INTRODUCTION

The formation of a distributed biosphere hydrological model (DBHM) for use in large river basin is presented. The model system DBHM is a continuous-time spatially distributed model, integrating hydrological processes and vegetation- atmosphere transfer processes at the river basin scale. DBHM incorporates a previously developed land surface model SiB2 (Sellers et al., 1996) and a distributed hydrological sub-model, including the use of satellite data to describe vegetation state and phenology and the use of Digital Elevation Model (DEM) data to describe geomorphological characteristics. The principle motivation for formulating DBHM was to provide more realistic estimates of evapotranspiration and runoff over large river basin. The model can be used to investigate runoff change according to land cover change and to assess water resource in large basin.

Hydrological models with a spatial structure are being increasingly based on DEM or Digital Terrain Model (DTM) (Moore et al., 1988a; Moore et al., 1988b). DEMs automatically extract topographic variables, such as basin geometry, stream networks, slope, aspect, flow direction, etc. from raster elevation data. Many of the existing models, such as SHE (Bathurst et al., 1995), TOPMODEL (Beven, 1995), GBHM (Yang et al., 1998; Yang et al., 2000), etc., use DEMs to represent geographical characteristics of a watershed. The adaptation of DEM improved the representation of runoff accumulation processes. However, the conceptual prescription of land surface and empirical evaporation calculation were recognized weaknesses of most of the distributed hydrological model. Distributed representation of land surface is still a challenge for large scale modeling because of the limitations of global observations. Satellite data gives a chance to describe the vegetation phenology. Vegetation index data acquired from meteorological satellites were processed to derive time series fields of the Fraction of Photosynthetically Active Radiation absorbed by green vegetation canopy (FPAR), the total Leaf Area Index (LAI), and the canopy greenness fraction (N). The remotely sensed data can be integrated into hydrological model and enhance the accuracy of evaporation calculation. Several scientific developments prompted a radical improvement of DBHM. First, meteorologists provided new insights into heat flux in the Soil-Vegetation-Atmosphere Transformation (SVAT) processes (Sellers, 1985; Sellers et al., 1986). Second, remote sensing became matured as a tool to get reliable land surface information. Third, topographic variables extracted from DEMs provided a plausible way to describe runoff accumulation on ground and in subsurface (Yang et al., 1998; Verdin and Greenlee, 1996). As more and more data have been collected in computerized database, in particular, in Geographical Information Systems (GIS), the data availability has improved significantly (Andersen et al., 2001). Readily accessible data is used in construction of large scale distributed models (Yang and Musiake, 2003; Vörösmarty et al., 1989; Abdulla and Lettenmaier, 1997).

The advantages of the DBHM can be summarized as follows:

1. The calculation of evaporation is more reliable by using a physical land surface scheme for simulation of heat flux in the SVAT processes.
2. Satellite data is used to describe time serial land cover change. It becomes possible to estimate hydrological responses in ungauged basin by using readily accessible data.
3. The meteorological observations from hydro- meteorological stations, such as precipitation, temperature, humidity, vapor pressure, sunshine duration, and wind speed, are used in the model. The model could be coupled with atmospheric model by using the outputs from atmospheric models instead of observations.