Şükrü Teoman GÜNER 1 *, Kürşad ÖZKAN 2 , Aydın ÇÖMEZ 3 1 Research Institute for Forest Soil and Ecology, 26160, Eskişehir, Turkey, *e-mail: stguner@hotmail.com (corresponding author) 2 Department of Soil Ecology, Suleyman Demirel University, 32260, Isparta, Turkey, e-mail: kursadozkan@gmail.com 3 Research Institute for Forest Soil and Ecology, 26160, Eskişehir, Turkey, e-mail: acomez@hotmail.com KEY FACTORS IN THE SITE SELECTION OF ROSA CANINA (L.) – APPLYING THE GENERALIZED ADDITIVE MODEL POLISH JOURNAL OF ECOLOGY (Pol. J. Ecol.) 59 3 475–482 2011 Regular research paper ABSTRACT: Determining suitable sites for species is of great importance in the choice of plant species to be used for ecosystem restoration. The number of plant species to be used in resto- ration of degraded ecosystems is restricted by cli- mate in arid and semi arid regions. Planting with economically important species in addition to ecological aspects is preferred by decision makers in Mediterrenian countries including Turkey. This study was addressed to determine the significant environmental factors influencing the distribu- tion of dog rose hip (Rosa canina) and to obtain its distribution model. The study was carried out in an area of 26,400 km 2 in The Inner Anatolia which has semiarid climate. 106 sample plots, 20  20 m in size, were selected from different sites with and without Rosa canina in the summer of 2007. Wil- coxon rank-sum statistic for continuous variables and Pearson Chi-square tests for categorical vari- ables were applied. Altitude, slope degree, avail- able water capacity, soil reaction and soil K, Ca, Mg content as well and total calcium carbonate amount as the continuous variables and north- west-northeast group of aspect, metamorphic group of bedrock as the categorical variables were determined as the significant factors influencing the presence of dog rose. Generalized Additive Model (GAM) was performed for modeling the distribution of the species. To obtain the best model, all significant environmental factors were evaluated. The best model result was obtained (training Area Under Curve (AUC) = 0.902 and cross-validation AUC = 0.841) by means of altitude, slope degree, aspect, and bedrock. Interspecific correlation analysis (ICA) was applied to define indicator species accompanying the dog rose. Dog rose distribution was positively correlated with Cistus laurifolius, Pyrus elaeagni- folia, Lonicera caucasica subsp. orientalis, Quercus vulcanica and Amelanchier rotundifolia subsp. ro- tundifolia. The results obtained from this study are cru- cial for estimation of the suitable areas for Dog rose in restoration studies of natural ecosystems in the Mediterranean region, including Anatolia in Turkey. KEY WORDS: habitat suitability, distribu- tion models, GAM-model, Rosa canina species- environmental relations 1. INTRODUCTION The genus Rosa L., one of the most wide- spread members of the Rosaceae family, is widely distributed in Eurasia, North Ameri- ca, and Northwest Africa (D av i s 1972). Dog rose (Rosa canina L.) is a widespread shrub species in Europe including Turkey, planted massively in the open landscape and in gar- dens (Özbek et al. 1999, Demir and Özcan 2001, Jürgens et al. 2007). journal 27.indb 475 journal 27.indb 475 2011-10-06 10:09:04 2011-10-06 10:09:04