Comparison of sampling schemes for the spatial prediction of soil organic matter in a typical black soil region in China

Data on the spatial distribution of soil organic matter (SOM) are important for the spatio-temporal modeling of soil organic carbon dynamics and soil carbon sequestration potential estimates. A total of 175 topsoil samples (0-20 cm) were collected from a typical black soil area in central Hailun County in northeastern China. Seven sampling design schemes, ordinary kriging (OK), and regression kriging (RK) were applied to the re-sampled SOM data for predicting the spatial distribution of SOM. The results showed that single sampling designs, such as simple random, stratified random (STR), and conditional Latin hypercube (CLH), produced poor estimates of SOM, while hybrid sampling designs, such as uniform distribution of point pairs for variogram estimation combined with spatial coverage, STR combined with spatial coverage (STRC), and CLH combined with spatial coverage (CLHC), had a higher predicting accuracy when the sample size was relatively small (B262). For square grid sampling, a higher predicting accuracy could be achieved only when the sample size was sufficiently large (i.e., C402). The inclusion of prior knowledge or SOM-related secondary data in the sampling design and the trade-off between the even and uneven distribution of sampling points are especially important for designing a small-size sampling scheme. Moreover, although the SOM-predicting accuracy of RK was not as good as OK in this study, increasing the sample size may improve the predicting accuracy of SOM. Therefore, the optimal sampling design and spatial predicting method are both important for the predictive mapping of SOM spatial distribution in this area.