As a core index of soil fertility, soil organic matter has played an important role in global carbon recycling and agricultural production. Field sampling, laboratory analysis and spatial interpolation are used as chief means of assessing content of soil organic matter in farmlands. Due to differences in soil forming factors and human activity, spatial distribution of soil organic matter contents and its driving force may vary from region to region under study, which in turn causes variation of applicable interpolations with the region. Fujian is a province very complicated in landform, as affected by the jointly actions of tectonic movement and long-term exogenic agencies, so the farmlands there are very high in land-use intensity and in spatial variability as well. Therefore, efficient site-specific point-plane expanding models will help expose laws of the spatial distribution of soil organic matter in farmlands of the province and realize precision agricultural nutrient management. In this study, nine counties, namely Minhou, Jian’ou, Yongding, Datian, Hanjiang, Fuding, Nan’an, Tong’an and Zhangpu, representative of 9 different types of landforms in Fujian Province, China, were selected for case study. Based on the data collected from the 29 320 sampling sites, scattered in these nine counties, of the soil testing-based fertilizer recommendation project sponsored by the Ministry of Agriculture of China, statistical characteristics for soil organic matter contents in farmlands were analyzed using the statistical software SPSS and GS+ and a semi-variance model established. In line with the principle of “for mean errors (ME), the closer to zero, the better and for root mean square errors (RSME), the smaller the better, and RSME should be given the first priority” impacts of different interpolation methods on accuracy of the reckoning of soil organic matter contents were evaluated and eventually a point-plane expanding model with the higher accuracy in predicting soil organic matter contents in the farmlands of Fujian Province was defined. Results show that the mean soil organic matter content of farmlands in Fujian Province was 27.83 g kg^-1, and its coefficient of variation was 37.87%, displaying a strong spatial autocorrelation and for semi-variance function the exponential model fitted the best. For prediction of organic matter contents in farmlands of Minhou County, Jian’ou City, Yongding County, Hanjiang District and Fuding City, the optimal geo-statistical interpolation method was the pan-kriging method, while the optimal deterministic interpolation method was the inverse distance weighting method, however, the former was higher than the latter in prediction precision. For Datian County, the optimal geo-statistical interpolation method was the simple kriging method, while the optimal deterministic interpolation method was the inverse high-order surface function method, however, the former was still higher than the latter. For Nan’an City, Tong’an District and Zhangpu County, the optimal geo-statistical interpolation method was the ordinary kriging method, while the optimal deterministic interpolation method was the tension spline function method, nevertheless, the former was higher, too, except for Zhangpu County. On the whole, in regions, like Minhou, Jian’ou, Yongding, Hanjiang and Fuding, that have similar distribution areas of plain or basin, hills and mountains, the use of the pan–kriging method in spatial interpolation of soil organic matter contents is relatively high in prediction precision, while in regions, like Datian, that are dominated with mountains the use of the simple kriging method is more advisable and in regions, like Nan’an, Tong’an and Zhangpu, that are dominated with plains or table lands, the use of the ordinary kriging method or the tension spline function method is.