Taking DengfengCounty as a case for study, effects of scale of the model on prediction of soil organic matter in complex landscape region were studied and at the same time the issue of how to comprehensively utilize spatial data of various scales explored. Soil organic matter of Dengfeng was predicted using the Regression Kriging method for analysis of roles and effects of spatial data of various scales on modeling. It was found that the soil organic matter was macroscopically related to elevation factor and large scale spatial data were the most optimal for use in fitting the trend value, while small scale spatial data were not proper for use in fitting. However, they revealed details of the spatial variability of small scale residual value and enhanced the spatial structure of large scale residual value, which effectively improved precision of the prediction. Therefore, in predicting soil organic matter in complex landscape, it is necessary to use large scale data in combination with small scale data to fit spatial variogram of residual value based on simulated trend value of large scale data and hence to predict residual value. Then by adding the trend value onto the residual value, soil organic matter predicted value is thus obtained.