Estimating Aboveground Biomass of a Regional Forest Landscape by Integrating Textural and Spectral Variables of Sentinel-2 Along with Ancillary Data.

Autor: Behera, Debabrata, Kumar, Vinjumuri Ashok, Rao, J. Prakasa, Padal, S. B., Ayyappan, N., Reddy, C. Sudhakar
Zdroj: Journal of the Indian Society of Remote Sensing; Apr2024, Vol. 52 Issue 4, p917-929, 13p
Abstrakt: Aboveground biomass (AGB) estimation is crucial for assessing forest productivity, carbon sequestration, and functional diversity. Integrating optical remote sensing data with field inventory to estimate AGB saturates at high biomass. However, transformed optical imagery is effective in tropical forest AGB estimation, especially for eliminating saturation effects. Recent AGB mapping research uses diverse predictor data fusion and advanced machine learning models. In this research, we extracted texture parameters using the Gray-level co-occurrence Matrix from the first two principal components of Sentinel-2's multispectral bands. We also calculated the Normalized Difference Vegetation Index, Visible Atmospheric Resistance Index, and Leaf Area Index for the analysis. Elevation, slope, and aspect data from SRTM DEM and GEDI-Landsat tree height product were used as ancillary datasets due to their significant impact on AGB. Neighborhood statistics (3 × 3 pixels) of predictor variables were calculated to account surrounding contributions of the focal plot. A total of 60 plots of 0.1 ha were established across the landscape where 70% and 30% of randomly selected plots were used for random forest model development and validation respectively. The final model explained AGB variability significantly (correlation coefficients = 0.72, root mean square error (RMSE) = 69.18 Mg/ha, mean absolute error (MAE) = 58.22 Mg/ha) with an uncertainty observation of 41.3 percent relative RMSE (rRMSE). The study concluded that combinations of texture and spectral variables derived from Sentinel-2 optical imagery along with physical variables are found effective in AGB mapping. The method and obtained results were promising and appeals to its replicability for building a generalized AGB model. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: