ABSTRACT Forest biomass estimates are required for many applications, including accounting for the role of forests in the global carbon cycle, supporting sustainable forest management and making informed decisions. For all these applications and others, accurate and reliable forest biomass estimates are required. This study evaluated the potential of Sentinel-2 data for predicting and mapping aboveground biomass (AGB) of a dry Afromontane Forest in Tigray, Northern Ethiopia. Multiple linear regression was employed to model the relationship between AGB and Sentinel-2-derived spectral variables. The evaluation criteria for best-fit model selection were based on the coefficient of determination (R2), root mean square error (RMSE, %), and bias (Bias, %). All the spectral variables evaluated here were significantly correlated with AGB (p < 0.01). The model that includes a fraction of photosynthetically active radiation (FAPAR), leaf area index (LAI), Band 2 and Band 3 as predictor variables provided the best predictive performance for AGB in the study area (R 2 = 0.38, RMSE = 13.62 Mg ha−1 and Bias = -3.10 Mg ha−1). The predicted AGB of the study area ranges from 0.1 to 141.8 Mg ha−1, with a mean value of 12.4 Mg ha−1. The results from this study suggested that Sentinel-2 data can be potentially applied for estimating and mapping AGB in dry Afromontane forests.
Read full abstract