%0 Journal Article %T 延河流域典型物种分布预测模型比较研究 %T Comparison of Predictive Models for Representative Species Distribution in Yanhe River Basin %A 曹剑侠 %A 温仲明 %A 李锐 %A CAO,Jianxia %A WEN,Zhongming %A LI,Rui %J 水土保持通报 %J 水土保持通报英文版 %@ 1000-288X %V %N 3 %D 2010 %P 134-139 %K 物种分布预测;R—BIOMOD;最优模型 %K predictive distribution of species;R-BIOMOD;best model %X 物种分布预测一直以来都是生态学研究的重要内容之一。应用生态学的发展为物种分布预测提供了众多强有力的模型,在推进物种分布预测进展的同时,也增加了合适模型选择的难度。评价和比较不同模型的预测效果,对于模型的选择和应用具有非常重要的意义。以黄土丘陵区延河流域为研究区,采用R语言和BIOMOD程序包为平台,选择人工神经网络(artificial neural networks,ANN)等9个较常用的物种分布模型,比较它们在物种分布预测精度上的差异,为物种分布预测模型的选择提供依据,也为进一步预测未来气候变化情景下物种空间分布的变化奠定基础。研究结果表明,不同模型对不同物种的模拟精度差异明显。根据Kappa,TSS和Roc评价方法,9个模型对百里香(Thymus mongolicus)分布的预测精度最高;对铁杆蒿(Artemisia gmelinii)分布的模拟精度最差;而对其余物种分布的模拟精度均比较理想,其中以随机树RF模型最好。 %X The spatial distribution of species is one of the important research projects in ecology. In recent years,much progress has been made in predictive models of species with the development of applied ecology, but on the other hand, it makes more difficult to use the species distribution models because the number of available techniques or models is large and is increasing steadily, making it confused for users to select the most appropriate methodology for their needs. So evaluat ing and comparing the predictive accuracy of different distribution models has great significances for selecting and using these models. BIOMOD ( BIOdiversity MODelling ), a new computation framework based on R language, is presented. Yanhe River basin was selected as the study area. In order to select a suitable model for simulating and predict ing the typical species distributions in the study area, nine widely used modeling techniques in species predictions, such as artificial neural networks ( ANN), were compared to predict the representative species distribution. The comparison of their dif ferences in predict ion accuracy can not only provide evidence for select ing species dist ribution models, but also lay a foundation for projecting the species spatial distributions into different environmental conditions ( e. g. climate change scenarios) . Three available techniques of Roc, Kappa, and TSS were used to assess each model's performance. Results from the model comparison showed that the relative model performance and simulation accuracy of different models were quite different across species. The evaluation indicated that the nine models had the highest predictive accuracy for the Thymus mongolicus dist ribution, yet the predictive accuracy of the Artemisia gmelinii distribution was the lowest. Furthermore, the nine models can predict the distribution of the rest species very well. It is concluded that the RF model which has the highest predictive accuracy using the three methods is the best one among the nine models. %R %U http://stbctb.alljournal.com.cn/stbctb/home %1 JIS Version 3.0.0