This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 140.180.241.115

This content was downloaded on 09/03/2016 at 19:14

Please note that terms and conditions apply.

Contrasting impacts of urban forms on the future thermal environment: example of Beijing

metropolitan area

View the table of contents for this issue, or go to the journal homepage for more

2016 Environ. Res. Lett. 11 034018

(http://iopscience.iop.org/1748-9326/11/3/034018)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 11 (2016) 034018 doi:10.1088/1748-9326/11/3/034018

LETTER

Contrasting impacts of urban forms on the future thermalenvironment: example of Beijing metropolitan area

LongYang1,2,7, DevNiyogi1,3,Mukul Tewari4, Daniel Aliaga5, Fei Chen6, Fuqiang Tian2 andGuanghengNi2

1 Department of Earth, Atmospheric, and Planetary Sciences, PurdueUniversity,West Lafayette, Indiana 47907,USA2 Department ofHydraulic Engineering, TsinghuaUniversity, Beijing 100084, People’s Republic of China3 Department of Agronomy- Crops, Soils, Environmental Sciences, PurdueUniversity,West Lafayette, Indiana 47907,USA4 IBM, T JWatsonResearchCenter, YorktownHeights, NewYork 10598,USA5 Department of Computer Science, PurdueUniversity,West Lafayette, Indiana 47906,USA6 National Center for Atmospheric Research, Boulder, Colorado 80307,USA7 Current address: Department of Civil and Environmental Engineering, PrincetonUniversity, Princeton, New Jersey 08540,USA

E-mail: longyang@princeton.edu

Keywords: urbanization, urban form, climate change,mitigation and adaptation, UHI, Beijing

Supplementarymaterial for this article is available online

AbstractThis study investigated impacts of urban forms on the future thermal environment over Beijing, thecapital city of China. Beijing is experiencing remarkable urban expansion and is planned to undergothe transformation of urban forms from single-centric (compact-city) to poly-centric city (dispersed-city). Impacts of urban forms on the future thermal environment were compared and evaluated byconducting numerical experiments based on a regional atmosphericmodel coupledwith a single-layer urban canopymodel as well as future climate forcing output from a global climatemodel. Resultsshow that a dispersed city is efficient in reducingmean urban heat island intensity, but produces largerthermal loading and deeper thermal feedback at the regional scale compared to a compact city.Thermal comfort over downtown areas is reduced in compact-city scenario under future climateconditions. Future climate contributes almost 80%of the additional thermal loading over urban areas,with the remaining 20%contributed by urbanization (for both the compact-city and dispersed-cityscenarios). The thermal contrast between the two urban forms is dominated by the expected futureclimate change. This study leads to two complementary conclusions: (i) for developing assessmentsrelated to current climate comfort, urban formof the city is important; (ii) for assessing future climatechange impacts, the areal coverage of the city and urbanization extent emerges to bemore importantthan the details related to how the urbanizationwill evolve.

1. Introduction

Urbanization, characterized by the conversion ofnatural (e.g., vegetation) to artificial surfaces (withcontrasting physical and thermal properties, e.g., sur-face albedo, heat capacity) and boom in urban popula-tion, contributes to the deterioration of urban thermalenvironment. Urbanization impacts transcend farbeyond its physical boundary and could also affectsevere weather (Niyogi et al 2006, 2011), hydrologicalcycles (e.g., Shepherd 2005, Yang et al 2013a, 2013b),biodiversity (Seto et al 2012), air quality (Jacobson

et al 2015), energy consumption (Martilli 2014) andhumanhealth (Hondula et al 2014).

Transformation of urban forms is regarded as aneffective tool for alleviating negative impacts of urba-nization on the thermal environment (Lands-berg 1981, Oke 1984, Scherer et al 1999, Stone andRodgers 2001, Hathway and Sharples 2012, Schwarzet al 2013). Stone et al (2010) investigated correlationsbetween urban forms and the frequency of extremeheat events over selected US cities, and concluded thatincreasing rates for extreme heat events in sprawlingcities is twice as large as that in compact cities.

OPEN ACCESS

RECEIVED

12November 2015

REVISED

5 February 2016

ACCEPTED FOR PUBLICATION

16 February 2016

PUBLISHED

7March 2016

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2016 IOPPublishing Ltd

However, Adachi et al (2014) found that a compactcity has the potential tomoderate themean urban heatisland effect (i.e. urban–rural contrast of air temper-ature) over the entire Tokyo metropolitan area, butmight further increase total thermal loadings overdowntown Tokyo. Considering that changes in urbanstructures are varied in cities and regions, the effective-ness of urban forms in mitigating heat-related risksalso varies (Frolking et al 2013). For cities in develop-ing countries, transformation of urban forms is com-monly associated with the decrease of surroundingrural/vegetated surfaces (outward expansion); while itmay not be the case for most cities in developed coun-tries where available rural spaces are already limited.Cities in developed countries tend to increase verti-cally and in capacities within existing urban frames(upward expansion) (Adachi et al 2014). The expan-sion of urban land cover globally is projected to beaccelerated into 2050s, with most expansion takingplace in the developing world (Angel et al 2011). In thisstudy, we will focus on urban development in the Beij-ingmetropolitan area (BMA), China.

As one of the largest mega-cities in the world, Beij-ing has witnessed astonishing urbanization rate inrecent decades (Bureau 2012, Zhao et al 2015). Analy-sis of Landsat images from 1990 to 2005 indicates a7-fold increase in the urban footprint of the Beijingurban region (Miao 2015). Rapid expansion of metro-politan area will continue to play a vital role in thedevelopment of Beijing metropolitan area in future.The current urban population is 16 million over theentire BMA, and this number is projected to double by2050s (see figure S1 formore details). According to thecitymaster plan approved by theNational Council, theurban spatial structure for BMA will gradually shiftfrom a mono-centric (compact) city to a poly-centric(dispersed) city (more details about the city masterplan are available at http://zhengwu.beijing.gov.cn/ghxx/ztgh/). It is therefore an interesting opportunityto examine the impacts of the projected transforma-tion in urban forms on regional environment (urbanheat island, total thermal loading, etc). We aim to pro-vide references regarding the potential environmentalfeedbacks related to urban transformation for cityplanners and policymakers.

Another related aim of this study is to make pro-jections on relative contributions of urbanization andclimate change to future warming over BMA underthe compact versus dispersed urban forms. The histor-ical contribution rate to regional warming has beenexamined in previous studies (Zhou et al 2004, Renet al 2007). For instance, Ren et al (2008) derived thatwarming rate due to urbanization is about 0.11 °C perdecade over North China (including BMA). Zhouet al (2004) estimated that urbanization contributed0.05 °C per decade increase of mean near-surfacetemperature in southeast China. However, future pro-jection of warming rate due to urbanization is stilllacking for the BMA region.

2.Methodology

We designed two urban forms for future Beijingaccording to the master plan: compact-city anddispersed-city. The dispersed-city design correspondsto the poly-centric urban pattern in the master plan.For the two urban forms, we examine the future urbanwarming effect based on the Weather Research andForecasting model-Advanced Research Version(WRF) coupled with an explicit urban canopy model(Chen et al 2011). The model was driven by down-scaled future climate forcing scenarios as well as landuse scenarios.We focus on a heat wave case during July2010 over BMA. Daily Maximum air temperatureattained over 40 °C for a large sample of temperaturestations over BMA (seefigure S2 formore details).

2.1. Projections of future urban developmentWe made projections on future urban coverage overBMA based on the projected increase in urbanpopulation, assuming that the total urban coverageshould be at least doubled so as to accommodate thenew urban inhabitants (since logistic regressionmodelshow urban population is to be doubled in 2050s). Inthis study, we assumed urban coverage sprawls at theexpense of converting suburban/rural areas in thesurroundings, instead of increasing capacities withincurrent urban regions (e.g., shrinking green spaces,increasing heights/densities of buildings, etc), whichis typical especially for city growth in developingcountries.

Two contrasting urban forms were designed inthis study: compact- and dispersed-city. For compact-city, additional urban coverage was distributed aroundthe old urban core region (downtown Beijing, seefigure 1), which maintained a mono-centric urbanform (figure 1(c)); while for dispersed-city, the addi-tional urban coverage was distributed onto five sub-regions (including Changping, Fangshan, Daxing,Tongzhou and Shunyi) surrounding the old urbancore region, exhibiting a pattern of poly-centric urbanform (figure 1(d)) as designed in the master plan. Wenote that the additional urban coverage mainlybelongs to the category of high-density residential landuse (see section 2.2 for numerical representations ofland use in themodel).

2.2.Model descriptions and configurationsSimulations in this study were conducted using theAdvanced Research version of WRF, ARW V3.4(Skamarock et al 2008). We configured three one-waynested model domains. The inner domain (D03), withhorizontal spatial resolution of 1 km, was centeredover Beijing metropolitan region (figure 1(a)). TheWRF model was coupled with a Single-Layer UrbanCanopy Model (SLUCM), to capture the character-istics and dynamics of heat/moisture fluxes overurban areas (Chen et al 2011). A three-category urban

2

Environ. Res. Lett. 11 (2016) 034018

land use (representing low-intensity residential, high-intensity residential and industrial/commercial,respectively) was incorporated into the model forbetter representations of surface heterogeneity overurban areas. The three-category urban land use wasdeveloped based on remote sensing images of LandsatTM5 (see Hou 2012 for more details) and has beenused in previous urban climate studies (Hou et al 2012,Yang et al 2014). The default parameters in SLUCMwere used in this study (see Chen et al 2011 for moredetails). Other physical options were summarized intable S1. Previous studies have extensively examinedthe utility of current configurations of WRF/SLUCMmodel over BMA (Miao et al 2009, 2011, Zhanget al 2009, Yang et al 2014).

Uncertainty for the simulations could originatefrom the initial/boundary conditions for the model(Wu et al 2005).We conducted three ensembles drivenby different datasets, including JRA-55, GFS-FNL andERA-interim (see table S2). We validated the simula-tion results against corresponding 2 m air temperatureobservations. The mean bias of simulated 2 m airtemperature is −0.02 °C (with the standard deviationof 2.4 °C) with ERA-interim dataset, compared to

−1.6±2.7 °Cwith JRA-55, and−0.24±2.6 °CwithFNL (see figure S3 formore details). In addition, simu-lation results with ERA-interim exhibited better spa-tial distribution of temperature bias than the other twodatasets (see figure S4). Spatial-average temperatureshows smallest bias with ERA-interim data (see figureS5). Accordingly, we chose ERA-interim as initial andboundary conditions for WRF/SLUCM simulationsin the following analyses. Simulation period of themodeling system covers the entire period of heat wavecase (from 0000 UTC 1 July to 0000 UTC 7 July 2010).Results for the peak heat wave period (4–6 July 2010)were analyzed and presented, placing 00 UTC 1 July to00 UTC 4 July as the spin-up period. Time interval formodel output was one hour, so as to match up the fre-quency of observations.

2.3. Experiment designsWe analyzed changes in thermal environment underboth future climate change at 2050s and urbandevelopment (as represented by changes in urbanforms and size). Climate forcing corresponding to2050s was provided based on the Community EarthSystemModel (CESM, http://www2.cesm.ucar.edu/)

Figure 1. (a)Topographic features of Beijingmetropolitan region. Elevation is shaded.Grey area represents the urban core region (i.e.downtown region). Black dots are surfacemeteorological stations that are used in theWRFmodel validation. The black box is themost inner domain ofWRF simulations; (b) current land use/land cover of Beijingmetropolitan region, green dots represent five sub-regions of Beijingmetropolitan region, include: CP-Changping, FS-Fangshan, DX-Daxing, TZ-Tongzhou and SY-Shunyi; the blackline highlights the boundary of downtownBeijing; the yellow straight line represents a cross-section though themetropolitan region;(c) land use/land cover in compact-city scenario, additional urban coverage (consists of high density urban land use) is highlighted bycrosshatches; (d) same as (c) but for dispersed-city scenario.

3

Environ. Res. Lett. 11 (2016) 034018

under one future emission scenario which is Repre-sentative Concentration Pathways 8.5 (RCP8.5) (Riahiet al 2011). Climate forcing from CESM was further‘downscaled’ into a finer spatial and temporal resolu-tion based on the pseudo-global warming downscaling(PGW-DS) approach (see Kawase et al 2009, Laueret al 2013 for more details about the downscalingapproach). The downscaled forcing was then used toprovide initial/boundary conditions for WRF/SLUCM model under future climate scenarios, whileERA-interim reanalysis data was used for currentclimate scenarios.

In total, an ensemble of two climate forcings andthree urban scenarios (current urban coverage, com-pact- and dispersed-city) were examined, resulting insix numerical experiments (table 1). We focus espe-cially on the contrasting behaviors between two urbanscenarios under future climate conditions, and thesimulation results based on ctl_fut, com_fut and dis_-fut runs will be analyzed and presented in the follow-ing section. Simulation results from other numericalexperiments (e.g., com_curr, dis_curr) will be mainlyused to determine relative contributions of future cli-mate and urban forms to the changes in thermalenvironment over BMA.

3. Results

3.1. Contrasting impacts on thermal environmentWe analyzed contrasting impacts on thermal environ-ment mainly from three perspectives: (i) thermalcontrast between urban and surrounding rural regionsthat is the urban heat island intensity, (ii) total thermalloading (as represented by air temperature and heatstress index) at downtown and broader regional scale,and (iii) the vertical extent of the thermalperturbation.

3.1.1. Urban heat island intensity (UHII)We define UHII as the difference between 2 m airtemperature averaged over urban region and that oversurrounding rural region (with a similar spatial scalewith urban region) (Oke 1982). Figure 2(a) shows timeseries of UHII for three future scenarios (ctl_fut,com_fut and dis_fut). Relative differences betweentwo future urban scenarios (R5 and R6) and currenturban scenario (R4) are shown in figure 2(b). TheUHII values range from 1 °C to 4 °C for the three

simulations. There is a notable contrast between thesimulations close to sunrise and sunset. ThemaximumUHII (about 4 °C) under the heat-wave period iscomparable but relatively larger in this study com-pared to previous studies (with a range of 2∼3 °C,not specifically focus on heat-wave periods) (Ryu andBaik 2012, Zhou et al 2013). The increasedUHII underthe heat-wave period could possibly be attributed tothe synergistic effects between heat wave and urbanheat island as discussed for instance in Li et al [2015].

The thermal contrast evolves quite differently forthe two urban forms. Compact-city scenario furtherenhances UHII by 0.1 °C (averaged over the peak heatwave period) compared to current urban scenario;while UHII is reduced by 0.4 °C on average in dis-persed-city scenario. The anomalies of UHII inducedby two different urban forms (with respect to the con-trol scenario) are statistically different at the sig-nificance level of 1% based on the student’s t-test.Contrast inUHI shows that dispersed-city ismore effi-cient inmoderating summertime heat island phenom-ena than compact-city. We note that both the urbanforms are spread over an area of total urban coveragetwice as large as the current urban scenario. The ther-mal contrast between urban and surrounding ruralregions could be directly related to the transforma-tions of surrounding rural areas (mainly in the formsof crops, grassland and other vegetated surfaces) intourban coverages (paved road, roofs, etc) under futureurban development.

3.1.2. Thermal loading at downtown and at the regionalscaleWe evaluated impacts of urban forms on thermalloading as represented by both 2 m air temperatureand heat stress index, expecting to provide a broaderimplication for meteorological perspectives as well ashuman health. The heat stress index is based on ameasure of apparent temperature that reflects bothtemperature and humidity (Steadman 1984), and takesthe formof:

A T e1.3 0.92 2.2 1( )= - + +

where T is 2 m air temperature (°C) and e is watervapor pressure (kPa) at 2 m. BothT and e are simulatedby the WRF/SLUCM modeling system. Temporalaverage (over the peak wave period of 4th–6th July2010) air temperature and heat stress index areaveraged at different spatial scales.

Table 1.Experimental design.

No. Experiment name Urban scenarios Initial/boundary conditions

R1 ctl_curr current urban ERA-interim in 2010

R2 com_curr compact-city ERA-interim in 2010

R3 dis_curr dispersed-city ERA-interim in 2010

R4 ctl_fut current urban CESM in 2050s

R5 com_fut compact-city CESM in 2050s

R6 dis_fut dispersed-city CESM in 2050s

4

Environ. Res. Lett. 11 (2016) 034018

Both changes of air temperature and heat stressindex fit a quadratic function of spatial extent(figure 3). In this study, the spatial extent is repre-sented by ratios of the buffering area surroundingurban coverage to that of original urban core region(downtown region, see figure 1(b)). For instance, arearatio equal to one means spatial scale covering

downtown area only, while area ratio equal to 2meanstwice as large as downtown coverage (figures 1(c) and(d)) (see also Zhou et al 2015).

Both air temperature and heat stress index decayquadratically from downtown area towards surround-ing rural area (figures 3(a) and (b)). The decaying trendof thermal loading has been recorded in previous

Figure 2.Time series of (a)UHI intensity (°C) for control, compact-city and dispersed-city scenarios under the 2050s climate and (b)difference of UHI intensity between compact-city and control scenario (com_futminus ctl_fut) aswell as dispersed-city and controlscenario (dis_futminus ctl_fut).

Figure 3.Horizontal spatially averaged (a) 2 m air temperature and (b) heat stress index over and beyond the extent of urban coverageunder future climate condition.X-axis is the horizontal extent. The differences between compact- and dispersed-city scenario areshown in (c) and (d) for 2 m temperature and heat stress index, respectively. Dots are based on results fromWRF simulations. Linesshown in (a) and (b) arefitted using power-law functions. The legend of (a) is the same as (b), and also the legend of (c) is the same as(d).

5

Environ. Res. Lett. 11 (2016) 034018

studies based on either in situ or remote sensing obser-vations (Jin et al 2005, Zhou et al 2015). In this study,we are particularly interested in the contrasting spatialdecaying pattern between two urban forms of com-pact- and dispersed-city. Because of the increasedurban area, it is expected that both compact- and dis-persed-city produce larger thermal loading at all spa-tial scales than current urban scenario. However, thecontrast between compact- and dispersed-city scenar-ios is also noticeable even though both scenarios areover twice the urban coverage compared to the currenturban. Dispersed city produces a larger regionalwarming effect (beyond downtown area, corresp-onding to area ratios larger than 1), with 2 m air temp-erature and heat stress index higher by 0.25 °C and0.15 °C than compact-city scenario, respectively. Thecontrast between the two scenarios gradually decayswith the increased spatial extent (figures 3(b) and (d)).There is no noticeable difference between two urbanforms at spatial scales of 7∼8 times of downtownarea, which indicates that different perturbations onthermal loading induced by urban forms are confinedwithin certain spatial scale (around 40 km way fromthe city center).

Unlike the impacts on thermal loading at spatialscales beyond downtown region, compact city pro-duce a larger warming effect than dispersed city, withair temperature and heat stress index larger by 0.2 °Cand 0.4 °C as compared to the dispersed city respec-tively (figures 3(c) and (d)). Larger thermal loading inthe downtown area indicate deteriorated urbanenvironment for city inhabitants with higher potentialof health-related risks under future climate (Hondulaet al 2014). We note that heat stress index will exceed35 °C in downtown Beijing for all three future scenar-ios. The minimum spatial-average heat stress index isabove 33 °C, which is even larger than maximum heat

stress index under current climate conditions (seefigure S6 for more details). Sherwood and Huber(2010) put forth that any prolonged exposure beyond35 °C could induce hyperthermia in humans andother mammals due to the impossibility of dissipationof metabolic heat. The total amount of emergencydepartment visits are 10% more in Beijing during the2010 heat wave period than the same historical period(Liu et al 2014).

3.1.3. Thermal perturbation in vertical extentFigure 4 shows the relative differences of potentialtemperature profiles (averaged over the entire periodof 4–6 July 2010) for compact-city/dispersed-cityscenarios to a control scenario. The vertical impactdecreases at higher altitudes for both the urban forms.However, the vertical extents of thermal impact arecontrasted in two scenarios. A dispersed city producesa deeper perturbation than compact city. Morespecific, the potential temperature difference reaches0.08 °C at the height of 250 m for the compact-cityscenario, while the same perturbation as 0.08 °C waspushed to a higher altitude of around 400 m for thedispersed-city scenario. The perturbation on potentialtemperature is not confined within vertical extentsover the additional urban coverage for compact- anddispersed-city scenarios, but could extend to a largerhorizontal extent covering the entire metropolitanregion.

The deeper thermal influence noted in the verticalprofile of potential temperature in dispersed-city sce-nario highlights a higher potential for increasedinstability leading to higher turbulent kinetic energyand planetary boundary layer height, which mightprovide favorable conditions for vertical mixing ofurban pollutants and influence assessment of air qual-ity over urban regions (Georgescu 2015).

Figure 4.Differences of potential temperature profiles along the cross section (shown as the yellow straight line in figure 1(b)): (a)compact-city (com_futminus ctl_fut) (b) dispersed-city (dis_futminus ctl_fut). The downtown area is shaded in grey.

6

Environ. Res. Lett. 11 (2016) 034018

The contrasting impacts for the two urban formsare illustrated in figure 5. Both compact- and dis-persed-city scenario increase thermal loading underfuture climate, but are contrasted by the magnitudesand spatial extents. The increased thermal loading ismore concentrated over the urban core region for thecompact-city scenario, while for the dispersed-cityscenario the increased thermal loadings are relativelyspread out over the region. The contrasting impacts ofurban forms on thermal environment are not restric-ted within downtown area or urban coverage, butextend into a larger spatial scale both horizontally andvertically beyond the urban landscape.

3.2. Relative contributions of future climate andurbanization to temperature increaseTo extract the relative contributions of future climate(Fclimate) and urbanization (Furbanization) a factorseparation analysis (Stein and Alpert 1993)was under-taken. The analysis considers the following equations:

F R R 2climate 4 1 ( )= -

F R Rcompact_city: 3urbanization 2 1 ( )= -

F R R R R 4interactions 5 2 4 1( ) ( ) ( )= - - -

F R Rdispersed_city: 5urbanization 3 1 ( )= -

F R R R R 6interactions 6 3 4 1( ) ( ) ( )= - - -

where R1 to R6 correspond to different numericalexperiments as listed in table 1.

Urbanization contributes to the temperatureincrease by 0.5 °C for compact-city scenario R R2 1( )-and 0.4 °C for dispersed-city scenario R R ,3 1( )- at therates of 0.1∼0.13 °C per decade. Future climate con-tributes 2.44 °C of temperature increase, taking up

more than 80% of the total increased thermal loading(see table 2). The projected warming rate due to urba-nization is comparable to historical evaluation(0.11 °C per decade) over BMA, but relative contrib-ution is smaller than historical estimation (which is40∼61% during 1981–2000 as estimated in previousstudies as Ren et al (2007, 2008)). From urban climatemodeling perspective, the difference between com-pact- and dispersed-city scenarios is thus relatively asmaller consideration compared to the overwhelmingimpact of future climate on temperature increase,although thermal comfort over downtown Beijing isslightly better in dispersed-city scenario than com-pact-city scenario.

4. Summary anddiscussions

In this study, we compared the contrasting impacts ofurban forms on thermal environment under futureclimate conditions taking the Beijing metropolitanregion as an example. Future urban forms are repre-sented by compact- and dispersed-city scenarios, withthe latter being specifically designed as a future cityform by the local government. Our analyses show thatdispersed city is effective in alleviating mean urbanheat island intensity, as well as producing a smallerthermal loading over downtown region compared tocompact-city scenario. However, the regional warm-ing effect is more significant in the dispersed-cityscenario, due to the conversion of surroundingvegetated surface to urban coverage. For the compact-city scenario, the additional urban coverage is only atthe expense of sub-urban area that surrounds the

Figure 5. Schematic contrasts of the impacts on thermal environment between compact- and dispersed-city scenarios.Horizontaldashed lines representmaximum temperature for compact and dispersed-city, andmean temperature for dispersed- and compact-city, respectively. Red: compact-city and blue: dispersed-city.

Table 2.Temperature increase and relative contributions. Percentages are shown in parentheses. The temperature is evaluated over down-town area.

Urban scenarios Total increase (°C) Attributable to climate (°C) Attributable to urbanization (°C) Due to interactions (°C)

compact-city 2.98 2.44 (82%) 0.5 (17%) 0.04 (1%)dispersed-city 2.89 2.44 (85%) 0.4 (13%) 0.07 (2%)

7

Environ. Res. Lett. 11 (2016) 034018

downtown region. The larger regional warming in thedispersed-city scenario might also potentially increasethe depth of vertical circulation by producing a deeperperturbation on vertical temperature profile underfuture climate over the domain.

Contrasts between compact- and dispersed-cityscenarios indicate that no one scenario outperformsthe other by all means. The ‘superiority’ of urbanforms is conditioned on perspectives in evaluation.For instance, the dispersed-city scenario might beeffective in moderating the urban heat island or heatstress increase over the downtown region, but this sce-nario might increase the potential of more frequentwarm-season thunderstorms over urban areas, whichis typically not considered by city dwellers and urbanplanners (e.g., Yang et al 2014). In addition, the warm-ing effect might have broader implications beyondurban climate and human comfort (Grimmet al 2008), for instance, the phenology of vegetationsurrounding urban areas could be significantly dis-turbed by the urban heat island effect (Zhanget al 2004, Zhou et al 2014), impact on biodiversity andcarbon pools (Seto et al 2012), and regional air quality(Jacobson et al 2015). The thermal impact beyondurban coverage is also known as urban footprint(Zhang et al 2004). More recently, Zhou et al (2015)conducted a regional analysis of the urban footprintfor 32 major cities in China, which included the Beij-ingmetropolitan region. Our study finds that the ther-mal contrast between different urban formsdiminishes at the spatial extent with the entire areabeing 7∼8 times of the urban core region. That is, theurban thermal footprint may influence about 7∼8times the size of the urban area (around 40 km awayfrom the city center). Schmid and Niyogi [2013] dis-cussed the threshold of city size for urban modifica-tion on thunderstorms, and found the magnitude ofprecipitation modification increased linearly with theradius of a city up to 20 km, where it became moreconsistent with increasing city size.

Cities evolve both in size and forms by transform-ing surrounding rural or vegetated areas. This is a typi-cal development scenario for cities in developingcountries. As such, findings from our study directlyapply for city planners in developing countries. Resultsalso show that future climate warming contributes tomore than 80% of total thermal loading over BMA,with the rest effect from urbanization. Thus resultsfrom this study indicate that the impact of regionalwarming or climate change is larger than the impact ofurban extent for urban thermal loading, followed byurban form as the third important factor.

There are no ‘one-size-fits-all’ type solutions formaintaining sustainable cities (Georgescu et al 2015).Comprehensive mitigation tools should be examinedin conjunction with urban forms to enhance theadaptability of cities to potential heat-related risksunder the context of global warming. Potential mitiga-tion tools include: reducing surface albedo through

configuration of green/white roofs (Georgescuet al 2014, Li et al 2014), expanding green vegetationcoverage within downtown region (Zhang et al 2009),and also reconstructing the structures of industriesand energy consumption.

The issue being addressed has multiple scales andindeed it is possible to use multiple approaches toinvestigate the interactions. Every study and approachhas certain limitations, as are also inherent to ourstudy. For example, anthropogenic heat is an impor-tant but an uncertain component of the urban energybudget. In the WRF/SLUCM modeling systems usedin this study, anthropogenic heat is represented bycharacteristic diurnal cycles that are directly linked tourban land use categories (Chen et al 2011). Anthro-pogenic heat is generally contributed by three mainsectors: building, human metabolism and traffic (Sai-lor 2011, Sailor et al 2015). Current consideration inanthropogenic heat underestimates contributionsfrom traffic sector over cities with different urbanforms. For instance, in dispersed-city scenario,anthropogenic heat from traffic sectors should bemost influenced by frequent commute between newresidential centers and downtown region. Emissionsfrom transportation in dispersed-city scenario wouldfurther enhance regional thermal loading as identifiedin this study. For compact-city scenario, variation ofanthropogenic heat depends on outside air temper-ature, which is determined by the positive feedbackbetween temperature and air-conditioning energydemand (Kikegawa et al 2003, Gago et al 2013). Wenote that total thermal loading over downtown regionshould be even larger in compact-city scenario. Eventhough we used the default configurations related toanthropogenic heat in all the simulations, the contrastbetween two urban forms should not change but mayintensify to some extent when dynamic (i.e. consider-ing traffic flow and air conditioning) anthropogenicheat components are considered. The relative con-tributions to temperature increase from the urbaniza-tion side should take a larger portion than 20%, whichhighlights the importance of extra mitigations tools inmoderating future thermal risks. Future studies needto come up with comprehensive urban scenarios withboth changes in urban coverage and socioeconomicvariations (such as changes in energy structures, greenplanning, etc) considered.

Acknowledgement

This study is financially supported byNational ScienceFoundation (CDSE 1250232, AGS 1522 494, AGS0847472). DN acknowledges support from NSFCAREER,NSF STRONGCity grants. LY also acknowl-edges support from Chinese Postdoctoral funding(2015M570110). The authors would like to thank DrM Georgescu for his pre-submission review andcomments which improved the paper. Two

8

Environ. Res. Lett. 11 (2016) 034018

anonymous reviewers provided useful commentswhich helped improve the paper.

References

Adachi SA, Kimura F, KusakaH,DudaMG, Yamagata Y, SeyaH,Nakamichi K andAoyagi T 2014Moderation of summertimeheat island phenomena viamodification of the urban form inthe TokyoMetropolitan area J. Appl.Meteorol. Climatol. 531886–900

Angel S, Parent J, CivcoDL, Blei A and PotereD 2011Thedimensions of global urban expansion: Estimates andprojections for all countries, 2000–2050Prog. Plann. 7553–107

BureauB S 2012Beijing Statistics Yearbook (Beijing, China: ChinaStatistics Press)

Chen F et al 2011The integratedWRF/urbanmodelling system:development, evaluation, and applications to urbanenvironmental problems Int. J. Climatol. 31 273–88

Frolking S,MillimanT, SetoKC and FriedlMA2013A globalfingerprint ofmacro-scale changes in urban structure from1999 to 2009Environ. Res. Lett. 8 024004

Gago E J, Roldan J, Pacheco-Torres R andOrdóñez J 2013The cityand urban heat islands: a review of strategies tomitigateadverse effectsRenew. Sustain. Energy Rev. 25 749–58

GeorgescuM2015Challenges associatedwith adaptation to futureurban expansion J. Clim. 28 2544–63

GeorgescuM,Morefield P E, Bierwagen BG andWeaverCP 2014Urban adaptation can roll backwarming of emergingmegapolitan regionsProc. Natl Acad. Sci. USA 111 2909–14

GeorgescuM,ChowWTL,WangZH, Brazel A, Trapido-Lurie B,RothMandBenson-Lira V 2015 Prioritizing urbansustainability solutions: coordinated approachesmustincorporate scale-dependent built environment inducedeffectsEnviron. Res. Lett. 10 061001

GrimmNB, Faeth SH,GolubiewskiNE, RedmanCL,Wu J,Bai X andBriggs JM2008Global change and the ecology ofcities Science 319 756–60

Hathway EA and Sharples S 2012The interaction of rivers andurban form inmitigating theUrbanHeat Island effect: a UKcase studyBuild. Environ. 58 14–22

HondulaDM,GeorgescuMandBalling RC Jr 2014Challengesassociatedwith projecting urbanization-induced heat-relatedmortality Sci. Total Environ. 490 538–44

HouA2012 Impacts of Urbanization on LocalHydrometeorologicalVariables—ACase Study in Beijing (Beijing, China: TsinghuaUniversity)

HouA,NiG, YangH and Lei Z 2012Numerical analysis on thecontribution of urbanization towind stilling: an exampleover theGreater Beijingmetropolitan area J. Appl.Meteorol.Climatol. 52 1105–15

JacobsonMZ,Nghiem SV, Sorichetta A andWhitneyN 2015Ringof impact from themega-urbanization of Beijing between2000 and 2009 J. Geophys. Res. Atmos. 120 5740–56

JinML,DickinsonRE andZhangDL 2005The footprint of urbanareas on global climate as characterized byMODIS J. Clim. 181551–65

KawaseH, Yoshikane T,HaraM,Kimura F, Yasunari T, Ailikun B,UedaH and Inoue T 2009 Intermodel variability of futurechanges in the Baiu rainband estimated by the pseudo globalwarming downscalingmethod J. Geophys. Res. 114D24110

Kikegawa Y,Genchi Y, YoshikadoHandKondoH2003Development of a numerical simulation system towardcomprehensive assessments of urbanwarmingcountermeasures including their impacts upon the urbanbuildings’ energy-demandsAppl. Energy 76 449–66

LandsbergH 1981TheUrbanClimate (NewYork: Academic)Lauer A, ZhangCX, Elison-TimmO,WangYQ andHamiltonK

2013Downscaling of climate change in theHawaii regionusingCMIP5 results: on the choice of the forcing fieldsJ. Clim. 26 10006–30

LiD, Bou-Zeid E andOppenheimerM2014The effectiveness ofcool and green roofs as urban heat islandmitigation strategiesEnviron. Res. Lett. 9 055002

LiD, SunT, LiuM,Yang L,Wang L andGaoZ 2015Contrastingresponses of urban and rural surface energy budgets to heatwaves explain synergies between urban heat islands and heatwavesEnviron. Res. Lett. 10 054009

Liu Y,DuZ,Wang Y, ZhangW, Fei T and Li T 2014 Impacts of heatwaves on emergency department visits in Beijing 2010 SouthChina J. Prev.Med. 40 322–6

Martilli A 2014An idealized study of city structure, urban climate,energy consumption, and air qualityUrbanClim. 10 430–46

Miao S, Chen F, LeMoneMA, TewariM, LiQ andWang Y 2009Anobservational andmodeling study of characteristics of urbanheat island and boundary layer structures in Beijing J. Appl.Meteorol. Climatol. 48 484–501

Miao S, Chen F, Li Q and Fan S 2011 Impacts of urban processes andurbanization on summer precipitation: a case study of heavyrainfall in Beijing on 1 august 2006 J. Appl.Meteorol. Climatol.50 806–25

Miao S 2015Beijing Institute of UrbanMeteorology personalcommunication

Niyogi D,Holt T, Zhong S, Pyle PC andBasara J 2006Urban andland surface effects on the 30 July 2003mesoscale convectivesystem event observed in the southernGreat PlainsJ. Geophys. Res. Atmos. 111D19107

Niyogi D, Pyle P, LeiM, Arya S P, Kishtawal CM, ShepherdM,Chen F andWolfe B 2011Urbanmodification ofthunderstorms: an observational storm climatology andmodel case study for the Indianapolis urban region J. Appl.Meteorol. Climatol. 50 1129–44

OkeTR 1982The energetic basis of the urban heat islandQ. J. R.Meteorol. Soc. 108 1–24

OkeTR 1984Towards a prescription for the greater use of climaticprinciples in settlement planning Energy Build. 7 1–10

RenG, ZhouY,ChuZ, Zhou J, ZhangA,Guo J and LiuX 2008Urbanization effects on observed surface air temperaturetrends inNorthChina J. Clim. 21 1333–48

RenGY, ChuZY,ChenZH andRenYY 2007 Implications oftemporal change in urban heat island intensity observed atBeijing andWuhan stationsGeophys. Res. Lett. 34 L05711

Riahi K, Rao S, KreyV, ChoC,ChirkovV and Fischer G 2011RCP8.5—a scenario of comparatively high greenhouse gasemissionsClim. Change 109 33–57

RyuY-H andBaik J-J 2012Quantitative analysis of factorscontributing to urban heat island intensity J. Appl.Meteorol.Climatol. 51 842–54

SailorD J 2011A review ofmethods for estimating anthropogenicheat andmoisture emissions in the urban environment Int. J.Climatol. 31 189–99

SailorD J, GeorgescuM,Milne JM andHartMA2015Development of a national anthropogenic heating databasewith an extrapolation for international citiesAtmos. Environ.118 7–18

SchererD, FehrenbachU, BehaH and ParlowE 1999 Improvedconcepts andmethods in analysis and evaluation of the urbanclimate for optimizing urban planning processesAtmos.Environ. 33 4185–93

Schmid PE andNiyogi D 2013 Impact of city size on precipitation-modifying potentialGeophys. Res. Lett. 40 5263–7

SchwarzN, PhD,Manceur AMandPhD2013 Analyzing theInfluence of urban forms on surface urban heat islands inEurope J. Urban Plann. Dev. 141A4014003

SetoKC,Guneralp B andHutyra LR 2012Global forecasts of urbanexpansion to 2030 and direct impacts on biodiversity andcarbon poolsProc. Natl. Acad. Sci. USA 109 16083–8

Shepherd JM2005A review of the current investigations of urbanindeuced rainfall and recommendations for the futureEarthInteract. 9 1–27

Sherwood SC andHuberM2010An adaptability limit to climatechange due to heat stressProc. Natl Acad. Sci. USA 1079552–5

9

Environ. Res. Lett. 11 (2016) 034018

SkamarockWC,Klemp J B,Dudhia J, Gill DO, BarkerDM,DudaMG,HuangX-Y,WangWandPowers J G 2008ADescription of the Advanced ResearchWRFVersion 3 (Boulder,Colorado,USA:National Center for Atmospheric Research)

SteadmanR 1984Auniversial scale of apparant temperature J. Appl.Meteorol. 23 1674–87

SteinU andAlpert P 1993 Factor separation in numericalsimulations J. Atmos. Sci. 50 2107–15

Stone B andRodgersMO2001Urban form and thermal efficiency:how the design of cities influences the urban heat island effectJ. Am. Plan. Assoc. 67 186–98

Stone B,Hess J J and FrumkinH2010Urban form and extreme heatevents: are sprawling citiesmore vulnerable to climate changethan compact cities? Env. Heal. Perspect 118 1425–8

WuW, LynchAH andRivers A 2005 Estimating the uncertainty in aregional climatemodel related to initial and lateral boundaryconditions J. Clim. 18 917–33

Yang L, Smith J A, BaeckML, Bou-Zeid E, Jessup SM,Tian F andHuH2013a Impact of urbanization on heavy convectiveprecipitation under strong large-scale forcing: a case studyover theMilwaukee–LakeMichigan region J. Hydrometeorol.15 261–78

Yang L, Smith J A,Wright DB, BaeckML,Villarini G, Tian F andHuH2013bUrbanization and climate change: anexamination of nonstationarities in urbanfloodingJ. Hydrometeorol. 14 1791–809

Yang L, Tian F, Smith J A andHuH2014Urban signatures in thespatial clustering of summer heavy rainfall events over theBeijingmetropolitan region J. Geophys. Res. Atmos. 1192013JD020762

ZhangCL, Chen F,Miao SG, LiQC,Xia XA andXuanCY 2009Impacts of urban expansion and future green planting onsummer precipitation in the Beijingmetropolitan areaJ. Geophys. Res. 114D02116

ZhangX, FriedlMA, Schaaf CB, Strahler AH and Schneider A 2004The footprint of urban climates on vegetation phenologyGeophys. Res. Lett. 31 10–3

ZhaoS,ZhouD,ZhuC,QuW,Zhao J, SunY,HuangD,WuWandLiu S 2015Rates andpatterns of urban expansion inChina’s 32major cities over thepast threedecadesLandsc. Ecol.30 1541–59

ZhouB, RybskiD andKropp J P 2013On the statistics of urban heatisland intensityGeophys. Res. Lett. 40 5486–91

ZhouD, Zhao S, Liu S andZhang L 2014 Spatiotemporal trends ofterrestrial vegetation activity along the urban developmentintensity gradient in China’s 32major cities Sci. TotalEnviron. 488–489 136–45

ZhouD, Zhao S, Zhang L, SunG andLiu Y 2015The footprint ofurban heat island effect inChina Sci. Rep. 5 11160

Zhou L,DickinsonRE, TianY, Fang J, LiQ, KaufmannRK,Tucker C J andMyneni RB 2004 Evidence for a significanturbanization effect on climate inChinaProc. Natl Acad. Sci.USA 101 9540–4

10

Environ. Res. Lett. 11 (2016) 034018