Traditional soil mapping methods, relating soil survey profiles with soil type maps under some criterions, based on soil surveyors’ experience and manual delineation process are time-consuming and cost-prohibitive. The digital soil mapping (DSM), which takes advantage of the advancement in spatial information processing techniques and the increasing availability of high resolution spatial data, is feasible in providing soil distribution information at the required level of accuracy and spatial details. Field samples are the most important data sources for DSM. There are three DSM methods based on samples which are the ones based on spatial autocorrelation, the ones based on relationship between soil properties and environment co-varieties as well as the hybrid ones based on both spatial auto-correlation and the relationship between soil properties and environment co-variates. In order to capture the global representativeness of samples for characterizing the spatial variation of soil property, these kinds of methods all have serious requirement for the size, distribution and typicality of field samples. So, sampling design is an important process for DSM. Schemes for sampling design can be grouped into two categories based on whether the existing samples can be integrated in the sampling process: simple sampling scheme and integrative sampling scheme. This paper provides an overall review over these aspects of digital soil mapping.