8256

7654

كارشناسي ارشد

مرتعداري

اصفهان: دانشگاه صنعتي اصفهان، دانشكده منابع طبيعي

1392

چهارده،100ص.: مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

مصطفي تركش اصفهاني

محمدرضا وهابي

19/9/92

حسين بشري، محمودرضا همامي

مهندسي منابع طبيعي

ID7654

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Potential Habitat Modelling of Ferula ovina Boiss using Artificial Neural Network Multivariate Adaptive Regression Splines in Fereydunshahr Region Isfahan province Zahra Rahmati Email Address Z Rahmati@na iut ac ir 22 September 2013 Department of Natural Resources Isfahan University of Technology Isfahan 84156 83 111 Iran Degree M Sc Language Farsi Supervisor Mostafa Tarkesh Isfahani Email Address M Tarkesh@cc iut ac ir Abstract Predicting species spatial distribution requires an appropriate understanding of effective ecological factors which play substantial role in evaluation protection and development palns The lack of appropriate environmental data is a challenging issue in Species Distribution Modeling SDM SDM approaches are performed by various mathematical algorithms This study aimed to generate potential distribution map of Ferula ovina Boiss a valuable medicinal plants in Fereydunshahr rangelands in Isfahan province Artificial Neural Networks ANN and Multivariate Adaptive Regression Splines MARS methods were used to map the potential habitat of the species Some topographic edaphic and climatic maps were prepared using Kriging and inverse distance weighting methods The occurrence data of Ferual ovina were collected from 278 sites 137 presence and 141 absence sites and the relationships between environmental variables and species occurrence were determined using ANN and MARS methods These models were then transferred from ecological space to geographical space using mapping function in geographic information system The results of the two studied models were comparable because similar input variables for both models were used Five independent variables sand silt organic matter elevation and slope out of 31 produced environmental maps were selected as models inputs according to the results of principle component analysis correlation matrix and box plots The evaluation of produced models by area under curve of ROC plot and kappa coefficient indicated that both models had good performance The AUC of ANN and MARS were 0 9 and 0 84 respectively According to the results both models were enable to predict spatial distribution of Ferula ovina in a local scale appropriately Habitat suitability maps were generated using optimal threshold maximum kappa ANN and MARS models revealed that 36713 and 27379 hectares 37 and 27 of the studied area have potential habitat of the species respectively According to species response curves the most abundance of Ferula ovina occurs in the elevation of 2400 to 2800 meters from sea level the slope between 15 to 40 degree sand 9 to 20 25 silt 22 to 28 and organic matter 3 5 to 4 75 The studied modeling approaches can be used broadly to achieve appropriate management plans for allocating appropriate areas for rehabilitation and protection of other valuable native plant species in rangeland ecosystems Key words Species Distribution Model Potential habitat of Ferula ovina Artificial Neural Network Multivariate Adaptive Regression Splines Geo statistic Geographic information system Fereydunshahr PDF created with pdfFactory trial version www pdffactory com

مصطفي تركش اصفهاني

محمدرضا وهابي

حسين بشري، محمودرضا همامي