A novel sampling design considering the local heterogeneity of soil for farm field-level mapping with multiple soil properties

Soil sampling is critical to obtaining reliable input for farm field-level digital soil mapping (DSM). Sample size and location are the key issues for soil sampling. However, sample size is often restricted by available budgets. In this case, recognizing the key sample locations is necessary. Existing methods have optimized the sample locations in a global manner without considering the impacts of local heterogeneity of soil. In this paper, a novel sampling approach based on the local heterogeneity of soil with a limited sample size (40 samples in this research) was developed. First, the local heterogeneity of soil was inferred. Second, the sub-regions were divided based on the level of local soil heterogeneity and the corresponding sample numbers were determined. Finally, the key sample locations were determined based on the fuzzy memberships. To validate the proposed method, it was compared with stratified random sampling, k-means sampling and conditional Latin hypercube sampling. The ordinary kriging method was applied to map five soil properties, including soil organic matter, pH, total nitrogen, available phosphorus and available potassium. The comparative experiments showed that the proposed method has better robustness in satisfying good mapping accuracy for multi-soil properties at the farm field level compared with the competing sampling methods, as indicated by the relatively lower and more stable mean bias error (MBE) and root mean square error (RMSE) values. It can be concluded that the consideration of local heterogeneity of soil is helpful to recognize the key sample locations for limited sample sizes.