农作物面积抽样单元空间变异性及其对抽样外推效率的影响

SPATIAL VARIABILITY OF SAMPLING UNIT OF CROP AREA AND ITS EFFECT ON SAMPLING EXTRAPOLATION EFFICIENCY

关键词:  农作物面积抽样单元尺度空间变异性空间抽样抽样效率

Key words:  crop area  sampling unit scale  spatial variability  spatial sampling  sampling efficiency

摘要:

［目的］联合遥感和传统抽样方法的空间抽样技术是进行大区域农作物面积监测的有效手段。但传统抽样要求抽样单元间相互独立，并未考虑到区域农作物空间变异性的存在对农作物面积空间抽样效率是否有影响及其影响程度如何，从而限制了农作物面积监测的效率和精度。［方法］文章以吉林省德惠市为研究区，利用变异函数分别对10种抽样单元尺度内玉米和水稻种植面积的空间变异性进行定量评价。选择3种空间抽样方案（简单随机抽样、系统抽样、分层抽样），分别计算不同空间变异程度下的样本容量（n）、抽样相对误差（Re）和总体总值估计量的变异系数（CV（Y︿））作为空间抽样效率的评价指标，定量分析空间变异性对农作物面积空间抽样效率的影响，提出适宜空间变异农作物的空间抽样优化方案。［结果］(1)玉米和水稻面积的空间变异性（反映为基台值C0+C）均随着抽样单元尺度的增加逐渐而减小，其中500m×500m和2 500m×2 500m抽样单元尺度下农作物面积的空间自相关程度相对最小。(2)相同抽样比下，农作物面积在3种抽样方法下的相对误差和变异系数均随着空间变异性的减小（抽样单元尺度增加）逐渐增大，其中玉米和水稻在分层抽样条件下的相对误差和变异系数具有相对最小值，分别限制在10%和20%以内，其次为系统抽样（Re<40%，CV（Y︿）<88%）和简单随机抽样（Re<50%，CV（Y︿）<75%）。 在抽样方法被选定的情况下，两种农作物的分层抽样相对误差变化率随着抽样比的增加迅速降低，当抽样比增至5%时，其相对误差和变异系数的变化率基本趋于稳定，继续增加抽样比来提高抽样精度的效果甚微。 在3 500m×3 500m抽样单元尺度内，按5%抽样比确定的样本容量对玉米和水稻面积进行分层抽样时，可达到95%的抽样精度（Re<5%）。 综合农作物在不同抽样单元尺度下的空间变异特征和空间抽样效率，基于2 500m×2 500m的抽样单元尺度和5%抽样比进行分层抽样的空间变异性农作物空间抽样方案是一种高精度、低成本、代表性强的空间抽样方案。［结论］该研究为区域农作物面积抽样调查方法的改善和抽样效率的提高提供技术支持和参考依据。

Abstract:

The spatial sampling technique combined with remote sensing and traditional sampling method is an effective method for monitoring crop area in large areas. However, traditional sampling requires that sampling units should be independent of each other, and does not consider whether the spatial variability of regional crops has an impact on the sampling efficiency of crop area and its degree of influence, thus limits the efficiency and accuracy of crop area monitoring. In view of the above problems, this study took Dehui city, Jilin province as the research area, the spatial variogram was used to quantitatively evaluate the spatial variability of crop area in 10 sampling units. To quantitatively evaluate the effect of spatial variation on sampling efficiency and to optimize sampling scheme for spatial variability crops, three sampling methods, i.e., simple random (SR), systematic sampling (SY), and stratified sampling (ST), were selected to calculate sample size (n), overall relative error of extrapolation (Re), and coefficient of variation (CV(Y︿） )) as indices of sampling efficiency under different spatial variation. The results indicated that the spatial variability of the area of maize and rice (reflected as the base value C0+C) decreased with the increase of the sampling unit scale, among them, the spatial autocorrelation of crop area was relatively small at the sampling scale of 500m×500m and 2 500m×2 500m. Under the same sampling ratio, the relative error and coefficient of variation of crop area under the three sampling methods all increased with the decrease of spatial variability (the sampling unit scale increases). The stratified sampling relative error and stratification coefficient of variation of maize and rice had relative minimum, which were limited to 10% and 20%, respectively, followed by systematic sampling (Re<40%, CV(Y︿）)<88% ) and simple random sampling (Re<50%, CV(Y︿）)<75%). When the sampling method was selected, the relative error rate of stratified sampling of the two crops decreased rapidly with the increase of the sampling ratio, when the sampling ratio increased to 5%, the relative error and the coefficient of variation of the coefficient of variation tended to be stable. Within the sampling unit scale of 3 500m ×3 500m, the sampling precision (Re<5%) could be achieved when stratified sampling was conducted on the area of maize and rice according to the sample size determined by 5% sampling ratio. Considering the spatial variation and spatial sampling efficiency of crops at different sampling unit scales, the stratified sampling method with 2 500m×2 500m scale and 5% sampling ratio was a high precision, low cost and representative program for crop area with spatial variation. This study provides technical support and reference for the improvement of sampling method and sampling efficiency of regional crop area.