城市热岛效应的卫星遥感分析
2020-08-28
来源:年旅网
Vo1.10 N0.2 Marine Science Bulletin Oct.2008 An Assessment of Urban Heat Island Effect using Remote Sensing Data WANG Guiling’JIANG Weimei ,WEI Ming。 ,1.Institute ofMeteorology,PLAUST,Nanjing 211101.Jiangsu,China; 2.Department ofAtmospheric Sciences。Nanjing Universit Nanjing 210093.Jiangsu.China; 3.Sino-American Cooperative Remote Sensing Center,NUIST,Nanjing 210044,Jiangsu,China Abstract:Characteristics of urban heat island(UHI)effect and its cause are investigated by using MODIS data in April 2004.Surface parameters fr0m the MODIS data have surface temperature(fs) albedo( ,and normalized diference vegetation index(NDVI)Their heterogeneities over urban and rural area are analyzed based on land cover classiifcation,and their relations are also presented in order to explain the UHI effect.The results show that there exists obvious the UHI effect. over urban areas are by 10.83%higher than those over rural area,and NDVI and dover urban area are by 62%and 1 8_75%less than those over rural area.respectively.Surface temperature has signiifcantly negative correlation with NDVI and their correlation coefficient is -0 73 Correlation between NDVI and albedo is determined by the spectrum of light.Diference in vegetation cover is the primary cause of the UHI effect. Keywords:remote sensing;MODIS;urban heat island;surface temperature;NDVI;albedo Introduction Urban development usually gives rise to a dramatic change of the Ea rth’S surface,as natural vegetation is removed and replaced by non—evaporating and non-transpiring surfaces such as metal, asphalt,and concrete.This alteration inevitably results in the redistribution of incoming solar radiation,and induces the urban.ruraI contrast in surface radiance and air temperatures.The diference in ambient air temperature between an urban area and its surrounding ruraI area iS known as the effect of UHI.1t iS a meteorological phenomenon developing with the growth of urbanized areas,are increasingly affecting citizen’S Iives and health.The study wilI help US to better understand the UHI aspects and lts Causes。 providing an important addition to conventional methods of monitoring the urban environment.It is important to urban expand and layout in order to Iessen the UHI. Studies on UHI phenomenon using satellite remote sensing data have been conducted for more than ten years.The methods are included the following types。Firstly,UHI actuality and its dynamic change are analyzed using long term meteorological record over urban and suburban area.Secondly。the causes of UHI characteristics are explained by the remote sensing or aerial photos.Thirdly,the numerical model is used to simulate the UHI effect.Fourthly,some stations are set to measure UHI.The recent Received on January 23,2008 No.2 WANG Gulling et al:An Assessment of Urban Heat Island Effect using Remote Sensing Data 1 5 increasing availability of remote sensing technique is an efficient way to survey UHI periodically for long time.Roth(1989)and Gallo(1993)used remote sensing techniques to compare the UHI effect to vegetation index 。。 .Owen(1 9981 used fractional vegetation cover and surface moisture availability to study the impact of urbanization in and around State College,PA .Streutker(2003)analyzed the growth of UHI and successfully quantiifed the UHI of Houston.TX taken 1 2 years apart .All the above studies used National Oceanic and Atmospheric Administration(NOAA)AVHRR data.The 1.1 km spatia} resolution of these data was found suitable for the su rface temperature mapping over urban area and used to study the surface UH1.UHI has been also studied with more fine resolution satellite data such as Landsat TM,SPOT in many cities such as the Zhujiang Delta in China,L6d之in Poland,Granada in Spain over the past few years.But such remote sensing data have low temporal resolution and a shorter data record[5-8]. MODIS is a key instrument aboard the Terra(EOS AM)and Aqua(EOS PM)satellites.Terra MODIS and Aqua MODIS are viewing the entire Earth’s surface every 1 to 2 days,acquiring data in 36 spectral bands.These data will improve our understanding of global dynamics and processes occurring on the land,in the oceans,and in the lower atmosphere.MODIS is playing a vital role in the development of validated,global,interactive Earth system models able to predict global change accurately enough to assist policy makers in making sound decisions concerning the protection of our environmentMost recent .researches were about the regional or globaI scale climatic and environmental changes and seldom on the UHI using MODIS data[9-13]. Surface temperatures derived from satellite are believed to correspond more closely with the canopy layer heat island,although a precise transfer function between the surface temperature and the near ground air temperature is not yet available.Satellite information is the only source for determing of the surface temperature(fS)at macro level over a region.In this study,we use the MODIS and other supplementary data to study the characteristics of the UHI and its causes. evergreen forest 32 2 deciduous forest shrupIandS 32 O A B grasslands croplands urban and built—up 31 8 cropland/vegetation mosaic barren or sparsely vegetated 1184 118 8 Longitude/。E 119 2 Fig.1 Land cover classiicatifon from MODIS in 2004 and locations of automatic weather station 16 Marine Science Bulletin Vl01.10 MODIS data are pretreated firstly by the method of ovedapped mobile window in Geo—statistics to eliminate abnormal values due to cloud[14,15】MODIS land surface temperature products are compared .with monthly mean fs from seven Automatic Weather Stations(AWSs)to validate them.Then,the parameters in three broadbands(0.3-0.7 lJm.O.7-5.0 IJm,and 0.3—5.0 m)are used to compute the surface albedos in the broadbands with BRDF mode1.Finally。the diference and correlation analysis of , NDVI and atbedos between the urban and rural area are performed depending on land use pa ̄ern.The urban heat island in the study area is synthetically analyzed by comparing the distribution of ,NDVI and albedos over Urban areas and non.Urban areas. 1 Studyarea The study area considered in this paper encompasses an area of 83×83 km2 centered at coordinates 1 1 8.8。E,32.0。N(Nanjing,China).This area includes Nanjing urban area and its suburban counties.Most of the terrain heights are very low in the region and mountains are distributed mainly in the east and southwest areas. The study area is very heterogeneous,comprising of water,cropland,and forest as well as urban and built-up.Fig.1 shows the land cover classification map in study area(2004)and the locations of seven AWSs.Fr0m Fig.1 it is easily seen that the Yangtze River runs across the urban zone。and the Xuanwu Lake and the Purple Mountains locate in the central part of the area。with large green cover of semi-natural forest.A transect crossing the centre of the area(AB)is chosen fr0m west to east. 2 Data set description 2.1 Satellite remote sensing data Satellite remote sensing has potential utility for this landscape scale characteristic analysis because of the global continuity of observation.Here we choose MODIS satellite data.A series of high-level land surface products have been generated by the MODIS land science team,which are all operational and an be downlcoaded frOm the following web site:http://edcimswww.cr.usgs.govlpublimswelcome/. Tab.1 Selected MODIS produc ̄of 2004 and their spatial and temporal resolutions The selected products are listed in Tab.1.ESDT is the short name for Earth science data type.DOY No.2 WANG Guiling et a1.:An Assessment of Urban Heat Island Effect using Remote Sensing Data 17 means Julian day of the year.Temporally,these data cover April 2004(spring season).We have included MODIS data from the period March 21 (DOY 81)to May 8 (DOY 121).The time period covers 4O days rifve 8-day periods and three 1 6一day periods).The 96-day land cover product included here does not specifically overlap with the spring season.The product“MOD1 1A2”in the MODIS Land Discipline is defined as the“land surface temperature(LST)”over the global land surface every 8 days with 1 km—resolution.Other products used include the land cover product“MOD12Q1”and NDVI product “MOD1 3A2”. The accuracy speciifcation for MODIS LST iS 1 K at 1 km resolution under the clear sky conditions.1t can be validated by field measurements over flat uniform Iand surfaces.The accuracy speciifcation for Iand.surface emissivity retrieved fr0m MODIS data iS 0.O2 f0r bands 29,31 and 32,and 0.05 f0r bands 20, 22.and 23.The MOD1 3A2 are standard products designed to be fully operational at launch.The MOD12Q1 product provides a suite of land covers with the primary classiifcation in the IGBP(International Geosphere Biosphere Programme)scheme.Each of these classiifcation schemes is accompanied by the assessments of its quality or confidence.More information about the MODIS products call be found from the web site http://modis.gsfc.nasa.gov/data. All data are in HDF(Hierarchical Data Format)format and are provided in ISG(an Integerized Sinusoidal Grid1 projection.While the projection becomes increasingly sheared with distance fr0m the Greenwich meridian.the data are converted to Lambert conformal projection and the format from original HDF to a flat binary format by the MODIS Reprojection Tool(MRT v2.4 Beta). 2.2 Meteor0IogIcal Data In this study,seven AWSs are selected and their locations are listed in Tab.2.Monthly surface temperature in each hour in April 2004 are obtained fr0m the AWSs.STI D is the station code number of AWS. 1-ab.2 Locations of the automatic weather Stations 18 Marine Science Bulletin Vo1.1O 3 Surface temperature from MODIS product and its validation Some hot spots,or urban heat islands,can be easily identiifed(shown in Fig.2).The terrain height is also drawn in the same map jn order to ex plain the distribution pattern of the surface temperature.The z, n】口q一 most extensive UHIiS distributed Ieft of the centraI part of the area which iS business zone and can be seen in the centre of some smalI town around the area.However.no extensive UHl exists in the east of the city,where the Purple Mountains are located.Forest and agricultural landcover there restrict the development of UHI.The maximum diference of ts between urban and suburban iS about 1 0 oc.and the average diference of fs iS about 2.29。C in daytime. Interpreting thermal data and images of distribution over an area is often not an easy job,because of many complex factors involved.The most influential factors affecting the heat island are the distribution of surface cover characteristics,and urban morphology,such as building materials,geometry,and density Each component surface in urban landscapes possesses a unique radiative,thermal,moisture,and aerodynamic properties,and relates to their surrounding environment.The myriad of the component surfaces and the spatial complexity,when mosaicked,create a limitless array of energy balance and microclimate systems.confiscating urban meteorologists fr0m drawing any general conclusion[ZJ Land covers are more effective In developing UHI. 118 8 Longitude/0E 13 17 21 Fig-2 Terrain height(m)and monthly average surface temperatures f from MODIS at about 1 0:30 LT(Iocal time1 in daytime in April 2004。 The monthly average fS from AWS and MODIS in daytime are compared.Tab.3 is the comparison of the monthly average fs derived from MODIS with AWS valid at 1 0:30 hour local time in day.Ats is the : 『A Q 鱼et al-.A—n—Assessment of Urban Heat Island Effect using Remote Sensing Data 1 9 difference of fS between MODIS and AWS.The monthly average fS derived fr0m MODIS are generally in good agreement with those from AWSs.The relative error is about-8.76%一2.39%.most of which is negative.It can be concluded that the monthly average from MOD S is somewhat lower than the values ofAWS. Tab.3 Comparison of monthly average fs at about 10:30 LT in April 2004 from MODIS with AWS 4 Relations of Tsand NDVI 4.1 NDVl This research uses NDVI to analyze thedistribution of temperature in green area and non.green .area.Generally speaking,NDVI has the value of 0(Min)-1(Max)through the following equation. NDVI=( f,一r,)/( + )) r.ir and are surface reflectances in the NIR and red bands,respective ̄ 吾32.D 31 8 118 8 Longitude/0E 一0 1 O 0 1 0 2 0 3 04 0.5 0 6 0 7 Fig.3 Monthly average NDVI from MODIS in April of 2004 Marine Science Bulletin Vo1.10 It carl be seen fr0m the NDVI map(Fig.3)that all the urban or built—up areas have a relatively low NDVI,ranging from 0.2 to 0.4.The same is true in small towns,which are related to the sparse vegetation. The mean diference of NDVI between urban and ruraI area is 0.31. 4.2 albedo(0‘) The parameters in three broadbands(0.3—0_7l|m,0_7—5.0 lJm.and 0.3—5.0 pm)are used in this paper in a fo rward version of the model to compute the integrated black—sky(at some solar zenith angle) and white-sky albedos.Alternately,the parameters can be used with a simple polynomial to easily estimate the black-sky albedo(completely direct beam)with good accuracy for any desired solar zenith angle l,J_The polynomial is as follows: I ( , ): 。( )(g。哪+g. 。 +g 。 )+ ( )(g。 ,+g。 , +gz )+ ( )(go +gt +gz ) r11 、。, where , are the solar zenith angle,the spectral band.Similarly,the white—sky albedo(completely difuse beam)can be computed by using the equation: ( ): 。( )g 。+ 。,( )gv0,+/ (2)gg ̄o (2) The appropriate constants are from the MODIS BRDF/Albedo Product(MOD43B)User's Guide as mentioned above.The albedo, (A, ,under actual atmospheric conditions can also be modeled quite accurately as an interpolation between the black—-sky albedo and white--sky albedo as a function of the fraction of difuse skylight s( ̄i,彳( )),which,in turn,is a function of optialc depth彳( ). a(A, 。):(卜 ( ,f( ))口 ( , )+ ( , ( ))口w ( ) (3) The fraction of difuse skylight is calculated through a look—up table downloaded fr0m http://geography.bu.edulbrdfluserguideltools.htm1.which was precalculated with the 6S code for seven MODIS land spectral bands plus three broadbands。The optialc depth r(A)is assumed ofr the three broadbands as a constant 0.5.and the solar zenith angle is at IocaI solar noon.The error of albedo aused by (c )is very small,about O.003[191.The urban albedo is the area—averaged values over all urban areas,and the rural albedo is the area—average albedo over other land cover type areas.Fig.4 is the monthly average albedo in total broadband(0.3—5.0 pm)fr0m MODIS in April 2004.The average albedo over urban area is about 0.1 3 jn the study.which is consistent with other researches.Validating albedo products is important because their accuracy is critiaIc to the scientiifc community for their applications.Some scientists have performed initiaI validations of land surface albedo products based on the up-scaling methods.Some preliminary validation results show that these products are reasonably accurate.with typialc absolute error of Iess than 5%I J. 4。3 Analysis of urban heat island To show the distribution of the urban heat islands,albedos in three broadbands(0.3—0.7 IJm,0_7一 No.2 WANGGulling et a1.:AnAssessmentOfUrban Heat sand Efect usingRemoteSensingData ——5.0 pm,and 0.3.5.0 pro),fs and NDVI in daytime variations along the transect AB(Fig.1)are drawn in Fig.5.It should help US to understand how industriaI development and jnfrastructure construction have altered the urban thermal characteristics.It exhibits numerous“peaks,”“valleys,”“plateaus,”and“basins,’’ indicating the heterogeneous nature of surface temperature along the line.Factors such as the spatial pa ̄ern of diferent land.cover classes.the occurrence of water bodies,building,and the division of the z。『8D暑 -1 city’S functional districts,among many others,may have affected the development of UHI.Visually reviewing the variations among the land covers reveals that the fs“peaksI’occur over urban and building areas.The UHl pa ̄ern becomes visible。 32.0 31.8 118.4 1188 1192 Longitude/0E Fig.4 Monthly average albedo in total broadband(0.3-5.0 pm)from MODIS in April 2004 AIbedo over urban area iS IeSS than that over the rura1.Correlation between albedo in near infrared broadband(0.7—5.0 pm)and NDVI iS positive,their correlation coefficient is 0.83,but it is-0.80 in the visible(0.3-0.7 pm).Albedo in the visible over urban area is larger than that over rural area,but is less in the N I R and in the total(0.3—5.0 pm).The result can be explained that the reaction of vegetation to sun light in wavelength is quite diferent.Albedo in the visible(0.3—0.7 pm)over the area with high density vegetation is lower than that over sparse vegetation.Because vegetation absorbS the light in the visible for photosynthesis,but it absorbs hardly in near infrared broadband and reflect most of it.Therefore albedo in near infrared broadband is larger.There is few vegetation over urban area and NDVI over urban area is always less than that over the rura1.We can see fr0m the figure that there is a negative correlation between NDVI and fs which has also been obsewed by several researchers for various types of ecosystems[14-17].The correlation coefifcient between and NDVI reaches.0.73.NDVI presents spectral reflectance properties and densiyt of vegetation,surface temperature,on the other hand,is related to water and energy balance determined by soil moisture availabiliyt,amount of evapotranspiration and local climate factors. Marine Science Bulletin Vo1.1.0 The inhomOgeneitles of urban/rural surface characteristics(such as ts,albedo,NDVI)are analyzed based on MODIS data.and conclusions are shown in Tab.4. 、 U 4 O 3 O 2 0 l O 的 他 Longitude/。E 一一 。咐 口n 一 f —NDVl (a) 0 8 一118.4 1188 0 6 04 舌 0:2 Z 0 119 2 Longitude/。E 一 — 一NDVI _0 2 8 6 4 2 ∽。 . 2 (b) Fig.5 Zonal profiles of NDVI and(a)albedos in three broadbands and(b)surface temperature ts along 舌Z transect AB in Fig.1 Tab.4 Comparison of monthly average ,NDVI and over urban area with those over rural area fs over urban area increase by 1 O.83%,and diference between urban and rural area is about 2.29。C:NDVI over urban area is less than that over the rura1.On the contrary NDVI and albedo over the - urban are by 62%and 1 8.75%less than the rura1.and the mean diference between urban and rural area is about 0.31 and 0.O3.respectively. 5 Conclusion (a)£s from MODIS product well agrees with that of AWS,the and the UHI effect around Nanjing urban area is obvious by MODIS images. (b)NDVI and fs has correlation coeficifents of一0.73.Correlation coeficifent of NDVI and albedo is determined by spectrum of light,and that of albedo in near infrared broadband(0.7 pm一5.0 pm)and No 2 WANG Gulling et al:An Assessment of Urban Heat Island Effect using Remo ̄Sensing Data 23 NDVI is 0.83,while that is-0.80 in the visible(0.3pm一0.7 pm) (c) over urban area is larger than that over the rural area.fs diference between urban and rural area is about 2.29。C.and NDVI and albedo over the urban area are by 62%and 1 8.75%less than those over the rural area.and the mean diference between urban and rural area is about 0.31 and 0.03, respectively. The results indicate that surface characteristic parameters derived frOm remote sensing data will be regarded as one of effective methodologies to manifest urban heat island.The vegetation cover appears tO be a primary cause of surface properties between two environments that are responsible for the diference between urban and ruraI surface.It can decrease the surface temperature over urban area and lessen effectively the UHI effect by enlarging vegetation cover and albedo.This study can also be the foundation for the further research on the UHI formation mechanism in term of the surface energy exchange using remote sensing data. Acknowledgments This work was supported by the project of National Natural Science Funding of China under grant No.40075004. Reference 【1】 Roth M,Oke T R,Emery W J.Satellite derived urban heat islands from three coastaI cities and the utilization of such data in urban climatology【J】.International Journal of Remote Sensing,1 989,1 0(1 1):1 699—1 720. Gallo K P The use of a vegetation index for assessment of the urban heat island effect[J】.International Journal of Remote Sensing,1993,14(11):2 223—2 230. 【3】 Owen T W,Carlson T N,Glilies R R.An assessment of satellite remotely—sensed land cover parameters in quantitatively describing the climatic effect of urbanization[J】.International Journal of Remote Sensing,1998,19 (9):1 663—1 681. Streutker D R.Satellite-measured growth of the urban heat island of Houston,Texas[J】.Remote Sensing of Environment,2003。85(3):282-289. [55】 Klysik K,Fo ̄uniak K.Temporal and spatial characteristics of the urban heat island of L6d乏.Poland[J】. Atmospheric Environment,1 999,33(24/25):3 885-3 896. Monta0ez J P,Rodriguez A,Jim ̄nez J I.A study of the urban heat island of Granada[J】.International Journal of Climatology,2000,20(8):899—91 1. Weng Q.Fractal analysis of satellite—detected urban heat island effe ̄【J】.Photogrammetric Engineering& Remote Sensing,2003,69(5),555—566. 【81 Weng Q,Lu D,Schubring J.Estimation of land sudace temperature-vegetation abundance relationship for urban heat island studies[J】.Remote Sensing of Environment,2004,89(4):467—483. Zhu A,Li C,Liu G,et al Validation and application of MODIS remote sensing aerosol information in Beijing【J】. 24 Marine Science Bulletin Vo11O .Acta Scientiae Circumstantiae,2004,24(1):86—9O. 【1 O】 Keith W O,Gordon B B,Schaaf C B,et a1.Assessment of global climate model land surface albedo using MODIS data【J】.Geophysical Research Letters,2003,30(8):1 443—1 447. 【11】 Wei M,Wang G,Jiang W M,et a1.Surface temperature analysis in urban area using MODIS[J】,Proceeding of SPIE.2006.61 99:32・40 【121 Wang G,Jiang W,Wei M,et a1.Analysis of urban/rural albedo from MODIS over East China[J】,Proceeding of SPIE,2006.6199:123—134. 【1 3】 Wang G.Study on Satellite Remote Sensing Parameterization of Surface Characteristics and Atmospheric Boundary Layer over Urban/Rural Areas【D1,Nanjing:Department of Atmospheric Sciences Nanjing University, 2005. 【14] wang z,Geo-statistics and its application in ecology【M】Beijing:Science Press.1999. 【1 5] Sun H,Sun Z,Li Meteorological Satellite Remote-sensing Spectral Characteristics of Fog[Jl,Journal of Nanjing Institute of Meteorology,2004,27(3):289・301. 【16] Yang J,Yin Q,Gu S.Band combination research of city region hyperspectral remote sensing data in false color synthesis[J],Journal of Nanjing Institute of Meteorology,2005,28(3):289—295 【17]Lucht W,Schaaf C B.Strahler A H.An algorithm for the retrieval of albedo from space using semi-empirical BRDF models【J】.IEEE Trans Geosci Remote Sens.2000,38(1):977・998 【1 8】Luo Y.Characteristics of atmospheric aerosol optical depth and its simulation of forcing on direct radiation and its climate effect over China[D],Beijing:Department of Geophysics Beijing University,1 998. 【19]Wang K,Liu J,Zhou X,et a1.Retrieval of the surface albedo under clear sky over china and its characteristics analysis by using MODIS satellite[J】.Chinese Journal of Atmospheric Sciences,2004,28(6):941—949. [20】Liang S.Validating MODIS land surface reflectance and albedo products:methods and preliminary results【J】. Remote Sensing of Environment,2002.83(1/2):149—162. 【21】Xu X.A study of monthly—mean surface albedo of the Yangtze River Delta[J]Chinese Journal of Atmospheric Sciences,2002,26(3):394-400. 【221 Zhou S.Shu J.Urban Clima ̄logy.Beijing:Meteorological Press 1994. [23】Nemani R.Running S.Estimation of regional surface resistance to evapotranspiration fr0m NDVI and thermal-IR AVHRR data【J】.Journal of Applied Meteorology,1 989,28(4):276—284. [241 Smith R C G,Choudhury B J.Analysis of normalized difference and surface temperature observations over southeastern Australia【J】.International Journal of Remote Sensing,1991,12(1O):2 021—2 044. [25】 Boegh E,Soegaard H,Thomsen A,Evaluating evapotranspiration rates and surface conditions using Landsat TM to estimate atmospheric resistance and surface【J】.Remote Sensing of Environment,2002,79(2/3):329- 343. 【26】 Yang B,Ding Y.An Experiment on Soil Moisture Monitoring Using an Improved Thermal Inertia Method[J], Journal of Nanjing Institute of Meteorology,2004,27(2):218—223. No.2 WANG Guiling et a1.:An Assessment of Urban Heat Island Effect using Remote Sensing Data 25 城市热岛效应的卫星遥感分析 王桂玲’, 蒋维楣2,魏鸣3 (1.解放军理工大学气象学院,江苏南京 211101:2.南京大学大气科学系,江苏南京210093 3.南京信息工程大学中美合作遥感中心,江苏南京210044) 摘要:利用MODIS资料研究了2004年4月南京城市热岛特征及其影响因子,结合地表覆盖类型分析了植被 归一化指数(Normalized Diference Vegetation Index,NDVI)、地表温度( )、地表反照率( 的城乡差异及其相 互关系,探讨了城市热岛(Urban Heat Island,UHI)效应形成的机制。结果表明:南京城区存在着明显的城市热 岛效应;城市平均 比乡村高约10.83%;城市NDVI和 分别比乡村低约为62%和18-75%;NDVI与『s呈 负相关,相关系数为-0.73,而NDVI与 之间关系与波段有关;城乡植被覆盖差异是造成UHI的主要原因, 其次是地表反照率,通过提高植被覆盖率和地表反照率可以减小城市热岛效应。 关键词:卫星遥感;MODIS;城市热岛;地表温度;植被归一化指数;地表反照率