تحقیقات جنگل و صنوبر ایران (Feb 2014)
Modeling distribution of forest types of Armardeh forests at Baneh, using logistic regression method
Abstract
This research was carried out to predict potential distribution of Armardeh forests types (16482.44 hectares). Determination and classification of forest types was made, based on data of 448 circular sample plots (0.1 hectare area). Eight forest types were identified, in which four types, including “Quercus infectoria- Quercus brantii”, “Quercus brantii-other species, mixed with Quercus infectoria”, “Quercus brantii, Quercus libani mixed with other species” and“Quercus brantii” were modeled using physiographical factors. Modeling was performed by both logistic regression and stepwise methods (likelihood ratio), using 70% of the samples for modeling and 30% of them for model validation. Results showed that the achieved models for the forest types with limited distribution range, had more accuracy than the other types. According to ROC curve test, the greatest precision was allocated to models related to “Q .brantii”, “Q. brantii, Q. libanii mixed with other species”, “Q. brantii -other species, mixed with Q. infectorai” and “Q. infectoria- Q. brantii” forest types, respectively. Furthermore, due to presence of aspect in most of the models, it was distinguished as an important physiographical parameter in local forest type's distribution. Overall, according to positive and negative correlation between presence of each forest type with variables which take part in the logistic model process, its accordance with results obtained from forest type map adaptation with other studied parameters, and further similar research results, it might be concluded that logistic regression is an appropriate method to study effects of different factors on spatial distribution of various forest types. Forest types predicted probability maps, could be used as a management tools for development and rehabilitation of forest ecosystems.
Keywords
