Effects of short distance sampling on the prediction accuracy of the spatial variability of soil respiration

Sampling design is important for the prediction accuracy of the spatial variability of soil respiration. In this study, a plot of 1 km×1 km was selected in a summer maize field from the northern part of the Huang-Huai-Hai Plain. Each of the forty-nine sampling sites were set on the basis of three different sampling designs, including a regular grid of 7×7 unit rule (with a spacing of 167 m), completely random (with an average spacing of 433 m), and a regular grid of 3×3 unit rule combined with completely random (with an average spacing of 405 m). To optimize the layout, based on the 3 designs, we maintained the total number of samples (49) and replaced the original sampling with short-distance sampling points for 2% to 14% of the total number of samples (with a spacing of 4 m). The spatial interpolation was finished with the ordinary Kriging interpolation method. The root mean square error (RMSE) and determination coefficient (R²) were chosen as indicators to investigate the effects of short distance sampling on the prediction accuracy of the spatial variability of soil respiration. The results showed that the spatial distribution of soil respiration under the three sampling designs was high in the west and low in the east, with moderate variation. Different sampling designs had significant impacts on the prediction accuracy of the spatial variability of soil respiration. The short distance sampling under the three sampling designs increased the prediction accuracy of the spatial variability of soil respiration by 7%-13%. Without short distance samples, the sampling design of the regular grid combined with completely random had the highest prediction accuracy, which was 10% and 22% higher than the regular grid and completely random sampling designs, respectively. Upon the replacement with short distance sampling, the prediction accuracy of the optimal sampling design (regular grid combined with completely random) was increased by 4%-7%. The prediction accuracy of the spatial variability of soil respiration was most obviously improved when the proportion of short distance samples was 10% of the whole size. This study found that setting short distance samples based on the same sample size could increase the sample density within a region and improve the prediction accuracy of soil respiration spatial variation and the reliability of experimental results. Therefore, a completely random sampling design combined with a regular grid and 10% short distance samples is a better choice for the soil respiration spatial variation estimation of a 1 km×1 km plot in a summer maize field from the northern part of the Huang-Huai-Hai Plain. The results of this study provide guidance for relevant research and field sampling designs.