In the headwater regions of most large rivers in China, the glacier, permafrost and seasonal frozen soil have degraded largely, and the snowline has also risen in the recent 50 years, under global warming, especially in an inland river basin, where the main water resources come from mountainous river basin. Thus, it is very necessary to quantify the hydrological processes in these mountainous river basins, according to the fieldwork results and using hydrological modeling. However, the distributed hydrological model that describes the water-heat coupled processes is very rare, at the present time. Took the mountainous watershed of Heihe river basin, with an area of 10009km², as an example, the authors designed a distributed water-heat coupled (DWHC) model. The minimal computing unit is a DEM (Digital elevation model) grid, with a spatial resolution of 1 km×1 km, and the reference frame is Alberts. The time step of the model is one day. The soil and vegetation of the watershed is divided into 18 and 9 types, respectively. In each computing grid, the soil is divided into 3~5 layers, according to the landuse types. The DWHC model included 8 sub-models, which were meteorological model, vegetation interception model, snow and glacier melting model, soil water-heat coupled model evapotranspiration model, runoff production model, infiltration model and flow concentration model. The water-heat coupled processes, based on the continuous water and heat equation, runs through the runoff production processes, infiltration processes and evapotranspiration processes. The DWHC model gave a simple numerical solution to the continuous water and heat equation, according to the soil frozen states. The meteorological inputs are daily precipitation, daily averaged air temperature, and potential evapotranspiration, which come from the meteorological stations, or from the climatic models such as MM5. The soil and vegetation characters should also be described. At that time, the model could calculate the soil temperature, soil liquid water content, soil solid water content, sense heat, latent heat, soil water tension and runoff amount, etc., given the initial soil water content and soil temperature. This paper just described the model principles, and the model results using the data at the meteorological stations as inputs, or using the MM5 results as inputs, would be discussed in the following papers.