Abstract
This study investigates the above-ground biomass estimation of living trees through the application of remote sensing and field data in the plantation forest of Amhara regional state, Tehuledere woreda. Aboveground biomass of trees above 10 cm diameter is considered in this study.
The study used field inventory data in 26 sample plots and Landsat 8 OLI TIRS satellite imagery to extract information and develop biomass prediction. The process of biomass predicting includes: calculating of plot-based biomass using species-specific allometric equations and extracting different vegetation indices namely NDVI, SAVI, GNDVI, EVI, and NDMI. Accordingly, Pearson correlation analysis was done to see if the selected VI’s have a significant correlation with AGB and found all VI’s have a Pearson correlation value > 0.05. The relation between plot biomass and vegetation indices were also investigated; for every VI independently using simple linear regression and for overall using multiple linear regression equation in R-statistical software version 3.3.3. Using the backward elimination process, the individual VI’s and above-ground biomass have low to moderate relation and the overall best correlation was found with a combination of NDVI and EVI with the predictive equation Y= -34.85+110.82(NDVI) +250.31(EVI), R2 = 0.54 and adjusted R 2= 0.50. Using this model the mean predicted biomass calculated to be 63.68 ton/ha + 10.2 ton/ha with RMSE%= 14.5 and the biomass distribution was mapped using the equation.
Keywords: Biomass, Vegetation indices, regression, Model, Landsat, Root mean square error.
