Estimation of surface energy fluxes under complex terrain of Mt. Qomolangma over the Tibetan Plateau

Surface solar radiation is an important parameter in surface energy balance models and in estimation of evapotranspiration. This study developed a DEM based radiation model to estimate instantaneous clear sky solar radiation for surface energy balance system to obtain accurate energy absorbed by the mountain surface. Efforts to improve spatial accuracy of satellite based surface energy budget in mountainous regions were made in this work. Based on eight scenes of Landsat TM/ETM+ (Thematic Mapper/Enhanced Thematic Mapper+) data and observations around the Qomolangma region of the Tibetan Plateau, the topographical enhanced surface energy balance system (TESEBS) was tested for deriving net radiation, ground heat flux, sensible heat flux and latent heat flux distributions over the heterogeneous land surface. The land surface energy fluxes over the study area showed a wide range in accordance with the surface features and their thermodynamic states. The model was validated by observations at QOMS/CAS site in the research area with a reasonable accuracy. The mean bias of net radiation, sensible heat flux, ground heat flux and latent heat flux is lower than 23.6 W m−2. The surface solar radiation estimated by the DEM based radiation model developed by this study has a mean bias as low as −9.6 W m−2. TESEBS has a decreased mean bias of about 5.9 W m −2 and 3.4 W m−2 for sensible heat and latent heat flux, respectively, compared to the Surface Energy Balance System (SEBS).


Introduction
Mountainous area covers about one-fifth of the earth's continental areas (X.Yang et al., 2011).Accurate surface solar radiation estimations are essential for studies of solar energy resource, hydrological processes, and climate change.Solar radiation exerts strong control on available energy exchanges at the surface.Knowledge of the spatial distribution of solar radiation in mountainous area is therefore vital for the energy exchange process between the atmosphere and the mountain land surface.Terrain determines whether a surface receives direct radiation or if it is shaded.In zones of complex topography, variability in elevation, surface slope and aspect create strong spatial heterogeneity in solar radiation distribution, which determines air temperature, soil temperature, evapotranspiration, snow melt and land-air exchanges.The spatial and temporal distribution of surface radiation exerts a fundamental control on mass and energy exchange between air and land.The mountainous areas are often remote and inaccessible to carry out measurement of land-air interactions.The zones of complex topography therefore form interesting but little studied areas for land-air exchange studies.
Recent studies have explored approaches to estimate the regional distribution of surface heat fluxes with observational data of different satellite sensors (Ma et al., 2006(Ma et al., , 2011;;Oku et al., 2007;W. Ma et al., 2009).Remote sensing based turbulent flux algorithms can be divided into two-source (like TSEBS of Anderson et al., 2008) and single source models (like SEBS of Su, 2002).Yang et al. (2003) pointed out that a single source heat transfer model is applicable on the Tibetan plateau; SEBS was adopted here.The Surface Energy

X. Chen et al.: Estimation of surface energy fluxes under complex terrain
Balance System (SEBS) developed by Su (2002) has been designed to estimate energy partitioning by using satellite and meteorological data.While most of the studies using SEBS derive surface energy balance items located at flat areas (Su et al., 2005;Yang et al., 2010), none of them consider the influence of topographical influence.With the development of satellite sensor grid resolution, when applying SEBS to the high resolution satellite dataset, the topographic influences become increasingly important.Terrain controls how much sky is visible and therefore influences incident diffuse and reflected sky radiation.Since surface solar radiation measurement is very sparse in the mountainous region, the knowledge of the terrain is thus important for the radiation balance and further for the surface energy balance in complex terrain (Tovar-Pescador et al., 2006;Aguilar et al., 2010;Long et al., 2010).The aim of this research was to combine a topographically corrected solar radiation (the amount of shortwave radiation received under clear-sky conditions) with SEBS over the Tibetan Plateau mountain area.A topographically enhanced surface energy balance system (TESEBS) was developed to generate a series of distributions of surface energy balance in a meso-scale area on the north area of Mt.Qomolangma over the Plateau.Small lakes, rivers, glacier, and surfaces with short canopies are all included in the study area (Fig. 1).
The surface energy balance analysis around Mt. Everest was studied with measurement at point scale (Zhong et al., 2009;Zou et al., 2009).The aim of this research is to upscale in situ point observations of land surface variables and land surface heat fluxes over regional scale using high resolution remote sensing data.In mountainous regions, due to the complex topography, high-resolution data are needed.Landsat TM/ETM+ sensors include optical and thermal sensors with higher image resolution.Here we use Landsat data to determine regional land surface heat fluxes around the area.
In this study, we make efforts to improve solar radiation estimation under complex terrain.The shortwave radiation reaching the surface of the earth is divided into direct, diffuse or reflected radiation.Direct radiation reaches the surface of the earth from the solar beam without interactions with particles in the atmosphere.Diffuse radiation is scattered out of the solar beam by gases and aerosols before reaching the surface.Reflected radiation is mainly reflected to the surface from surrounding terrain and is therefore important in mountainous areas.A knowledge of the values for each component is often required when considering the topographic effects on each radiation component separately (Aguilar et al., 2010).To get an accurate incoming solar radiation flux in mountainous terrain, a radiation model which considers the shading and reflecting effects of adjacent features is needed by SEBS.At each point, the direct, diffuse, and reflected solar radiations were estimated.The global radiation was obtained by adding the direct, diffuse and reflected radiation.The intention of this study is to compute the instantaneous solar radiation with the above three radiation variables for various slopes and azimuth terrains in order to make each component of the energy balance system more accurate.
A topography dependent incoming radiation model combined with SEBS is introduced in Sect. 2. The study area and remote sensing data preparations are included in Sect.3. Remote sensing applications over the Qomolangma area were described in Sect. 4. The relationship between surface energy items and land cover types, and the spatial distribution of surface energy balance under complex terrain are examined in Sect. 4. Discussion and conclusions are given in Sect. 5.

Model formulation
The surface energy balance equation is written as where R n is the net radiation; G 0 is the ground heat flux; H is the turbulent sensible heat flux, and LE is the turbulent latent heat flux.Latent heat flux, LE, is computed as the last item of the surface energy balance equation after derivations the of other three variables in Eq. (1).

The instantaneous net radiation
Net radiation is a critical input variable in the energy balance equation and the most sensitive variable in latent heat flux estimate (Zhang et al., 2005).Therefore, the accuracy of retrieved net radiation determines the accuracy of estimates of latent heat flux and ET (evapotranspiration) to some extent.
For the computation of surface energy balance in complex terrain, a highly detailed surface radiation balance model is necessary.Topography is well known to alter the shortwave radiation balance at the surface.In order to use the surface energy balance equation over the complex topography of the Plateau, further efforts are needed to improve spatial accuracy of satellite based surface energy budgets in mountainous regions.A detailed radiation balance model is therefore required by this study.Thus here we introduce the surface radiation model as follows.
The net radiation flux is computed with the following equation: where α is the broadband albedo, SWD is downward surface shortwave radiation.LWD and LWU are downward and upward surface longwave radiation.On flat terrain and under clear-sky conditions, the downwelling shortwave radiation is nearly the same from point to point over relatively large areas and so one measurement can be taken to be representative of the entire regional area (Bastiaanssen, 2000;van der Kwast et al., 2009).However, direct measurements are rarely available to represent the shortwave radiation over most mountainous areas.Therefore, in mountainous regions a detailed solar radiation balance solution is required by surface energy balance equation.Parameterization models are often used to make predictions of individual solar radiation components under clear sky conditions (e.g., Yang et al., 2001;Liang et al., 2012).Meanwhile, the topographic effects are rarely considered.Here we employ surface radiation parameterization models and solar radiation transfer above an inclined surface to account for the impact of complex terrain, which follow the simple form of the Angstrom-Prescott model (Prescott, 1940), and its inputs (air temperature and relative humidity) are easily accessible from routine surface meteorological observations.

The instantaneous downward shortwave radiation
The surface downward shortwave (solar) radiation is divided into three parts over complex terrain: direct radiation (I b ), diffuse (I d ) and reflected (I r ) insolation.
The downward shortwave radiation varies in response to altitude, surface slope and aspect.The parameterization schemes for calculating the instantaneous solar radiation were improved by accounting for the three part variations to slope and azimuth of land surface and terrain shadow in mountainous areas.Studies have described how to use a digital elevation map dependent model to compute direct solar radiation, diffuse and reflected insolation (Kumar et al., 1997).According to the knowledge, the method used to compute distribution of downward shortwave radiation over mountainous areas is as follows: where θ is solar incidence angle; a is the solar altitude angle; s is the tilt angle of the surface (slope); τ c solar beam radiative transmittance; τ d solar diffuse radiative transmittance, and τ r is the reflectance transmittance; r is the ground reflectance.
The applications of optical remote sensors by SEBS are limited to conditions of cloudless sky; therefore, in this paper we consider atmospheric transmittances under conditions of cloudless sky.The clear sky radiative transmittance is based on local geographical and meteorological conditions.The clear sky surface solar radiation is affected by a number of extinction processes in the atmosphere.Although Kumar et al. (1997) suggested a method for computation of τ c , τ d and τ r , their method did not consider the difference in Rayleigh scattering, aerosol extinction, ozone absorption, water vapor absorption and permanent gas absorption.Yang et al. (2001) developed a broadband radiative transfer model based on Leckner's (1978) spectral model.Evaluated as one of the best broadband models (Gueymard, 2003a, b), here we used the model to calculate solar beam radiative transmittance, τ c , and solar diffuse radiative transmittance, τ d , under clear skies.τ c is computed as function of radiative transmittance due to ozone absorption, water vapour absorption, permanent gas absorption, Rayleigh scattering, and aerosol extinction, respectively.The detailed solution of τ c is described in Appendix A. τ d and τ r are computed with τ c .The detail solutions for I d and I r are presented in Appendix B.
The influence of tilted surface on surface radiation is expressed by solar incidence angle, solar altitude angle and topographic information shown in Eqs. ( 4), ( 5), and (6).A high resolution DEM map (obtained from the US Geological Survey Earth Resources Observation and Science center) of SRTM (Shuttle Radar Topography Mission) in the region was used to calculate slope and aspect of each pixel.The slopes and aspects were then used in subsequent executions to generate solar radiations in complex mountainous areas.
Longwave emission from different terrains was taken as isotropic here.The upward longwave radiation is computed using the Stefan-Boltzmann equation: where ε is the "broad-band" land surface emissivity, derived from a satellite based method referred to Chen et al. (2013).
When NDVI < 0 and α < 0.47, the pixel was taken as water surface, where ε = 0.985; if α ≥ 0.47 the pixel was assumed as snow, where ε = 0.99; LST is land surface temperature.

The instantaneous soil heat flux density
The regional ground heat flux G 0 in Eq. ( 1) cannot directly be mapped from satellite observations.The ground heat flux, as an indirect variable in surface energy balance, was calculated through net radiation according to a different surface dependent ratio value.The relationship between G 0 and R n (Kustas and Daughtry, 1990) over bare soil in this area is For water area (NDVI < 0 and α < 0.47), we use the equation: G 0 = 0.5R n (Gao et al., 2011).For glacier area, G 0 is negligible according to Yang et al. (2011) and we use an equation of G 0 = 0.05 R n .The glacier is distinguished according to LST < = 273 K. Daughtry et al. (1990) investigated that the midday G 0 / R n ratio is predictable from vegetation indices.For the canopy coverage area, the following equation is adopted: where r s and r c are the ratios between ground heat flux and net radiation for bare soils and surfaces with fully covered vegetation respectively.The r s in this area is given a value of 0.315 (Kustas and Daughtry, 1990), and r c has a value of 0.05 (Monteith, 1973).The fractional vegetation cover f c is determined using the normalized difference vegetation index NDVI in Eq. ( 20).

The instantaneous sensible heat flux density
The sensible heat flux was computed by means of Monin-Obukhov similarity theory (MOST) theory with (Eq.12): where H is the sensible heat flux.k is the von Karman constant.u * is the friction velocity.ρ is the density of air.C p is specific heat for moist air.θ 0 is the potential temperature at the surface, θ a is the potential air temperature at height z, d is the zero plane displacement height, h is the stability correction functions for sensible heat transfer (Brutsaert, 1999), and L is the Obukhov length.The roughness height for heat transfer (z 0h ) or kB −1 must be accurately determined before the computation of sensible heat flux.Based on the fractional canopy coverage, Su et al. (2001) and Su (2002) give kB −1 as the following equation: where f c is the fractional canopy coverage and f s is that of soil.kB −1 c is the kB −1 of the canopy and kB −1 s is that of the bare soil.Being testified as a novel one over the Tibetan Plateau (Yang et al., 2002;Chen et al., 2010), the parameterization method of Yang et al. (2002) was introduced into kB −1 s equation.The soil part of kB −1 s in Brutsaert (1982) in Eq. ( 13) is revised as follows according to Chen et al. (2013): where ϑ is the kinematic viscosity of air (1.5×10 −5 m 2 s −1 ).β equals 7.2 s 0.5 m −0.5 K −0.25 .u * is the surface friction velocity (m s −1 ), and θ * is the surface friction temperature (K).z 0m is momentum roughness length (m) referred to Eq. (7d) in Chen et al. (2013).

Study site and data processing
Two scenarios have been used to assess the TESEBS model.One is to use all available in situ measurements at QOMS/CAS as forcing data.The other is to use operational meteorological and satellite data.The model outputs of surface energy fluxes are then compared with flux tower measurement.

Study area
The Himalaya, as the south barrier of the Plateau with a large area of high mountains pierces directly to the middle troposphere of earth; consequently, it has great influence on the weather and climate of the region (Ye and Gao, 1979;Gao, 1981;Bollasina and Benedict, 2004;Ueno et al., 2008;Zhou et al., 2008;Zhong et al., 2009;Zou et al., 2009).The Himalaya exerts profound thermal and dynamical influence on atmospheric circulation (Bollasina and Benedict, 2004).The Himalaya mountains provide the water sources for the Indus, Ganges, and Brahmaputra rivers, which supply water to billions of people in Asia (Immerzeel et al., 2010).Considering the importance of the Asian monsoon (Boos and Kuang, 2010), Bollasina and Benedict (2004) pointed out that land-atmosphere interactions over Himalaya requires particular attention.The study was conducted in 28 • 0 42 N to 29 • 0 54.9 N and 86 • 9 5 E to 87 • 1 42.8 E, located around QOMS/CAS station (Fig. 1), a comprehensive observation and research station on the north slope of Himalaya.Elevation in the study area changes between 3700 to 7107 m.This region has been chosen because it is representative of a high-alpine environment and because of the glaciers, lakes, rivers, and short canopies present in this area (see bottom left picture of Fig. 1).

Field measurements
The For all fluxes in this paper, the same sign convention applies: fluxes transporting energy towards the surface are negative, and fluxes transporting energy away from the surface are positive.

Local-scale evaluation
Standard meteorological forcing data at QOMS/CAS, including wind velocity, vapor pressure, air temperature, and atmospheric pressure, were used to run the TESEBS model.Land surface temperature is derived from measured upward longwave radiation.A constant emissivity of 0.97 is assigned to estimate the surface temperature.Time series comparison of the model output and observations show that TESEBS correctly interprets variability and is capable of accurately representing the temporal development of surface energy balance items at the local scale (Chen et al., 2013).
The accuracy of the model was evaluated using mean bias (MB) and root mean square error (RMSE).These statistical indicators are defined as where x i stands for simulations of TESEBS.obs i stands for observations, and N is the sample number.There were 4616 data points with a data quality level < = 3 used to evaluate TESEBS in Fig. 2. Linear regression between modeled and measured values was also computed.The slope value (= 1.01) in Fig. 2 demonstrate that TESEBS provides a good estimation of sensible heat flux with the new kB −1 equations.The correlation coefficient R is as high as 0.91, with MB = −7.3W m −2 , and RMSE = 41.76W m −2 .

Remote sensing data preparation
To capture the heterogeneity of the land surface over the domain, high-resolution satellite data are required.The fraction of cloud cover is not more than 3 % on these days.In mountainous areas, topography also strongly influence the signal recorded by space-borne optical sensors.
The topographic influence on the satellite received signal were eliminated by the method of Richter et al. (2009).The reflectivity for each band (ρ λ ) is calculated as where L λ is the spectral radiance for each band.ESUN λ is the mean solar exo-atmospheric irradiance for each band.θ is solar incidence angle.s is the surface slope.The solar exo-atmospheric irradiance for TM and ETM+ 1, 2, 3, 4, 5, 7 band in Markham and Barker (1987) and Chander and Markham (2003) were used separately.θ is the solar zenith angle (from nadir), and d r is the inverse squared relative Earth-Sun distance.The surface albedo (α) for shortwave radiation is retrieved from converting narrowband to  broadband planetary reflectivity which is obtained as the total sum of the different narrow-band reflectivity according to weights for each band.The weights for the different bands are given by Teixeira (2010).Broadband shortwave surface albedo was calculated from the normalized reflection values of channels 1, 2, 3, 4, 5 and 7, using the following equation: where ρ 1 -ρ 7 are the reflectivity for band 1 to 7. The spectral radiance of band 6 is converted into a brightness temperature applicable at the top of the atmosphere by inversion of Plank's law (Teixeira, 2010).The LST is calculated by monowindow algorithm (Qin et al., 2001;Sobrino et al., 2004).
The NDVI is computed as the ratio of the differences in reflectivities for the near-infrared band (ρ 4 ) and the red band (ρ 3 ) to their sum.The vegetation fractional coverage is estimated using formulation: The value of NDVI min and NDVI max is about 0.2 and 0.5 (Sobrino et al., 2004).In the case of NDVI < 0.2, f c = 0.The broad band emissivity ε is used to calculate total longwave radiation emission from the surface.This broadband emissivity was calculated from the NDVI according to the method of Sobrino et al. (2004).To maintain a spatial consistency, NDVI, albedo and other data were interpolated to corresponding thermal infrared band using a linear technique.

Weather data
To compute surface fluxes over the area of one satellite image, the spatial distribution of meteorological data (air temperature, atmospheric pressure, relative humidity etc.) at PBL height or near-surface height at satellite pixel scale is required.The spatial interpolation method is often used to get these meteorological data from meteorological stations or atmospheric reanalysis data (Ma et al., 2006;Oku et al., 2007;McVicar et al., 2007).Xie et al. (2007) pointed out that meteorological elements above Mt.Everest coincided with measurements at a meteorological station 60 km away.Then we assume the meteorological measurement at QOMS/CAS station can represent the synoptic situation of our research area.Air temperature for each grid cell was adjusted with respect to elevation, assuming a standard air temperature lapse rate of 6 K km −1 (Chen et al., 2007).Atmospheric boundary layer height of 600 m is used according to the results of Y. Ma et al. (2009).Due to the elevation changes from 4000 to higher than 8000 m, the corresponding surface pressure changes significantly.Thus the surface pressure is estimated by where p 0 = 101 325 Pa, z is DEM data in a unit of m.

Evaluations of TM/ETM+ based TESEBS results
Instantaneous surface energy balance items at the satellite overpass time are highly dependent on the estimation of key variables, namely, land surface temperature, albedo, downward and upward shortwave/longwave radiation, etc.Hence, we evaluate these variables by comparison with sitespecific ground-based measurements (Fig. 3).The correlation coefficient of downward shortwave radiation (SWD) between TESEBS simulated and measured by radiometers is as high as 0.99, with MB of −9.62 W m −2 and RMSE of 45.4 W m −2 .The LST has a mean bias of 1.48 K with a high R value.The albedo derived from TM/ETM+ is a little bit lower than the in situ true values.The SWU of TESEBS have a mean value of 42.55 W m −2 higher than the observation.The final surface energy balance items were also evaluated with in situ measurement (Fig. 4).The MB values of R n , H , G 0 and LE is about −23.6, −15.8, 7.7 and −6.8 W m −2 , respectively.Overall, the values derived from TM/ETM+ by TESEBS agree well with ground measurements.The comparison of TESEBS and SEBS statistical variables were listed in Table 1.Both RMSE and MB of TESEB are lower than SEBS, which shows the better performance of TESEBS.
Figure 5 shows the distribution maps (1605 × 1882 pixels) of each surface energy balance item based on TM remote sensing data obtained on 9 April 2010.The overpass time is 10:32 LT (local time), the sun azimuth angle is 127.2 degree, and the sun elevation is 58.15 degree.To maintain consistency in spatial resolution, the remote sensing data of band 1, 2, 3, 4, 5, 7 were interpolated to 120 m × 120 m.The experimental area presents extreme variability characterized not only by steep slopes and altitude variations of thousands of meters, but also by a variety of land surfaces such as grassy marshland, several small rivers and lakes, bare soil and glaciers.Therefore, these derived parameters show a wide range due to the strong contrast of surface features.Due to low albedo of the water surface, the net radiation and  ground heat fluxes over the small lakes are relatively higher than in other places (Fig. 5c, d), and the corresponding sensible heat flux is lower than in other places around the bodies of water, while the latent heat flux is higher (Fig. 5a).The sensible heat flux over the glacier area is dominated by negative values.Slopes which are exposed to the East consequently have a higher net radiation and sensible heat flux than slopes with an exposition to the West.
The processes and mechanisms of energy and mass transfers become complicated under the complex terrain due to the substantial differences in radiation availability caused by various slopes and azimuths of surfaces.The surface net radiation in the region changes from −155 to 461.5 W m −2 , sensible heat flux from −25.2 to 265.6 W m −2 , and latent heat flux from −4.5 to 257.9 W m −2 .The surface flux maps reflect distinct mechanisms of energy partition and water evaporation of various land cover types, showing differences in the spatial distribution pattern of surface turbulent heating.
West facing slopes are receiving only about half of the shortwave radiation of the eastern slopes (Fig. 7a).Such variations would surely have a significant effect on the heat budget of different places, thus influencing latent and sensible heat fluxes.It can be seen that the values in the region present a huge difference visible in the distributed values of the variable cell by cell, but also noticeable was the wide range of global values in the region, when topographic factors are taken into account, as shown by the maximum, minimum values.The spatial gradient in each part of solar radiation is evident (Fig. 6).It can be seen that the locations receiving more radiation are those in the highest part of the region, with an east-facing orientation that remains unobstructed during the hours of satellite pass over.While the west-facing slopes which are shaded by the terrain at the satellite pass over time receive a relative low solar beam radiation (Fig. 6b), and a relative high diffuse radiation (Fig. 6c).The surface energy balance has the greatest influence on environmental processes, especially if snow is present.Over the snow covered glacier area on the bottom left corner of the map (Fig. 6d), the reflected solar radiation can be around 100 W m −2 , which is significant compared to low values of other surface energy variables over there.This also demonstrate that the glacier topography does play a fundamental role in determining the surface energy balance (Arnold et al., 2006).

Discussion and conclusions
Dealing with regional land surface heat fluxes over heterogeneous landscapes is not an easy job.In order to analyze the interactions between the land surface around Mt. Qomolangma and the atmosphere above it, a Topographical Enhanced Surface Energy Balance System (TESEBS) was developed to upscale energy and turbulent heat fluxes from point to a meso-scale.When using high resolution satellite data over mountainous areas to get the surface energy balance items, the terrains effects must be considered.In this study, a radiation parameterization scheme for grid topography accounting for shading, and terrain reflections is used to get the surface radiation and energy balance system.DEM information was used to characterize the topographic role  in the spatial distribution of surface energy balance items in Qomolangma region's complex topography.Each radiative flux is parameterized individually as a function of slope, sun elevation angle, and albedo.We quantify the topographic impacts on each individual shortwave radiation (solar beam, diffuse and reflected radiation) with real topographies.Variations in atmospheric transmissivity resulting from actual column water vapor, ozone, and aerosol have been considered in our clear-sky satellite applications.
TESEBS was evaluated from site point and a regional scale.Firstly TESEBS was forced by a long time meteorological observation data at point scale.The performance of TESEBS has been evaluated by comparisons between its output fluxes and a high quality dataset of observed turbulent fluxes.Then TESEBS was expanded to a regional scale, where glaciers, bare soil and small lakes all present.It is difficult to get spatial distributions of meteorological inputs, hereby we spatialized the meteorological measurements at the station to present synoptic situation over the complex region.The distributions of surface energy balance correspond well with the land surface class.The significant incidence of topography on the values of surface energy balance throughout the region has been demonstrated by the proposed topographic solar radiation algorithm.Results indicate that surface flux predictions from TESEBS perform well at local scales on a flat terrain, when assessed against in situ flux measurements derived from eddy covariance approaches, and provides realistic outputs at regional scale for more complex topography.Due to lack of measurement on different aspect and slopes, it is impossible to do fully evaluations for this complex topography.
This work helps us to analyze the possibility and suitability of TESEBS to model surface turbulent heat flux over typical land covers of the Plateau by remote sensing technique.The performance of TESEBS over the glaciers also makes it possible to study the energy balance of the snowpack, and validate snowmelt runoff models in the future.Opportunities also exist for improving the performance of both models via data assimilation and model calibration techniques that integrate remote sensing based surface energy flux predictions.
However, the topography effects on the roughness length still remain a blank area at present.Advection could be formed under complex terrain, which may complicate the energy balance at point scale.In the future, further validation of the parameterization method needs to be made over the Plateau water surface and other land covers.

Figure 1 .Fig. 1 .
Figure 1.The landscape of the Tibetan Plateau (top), study area (bottom-left picture, composites 841 Qomolangma Station for Atmospheric and Environmental Observation and Research, Chinese Academy of Sciences (QOMS/CAS) is located at 28 • 21.63 N, 86 • 56.93 E, with an elevation of 4276 m, and 30 km away from Mount Everest.It was established by the Institute of Tibetan Plateau Research (ITP), the Chinese Academy of Sciences(Ma et al., 2008).The dataset of the QOMS/CAS station consist of surface radiation budget components (CNR-1, Kipp & Zonen), vertical profiles of air temperature, humidity, wind speed and direction (MILOS520, Vaisala), turbulent fluxes measured by eddy correlation technique.Sensors of wind speed, wind direction, air temperature, and relative humidity at five levels (1.0, 2.0, 4.0, 10.0, and 20.0 m) are installed on a 40 m p.b.l.tower.The soil temperature was measured at depths of 0, 10, 20, 40, 80 and 160 cm.The soil moisture content was measured at depths of 10, 20, and 30 cm.The soil heat flux was measured using soil heat flux plates (HFP01) buried at a depth of 10 cm.The soil heat flux at the surface is calculated by adding the measured flux at 10 cm depth to the energy stored in the layer above the heat flux plates(Chen et al., 2012).An open path eddy covariance turbulent measurement system (CSAT3, Campbell; LI 7500, Licor) is set up at 3.5 m height.The high frequency turbulent data are processed by TK2 software(Mauder and Foken, 2004) to control the quality of every half-hour flux.Calibrations are also done in the process.TK2 also produces a quality level of turbulent fluxes with the definition of the quality level given by Rebmann et al. (2005) divided into 1-5 levels.The high quality fluxes (quality level < = 3) were used in this study.

Figure 2 .
Figure 2. Evaluation of TESEBS based on the point sensible heat flux data of high quality (with 850

Fig. 2 .
Fig. 2. Evaluation of TESEBS based on the point sensible heat flux data of high quality (with quality level < = 3) in 2007 at QOMS/CAS station.Scatter plot of sensible heat flux (H , unit of W m −2 ) between the measurements of eddy covariance (ED) and outputs of TESEBS.The thick line is 1 : 1 line.The linear fitting line is dashed line.Root mean square error (RMSE).Correlation coefficient (R).Mean bias (MB).

Figure 3 .Fig. 3 .
Figure 3.Comparison of derived results with field measurements for the (a) land surface 859

Figure 4 .
Figure 4. Comparison of derived results with field measurements for the (a) net radiation (Rn, 864

Figure 5 .Fig. 5 .
Figure 5. Distribution of each surface energy balance item over the study area of Fig. 1 (b) at 877

Figure 6 .Fig. 6 .Figure 7 .
Figure 6.Distribution of each surface solar radiation item over the study area of Fig. 1 (b) at 885

Fig. 7 .
Fig. 7. Distribution of each surface radiation balance item over the study area of Fig. 1b at 10:30 LT on 9 April 2010, the cross line is the location of the station, SWD and SWU are downward and upward surface shortwave radiation.LWD and LWU are downward and upward surface longwave radiation.The geographical information is the same as bottom-left picture of Fig. 1.

Table 1 .
The comparison between TESEBS and SEBS statistical variables.