Investigation on the adequacy of ordinary least square regression in modeling the oriental beech tree height in relation to environmental predictors in experimental and educational forest of Tarbiat Modares University

Dormann, Carsten

Investigation on the adequacy of ordinary least square regression in modeling the oriental beech tree height in relation to environmental predictors in experimental and educational forest of Tarbiat Modares University

Document Type : Research Paper

Author

Carsten Dormann

Abstract

In the current research, the performance of ordinary least square regression model was studied for the task of predicting a key forest structural parameter – tree height – across a study area in Tarbiat Modares University forest research station using a series of edaphic and topographic variables. For this purpose, 123 0.1 ha circular sample plots were established and total height and diameter of Fagus orientalis Lipsky trees with DBH ≥ 7.5 cm within each plot was recorded along with elevation, azimuth and slope of the ground. Also, at the center and four geographical aspects of sample plot, soil samples from first layer (0-10 cm) were taken and mixed for analyzing several soil variables. The results showed the OLS model performed moderately well based on R-squared, but exhibited clear signs of spatial autocorrelation (Moran’s I =0.168). Adding spatial weighted matrix in spatial simultaneous autoregressive models resulted in removing autocorrelation and statistically significant improvement in model fit. Comparison of spatial lag and error SAR models using AIC, log-likelihood and R squared indicated that error SAR model performs better than lag SAR model. These analyses underscore the importance of controlling for spatial autocorrelation in forest ecology studies and furnish guidelines for future modeling of species performance in relation to environmental predictors.

Keywords

Moran’s I statistic

Neighborhoods’ structure

Spatial error model

Spatial lag model

Spatial Autocorrelation

Subjects