Estimation of Forest Structural Attributes Using ASTER Data

Estimation of forest structural parameters is one of major basic information in sustainable management and planning in forest stands. In this study، relationship between ASTER satellite data and three forest structure factors including stand volume، basal area and number of trees per hectare were investigated in Darabkola forest، northern Iran. A multivariate linear regression approach was used to analyze and evaluate relationship between mentioned characteristics and ASTER satellite data. Relevant preprocessing and methods were conducted on spectral data. After gathering terrestrial information، stand volume، basal area and number per hectare were calculated for sample plots. Using some plots، performance of the best models examined by relevant evaluation criterions. The results showed that a combination of MSAVI2، NDVI and Green bands could predict stand volume characteristics better with R2adj=59. 2; and RMSE=116. 5 m3/h-1 in comparison with other indices and band combinations. For basal area، the best results were obtained using combination of MSAVI2، NDVI and simple ratio of SWIR12 with R2adj=73. 5 and RMSE=5. 14 m2/h-1. In addition، combination of MSAVI2، SWIR1 and SWIR2، was a better predictor for number per hectare rather than the other combinations by R2adj equal to 0. 85 and RMSE about 50. 95 number per hectare. Generally، this research showed that using linear regression approach by the ASTER data presents only general status of forest structure attributes in the study area and having more precise estimation of these attribute needs investigating other approaches such as nonlinear or nonparametric and learning machines approaches.