Estimation of forest stand volume, tree density and biodiversity using Landsat ETM + Data, comparison of linear and regression tree analyses

Estimation of forest attributes using remotely sensed data has being as a new potential for continuous management of natural resources. Simple statistical models such as linear regressions are most used approach that has been used by researchers. Applying other regression types in forest attribute estimations and their spatial modeling using decision tree analysis such as regression tree may be more usefulness compare to linear regression. In a case study in the Hyrcanian forests, northern of Iran, the capability of linear and regression tree analyses were compared to estimation of stand volume, tree density and tree diversity. Stepwise multiple regression and regression tree analyses were conducted to evaluate relationships between forest characteristics as dependent and ETM + bands and vegetation indices as independent variables. Performance assessment of models was examined using RMSE and Bias on the unused validation plots. The results of analysis showed that statistical models of stand volume, tree density, species richness and reciprocal of Simpson indices using tree regression analysis had higher adjusted R2 and CE compare to linear regression models. In addition, the performance results showed that RMSE of models using tree regression were 88.7 m3/ha, 157n/ha, 1.15 and 0.61 respectively for stand volume, tree density, species richness and Simpson index, Whereas, the RMSE of obtained models using linear regression were computed about 97m3/ha, 170n/ha, 1.51 and 1.15, respectively.