Abstract:Four sampling point layouts at different scales based on different grid sizes:scale a, 7 050 sampling points in 0.5 km×0.5 km; scale b, 1 757 sampling points in 1 km×1 km; scale c, 444 sampling points in 2 km×2 km; scale d, 110 sampling points in 4 km×4 km were set up. Then, the optimized selection of the original soil sampling points at the four sampling scales were conducted using the simulated annealing method, and the optimum sampling numbers of scales a, b, c and d were 956, 751, 283 and 95, respectively, and were uniformly distributed in the space. Relative to the decrease in the sampling scale, the number of sampling points increased multiplicatively, but the predicted accuracy of soil properties did not increase proportionately. With the increased sampling point number, the predicted accuracy of soil properties gradually decreased. Considering the number of samples and the predicted accuracy of soil properties, a sampling scale of 2 km×2 km is optimal for soil organic matter.