中文摘要：

运用植物植硅体的形状系数，对不同禾本科植物在亚科和种的水平上进行分类，可为深入理解植硅体的环境指示意义，提高古植被重建精度提供参考。本文选取长白山常见的10种禾本科植物，对其叶片进行植硅体分析，共统计并鉴定植硅体3037粒，对其中哑铃型、棒型、尖型、齿型、鞍型和硅化的气孔等植硅体的7个形状系数进行非参数检验，并在亚科水平和种的水平上分别进行判别分析。结果表明，3个亚科之间的形状系数sF。相差较大，故据sFd的平均值可在亚科水平上对植物进行区分。上述植硅体的形状系数在亚科水平上判别分析的正确率可达88．51％，且检验结果表明文中根据典型亚科植硅体形状系数所建立的判别方程是有效的；利用哑铃型植硅体的7个形状系数则可以将黍亚科的植物区分到种，其判别分析结果显示总的正确率为73．12％；根据棒型、尖型和齿型植硅体的7个形状系数也可以将早熟禾亚科的植物区分到种，其判别分析结果显示正确率分别为60．87％，59．21％和87．07％。因此，通过典型植硅体的形状系数对植物进行分类，不但可以提高禾本科植物在亚科水平上分类的精确度，还可以根据植硅体的形状系数对禾本科进行进一步的分类。

英文摘要：

Phytolith as a climate proxy is one of the most important and effective tool for reconstruction of paleoenvironment. Thus it＇s very significant to research further how to distinguish between phytoliths for accurately reconstructing paleoenvironment. In this paper, 10 species plants （ Miscanthus sacchariflorus, Echinochloa phyllopogon, Echinochloa var. caudata, Arthraxon hispidus, Calamagrostis angustifolia, Calamagrostis langsdorffii, Beckmannia syzigachne, Millium effusum, Calamagrostis arundinacea, Phragmites communis） are selected and 3037 phytoliths are identified. 7 shape factors of Bilobate,Elongate,Lanceolate, Crenate, Saddle and Stomata are measured and analyzed by nonparametric testing and discriminant analysis. Results are following： The result of the nonparametric testing indicates that average value of the SF4 among the 3 subfamilies is obviously different. Discriminant functions for the 3 subfamilies, also reveal that the discriminant accuracy is high （88.51%）. Total discriminant accuracy for Panicoideae is 73.12%. Discriminant accuracies for Pooideae are respectively 60. 87%, 59.21%, and 87.07% based on shape factors of the Elongate, Lanceolate and Crenate phytoliths. At the same time, Leymus chinensis and Setaria viridis are used to test whether the discriminant functions are effective or not. Result shows discriminant functions classify precisely them, thus verify discriminant functions are useful. In a word,the shape factors not only may improve the accuracy of plant classification,but also provide a new profile for the plant classification.