样本量对叶面积指数遥感经验建模精度影响研究

Effects of Sample Size on LAI Accuracy by Remote Sensing Empirical Modeling

遥感经验模型反演方法是叶面积指数(LAI,Leaf Area Index)反演的重要途径,而实测数据又是经验统计模型的基本数据来源之一,但对于地面采样样本量对LAI遥感经验建模的影响研究较少。文章对单一地表覆盖类型的研究区采用不同样本量多次随机采样获取的采样数据来构建经验模型反演LAI,探究样本量对于建模精度的影响。研究结果表明:1遥感模型精度评估指数(RMAI,Remote Sensing Model Accuracy Index)随样本量呈幂函数形式逐渐减小;2在样本数量小于30时RMAI较为敏感,建模精度较差,当样本量达到45左右时,其精度达到稳定状态;3样本量越大的采样方案构建的遥感经验模型越稳定;4综合RMAI平均值及标准差的变化趋势,当样本量达到40时采样数据即可构建精度较高且稳定的LAI经验统计模型。

Empirical-model-based inversion is an important approach to the retrieval of leaf area index （LAI）. The ground sampling data is the primary data resource of Empirical modeling, and the sample size affects directly the precision of the empirical model. However, few studies were investigated about the effects of sample size on remote sensing empirical modeling accuracy of LAI. This article, based on the ground sample data sampled repeatedly with different sample sizes, built an empirical model of LAI to explore the effects of sam- ple size on modeling accuracy. The results showed that ： ①remote sensing model accuracy index （ RMAI ） decreases with increase in sample size in the power function form ; ② when the number of samples is less than 30, RMAI is more sensitive and the modeling accu- racy is lower,while the sample size approaches 45, the modeling accuracy reaches a steady state; ③ the larger the sample size, the more stable the modeling based on sample data ; ④ giving consideration to the change trends of the mean and standard deviation of RMAI,it comes to the conclusion that the sample data,whose sample size reaches 40 ,can build a stable empirical model with high precision.