3.2. SWC semivariance characteristics of sampling plots
We used Variogram analysis and Kriging interception to characterize and generate spatial structures of soil moisture for each plot area (Table 3). RSS for most plot areas was < 1 (with an average value of 0.23 for plots #2 and #7), indicating a clear spatial auto-correlation relationship among data distribution in their effective ranges. Further, the models which we used for semivariance analysis were appropriate and the parameter combination for each fitting function model can be used for explanation of the spatial pattern in the study plots. RSS in plot #1 was 12.9 suggesting that the model used for plot #1 was not as fitting as theoretically needed, and the interception map should be carefully considered in the following analysis.
We used the Gaussian model for plot #3, spherical for plots #2, #5, #6 and #10, and exponential for plots #4, #7 and #9, indicating that the fitting model was robust and interpolation results were based on a function that can represent the actual spatial pattern of SWC. The optimal function was linear for plot #1 and #8 which indicated that the model was not as ideal as the RSS showed, and that may be a limitation due to the size of the dataset. A spatial correlation of < 0.5 between SWC in plot #1 and #8 indicated that there was no spatial correlation, closely with what was reflected by RSS (Yang et al., 2017). Other sampling plots showed a strong spatial auto-correlation, indicating that parameters based on the models can be used for spatial interpolation of SWC at each sampling plot.