Optimal design on samples layout of spatial sampling schemes for estimating winter wheat planting acreage

Sample layout is one of key factors in spatial sampling schemes for estimating crop planting acreage. It plays an important role that optimizing sample layout for improving the representativeness of samples versus population and the accuracy of population extrapolation, decreasing the cost of survey sampling. In this study, focusing on the problem that the samples layout design is not reasonable (e.g. samples units are not all independent of each other, when simple random sampling method is used to design samples layout; sampling intervals are not defined reasonably, when systematic sampling method is used to distribute samples in space), we tried to propose a optimal scheme of samples layout to improve the spatial sampling efficiency. Mengcheng County in Anhui Province, China was chosen as the study area, winter wheat planting acreage as the study object, and square girds as the shape of sampling units. Geostatistics, “3S” technology (Remote Sensing, Geographic Information Systems and Global Positioning systems) and traditional sampling methods are used in this paper. Firstly, 8 kinds of sampling unit sizes are formulated, and then the study area is subdivided by the sampling units with the 8 kinds of sizes to construct the sampling frame. The winter wheat acreages in all sampling units are calculated based on the spatial distribution data of winter wheat in 2009 and 2010(derived by ALOS AVNIR-2 and Landsat5 TM image, respectively); Secondly, in order to build the Variogram theoretical model of winter wheat acreage proportion within one sampling unit (WPS), simple random sampling method is used to draw the initial samples. Spatial correlation and variability of sampling units are analyzed, and spatial correlation threshold is quantitatively determined by the Variogram model; Thirdly, the equal interval pattern (sampling intervals are the same in vertical and horizontal directions, and spatial correlation threshold of samples is chosen as the sampli- g interval) is used to reasonably formulate the samples layout; Finally, the extrapolation accuracy, stability and sampling cost are estimated based on the samples after the layout are reasonably designed. In order to evaluate the design effect of samples layout, relative error, coefficient of variation (CV) and sampling size are selected as the indices, and simple random sampling method as the control treatment. The experimental results demonstrate that, the variability of WPS increases with sampling unit size increasing. CV of WPS varies from 32.75% to 43.46% under 8 sampling unit size levels; Spatial correlation thresholds of WPS increase with sampling unit size increasing; The relative error and CV of population extrapolation that samples layout is optimized are obviously less than those of simple random sampling method, when sampling unit size is small (500m×500m~2000m×2000m); Although the relative error and CV are not reduced after optimized design of sample layout, they occur an obvious decrease on sample size, when sampling unit size is larger (2500m×2500m~4000m×4000m). In this way, this research can provide a solution for improving the spatial sampling efficiency to estimate crop planting acreage.