摘要
通过概形分析模型(profile technique)——DOMAIN生成物种生境适宜分布图,选取低适宜性的地区作为物种不存在区,然后应用分类判别分析模型(group discrimination technique)——NeuralEnsembles预测我国毛竹潜在分布。结果表明:通过耦合DOMAIN和NeuralEnsembles模型可以改进NeuralEnsenbles模型预测精度;AUC和敏感度对用于建模的物种不存在数据取样数量不敏感,而最大Kappa值随着不存在数据取样数量的增大逐渐减小;未来气候变化将导致毛竹向北迁移33～266km,面积增加7.4%～13.9%。
In this paper a profile technique- DOMAIN was used to map potential habitat suitable for moso bamboo （Phyllostachys edulis）. and to select the areas with low suitable habitat as pseudo-absences. Then a group discrimination technique-NeuralEnsembles was employed to predict the potential distribution of moso bamboo （hereafter termed hybrid model） based on pseudo-absences and true presences data. Sensitivity, Kappa and the area under the curve （AUC） values of receiver operator characteristic （ROC） curve were employed to assess model predictive accuracy. Meanwhile, we investigated the sample size effects of pseudo-absences generated by DOMAIN on model performance. We also compared model performance of hybrid model with single model-NeurnalEnsembles. Results indicated that the hybrid model could achieve a higher accuracy in simulating current distribution of moso bamboo in comparison to single model. Sensitivity and AUC were relatively independent from pseudo-absence sample size, but Kappa declined with the increasing pseudo-absence sample size. Climate change is likely to have dramatic effects on the potential distribution of moso bamboo, with the northward migration ranging from 33 to 266 km, and the area expansion by 7.4% to 13.9%.
出处
《林业科学》
EI
CAS
CSCD
北大核心
2011年第7期20-26,共7页
Scientia Silvae Sinicae
基金
国家自然科学基金重大项目课题(30590383)
林业公益性行业重大科研专项(200804001
201104006)
中国林业科学研究院院所基金海外人才专项(CAFYBB2008007)
"十一五"科技支撑项目(2006BAD03A04)
国家科技部国际科技合作项目(2008DFA32070)资助
