Estimation of aboveground biomass using optical and radar images (Case study: Nav-e Asalem forests, Gilan)

Abstract

Using remote sensing data is an applied method to estimate above ground biomass. In this study, satellite radar data of ALOS-2, with the full polarization and the optical data of Sentinel-2, has been used to estimate the aboveground biomass in the Nav-e Asalem forests, Gilan province. The backscattering coefficients at different polarization, the texture measures and target decomposition features of SAR images, obtained original and synthetic bands from optical images in three different combinations of radar images, optical images and the composition of radar and optical images, as inputs to the Artificial Neural Network (ANN) and Multiple Linear Regression (MLR) models were used. In order to measure aboveground biomass, 149 sample plots were laid out. Evaluation of ANN and MLR models using R2 and RMSE statistics showed that in all cases the ANN was better performance to estimate the aboveground biomass than MLR. The best results showed that the ANN from combined optical and radar data with R2 and RMSE, 0.86 and 31.62 Mg/ha (15.34%), respectively, can be the best applied method to estimate the aboveground biomass. The results of radar images and optical separately, with the R2 and RMSE for the modeling of aboveground biomass have been shown, respectively, 0.57 and 49.17 Mg/ha (23.85%) by radar images and 0.7 and 39.53 Mg/ha (19.17%) by the optical images, superior modeling to estimate aboveground biomass represents by optical imaging. The overall and more accurate results to estimate of aboveground biomass have been shown when we used combined radar and optical images with the ANN model.

Keywords

• ALOS
• Artificial neural network
• Remote Sensing
• SAR
• Sentinel