中文摘要：

该文采用主成分分析和聚类分析方法,运用扫描电镜对新疆胡颓子属（Elaeagnus Linn.）落叶组大果沙（E.moorcroftii）、尖果沙枣（E.oxycarpa）和沙枣（E.angustifolia）共计18个样品的花粉形态进行了观察和研究;全面描述了新疆胡颓子属3个种花粉的极轴长、赤道轴长、极面观、赤道面观、孔沟形态和外壁纹饰等形态特点,并对其进行数据分析,通过对聚类结果的比较,讨论不同孢粉学特征对分类结果的影响。结果表明：（1）3个种的花粉均为中等大小;萌发孔类型均为三孔沟型。（2）花粉极面观形状从三角形、钝三角形到圆三角形;赤道面观形状从菱角形、半圆形到扁圆形;外壁纹饰从皱波状到穴状。（3）以筛选出的5个反映花粉形态的主成分指标进行聚类,结果显示所选指标不能区分种;以代表花粉形态的极面观形态和赤道面观形态为指标进行聚类同样也无法区分3个种;以花粉粒大小为指标进行聚类发现,大果沙枣的花粉大小可以区别于其他2个种。研究表明除了大果沙枣花粉的大小,尖果沙枣和沙枣的花粉形态不宜直接应用于种的划分。该研究结果为新疆胡颓子属落叶组植物花粉的种间鉴定和品种划分以及进化关系的研究提供了孢粉学依据。

英文摘要：

The pollen morphology of Elaeagnus Linn. in Xinjiang namely E. moorcroftii, E. oxycarpa and E. angustifolia was observed by scanning electron microscopy,18 samples were analyzed using the method of principal component analy-sis and cluster analysis. We conducted an investigation and collection of 3 species of Elaeagnus,collection locations in-cluding Urumqi ,Hami, Turpan, Akesu, Yili and Kashi area. Specimen collection was divided into two periods,we col-lected flowering specimens in May and gathered fruit in the following year,and then identification. The identification of 3 species based on the description in China and Desert Flora, Flora of China, РАСТИТЕЛЬНОЕ СЫРЬЕСССР, ФЛОРАСССР and Флора Таджикской ССР. We selected 6 samples each kind from the identification results, a total of 18 kinds. The pollen was spread evenly on the sample stage sticked double-sided adhesive tape, placed in the Hitachi HUS-5GB, metal spraying and ion sputtering coating in vacuum about 10 minutes, and removed the sample and observed them by scanning electron microscope. Firstly, we observed whether there existed differences in morphological characters,then expanded the multiples to measure polar axis and equatorial axis. We described the three kinds of Elaeagnus Linn. pollen of the length of polar axis,the length of equatorial axis, the polar view,the equatorial view, the hole groove shape and the exine ornamentation,etc. morphology characteristics comprehensively, and carried on the data analysis.We selected 8 indicators with the specified character encoding,namely 4 quantitative indexes,4 qualitative indexes,each kind was taken the minimum, maximum and average,represented variation amplitude. In order to obtain the reliable result of cluster a-nalysis, we classified significant differences of the measurement indexes, and then through principal component analysis, calculated the principal component scores respectively, according to the selection index,Q type clustering,obtained re-spect