Species Potential Distribution Models and Evaluation Based on Dissimilarity Index of Variables of Qinghai Spruce (picea Crassifolia) in Qilian Mountains

XU Zhong-lin,ZHAO Chuan-yan,FENG Zhao-dong
2011-01-01
Abstract:Three species potential distribution models(border function(BF),maximum entropy(Maxent) and genetic algorithm for rule set production(GARP)) and four environmental variables(aspect,elevation,annual mean temperature and precipitation) were selected to predict the potential distribution of Qinghai spruce(Picea crassifolia) in Qilian mountains national reserve.A new approach was introduced which could extract environmental variables above the actual and potential distribution areas;then differences between two sets of environment variables were compared and the differences were demonstrated by computation of marginal dissimilarity index of variables(MDIV) and dissimilarity index of variables(DIV).Our results indicate that GARP obtained four minimum MDIVs in all the nine variables,Maxent and BF obtained three and two minimum MDIVs,showing the best performance of GARP among the three models.Total dissimilarity indexes of variables(TDIVs) for Maxent,GARP and BF were 15.145 5,10.996 5 and 18.874 7 respectively,also indicating that GARP is better than BF and Maxent in prediction of distribution of species.The middle part of the study area,administerially including Sunan,Minle and Qilian counties,are the most suitable growth area for Qinghai spruce(Picea crassifolia).The potential distribution is larger than the real distribution at southern part of thestudy area(Tianzhu county),which indicates that there has been an improvement of potential ecology at this part of the study area under current environmental conditions.
What problem does this paper attempt to address?