Abstract: For the problems of reasonable sampling point number and spatial prediction optimization in fine soil management system, this paper explored the possibility of finite sample points with different sampling methods based on soil-terrain relationship. Specially, the effects of spatial distribution of optimal sampling points on the prediction of soil organic matter (SOM) were studied based on the sampling methods of grid system, terrain unit partition expression and optimal statistical unit for relief degree, combined with the geo-statistics, soil-topography relationships, and geographically weighted regression kriging model. The results showed that: (1) The optimal unit size for topographic relief was 10 × 10 pixels in the plain, hilly and platform areas, and the spatial distributions of sampling points were mostly dense in the statistical units of plain area. The optimal sampling point numbers was 1656 based on the evaluation of standard samples. (2) The main influencing factors on the spatial variation of SOM were elevation, slope direction, topographic position, relative position, topographic relief, and the other topographic factors, which could explain 69.2% of the spatial variation of SOM in the study area. (3) The accuracy of GWRK model was greater than that of OK interpolation model based on different sampling points or terrain units. The optimal number of sampling points were 39, 481, 9, and 28 for the positions of mountain ridge, backslope (N/EN), steep slope (N/EN), and slope base, respectively. The improvement of GWRK model on SOM predictions was limited under the maximum sampling points (n = 2806). The local variation of the spatial distribution of the predicted SOM content was relatively decreased with the sampling point number decreasing Therefore, the effect of reasonable sampling points on predict SOM content was significantly different among different sampling methods, while the prediction trend was similar among them. The mapping of predicted SOM contents could demonstrate the basic spatial distribution pattern of SOM.