Abstract
This study aims to comparatively assess the effectiveness of spectral and structural parameters of vegetation in estimating aboveground carbon (AGC) stock by vegetation type. A total of 38 structural metrics (including 21 percentiles) derived from LiDAR data, and 105 spectral indices and reflectance (including 10 percentiles for each of them) from Landsat 8 imagery were comprehensively assessed. It is found that the best-performing structural parameters vary with vegetation type. Namely, standard deviation of height is the best predictor for trees (R-2 = 0.83) while the mode of height is the best for shrubs (R-2 = 0.64). Of the spectral parameters, GNDVI(p80) (80 percentile of normalised difference vegetation index of green band) is the best for trees (R-2 = 0.51), and Green_Max is the best for shrubs (R-2 = 0.44). Furthermore, the estimation models based on structural parameters (R-2 >= 0.83 and RMSE >= 46.8 Mg C ha(-1) for trees, and R-2 >= 0.57 and RMSE >= 9.3 Mg C ha(-1) for shrubs) are more accurate than those based on spectral parameters (R-2 >= 0.46 and RMSE >= 54 Mg C ha(-1) for tress, and R-2 >= 0.44 and RMSE >= 11.9 Mg C ha(-1) for shrubs) despite the identified inaccuracy in LiDAR-derived height. Nevertheless, the joint consideration of both spectral and structural parameters in the same estimation model does not make it markedly more accurate than those involving either structural or spectral parameters.
NSTL
Elsevier