Estimation of the all-wave urban surface radiation balance by use of ASTER multispectral imagery and in situ spatial data

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D18, 4582, doi: /2003jd003396, 2003 Estimation of the all-wave urban surface radiation balance by use of ASTER multispectral imagery and in situ spatial data Nektarios Chrysoulakis Foundation for Research and Technology-Hellas, Regional Analysis Division, Institute of Applied and Computational Mathematics, Herklion, Crete, Greece Received 10 January 2003; revised 16 April 2003; accepted 12 June 2003; published 23 September [1] In this study, Terra/Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) imagery was analyzed together with in situ spatial data to examine the potential of high spatial resolution multispectral remote sensing to support the definition of energy exchanges in urban environment. The spatial distributions of physical parameters with relation to the urban energy budget were determined, and finally, taking into account the effects of different types of built-up areas and different land uses, the spatial distribution of all-wave surface net radiation balance for the center of the metropolitan city of Athens was estimated within about ±44.5 Wm 2. The atmospheric downward longwave flux was estimated with the use of radiosonde measurements. The Santa Barbara DISORT atmospheric radiative transfer model was used for the simulation of the radiative transfer in the atmosphere. The results indicate that ASTER multispectral imagery enables a better understanding of energy aspects and their causes and effects, providing an important addition to conventional methods of monitoring the urban environment. INDEX TERMS: 3322 Meteorology and Atmospheric Dynamics: Land/atmosphere interactions; 3359 Meteorology and Atmospheric Dynamics: Radiative processes; 3360 Meteorology and Atmospheric Dynamics: Remote sensing; 7538 Solar Physics, Astrophysics, and Astronomy: Solar irradiance; KEYWORDS: remote sensing, urban environment, energy budget Citation: Chrysoulakis, N., Estimation of the all-wave urban surface radiation balance by use of ASTER multispectral imagery and in situ spatial data, J. Geophys. Res., 108(D18), 4582, doi: /2003jd003396, Introduction [2] The main driving force of the Earth system is radiation forcing. A detailed and quantitative knowledge of the Earth radiation field is crucial for understanding and predicting the evolution of the components of the Earth system. Surface energy flux measurements play an important role in understanding energy exchanges in urban environment. Urbanization has led to a high increase of energy use. Increased urban temperatures have a direct effect on the energy consumption of buildings increasing the electricity demand and the production of carbon dioxide and other pollutants. It has been reported that the rate of change in energy use is twice the rate of change in urbanization [Santamouris et al., 2001]. [3] The use of satellite imagery allows to instantaneously observe large areas providing an important addition to conventional methods of monitoring urban environment. Satellite-derived surface temperature data have been utilized for urban climate analysis in several studies [Price, 1979; Vucovich, 1983; Ackerman, 1985; Caselles et al., 1991; Kim, 1992; Lee, 1993; Boldo and Le Men, 2001; Stefanov et al., 2001; Small, 2002]. Temperature distribution in urban areas is highly affected by the urban radiation budget. Solar radiation incident on urban surfaces is Copyright 2003 by the American Geophysical Union /03/2003JD absorbed and then transformed to sensible and latent heat. Most of the solar radiation impinges on roofs, and the vertical walls of the buildings, and only a relatively small part reaches the ground level. Walls, roofs and the ground emit longwave radiation to the sky. The intensity of the emitted radiation depends on the view factor of the surface regarding the sky [Boldo and Le Men, 2001; Dousset and Gourmelon, 2001]. [4] Urban surface and atmosphere interactions are governed by the surface heat fluxes, which are directly related to the all-wave surface net radiation balance (NRB). The distribution of these fluxes is drastically altered by urbanization. Main contributing factors are: changes in physical characteristics of the surface (albedo, thermal capacity, heat conductivity) due to the replacement of vegetation by asphalt and concrete; decrease of surface moisture available for evapotranspiration; changes in the relative fluxes and in the near surface flow due to complex geometry of streets, tall buildings and anthropogenic heat. [5] The objective of this study is to examine the potential of high resolution multispectral remote sensing to support the definition of energy exchanges in urban environment by estimating the spatial distribution of surface NRB for a metropolitan city center with the use of Terra/ASTER imagery and in situ spatial data. The entire energy budget cannot be calculated solely from satellite imagery because for the estimation of the atmospheric contribution the vertical profiles of temperature and water ACL 8-1

2 ACL 8-2 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY Table 1. Spectral Ranges of ASTER Channels Subsystem Channel Spectral Range, mm VNIR N B SWIR TIR vapor are needed. In this study, the net horizontal advection has been ignored because the horizontal scale is of the order of 10 km (local scale). Therefore the transformation of the solar energy at the surface was considered in one dimension. The all-wave surface NRB (Wm 2 ) is given as a balance between the shortwave radiation intake plus the atmospheric downward longwave flux and Earth s thermal emission: NRB ¼ ð1 a short ÞE þ F # F " ; ð1þ where a short is the surface total shortwave albedo; E is the shortwave irradiance (direct and diffuse) on the surface (Wm 2 ); F # is the atmospheric downward longwave flux (Wm 2 ); and F " is total surface radiant exitance (Wm 2 ). 2. Data and Methodology [6] The data used in this study is an ASTER image, acquired on 30 October 2001 at LST, over the region of Attica, Greece and vectors of the road network for the city of Athens, derived from in situ spatial information. ASTER is an advanced multispectral imager on board NASA s Terra spacecraft. It covers a wide spectral region with 14 bands from the visible to the thermal infrared with high spatial, spectral and radiometric resolution. It consists of three separate instrument subsystems, each operating in a different spectral region and each using a separate optical system. These subsystems are the visible and near infrared (VNIR), the short wave infrared (SWIR) and the thermal infrared (TIR). The spatial resolution varies with wavelength: 15 m in the VNIR, 30 m in the SWIR and 90 m in the TIR. The VNIR subsystem consists of two telescopes, one nadir looking with a three-band detector (channels 1, 2 and 3N) and the other backward looking (27.7 off-nadir) with a single band detector (channel 3B). The SWIR subsystem consists of one telescope with a six-band detector (channels 4, 5, 6, 7, 8 and 9) and the TIR subsystem consists of one telescope with a five band detector (channels 10, 11, 12, 13, and 14). The spectral ranges of ASTER channels are given in Table 1 [Abrams and Hook, 2001]. [7] Athens, the capital of Greece contains about half of the population of the country. The area under study is the center of city and it is presented in Figure 1 as a pseudo-colored Figure 1. The study area as ASTER pseudo-colored composition red-green-blue: 3N-2-1. The vegetation areas are shown in red, the built-up areas in blue tones, and the main road network in black tones.

3 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY ACL 8-3 composition red-green-blue: 3N-2-1 of ASTER channels 1, 2 and 3N. Vegetation areas are presented in red because of strong reflection of vegetation in near infrared. The built-up areas are presented in blue tones due to the reflection of building materials in visible spectrum, whereas the main road network is in black tones because the asphalt strongly absorbs the incoming solar radiation. [8] It has been recognized that surface albedo is among the main radiative uncertainties in current climate modeling. Broadband surface albedo is usually estimated from broadband satellite sensors [Jacobowitz et al., 1984; Smith et al., 1986; Kandel et al., 1998]. However, ASTER narrowband multispectral observations have much finer spatial resolution that allow to characterize surface heterogeneity. For this reason, ASTER imagery was used in this study for the estimation of the spatial distribution of the shortwave ( mm) broadband albedo (a short ) at the center of Athens. The nine NVIR and SWIR channels allow to effectively convert narrowband to broadband albedos. The linear conversion formula proposed by Liang [2000] was used: a short ¼ 0:484a 1 þ 0:335a 3 0:324a 5 þ 0:551a 6 þ 0:305a 8 0:367a 9 0:0015; ð2þ where a i (i = 1, 3, 5, 6, 8, 9) are the spectral albedos of the respective channels. [9] The surface spectral albedo, defined as the ratio of upwelling to downwelling spectral irradiance at the surface, can be estimated when the reflected solar energy is known for each spectral channel. The reflected energy was derived from the total spectral radiance measured by each channel (radiance at the sensor) by applying an atmospheric correction procedure. The radiance at the sensor for each pixel was calculated from the digital numbers (DN) of VNIR and SWIR channels. The unit conversion coefficients defined as radiance per 1 DN [Abrams and Hook, 2001] were used to convert from DN to radiance at the sensor. The radiance at the sensor was related to the surface spectral reflectance and the incoming shortwave irradiance using the equation [Lillesand and Kiefer, 2000]: L i ¼ a ie i t i p þ L pi; ð3þ where i is the ASTER channel (i = 1 to 9); L i is the radiance at the sensor (Wm 2 sr 1 mm 1 ); a i is the surface spectral albedo; E i is the total (direct and diffuse) solar irradiance (Wm 2 mm 1 ); t i is the transmission of the atmosphere; and L pi is the atmospheric path radiance (Wm 2 sr 1 mm 1 ). [10] As long as the unknowns E i, t i and L pi are determined, it is straightforward to calculate surface spectral albedo for each pixel by using equation (3). For the estimation of the aforementioned parameters the Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART) model [Ricchiazzi et al., 1998] was used. The central coordinates of the study area and the ASTER acquisition time determined the solar zenith and azimuth angles (64.14 and respectively). An urban aerosol model [Ricchiazzi et al., 1998] with horizontal visibility higher than 30 km at 0.55 mm was used. Moreover, typical trace gases concentrations were Table 2. Albedo and Thermal Emissivity Mean and Variance Urban Surface Type Albedo Thermal Emissivity Roads 0.07 ± ± 0.03 Mixed areas 0.12 ± ± 0.04 Vegetated areas 0.16 ± ± 0.03 Medium built areas 0.20 ± ± 0.03 Densely built areas 0.25 ± ± 0.03 used (360 ppm for CO 2, 1.74 ppm for CH 4 and 0.32 ppm for N 2 O), whereas the method proposed by Cartalis and Chrysoulakis [1997] was applying for the calculation of precipitable water (PW) values with the use of radiosonde measurements. Radiosonde data were available from the National Meteorological Service (Hellinicon station: N, E). [11] There are five standard surface types in SBDART for the parameterization of the surface spectral albedo. Instead of using these surface types, the urban land cover was analyzed on the basis of ASTER imagery and the following urban surface types were used in SBDART calculations: (1) roads; (2) densely built areas; (3) medium built areas; (4) vegetated areas and (5) mixed areas. The latter correspond to open areas where asphalt, concrete, scarce vegetation and bare soil coexist, resulting in subpixel variations of the surface cover. SBDART simulation was started using the albedo and thermal emissivity values which are shown in Table 2 [Lillesand and Kiefer, 2000]. [12] Urban areas involve spectrally heterogeneous land cover types, making impractical a correct classification based solely on spectral information as in traditional classification algorithms. The heterogeneity and small spatial size of the surface materials lead to significant subpixel mixing [Ridd, 1995; Foody, 2000]. The location of each surface type in the study area was determined by combined use of ASTER imagery and in situ spatial data. More specifically, road network vectors were used to mask the VNIR image by setting the pixels corresponding to roads to the value of zero. Vegetated areas were defined on the basis of NDVI (Normalized Difference Vegetation Index), which was calculated from DN of channels 2 and 3N using the formula: NDVI ¼ DN 3N DN 2 DN 3N þ DN 2 : [13] An experimentally derived threshold was applied to NDVI values to separate pixels corresponding to vegetation in the raw image. Following, a digital mask was applied by setting these pixels to the value of 255 (ASTER VNIR raw data are coded in 8 bits, thus 255 is the maximum value). This double masking process resulted in the identification of the pixels corresponding to roads, as well as of the pixels corresponding to urban vegetation. Next, an unsupervised classification and a postclassification sorting were applied to classify the remaining pixels in three categories: Densely built areas, medium built areas and mixed areas (mainly unclassified pixels). Figure 2 shows the five classes of urban surface types used in SBDART simulation. [14] SBDART can be run without using real surface reflectance spectra. According to Liang [2000], it was run three times with three different surface reflectance values (0.0, 0.5 and 0.8) for the specific atmospheric conditions ð4þ

4 ACL 8-4 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY Figure 2. The five urban surface types used in this study. and solar geometry. Results from the three runs of the SBDART allowed to determine the atmospheric transmittance and the atmospheric path radiance for each channel using the equation (3). The derived t i and L pi values are shown in Table 3. As long as t i and L pi are known, the spectral albedo can be calculated using the formula: a i ¼ p L i L pi : ð5þ E i t i [15] The only unknown in equation (5) is now the E i, which also depends on the surface type. Therefore E i was calculated by running the SBDART for each channel and for each surface type (45 runs). Next, the spatial distribution of each spectral albedo was estimated using equation (5) and finally, the broadband shortwave albedo was calculated using equation (2). [16] Temperature is not an intrinsic property of the surface; it varies with the irradiance history and meteorological conditions. Emissivity is an intrinsic property of the surface and it is independent of irradiance. Brightness temperature differs from actual surface temperature due to several effects: (1) partial absorption of blackbody radiation by the atmosphere; (2) land surface emissivity being less than unity and spatially and spectrally variable [Becker, 1987]; (3) subpixel variations of surface temperature being averaged nonlinearly due to Plank s law [Dousset et al., 1993]; (4) urban geometry trapping radiated and incident energy in urban canyons, effectively increasing the pixel average emissivity; and (5) nonvertical satellite viewing angles biasing toward vertical wall and hiding horizontal surfaces [Voogt and Oke, 1997]. [17] Surface temperature can be estimated using thermal infrared satellite data provided that the variability of the emission coefficients as well as the atmospheric absorption are taken into account [Price, 1984; Becker and Li, 1990; Vidal, 1991; Sobrino et al., 1991; Kerr et al., 1992; Ottle and Vidal-Madjar, 1992; Coll et al., 1994; Prata, 1993; Prata et al., 1995; Caselles et al., 1997; Chrysoulakis and Cartalis, 2002]. [18] Assuming uniform atmospheric transmissivity for all TIR ASTER channels, the radiance at the sensor (L ij ), which is recorded at pixel j and channel i, is a function of the radiance emitted from the surface [Gillespie et al., 1998]: S # L ij ¼ t e ij L B l i ; T j þ t rij p þ S "; where t is the atmospheric transmissivity in thermal infrared; S # is the downwelling atmospheric irradiance (Wm 2 mm 1 ); S " is the upwelling atmospheric path radiance (Wm 2 sr 1 mm 1 ); ij is the emissivity, in wavelength l i, of the surface corresponding to pixel j; r ij is the reflectivity, in wavelength l i, of the surface corresponding Table 3. SBDART-Derived Atmospheric Transmittance and Path Radiance Over Athens a ASTER Channel Atmospheric Transmittance, % ð6þ Atmospheric Path Radiance, Wm 2 sr 1 mm N a 30 October 2001, at LST.

5 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY ACL 8-5 to pixel j (r ij and ij are relate by Kirchhoff s law: r ij =1 ij ); L B (l i,t j ) is the equivalent blackbody radiance at wavelength l i and temperature T j (Wm 2 sr 1 mm 1 ); l i is the central wavelength of the ASTER TIR channel i; and T j is the kinetic temperature of the surface corresponding to the pixel j. [19] Incident radiance from adjacent pixels is typically less than S # and thus can be ignored. According to Gillespie et al. [1998], typical values of the atmospheric variables estimated by the MODTRAN3 atmospheric model can be used in equation (6): t 70%; S " 2.4 Wm 2 sr 1 mm 1 and S # 11.7 Wm 2 mm 1. equation (6) indicates that if radiance is measured in N spectral channels, there will be N + 1 unknowns: the N emissivities ij and the surface temperature T j. Thus additional information is needed to extract either the temperature or emissivity information. [20] A complicated temperature and emissivity separation algorithm has been proposed by Gillespie et al. [1998], able to recover temperatures within about ±1.5 K and emissivities within about ±0.015 K. It is difficult to achieve such accuracy in urban environmental conditions, where the surface heterogeneity causes subpixel variations of temperature and emissivity. Therefore for the estimation of temperature spatial distribution, the atmospherically corrected channel 13 records were used in combination with the emissivity values assigned to each urban surface type (Table 2). Therefore equation (6) can be written as: L ij þ 2:59 e k 0:19 L B l i ; T j ¼ ; ð7þ 0:7 e k where k indicates the respective urban surface type (k = 1 to 5) and i = 13. [21] The equivalent blackbody radiance (Wm 2 sr 1 mm 1 ) at wavelength 10.6 mm and temperature T j was calculated for each pixel j using equation (7); in the sequel the spatial distribution of temperature was estimated. The central wavelength method was used as the most straightforward approximation since Plank s equation cannot be inverted explicitly. This method is accurate for the observed radiance that yields the standard temperature, but it becomes increasingly inaccurate for temperatures further away from the standard temperature [Alley and Jentoft-Nilsen, 1999]. Finally, the Stefan-Boltzmann law was used for the estimation of the spatial distribution of the total surface radiant exitance. [22] It is important to know the value of the downward flux of the longwave atmospheric radiation (F # ) at Earth s surface because is a significant factor for the surface energy budget. Many empirical and theoretical methods have been developed, by means of which the F # at the surface can be calculated using standard surface observations of temperature and humidity or upper air data [Swinbank, 1963; Idso and Jackson, 1969; Exell, 1978]. In this study, F # under clear sky in Athens was calculated using the method proposed by Atwater and Ball [1978]: F # ¼ Z 1 0 s T 4 ðz z0þ 0 dz 0 ; ð8þ where s is the Stefan-Boltzmann constant; T is the atmospheric temperature at height z 0 ; and P is the transmission function. [23] According to Pissimanis and Notaridou [1981] the transmission function can be calculated based on the fact that that the CO 2 emission may be added to that of H 2 O. Thus: P(z, z 0 ) = 1 ( CO2 (z, z 0 ) + H20 (z, z 0 )), where CO2 (z, z 0 ) is the emission function of the CO 2 and H20 (z, z 0 ) is the emission function of the H 2 0. Emission functions are defined using empirical relations when PW and CO 2 content U (g/cm 2 ) are known. For the calculation of H20 (z, z 0 ) the atmospheric column from surface to 500 hpa was divided in 11 layers with p = 50 hpa. For each layer, temperature and water vapor partial pressure were assumed to be constant and therefore PW was calculated from radiosonde data using the P11 1 formula: PW ¼ g ðm i p i Þ, where g is the acceleration i¼1 of gravity, M i is the mixing ratio and p i is the pressure difference between the top and the base of the i atmospheric layer [Cartalis and Chrysoulakis, 1997]. Using the Atwater and Ball [1978] method, H20 (z, z 0 ) was determined from PW values for each layer using a lookup table, whereas CO2 (z, z 0 )=1 exp( U 0.4 ). Measurements of the concentration of CO 2 were not available, so a typical concentration of 360 ppm constant with height was considered for CO 2. After dividing the atmospheric column 11 layers, equation (8) can be written as: " # F # ¼ s X11 T i þ T 4 iþ1 ðp i P iþ1 Þ ; ð9þ 2 i¼1 where T i is the temperature and P i is the transmission function of the layer extended from surface to the level i. 3. Results and Discussion [24] The distribution of VNIR irradiance at the surface, for a spectral albedo range 0 80%, is presented in Figure 3a. The linear relationship is evident in all cases, but the slope increases as the wavelength decreases. This means that the surface contribution to the downward flux is higher in shorter wavelengths. The intercept with vertical axis indicates the downward flux without surface contribution for each channel. SWIR irradiance at the surface is practically independent from the spectral albedo as it can be seen in Figure 3b. [25] The spatial distribution of the broadband shortwave albedo is shown in Figure 4. Densely and medium built areas yield higher a short values, consequently they absorb lower percentages of incoming solar energy than their surroundings. It should be noticed that there are some specific areas in which the spatial distribution of a short reaches its maximum value. It is difficult to determine what every pixel contains when the urban environment is concerned. However, in some cases these maxima can be easily explained, as for example in areas A and B, which correspond to Panathenaic stadium and Acropolis, respectively. Both contain ancient constructions, thus the white marble (a hort > 50%) dominates in these areas. High albedo values in other specific areas may be attributed to: (1) white paints, mainly in the industrial area (left part in Figure 4) and (2) bare soil at the Olympic Games of 2004 construction sites. [26] SBDART was used to estimate the spatial distribution of absorbed shortwave energy at the surface. Initially, total

6 ACL 8-6 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY Figure 3. Surface (a) VNIR and (b) SWIR irradiance. The intercept with vertical axis indicates the value of the downward flux without surface contribution. shortwave irradiance reaching each surface type was simulated. Further, spatial distribution values for a short were used to estimate the reflected shortwave energy for each pixel. Finally, the absorbed shortwave energy was calculated by subtracting the upwelling from the downwelling irradiance. [27] Figure 5 presents the spatial distribution of temperature for the center of Athens. The spatial distribution of the total surface radiant exitance was estimated on the basis of this temperature distribution using the Stefan-Boltzmann law and the thermal emissivity of each surface type. In general, maximum temperatures at densely built areas and lower temperatures at vegetated areas were expected. However, it is obvious in Figure 5 that maximum temperatures are observed in mixed and vegetated areas at this time of the

7 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY ACL 8-7 Figure 4. Spatial distribution of the broadband shortwave albedo for the center of Athens. White marble dominates in areas indicated by A and B; therefore high albedo values are observed there. day. Likewise, the central commercial area, as well as the extended industrial area (left) obtain higher temperature values than the densely built areas. This fact may be attributed to: [28] 1. Low solar height at the time of the satellite pass, because in densely built areas sky is partly covered up by surrounding buildings. In the streets, sky illumination is only partial and sun can be occluded creating shadow areas. Figure 5. Surface temperature spatial distribution for the center of Athens. The central commercial area, as well as the extended industrial area (left) obtain higher temperature values than the densely built areas. In mixed and medium built areas the maximum temperature values are observed along the main roads.

8 ACL 8-8 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY Figure 6. NRB spatial distribution for the center of Athens. The maximum values are observed over the densely built areas. [29] 2. The thermal capacity of the buildings, which lessens heating of densely built areas as compared to the respective heating of the medium built and mixed areas. [30] 3. The high shortwave absorptivity of the asphalt which affects mainly medium built and mixed areas, because in densely built areas only a relative small part reaches the ground level. [31] Temperature and humidity vertical profiles for 30 October 2001, derived from radiosonde data, were used to estimate the atmospheric downward longwave flux at the surface. On the basis of these profiles and assuming uniform CO 2 mixing, F # was estimated at Wm 2. As long as the spatial distributions of a short,e,f " and F # were known, the calculation of the surface NRB was straightforward from equation (1). Figure 6 presents the spatial distribution of the surface NRB for the center of Athens. NRB is positive everywhere, as it was expected, because at that time of the day the surface absorbs more energy than it emits. [32] There does not exist any operational energy balance measurement network in Athens, therefore in situ measurements, with the appropriate spatial resolution, were not available. However, incoming solar irradiance and surface temperature data were available from three meteorological stations: Station 1, National Technical University of Athens ( N, E). Station 2, National Observatory of Athens ( N, E). Station 3, National Observatory of Athens ( N, E). Station 2 is located in a vegetated area, whereas Stations 1 and 3 are located in mixed areas. Using these measurements NRB values were estimated: NRB station1 = 395 ± 25 W/m 2, NRB station2 = 353 ± 18 W/m 2 and NRB station3 = 358 ± 27 W/m 2. Consequently, a correspondence to NRB spatial distribution (Figure 6) is observed. [33] Visual inspection and comparison of Figures 2, 4, and 6 indicate that densely built areas obtain higher NRB values than medium built or mixed areas. The densely built areas obtain higher albedo values than open areas in which asphalt or vegetation dominates. For this reason densely built areas absorb lower parts of incoming solar energy than their surroundings and therefore NRB values should be lower. NRB maxima in densely built areas can be explained by the fact that lower temperatures are observed at the time of satellite pass (Figure 5) in these areas, therefore they lose lower percentages of energy than their surroundings through infrared emission. However, it should be mentioned that higher temperatures are expected in densely built areas during the course of the day due to the high thermal capacity of buildings, which increases storage of sensible heat in these areas. [34] The overall accuracy of the proposed method may be analyzed as follow: a short and L B errors were calculated by differentiating equations (5), (2) and (7) with the use of the initial spectral albedo and emissivity accuracies (Table 2). The error for both parameters was estimated around 5%. Therefore surface temperatures were recovered within about ±2.26 K and the total surface radiant exitance, taking into account the error introduced in assuming typical atmospheric parameters, was recovered within ±11.90 Wm 2. The atmospheric downward longwave flux was estimated within about ±35 Wm 2 [Pissimanis and Notaridou, 1981; Chrysoulakis and Cartalis, 2002]. The total shortwave irradiance on the surface was estimated by SBDART within about ±20 Wm 2 [Ricchiazzi et al., 1998]. Therefore surface NRB was estimated within about ±44.5 Wm 2, which means that the overall error of the method is around 15%.

9 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY ACL 8-9 [35] The methods for retrieving radiative fluxes from satellite data have been assessed by Eymard and Taconet [1995]. The accuracy of the retrieved downward solar irradiance lay in the range Wm 2. The accuracy of retrieved longwave surface fluxes was ranged between 15 and 30 Wm 2. In this study, the 44.5 Wm 2 error is mainly introduced due to the uncertainties in urban surface albedo and emissivity definition at the scale of 15 m. However, experimental measurements and investigation of the spatial distribution of radiative fluxes and albedo in urban environment are still rare. Therefore results of this study may be useful for urban applications. For instance, information about the spatial distribution of NRB is required in advanced dispersion models in order to estimate the flux of sensible heat, which regulates the boundary layer evolution and in turn determines vertical mixing of air pollutants in cities like Athens. 4. Conclusions [36] Urban surface NRB spatial distribution data is required for a wide range of studies in the physical and social sciences, as well as by municipalities for urban planning purposes. This study uses an integrated approach in order to examine the potential of satellite remote sensing for urban environmental monitoring, by estimating the spatial distribution of physical parameters related to the urban radiation balance. ASTER multispectral observations over the center of the metropolitan city of Athens were analyzed together with in situ ancillary spatial data in order to extract surface temperature and albedo values. SBDART model was used for the simulation of the radiative transfer in the atmosphere. [37] Products as spatial distributions of broadband shortwave albedo and temperature are very important for studying microclimate in urban areas and its variations, for supporting the energy budget definition in a heavily populated area such Athens, for assisting energy demand and management studies and for examining the links between the prevailing microclimatic conditions and air pollution levels. The spatial coverage offered by ASTER provides an opportunity to study various aspects of the urban radiation balance which is the driving force of urban microclimate, having significant practical implications on energy conservation, human comfort, air pollution dispersion and local air circulation. [38] Surface NRB was estimated within ±44.5 Wm 2 and it was found that the densely built areas obtain higher surface NRB and lower temperature values than the other urban area types. This may be attributed to the low solar height at the time of the satellite pass, to the high thermal capacity of the buildings and to the high shortwave absorptivity of asphalt, which affects mainly the medium built and the mixed areas. It should be noted that an underestimation of surface NRB values is inherent in the proposed method, because the spectral albedo values are overestimated from equation (5). The reason is that during the atmospheric correction of satellite data the radiation reflected by the neighborhood and scattered into the direction of the view (adjacency effect), was not taken into account. [39] This study shows that ASTER imagery can be a valuable tool in the field of urban climatology enabling a better understanding of energy aspects, their causes and effects and providing an important addition to conventional methods of monitoring the urban environment. The various thematic layers presented in this study can be integrated with other satellite or in situ derived information in a GIS platform capable of supporting urban planning. Finally, it may be used to assist researchers and engineers as a basis for creating energy budgets in urban environment. [40] Acknowledgments. The author is grateful to P. Pastacos (Scientific Director of Infocharta Ltd., Science and Technology Park of Crete), for providing the road network data for the city of Athens. References Abrams, M., and S. Hook, ASTER User Handbook, Jet Propulsion Lab., Pasadena, Calif., Ackerman, B., Temporal March of the Chicago Heat Island, J. Clim. Appl. Meteorol., 24, , Alley, R. E., and M. Jentoft-Nilsen, Algorithm Theoretical Basis Document for Brightness Temperature, Jet Propulsion Lab., Pasadena, Calif., Atwater, M. A., and J. T. Ball, Computation of IR sky temperature and comparison with surface temperature, Sol. Energy, 21, , Becker, F., The impact of spectral emissivity on the measurement of land surface temperature from a satellite, Int. J. Remote Sens., 8, , Becker, F., and Z. L. Li, Towards a local split window method over land surfaces, Int. J. Remote Sens., 11, , Boldo, D., and H. Le Men, Remote sensing model adaptation to very high resolution digital images of urban areas, paper presented at IEEE/ISPRS Joint Workshop on Remote Sensing and Data Fusion over Urban Areas, Univ. of Rome La Sapienza, Rome, 8 9 Nov Cartalis, C., and N. Chrysoulakis, Estimation of precipitable water in Greece on the basis of radiosondes and satellite data, Toxicol. Environ. Chem., 58, , Caselles, V., M. J. Lopez-Garcia, J. Melia, and A. J. Perez-Gueva, Analysis of the heat island effect of the city Valencia, Spain through air temperature transects and NOAA satellite data, Theory Appl. Climatol., 43, , Caselles, V., C. Coll, and E. Valor, Land surface emissivity and temperature determination in the whole HAPEX-Sahel area from AVHRR data, Int. J. Remote Sens., 18, , Chrysoulakis, N., and C. Cartalis, Improving the estimation of land surface temperature for the region of Greece: Adjustment of a split window algorithm to account for the distribution of precipitable water, Int. J. Remote Sens., 23, , Coll, C., V. Caselles, J. A. Sorbino, and E. Valor, On the atmospheric dependence of the split-window equation for land surface temperature, Int. J. Remote Sens., 15, , Dousset, B., and F. Gourmelon, Remote sensing applications to the analysis of urban microclimates, paper presented at IEEE/ISPRS Joint Workshop on Remote Sensing and Data Fusion over Urban Areas, Univ. of Rome La Sapienza, Rome, 8 9 Nov Dousset, B., P. Flament, and R. Bornstein, Los Angeles fires seen from space, Eos Trans. AGU, 74(33), 37 38, Exell, R. H. B., The atmospheric radiation climate of Thailand, Sol. Energy, 21, 73 79, Eymard, L., and O. Taconet, The methods for inferring surface fluxes from satellite data, and their use for atmosphere model validation, Int. J. Remote Sens., 16, , Foody, G. M., Estimation of sub-pixel land cover composition in the presence of untrained classes, Comput. Geosci., 26, , Gillespie, A., S. Rokugawa, T. Matsunanga, J. S. Cothern, S. Hook, and A. B. Kahle, A temperature and emissivity separation algorithm for Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Images, IEEE Trans. Geosci. Remote Sens., 36, , Idso, S. B., and R. D. Jackson, Thermal radiation from the atmosphere, J. Geophys. Res., 74, , Jacobowitz, H., and R. J. Tringhe, the Nimbus-7 ERB Experiment Team, The Earth radiation budget derived from the Nimbus-7 ERB experiment, J. Geophys. Res., 89, , Kandel, R., M. Viollier, P. Raberanto, J. P. Duvel, L. A. Pakhomov, V. A. Golovko, A. P. Trishchenko, J. Mueller, E. Rashke, and R. Stuhlmann, The ScaRaB Earth radiation budget dataset, Bull. Am. Meteorol. Soc., 79, , Kerr, Y. H., J. P. Lagouarde, and J. Imbernon, Accurate land surface temperature retrieval from AVHRR data with the use of an improved splitwindow algorithm, Remote Sens. Environ., 41, , 1992.

10 ACL 8-10 CHRYSOULAKIS: URBAN NRB ESTIMATION USING ASTER IMAGERY Kim, H. H., Urban Heat Island, Int. J. Remote Sens., 13, , Lee, H. Y., An application of NOAA AVHRR thermal data to the study of urban heat island, Atmos. Environ., 278, 1 3, Liang, S., Narrowband to broadband conversions of land surface albedo I algorithms, Remote Sens. Environ., 76, , Lillesand, T. M., and R. W. Kiefer (Eds), Remote Sensing and Image Interpretation, 4th ed., 22 pp., John Wiley, Hoboken, N. J., Ottle, C., and D. Vidal-Madjar, Estimation of land surface temperature with NOAA-9 data, Remote Sens. Environ., 40, 27 41, Pissimanis, D. K., and V. A. Notaridou, The atmospheric radiation in Athens during the summer, Sol. Energy, 26, , Prata, A., Land surface temperatures derived from the Advanced Very High Resolution Radiometer and the Along-Track Scanning Radiometer: 1. Theory, J. Geophys. Res., 98, 16,689 16,702, Prata, A., V. Caselles, C. Coll, C. Ottle, and J. Sorbino, Thermal remote sensing of land surface temperature from satellites: Current status and future prospects, Remote Sens. Rev., 12, , Price, J. C., Assessment of the urban heat island effect through the use of satellite data, Mon. Weather Rev., 107, , Price, J. C., Land surface temperature measurements from the split window channels of the NOAA-7 AVHRR, J. Geophys. Res., 89, , Ricchiazzi, P., S. Yang, C. Gautier, and D. Sowle, SBDART: A research and teaching software tool for plane-parallel radiative transfer in the Earth s atmosphere, Bull. Am. Meteorol. Soc., 79, , Ridd, M. K., Exploring a V-I-S (vegetation-impervious surface-soil) model for urban ecosystem analysis through remote sensing: Comparative anatomy for cities, Int. J. Remote Sens., 16, , Santamouris, M., N. Papanikolaou, I. Livada, I. Koronakis, C. Georgakis, A. Argiriou, and D. N. Asimakopoulos, On the impact of urban climate on the energy consumption of buildings, Sol. Energy, 70, , Small, C., Multitemporal analysis of urban reflectance, Remote Sens. Environ., 81, , Smith, G. L., R. N. Green, E. Raschke, L. M. Avis, J. T. Suttles, B. A. Wielicki, and R. Davies, Investigation methods for satellite studies of the Earth s radiation budget: Development of algorithms for the ERBE mission, Rev. Geophys., 24, , Sobrino, J. A., C. Coll, and V. Caselles, Atmospheric correction for land surface temperature using NOAA-11 AVHRR channels 4 and 5, Remote Sens. Environ., 38, 19 34, Stefanov, W. L., M. S. Ramsey, and P. R. Christensen, Monitoring urban land cover change: An expert system approach to land cover classification of semiarid urban centers, Remote Sens. Environ., 77, , Swinbank, W. C., Longwave radiation from clear skies, Q. J. R. Meteorol. Soc., 89, , Vidal, A., Atmospheric and emissivity correction of land surface temperature measured from satellite using ground measurements or satellite data, Int. J. Remote Sens., 12, , Vucovich, F. M., An analysis of the ground temperature and reflectivity pattern about St. Louis, Missouri, using HCMM satellite data, J. Appl. Meteorol., 22, , Voogt, J. A., and T. R. Oke, Complete urban surface temperatures, J. Appl. Meteorol., 36, , N. Chrysoulakis, Foundation for Research and Technology-Hellas, Regional Analysis Division, Institute of Applied and Computational Mathematics, Vassilika Vouton, P.O. Box 1527, Heraklion GR-71110, Crete, Greece. (zedd2@iacm.forth.gr)