Potential of Landsat-8 spectral indices to estimate forest biomass

Forest ecosystems are among the largest terrestrial carbon reservoirs on our planet earth thus playing a vital role in global carbon cycle. Presently, remote sensing techniques provide proper estimates of forest biomass and quantify carbon stocks. The present study has explored Landsat-8 sensor product and evaluated its application in biomass mapping and estimation. The specific objectives were estimation of above ground biomass and carbon stocks using field data, assessing relationships of Landsat-8 spectral indices and field data and modeling of biomass and carbon stocks based on best linear regression model. Results showed that the highest aboveground biomass and below ground biomass was recorded as 246 t/ha and 64 t/ha whereas the lowest aboveground biomass and below ground biomass was 55 t/ha and 14 t/ha, respectively. Similarly, the highest above ground carbon and below ground carbon (t/ha) were 116 t/ha and 30 t/ha respectively while the lowest above ground carbon and below ground carbon (t/ha) were estimated as 26 t/ha and 6.7 t/ha respectively. Indices computed from Landsat-8 included normalized difference vegetation index, difference vegetation index, soil adjusted vegetation index, perpendicular vegetation index and atmospherically resistant vegetation index. Regarding relationship between aboveground biomass and vegetation indices, the coefficient of correlation (R2) were 0.67, 0.68, 0.65, 0.58 and 0.23 for normalized difference vegetation index, soil adjusted vegetation index, Perpendicular vegetation index, difference vegetation index and atmospherically resistant vegetation index respectively. The stepwise correlation between aboveground biomass (dependent variable) and five indices (Normalized difference vegetation index; soil adjusted vegetation index; Perpendicular vegetation index; difference vegetation index; atmospherically resistant vegetation index). Among five vegetation indices, only soil adjusted vegetation index was selected in stepwise method, satisfying the criteria and the overall model R^2 was 0.63 and its adjusted R2 was 0.60. Simple linear regression model between aboveground biomass and single predictor index was better than stepwise regression model with (R^2= 0.68) and (Root mean square error = 33.75 t/ha). Thus, soil adjusted vegetation index was considered best for biomass mapping. The study concluded that Landsat-8 product has considerable potential for biomass and carbon stocks estimation and can be expanded to national and regional forest inventories, modeling and future reducing emission from deforestation and forest degradation+ implementation.