Theor. Appl. Climatol. **, *** *** (****)

DOI **.1007/s00704-006-0273-1

Printed in The Netherlands

1

Institute of Remote Sensing Applications, Chinese Academy of Sciences, Beijing, China

2

Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

3

Institute of Atmospheric Physics, Chinese Academy of Sciences, Chao yang district, Beijing, China

4

School of Forestry and Wildlife Sciences, Auburn, Alabama, USA

Assessing the effect of land use= land cover change

on the change of urban heat island intensity

J. F. He1, J. Y. Liu2, D. F. Zhuang2, W. Zhang3, and M. L. Liu2;4

With 6 Figures

Received January 24, 2006; revised May 11, 2006; accepted August 31, 2006

Published online February 28, 2007 # Springer-Verlag 2007

Summary (Esserl, 1989; Grunblatt et al., 1992; Wei and Fu,

1998; Lambin et al., 1999; Hillel and Rosenzweig,

Due to rapid economic development, China has experienced

2002; Gao et al., 2003). The urban heat island

one of the greatest rates of change in land use=land cover

(UHI) effect, the phenomena where air and surface

during the last two decades. This change is mainly urban

expansion and cultivated land reduction in urban growth temperatures of urban areas are higher than those

regions, both of which play an important role in regional of its surrounding rural areas, has been rigorously

climate change. In this paper, the variation of the urban heat

researched (e.g. Oke, 1982; Owen, 1998). UHI

island (UHI) caused by urbanization has been evaluated with

was described by Roth et al. (1989) as one of

an analysis of land use change in China. First, meteorological

the most clearly established examples of inadver-

observation stations were grouped by different land cover

tent modi cation of climate. Numerous studies

types (dry land, paddy eld, forest, grassland, water eld,

urban, rural inhabitable area, industrial and mineral land, and have shown that urbanization can produce radi-

waste land) throughout China. These stations were sub- cal changes in the radiative, thermodynamic, and

divided into urban and non-urban classes. Then, a new

water processes of the land surface and modify

method was proposed to determine the UHI intensity from

local climate in terms of for example, tempera-

the difference between the observed and the interpolated air

ture, precipitation, and cloudiness (Carlson and

temperature of urban type weather stations. The results

indicate that the trends of UHI intensity in different land use Arthur, 2000; Huff and Changnon, 1972; Oke,

change regions are spatially correlated with regional land use 1974). There is increasing evidence that global

and its change pattern. During 1991 2000, the estimated

warming is a result of anthropogenic activity dur-

UHI intensity has increased by 0.11 C per decade in the

ing the past fty years (IPCC, 2001). Global warm-

spring and has uctuated in other seasons throughout China

ing can be partitioned into (1) the urban heat

resulting from land use change.

island effect, (2) the effect of deforestation, (3)

the effect of secular micro-climate shift, (4) the

1. Introduction

in uence of general global warming with par-

ticular reference to the tropics (Harger, 1995). It

In recent years, scientists have recognized that

land cover=land use changes induced by human is therefore important to eliminate the urban heat

island effect from observed temperature records

activities have large impacts on regional climate

218 J. F. He et al.

in order to assess monthly, seasonal, and annual which is derived from high resolution remote

averaged long-term temperature trends (Karl sensed data and covers the period from the end

et al., 1988; Jones et al., 1990; Wang et al., 1990; of the 1980s to 2000 (Liu et al., 2003, 2005a), this

Matsushita et al., 2004). For example, a reanalysis study develops a new method of estimating how

of global temperature data over the past 50 years the intensity of the UHI has changed during this

that is insensitive to surface dynamics has been ten year period and how much was the result of

produced (Kalnay et al., 1996; Kistler et al., 2001) land use change.

and the data have been used to assess the long-term

climate change, its causes and its impact on carbon

2. Data sources

and water cycles.

The intensity of the urban heat island (UHI) is The intensively validated NLCD are used as land

directly related to the rate of urbanization, land use use change background for this study. The NLCD

patterns, and building density. It can be assessed in contains land use data in vector format with a

several ways: measuring the UHI by comparing scale of 1:100,000 and gridded land use percen-

present to earlier pre-urban conditions (Lowry, tage data at a one square kilometre resolution.

1977), measuring the UHI through the urban-to- The NLCD, constructed by the Chinese Academy

rural temperature difference (Lowry, 1977; Karl of Sciences and the Ministry of Science and

et al., 1988); and measuring the UHI in terms of Technology, is based on remote sensed data

surface (or skin) temperature through the use of Landsat TM=ETM (Thematic Mapper=Enhanced

airborne or satellite remote sensing (Roth et al., Thematic Mapper) covering three periods: the

1989; Gallo and Tarpley, 1996; Streutker, 2003). end of the 1980s (1985 1990), 1995=1996, and

The selection of the most appropriate method 2000 (Liu et al., 2003). According to quality as-

depends mainly on the observed data available. sessment, the accuracy of classi cation for urban

China, the biggest developing county in land areas is 96.32% and the bias of polygon bound-

area, has experienced rapid urbanization in recent aries for each land use type is less than 45 m

years (Liu et al., 2005a, b). Some recent studies (Liu et al., 2003). The gridded percentage of land

have shown the effect of large scale urbanization use represents the areal composition of each land

on regional temperature change in China (Zhang use type within each one square kilometre unit

et al., 2003; Zhou et al., 2004), but they failed to (Liu et al., 2002). To characterize the spatial pat-

directly estimate the temporal and spatial variabil- tern of variability in intensity of UHI from land

ity of the UHI intensity impacted by the real land use change, the map of national land use change

use change (especially urban sprawl). Using the zoning (Liu et al., 2003) was used as regional

National Land Cover Dataset (NLCD) of China, divisions.

Fig. 1. Spatial distribution of

meteorological observation sta-

tions and their corresponding

ground surface land use types

Assessing the effect of land use=land cover change on the change of urban heat island intensity 219

Daily temperature data used in this study are improvement and extension in order to eliminate

collected from the 673 meteorological stations uncertainty and to discover the effect of regional

of the China Historical Climatological Network land use change over a long time period by using

and include records from 1991 to 2000 (China high resolution data. In general, signi cant spa-

Meteorological Administration, http:==211.147. tial differences (elevation and latitude) would

16.25=ywwz=constitute=wls01.php). At rst, the exist between urban and non-urban pair stations

tabular air temperature data are converted into if their distance is over 100 kilometres (km). Un-

spatial data with point vector format in Geo- fortunately, many urban stations in China are far

graphical Information Systems (GIS). Then, by from the surrounding non-urban stations. It is dif-

overlaying the station point vector layer with cult to nd enough urban and non-urban pairs to

the NLCD land use layer, the satellite-based land estimate the intensity of UHI at a national scale.

use classes (dry land, paddy eld, forest, grass- Here, a new method is proposed to assess the

land, water eld, urban, rural inhabitable area, intensity of UHI. It identi es the difference be-

industrial and mineral land, and waste land) are tween the observed air temperature by an urban

designated to meteorological observation stations. weather station and the interpolated temperature

Stations with the urban land use feature are de- by non-urban stations. The interpolated tempera-

ned as urban type weather stations, and those tures from non-urban stations are set as back-

with other land use features are classed as non- ground temperature of the urban station. The

urban types. According to this de nition, 273 estimated intensity of UHI (HI) for one urban

stations (40% of total numbers) were urban in station can be expressed as

2000. There were 46 weather stations experienc- HI Tu Tb 1

ing urbanization (from non-urban to urban) from

where Tu is the observed temperature of the

1991 to 2000, accounting for 6% of the total

urban station; Tb is the background temperature

stations (Fig. 1). In this study, observations are

of the urban station. Figure 2 describes the de -

used to identify which stations (urban type) are

nition of urban background temperature of the

affected by the UHI effect and which are not,

urban station. The interpolation method is from

with the exception of the 46 weather stations ex-

Thornton et al. (1997) which is based on the spa-

periencing urbanization.

tial convolution of a truncated Gaussian weight-

ing, i.e.

Pn

3. Methods Wi;r Ti 0 1 zb zi

Tb i 1 Pn 2

i 1 Wi;r

3.1 Estimation of UHI intensity 8

at the regional scale Rb

2

Wi;r 3

: exp Rb e ; r Rb

As previously described, the traditional methods r

used to estimate the intensity of UHI required

Fig. 2. (a) Urban and its neigh-

bor non-urban stations (b) cal-

culation of urban background

temperature of urban station

220 J. F. He et al.

where Ti is the observed temperature around the Finally, the probability distributions of the

number of UHI affected sites Pk for each sea-

non-urban station i; zb is the elevation of the pre- m

dicting urban station; zi is the elevation of the son at the regional scale is calculated.

X 1; if F k m

non-urban station i; Wi,r is the lter weight of k j

Pm 6

the non-urban station i that related with radial 0; else

j

distance r from the predicting site; Rb is the trun-

where m represents the number of years that site j

cation distance from the predicting site; is a

represented positive UHI intensity in season k, the

unit-less shape parameter of the Gaussian func-

value scope of m is from 0 to 10. The probability

tion; n is the total number of non-urban weather

of Pk is equal to Pk =sum of urban stations.

stations that have a distance less than Rb from the m m

predicting urban station; 0 and 1 are the coef-

cients of linear regression, which is calculated 3.3 Validation of the method

from station pairs by the 2-stage least squares ap-

To validate this method, we camed out experi-

proximation method to re ect the altitude effect

ments to randomly select urban and non-urban

on temperature. Rb is assigned as 150 km and

stations upon which to calculate UHI intensities

as 3.0 after many tests to achieve the least extra-

for the study period. A station is de ned as an

polation error. In the processing of interpolation,

urban type based on the possibility of an urban

weighting averages of the temperature of non-

effect within a certain distance from the station.

urban stations within 150 km of the urban station

A one kilometre distance is selected as the buf-

will remove the effects of latitudinal differences,

fering radius around the station to run the overlay

and the calculated vertical temperature reduction

procedure with NLCD urban layer by GIS. If the

rate of elevation will reduce the elevation effects

urban area covers more than 75% of the buffering

on the interpolated temperature.

zone of the station, the station is re-classi ed as

In this paper, we calculated the variability

an urban type, otherwise, as a non-urban station.

of UHI intensity for each urbanization affect-

ed weather station by seasonal average air tem-

perature for the periods: Dec. Feb. (winter),

Mar. May (spring), June Aug. (summer) and

Sep. Nov. (autumn).

3.2 Analysis of statistical features

of estimated UHI intensity

In order to estimate regional UHI intensity effec-

tively, a discrete mathematics method is used. At

rst, for each site in each season, the estimated

UHI intensity is transformed from oat format

into integer format so that only 0 or 1 is assigned

(Eq. 4).

(

0; HIik; j 0

Dk; j 4

1; HIik; j > 0

i

where k represents each season; i represents each

year; j represents each urban weather station.

Then, the frequency of the UHI effect Fjk (i.e.

UHI intensity is 1) at each station for each season

during 1991 2000 is calculated.

X

Fjk Dk; j 5 Fig. 3. Statistical distributions of Pk for both urban type

i m

(a) and non-urban type (b) stations

i

Assessing the effect of land use=land cover change on the change of urban heat island intensity 221

The statistical distributions of Pk for both urban ability of UHI intensity at the station level with

m

and non-urban type stations are then calculated. land use change zoning maps (Liu et al., 2003)

Signi cant differences between them are found. (Fig. 5). Table 1 describes the change area of

For urban stations, when m 10, the probability main land use types in different land use change

of Pk is the largest in all m and when m 0, the zones. From the percentage of urban stations in-

m

probability of Pk is small; while for the non- tensively affected by UHI (we refer to the value

m

of Pk from formula 6) in each zone, the spatial

urban stations, the probability distribution of 10

Pk is close to the ideal situation that Pk is equal differences in urbanization effect on temperature

m m

to m 0 and m 10 (Fig. 3). This indicates that would be discovered. Our results show that Zone 4

our method effectively detects the UHI effect. (Huang-Huai-Hai and Yangtze River delta culti-

vated land to urban land transform area) has the

strongest UHI effect from 1991 to 2000 (Figs. 5

4. Results

and 6).

Our study also shows that the variability of

4.1 The trend and seasonal variability of UHI

UHI intensity is correlated with land use change

intensity in China

during the period of 1991 to 2000 (Fig. 6). The

The intensity of UHI varies seasonally. The urban

greatest increase in UHI intensity took place in

effect on temperature is more signi cant in winter

the area with rapid urbanization. Urbanization is

and autumn than that in spring and summer. For

a widespread phenomenon in China due to its

example, more stations have large UHI intensity

fast economic growth and the increasing desire

in the rst two seasons (Fig. 4). For each season,

by people for an urban life style. During this study

we found no signi cant trend in HI except in spring

period, remote sensing data discovered that more

with an increasing rate of 0.11 C per decade

than 70% of the increased urban area is located

(y 0.011x 0.605, R2 0.517) (Fig. 4). HI uc-

in Region 4 (Huang-Huai-Hai and Yangtze River

tuated from 1.0 to 1.2 C (mean 1.1 C) in winter,

delta region), Region 5 (Sichuan basin region),

from 0.6 to 0.8 C (mean 0.66 C) in spring, from

and Region 11 (East-southern seaside of China)

0.6 to 0.7 C (mean 0.65 C) in summer, and from

(Liu et al., 2003, 2005a). Table 2 gives the urban

0.7 to 0.8 C (mean 0.76 C) in autumn.

area change ratios of the four main urban growth

regions every ve years during 1990 2000. Most

4.2 The spatial pattern of UHI intensity of the newly increased urban area in Region 4

and its relationship with land use (about 1100 thousand hectares) is from farmland

that has been cultivated for a long time. The ur-

To study the relationship between UHI inten-

banization process has greatly affected the local

sity and land use change, we linked the vari-

environment and climate systems in the last de-

cade as has been shown in this study in the case

of temperature in Regions 4, 6 and 7. The urban

weather stations with greatest UHI effect are

broadly distributed across these areas. In North

China and the Loess Plateau region (Region 6),

a large area covered by natural vegetation (re-

duction 1,695,820 ha.) and water (reduction

116,348 ha.) has been transformed into cropland

and urban areas. This would induce signi cant

environmental risks because of the fragile phys-

ical background (Zhang et al., 2003). The main

feature of land use change in Central China

(Region 8) is the transformation of paddy elds,

pothole pools, bottomland and lakes and the ex-

pansion of urban land. In the southeastern hill

forest, the land=cultivated land transformation

Fig. 4. Urban heat island intensity distributions in different

seasons during 1991 2000 (unit: C) area (Region 9), farmland and forestland re-

222 J. F. He et al.

Fig. 5. The distribution map of urban stations which are affected by UHI overlaying with land use change zoning map in four

seasons. The black dot represents urban station; the size of them represents the seasonal average UHI intensity of urban

station; the number of polygon is the code of land use change zone, which are, 1 north-east China Da=xiao xing an mountains

forestry=grass land ! cultivated land transform area, 2 east part of north east China forestry=grass land ! cultivated land

transform area, 3 north-east China plain dry land ! paddy eld transform area, 4 Huang-Huai-Hai and Yangtze river delta

cultivated land ! urban land transform area, 5 Sichuan basin cultivated land ! urban transform area, 6 north China=Loess

Plateau farming and herding grass land ! cultivated land transform area, 7 north-west China cultivated land expansion and

waste land transform area, 8 central China interchange between paddy elds, pothole pools, bottomland and lakes and

urbansprawl area, 9 south eastern hill forest land ! cultivated land transform area, 10 south eastern seaside grass land ! man

made forestry transform area, 11 south-east seaside urban sprawl area, 12 south-western forest-grass land and forest grass

land ! cultivated land transform area, 13 Qinghai-Tibet Plateau little change area

veal a mixed distribution. Water area changes fre- Plateau region, belonging to stable and less change

quently. Though urbanization expands in these areas), urban stations are either not or only very

regions, due to the temperature regulation func- slightly affected by the UHI effect. However, be-

tion of water, the climate environment is hard- cause of development, some urban stations in

ly affected by land use change. Therefore, in big cities in these regions have experienced the

Regions 1, 2, 3, 5, 8, 9 and 13 (Qinghai-Tibet UHI effect.

Assessing the effect of land use=land cover change on the change of urban heat island intensity 223

Table 1. Land use change area in different land use change zone

Zoning code Paddy eld Dry eld Forestry Grassland Unused eld Water eld Urban

758***-****** 42924

1-242**-****** 10260 6310

319***-****** 106***-****

2-135***-****** 17192

391**-****** 649**-****** 50233

3-383***-*****

349***-****** 219**-****** 117955

4-616**-*******

540**-*****-****

5 969 23 599 80747

169****-****** 116348

6-271***-******* 969**-*****

872812

7-132**-****** 527*-****** 126***-*****

531**-*****-****

8-134*-****-***** 48079

246**-*****-**** 20030 66

9-277**-*****

56231 634*-****** 7268

10-249***-***** 59455

131***-*****-**** 789

11-278*-***** 192767

176**-******-***

12-117**-****** 159**-*****

180**-*****

13 0-906*-****-***** 2987

Unit: Ha. Positive value means increase, negative value means decrease.

and the scale of study. At the regional scale, if

urban expansion cannot be successfully detected

and considered in analyzing the station data, the

bias in estimating the UHI intensity and the tem-

perature trend would be large. The method we

used in this study is appropriate in studying the

UHI effect when the availability of urban rural

pairs and meteorological stations is limited. At

the regional scale, urban land use changes cannot

be detected and may cause some bias in UHI

studies (Brian et al., 1998). However, we selected

the non-urban stations free of the in uence of

Fig. 6. The relationship between UHI spatial distribution

UHI, in some cases; microclimate differences

and the land use dynamic change spatial distribution

cannot be removed completely within a range of

150 km. Also, the interpolation method itself

5. Discussion

may introduce some uncertainties because of the

sparse distribution of the stations.

UHI is a special meteorological phenomenon in

In order to evaluate our results we compared our

urban areas but is controlled by the physical

study with that of Weng and Yang (2004) which

characteristics and structure of the land surface

examines the impact of urban development on

within and around the urban area such as soil

UHI through a historical analysis of urban rural

moisture, vegetation communities and imper-

air temperature differences. Weng and Yang (2004)

vious surface coverage. Detecting the UHI effect

indicates that during 1985 1995 and 1996 2000,

relies on the availability of meteorological data

Table 2. Urban area change ratio in main 4 urban growth regions every ve years during 1991 2000

Code Urban area Increased area Increased area Ratio during Ratio during

of 1990s during 1990 1995 during 199*-****-**** 199*-****-****

4 1143462.12 418811.35 69101.79 36.63 4.42

5 73057.95 19332.61 21261.72 26.46 23.01

8 135546.88 22887.80 3187.50 16.86 2.01

11 254623.56 101871.23 6541.32 40.01 1.83

Area unit: Ha.

224 J. F. He et al.

Table 3. The UHI intensity of Guangzhou city measured in two ways

Year 199*-****-**** 199*-****-**** 199*-****-**** 2000

0.308 0.125 0.100 0.050

Weng s result 0.542 0.458 0.492 0.400 0.658 0.058

0.203 0.2 0.025

Estimated in our paper 0.548 0.608 0.593 0.718 0.703 0.025 0.12

0.006 0.150 0.101 0.318 0.045 0.106 0.075 0.170

Diff. 0.075 0.033

R2: 0.822

the average intensity of the UHI is 0.7 C=year If the weather station is located at the bound-

and 0.13 C=year, respectively, in the station lo- ary of an urban area, it may not be detected even

cated in Guangzhou, a big city in southeast China though it is surely affected by the UHI. The noise

(Table 3). The intensity measured in Weng s paper produced by the microclimate of the urban area

is lower than ours and the R2 0.822. The main where weather stations are located may disguise

reason for this difference is that the rural station the UHI effect (Karl et al., 1988). Also, environ-

has been affected by UHI after 1990 due to urban mental variables, such as wind speed, cloud cov-

area enlargement and the rapid increase of its tem- er, atmospheric aerosol, air pollutions, and water

perature. The high R2-value, close to 1, shows that vapor content, would have an impact on the UHI

the precision of our estimated UHI intensity at this effect. This indicates that many factors can con-

station is relatively high and reasonable. Those trol the UHI effects occurring in an urban area.

urban stations with negative HI during 1990 Our study has indicated that the pattern of land

2000 are not considered as UHI affected stations use change is very important for the dynamics of

in our study. For example, we did not select the local climate. For example, both the growth of

weather station located in Guangzhou City as a UHI the urban area and the drastic reduction in lake

affected station because its HI was not always areas produced signi cant UHI effect in the ur-

positive during 1990 2000. The negative HI may ban area (Jazcilevich et al., 2000). In contrast, the

be mainly due to the impact of other land-use expansion of the water surface can buffer the

changes on non-urban stations (Kalnay and Cay, variation of surface temperature and so decrease

2003), or may potentially be related to: (i) the the UHI effect.

difference of geographic environment (longitude

and latitude, elevation, land use type) of the

6. Conclusions

weather stations, or (ii) the location of the weather

stations. In Fig. 3, we nd that about half of the It is feasible to estimate the UHI effect on urban

urban stations show no urban heat island effect weather stations at the regional scale by com-

(i.e. there are several negative HI in those sta- paring observed temperature with interpolated

tions). We did not discuss how to identify and background temperatures from near non-urban

detect whether an urban station was affected by stations. Our results show that many weather sta-

UHI when the station revealed negative HI. Ap- tions were under the UHI effect during all sea-

parently, the estimation of UHI intensity is sensi- sons, especially in winter and autumn, and they

tive to methodology. For different countries and were broadly distributed during the period of

regions, the land use change, especially urbaniza- 1991 2000. The intensity of UHI revealed sea-

tion, effect on the intensity of the UHI effect is sonal variation. In winter and autumn the urban

estimated to be 0.11 C=year 0.91 C=year (Karl effect on temperature was more signi cant than

et al., 1988; Wang et al., 1990; Streutker, 2003). in spring and summer. The average UHI intensity

was 1.1 C in winter, 0.66 C in spring, 0.65 C in

We do not assess the effect of UHI on temperature

summer, and 0.76 C in autumn during this ten-

trend series warming as do many studies such as

Kalnay and Cai (2003), Zhou et al. (2004), Li et al. year period.

(2004). These studies indicate that the UHI ef- The urban stations affected by UHI are

fect causes the temperature warming trends of broadly distributed in the Huang-Huai-Hai River

0.05 C=decade 0.011=decade depending on dif- and Yangtze River Delta areas, Southeast China,

ferent regions and analysis methods. the Loess Plateau of North China, and Wulumuqi

Assessing the effect of land use=land cover change on the change of urban heat island intensity 225

Harger JRE (1995) Air-temperature and ENSO effects in

city-Shihezi city in Northwest China. These areas

Indonesia, the Philippines and El Salvador. ENSO pat-

have experienced rapid urbanization and other

terns and changes from 1866 1993. Atmos Environ 29:

intensive land use change during this study pe- 1919 1942

riod. We found that the variability of UHI inten- Hillel D, Rosenzweig C (2002) Deserti cation in relation to

sity has a strong spatial connection with the climate variability and change. Adv Agron 77: 1 38

pattern of land use, i.e. the in uence of the UHI Huff FA, Changnon SA Jr (1972) Climatological assessment

of urban effects on precipitation at St. Louis. J Appl

effect on urban stations is not only related to land

Meteor 11: 823 842

use change type, but also to land use change

IPCC (2001) Climate change 2001: Working Group I: The

structures. In regions where land use change type Scienti c Basis, IPCC third assessment report

represented the sharp expansion of an urban area Jones PD, Groisman PY, Coughlan M, Plummer N, Wang

and the reduction in water area, the urban sta- WC, Karl TR (1990) Assessment of urbanization effects

in time series of surface temperature over land. Nature

tions were often easily affected by UHI. On the

347: 1669 1772

other hand, in regions where land use change type

Kalnay E, Cai M (2003) Impact of urbanization and land-use

was the expansion of urban area accompanied by change on climate. Nature 423: 528 531

an increase in water area or vegetation area, the Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D,

urban stations were seldom affected by UHI. So, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y,

reductions in natural vegetation (forest and grass) Celliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC,

Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R,

and water area (such as wetland and lake) would

Joseph D (1996) The NCEP=NCAR 40-year reanalysis

change microclimate patterns of temperature and

project. Bull Amer Meteor Soc 77: 437 471

make the environment more fragile and climate Karl TR, Diaz HF, Kukla G (1988) Urbanization: its de-

elements more sensitive to global temperature tection and effect in the United States climate record.

change. J Climate 1: 1099 1123

Kistler R, Kalnay E, Collins W, Saha S, White G, Woollen J,

Chelliah M, Ebisuzaki W, Kanamitsu M, Kousky V, et al

(2001) The NCEP-NCAR 50-year reanalysis: monthly

Acknowledgments

means CD-ROM and documentation. Bull Amer Meteor

This research was supported by China s Ministry of Science Soc 82: 247 268

and Technology (MOST) 973 Program (2002CB412500), Lambin EF, Baulies X, Bockstael N, Fischer G, Krug T,

and the Resources and Environment Scienti c Data Center, Leemans R, Moran EF, Rindfuss RR, Sato Y, Skole D,

Chinese Academy of Sciences. We thank three anonymous Turber BL II, Vogel C (1999) IGBP Report No.48=IHDP

reviewers and Dr. P. Gong for their very helpful comments Report No. 10: Land-use and land-cover change imple-

and suggestions. mentation strategy. IGBP Secretariat, The Royal Swedish

Academy of Sciences, pp 75 85

Li Q, Zhang H, Liu X, Huang J (2004) Urban heat island

References effect on annual mean temperature during the last 50 years

in China. Theor Appl Climatol 79: 165 174

Brian P, Dan Moses S-K, Craig KS, Hashem A (1998)

Liu J, Tian H, Liu M, Zhuang D, Melillo JM, Zhang Z

Existing Climate Data Sources and Their Use in Heat

(2005a) China s changing landscape during the 1990s:

Island Research, http:==eetd.lbl.gov=EA=Reports=41973=

Large-scale land transformations estimated with satel-

Existing Climate Data Sources and Their Use in Heat

lite data. Geophys Res Lett 32: L02405, doi:10.1029=

Island Research

2004GL021649

Carlson TN, Arthur ST (2000) The impact of land use=land

Liu J, Liu M, Tian HQ, Zhuang D, Zhang Z, Zhang W, Tang X,

cover changes due to urbanization on surface microcli-

Deng X (2005b) Spatial and temporal patterns of China s

mate and hydrology: A satellite perspective. Global Planet

cropland during 1990 2000: An analysis based on Landsat

Change 25: 49 65

TM data. Remote Sensing Environ 98: 442 456

Esser G (1989) Global land-use changes from 1860 to

Liu JY, Zhang ZX, Zhuang DF, Wang YM, Zhou WC, Zhang

1980 and future projections to 2500. Ecol Model 44:

SW, Li RD, Jiang N, Wu SX (2003) A study on the spatial-

307 316

temporal dynamic changes of land-use and driving forces

Gallo KP, Tarpley JD (1996) The comparison of vegetation

analyses of China in the 1990s. Geographical research

index and surface temperature composites for urban heat-

(Chinese version) 22: 1 12

island analysis. Int J Remote Sensing 17: 3071 3076

Liu JY, Liu ML, Zhuang DF, Zhang ZX, Deng XZ (2002)

Gao XJ, Luo Y, Lin WT, Zhao ZC, Giorgi F (2003) Simula-

Analyses of the recent spatial land use change pattern in

tion of effects of land use change on climate in China by a

China. Sci China Ser D 32: 1031 1040

regional climate model. Adv Atomos Sci 20: 583 592

Lowry WP (1977) Empirical estimation of urban effects

Grunblatt J, Ottichilo WK, Sinange RK (1992) A GIS

on climate: A problem analysis. J Appl Meteor 16:

approach to deserti cation assessment and mapping.

129 135

J Arid Environ 23: 81 102

J. F. He et al.: Assessing the effect of land use=land cover change on the change of urban heat island intensity

226

Matsushita B, Xu M, Chen J, Kameyama S, Tamura M (2004) Wei HL, Fu CB (1998) Study of the sensitivity of a regional

Estimation of regional net primary productivity (NPP) model in response to land cover change over northern

using a process-based ecosystem model: How important China. Hydrol Process 12: 2249 2265

is the accuracy of climate data? Ecol Model 178: 371 338 Weng QH, Yang SH (2004) Managing the adverse thermal

Oke TR (1974) Review of urban climatology 1968 1973. effects of urban development in a densely populated

Tech Note No. 134, WMO, 132 pp Chinese city. J Environ Manage 70: 145 156

Oke TR (1982) The energetic basis of the urban heat island. Zhang GY, Dong WJ, Ye DZ, Fu ZB (2003) New proof to

Quart J Roy Meteor Soc 108: 1 24 validate the effect of vegetation on summer climate in

Owen TW, Carlson TN, Gillies RR (1998) An assessment of China. Chinese Sci Bull 48: 91 95

satellite remotely-sensed land cover parameters in quan- Zhou LM, Dickinson RE, Tian YH, Fang JY, Li QX,

titatively describing the climatic effect of urbanization. Int Kaufmann RK, Tucker CJ, Myneni RB (2004) Evidence

J Remote Sensing 19: 1663 1681 for a signi cant urbanization effect on climate in China.

Roth M, Oke TR, Emery WJ (1989) Satellite-derived urban PNAS 101: 9540 9544

heat islands from three coastal cities and the utilization of

such data in urban climatology. Int J Remote Sensing 10: Authors addresses: J. F. He (e-mail: abpi27@r.postjobfree.com),

1699 1720 Institute of Remote Sensing Applications, Chinese Academy

Streutker DM (2003) Satellite-measured growth of the urban of Sciences, P.O. Box 9718, Beijing 100101, China; J. Y. Liu

heat island of Houston, Texas. Remote Sensing Environ (correspondence; e-mail: abpi27@r.postjobfree.com), D. F. Zhuang,

85: 282 289 Institute of Geographical Sciences and Natural Resources

Thornton PE, Running SW, White MA (1997) Generating Research, Chinese Academy of Sciences, P.O. Box 9719,

surfaces of daily meteorological variables over large Beijing 100101, China; W. Zhang, Institute of Atmospheric

regions of complex terrain. J Hydrol 190: 214 251 Physics, Chinese Academy of Sciences, Chao yang district,

Wang WC, Zeng Z, Karl TR (1990) Urban heat islands in Beijing 100029, China; M. L. Liu, School of Forestry and

China. Geophys Res Lett 17: 2377 2380 Wildlife Sciences, Auburn, Alabama, USA.

Contact this candidate