Optimization of spatial sampling schemes and elements for estimating farmland area

Information on crop sown acreage is an important basis for a big agricultural country like China to formulation of national food policies and economic planning. The accuracy statistics of the crop planting area helps the macro decision-making departments a lot in tracing the condition of crop production and making proper management measures. Currently, sampling combine with remote sensing techniques is often used in conducting statistics and monitoring of the crop planting areas at regional scale. Scholars both at home and abroad have achieved very great success in the process of sampling research and study on the crop planting areas. However, some defects still exist in the process and indicate in two aspects: one is that the optimization design for elements (mainly size of samples) of space sampling have not been realized; the other is that the selection of sampling methods tend to be simple and lack of the comparison of various sampling techniques, and the optimization of spatial sampling schemes have not been realized. [Objective] experiments were conducted to optimize spatial sampling scheme and elements for estimating crop area. [Method] in the experiments, total farmland area in Beijing City was regarded as sampling population and Remote Sensing (RS), Geographic Information Systems (GIS) and classic sampling methods were applied as the study method in order to make an optimization on spatial sampling scheme and elements for crop planting areas estimation. This study adopted the Moran command in ARCGIS to calculate the global spatial autocorrelation (Moran I) and drew a scatter plot for different pixel scale and corresponding Moran I and analyzed the relationship between them. Eventually, 8 sampling quadrate size were selected. Then, 4 sampling methods (simple random sampling, systematic sampling and stratified sampling) were used to estimate farmland area. [Result] the experimental results demonstrated that, when the relative error and the cost of sampling (- hown through the sample capacity) are served as the evaluation indicators of sampling efficiency, the sampling efficiency of stratified sampling (the stratification symbol was the ration of farmland accounting for a gird area) was the highest (the average sampling error was 0.8%; sample capacity varied from 17 to 43), and then that of systematic sampling was the higher, the efficiency of simple random sampling was the lowest among 4 sampling methods. [Conclusion] the sampling error from the sampling frame which the sampling survey unit was a sampling quadrate with the size of 5000m×5000m was the least among 8 sampling quadrate size. This research can provide a theoretical basis for improving the system of spatial sampling survey for crop sown acreage estimation.