Document Type : Research Paper
Authors
1
M.Sc. Student, Forestry Department, Faculty of Forest Science, Gorgan University of Agriculture and Natural Resources, Gorgan, I.R.Iran.
2
Asso. Professor, Forestry Department, Faculty of Forest Science, Gorgan University of Agriculture and Natural Resources, Gorgan, I.R.Iran.
3
Assist. Professor, Remote Sensing and Geographical Information System Department, Faculty of Geography, University of Tehran, Tehran, I.R.Iran.
4
Assist. Professor, Forestry Department, Faculty of Forest Science, Gorgan University of Agriculture and Natural Resources, Gorgan, I.R.Iran.
Abstract
Quantifying the forest biophysical variables such as biomass in local, zonal, and regional scales is necessary for evaluation, monitoring, and management of carbon sequestration conditions. Polarized SAR images in L band (like Alos-Palsar) have been used for estimating AGB due to its ability in penetration into canopy and extracting the trunk information. The capability of Alos-Palsar data for above-ground biomass estimation of trees was studied in some part of mixed hardwood forest of Dr. Bahramnia forestry plan. The aboveground biomass of trees was computed in 308 0.1 circular sample plots using local allometric equations. In this study, we used radiometric and geometric processed radar data and according that, normalized backscatter coefficients in HH and HV polarization, ratio and difference of them, alpha and entropy components from Cloud-Pottiers target decomposition approach and GLCM texture features were extracted from Alos-Palsar images on sample plots. Biomass modeling and estimation were done using 70 percent of sample plots by KNN, SVR and Random Forest nonparametric algorithms as well as multiple linear regression algorithms. The validity assessment was done by using the remaining 30 percent samples. According to the results, KNN algorithm had better performance than the other algorithms in estimation of aboveground biomass of trees. RMSE (57.189%) and adjusted R2 (0.032) showed the weak performance of this approach in aboveground biomass estimation of trees.
Keywords
)