Provincial spatial sampling method for crop sown acreage estimation

Timely and accurate estimating crop sown acreage is one of key technologies of crop yield monitoring by remote sensing, it has become an important subject in national agricultural condition monitoring field. In view of the problems occurred in the current sampling survey technology system, such as the formulation of sample size is lack of scientific, the samples are optionally distributed in space, and the accuracy of population extrapolation are still poor. In this study, Remote Sensing, Geographic information systems technology and traditional sampling methods are used to conduct a study on provincial spatial sampling methods for crop sown acreage estimation, in order to improve the existing sampling survey efficiency. Shandong Province, China is chosen as the study area and winter wheat sown area as the study object. The basic geographic information data (1:250000, province and county boundaries), land use data (1:250000) in 2010, the spatial distribution data of winter wheat (derived by MODIS image) in 2010 and the winter wheat planting regionalization data of the study area are used in this paper. Firstly, two-stage sampling method is employed as the main sampling scheme, considering the convenience of sampling frame construction and samples field investigation; Secondly, all counties in Shandong Province are served as the sampling units of the first stage, 4 kinds of sampling methods (simple random sampling, stratified sampling with the winter wheat planting regionalization as stratification sign, stratified sampling with cultivated land types as stratification sign, stratified sampling with winter wheat sown area in each county as stratification sign) are selected to draw the samples of the first stage, and then sampling cost and errors are evaluated to optimize spatial sampling method of the first stage, based on the samples drawn by the 4 sampling methods; Thirdly, 500 m×500 m are selected as the sampling units size, and simple random sampling method is used to draw the samples of the second sampling stage. 8 levels of sampling size are formulated in the second sampling stage to optimize sample size; Finally, the experiment on population extrapolation and error estimation is conducted based on the samples drawn at the first and second stage. The experimental results demonstrate that the efficiency of the stratified sampling which the winter wheat area in each county was selected as stratification sign is the highest among 4 sampling methods, when the relative errors of population extrapolation are nearly equivalent; The relative errors decrease with the sample size of the second sampling stage increasing, but the coefficient of variation (CV) is still higher. Comprehensively evaluating the relative error and CV of population extrapolation at 8 kinds of sample size levels, 9 samples drawn from the every sampled county are considered as the optimal sample size at the second sampling stage. In this way, this research can provide a theoretical basis for improving the efficiency of spatial sampling survey for crop sown acreage estimation.