18, 681–700.

Voogt, J. A., & Oke, T. R. (2003). Thermal remote sensing of urban climate. RemoteSensing of Environment, 86, 370–384.

Weng, Q. (2009). Thermal infrared remote sensing for urban climate andenvironmental studies: Methods, applications, and trends. ISPRS Journal of Pho-togrammetry and Remote Sensing, 64, 335–344.

Weng, Q., & Larson, R. C. (2005). Satellite remote sensing of urban heat islands:Current practice and prospects. In R. R. Jensen, J. D. Gatrell, & D. D. McLean (Eds.),Geo-spatial technologies in urban environments (pp. 91–111). Berlin: Springer.

Weng, Q., & Lu, D. (2008). A sub-pixel analysis of urbanization effect on land surfacetemperature and its interplay with impervious surface and vegetation coveragein Indianapolis, United States. International Journal of Applied Earth Observationand Geoinformation, 10(1), 68–83.

Weng, Q., Lu, D., & Schubring, J. (2004). Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies. Remote Sensingof Environment, 89, 467–483.

Weng, Q., Lu, D., & Liang, B. (2006). Urban surface biophysical descriptors and landsurface temperature variations. Photogrammetric Engineering & Remote Sensing,72(11), 1275–1286.

Weng, Q., Liu, H., & Lu, D. (2007). Assessing the effects of land use and land coverpatterns on thermal conditions using landscape metrics in city of Indianapolis,United States. Urban Ecosystem, 10(2), 203–219.

Wilson, J. S., Clay, M., Martin, E., Stuckey, D., & Vedder-Risch, K. (2003). Evaluatingenvironmental influences of zoning in urban ecosystems with remote sensing.Remote Sensing of Environment, 86(3), 303–321.

Wong, N. H., Jusuf, S. K., La-Win, A. A., Thu, H. K., Negara, T. S., & Wu, X. C. (2007).

Environmental study of the impact of greenery in an institutional campus in thetropics. Building and Environment, 42, 2949–2970.

Xian, G., & Crane, M. (2006). An analysis of urban thermal characteristics and associ-ated land cover in Tampa Bay and Las Vegas using Landsat satellite data. RemoteSensing of Environment, 104, 147–156.

Urban

X

Y

F. Kong et al. / Landscape and

iao, R., Weng, Q., Ouyang, Z., Li, W., Schienke, E. W., & Zhang, W. (2008). Land sur-

face temperature variation and major factors in Beijing, China. PhotogrammetricEngineering & Remote Sensing, 74(4), 451–461.

uan, F., & Bauer, M. E. (2007). Comparison of impervious surface area and normal-ized difference vegetation index as indicators of surface urban heat island effectsin Landsat imagery. Remote Sensing of Environment, 106, 375–386.

Planning 128 (2014) 35–47 47

Zhou, W., Huang, G., Mary, L., & Cadenasso, L. (2011). Does spatial configuration mat-

ter? Understanding the effects of land cover pattern on land surface temperaturein urban landscapes. Landscape and Urban Planning, 102, 54–63.

Zoulia, I., Santamouris, M., & Dimoudi, A. (2009). Monitoring the effect of urban greenareas on the heat island in Athens. Environmental Monitoring and Assessment, 156,275–292.