样本容量和物种特征对BIOCLIM模型模拟物种分布准确度的影响--以12个中国特有落叶栎树种为例

EFFECTS OF SAMPLE SIZE AND SPECIES TRAITS ON PERFORMANCE OF BIOCLIM IN PREDICTING GEOGRAPHICAL DISTRIBUTION OF TREE SPECIES——A CASE STUDY WITH 12 DECIDUOUS QUERCUS SPECIES INDIGENOUS TO CHINA

物种分布模型被广泛应用于生态学、生物地理学及保护生物学等领域的研究。由于难于取样或标本记录不完善等原因,真正能够用于模型预测的物种分布数据非常有限。因此,有必要搞清楚样本容量和物种特征对模型模拟准确度的影响,为确定以物种特征为区分条件的最小样本容量奠定基础。为了探讨应用BIOCLIM模型预测中国特有植物种的效果,以12个落叶栎树种为例,从不同的样本容量和生态特征两方面研究其对BIOCLIM模型模拟准确度的影响。结果表明:BIOCLIM模型模拟准确度随样本容量的增加在初期几乎呈直线增加趋势至样本容量达到25,随后渐变平缓至样本容量为75～100时达到最大值。此外,生态幅窄和环境特化物种比生态幅宽和对环境耐受性强的物种更容易获得较高的准确度。结果说明,BIOCLIM可有效地用于样本数量较小的狭域型物种分布预测。

Aims Our objective was to evaluate the performance of BIOCLIM in predicting geographical distributions of tree species in China and to determine the impacts of sample size and species traits on predictive accuracy. Methods BIOCLIM model was used to predict the geographical distribution of 12 deciduous Quercus species differing in ecological traits and sample sizes. We used data from museum or herbarium collections and 18 layers of grid-data of meteorological variables at 6 ×6 resolution to run the model. We used data from museum or herbarium collections and 18 layers of grid-data of meteorological variables at 6′×6′ resolution to run the model. We evaluated the quality of model predictions with area under the receiver operating characteristic curve （AUC） and Kappa statistic. Important findings Model performance was improved and the variability in predictive accuracy was reduced with increasing sample size. Reasonable predictions to county-level resolution can be achieved with a sample size of 25 data points. On average, 75-100 observations were found to be sufficient to obtain maximal accuracy. Furthermore, we found that accuracy of BIOCLIM was greater for species with narrower geographical range and limited environmental tolerance than those with broader range and greater tolerance.