Estimating Community-Level Plant Functional Traits in a Species-Rich Alpine Meadow Using UAV Image Spectroscopy

Plant functional traits at the community level (plant hereafter) are commonly used in trait-based ecology for study of vegetation–environment relationships. Previous studies have shown that a variety plant species or can be successfully retrieved by airborne spaceborne imaging spectrometer homogeneous, species-poor ecosystems. However, findings from these may not apply to heterogeneous, species-rich Here, we aim determine whether unmanned aerial vehicle (UAV)-based hyperspectral could adequately estimate alpine meadow ecosystem on Qinghai–Tibet Plateau. To achieve this, compared performance four non-parametric regression models, i.e., partial least square (PLSR), generic algorithm integrated with PLSR (GA-PLSR), random forest (RF) and extreme gradient boosting (XGBoost) retrieval 10 using visible near-infrared (450–950 nm) UAV imaging. Our results show chlorophyll a, b, carotenoid content, starch specific leaf area thickness were estimated good accuracies, highest R2 values between 0.64 (nRMSE = 0.16) 0.83 0.11). Meanwhile, estimation accuracies nitrogen phosphorus height dry matter content relatively low, varying 0.3 0.24) 0.54 0.20). Among tested algorithms, GA-PLSR produced accuracy, followed XGBoost, RF showed poorest performance. Overall, our demonstrates UAV-based has potential accurately multiple natural grassland fine scale.