Aboveground Biomass Prediction Model Using Landsat 8 Data: A Test on Possible Approaches for Seasonally Dry Forests of Northern Ethiopia

Forest biomass is an important vegetation attribute for assessing forest ecosystem productivity and function. Estimation of tropical forest biomass has gained importance due to the significance of tropical forests in the global carbon balance. Using remote sensing has been suggested as a potential aboveground biomass (AGB) estimation approach for its spatial coverage and cost effectiveness. Application of such techniques for seasonally dry tropical forests in developing countries is, however, less explored. To this end, the aim of this study is to test the potential of Landsat 8 data for AGB prediction for dry forests in northern Ethiopia. A protected dry Afromontane forest (exclosure), representing a common vegetation type in the region, is selected. The forest is dominated by early successional species, mainly Acacia etbaica. The data necessary for ground-based AGB estimation is collected from 60 circular plots of 200 m. This study operates with a Landsat 8 image acquired in December 2017, corresponding to the year of vegetation inventory. After correcting the image for sensor, solar, atmospheric, and topographic effects, seven commonly used vegetation indices are derived from the spectral bands. AGB is significantly correlated to all of the Landsat 8 derived spectral variables. One important finding is a linear relationship between AGB, and all of the spectral variables, which allow the development of a linear model using a normalized difference vegetation index (NDVI) as a predictor. The root mean square error of the values predicted using the model is 4 Mg/ha. Finally, the findings allow to conclude that the remotely sensed data have the potential for predicting AGB of seasonally dry forests.