Estimation of aboveground vegetation biomass based on Landsat-8 OLI satellite images in the Guanzhong Basin, China

This paper presents a method to estimate the aboveground biomass (AGB) through the selection of different estimation methods based on numerous vegetation types (i.e., broadleaf forest, coniferous forest, shrub and grassland) at a regional scale. The proposed method is based on three models, namely, the stepwise regression, an improved back-propagation neural network (Improved BBPNN) model based on the Gaussian error function, and the support vector machine (SVM) technique, Meanwhile, Landsat 8 Operational Land Imager (OLI)/Thermal Infrared Sensor (TIRS) image data and geo-parameters are employed to select 68 feature variables and optimize 213 data samples. Our results reveal that, the stepwise regression method provides the best AGB estimation performance for broadleaf forests and coniferous forests, while the SVM technique shows the best performance for grasslands and shrubs. Different vegetation types should be selected for additional biomass estimation models that have been proven to enhance the biomass estimation. This study on the AGB not only promotes research on the net primary productivity (NPP), but also plays a key role in global carbon cycle research.