遥感辅助的农作物种植面积小域估计方法研究

遥感影像是大数据的一种，利用遥感对农作物播种面积进行估算常采用回归估计量或校准估计量，通常都需要将地面样本数据与遥感分类信息相结合。但对于大多数回归估计量，对省级总体的农作物面积估算只能满足对省级总体的精度要求而不能分解到更小区域，比如县和乡级。本文利用黑龙江省2011年的地面实测样本数据结合遥感分类结果，构建了单元层次的多响应变量的多元回归形式的小域模型，并将小域效应设定为固定形式。这样基于回归估计方法，既可以估算分县的主要作物播种面积，也可以使得各县播种面积估计结果相加就等于回归模型含义下的省级总体的总量估计。对黑龙江省玉米、水稻、大豆分县小域估计结果的精度评价（变异系数C.V），平均而言均可以满足县级精度要求。本文的结果表明小域估计方法在解决省级总体对全省和分县的农作物种植面积多级估算问题中具有很好的应用。

A Study on Small Area Estimation for Crop Acreage in Remote Sensing Assisted Crop Survey

Remote sensing imagery is one form of the big data. It typically uses regression estimator or calibration estimator to estimate the planted acreage by applying remote sensing. Therefore, it’s necessary to combine the sample data from ground survey with satellite image classification. For most cases of regression estimator, the estimation of provincial crop acreage only satisfies the precision target for a province but could not disaggregate to small areas, such as county and town level statistics. This paper is adopted small area estimation approach to estimate crop acreage at county level in Heilongjiang province by combining ground survey data with image classification of the year 2011. A basic level small area model in the form of multi-response multiple regression with fixed effects is adopted to estimate acreage for major crops. Based on regression estimator, not only all county level estimation is produced but also the constraint that the aggregate of county level estimates could be equal to the estimate for entire province under linear model is satisfied.