The occurrence of hurricanes in the Southern U.S. is increasingly frequent and quantifying the damage caused to forests is crucial to assist in protection measures and understanding the dynamics of recovery. The aim of this study is to develop a data fusion framework based on NASA’s GEDI (Global Ecosystem Dynamics Investigation) and Landsat 8 OLI for mapping aboveground biomass density (AGBD, Mg/ha) that can be further used to damage severity and recovery in forested ecosystems impacted by Hurricane Ian in Florida. We used GEDI level 4A and L8 data for calibrating a Random Forest (RF) for predicting and mapping AGBD at four-months pre-Hurricane Ian disturbance across areas impacted by Hurricane Ian. The RF model showed good performance with R2= 0.79, absolute and relative RMSE of 29.17 Mg/ha (64.27%) and Bias of −1.14 Mg/ha (2.66%), respectively. This research highlights methodological opportunities for fusing GEDI and L8 data streams toward improved AGB mapping and for assessing the impact of Hurricane Ian disturbance in Florida through data fusion.

Terrestrial Laser Scanning (TLS) has been used on forest inventories as an alternative to conventional in-field measurements given how effectively it can collect data. However, not enough emphasis has been given on using these sensors to describe different pine forests in the South. We aimed to analyze the effectiveness of TLS in characterizing even- and uneven-aged forest stands. After the data collection with the TLS, we developed a framework based on open-source tools in R for computing individual tree-level metrics (e.g. diameter at breast height (DBH), height, and leaf area index (LAI)). The TLS system showed great potential in capturing the differences between forest stands across those two silvicultural systems. The diameter class distributions of the even- and uneven-aged stands followed bell-shaped and J-distributions, respectively. This study demonstrates TLS as a powerful asset for obtaining forest inventory metrics with a reduced need for field data collection.

Continuous monitoring and quantification of aboveground biomass (AGB) using in situ methodologies are limited by cost and time. UAV-lidar has been used as an efficient tool for estimating AGB, however, up to date, no study has attempted to estimate total AGB (TAGB) change detection using UAV-lidar in tropical savannas. This study aimed to estimate TAGB stock and changes in the Brazilian savanna (Cerrado) using UAV-lidar data and Support Vector Machine (SVM) model. We used four canopy-level derived metrics from the UAV-lidar data (COV, H99TH, HSKE, and HKUR) for modeling TAGB with an R2of 0.62, RMSE of 26.62 Mg/ha (46.5%), and bias -3.85 Mg/ha (6.73%), respectively. Our results showed an increase in the average TAGB of 5.07 mg/ha for the SCNPK site in 2021 when compared to 2019. The Kolmogorov-Smirnov (KS) test confirmed a statistical difference in the distribution of TAGB (p-value =< 0.5).

Lidar (light detection and ranging) has been used for mapping fuel loads in Longleaf Pine (Pinus palustris Mill.) forests ecosystems. However, there are sources of bias and uncertainty associated with estimating crown-bulk density (CBD) from either Airborne Laser Scanners (ALS) and Terrestrial Laser Scanners (TLS) data. Therefore, the aim of this study was to assess the utility of ALS and TLS systems and their combination (ALS+TLS) in predicting CBD in a longleaf pine forest ecosystem in Florida. In the field, tree attributes, such as tree height (HT), crown width (CW), crown base height (CBH) and diameter at breast height (DBH) in three plots of ~ 0.19 ha were measured and CBD (kg/m3) was calculated. Individual trees were detected from ALS, TLS and ALS+TLS, and lidar-derived crown-level metrics were computed for CBD modeling. The results show that CBD can be accurately predicted from ALS, TLS and ALS+TLS. However, the ALS + TLS improved CBD prediction accuracy only slightly. Given that ALS+TLS fusion is less practical and more expensive, our comparison suggests that either ALS or TLS measurements are still reasonable for CBD prediction and their usefulness is justified.

Forest attributes at the tree-level could be quickly and precisely acquired using terrestrial laser scanning (TLS). For a quantitative analysis of forest attributes, including timber volume and leaf area index, the leaf and wood part of the high resolution TLS-derived 3D point cloud should be separated. There are several strategies used for classifying these two components, but no studies so far have tested them in southern pine forest in the U.S. Herein, one longleaf pine tree was used to test three geometric feature-based (TLSeparation, lidUrb-graph and lidUrb-dbscan) wood/leaf classification algorithms from the TLS data. TLSeparation's overall accuracy is 74.65%, lidUrb-graph is 82.74%, and lidUrb-dbscan is 83.03%. The performance of a segment-wise approach (e.g. lidUrbAr-dbscan) is superior to the point-wise algorithm (e.g. TLSeparation and lidUrb-graph), but none of them achieves the same high precision as the results with broadleaf trees. Southern pine trees have not been well studied or treated using current approaches.

In the context of global climate change, the quantification of carbon stocks in forests is essential, mainly because forests play a key role in balancing global carbon cycles. Existing methodologies for measuring carbon stocks in forests are constrained by budgetary issues and time, making it difficult to deliver large scale full inventories in short periods of time. ALS (Airborne Laser Scanning) technologies have been used as an efficient and flexible alternative to estimate carbon stocks in forests, due to its accuracy and efficiency when compared to conventional methods. This study evaluates the use of LIDAR (Ligth Detection and Ranging) ALS to estimate the amount of carbon in aboveground biomass of Eucalyptus plantations. We have used a multiple linear regression model and a suite of 68 predictor variables derived from discrete-return LIDAR data to create the carbon stock model. Six variables related of height and intensity from LiDAR cloud points were selected to build the final model (R2=0.93, Pearson's correlation r= 0.97 and RMSE = 1.93 m3).