Compositional data is very common in geosciences, which must meet four conditions in spatial interpolation, including ensuring positive definiteness and a constant sum of interpolated values at a given position, error minimization and lack of bias. This study took a case of fuzzy membership values of soil continuous classification, applied three methods of data transformation prior to kriging, i.e., logarithm transformation (LN), asymmetry Logratio transformation (ALR) and symmetry Logratio transformation (SLR). The performance of the transformed values by ordinary kriging was compared with the spatial prediction of the untransformed data using ordinary kringing (UTok), compositional kriging (CK). The results showed that the sum of interpolated values at a given position wasn’t equal to constant 1 by UTok and LN. Obviously, the above predictive result was theoretically unauthentic. Contrarily, membership values of all the spatial predicted sites summed to 1 when the membership values of the known soils were transformed by asymmetry Logratio and symmetry Logratio approaches and compositional kriging. Comparatively, symmetry Logratio transform could lead to a better spatial continuous distribution pattern. Interpolation results by compositional kriging could keep membership values either unbiased predictions or minimum prediction error variances.