Keywords: Energy distribution, Remote sensing, GIS, China

Abstract
A land surface feature model (LSFM) and energy exchange model based on remote sensing and GIS have been developed, including databases of land cover type, soil texture, elevation, climate planning, phenology and NOAA-AVHRR Images etc.. Monthly mean roughness length of land surface and albedo was calculated on basis of the databases applying statistic model and BRDF model. Land surface temperature is derived using NOAA-AVHRR data and split window models aiming to different climate region, then distribution of energy was calculated. Complementary relationship model, Penman-Monteith model and other statistic models were employed to calculate monthly mean evapotranspiration for whole China

The temporal and spactial distribution of land surface feature and energy was studies and discussed. The results shown that different climate and land cover intensity were main factors that affected physical parameters of land surface, and effect of snow was more stands out. The land cover type was the most important factor that affectted the distribution of land surface feature and energy exchange. Meanwhile, Distribution of cold and heat sources in 1997 for was studied. There was larege cold regions in Northeast, Xinjiang and Qingzang plateau. The factors that affect temporal-spatial distribution of albedo and latent heat were analyzed.

Introduction
Earth is a complicated great system. Energy source of his movements and living process comes from Sun directly or indirectly, but the energy distribution is distinct inhomogeneous at temporal and spatial scales due to Earth’s movement and different latitude, so that the importance of this effect excesses the effect of solar activity on the Earth’s system.

Temporal and spatial distribution of physical characteristics of land surface changes easy. Dynamic and thermodynamic actions of land surface with atmosphere due to the diversity of land surface feature are great difference. Every type of land surface has distinguished way of energy distribution and mass exchange. Change of land surface feature impacts on the balance of energy, momentum and mass between land and atmosphere, thereby affects local, regional even global climate changes. Therefore research on temporal and spatial distribution of energy of land surface is important significance for the research on interaction of land with atmosphere, global climate change and global change.

It is necessary to build data bases of land surface type in order to provide priori knowledge for global climate model (GCM) and to analyze land surface energy distribution due to differences of its physical characteristics and its role in exchanges of energy, momentum and mass. This is also prerequisite condition for inversion of albedo, land surface temperature and land surface roughness. Albedo and temperature of land surface reflect information of structure in vegetation and energy distribution, roughness describes intensity of turbulence exchange, and finely determine energy distribution in land – atmosphere system. China as a large country has very complicated land type and cross several climate zone. Methods of conventional measurements are not meet practical requirement in simultaneous and representative nature. Remote sensing and GIS provide power tools for study on land surface feature and temporal and spatial distribution of its energy.

Methods
Remote sensing can provide temporally land information with local, national and global scales. Data bases and spatial analysis models based on GIS can realize extraction of land surface feature, calculation and analysis of energy distribution of land surface.

Pre-processing of remotely sensed data
Pre-processing of remote sensing data must be conducted using NOAA-AVHRR data to inverse monthly parameters, albedo and temperature of land surface, including rectification, mosaic of different strap and projection change, reducing clouds and discrimination of cloud and snow because we need cloudfree data and also retaining snow information. A progressive approach method was used to distinguish cloud and snow as following:

• Reflectance of channel 1 R_ch1 >W₁ ;
• Normalized difference vegetation index NDVI£W₂ ;
• Brightness temperature of channel 3 and channel 4 T_ch3 - T_ch4 ³W₃ ;
• Brightness temperature of channel 4 T_ch4 >W₄ ;
• Reflectance of channel 3 R_ch3< W₅ .

Where W_n (n=1, 2, 3, 4, 5) is threshold. Corresponding threshold was calculated for different climate zone and month because different climate zone affects the thresholds. Atmosphere effect on NOAA-AVHRR data was corrected for calculation and comparison of monthly parameters of land surface ( Qin and Tian, 1994).

Building databases
It needs the data of climate, land cover, soil, phenology and topography for development of land surface feature model and calculation of land surface energy distribution. The following data bases have been built:

• data base of climate planning in China;
• data base of monthly land cover type in China;
• data base of soil texture in China;
• data base of phenological distribution in China;
• data base of land elevation in China;
• data base of climate in China.

Extraction of land surface feature in China
Exchange of land surface energy, momentum, mass is sensitive to land cover. Yearly mean type of land cover is not representative to monthly type of land cover because of difference of large area and climate change. Daily NOAA-AVHRR data were used to extract monthly land cover and build data base of monthly type of land cover based on data base of national resources at the scale of 1:4,000,000. The land cover was divided into 20 types, which included land cover type of IGBP required.

Inversion of land surface albedo in China
The semi-spherical albedo of land surface, as a part of energy balance model, decide s energy distribution in energy exchange between land and atmosphere. Climate model of computation albedo of land surface is not meet the requirement in spatial resolution (Dicknson, 1986). Calculation of albedo by using remote sensing has much more advantage. There are two models: direct inversion and indirect inversion models.

The direct inversion model is based on distribution weight of value in solar band observed by satellite to establish statistical model aiming different land cover (Brest, 1987). The most representative model of indirect inversion of albedo is kernels-driven model. The basic consideration of this model is to extract feature quality – “kernel” which closely relates vegetation type. The linear correlation between kernel and BRDF of land cover has been established (Walthall, et al, 1985, Wolfgang, et al, 1995). In this paper, a combination of the two models was employed to calculate monthly albedo of land surface in China.