Soil is a continuous spatial-temporal heterogeneity with high spatial variability. Soil nutrients are important parts of soil，and scientific and effective spatial prediction of the spatial variability of soil nutrients is the foundation of sustainable utilization of soil. The multi-dimensional fractal (multifractal) Krige method (Mkrige) with Krige method as a reference was used to simulate four types of soil nutrients, i.e., soil organic matter, nitrogen, phosphorus and potassium. Effectiveness of spatial prediction of Mkrige method was deeply analyzed in terms of five aspects, i.e., the traditional statistics, accumulation curve, multi-dimensional fractal images, multi-dimensional parameters and specific values. The results showed that traditional statistics parameters, accumulated value curve, multi-dimensional fractal parameters and multi-dimensional fractal image of the predicted values by Mkrige method were closer to the measured values than that of the predicted values by Krige method for the considered four soil nutrients. Mkrige method can better maintain the specific value area of the original samples, which truly reflected the chaotic state of the spatial distribution of soil nutrients and had a better prediction effect. The fractal degrees of phosphorus, organic matter, total nitrogen and potassium were reduced in order. For Mkrige method, the higher the fractal degree was, the more excellent the prediction effects were.