Evaluation of capability of IRS-P6 satellite data for predicting quantitative attributes of forests (case study: Northern Zagros forests)

The aim of this research was to predict tree density and basal area using IRS-P6 satellite data in the northern Zagros forests. A random-systematic sampling grid consists of 312 circular sample plots (each plot, 0.1 ha) were used to collect field data. The images were georeferenced using 29 ground control points. Spectral values related to field plots were extracted from original and synthetic bands composed of vegetation indices, principle component analysis and data fusion. Ancillary data such as slope, aspect and elevation are also extracted. Multiple regression and stepwise method were used to predict tree density and basal area from original and synthetic bands as independent variables. The best models (at first just original bands and then combined of original and synthetic bands) were selected using RMSE, Bias, Correlation and the F values (the best model for tree density: R²_adj = 0.31 & for basal area: R²_adj = 0.38). Using slope, aspect, and elevation ancillary data did not improve the results.