Optimization Study of Crop Area Spatial Sampling Method Based on Kriging Interpolation Estimation

Timely and accurate estimation of crop area are critical for enhancing agriculture management and ensuring national food security. Spatial sampling can take advantage of both remote sensing and field sampling, it has been widely used in large-scale crop area estimation. A large number of existing studies used a single traditional sampling method for sampling extrapolation without considering the optimization of sampling method. they are limited by the traditional sampling method and not capable to estimate the spatial distribution of crops effectively. For this reason, this paper selected Dehui County in Jilin Province as research area, and constructed the sampling frame using GF-1 PMS image at 8-m spatial resolution to extract the maize and rice area and distribution as the overall prior knowledge. Three spatial sampling methods (spatial simple random method, spatial system method and spatial stratification method) were selected for sample selection according to the same sampling ratio, and established variogram models of maize and rice based on the sample, respectively. Kriging method was used to estimate the crop area in the unsampled unit and the error between estimated and actual crop area in all sampling units (selected and unselected) was evaluated by cross validation method, to select the best sampling method suitable for estimating the spatial distribution of crop area. The experimental results demonstrate that the nugget coefficient $C_{0} /\left(C+C_{0}\right)$ of maize and rice variogram models established by three spatial sampling methods was less than 20%, indicating that the two kinds of crop have strong spatial variability, which is mainly structural variation (caused by natural factors such as climate and soil). Therefore, Kriging method can be used to estimate the spatial distribution of crops. Under the 3 sampling methods, the overall variation trend of kriging interpolation of maize and rice is roughly the same, but the interpolation effect of spatial system method is more consistent with the real spatial distribution trend of crops. The cross-validation results of all sample units show that the error terms ME (0.0059), MSE (0.0337) and RMSSE (0.9891) of the sample interpolation results sampled from the spatial system method are all the best, and the results from spatial random method are the worst. Considering the spatial distribution trend and accuracy of estimation, spatial system method is optimal for estimating the spatial distribution of crops. This study can provide an effective reference for improving the estimation accuracy of crop area.