白马鸡冬季群体栖息地选择的经验模型

于2003年1-4和10-12月考察了四川省稻城县著杰寺、格瓦村、冉子村等地白马鸡(Crossoptilon crossoptilon)群体的栖息地选择.采用样线法进行系统取样或随机取样调查,通过Mann-Whitney U检验判定遇见样方(有白马鸡出现)和未遇样方(无白马鸡出现)间呈显著差异的环境变量,对这些变量进行数据转换(同一类型的变量相互乘积)并以它们为自变量,以白马鸡的出现赋值为1或不出现赋值为0为因变量进行逐步线性回归,结果中保留相关显著的变量(t检验,P＜0.05),从而得到白马鸡栖息地选择的经验模型.结果表明:距水源距离、灌木盖度、灌木高度、乔木盖度、乔木高度、乔木胸径和草本盖度等变量在遇见样方与未遇样方间呈显著差异(P＜0.05),数据转换后引入的3个新变量亦呈显著差异(P＜0.05);回归等式的决定系数较高且回归相关显著(r2=0.832,F=17.125,df=4,P=0.000),各自变量也相关显著(t检验,P＜0.05),而灌木盖度、灌木高度、乔木盖度、乔木高度、乔木胸径和草本盖度等变量相关不显著(t检验,P＞0.05).在3个乡的考察中对该经验模型进行了验证,采用同样方法进行数据变换,以与经验模型相同的自变量和因变量进行逐步线性回归,最后根据得到的各变量系数利用经验公式对白马鸡的栖息地选择进行了预测.以0.5为切断点将预测结果转化为0/1二态数据,利用成对样本t检验来验证经验模型的预测准确性.结果表明,在3个不同区域利用经验模型进行回归,回归等式均具有较高的决定系数,并回归显著,各自变量均显著相关(t检验,P＜0.05),成对样本t检验均未呈现显著差异(P＞0.05),具有较好的预测准确性.

EXPERIENTIAL MODEL FOR THE WINTERING HABITAT SELECTION OF WHITE EARED-PHEASANT FLOCKS

The field surveys on the wintering habitat selection of White Eared-pheasant (Crossoptilon crossoptilon) flocks were carried out from Jan. to Apr. and from Oct. to Dec.2003 around Zhujie Monastery, Gewa village and Ranzi village in Sichuan province, China. Line transects of systemic sampling and random transects were used in the surveys. In the first survey in Zhujie Monastery, Mann-Whitney U test is applied to decide the variables with significant differences between active plots(presence of White Eared-pheasant)and inactive ones (absence of the species) (P<0.05). Some new variables are introduced by multiplying the variables presenting the vegetation types. Stepwise line regression is conducted with presence/absence(0/1)of the species as dependent variable and the above variables as independent ones, respectively. In the regression model, the independent variables with significant correlation are remained (t test, P<0.05). Results show that distance to nearest water, shrub cover, shrub height, wood cover, wood height, wood diameter at breast height and herb cover are significantly different between active and inactive plots(Mann-Whitney U test, P<0 05). Three new variables are introduced after the transformations. They are also significantly different between active and inactive plots(Mann-Whitney U test, P<0 05). The experiential model can be formally expressed. The equation has relatively high value of coefficient of determination and shows significant correlation(r 2=0.832, F=17.125, d f=4, P=0.000), and the independent variables also show significant correlation(t test, P<0 05). Shrub cover, shrub height, wood cover, wood height, wood diameter at breast and herb cover are insignificant and deleted(t test, P>0 05). The above model is tested in the second survey in three other study areas. The same data transformations, independent and dependent variables as those in experiential model are used to conduct the regression models for the 3 study areas, respectively. Finally, the obtained coefficients of independent variables are applied to predict the occurrence of the bird. The results show that 3 regression models all had relatively high values of coefficient of determination and are all significant correlation. The predictive values(0.5 as the cut-off point) are not significantly different from the observed ones(paired-samples t test, P> 0 05). The experiential model could well predict the habitat selection of White Eared-pheasant.