Multiple studies have analyzed the error budget of biomass predictions from airborne laser scanning (ALS) data utilising the area-based approach (ABA). While some error sources, such as high-incidence scan angles, can be excluded from analyses, the main error sources of (a) geolocation errors, (b) quality of the reference data, and (c) residual modelling errors remain. In this work, we aim to focus on the contribution of inherent uncertainty of the trees themselves, e.g., caused by wind during the ALS data acquisition. At Petawawa Research Forest in Ontario, Canada, we analysed overlapping pairs of ALS flight lines and trained a Random Forest to estimate biomass from 223 forest inventory plots, both on the merged dataset and separated by flight strips using a set of state-of-the-art LiDAR metrics as independent variables. Quantifying inference performance on withheld validation data, we compared single-strip predictions with estimates from the merged dataset to examine how much of model error is due to the model’s capacity and reference data quality, and how much is caused by different representations from the separated strips. Our results show that after excluding problematic scan angles, the RMSE differences between merged and separated strips amount to approximately 50-80% of the predictor RMSE. These errors tend to be larger for taller trees, especially when close to clearings, where effects of wind on the point cloud metrics are larger. Furthermore, we analysed multiple data acquisitions acquired from several sensors. Overall, increased footprint size and increased point density corresponded to a decrease in biomass RMSE. Considering current developments towards small-footprint (UAV-based) laser scanning, uncertainty caused by movement of tree crowns may be an increasingly limiting factor for accurate biomass- and other forest-related metric extraction from laser scanning. We therefore strongly suggest that this uncertainty should be quantified when analysing laser scanning data of forests.

Southern U.S. forests are among the most productive on Earth and represent a significant portion of the U.S. terrestrial carbon sink. Commercially, these forests account for over 60% of U.S. timber production. However, they are frequently affected by hurricanes, which can result in significant damage to forest structure and associated ecosystem functions and services. This damage includes the loss and disruption of timber supply, an increased risk of wildfire, and a reduction in recreational opportunities. As hurricanes are projected to increase in frequency and magnitude in the Southern U.S., accurately and timely quantification of damage to these forests is essential for developing effective protection and restoration measures and understanding the dynamics of forest recovery. This study aimed to develop a data fusion framework based on NASA’s GEDI (Global Ecosystem Dynamics Investigation) and Sentinel for mapping aboveground biomass density (AGBD, Mg/ha) that can be further used to evaluate damage severity and recovery in forested ecosystems impacted by Hurricane Ian in Florida. We used GEDI level 4A and Sentinel level 2a and 1c data to calibrate a Random Forest (RF) model for predicting and mapping aboveground biomass density (AGBD) for four months prior to the Hurricane Ian disturbance. The study area covered regions affected by the hurricane. The results showed that the combination of the spectral/radar bands and vegetation indices derived from Sentinel 2a and 1c data explained more than 79% of the variation in AGBD. The absolute (and relative) root mean square error (RMSE) were 29.17 Mg/ha (64.27%) and Bias was -1.14 Mg/ha (2.66%). This research highlights a novel approach for improving aboveground biomass (AGB) mapping by fusing GEDI and Sentinel data streams. This advancement in data fusion provides an opportunity for more accurate and timely assessments and targeted management of natural disaster impacts such as hurricanes on forest ecosystems.

The Global Ecosystem Dynamics Investigation (GEDI) instrument has been designed to monitor forest ecosystems and improve the understanding of forest carbon dynamics and is collecting billions of forest canopy measurements globally. In this study, we evaluate whether GEDI observations can be used in conjunction with Landsat-8 and Sentinel-2B optical data for the high resolution, spatially explicit estimation of the biomass consumed by forest fires, one of the key needs of the fire ecology and fire emissions community, unmet by the current generation of remotely sensed burned area products.

The sampling configuration of the GEDI instrument poses unique challenges for mapping changes compared to synoptic instruments, such as most optical sensors and Airborne LiDAR system. This poses a challenge for monitoring biomass consumption due to fire, as pre-fire and post-fire GEDI footprints are only sporadically coincident.

Time series of atmospherically corrected Landsat-8 and Sentinel 2-B data are first used to map the extent of burned area perimeters, as well as to predict a pixel-level estimate of the combustion completeness and fraction of area burned (f*cc), using the approach proposed by Roy et al. (2019); Level 4 footprint-level GEDI biomass estimates (Duncanson et al., 2022), acquired pre-fire and post-fire are then spatially interpolated for the extent of the fire using the 30-m f*cc map as auxiliary variable.

Three fires in the Western United States, which occurred within the timespan of the GEDI mission, and where pre-fire and post-fire airborne LiDAR data are available, as well as pre-fire and post-fire field measurements, are used for the accuracy assessment of the results.

The assessment of tree aboveground biomass (AGB), and consequently carbon stock, is a vital component of carbon budgeting and monitoring. This is traditionally done using allometric equations which estimate biomass from metrics such as DBH and height. However, recent studies have shown that current allometries can lead to significant errors in the biomass estimates. The training datasets on which these allometric models are built are often biased towards smaller trees, due to the logistical issues that are associated with the destructive sampling of large trees. Because of this limitation, allometries are often extrapolated beyond the intended tree sizes, assuming that the relation between tree size and biomass remains constant. However, this assumption does not hold, especially for large trees. Using terrestrial laser scanning (TLS), more evenly distributed and complete training datasets can be collected as it enables the non-destructive estimation of biomass for trees of all sizes.

In this work, we present a case study on dynamic allometric models using the sycamores (Acer pseudoplatanus) of Wytham Woods. Leaf-off TLS data was collected at this site in 2015, individual trees were segmented and their AGB was estimated. Sycamores were the dominant species in this forest and a wide size range of sycamores is included in the dataset. We will explore several methods to construct sycamore specific dynamic allometric models, which are able to cover the full size range of the species. The relation between size and biomass in these dynamic models is not constant, but varies based on tree size. We will test how these dynamic models perform and how they compare to traditional allometries. Furthermore, we will explore how these new models, based on the Wytham Woods dataset, perform on sycamore data from other sites in the UK and Western Europe.

The Global Ecosystem Dynamics Investigation (GEDI) is a sampling instrument on the International Space Station and does not provide data on a regular, systematic basis. Reducing Emissions from Deforestation and Degradation and enhancement of carbon stocks (REDD+) projects require forest above ground biomass (AGB) inventories to quantify avoided carbon emissions achieved by conserving forest biomass. Although there is high confidence that GEDI can retrieve measurements that allow estimation of AGB at scale, less is known about how well its operational deployment performs for measurement of AGB to support REDD+ projects. This includes an understanding of the appropriate period required to collect sufficient GEDI observations for reliable forest AGB assessment. We describe the first study to examine the amount of GEDI data needed to characterize tropical forest AGB at REDD+ project scale. A global 30m percent tree cover product, updated with contemporary mapped forest cover loss, was used to map the intact forest across the Mai Ndombe province in the Democratic Republic of the Congo, and to help select 50 × 50 km test sites each with >80% forest cover and forest good quality AGB GEDI footprint data distributed across them. The overall mean GEDI AGB (OMGA) was derived from all the good quality forest GEDI footprint AGB values acquired over several years at each site. The expected minimum number of GEDI orbits required to characterize the OMGA to within ±5%, ±10%, ±20% was derived by considering different combinations of randomly selected GEDI orbits. These numbers were converted to days using scaling factors based on analysis of the number of days required to obtain GEDI orbits at 50 × 50 km scale. The results are presented with implications for the use of GEDI for AGB monitoring in support of REDD+ activities and are contextualized in terms of IPCC AGB uncertainty reporting recommendations.

More of the Amazon rainforest is disturbed annually than completely deforested, but the impact of disturbances on the carbon cycle remains poorly understood. Recent advances using optical (LandSat) and radar (Sentinel-1) remote sensing have improved detection of small-scale disturbances, but quantifying changes in forest structure and biomass associated with these disturbances has proven challenging.

The Global Ecosystem Dynamics Investigation (GEDI) spaceborne LiDAR provides an opportunity to address this problem. GEDI captured billions of measurements of forest height, leaf area, and understory structure within ~25-meter diameter footprints scattered across the tropics. Though the instrument had no guaranteed repeat cycle, it sometimes sampled nearby locations twice; some of these spatially coincident footprints happened to measure forest structure before and after a detected disturbance.

In this study, we developed an efficient open-source pipeline for identifying spatially coincident footprints and used it to find over 7,100 footprint pairs with intervening disturbance events across the Amazon basin. We also identified ~34,000 spatially coincident footprint pairs that lacked an intervening disturbance but came from regions with similar disturbance threat; these provided a control dataset to evaluate the effectiveness of estimating forest structure and biomass changes with coincident footprints.

Analysis of this Amazon-wide dataset demonstrated that GEDI could detect canopy and biomass losses in non-stand-replacing disturbances as small as 0.09 ha. GEDI’s unique three-dimensional view of forest structure allowed us to identify effects of different intensities of fire disturbance, including areas where the upper canopy retained most of its height but the understory suffered substantial foliage losses. Finally, we identified temporal trends in biomass loss and recovery following disturbance and showed that certain satellite-derived intensity metrics are correlated with increasing biomass loss. This work represents an important step towards the development of a pan-tropical, spatially explicit estimate of carbon losses and structural changes due to forest disturbance.