Forest Biomass Inversion in Jilin Province of China Based on Machine Learning and Multi-source Remote Sensing Data

Abstract

Forest is an important part of natural ecosystems and the environment in which living things depend. Forest is constantly changing under the influence of natural and man-made forces. Therefore, monitoring and estimating forest parameters play a key role in forest resource management, while forest growth situation is most directly reflected by the forest biomass and stock. The study area is located in Jilin Province of China, belonging to a national forest park. Combining the spectral data of Sentinel-2 with the radar data of GF-3, an inversion method of forest biomass and stock was proposed in this paper. The actual data of the field experiment in 2018 was used in this paper to validate the proposed method. Twelve vegetation indexes of spectral features were extracted, for example, ratio vegetation index (RVI), normalized difference vegetation index (NDVI) and soil adjustment vegetation index (SAVI). For the radar data of GF-3, the backscattering coefficients were extracted. Taking sixteen extracted eigenvalues as the independent variables, these variables with high importance to the model were further selected to participate in the construction of the inversion model of forest biomass based on random forest. Finally, the biomass of the research area was effectively obtained.