地理空间分层抽样对大尺度草地估产精度的改善--以低地草甸为例

Accuracy Improvement in Yield Estimation of Large-scale Grassland by Stratified Sampling——A Case Study of Lowland Meadow

为优化天然草地监测策略,提高样本点的地域空间代表性和分层估产模型的稳定性,改善全国天然草地地上生物量的估测精度,以低地草甸草原为研究对象,结合地面采样数据和MODIS-NDVI数据,根据统计学的分层抽样方法和全国草地资源区划原理为理论依据,通过分层效率检验、总体平均值误差对比验证和模型分析等方法,以低地草甸为例进行分析。结果表明:各分层区域的分层效率都大于1;采用分层抽样方法相对于简单随机抽样方法地面采样样本的总体平均值精度提高9.672%,模型的估产精度提高13.827%。从而,验证了某一类型草地的区域分层抽样方法比简单随机抽样方法更可取。

There are various types of grassland in China and the geographical spatial span of a given grassland type is very large.Due to different climate and soil condition in different areas,the situation of vegetative development is different from each other,thus these elements affect the precision of yield estimation by conventional methods.In order to optimize the monitoring strategy of natural grasslands,the aboveground biomass yield of lowland meadow was evaluated according to the principles of stratified sampling method and grassland zoning in China using stratification efficiency test,overall average error comparing analysis,and model analysis based on the ground measured data and MODIS-NDVI data.The results show that： the stratification efficiency of each strata area was more than 1;in comparison with random sampling method,the stratified sampling method improved the overall average accuracy by 9.672% and the regression mode accuracy by 13.827%.These results verify that stratified sampling method could be more recommendable than random sampling method for the biomass yield estimation of a given grassland type.