Natural disturbances like hurricanes can cause extensive disorder in forest structure, composition, and succession. Consequently, ecological, social, and economic alterations may occur. Terrestrial laser scanning (TLS) and deep learning have been used for estimating forest attributes with high accuracy, but to date, no study has combined both TLS and deep learning for assessing the impact of hurricane disturbance at the individual tree level. Here, we aim to assess the capability of TLS and convolutional neural networks (CNNs) combined for classifying post-Hurricane Michael damage severity at the individual tree level in a pine-dominated forest ecosystem in the Florida Panhandle, Southern U.S. We assessed the combined impact of using either binary-color or multicolored-by-height TLS-derived 2D images along with six CNN architectures (Densenet201, EfficientNet_b7, Inception_v3, Res-net152v2, VGG16, and a simple CNN). The confusion matrices used for assessing the overall accuracy were symmetric in all six CNNs and 2D image variants tested with overall accuracy ranging from 73% to 92%. We found higher F-1 scores when classifying trees with damage severity varying from extremely leaning, trunk snapped, stem breakage, and uprooted compared to trees that were undamaged or slightly leaning (<45°). Moreover, we found higher accuracies when using VGG16 combined with multicolored-by-height TLS-derived 2D images compared with other methods. Our findings demonstrate the high capability of combining TLS with CNNs for classifying post-hurricane damage severity at the individual tree level in pine forest ecosystems. As part of this work, we developed a new open-source R package (rTLsDeep) and implemented all methods tested herein. We hope that the promising results and the rTLsDeep R package developed in this study for classifying post-hurricane damage severity at the individual tree level will stimulate further research and applications not just in pine forests but in other forest types in hurricane-prone regions.

Effective estimation of fuel load is critical for mitigating wildfire risks. Here, we evaluate the performance of mobile laser scanning (MLS) and terrestrial laser scanning (TLS) to estimate fuel loads across multiple vegetation layers. Data were collected in two forest regions: the North Kaibab (NK) Plateau in Arizona and Monroe Mountain (MM) in Utah. We used random forest models to predict vegetation attributes, evaluating the performance of full models and transferred models using R², RMSE, and bias. The MLS consistently outperformed the TLS system, particularly for canopy-related attributes and woody biomass components. However, the TLS system showed potential for capturing canopy structure attributes, while offering advantages like operational simplicity, low equipment demands, and ease of deployment in the field, making it a cost-effective alternative for managers without access to more complex and expensive mobile or airborne systems. Our results show that model transferability between NK and MM is highly variable depending on the fuel attributes. Attributes related to canopy biomass showed better transferability, with small losses in predictive accuracy when models were transferred between the two sites. Conversely, surface fuel attributes showed more significant challenges for model transferability, given the difficulty of laser penetration in the lower vegetation layers. In general, models trained in NK and validated in MM consistently outperformed those trained in MM and transferred to NK. This may suggest that the NK plots captured a broader complexity of vegetation structure and environmental conditions from which models learned better and were able to generalize to MM. This study highlights the potential of ground-based LiDAR technologies in providing detailed information and important insights into fire risk and forest structure.

Developing the capacity to monitor species diversity worldwide is of great importance in halting biodiversity loss. To this end, remote sensing plays a unique role. In this study, we evaluate the potential of Global Ecosystem Dynamics Investigation (GEDI) data, combined with conventional satellite optical imagery and climate reanalysis data, to predict in situ alpha diversity (Species richness, Simpson index, and Shannon index) among tree species. Data from Sentinel-2 optical imagery, ERA-5 climate data, SRTM-DEM imagery, and simulated GEDI data were selected for the characterization of diversity in four study areas. The integration of ancillary data can improve biodiversity metrics predictions. Random Forest (RF) regression models were suitable for estimating tree species diversity indices from remote sensing variables. From these models, we generated diversity index maps for the entire Cerrado using all GEDI data available in orbit. For all models, the structural metric Foliage Height Diversity (FHD) was selected; the Renormalized Difference Vegetation Index (RDVI) was also selected in all species diversity models. For the Shannon model, two GEDI variables were selected. Overall, the models indicated performances for species diversity ranging from (R² = 0.24 to 0.56). In terms of RMSE%, the Shannon model had the lowest value among the diversity indices (31.98%). Our results suggested that the developed models are valuable tools for assessing species diversity in tropical savanna ecosystems, although each model can be chosen based on the objectives of a given study, the target amount of performance/error, and the availability of data.