The Integrated Biosphere Simulator (IBIS) model is an important tool for terrestrial carbon cycle simulation. One of the key ecological processes of the terrestrial carbon cycle is soil respiration. Modeling soil respiration and revealing its space-time pattern using the IBIS model is of great significance to the study of the terrestrial carbon cycle and global change. However, the application of the IBIS model to a catchment with complicated terrain has not been attempted because of problems with the model mechanism. In this study, the IBIS model was improved by addition of a terrain analysis module, and modification of the redistribution module of soil water and the calculation module of solar radiation received at the ground surface. Soil respiration of the five forest types in the Zhangjiagou catchment of Heilongjiang province in northeast China was then simulated in 2004 using the improved IBIS model. The model was driven by terrain data, vegetation parameters, soil texture parameters and climate variables. The simulated values were validated with measurements, and the temporal-spatial patterns of forest soil respiration and the relationship between soil respiration and soil temperature and humidity were analyzed. The five forest types were: Larch plantation, Oak forest, Aspen-birch forest, Hardwood forest and Mixed deciduous forest. Results showed that: (1) The simulated daily soil respiration by the improved IBIS model had a significant relationship with the measured daily soil respiration. The improved IBIS model can be better used in simulating and estimating soil respiration of forests at the catchment scale, which may provide a favorable tool for carbon cycle simulation at the catchment scale. (2) The annual highest soil respiration of forests was 700 gC m^-2 a^-1 and the annual mean soil respiration of forests was 571 gC m^-2 a^-1 in the Zhangjiagou catchment. The spatial pattern of the annual soil respiration was similar to the spatial pattern of soil respiration for the growing season, which showed that high values of soil respiration were distributed in the north, southwest and southeast of the catchment, and the low values were largely in nearby valleys. The spatial pattern of soil respiration was closely related to topography, vegetation and other factors. (3) During the growing season, the five forest types showed a similar seasonal pattern in simulated soil respiration, characterized by a mono-peak curve, with a summer maximum and an early and late growing season minimum. The peak soil respiration value of the five forest types occurred in July, and was 146.3 gC/m² in the Mixed deciduous forest, >121 gC/m² in the Oak forest, >118.3 gC/m² in the Aspen-birch forest, >114.9 gC/m² in the Hardwood forest and >85.5 gC/m² in the Larch plantation. (4) In the growing season, the simulated soil respiration of the five forest types showed a significant index correlation (P<0.001) with the simulated soil temperature at 5 cm depth, and could explain approximately 70% of the seasonal variation in soil respiration. The correlation coefficients varied from 0.6873 in the Hardwood forest to 0.7205 in the Larch plantation. There was no obvious correlation between the simulated soil respiration and the simulated soil moisture for all forest types at 5 cm depth. This showed that the dominant factor of seasonal variation in soil respiration was soil temperature and the effect of soil moisture on soil respiration was not notable.