中文摘要：

结合Landsat TM影像和毛竹林样地调查数据,分别采用原始样本及基于太阳入射角余弦值cos（β）分层后样本建立多元线性回归模型估算安吉县毛竹林地上部分碳储量,并对利用这2种不同样本集所建模型的精度作比较。结果表明：通过cos（β）对样本分层能够降低样本点离散程度,提高遥感因子与毛竹林地上部分碳储量相关性;采用cos（β）分层后样本集所建的模型拟合（RMSE=4.07 MgC·hm-2）和预测精度（RMSE=4.63 MgC·hm-2）都高于基于原始样本集所建的模型;样本分层能够显著改善高和低水平地上部分碳储量精度。本研究提出的样本分层方法还需应用于不同影像及森林类型作进一步验证。

英文摘要：

Based on a combined use of landsat thematic mapper （TM） and field measurements, original plots and stratified plots according to cosine of incidence angle, which can represent correlation of terrain and sun light, were respectively used for variables of muhiple regression models to estimate the aboveground carbon storage of moso bamboo （ Phyllostachys edulis） forest. Results showed that ： dispersion degree of sample plots decreased and correlation coefficients between remote sensing variables and aboveground carbon storage increased through stratified sample plots; The fitting （RMSE =4.07 MgC. hm-5） and predicted （ RMSE = 4.63 MgC · hm-2 ） accuracy of model with stratified sample plots were high than that of model with original samples; the improvement was of statistical significance for those high and low aboveground carbon storage value based on stratified plots strategy. The stratified sample plots method provided in this study is still needed to be tested with different images and forest types in future studies.