Evaluation on extrapolation efficiency of multiple estimators for winter wheat planting acreage estimation

In view of the problem occurred in those studies on crop planting acreage estimation using estimators at home and abroad, such as just one single estimator (simple estimator or regression estimator) is used to extrapolate the population values and estimate sampling errors; the efficiencies of population extrapolation are not evaluated among multiple estimators quantitatively; furthermore, the estimator for crop planting area estimation is not optimized, it limited the improvement of the efficiency of sampling survey consequently. In this study, “3S” technology (Remote sensing, Geographic Information Systems and Global Positioning Systems) and traditional sampling methods are used to conduct a research on evaluating the extrapolation efficiencies of multiple estimators for crop planting acreage. Mengcheng County in Anhui Province, China is selected as the study area and winter wheat planting acreage as the study object. Firstly, 500m×500m is chosen as the sampling unit size and the physical boundaries of cultivated land plots as the sampling unit boundary to facilitate field investigation of the samples. Secondly, the study area is subdivide by the sampling unit with the size of 500m×500m, and the sampling frame is constructed based on the spatial distribution data of winter wheat in 2009 (derived by ALOS AVNIR-2, the spatial resolution is 10m) and the basic geographic information(1:250000, county boundary). Thirdly, the simple random sampling method is used to draw the samples, and the winter wheat planting acreage in the samples are measured by field investigation and remote sensing image. Fourthly, 3 kinds of estimators (they are simple estimator, ratio estimator and regression estimator, respectively) are constructed by combining the data of winter wheat planting acreage surveyed by remote sensing and the field investigation data of the sampled units, then 3 estimators are used to extrapolate population values and estimate the sampling errors, respectively; Finally, the relative error and coefficient of variation (CV) are selected as the index to quantitatively evaluate the extrapolation efficiencies of every estimator, and the experiment on optimizing the estimator for winter wheat acreage estimation is conducted. The experimental results demonstrate that, the ground survey accuracy of winter wheat planting acreage in the samples is higher than that of remote sensing investigation, but there are no significant differences between the two survey results; There is a significant direct proportional relationship between the ground survey results of winter wheat planting acreage in all samples and those of remote sensing investigation; The extrapolation efficiency of ratio estimator is the highest (relative error and CV are the minimum, they are 11.01% and 6.20%, respectively), the second is that of regression estimator, and the extrapolation efficiency of simple estimator is the lowest in 3 kinds of estimators. Therefore, the ratio estimator has certain effect on improving the estimation accuracy of winter wheat planting area by remote sensing. In this way, this research can provide a solution for improving the extrapolation efficiency of crop planting acreage at the regional scale.