Assessing the crop acreage at county level on the North China Plain using an adapted regession estimator method

Image classifications including sub pixel analysis are often used to estimate directly the crop acreage, while ground data collected during field surveys play a secondary role. This type of crop area assessment using image classifications often leads to a biased estimation due to non-representative selection of training data and subjective a-priori knowledge. Instead regression estimator approach combining remote sensing information with a rigorous ground sampling can result in an accurate assessment of crop acreage. In this study to produce the crop statistics, the area frame sampling approach is adapted to the strip-like cropping pattern on the North China Plain. Remote sensing information is used to perform a cost-efficient stratification from which no-agricultural areas are excluded from ground survey. This information is also included in a later stage as an auxiliary estimator in regression analysis. The results showed that the integration of remote sensing information as an auxiliary estimator can improve the confidence of estimation by reducing the variance of the estimates.