Optimizing KNN and multi-features extracted from GF-1 and GF-3 for forest above-ground biomass (AGB) estimation

China is advancing toward its carbon peak and neutrality goals, in which accurate above-ground biomass (AGB) estimation is crucial. In this study, GF-1, GF-3, and their combined data from Yiliang and Puer were used, testing KNN and two enhanced algorithms (KNN-FIFS and GA-KNN) for AGB estimation. We found that (i) KNN-FIFS selected more significant features, primarily GF-1 textures and GF-3 polarimetric decompositions. (ii) GF-1/GF-3 combination features achieved the highest accuracy across four forest types. (iii) KNN-FIFS outperformed others in total and component AGB estimation. (iv) Component-based AGB summation improved accuracy by 3% over direct estimation. In summary, to improve the efficiency of forest AGB estimation, the optimizations of introducing GA and FIFS for optimal feature selection into the KNN algorithm are effective. Our results confirmed the effectiveness of KNN-FIFS and GA-KNN applied to GF-1/GF-3 fusion for both total and component AGB estimation and also compared their performance. Meanwhile, a 3% improvement using component-based estimation suggests a viable strategy for accurate biomass monitoring.