HighlightsWhat are the main findings?A Multi-Stage Feature Selection (MSFS) framework integrating Pearson correlation, mutual information, and Boruta was developed to optimize high-dimensional airborne LiDAR point cloud features for individual-tree DBH estimation.The proposed MSFS–XGBoost model significantly improved prediction accuracy, achieving an of 0.901 and an RMSE of 1.647 cm, outperforming DTR, RFR, and GBM models.What are the implications of the main findings?The proposed feature optimization strategy effectively reduces redundancy in LiDAR-derived features and enhances model stability for forest structural parameter estimation.The MSFS–XGBoost framework provides a reliable approach for accurate individual-tree DBH estimation and supports refined forest resource monitoring using airborne LiDAR data.Diameter at breast height (DBH) is a fundamental structural parameter for forest inventory and ecological analysis. However, field-based measurements (e.g., diameter tape surveys) are labor-intensive and inefficient for large-scale applications. Airborne light detection and ranging (LiDAR) provides an efficient alternative for individual-tree DBH estimation. Nevertheless, LiDAR-derived features—defined as statistical descriptors of point cloud structure and radiometric properties—are typically high-dimensional and redundant, which may degrade model performance. To address this issue, this study proposes an integrated framework combining Multi-Stage Feature Selection (MSFS) and Extreme Gradient Boosting (XGBoost) for DBH estimation. A total of 104 variables, including LiDAR-derived features (height, density, intensity, and canopy structure metrics) and structural parameters (tree height, crown diameter, and crown area), were used as predictors. The MSFS framework was applied to progressively reduce feature redundancy and identify an optimal subset, which was then used to train the XGBoost model. The results demonstrate that the MSFS–XGBoost model achieved the best performance, with a coefficient of determination (R2) of 0.901 and a root mean square error (RMSE) of 1.647 cm. Compared with models using the original feature set, R2 increased by 0.384 and RMSE decreased by 1.146 cm. These findings indicate that the proposed framework effectively improves DBH estimation accuracy and provides a reliable approach for individual-tree parameter estimation and large-scale forest resource monitoring using airborne LiDAR data.

Forest Aboveground Biomass Density (AGBD) estimation supports forest carbon accounting and informs carbon monitoring, reporting, and verification. Despite the demonstrated potential of airborne light detection and ranging (lidar) and satellite imagery, accurate AGBD estimation in disturbance-prone, mixed forests remains challenging. To better understand the applicability of these data in disturbance-prone forests, we sought to determine an optimal modeling framework for AGBD estimation. We utilized 70 PlanetScope (3 m), 34 airborne lidar, and 3 ancillary predictors (elevation, slope, and aspect) with field-estimated AGBD across 5 sites in the southeastern United States (US) with different kinds of wind disturbance (tornado, hurricane, straight-line wind): Bankhead, Mountain Longleaf, Oakmulgee, Weeks Bay, and Flagg Mountain. We evaluated: (1) five established variable selection methods; (a) all predictors, (b) top 5 predictors from Random Forest (RF), (c) top 10 predictors from RF, (d) Least Absolute Shrinkage and Selection Operator (lasso), and (e) Recursive Feature Elimination (RFE), and (2) compared 2 modeling algorithms; (a)RF and (b) Bayesian-based Gaussian Process Regression (GPR) for AGBD estimation. Results show that lasso and RF-based variable selection methods outperformed RFE, while GPR outperformed RF. Model accuracy (R 2 = 0.29–0.73; Root Mean Squared Error (RMSE) = 16.29–75.14 Mg/ha) was highest in the undisturbed (Bankhead) and lowest in the wind-disturbance-prone site (Oakmulgee). Findings demonstrate that AGBD estimation is more reliable in undisturbed landscapes, while such frameworks may be inadequate in disturbance-prone, dynamic landscapes. Our study offers optimal modeling frameworks for disturbance-prone mixed forests and advances the synergistic use of lidar and PlanetScope for AGBD estimation. • Built site-specific AGBD frameworks for disturbance-prone sites, fusing airborne lidar and PlanetScope (20 m). • Model accuracies are highest in undisturbed southeastern US forests, and lowest in disturbed forests. • RF-based and lasso feature selection outperformed RFE across sites.

Automatic tree detection in varying urban environments using airborne LiDAR data

Landslide areas and depths were investigated using a digital elevation model (DEM) of difference (DoD) analysis based on pre- and post-rainfall airborne laser survey data. Three regions, consisting predominantly of granite or granodiorite bedrock, were selected as study sites: (A) the Noborikawa River Basin, (B) the Serizawa and Iriyamazawa Basins, and (C) the Uchikawa River Basin. Site C features gentler slopes than A and B. Although, assuming scale-invariant geometry, landslide depth is theoretically expected to depend on area, this was not observed, especially for shallow landslides. However, in all three regions, the landslide area was not a dominant determining factor for the depth of particularly shallow landslides. At site C, the landslide depth did not correlate with area, while at sites A and B it showed only a weak dependence. In both regions, gentle slopes showed a decreasing lower envelope of the landslide depth with increasing slope angle, while steep slopes showed a decreasing upper envelope. The results align with infinite slope theory, suggesting that landslide depth is controlled by mechanical equilibrium and maximum soil depth on steep slopes, rather than landslide area.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-46714-4.

ABSTRACT Accurate alignment between overlapping lidar swaths and ground checkpoints is essential for generating seamless digital elevation models (DEMs), particularly in dynamic coastal environments. This study assessed the geometric integrity of airborne lidar datasets acquired by researchers from the Bureau of Economic Geology (BEG) along the Texas Gulf Coast in 2024 through combined vertical and horizontal accuracy evaluations. The airborne lidar system calibration was validated using 789 GNSS-derived ground control points (GCPs) along the coast, yielding an absolute mean elevation difference of 0.012 m and root-mean-square error (RMSE) of 0.025 m. High-resolution DEMs were validated against the United States Geological Survey (USGS) 3-Dimensional Elevation Program (3DEP) products and legacy datasets along the coast. The results revealed minor systematic differences, primarily influenced by surface-type variability, ground modelling approaches, and variations in GNSS base station configuration and post-processing methods. Additional analyses and practices demonstrated that lidar swath misalignments and lever arm offset inconsistencies were effectively reduced through algorithm corrections, while geoid model differences introduced minor but measurable offsets. To isolate geomorphic influences, validation transects were divided into inland and beach segments. Inland sites showed strong agreement (mean R2 = 0.96), with mean elevation differences <0.01 m and RMSE < 0.08 m. In contrast, beach segments exhibited greater variability (R2 = 0.87), with absolute mean differences of 0.12 m and RMSE up to 0.44 m, demonstrating variability outside natural error thresholds, attributable to detections of geomorphic change which is characteristic of highly morphodynamic near-shore environments like those surveyed along the Texas Gulf Coast. Overall, the findings confirm that BEG and current USGS DEMs exhibit high geometric fidelity and are suitable for coastal applications, including shoreline change detection, flood risk assessment, and habitat monitoring. In addition, the study emphasizes the importance of achieving sub-decimetre accuracy in beach and dune environments, highlighting the significance of standardized calibration, alignment, and processing workflows to ensure reliable long-term coastal and environmental monitoring.

Abstract Peatlands play a critical role in the global carbon (C)‐climate cycle, acting as vast long‐term stores of disproportionately large quantities of C relative to their land area. In recent decades, climate‐driven shifts in fire regimes and peatland hydrophysical properties have occurred across Canada's boreal regions, increasing concerns about the vulnerability of peatland C to combustion losses. Understanding the magnitude and vulnerability of C lost during wildland fires in peatlands is therefore essential but remains highly uncertain. This study was conducted in the Athabasca Oil Sands Region of Alberta's Boreal Plains. C losses from peatlands during the 2016 Horse River Wildfire were estimated based on field‐collected soil C data and pre‐ and post‐fire airborne LiDAR data. C Losses were quantified across peatland types and ecotones and separated into above‐ and below‐ground combustion. Soil C losses were nearly an order of magnitude greater than vegetation C losses (2.11 ± 5.09 kg C m −2 vs. 0.38 ± 0.32 kg C m −2 , respectively). Bog ecotones were zones of significant soil C loss, with average losses of 16.5 kg C m −2 . LiDAR‐derived burned area and C losses were compared with the spectral burn severity index, dNBR. A binary burned/unburned classification showed strong agreement in bogs (88%) but poor agreement in swamps (48%). Vegetation C loss correlated moderately well with dNBR strength, whereas the relationship between soil C loss and dNBR was very weak. Comparisons between LiDAR‐derived soil C losses with estimates of C loss based on the fire disturbance module of the national C loss model, the Canadian Model for Peatlands (CaMP), indicated that C losses from bogs were greater than expected, particularly when ecotones were included, while fens and swamp C losses were on the low end of model expectations.

Mapping three decades of forest structural changes in Japan using Landsat time series and airborne LiDAR data

Sustainable forest management requires current, territory-wide data, which is difficult to obtain in vast regions like Quebec, Canada. To complement ground inventories and photo-interpretation, the province developed an airborne laser scanning (ALS)-based model that performs well in coniferous stands, but its accuracy in hardwood stands remains untested. This study aims to evaluate the accuracy of the ALS-based prediction of stand basal area and then test new approaches to increase its performance. Airborne LiDAR data from 2011 to 2020 and 12,506 validation plots from sample plots were used. The ALS model accuracy was initially compared across the stand types, revealing lower accuracy in shade-tolerant deciduous stands. Three inputs were found to increase prediction accuracy: proportion of each species basal area in the stand, geographical coordinates, and meteorological data associated with location. Parametric and auto machine learning (AutoML) methods were employed using those inputs to improve accuracy, with AutoML achieving the highest improvement with initial R2 of 0.27, 0.47 and 0.54 and after correction R2 of 0.31, 0.56 and 0.67, respectively, for shade-tolerant deciduous, shade-intolerant deciduous, and coniferous stand. Even with the advancements made, further improvements will be necessary to consider using an ALS-based model for shade-tolerant deciduous species.

Accurate alfalfa biomass prediction is crucial for pasture management and sustainable livestock production. However, traditional methods often perform poorly under complex field conditions. To address the limited prediction accuracy of traditional methods under complex planting environments, this study proposes an alfalfa biomass prediction method combining multispectral and LiDAR data with ensemble learning model. Based on the multispectral images acquired by unmanned aerial vehicle (UAV) and airborne LiDAR data, the spectral features, three-dimensional structural features, and their interaction features are systematically extracted at the quadrat scale, and a high-quality modeling dataset is constructed by feature selection. Secondly, an ensemble model for alfalfa biomass prediction was constructed, which was composed of random forest, extra trees, and histogram gradient boosting. After model training, the coefficient of determination (R2) of the integrated model on the test set reached 0.813, and the root mean square error (RMSE) and mean absolute error (MAE) were 0.178 kg m-2 and 0.146 kg m-2, which were significantly better than those of similar single models. Under feature combinations, the fusion model was better than that of spectral indices only (R2 = 0.773) and LiDAR traits only (R2 = 0.576), and the model achieved the highest accuracy from bud emergence to early flowering (R2 = 0.917). The overall prediction error of the model was approximately normal distribution, and the absolute error of more than 65% of the samples was less than 0.2. However, there was still a trend of underestimation in the high biomass interval. This research showed that the multimodal data fusion and ensemble learning method could achieve high-precision prediction of alfalfa biomass, which provided reliable technical support for pasture resources monitoring and precision agriculture management.

The role of tidal range and seawater pollution in shaping mangrove biomass and carbon stocks.

Abstract Background Surface and ladder fuels play a significant role in controlling fire behavior, and their estimation is critical for fire modeling and management. Although airborne laser scanning (ALS) provides cost-effective, spatially explicit data on forest 3D structure, its utility for surface fuel estimation remains uncertain due to canopy occlusion and the presence of tree trunk points. We assessed the impact of tree trunk point filtering (TPF) on model performance for estimating surface fuel loads in strata within a vertical gradient of 0.0–2.0 m, which includes litter, herbaceous, and shrub layers. We used high-density ALS data (~ 2500 points m −2 ) from boreo-nemoral mixed forests in southern Sweden. We compared the performance of 438 LiDAR (lidar) metrics in characterizing surface fuels using parametric (linear and non-linear) and nonparametric (random forest — RF) regressions. Results There was no significant impact of TPF when comparing lidar-derived metric distributions and model performance under filter types, although a minor improvement was observed in the 0.5–2.0-m stratum. The performance of surface fuel strata modeling was the highest for the litter layer depth ( R 2 = 0.39) and moderate for the herbaceous layer and branch biomass ( R 2 = 0.26–0.28). The linear regression model consistently outperformed the RF model and showed slightly better performance than the nonlinear regression. We obtained a negligible positive impact of TPF ( ΔR 2 = 0.02) on predicting the litter layer depth utilizing the parametric regression approaches. Intensity-based metrics calculated using a minimum 5-m buffer radius were instrumental in modeling fuel layers within the 0.0–0.5-m stratum. Conclusions Removing tree trunk points did not affect the representation of surface fuels in airborne lidar data. We suggest, however, that the correct classification of ground and no-ground points and detection of objects such as boulders and deadwood can have a major effect on the adequate prediction of surface fuels.

Abstract. In the past decade, several fatal landslides have impacted Southeast Alaska, highlighting the need to advance our understanding of regional geomorphic and atmospheric controls on triggering events and runout behaviour. A large and long runout landslide on Wrangell Island, with area in the top 0.5 % of &gt;14 760 slides mapped in the Tongass National Forest, initiated during an atmospheric river event in November 2023 and travelled &gt;1 km downslope, causing six fatalities. We used field observations, sequential airborne lidar, geotechnical analyses, and climate data to characterize the geomorphic, hydrologic, and atmospheric conditions contributing to the landslide. Rainfall intensities recorded at the Wrangell airport were modest (∼ 1 year recurrence interval), but rapid snowmelt and drainage from a ridgetop wetland may have contributed to rapid saturation of the landslide. Although strong winds were recorded, we did not observe extensive windthrow, which may downgrade its contribution to slope failure. The landslide mobilized a steep, thick (&gt;4 m) wedge of colluvium that accumulated below a resistant bedrock ledge and entrained additional colluvial deposits as it travelled downslope across cliff-bench topography. The substantial entrainment resulted in an unusually large width, extensive runout, and low depositional slope as the landslide terminated in the coastal environment. Our results suggest that the sequencing of rain- and snow-dominated storms, geologic controls on post-glacial colluvium production and accumulation, and ridgetop hydrology contributed to landslide initiation and runout. Advances in post-glacial landscape evolution models that include colluvium production, frequent lidar acquisition, and additional climate data are needed to inform regional landslide hazard assessment.

Abstract Traditional nadir altimeters struggle with coastal water surface elevation (WSE) measurement and fine‐scale river‐estuary interactions, due to land‐water signal interference and their wide inter‐track spacing. The wide‐swath Surface Water and Ocean Topography (SWOT) mission, using a new Ka‐band radar interferometer, aims to address these issues by delivering 2D WSE measurements with unprecedented spatial resolution, accuracy, and precision. However, the mission's effectiveness in coastal WSE retrieval and its error characteristics remain unverified. This study leverages gauge and airborne LiDAR data to validate SWOT's WSE in the Bristol Channel and Severn Estuary. Assuming error‐free in situ data, SWOT ocean products exhibit a standard deviation of difference (STD) of 13 cm within a 3 km radius of tide gauges. Compared to LiDAR, SWOT's PIXC measurements have STD of 37 cm, improving to 14 cm over 100 m grids and 9 cm over 1 km 2 areas. This meets the SWOT science requirement of 10 cm STD at 1 km 2 scale and extends satellite‐based WSE monitoring into complex coastal environments.

HighlightsWhat are the main findings?A framework comprising ground point classification, building Region of Interest filtering, and refined extraction is proposed based on airborne LiDAR data for building classification purposes in complex rural scenes.Ground points form the foundation for building classifications. Region of Interest filtering based on geometric features further confirms building scopes, then a set of morphological features primarily based on local dimensionality models enables precise building classification.What are the implications of the main findings?The proposed framework employs a spatial hierarchical strategy to extract building data, thereby avoiding the substantial redundant computations caused by traversing the entire point cloud. By precisely capturing local features of point clouds within an optimal neighborhood range, it achieves high-precision building classification.This work provides a comprehensive approach for point cloud recognition of low-rise structures in rural areas, particularly suited for regions where vegetation and buildings are of similar height and exhibit an interlocking pattern. It demonstrates significant potential for use with other intelligent building classification methods.Research on extracting building from airborne point clouds is abundant, yet discussions regarding scenarios where vegetation and building structures are closely intertwined with similar height in rural areas remain relatively scarce. This thesis adopts a region representative of typical rural building features in China as an experimental site to conduct research on building classification procedures from airborne point clouds. Firstly, the multi-level grid size is dynamically determined through slope analysis to creatively segment and recognize terrain type, then differentiated filtering parameters are applied to various terrains to fully extract ground points, providing a ground reference for building classification. Secondly, the selection of building Region of Interest is conducted by multiple geometric feature differences between building and other objects based on watershed segmentation results, which eliminates interference from non-building points, significantly reducing redundant and unnecessary mathematical computation. Finally, refined building classification is achieved based on multiple morphological differences between buildings and other objects. The experimental results show that the precision, recall, and F1 of both datasets exceeded 93.37%, 97.05%, and 95.17%, respectively. The average precision, recall, and F1 reached 94.02%, 97.20%, and 95.58%, respectively. This method demonstrates successful building classification in rural areas, showing strong adaptability and practicality for the extraction of various building data.

Abstract. Snowpack dynamics play a key role in controlling hydrological and ecological processes at various scales, but snow monitoring remains challenging. Data assimilation techniques are emerging as promising tools to improve uncertain snowpack simulations by fusing state-of-the-art numerical models with information rich, but noisy observations. However, the occlusion of the ground below the forest canopy limits the retrieval of snowpack information from remote sensing tools. Remote sensing observations in these environments are spatially incomplete, impeding the implementation of fully distributed data assimilation techniques. Here we propose different experiments to propagate the information obtained in forest clearings, where it is possible to retrieve observations, towards the sub-canopy, where the point of view of remote sensors is occluded. The experiments were conducted in forests within Sagehen Creek watershed (California, USA), by updating simulations conducted with the Flexible Snow Model (FSM2) using airborne lidar snow data using the Multiple Snow data Assimilation system (MuSA). The successful experiments improved the reference simulations significantly both in terms of validation metrics (correlation coefficient from R=0.1 to R=0.8 on average) and spatial patterns. Data assimilation configurations using geographical distances and space of topographical dimensions, improved the reference run. However, those creating a space of synthetic coordinates by combining the spatiotemporal data assimilation with a principal components analysis did not show any improvement, even degrading some validation metrics. Future data assimilation initiatives would benefit from building specific localization functions that are able to model the spatial snowpack relationships at different resolutions.

Large old trees are widely recognized as ecologically important across forest landscapes and concern regarding the decline of these trees is well documented because of their role in maintaining biodiversity for a broad range of organisms. In response to a growing need to inventory such trees, we developed and present the methodology to map and quantify the occurrence of large trees based on height and dbh thresholds using airborne LiDAR data and associated canopy height models. The innovative, succinct, and flexible solution we offer is based on the integration and augmentation of several existing packages within the open‐source statistical software R. We use local tree‐height and crown diameter data to calibrate an algorithm to count individual trees above specified height thresholds, including supercanopy trees. To satisfy large‐tree definitions based on dbh, we used individual‐tree height and dbh data available from existing forest inventory plots to define height–dbh curves for dominant forest community types, which then allowed height thresholds to be used as a surrogate for specified dbh thresholds. We illustrate the use of these methods to efficiently map and quantify large tree distributions within 8 forest communities across a study area consisting of 1.65 million ha of productive, industrially managed forest in New Brunswick, Canada. Spatial maps are presented, along with large‐tree frequency statistics for specific communities, according to the definitions outlined in New Brunswick's provincial forest management guidelines. In excess of 37 million large trees are estimated to be broadly distributed across the study area. The methods developed identify patterns in the distribution of large trees across extensive areas (e.g., in millions of hectares) as one metric for maintenance of biodiversity at the landscape level. The methodology may be readily adapted to alternative forest‐specific definitions of large trees based on tree height or dbh.

Persistent Scatterer Interferometry (PSI) is widely used for surface deformation monitoring, but its accuracy is often limited by atmospheric path delays, orbit errors, and timing inconsistencies. To address these issues, the European Space Agency (ESA) released the Sentinel-1 Extended Timing Annotation Dataset (ETAD), providing refined timing corrections that account for instrument, atmospheric, geophysical, and Doppler-related effects. This study introduces PSI+ETAD: a sensor-native PSI processing framework that fully exploits ETAD’s range- and azimuth-direction timing annotations to jointly enhance interferometric phase quality, deformation time-series accuracy, and PS 3D localization precision — thereby delivering simultaneous, cost-effective improvements across all three pillars of operational PSI. Experimental results show that ETAD integration substantially improves phase quality, with reductions of 60.77% in the Sum of Phase Differences (SPD) and 40.3% in the Phase Standard Deviation (PSD). In deformation velocity estimation, PSI+ETAD achieves a standard deviation of 5.26 mm/year, compared with 18.64 mm/year for PSI (uncorrected), 2.82 mm/year for PSI+STF (PSI with spatio-temporal filtering), and 6.34 mm/year for PSI+GACOS (PSI with GACOS-based tropospheric delay correction), providing a better balance between noise suppression and signal preservation. In time-series analysis, ETAD corrections lower the Standard Deviation of Displacements (SDD) to 3.65 and the Spatio-Temporal Consistency (STC) to 2.23. Validation against GNSS confirms that PSI+ETAD yields the closest agreement with ground-truth deformation, with the lowest errors (MAE of 3.18 mm and RMSE of 3.91 mm). For geolocation, mean offsets of 5.566 m in range, 1.736 m in azimuth, and 0.974 m in cross-range are corrected, with improvements validated against airborne LiDAR data. These findings demonstrate that integrating ETAD into standard PSI workflows provides a cost-effective and operationally feasible correction strategy, significantly enhancing phase stability, deformation accuracy, and geolocation precision for large-scale InSAR monitoring in urban and tectonically active regions.

Abstract Carbon monitoring is needed to track the exchange of carbon between the land and atmosphere. However, comprehensive monitoring of terrestrial ecosystems is challenging, partially because of the fine-scale spatial heterogeneity of land cover and vegetation structure. While forest ecosystems are relatively well studied, knowledge of trees outside forests (TOF) is limited, even though TOF contribute substantially to many ecosystem services. In this study, we leveraged advances in remote sensing and modeling to quantify TOF and their contribution to the 2011-2023 Maryland state carbon budget. Specifically, the Ecosystem Demography model was initialized with 1 m airborne optical and LiDAR data. These datasets were combined with 30 m National Land Cover Database maps to identify and monitor TOF statewide. In the base year, TOF contributed to 22.86% (2,795.59 km2) of the total tree cover, 14.70% (15.98 Tg C) of the total aboveground live tree carbon stocks, and 22.56% (0.39 Tg C yr-1) of the corresponding carbon fluxes in the state, with variation by land cover class. TOF on Developed and Planted/Cultivated lands contributed approximately 90% of the total TOF tree cover and carbon stocks, as well as over 96% of the total TOF carbon fluxes. Most TOF occurred as polygons smaller than 1 km2 and were within 600 meters distance from forests. From 2011-2023, persistent TOF carbon stocks were estimated to increase by 5.15 Tg C, representing 23.71% of the state’s total increase in tree carbon stocks over the interval, with substantial interannual variability in carbon fluxes. These results suggest TOF are an important contribution to statewide tree cover and produce carbon stocks and fluxes that should be highlighted in carbon budgets for effective monitoring and assessment. Moving forward, efforts are needed to expand the availability and application of high-resolution remote sensing data and models needed for TOF quantification in carbon budgets.

Annual seasonal snow cover is found in high latitude and high-altitude regions of the globe, with snowpack depth capable of being determined with automated devices, measuring sticks, or other similar instruments. Direct upscaling the sparse, field measured snow depth to regional scales (∼100 km 2 ) using remote sensing observations leads to large uncertainties due to the large difference in spatial scales. Linking field measurements with airborne remote sensing data to generate an intermediary snow product offers an opportunity to upscale the limited measurements to regional scales using large scale satellite observations. Here we developed an approach to generate an intermediary airborne snow depth map product at the local scale (1 km 2 ) first using end of season snow depth field measurements along limited transects between 2016 to 2022 and airborne hyperspectral and LiDAR data, and then further upscaled the airborne snow depth product to a regional scale (100 km 2 ) using satellite observations. Multiple regression approaches were applied in the upscaling. Snow depth patterns were identified in areas containing a presence or absence of thick canopy cover along with differences in snow depth between deciduous and evergreen forests. By generating airborne-based products as intermediate layers for regional scale upscaling, it allowed for reduced uncertainties when upscaling field measurements directly to a regional scale. This approach may be expanded beyond snow depth regional scale modeling to estimate snow water equivalent and snow density at regional scales by first generating intermediary local scale products from field data as inputs. • Snow depth data were upscaled to local (1 km 2 ) and then regional (100 km 2 ) scales • The upscaling procedure was achieved with an object-based machine learning technique • Similar snow depth patterns were observed over multiple spatial and temporal scales • Forest snow depth was markedly lower with a dense than absent canopy cover

The accurate three-dimensional (3D) distribution of plant area density (PAD) within forests is crucial for understanding canopy structure and provides essential scene inputs for 3D Radiative Transfer Models (RTMs) to facilitate remote sensing interpretation. However, current lidar-based voxelization methods that estimate detailed PAD distributions often cover limited areas, constraining their applications in conducting broad forest studies and interpreting Earth Observation Satellite (EOS) data of various scales and resolutions. To address this, we developed the Large-Scale Path Volume Leaf Area Density (LS-PVlad), a novel forest 3D reconstruction workflow capable of producing extensive high-resolution 3D voxelized forest scenes (up to 100 km 2 with ≤2 m voxel size) from worldwide open-access airborne lidar scanning (ALS) data. By applying LS-PVlad to the ALS data acquired during the extensive NASA Goddard's LiDAR, Hyperspectral & Thermal Imager (G-LiHT) campaigns, we developed the first release of FoScenes—a high-fidelity PAD product comprising 40 seamless scenes from 28 diverse forest sites, with individual area ranging from ∼50 to ∼11,000 ha. The leaf area estimates of LS-PVlad have been validated by two-year field-measured leaf area index (LAI) from litter collection (best RMSE = 0.35 m 2 /m 2 ) and digital hemispherical photography (DHP) images (RMSE = 0.46 m 2 /m 2 ) across multiple plots at a deciduous forest site. Additionally, a broad comparison between FoScenes and MODIS plant/leaf area index product demonstrates high consistency (R 2 = 0.70, RMSE = 0.86 m 2 /m 2 ). By providing multi-dimensional forest characterizations, FoScenes enables temporal insights into structure dynamics. Its integration with the discrete anisotropic radiative transfer (DART) model underscores the potential of FoScenes for extensive 3D RTM applications at various scales. • We developed a large-scale ALS-data-driven 3D forest reconstruction workflow. • FoScenes product consists of 40 various forest scenes derived from NASA G-LiHT data. • The estimated leaf/plant area index strongly aligns with field data and EOS products. • FoScenes captures temporal structure variation by multi-dimensional characterization. • FoScenes can be integrated into DART for realistic simulations at varied scales.