Urban forest parks are vital ecological barriers that safeguard urban ecological security and provide essential ecosystem services. Aboveground biomass (AGB) is a key indicator for evaluating these services. This study targeted three tree species—Ligustrum lucidum, Camphora officinarum and Koelreuteria paniculata—in Haiwan National Forest Park of Shanghai, China. Based on field-measured individual tree AGB, high-density point clouds from terrestrial laser scanning (TLS), and features from UAV multispectral imagery, four machine learning models—Random Forest (RF), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Support Vector Regression (SVR)—were developed. SHapley Additive exPlanations (SHAP) analysis was conducted to identify key predictors and quantify their importance. The results show that: (1) Data fusion of TLS and multispectral imagery significantly improves estimation accuracy compared with single data sources, with RF consistently achieving the best performance across species (test set R² = 0.96, 0.92, and 0.91 for L.
Urban forest parks are vital ecological barriers that safeguard urban ecological security and provide essential ecosystem services. Aboveground biomass (AGB) is a key indicator for evaluating these services.
Note: The original text includes procedural prompts and metadata unrelated to the scholarly content. In the cleaned and normalized version, only the core research-related statements are retained, preserving quotes exactly as they appear where appropriate and removing extraneous instructions.
Data fusion of TLS and multispectral imagery significantly improves estimation accuracy compared with single data sources, with RF consistently achieving the best performance across species (test set R² = 0.96, 0.92, and 0.91 for L.
Summary of approach and findings are preserved in a concise form, suitable for HTML presentation, without introducing new facts or interpretations beyond the source material.
Авторское резюме: Совмещение TLS и спутниковых данных повышает точность оценки биомассы деревьев по сравнению с использованием единого источника данных, причем RF демонстрирует наилучшие результаты по всем видам.