Mapping 30 m Fractional Forest Cover over China’s Three-North Region from Landsat-8 Data Using Ensemble Machine Learning Methods

The accurate monitoring of forest cover and its changes are essential for environmental change research, but current satellite products coverage carry many uncertainties. This study used 30-m Landsat-8 data, aggregated 1-m GaoFen-2 (GF-2) images to construct the training samples multiple machine learning algorithms (MLAs) estimate fractional (FFC) in China’s Three North Region (TNR). In this study, MLAs were merged stacked generalization (SG) models based on idea SG, performances FFC estimation evaluated. results 10-fold cross-validation showed that all non-linear had a good performance, with an R2 value greater than 0.8 root-mean square error (RMSE) less 0.05. bagging ensemble, random (RF) (R2 = 0.993, RMSE 0.020) model performed best boosting light gradient boosted (LGBM) 0.992, 0.022) best. Although evaluation index RF is slightly better LGBM, independent validation show two have similar performances. datasets that, SG model, performance SG(LGBM) 0.991, 0.034) was single or non-ensemble model. Comparing estimates our those existing exhibited more spatial distribution details higher accuracy complex landscapes. Overall, method using high-resolution remote sensing (RS) extract feasible. Our demonstrate potential ensemble map FFC. research also among MALs, algorithm most suitable estimating FFC, which provides reference future research.