This study developed an integrated approach for estimating tropical forest aboveground biomass (AGB) by combining UAV–LiDAR structural metrics and Sentinel-2B spectral data, optimized through successive projections algorithm (SPA) feature selection and random forest (RF) regression. Field surveys across three tropical forest sites in Hainan Province (49 plots) provided ground-truth AGB measurements, while UAV–LiDAR (1 m resolution) and Sentinel-2B (10 m) data were processed to extract 98 and 69 features, respectively. The results showed that LiDAR-derived elevation metrics (e.g., percentiles and kurtosis) correlated strongly with the AGB measurements (r = 0.652–0.751), outperforming Sentinel-2B vegetation indices (max r = 0.520). SPA–RF models with selected features significantly improved accuracy compared to full-feature RF, achieving R² = 0.670 (LiDAR), 0.522 (Sentinel-2B), and 0.749 (coupled data), with the fusion model reducing errors by 46–54% in high-biomass areas. Despite Sentinel-2B’s spectral saturation limitations, its integration with LiDAR enhanced spatial heterogeneity representation, particularly in complex canopies. The 200-iteration randomized validation ensured a robust performance, with mean absolute relative errors of ≤0.071 for fused data. This study demonstrates that strategic multi-sensor fusion, coupled with SPA-optimized feature selection, significantly improves tropical AGB estimation accuracy, offering a scalable framework for carbon stock assessments in support of Reducing Emissions from Deforestation and Forest Degradation (REDD+) and climate mitigation initiatives.

For tropical rainforest regions with dense vegetation cover, the development of effective large-scale soil mapping methods is crucial to improve soil management practices to replace the time-consuming and laborious conventional approaches. While machine learning (ML) algorithms demonstrate superior predictability of soil properties over linear models, their practical and automated application for predicting soil properties using remote sensing data requires further assessment. Therefore, this study aims to integrate Unmanned Aerial Vehicles (UAVs)-based hyperspectral images and Light Detection and Ranging (LiDAR) points to predict the soil properties indirectly in two tropical rainforest mountains (Diaoluo and Limu) in Hainan Province, China. A total of 175 features, including texture features, vegetation indices, and forest parameters, were extracted from two study sites. Six ML models, Partial Least Squares Regression (PLSR), Random Forest (RF), Adaptive Boosting (AdaBoost), Gradient Boosting Decision Trees (GBDT), Extreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP), were constructed to predict soil properties, including soil acidity (pH), total nitrogen (TN), soil organic carbon (SOC), and total phosphorus (TP). To enhance model performance, a Bayesian optimization algorithm (BOA) was introduced to obtain optimal model hyperparameters. The results showed that compared with the default parameter tuning method, BOA always improved models’ performances in predicting soil properties, achieving average R² improvements of 202.93%, 121.48%, 8.90%, and 38.41% for soil pH, SOC, TN, and TP, respectively. In general, BOA effectively determined the complex interactions between hyperparameters and prediction features, leading to an improved model performance of ML methods compared to default parameter tuning models. The GBDT model generally outperformed other ML methods in predicting the soil pH and TN, while the XGBoost model achieved the highest prediction accuracy for SOC and TP. The fusion of hyperspectral images and LiDAR data resulted in better prediction of soil properties compared to using each single data source. The models utilizing the integration of features derived from hyperspectral images and LiDAR data outperformed those relying on one single data source. In summary, this study highlights the promising combination of UAV-based hyperspectral images with LiDAR data points to advance digital soil property mapping in forested areas, achieving large-scale soil management and monitoring.

This study proposes an Automatic Branch Modeling (ABM) framework that combines AdTree and AdQSM algorithms to reconstruct individual tree models and estimate timber volume from fused Hand-held Laser Scanners (HLS) and Unmanned Aerial Vehicle Laser Scanners (UAV-LS) point cloud data. The research focuses on two 50 × 50 m primary tropical rainforest plots in Hainan Island, China, characterized by dense and vertically stratified vegetation. Key steps include multi-source point cloud registration and noise removal, individual tree segmentation using the Comparative Shortest Path (CSP) algorithm, extraction of diameter at breast height (DBH) and tree height, and 3D reconstruction and volume estimation via cylindrical fitting and convex polyhedron decomposition. Results demonstrate high accuracy in parameter extraction, with DBH estimation achieving R² = 0.89–0.90, RMSE = 2.93–3.95 cm and RMSE% = 13.95–14.75%, while tree height estimation yielded R² = 0.89–0.94, RMSE = 1.26–1.81 m and RMSE% = 9.41–13.2%. Timber volume estimates showed strong agreement with binary volume models (R² = 0.90–0.94, RMSE = 0.10–0.18 m³, RMSE% = 32.33–34.65%), validated by concordance correlation coefficients (CCC) of 0.95–0.97. The fusion of HLS (ground-level trunk details) and UAV-LS (canopy structure) data significantly improved structural completeness, overcoming occlusion challenges in dense forests. This study highlights the efficacy of multi-source LiDAR fusion and 3D modeling for precise forest inventory in complex ecosystems. The ABM framework provides a scalable, non-destructive alternative to traditional methods, supporting carbon stock assessment and sustainable forest management in tropical rainforests. Future work should refine individual tree segmentation and wood-leaf separation to further enhance accuracy in heterogeneous environments.

Forest biomass, as a carrier of carbon, is an important indicator for judging forest productivity, stability and sustainable development capacity. Using the survey data of sample plots in eight forest areas in central Hainan, the biomass distribution of tropical mountain rainforests in National Park of Hainan Tropical Rainforest in different community sizes, diameter classes, altitudes and spaces was measured to explore the relationship between forest biomass and environmental factors. The results show that (1) the total area of tropical montane rainforests in National Park of Hainan Tropical Rainforest was about 983.70 km², distributed within an altitude range of 700–1300 m; the total aboveground biomass was about 25.208 million tons, which decreased first and then increased with increasing altitude, with an average aboveground biomass per unit area of 236.00 t·hm⁻²; (2) the primary forest accounted for 83.23% of the total aboveground biomass of the tropical mountain rainforest with only 29.84% of the total area, and the aboveground biomass per unit area was generally higher than that of the secondary forest; and (3) medium- and large-diameter trees were the main carriers of aboveground biomass in tropical mountain rain forests. More than 83.73% of the aboveground biomass was concentrated in large-diameter trees. The results of this study provide a reference for others aiming to perform measurement and evaluation of the carbon sink and the capacity for carbon neutrality in tropical rainforest ecosystems or to maintain regional biodiversity.

Monitoring the α-biodiversity indicators of mangrove forests and understanding their spatiotemporal trends can guide mangrove restoration strategies. Taking Qinglan Port in Hainan Province, China, as our study area, we compared multiple machine learning methods to predict the spatial distribution of α-biodiversity indicator Shannon’s diversity index (SHDI) by integrating LiDAR points and Worldview-2 images. In addition, the relationship between mangrove forests’ SHDI values and growth years was analyzed. The study extracted 28 spectral features and 99 LiDAR features from Worldview-2 and LiDAR data, respectively. The RReliefF method was adopted to select informative features. Four machine learning methods, including support vector machines (SVMs), extreme gradient boosting (XGBoost), deep neural networks (DNNs), and Gaussian process regression (GPR), were used to establish SHDI prediction models. The leave-one-out cross-validation (LOOCV) method was used to evaluate prediction accuracy, and the optimal model was adopted to generate a spatial map of SHDI. Based on Google Earth and Worldview-2 images, the spatial regions of mangrove forests in 2008, 2013, 2018, and 2023 were identified. The SHDI values within different restoration periods were statistically analyzed by using the mangroves’ spatiotemporal distributions. The results showed that RReliefF selected a total of 30 features, including 13 spectral features and 17 LiDAR features. Using preferred features, GPR had the highest prediction accuracy, with an LOOCV R² of 0.51, followed by SVM (R² = 0.44) and DNN (R² = 0.32); the accuracy of XGBoost (R² = 0.29) was relatively poor. The increased areas of rehabilitated mangrove forests in the periods of 2008–2013, 2013–2018, and 2018–2023 were 0.31 km², 0.13 km², and 1.35 km², respectively. Mangroves growing before 2008 owned the highest mean SHDI value of 0.74, followed by mangroves in 2008–2013 and 2013–2018; mangrove forests restored in 2018–2023 had the lowest mean SHDI value of 0.63. The results indicated that mangrove SHDI can be predicted by integrating LiDAR and Worldview-2. The mangrove population exhibited more diverse α-biodiversity characteristics as growth time increased. In subsequent mangrove restoration processes, planting mangroves of diverse species is beneficial to ensure the stability of the mangrove community.

Mangrove ecosystems are recognized for their exceptional carbon sequestration potential and crucial contribution to coastal ecological balance. However, the sharp decline in mangrove area necessitates efficient monitoring via remote sensing. While Deep Neural Networks (DNNs) have excelled in segmentation tasks, their robustness remains inadequate. This limitation stems from the lack of theoretical guarantees regarding the continuity of layer-by-layer discrete transformations, rendering models susceptible not only to man-made adversarial attacks but also to natural degradations. To address these vulnerabilities, this paper leverages Neural Ordinary Differential Equations (NODEs) to enhance the robustness of mangrove segmentation. We designed and integrated various NODE architectures, including a novel NODE-SE-Block inspired by adaptive feature recalibration, to achieve more stable feature representations. Crucially, our findings reveal that by employing an adversarial training framework based on known attacks, the NODE-integrated network demonstrates superior cross-domain robustness. It not only defends against malicious exploits but also exhibits significantly enhanced resilience toward natural degradations, such as Gaussian noise and sensor-induced artifacts. Experimental results on mangrove datasets verify that the proposed methodology provides a reliable and interference-resistant foundation for ecological management in mission-critical scenarios.