Application of topography and logistic regression in forest type spatial prediction

Abstract

This research was carried out for predicting probability of forest type's presence using topographic attributes in the Educational and Research Forest of Doctor Bahramnia, district 1, with 1714 ha area. Field sampling was performed, based on cluster random sampling and systematic random sampling and 249 plots were sampled with 0.1 hectare area (without plantation area). Diameter of trees larger than 12.5 cm, species information and geographic position were registered in each plot with GPS. Since, thick trees are dominated in the canopy cover, 10 thick trees in each plot were selected to determine the forest type. Using computing of frequency percent of species, forest type in each plot was determined and four types of Fagus-Carpinus, Quercus-Carpinus, Carpinus-Parrotia and Parrotia-Acer (mixed with Carpinus betulus) were determined in the study area. The topographic attribute maps were derived from a digital terrain model and these attributes were extracted in location of plots. Logistic regression was implemented to analysis of forest type’s correlation with attributes and to construct a predictive model. The analysis was performed using 70% of the plots and each forest type map was resulted by extrapolation of model in GIS. Validation of results was performed by 30% of the remained plots and total accuracy computed for each model. Validation result of accuracy showed that spatial prediction models of forest types for Fagus-Carpinus and Quercus-Carpinus have narrow amplitude than Carpinus-Parrotia and Parrotia-Acer (mixed with Carpinus betulus) types that spread in large extent region. This result showed that spatial predictive models of forest types, which have narrow amplitude, are more acceptable. The results also showed that consideration of altitude, solar radiation potential and aspect in the model were the main factors which are controlling forest types in the study area.

Keywords

• Spatial predictive map
• forest types
• primary and secondary topographic attributes
• Logistic regression