HYDROLOGICAL PROCESSESHydrol. Process. (in press)Published online in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/hyp.6389

Determination of regional land surface heat fluxes overa heterogeneous landscape of the Jiddah area of

Saudi Arabia by using Landsat-7 ETM data

Yaoming Ma,1,2* Hui Tian,2 Hirohiko Ishikawa,3 Ryohji Ohba,4 Hiromasa Ueda3 and Jun Wen2

1 Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100085, People’s Republic of China2 Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou 730000, People’s Republic of

China3 Disaster Prevention Research Institute, Kyoto University, Uji, Kyoto 611, Japan

4 Nagasaki R&D Center, Mitsubishi Heavy Industries, LTD, Fukahorimachi 5-717-1, Nagasaki, Japan

Abstract:

In this study, a parameterization method based on Landsat-7 Enhanced Thematic Mapper (ETM) data and field observations ispresented and tested for deriving the regional land surface variables, vegetation variables and land surface heat fluxes over aheterogeneous landscape. As a case study, the method and two Landsat-7 ETM images are applied to the Jiddah area of SaudiArabia. The regional distribution maps of surface reflectance, normalized difference vegetation index, modified soil adjustedvegetation index (MSAVI), vegetation coverage, leaf area index, surface temperature, net radiation flux, soil heat flux, sensibleheat flux and latent heat flux have been determined over the Jiddah area. The derived results have been validated by using the‘ground truth’. The results show that the more reasonable regional distributions of land surface variables (surface reflectance,surface temperature), vegetation variables (MSAVI and vegetation coverage), net radiation, soil heat flux and sensible heat fluxcan be obtained by using the method proposed in this study. Further improvement of the method is also discussed. Copyright 2006 John Wiley & Sons, Ltd.

KEY WORDS Landsat-7 ETM; regional land surface heat flux; arid area; Jiddah; Saudi Arabia

Received 5 October 2005; Accepted 22 February 2006

INTRODUCTION

Arid areas (e.g. Jiddah area of Saudi Arabia) with aninhomogeneous landscape are characterized by extremegradients in land surface properties, such as wetness,roughness and temperature, which have a significant butlocal impact on the atmospheric boundary layer (ABL).Observation of the actual extent over these areas isessential to understand the mechanisms through whichinhomogeneous land surfaces may have a significantimpact on the structure and dynamics of the overly-ing ABL. Progress in this research area requires spatialmeasurements of variables such as surface hemispheri-cal reflectance, radiometric surface temperature, normal-ized difference vegetation index (NDVI), modified soiladjusted vegetation index (MSAVI), vegetation cover-age, leaf area index (LAI), local aerodynamic roughnesslength, etc. On-board radiometric imaging by satellite canprovide useful estimates of most of these variables. Inusing these variables we can derive the distribution ofland surface heat fluxes and evaporation over an inho-mogeneous landscape.

A parameterization method based on Landsat-7 Enhan-ced Thematic Mapper (ETM) data and field observations

* Correspondence to: Yaoming Ma, Institute of Tibetan Plateau Research,Chinese Academy of Sciences, Beijing 100085, People’s Republic ofChina. E-mail: ymma@itpcas.ac.cn

(e.g. vertical profiles of air temperature, humidity andwind speed, surface radiation budget components and tur-bulent heat fluxes, etc.) will be presented and tested forderiving the regional land surface variables, vegetationvariables and land surface heat fluxes over the heteroge-neous landscape of the Jiddah area of Saudi Arabia.

LANDSAT-7 ETM DATA AND FIELDOBSERVATION DATA

Landsat-7 ETM provides a spectral radiance in sevennarrow bands, with a spatial resolution of 30 ð 30 m2

for three visible bands (bands 1, 2, 3) and three near-infrared bands (bands 4, 5, 7), and 60 ð 60 m2 for thethermal infrared band 6. The two ETM images used inthis study are at 10 : 00 (local time) 16 June 2002 and 10January 2003 over the Jiddah area of Saudi Arabia. Bothof the days were very clear when the Landsat-7 ETMoverpassed the Jiddah area.

The most relevant data, collected at the Jiddah surfacestation to support the parameterization of land surfaceheat fluxes and analysis of the ETM images, consist ofsurface radiation budget components, surface radiationtemperature, surface reflectance, vertical profiles of airtemperature, humidity, wind speed and direction mea-sured by radiosonde, turbulent fluxes measured by the

Copyright 2006 John Wiley & Sons, Ltd.

Y. MA ET AL.

eddy-correlation technique, soil heat flux, soil tempera-ture profiles, soil moisture profiles, and the vegetationstate, etc.

THEORY AND SCHEME

The general concept of the methodology is shown inFigure 1. The surface reflectance for shortwave radiationr0 is retrieved from the Landsat-7 ETM data withatmospheric correction by a four-stream radiative transferassumption in the solar spectral bands (Verhoef, 1997)using aerological observation data. The land surfacetemperature Tsfc is also derived from Landsat-7 ETMdata and aerological observation data. The radiativetransfer model MODTRAN (Berk et al., 1989) computesthe downward shortwave and longwave radiation at thesurface. With these results, the surface net radiation Rn

is determined. The soil heat flux G0 is estimated fromRn, Tsfc, r0 and MSAVI (Qi et al., 1994), which is alsoderived from Landsat-7 ETM data. The sensible heatflux H is estimated from Tsfc, surface and aerologicaldata with the aid of so called ‘blending height’ approach(Mason, 1988), and regional latent heat flux �E can bederived as the residual of the energy budget theorem forland surface.

Net radiation

The regional net radiation flux can be derived from

Rn�x, y� D [1 � r0�x, y�]K#�x, y�C εsL#�x, y�

� ε0�x, y��T4sfc�x, y� �1�

where ε0�x, y� is the surface emissivity, K is the short-wave (0Ð3–3 µm) radiation component and L is the long-wave (3–100 µm) radiation component. In Equation (1),εs is the longwave emissivity, and we have allowedfor an amount �1 � εs�L# of reflected longwave radia-tion, which equals zero if the surface is a black body

with εs D 1 (Garratt, 1992). Surface reflectance r0�x, y�can be derived from the integrated hemispherical plane-tary reflectance (e.g. Koepke et al., 1985; Menenti et al.,1989; Bastiaanssen, 1995; Wang et al., 1995). How-ever, the linear regression relationships are in doubt,because: (1) the ground measurement is only a pointvalue, whereas the satellite pixels are the average of manypoint values; (2) few ground observation data coincidewith satellite data (Wen, 1999). A four-stream radia-tive transfer assumption for atmospheric correction inthe solar spectral bands (Verhoef, 1997) is introduced toderive the surface reflectance over the Jiddah area in thisstudy.

The incoming shortwave radiation flux K#�x, y� inEquation (1) can be derived from the radiative transfermodel MODTRAN (Kenizys et al., 1996; Ma et al.,2002), where atmospheric shortwave transmittance �sw

is obtained. Hence, K#�x, y� can be obtained as

K#�x, y� D �swK#TOA�x, y� �2�

where the regional variation of radiation flux perpendic-ular to the top of atmosphere K#

TOA�x, y� is a spectrallyintegrated form of in-band radiation flux perpendicular tothe top of atmosphere K#

TOA (b), and

K#TOA�x, y� D K#

exo�b� cos �sun�x, y�

d2s

�3�

where K#exo (b) is the averaged in-band solar exo-

atmospheric irradiance undisturbed by �sun being zero, ‘b’is an abbreviation of in-band, ds is the relative sun–Earthdistance, and �sun is the sun zenith angle.

The incoming longwave radiation flux L#�x, y� inEquation (1) can also be derived from MODTRANdirectly (Ma and Tsukamoto, 2002). Surface tempera-ture Tsfc�x, y� in Equation (1) can be derived from theLandsat-7 ETM thermal infrared band-6 spectral radiance

Landsat-7 ETM data

MODTRAN surface and aero-logical data

LAI

Vegetationcoverage

MSAVI e0 r0 Tsfc K↓ L↓ Z0m

blending height approach

G0 Rn H

lE

kB−1 ,d0

Figure 1. Diagram of the parameterization procedure by combining Landsat-7 ETM data with field observations

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

REGIONAL LAND SURFACE HEAT FLUX OVER A HETEROGENEOUS LANDSCAPE

(Ma and Tsukamoto, 2002). In other words, the distribu-tion of land surface temperature is obtained as

Tsfc�x, y� D ε0�x, y��1/4TB�x, y� �4�

where TB is the brightness temperature at the surface,which can be derived from the surface upward longwaveradiation flux (Ma and Tsukamoto, 2002), and ε0�x, y�is the surface emissivity, which can be derived fromvegetation coverage Pv (Valor and Caselles, 1996), i.e.

ε0�x, y� D εv�x, y�Pv�x, y�C εg�x, y�[1 � Pv�x, y�]

C 4hεi[1 � Pv�x, y�]Pv�x, y� �5�

where εv�x, y� and εg�x, y� are the surface emissivitiesfor full vegetation and bare soil respectively, hεi is theerror item, and vegetation coverage (Carlson and Ripley,1997)

Pv�x, y� D[

NDVI�x, y�� NDVImin

NDVImax � NDVImin

]2

�6�

where NDVImin and NDVImax are the NDVI values forbare soil and full vegetation respectively.

Soil heat flux

The regional soil heat flux G0�x, y� is determinedthrough (Choudhury and Monteith, 1988)

G0�x, y� D �sCsTsfc�x, y�� Ts�x, y�

rsh�x, y��7�

where �s is the soil dry bulk density, Cs is the soil spe-cific heat, Ts�x, y� is the soil temperature of a determineddepth, and rsh�x, y� is the resistance of soil heat trans-portation. The regional soil heat flux G0�x, y� cannotbe mapped directly from satellite observations throughEquation (7). Many investigations have shown that themidday G0/Rn ratio is reasonably predictable from spe-cial vegetation indices (Daughtry et al., 1990). can beconsidered as a function F, which relates G0/Rn to othervariables (Ma and Tsukamoto, 2002). Some researchershave concluded that G0/Rn D D F�NDVI� (Clothieret al., 1986; Kustas and Daughtry, 1990). A better ratio ofG0/Rn D D F�r0, Tsfc,NDVI� was also found (Choud-hury et al., 1987; Menenti et al., 1991; Bastiaanssen,1995). The parameterization scheme based on MSAVIfound over the Jiddah area is

G0�x, y� D Rn�x, y�Tsfc�x, y�

r0�x, y��0Ð000 28 C 0Ð004 364r0

C 0Ð008 46r02�[1 � 0Ð978 92MSAVI�x, y�4]

�8�

where r0 is the surface reflectance averaged over theperiod when the soil is heated from the sun. Sur-face temperature Tsfc is expressed in degrees Celsius.

MSAVI(x, y) is derived from the band reflectance of theLandsat-7 ETM as (Qi et al., 1994).

MSAVI�x, y�

D2r4�x, y�C 1 �

√[2r4�x, y�C 1]2

�8[r4�x, y�� r3�x, y�]

2�9�

where r3 and r4 are the band reflectances of the Landsat-7ETM band-3 and band-4 on the land surface.

Sensible heat flux

The regional distribution of sensible heat flux iscalculated from

H�x, y� D �CpTsfc�x, y�� Ta�x, y�

ra�x, y��10�

where aerodynamic resistance is

ra�x, y� D 1

kuŁ�x, y�

ð{

ln[z � d0�x, y�

Z0m�x, y�

]C kB�1�x, y�� h�x, y�

}�11�

where k is the von Karman constant, uŁ is the frictionvelocity, z is the reference height, d0 is the zero-planedisplacement height, Z0m is the effective aerodynamicroughness, kB�1 is the excess resistance for heat trans-portation, and h is the stability correction function forheat. The friction velocity uŁ can be derived from

u�x, y� D uŁ�x, y�k

{ln

[z � d0�x, y�

Z0m�x, y�

]� m�x, y�

}�12�

where m is the stability correction function for momen-tum.

From Equations (10)–(12)

H�x, y� D �Cpk2u�x, y�

Tsfc�x, y�� Ta�x, y�{ln

[z � d0�x, y�Z0m�x, y�

]C kB�1�x, y�� h�x, y�

}{

ln[z � d0�x, y�Z0m�x, y�

]� m�x, y�

}�13�

One approach to simulate sensible heat flux on a largescale is to scale up or aggregate the regional sensible fluxby a weighted average of the contributions from differentsurface elements, based on the principle of flux conser-vation. The method of ‘blending height’ is proposed toderive the regional sensible heat flux in this study. If thelocal-scale advection is comparatively small during theperiod of the Landsat-7 ETM observation taking place,then the development of a convection boundary layer mayadjust the surface-disorganized variability at the ‘blend-ing height’, where the atmospheric characteristics becomeproximately independent of the horizontal position. Thecorresponding ‘effective’ surface variables can be deter-mined accordingly (Mason, 1988). This approach hasproved successful in calculating areally averaged surface

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

Y. MA ET AL.

fluxes in recent years (Lhomme et al., 1994; Bastiaassen,1995; Wang et al., 1995; Ma and Tsukamoto, 2002; Maet al., 2003). Based on this approach, the regional sensi-ble heat flux H�x, y� can be described as

H�x, y� D �Cpk2uB

Tsfc�x, y�� Ta�x, y�{ln

[zB � d0�x, y�Z0m�x, y�

]C kB�1�x, y�� h�x, y�

}{

ln[zB � d0�x, y�Z0m�x, y�

]� m�x, y�

}�14�

where ZB is the blending height and uB is the wind speedat the blending height. In this study, ZB and uB are deter-mined with the aid of field measurements of radiosondingsystem. Z0m�x, y� in Equation (14) over the Jiddah areaof Saudi Arabia includes the effect of topography, lowvegetation (e.g. shrubs), and taller plants (e.g. trees). Itis calculated using Taylor’s model (Taylor et al., 1989).kB�1 is the excess resistance, which can be determinedfrom the LAI by using a widely used model (Quallsand Brutsaert, 1995). d0 is the zero-plane displacement,which can be calculated from the LAI by using Raupach’smodel (Raupach, 1994). Ta�x, y� in Equation (14) is theregional distribution of air temperature at the referenceheight; it can be derived from an interpolation method(Ma and Tsukamoto, 2002). h�x, y� and m�x, y� inEquation (14) are the integrated stability functions. Foran unstable condition, the integrated stability functions h�x, y� and m�x, y� can be written as (Paulson, 1970){ m�x, y� D 2 ln[�1 C X�/2]

C ln[�1 C X2�/2] � 2 arctan�X�C 0Ð5 h�x, y� D 2 ln[�1 C X2�/2]

�15�

where X D f1 � 16[z � d0�x, y�]/L�x, y�g0Ð25. For a sta-ble condition, the integrated stability function m�x, y�and h�x, y� become (Webb, 1970)

m�x, y� D h�x, y� D �5[z � d0�x, y�]/L�x, y� �16�

The stability function [z � d0�x, y�]/L�x, y� is calculatedby Businger’s method (Businger, 1988):{ [z � d0�x, y�]/L�x, y� D Ri�x, y� (unstable)

[z � d0�x, y�]/L�x, y�D Ri�x, y�/[1 � 5Ð2Ri�x, y�] (stable)

�17�

where Ri�x, y� is the Richardson number; and accordingto the definition of the Richardson number, the approx-imate analytical solutions of Ri found by Yang et al.(2001) will be used here.

Latent heat flux

The regional latent heat flux �E�x, y� can be derived asthe residual of the energy budget theorem for land surfacebased on the condition of zero horizontal advection atz < zsur, i.e.

�E�x, y� D Rn�x, y��H�x, y��G0�x, y� �18�

CASE STUDY AND VALIDATION

Figures 2 and 3 show the distribution maps of surfacereflectance, surface temperature and surface heat fluxesaround the Jiddah area of Saudi Arabia. Their frequencydistributions are shown in Figures 4 and 5. The distri-bution maps of surface reflectance, NDVI, MSAVI andLAI in Figure 2 are based on 3200 ð 3200 pixels witha size of 30 ð 30 m2. The distribution maps of surfacetemperature in Figure 2 and land surface heat fluxes (netradiation, soil heat flux, sensible heat flux and latentheat flux) in Figure 3 are based on 1600 ð 1600 pix-els with a size of 60 ð 60 m2. Derived land surfaceheat fluxes are validated by field measurements. Sinceit is difficult to determine where the exact locationsof the experimental sites are, the values of a 5 ð 5pixel rectangle, surrounding the determined UniversalTransverse Mercator coordinates, are compared with thefield measurements. The mean absolute percent differ-ence (MAPD) can quantitatively measure the differencebetween the derived results (Hderived�i�) and measured val-ues (Hmeasured�i�):

MAPD D 100

n

n∑iD1

jHderived�i� �Hmeasured�i�jHmeasured�i�

�19�

The results show that:

1. The derived surface variables (land surface reflectanceand surface temperature) and surface heat fluxes (netradiation flux Rn, soil heat flux G0, sensible heat fluxHand latent heat flux �E) in two different months overthe study area are in good accordance with the landsurface status. These parameters show a wide rangedue to the strong contrast of surface features. Surfacereflectance is from 0Ð07 to 0Ð62 in June and from 0Ð06to 0Ð51 in January. Surface temperature ranged from24 to 78 °C in June and from 8 to 58 °C in January.Net radiation changed from 425 to 850 W m�2 in Juneand from 85 to 565 W m�2 in January. Soil heat fluxvaried from 100 to 250 W m�2 in June and from 20to 260 W m�2 in January. Sensible heat flux is from70 to 370 W m�2 in June and from 20 to 290 W m�2

in January, and latent heat flux varied from �20 toC550 W m�2 in June and �50 to C550 W m�2 inJanuary (Figures 3 and 5). The negative latent heatflux means that there is negative water vapour fluxover the Jiddah area. This phenomenon occurs overthe desertification area (Wang and Mitsuta, 1990,1992).

2. The derived pixel values (Figures 2 and 3) and averagevalues (Figures 4 and 5) of surface reflectance, surfacetemperature, net radiation flux, soil heat flux, sensibleheat flux and latent heat flux in June are higher thanthose in January.

3. The derived surface reflectance and surface temper-ature in this study are in good accordance withthe field measurements, with MAPD less than 6%(Table I).

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

REGIONAL LAND SURFACE HEAT FLUX OVER A HETEROGENEOUS LANDSCAPE

Surface reflectance (10 January 2003)

Surface temperature (10 January 2003) Surface temperature (16 June 2002)

Surface reflectance (16 June 2002)

20 km

0.06

0.15

0.24

0.33

0.42

0.51

0.51

0.07

0.18

0.29

0.40

0.62

8.0

18.0

28.0

38.0

48.0

58.0

24.0

33.0

42.0

51.0

60.0

69.0

78.0

N

Study area (around Jiddah City of Saudi Arabia)

Figure 2. Distributions of surface reflectance and surface temperature around the Jiddah area of Saudi Arabia

Table I. Comparison of the derived results (Calibrated) versusthose measured values (Measured) with MAPD

Calibrated Measured MAPD (%)

r0 0Ð180 0Ð171 5Ð26Ts (°C) 38Ð0 36Ð03 4Ð68Rn �W m�2� 307Ð0 316Ð2 2Ð91G0 �W m�2� 94Ð0 89Ð2 5Ð38H �W m�2� 149Ð0 159Ð5 6Ð60�E �W m�2� 64Ð0 67Ð5 5Ð18

4. The derived net radiation flux over the study area isvery close to the field measurement, with MAPD lessthan 3%(Table I).

5. Problems exist in the NDVI definition equationbecause of the external factor effect, such as soilbackground variations (Huete et al., 1985; Huete,1989). To reduce the soil background effect, Qi et al.(1994) proposed using MSAVI. Therefore, the param-eterization method based on MSAVI for soil heat fluxdensity is better than that based on NDVI on a hetero-geneous land surface. Although the derived regional

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

Y. MA ET AL.

20 km

N

Soil heat flux (10 January 2003) W/m2

20

60

100

140

180

260

220RED SEA

Net radiation (10 January 2003)85

165

245

325

405

485

565

W/m2

RED SEA

W/m2

RED SEA

Sensible heat flux (10 January 2003)

20

110

65

155

200

245

290

W/m2

RED SEA

−50

190

70

310

430

550

Latent heat flux (10 January 2003)

Soil heat flux (16 June 2002) W/m2

100

130

160

190

220

250

RED SEA

Net radiation (16 June 2002) W/m2

425

510

595

680

765

850

RED SEA

RED SEA

W/m2Sensible heat flux (16 June 2002)

70

170

120

220

270

320

370

W/m2

RED SEA

−20

170

75

265

360

455

550

Latent heat flux (16 June 2002)

Figure 3. Distributions of surface heat fluxes around the Jiddah area of Saudi Arabia

soil heat flux based on MSAVI is a bit higher than themeasured value, the parameterization method based onMSAVI for soil heat flux is suitable for the inhomoge-neous land surface of the Jiddah area, and the MAPDis about 5Ð38% (Table I).

6. The derived regional sensible heat flux and latentheat flux, with MAPD of around 6% at the valida-tion site, are in good agreement with the field mea-surements (Table I). It is pointed that the proposed

parameterization methodology for sensible heat fluxand latent heat flux is reasonable, and it can be usedover the Jiddah area of Saudi Arabia

CONCLUDING REMARKS

In this study, the regional distributions of land surfacevariables (surface reflectance and surface temperature),

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

REGIONAL LAND SURFACE HEAT FLUX OVER A HETEROGENEOUS LANDSCAPE

10 January 2003

0.450.350.250.150.05

Mean=0.148

0

5

10

15

20

Fre

quen

cy (

%)

Surface reflectance r0

16 June 2002

Mean=0.227

0.450.350.250.150.050

5

10

15

20

Fre

quen

cy (

%)

Surface reflectance r0

Mean=34.7 °C

10 January 2003

0

4

8

12

16

Fre

quen

cy (

%)

Surface temperature Tsfc(°C)

0 10 20 80 907060504030

Mean=53.3 °C

16 June 2002

0

4

8

12

16

Fre

quen

cy (

%)

Surface temperature (°C)

0 10 20 80 907060504030

Figure 4. Frequency distributions of surface reflectance and surface temperature around the Jiddah area of Saudi Arabia

vegetation variables (NDVI, MSAVI, vegetation cover-age and LAI) and land surface heat fluxes (net radiation,soil heat flux, sensible and latent heat flux) over hetero-geneous arid areas of Jiddah, Saudi Arabia, were derivedwith the aid of Landsat-7 ETM data and field observa-tions. This forms a sound basis for studying land surfacevariables, vegetation variables and land surface fluxes.

Dealing with the regional land surface heat fluxesover heterogeneous landscape is not an easy issue. Theparameterization method presented in this study is still inthe development stage:

1. From Equation (14), it is better to use air temperatureand wind speed at the reference height when wecalculate the sensible heat flux, and we only proposeone method to calculate regional air temperature atthe reference height. The wind speed at the referenceheight over the study area should be derived by usingsome numerical models, and the spatial resolution ofthe models should be 30 m ð 30 m.

2. Only two Landsat-7 images at a specific time of a spe-cific day are used in this study. To obtain more accurateregional land surface fluxes, their seasonal variationsand even day-by-day variation over the Jiddah areaof Saudi Arabia need to be examined and comparedwith other areas (e.g. the Tibetan Plateau area and thedesertification area over northwest China), more fieldobservations (PBL tower and radiation measurementsystem, radiosonde system, turbulent fluxes measuredby the eddy-correlation technique, soil moisture and

soil temperature measurement system, etc.), and othersatellites, such as the Moderate Resolution ImagingSpectroradiometer and National Oceanic and Atmo-spheric Administration Advanced Very High Resolu-tion Radiometer, have to be used. It is also worth tryingthe Surface Energy Balance Index method (Menentiand Choudhury, 1993) and the Surface Energy BalanceSystem (Su, 2002).

Vegetation variables cannot be validated in this studyas no such measurements were made during the exper-iments. More attention should be paid to the measure-ments of vegetation variables, such as NDVI, LAI andvegetation coverage, in future experiments.

ACKNOWLEDGEMENTS

This work was under the auspices of the National ScienceFoundation of China (40233035 and 40520140126) andthe cooperative research project between the Institute ofTibetan Plateau Research, Chinese Academy of Sciencesand Mitsubishi Heavy Industries, Ltd, Japan, ‘ResearchRevolution 2002’ of the Ministry of Education, Culture,Science and Technology of Japan (RR2002-6). Mostparts of this study were done as cooperative researchwork in the Disaster Prevention Research Institute, KyotoUniversity, Japan, and the Alterra Green World Research,Wageningen University and Research Centre. We thankDr L. Jia, Dr Y. Oku, and Dr K. Tanaka for their kindhelp and useful discussions.

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

Y. MA ET AL.

0 100 200 300 400 500 600 700 800 900

Mean=333 W/m2

10 January 2003

0

4

8

12

16

20

Fre

quen

cy (

%)

Net radiation Rn (W/m2)

0 100 200 300 400 500 600 700 800 900

Mean=735 W/m2

16 June 2002

0

4

8

12

16

20

Fre

quen

cy (

%)

Net radiation Rn (W/m2)

0 50 100 150 200 250 300

30

Mean=77 W/m2

10 January 2003

0

5

10

15

20

25

Fre

quen

cy (

%)

Soil heat flux G0 (W/m2)

0 50 100 150 200 250 300

Mean=167

16 June 2002

0

5

10

15

20

25

30

Fre

quen

cy (

%)

Soil heat flux G0 (W/m2)

0 50 100 150 200 250 300 350 400

Mean=122

10 January 2003

0

4

8

12

16

Fre

quen

cy (

%)

Sensible heat flux H (W/m2)

0 50 100 150 200 250 300 350 400

Mean=245

16 June 2002

0

4

8

12

16

Fre

quen

cy (

%)

Sensible heat flux H ( W/m2)

0 100 200 300 400 500 600

Mean=134

10 January 2003

0

5

10

15

20

25

Fre

quen

cy (

%)

Latent heat flux (W/m2)

0 100 200 300 400 600

Mean=323

16 June 2002

0

5

10

15

20

25

Fre

quen

cy (

%)

Latent heat flux (W/m2)

500

Figure 5. Frequency distributions of land surface heat fluxes around the Jiddah area of Saudi Arabia

REFERENCES

Bastiaanssen WGM. 1995. Regionalization of surface flux densities andmoisture indicators in composite terrain. PhD thesis, WageningenAgricultural University; 143–161.

Berk A, Berstein LS, Robertson DC. 1989. MODTRAN, a moderateresolution model for LOWTRAN 7. GL-TR-89-0122, GeophysicsLaboratory, Hanscom AFB, MA 01732.

Businger JA. 1988. A note on the Businger–Dyer profiles. BoundaryLayer Meteorology 42: 145–151.

Carlson TN, Ripley DA. 1997. On the relation between NDVI, fractionalvegetation cover, and leaf area index. Remote Sensing of Environment62: 241–252.

Choudhury BJ, Monteith JA. 1988. Four-layer model for the heatbudget of homogeneous land surfaces. Quarterly Journal of the RoyalMeteorological Society 114: 373–398.

Choudhury BJ, Idso SB, Reginato RJ. 1987. Analysis of an empiricalmodel for soil heat flux under a growing wheat crop for estimatingevaporation by infrared-temperature based energy balance equation.Agricultural and Forest Meteorology 39: 283–297.

Clothier BE, Clawson KL, Pinter PJ, Moran MS, Reginato RJ, Jack-son RD. 1986. Estimating of soil heat flux from net radiation during thegrowth of alfalfa. Agricultural and Forest Meteorology 37: 319–329.

Daughtry CST, Kustas WP, Moran MS, Pinter PJ, Jackson RD,Brown PW, Nichols WD, Gay LW. 1990. Spectral estimates of net

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp

REGIONAL LAND SURFACE HEAT FLUX OVER A HETEROGENEOUS LANDSCAPE

radiation and soil heat flux. Remote Sensing of Environment 32:111–124.

Garratt JR. 1992. The Atmospheric Boundary Layer . CambridgeUniversity Press: New York; 120.

Huete AR. 1989. Soil influences in remotely sensed vegetation-canopyspectra. In Theory and Applications of Optical Remote Sensing ,Asrar G (ed.). Wiley-Interscience: 107–141.

Huete AR, Jackson RD, Post DF. 1985. Spectral response of aplant canopy with different soil backgrounds. Remote Sensing ofEnvironment 17: 37–53.

Kenizys FX, Abreu LW, Anderson GP, Chetwynd JH, Shettle EP,Berk A, Bernstein LS, Robertson DC, Acharya P, Rothman LS,Selby JEA, Gallery WO, Clough SA. 1996. The MODTRAN3/2 reportand LODTRAN 7 model , Abreu LW, Andson GP (eds). Prepared byOntar Corp., North Andover, MA, for Phillips Laboratory, GeophysicalDirectorate, Hanscom AFB, MA.

Koepke P, Kriebel KT, Dietrich B. 1985. The effect of surface reflectionand of atmospheric parameters on the short wave radiation budget.Advances in Space Research 5: 353–354.

Kustas WP, Daughtry CST. 1990. Estimation of the soil heat flux/netradiation ratio from spectral data. Agricultural and Forest Meteorology39: 205–223.

Lhomme JP, Chehbouni A, Monteny B. 1994. Effective parameters ofsurface energy balance in heterogeneous landscape. Boundary LayerMeteorology 71: 297–310.

Ma Y, Tsukamoto O. 2002. Combining Satellite Remote Sensing withField Observations for Land Surface Heat Fluxes over InhomogeneousLandscape. China Meteorological Press: Beijing.

Ma Y, Su Z, Li Zh, Koike T, Menenti M. 2002. Determination ofregional net radiation and soil heat flux densities over heterogeneouslandscape of the Tibetan Plateau. Hydrological Processes 16:2963–2971.

Ma Y, Ishikawa H, Tsukamoto O, Menenti M, Su Z, Yao T, Koike T,Yasunari T. 2003. Regionalization of surface fluxes over heterogeneouslandscape of the Tibetan Plateau by using satellite remote sensing.Journal of the Meteorological Society of Japan 81(2): 277–293.

Mason P. 1988. The formation of areally averaged roughness lengths.Quarterly Journal of the Royal Meteorological Society 114: 399–420.

Menenti M, Choudhury BJ. 1993. Parameteraization of land surfaceevaporation by means of location dependent potential evaporation andsurface temperature range. In Exchange Processes at the Land Surfacefor a Range of Space and Time Scales , Bolle HJ, Feddes RA, Kalma JD(eds). IAHS Publication No. 212. IAHS Press: Wallingford; 561–568.

Menenti M, Bastiaanssen WGM, Van Eick D. 1989. Determination ofhemispheric reflectance with Thematic Mapper data. Remote Sensingof Environment 28: 327–337.

Menenti M, Bastiaanssen WGM, Hefny K, Abd El Karim MH. 1991.Mapping of ground water losses by evaporation in the Western Desertof Egypt . DLO Winand Staring Centre, Report no. 43, Wageningen,The Netherlands; 1–116.

Paulson CA. 1970. The mathematic representation of wind speed andtemperature profiles in the unstable atmospheric surface layer. Journalof Applied Meteorology 9: 856–861.

Qi JG, Chehbouni A, Huete AR, Kerr YH, Sorooshian S. 1994. Amodified soil adjusted vegetation index. Remote Sensing ofEnvironment 48: 119–126.

Qualls RJ, Brutsaert W. 1995. The effect of vegetation density on theparameterization of scalar roughness to estimate spatially distributedsensible heat fluxes. Water Resources Research 32: 645–652.

Raupach MR. 1994. Simplified expressions for vegetation roughnesslength and zero-plane displacements as functions of canopy height andarea index. Boundary Layer Meteorology 71: 211–216.

Su Z. 2002. The Surface Energy Balance System (SEBS) for estimationof turbulent heat fluxes. Hydrology and Earth System Science 6: 85–99.

Taylor PA, Sykes RI, Mason PJ. 1989. On the parameterization of dragover small scale topography in neutrally stratified boundary flow.Boundary Layer Meteorology 48: 409–422.

Valor E, Caselles V. 1997. Mapping land surface emissivity from NDVI:application to European, African, and South American areas. RemoteSensing of Environment 57: 167–184.

Verhoef W. 1997. Theory of radiative transfer models applied in opticalremote sensing of vegetation canopies . PhD thesis, Remote SensingDepartment of National Aerospace Laboratory, The Netherlands.

Wang J, Mistsuta Y. 1990. Peculiar downward water vapor flux over theGobi Desert in the daytime. Journal of the Meteorological Society ofJapan 68: 399–401.

Wang J, Mistsuta Y. 1992. Evaporation from the desert—a preliminaryresult of HEIFE. Boundary-Layer Meteorology 59: 413–418.

Wang J, Ma Y, Menenti M, Bastiaanssen WGM, Mistsuta Y. 1995. Thescaling-up of processes in the heterogeneous landscape of HEIFEwith the aid of satellite remote sensing. Journal of the MeteorologicalSociety of Japan 73: 1235–1244.

Webb EK. 1970. Profile relationships: the log–linear range and extensionto strong stability. Quarterly Journal of the Royal MeteorologicalSociety 96: 67–90.

Wen J. 1999. Land surface variables estimated from remote sensing andthe correction of atmospheric effects . PhD thesis, Lanzhou Instituteof Plateau Atmospheric Physics, The Chinese Academy of Sciences,Lanzhou (in Chinese with English abstract).

Yang K, Tamai N, Koike T. 2001. Analytical solution of surface layersimilarity equations. Journal of Applied Meteorology 40: 1647–1653.

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. (in press)DOI: 10.1002/hyp