Application of RF-KNN Optimal Technology for the Estimation of Forest Aboveground Biomass Using Multisource Remote Sensing Data

Abstract

Quantifying the forest above ground biomass is critical for accurate carbon accounting. Qualities of the nonparametric K Nearest Neighbour technique make it an attractive tool for forest aboveground biomass (FAGB) estimation based on remote sense data. However, a thorough analysis of the usability of model parameter and feature optimal selection for the estimation of FAGB is missing. This study based on multisource data which included optical and LiDAR data and used random forest feature selection algorithm to build optimal KNN model for the estimation of FAGB estimation. The result showed the RMSE and R2 of the optimal KNN model are 20.12 ton/hm2 and 0.8 respectively fitted for FAGB estimation.