Estimation of Dendrocalamus giganteus leaf area index by combining multi-source remote sensing data and machine learning optimization model

• DOI: 10.3389/fpls.2024.1505414
• Publication Date: 2025-01-15T12:39:26Z
• Authors: Zhen Qin; Huanfen Yang; Qingtai Shu; Jinge Yu; Zhengdao Yang; Xu Ma; Dandan Duan
• Keywords: remote sensing data; ICESat-2/ATLAS; sequential gaussian conditional simulation; Plant culture; optimization algorithm; Plant Science; sentinel data; LAI; SB1-1110

The Leaf Area Index (LAI) is an essential parameter that affects the exchange of energy and materials between vegetative canopy surrounding environment. Estimating LAI using machine learning models with remote sensing data has become a prevalent method for large-scale estimation. However, existing have exhibited various flaws, hindering accurate estimation LAI. Thus, new Dendrocalamus giganteus was proposed, which integrates ICESat-2/ATLAS, Sentinel-1/-2 data, refines through application Bayesian Optimization (BO), Particle Swarm (PSO), Genetic Algorithms (GA), Simulated Annealing (SA). First, spatial interpolation performed Sequential Gaussian Conditional Simulation (SGCS) method. Then, multi-source were leveraged to optimize feature variables Pearson correlation coefficient approach. Subsequently, optimization algorithms applied Random Forest Regression (RFR), Gradient Boosting Tree (GBRT), Support Vector Machine (SVR) models, leading efficient results showed BO-GBRT model achieved high accuracy in estimation, determination (R 2) 0.922, root mean square error (RMSE) 0.263, absolute (MAE) 0.187, overall (P 1) 92.38%. Compared methods, proposed approach demonstrated superior performance. This holds significant potential forest inversion can facilitate further research on other structure parameters.