Due to the spatial topographic heterogeneity in semi-humid and semi-arid regions, hydrological simulation is always a difficult problem as saturation-excess runoff and infiltration-excess runoff is coexisting during a storm. Most existing hydrological models cannot be customized to simulate flood processes of different areas because of their fixed structures and modes. This study developed a spatial combination model (SCMs) by combining the Xin'anjiang model (XAJ), Xin'anjiang-Green-Ampt model (XAJG) and Green-Ampt model (GA) for simulating runoff in the sub-watersheds where either saturation-excess or infiltration-excess runoff mechanism dominates. Curve number-topographic index method was used for dividing a watershed into infiltration-excess dominated and saturation-excess dominated sub-watersheds. The proposed models were tested for two selected watersheds:Dongwan watershed in semi-humid area and Zhidan watershed in semi-arid area. The results showed that the parameters of the Dongwan watershed are dominated by hydrological models, while the parameters of the Zhidan watershed are greatly influenced by the dominant runoff mechanism. The simulation results of the partial saturation-excess models are better than those of partial infiltration-excess models in Dongwan watershed. SCM2 model (combination model of XAJ and XAJG) simulated has the best performance (runoff depth error Re=75% and flood peak error Qe=75%), just as the SCM5 model (combination model of GA and XAJG) in Zhidan watershed (Re=53.3% and Qe =53.3%). In semi-arid and semi-humid watersheds, driven by topography and soil data, the SCMs have a more realistic model structure and parameterization with higher simulation accuracy and stability.