Bark beetle outbreaks pose a significant threat to European forest ecosystems, with early detection of their green attack phase being critical for implementing timely countermeasures. While traditional remote sensing approaches often focus on proxies representing vegetation structure, we aim to introduce a novel approach by emphasizing physiological proxies that respond near-instantaneously to stress. By bridging scales from leaf to tree, landscape, and satellite levels, the BeatTheBeetle project aims to develop a comprehensive framework for detecting early signs of bark beetle infestation.

In this contribution, we will present results from an intensive field campaign conducted at spruce trees in the Pitztal valley (Tyrol, western Austria) to characterize leaf-level physiological responses. Measurements included leaf gas exchange, active and passive chlorophyll fluorescence, and visible and near-infrared reflectance. Preliminary results present a comparison between leaf gas exchange data, leaf-level imaging spectroscopy, and initial observations from an uncrewed aerial vehicle (UAV) flight.

Our findings highlight the potential of physiological proxies in advancing remote sensing techniques for early bark beetle detection. They represent an important step towards integrating multi-scale physiological indicators into remote sensing workflows and pave the way for further work exploring the scalability of these proxies across other platforms, ranging from UAVs to the satellite scale, to enable large-scale forest health monitoring.

How to cite: Hänchen, L., Zähle, L., Wachter, H., Hammerle, A., Bremer, M., Czifersky, A., Geisler, T., Mössler, S., Spreitzer, S., Rutzinger, M., and Wohlfahrt, G.: Detecting Bark Beetle Infestation at the Green Attack Phase Using Multi-Scale Physiological Indicators, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8149, https://doi.org/10.5194/egusphere-egu26-8149, 2026.

EGU26-8149

Download Close

Dead trees are key indicators of biodiversity and forest health, recognized by both the UN Convention on Biological Diversity and the EU Biodiversity Strategy. Member states are required to monitor these indicators regularly, making accurate tree mortality detection essential. We utilized remote sensing data combined with AI-driven image analysis to improve dead tree detection and forest mortality mapping. In remote regions such as Lapland, where rapid changes caused by snow damage, windthrow, drought, or pest outbreaks occur, traditional inventories are often time consuming and difficult.

Our study area of 208 ha is located next to Pallasjärvi in Northern Lapland, Finland. The area is dominated by Norway Spruce (Picea abies), but Scots Pine (Pinus sylvestris), Silver Birch (Betula pendula) and Downy birch (Betula pubescens) are also present. In the training dataset we recorded 4380 tree segments, including 142 dead trees, and delineated their canopies using visual interpretation on an Altum multispectral drone imagery. We evaluated the integration of LiDAR-derived Canopy Height Models (CHM) with multispectral imagery for classifying living and dead standing trees. The training dataset consisted of 411 image tiles (256x256 pixels) with a pixel size of 5.2 cm, captured from drone in 2023 July using an Altum sensor. The CHM was interpolated with a 1m resolution from the low-pulse density Lidar data that is available via National Land Survey of Finland. The models’ performance was assessed using 10-fold cross-validation.

We applied pixel-level semantic segmentation using U-Net deep learning architecture to classify each pixel of the images into living tree, dead standing tree, and background (e.g. fallen dead trees and non-trees) pixels. The Basic model using only multispectral imagery achieved F1-scores of 0.33–0.44 for dead trees in different areas and up to F1-score of 0.83 for living trees. Incorporating the CHM improved dead tree detection by over 56%, providing F1-scores of 0.56–0.71 and 0.96 for living trees. Visual assessment confirmed that incorporating CHM improved crown delineation by producing more precise crown edges and enhanced the classification of standing deadwood by reducing misclassification of fallen deadwood. The resulting three-class map provides valuable data for qualitative measurements of deadwood, including total land area covered and the percentage of dead crown area.

Our current workflow relies on drone imagery and LiDAR data. However, future scalability through satellite data could enable large-scale, cost-effective monitoring beyond the Arctic region. By incorporating readily available canopy height models (CHM) as an additional input, we enhance tree classification accuracy and improve the detection of standing and fallen deadwood, furthermore, multiclass classification enables more precise tracking of tree mortality than binary classifications done by most studies, as classification could be extended by e.g. dying trees in future. This qualitative measurement supports forest conservation and biodiversity monitoring efforts and could provide a remote sensing-based estimate of dead wood volume to forest inventory.

How to cite: Waga, K., Rana, P., Kukkonen1, M., Kumpula, T., and Kuzmin, A.: Deep learning-based tree mortality detection using drone imagery and canopy height models in Northern Lapland , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19863, https://doi.org/10.5194/egusphere-egu26-19863, 2026.

EGU26-19863

Download Close

Sun‑induced chlorophyll‑a fluorescence (SIF) is used as a proxy for photosynthetic activity measurements from space‑ and airborne platforms. However, its signal is affected not only by the underlying leaf biochemistry but also by geometric factors such as canopy architecture and leaf orientation, which can lead to misinterpretations of SIF signals. To evaluate how these interacting influences affect SIF signals and subsequent drought‑stress detection, we combined SIF measurements with high‑resolution structural monitoring and UAV‑based thermal imaging during a four‑week drought experiment on seedlings of two ecophysiologically contrasting tree species, Pseudotsuga menziesii (Douglas‑fir) and Fagus sylvatica (European beech). Soil moisture was recorded continuously with SMT100 sensors, while top‑of‑canopy SIF spectra were captured under clear sky conditions on four days using a spectroradiometric setup (FLOX). Leaf‑level chlorophyll fluorescence (effective quantum yield)  was assessed with a Junior‑PAM fluorometer. Photogrammetric reconstruction of RGB images allowed for the analysis of 3‑D point clouds that permitted a quantitative comparison of structural parameters from the onset to the end of the treatment. After the drought period, a multi‑sensor UAV flight acquired LiDAR point clouds, multispectral reflectance, and thermal imagery to provide spatial context for the physiological observations. Structural changes were modest, whereas apparent SIF yields declined markedly in the drought‑stressed seedlings relative to well‑watered controls. Thermal maps showed slightly increased canopy temperature in stressed plants, corresponding closely with observed SIF reductions, particularly for F. sylvatica. By a combined analysis of temporal SIF dynamics and thermal signatures, we were able to jointly interpret observed signs of stress. From this plant level analysis, an outlook is given towards continuous observations, which were conducted over the temperate mixed ECOSENSE forest in SW Germany in 2025. In summary, the synergistic, multi-sensor approach presented here enhances the reliability of fluorescence-based remote sensing of plant stress and provides a scalable framework for monitoring drought impacts across heterogeneous forest ecosystems.  

How to cite: Göritz, A., Enriquez, A., Gassilloud, M., Stock, C., Haberstroh, S., and Werner, C.: Experimental assessment of drought‑induced changes in Sun‑Induced Fluorescence (SIF), 3‑D canopy structure and UAV‑based thermal imaging of Pseudotsuga menziesii and Fagus sylvatica seedlings, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20481, https://doi.org/10.5194/egusphere-egu26-20481, 2026.

EGU26-20481

Download Close

Understanding how tree provenances respond to local climate conditions is essential for predicting forest resilience under climate change. To investigate the performance of different European beech (Fagus sylvatica) provenances in the Dutch climate, a provenance trial was established in 1998 in Wageningen, the Netherlands. In this study, we evaluate the ability of UAV-borne LiDAR time series to capture temporal differences in spring and autumn leaf phenology among provenances.

Weekly UAV surveys were conducted from March to June and from October to December 2024 and 2025, with two additional flights during the summer period, over a 0.9 ha beech provenance trial consisting of 29 European provenances planted in three blocks (plot size 10 × 10 m). Data were acquired using a DJI M300 UAV equipped with a DJI L1 LiDAR sensor. From the LiDAR data, structural and radiometric canopy metrics were derived. These time series were compared with dendrometer measurements and physiological information related to the geographic origin of the provenances.

UAV-LiDAR structural metrics, such as canopy cover and height distribution, showed stable and consistent temporal patterns and were generally less sensitive to illumination and calibration effects than multispectral indices. However, LiDAR-derived metrics were highly sensitive to flight altitude, highlighting the importance of maintaining consistent acquisition settings throughout a time series. Differences in the onset and senescence of leaf phenology between provenances were observed from the LiDAR data, but clear relationships with provenance origin and dendrometer data are not yet conclusive.

How to cite: Bartholomeus, H., Kelly, N., and Copini, P.: The use of UAV-LiDAR time series to monitor spring and fall phenology of a beech provenance experiment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17321, https://doi.org/10.5194/egusphere-egu26-17321, 2026.

EGU26-17321

Download Close

Mangrove forests provide important coastal protection, biodiversity support and blue-carbon storage, yet they are highly sensitive to pollution-driven disturbances. The 2019 oil spill along the Brazilian northeast coast represents one of the largest recent environmental disasters in the South Atlantic, highlighting the need for scalable monitoring approaches capable of detecting both immediate impacts and delayed ecosystem responses. Beginning in late August 2019, petroleum residues reached hundreds of beaches and estuaries across multiple Brazilian states, affecting extensive intertidal habitats including mangrove-fringed shorelines; despite subsequent investigations, the spill’s source has not been conclusively established to date. Here we assess mangrove canopy dynamics following the spill using a multispectral, multi-resolution remote sensing framework that integrates satellite imagery and uncrewed aerial system (UAS) observations.

We analyzed time series of vegetation indices (NDVI, NDWI and chlorophyll-related indices) derived from PlanetScope, WorldView, Sentinel and UAS multispectral imagery for two affected mangrove areas in Pernambuco State, Brazil (Itamaraca and Carneiros, North and South coast respectively), covering the period 2018–2024. To enable cross-sensor comparison across different spatial resolutions, index distributions were harmonized relative to reference acquisitions. Pre- and post-spill windows were evaluated to capture short-term responses and longer-term trajectories. Tree-level structural data (height) were incorporated to test whether canopy condition changes were size-dependent such as other geographical parameters (zonation). Statistical analyses included parametric and non-parametric pre/post contrasts, trend evaluation across irregular acquisition intervals, and correlation and regression analyses linking tree height to spectral change metrics.

Across both sites, short-term analyses show no clear evidence of abrupt canopy degradation in moths immediately following the spill. In contrast, long-term trajectories reveal (years) a persistent decline in NDVI coupled with stable or slightly increasing NDWI, consistent with chronic physiological stress or progressive canopy thinning rather than acute dieback. The magnitude of long-term greenness loss is significantly greater in Itamaracá (North Coast) compared to Carneiros (South Coast), demonstrating spatial variation in exposure and/or ecosystem resilience. Additionally, emergency cleaning programs conducted primarily by local communities may have played an important role in influencing forest conditions. Height-dependent analyses further suggest that taller trees in Itamaracá experienced stronger post-spill declines, whereas responses in Carneiros were weaker and less structured by tree size. There was a slightly stronger decline in NDVI in the parts of the basin farthest from the tidal channels, likely because oil tended to linger longer in these areas.

These results demonstrate the value of multispectral, multi-resolution monitoring—combining frequent satellite coverage with targeted UAS surveys—for detecting subtle, delayed ecosystem responses to environmental disasters, supporting more effective impact assessment and evidence-based protection of sensitive coastal ecosystems.

How to cite: conti, L. A., Takahashi, K., and Barcellos, R. L.: Multi-Sensor Remote Sensing of Mangrove Resilience and Stress After the 2019 Northeast Brazil Oil Spill, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19047, https://doi.org/10.5194/egusphere-egu26-19047, 2026.

EGU26-19047

Download Close

How to cite: Schmid, J., Mosig, C., Vajna-Jehle, J., Mahecha, M., Cheng, Y., Hartmann, H., Montero, D., Junttila, S., Horion, S., Schwenke, M. B., and Kattenborn, T.: An update on deadtrees.earth: A community-driven infrastructure for tree mortality monitoring from local to global scales, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14805, https://doi.org/10.5194/egusphere-egu26-14805, 2026.

EGU26-14805

Download Close

Forest mortality is increasing globally under climate change, making detailed, large-scale monitoring essential for understanding ecosystem responses and guiding adaptive forest management. In this study, we present the first nationwide spatio-temporal assessment of standing deadwood in Switzerland, covering the period from 2018 to 2023, based on a semantic segmentation model applied to centimeter-scale high-resolution aerial imagery. We reveal a consistent upslope concentration of standing deadwood, with highest shares occurring around mid to high elevations (~1,500 m), despite declining forest cover. Relative increases of up to 43% were observed in overlapping survey areas, following the 2018 drought. Random forest models, interpreted using SHAP analysis, identified maximum temperature anomalies and conifer dominance as the key predictors of standing deadwood. The consistent accumulation of standing deadwood at higher elevations suggests increasing vulnerability of mountain forests, with implications for carbon storage, biodiversity, and disturbance susceptibility under ongoing climate change. Our results highlight the value of high-resolution remote sensing for large-scale monitoring of forest mortality. They advance understanding of the climatic and forest compositional drivers on forest mortality and offer a reproducible and transferable framework to support assessments of spatial patterns relevant for climate-adaptive forest management.

How to cite: Ferrari, L., Waser, L. T., Psomas, A., Mosig, C., Kattenborn, T., Ginzler, C., Griess, V. C., and Beloiu, M.: Nationwide deadwood mapping reveals rising mountain forests vulnerability, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4485, https://doi.org/10.5194/egusphere-egu26-4485, 2026.

EGU26-4485

Download Close

Long-term spatiotemporal mapping of landslides is crucial for understanding the dynamics of landslide, their impact on forest carbon stocks, and their interactions with environmental factors, climate variability, and disturbances. This study analyzed 33 years (1990-2022) of Landsat imagery and topography using machine learning (Random Forest) to map landslide dynamics in a 24,386-ha subtropical montane forest in Northeast Taiwan. We also quantified forest aboveground biomass (AGB) losses from landslides using temporally corresponding Landsat and lidar data. We observed pronounced interannual variability, with total landslide coverage ranging from 0.68% to 3.19%, and forest-to-landslide transitions driving annual AGB losses of 2 to 85 Gg yr⁻¹. Temporal analysis revealed exponential declines in landslide frequency (median = 2 events), persistence (one year), and reoccurrence (two times), indicating most landslides were short-lived. However, nearly half of affected sites reoccurred multiple times, indicating spatially persistent susceptibility. Topographic attributes, including elevation, aspect, slope, and local relief, exhibited greater sensitivity to extreme events. Crucially, typhoon-driven extreme rainfall, particularly daily maximum precipitation (r = 0.559, p = 0.004), showed a stronger relationship with newly formed landslides than daily maximum precipitation during the rainy season (r = 0.399, p = 0.026), emphasizing typhoon’s dominant triggering role. AGB losses from typhoon-triggered landslides were roughly 14-fold greater than in quiet years, profoundly impacting the regional forest carbon budget. Post-landslide vegetation recovery exhibited a highly variable trajectory and plateaued at ~63% of pre-disturbance biomass within 25 years, based on a non-linear asymptotic model. As climate change is projected to intensify typhoon activity and extreme rainfall, landslide risks and associated forest carbon losses will increase, particularly in vulnerable, typhoon-prone regions like Asia. These findings highlight typhoons are not only a principal driver of landslide activity but also a major disruptor of forest carbon budgets, underscoring their critical inclusion in carbon accounting frameworks for vulnerable montane ecosystems.

How to cite: Wang, H.-C., Wang, S.-F., Chung, C.-H., and Huang, C.: Spatiotemporal dynamics of typhoon-induced landslides and associated biomass loss over three decades, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15902, https://doi.org/10.5194/egusphere-egu26-15902, 2026.

EGU26-15902

Download Close

Rapid changes in climate and disturbance regimes are increasing uncertainty regarding the future vigour and productivity of boreal forests. This challenge is particularly relevant in Canada, where boreal forests cover more than 5.5 million km² and are predominantly composed of black spruce (Picea mariana Mill.) a species of central ecological and economic importance that appears increasingly sensitive to interacting climatic and biotic stressors. Drought, anomalous temperature extremes, frost damage and insect outbreaks can alter growth trajectories at annual to multi-decadal timescales. Quantifying the magnitude and spatial distribution of these growth changes is therefore crucial for anticipating impacts on timber supply, ecosystem service provision, and forest carbon balance.

Tree-ring data provide annual-resolution records of growth and have been instrumental in characterizing climate–growth relationships across the boreal forest. However, dendrochronological networks remain spatially sparse and often capture generalized sensitivities that do not fully reflect local growth responses driven by fine-scale environmental conditions, stand structure, and disturbance legacies. Critically, they do not provide a spatially continuous and regularly updated assessment of changes in forest productivity, nor do they readily identify the regions most vulnerable to emerging stressors. Time series of satellite observations offer a complementary and scalable perspective by providing spatially explicit, long-term measurements of canopy dynamics. In particular, Landsat imagery enables direct observation of forest canopy trajectories, capturing realized responses to multiple, interacting stressors and providing critical information to refine spatial assessments of growth dynamics beyond relationships based solely on climatic variability. Integrating Earth observation data with climate and tree-ring information therefore offers a powerful opportunity to leverage their complementary strengths and deliver timely, decision-relevant information for the stewardship of forest ecosystems.

In this study, we modelled the annual probability of severe growth decline in black spruce–dominated forests across Canada from 1988 to 2020 by integrating broad-scale climate data and Landsat time series with tree-ring–derived growth information from the CFS-TRenD repository. Tree-ring width series from 3,125 trees across 648 sites were used to characterize growth decline events and to train a probabilistic modelling framework that accounts for temporal dependence in growth responses and spatial heterogeneity in climate–growth relationships. The resulting model was then applied across key regions of Canada to examine spatiotemporal patterns in growth decline likelihood over recent decades and among major boreal ecozones. Results show that changes in the temporal trajectories of Landsat-derived spectral indices and forest structural attributes, together with indicators of climate extremes, were among the strongest predictors of growth decline probability, with spatial patterns and temporal trends in predicted likelihood consistent with observed growth variability in independent tree-ring series.

Although the mapped probabilities do not represent direct observations of severe growth decline, they provide continuous, spatially explicit information that is critical for identifying vulnerable regions, guiding targeted monitoring efforts, and anticipating future changes in boreal forest productivity under ongoing environmental change. More broadly, this study demonstrates how freely available climatic and satellite-derived datasets can be integrated with tree-ring information to extend growth-related insights to continental scales and support spatially explicit assessments of forest productivity and vulnerability.

How to cite: Morin-Bernard, A., Campbell, E. M., Hemosilla Gomez, T., Girardin, M. P., and White, J. C.: Mapping growth decline probability and trends across Canada’s black spruce forests from tree-ring, Landsat, and climate data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14546, https://doi.org/10.5194/egusphere-egu26-14546, 2026.

EGU26-14546

Download Close

Reliable global datasets of key forest variables are urgently needed to monitor forest dynamics both on regional and global scales. Forest canopy height is one of these key variables due to its close correlation to forest biomass and carbon stocks. Recently, new and promising datasets have been developed that utilize deep convolutional neural networks to predict canopy height from optical Sentinel-2 satellite data on a global scale. But how is this possible, given that there is no physical relationship between optical data and canopy height? To understand what the models have learned, we expanded upon the study of Lang et al. (2023) and quantified the contributions of geographical, spectral, and contextual features to the model's outcome. To evaluate the effect of geographical coordinates, the geo-locations of the model input scenes were systematically altered, while maintaining identical spectral features. The resulting canopy height predictions revealed consistent dependencies on geographic location, with mean increases of up to 10 m across entire Sentinel-2 scenes. Effect sizes for latitudinal shifts were large (Cohen's d ≈ 1), indicating that the model interprets spectrally identical input data differently at varying locations. This suggests that the subtle biases arose from the learned spatial priors of the model ensemble. Consequently, the accuracy of predictions decreases in areas where forest height substantially differs from the mean height typical for the respective biome or climate zone, e.g., due to local soil properties, climatic effects, or uncommon forestry management practices. To isolate the effect of spectral properties, we both increased and decreased values of single spectral bands in discrete steps while maintaining the same geographic locations. Mean differences in canopy height predictions, compared to those derived from unmanipulated input data, showed varying responses across different bands, manipulation degrees, and sample locations. The observed changes were not systematically connected to the manipulated spectral data, suggesting that spectral features did not significantly influence the model's output. By modifying the input data, we highlighted potentially significant obstacles to the further development of AI-driven models of key forest variables which need to be taken into account for applications thereof.

How to cite: Zerres, V., Sanchez, E., Nowosad, J., Meyer, H., and Lukas, L.: Global Canopy Height Models from Optical Satellite Data: What Has the AI Learned?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11803, https://doi.org/10.5194/egusphere-egu26-11803, 2026.

EGU26-11803

Download Close

How to cite: Zhang, H., Lang, N., Oehmcke, S., Mazurczyk, M., Brandt, M., Fensholt, R., Kariryaa, A., and Igel, C.: A Vertical Vegetation Structure Model of Europe, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14707, https://doi.org/10.5194/egusphere-egu26-14707, 2026.

EGU26-14707

Download Close

How to cite: Mukhopadhyay, R., Brede, B., and Bohlin, I.: Region-wide Prediction of Boreal Understorey Vegetation using Spaceborne Remote Sensing, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20765, https://doi.org/10.5194/egusphere-egu26-20765, 2026.

EGU26-20765

Download Close

The System of Environmental-Economic Accounting – Ecosystem Accounting (SEEA-EA) is a standard statistical framework developed by the United Nations and adopted in 2021. SEEA-EA aims to integrate the natural value of ecosystems in both physical and economic terms through ecosystem accounts. Physical accounts include ecosystem extent (e.g., extent of an ecosystem type), ecosystem condition (EC) (e.g., the health/ecological status of an ecosystem), and ecosystem services (e.g., carbon sequestration). Monetary accounts include ecosystem services (e.g., economic valuation) and assets. Forests play an important role in the socio-economic dynamics of many countries by providing multiple ecosystem services that support human well-being. In the context of SEEA-EA, forests are among the most-studied ecosystem types. However, most studies focus on ecosystem extent (e.g., forest cover changes) and ecosystem services (e.g., carbon sequestration). Less attention has been paid to EC, despite its importance in fully disentangling the link between ecosystem status and the supply of ecosystem services. In this study, we map and assess forest EC at the Lithuanian national scale and analyse changes over time by comprising two periods (2021 and 2024). In the SEEA-EA, EC is assessed based on abiotic, biotic, and landscape ecosystem characteristics, as defined by the SEEA-EA Ecosystem Condition Typology. Based on the literature, we defined three ecosystem variables for the ECT class, totalling 18 variables (e.g., tree cover density, soil organic carbon). The reference conditions for forest ecosystems in Lithuania were defined based on forests under strict protection. Based on these reference areas, the 18 variables were rescaled to 0-1 using the SEEA-EA methodological guidelines. The final EC index was calculated by overlaying the 18 indicators and assigning equal weights to each. The results showed higher EC values across Lithuania, particularly in the central and western parts of the country, which were associated with large, contiguous forest patches. Low EC was found in areas with smaller forest patches, mainly in the central, eastern, and western parts of the country. Regarding differences across years, the overall median EC index was higher in 2021 than in 2024. This can be attributed to changes in indicators that were not static (e.g., Leaf Area Index), which highlights both the advantages of remote sensing (e.g., large area cover and capacity to detect changes over time) but also influences the results (e.g., problems with cloud coverage for large areas such as national scale). Overall, this study is the first effort to map and assess forest EC beyond previous efforts to implement the SEEA in its experimental phase, serving as a basis for further development and improvement. The results obtained contribute to enhancing knowledge of the ecological status of Lithuanian forests, providing insights and guidance to support the implementation of SEEA-EA in Lithuania, which is envisaged within European environmental directives and policies.

How to cite: Inácio, M., Santos-Martín, F., and Pereira, P.: Mapping and assessing forest ecosystem condition in Lithuania, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13684, https://doi.org/10.5194/egusphere-egu26-13684, 2026.

EGU26-13684

Download Close

The Austrian Research Centre for Forests (BFW) uses a range of remote sensing (RS) data—including aerial imagery, airborne laser scanning, and high‑resolution Copernicus Sentinel data—to support the National Forest Inventory and produce national forestry products. These include tree species maps, timber stock maps, and a Sentinel‑2‑based anomaly detection map for identifying forest disturbances across Austria. The success of these products is grounded in three factors: (1) strong collaboration with researchers to ensure that recent scientific advances are translated into operational applications; (2) the integration of high‑resolution aerial data with extensive local expertise, enabling high‑quality training datasets for machine‑learning approaches; and (3) the Austrian National Forest Inventory, which provides a robust ground‑truth basis for validating RS‑derived results.

At the same time, the demand for precise, timely, and spatially detailed national forestry products is steadily increasing. This demand is driven by growing monitoring and reporting requirements, as well as by the increasing impacts of land-use and climate change on forest ecosystems. These developments highlight existing limitations of RS-based forestry products, particularly in complex alpine terrain, where terrain shadows and cloud cover can delay or obscure the detection of natural disturbances such as windthrows. Within the SAFIR project, we investigate how AI–based tools can be used to address these challenges and enhance the performance of existing forest disturbance monitoring tools.

As presented in our poster, the SAFIR project combines BFW’s anomaly detection map with ground-truth training data on windthrow events provided by the Österreichische Bundesforste (ÖBF). This fusion allows us to: (1) distinguish windthrow events from other disturbance types within the anomaly detection map; (2) assess the spatial and temporal agreement between modelled disturbances and inventoried windthrow events across Austria; and (3) quantify the detection rate of windthrow events in the existing product. Building on this assessment, identified windthrow sites can be specifically targeted with machine-learning approaches for cloud and terrain-shadow removal, thereby improving both the timeliness and accuracy of windthrow detection and providing validated inputs for developing new datasets and AI models specialized in de‑clouding and de‑shadowing windthrow areas.

By integrating established Copernicus-based forest disturbance products with an extensive, independently collected ground-truth dataset on windthrows, SAFIR enables a systematic evaluation of current windthrow detection capabilities and provides a pathway for targeted methodological improvement. By leveraging AI-based techniques to overcome known limitations, the project contributes to the development of more robust national forestry products for monitoring windthrow damage.

How to cite: Bjärhall, A., Fanta-Jende, P., Beltrame, L., Salzinger, J., Lampert, J., and Schumacher, B.: Leveraging Copernicus data, local expertise, and novel technology to create national forestry products, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11017, https://doi.org/10.5194/egusphere-egu26-11017, 2026.

EGU26-11017

Download Close

Forest biomass accumulation significantly alters three-dimensional structural characteristics. This study developed a physics-based allometric growth equation to estimate forest above-ground biomass (AGB) at the footprint level using GEDI L1B full-waveform LiDAR data. The model operates within a well-defined physical framework, systematically integrating canopy structural parameters and species-specific attributes through a three-component architecture. First, it constructs a Waveform Index (WI) characterizing the vertical energy distribution by combining canopy height (H) retrieved from waveform data with the typical crown architecture. The second component incorporates key canopy structural parameters: canopy gap fraction (P), which quantifies vertical openness, and leaf area volume density (LVD), describing the vertical distribution of foliar mass. The third component introduces wood density (ρ). In boreal coniferous forests, the model achieved an R² of 0.66 and an RMSE of 20.22 t/ha, explaining 83% of the observed variance in AGB.

The method revealed that biomass accumulation was closely related to canopy height and wood density. While canopy height was directly retrievable from the waveform, wood density data were not readily available at large regional scales. Therefore, this research utilized land cover types as a base map and inferred the distribution of diffuse-porous wood and ring-porous wood forests across China by integrating multiple factors—including climate, topography, and phenology. The species composition was further refined using provincial forest inventory data on dominant tree species, excluding species accounting for less than 5% of a province's forest area. Wood density grades were then classified and incorporated into the footprint-level allometric equation for AGB estimation. This estimation method enables direct parameterization and retrieves AGB directly from satellite observations, while also accounting for the physiological characteristics of trees. This study demonstrates the significant potential for forest AGB estimation by leveraging canopy height and wood density. The proposed approach provides a foundation for forest carbon monitoring in precision forestry.

How to cite: Zhihui, L. and Weimin, J.: A Physics-Based Estimation Model of Forest Aboveground Biomass Integrating Wood Density Classification and GEDI Waveform Retrieval, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6230, https://doi.org/10.5194/egusphere-egu26-6230, 2026.

EGU26-6230

Download Close

Canopy percentile height is a critical parameter for describing forest vertical structure and assessing terrestrial carbon sequestration. At large scales, it is typically measured using spaceborne LiDAR systems. The DQ-1 satellite differs from conventional spaceborne LiDAR systems by employing a multi-band, single-footprint design with full-waveform reception, similar to ICESat-1 and GEDI. However, DQ-1 canopy height inversion is limited by 1064 nm waveform saturation and cross-band differences in shape and resolution. To overcome these issues, we propose a multi-band fusion algorithm (MBFA-F), whose resulting products serve as ancillary outputs of the DQ-1 satellite. Multi-source validation shows that with Finnish ALS data, MAE and RMSE remain small and stable in the RH20–RH40 range (MAE: 3.05–2.46 m; RMSE: 2.84–1.99 m). The errors increase with higher percentiles (RH41–RH100), reaching maximum values at RH100, where the MAE is 6.30 m and the RMSE is 4.89 m. With the combined use of ICESat-2 and GEDI data, RH98 yields an MAE of 6.27 m and an RMSE of 8.51 m, while RH90 yields an MAE of 6.40 m and an RMSE of 8.49 m. Although the accuracy may have some limitations, they effectively fill the data gap in high-latitude boreal forests, offering useful supplementary information for related research. Compared to the canopy percentile height products of GEDI and ICESat-2, statistical analysis of boreal forest regions shows that DQ-1 provides additional pixel coverage ranging from 0.05% to 3.06% in various countries. The above results demonstrate that the DQ-1 satellite has significant potential for dynamic monitoring of forest canopies. 

How to cite: Zhang, H.: Global Mapping of Forest Vertical Structure with DQ-1 Multi-Wavelength LiDAR: Focus on Boreal Forests, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8619, https://doi.org/10.5194/egusphere-egu26-8619, 2026.

EGU26-8619

Download Close

Forest structural complexity is a key control on carbon storage, ecosystem functioning, and forest resilience, but its quantification across spatial scales remain challenging, even in managed tropical systems. Commercial plantations of macaúba (Acrocomia aculeata), a native Brazilian palm with increasing relevance for the vegetable oil and bioenergy markets, represent an emerging forest-based bioeconomy whose structural development is still insufficiently described using remote sensing techniques. In this study, we evaluate an integrated remote sensing framework that combine multi-platform LiDAR data (UAV-borne and airborne), multispectral satellite imagery, and field measurements to characterize forest structure and associated carbon stocks in commercial macaúba plantations. High-density LiDAR point clouds were used to derive three-dimensional structural attributes such as canopy height, vertical complexity, and spatial heterogeneity, assessed at both individual-tree and stand scales. These LiDAR-derived metrics were then integrated with satellite time series to support spatial extrapolation and the analysis of structural development and carbon accumulation over time. Relationships between remote sensing metrics and field observations were established using machine learning approaches, enabling robust estimation of aboveground biomass and carbon stocks while maintaining sensitivity to fine-scale structural variability. At the stand scale, the integrated LiDAR–satellite approach achieved coefficients of determination above 0.70 in independent validation, with biomass estimation errors on the order of 10 t ha⁻¹. These results indicate that reliable structural and carbon assessments can be obtained without rely on single-sensor datasets. The analysis highlight the complementary contribution of different LiDAR platforms. UAV-borne LiDAR provide detailed information on canopy and sub-canopy structure at the individual-tree level, whereas airborne LiDAR allow consistent and scalable mapping at the landscape scale. In addition, LiDAR acquisition characteristics, particularly point cloud density, was found to strongly influence the robustness and transferability of structural metrics. By linking individual-level measurements with landscape-scale observations, this multi-platform LiDAR framework advances the assessment of structure and carbon dynamics in planted tropical forests, supporting applications in forest inventory, ecological modeling, and the sustainable management of commercial forest systems associated with climate mitigation and the vegetable oil bioeconomy.

How to cite: Imbuzeiro, H., Rosário, D., and Filpi, H.: From Trees to Landscapes: Integrating Multi-Platform LiDAR for Structural Assessment of Commercial Macaúba Forests and Carbon Monitoring, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2632, https://doi.org/10.5194/egusphere-egu26-2632, 2026.

EGU26-2632

Download Close

Forest management shapes forest structure and thereby the habitat for wildlife. One group that relies on forests during at least part of its life cycle is European insectivorous bats. In the context of climate change, diversification of forest stands is increasingly promoted as a management strategy. However, robust and quantitative tools to evaluate the structural outcomes of different management regimes and their ecological consequences on forest-dwelling bats remain limited. Traditional forest habitat assessments rely largely on field surveys that are time-consuming and observer-dependent. Such surveys are unable to capture complex three-dimensional structural properties such as gap volume or foliage distribution.

In this study, we present a novel three-dimensional LiDAR-based forest habitat characterisation approach and assess how vegetation structure parameters relate to bat activity and bat species richness. High-resolution 3D point clouds were acquired using a handheld mobile laser scanner in beech and mixed forest stands under managed and unmanaged regimes in the Vienna Woods Biosphere Reserve, Austria. Unlike commonly used 2.5D raster-based methods, our approach exploits the full three-dimensionality of point clouds to quantitatively describe vegetation structure. For habitat characterisation, we calculated the number of potential habitat trees, gap availability (gap ratio), and foliage height diversity (FHD). To illustrate the ecological relevance of these structural parameters, we combined the 3D characterisation with acoustic monitoring of bat echolocation calls across 40 sampling plots. Activity data were collected in May and June 2024 and analysed in relation to forest type, management type, and the calculated vegetation structure parameters.

We found clear differences in vegetation structure between beech and mixed stands. Further, stand type and the three vegetation structure parameters (i.e. number of potential habitat trees, gap ratio, and FHD) significantly affected the activity of foraging groups (e.g. open-space foragers) and taxonomic groups (e.g. Myotis and the Nyctaloid group). In contrast, we did not detect significant effects of stand type, management type, or vegetation structure parameters on species richness. Our results suggest that forest structure primarily influences the intensity of habitat use rather than species presence in the Vienna Woods Biosphere Reserve.

Overall, this study demonstrates the added value of full 3D point cloud analysis for linking forest management practices to habitat characterisation. The proposed workflow, implemented in R and applicable across forested ecosystems, provides forest managers and researchers with a tool to assess and guide management decisions aimed at balancing timber production, climate adaptation, and biodiversity goals. The information gained from the habitat characterisation approach can support ongoing efforts within the Vienna Woods Biosphere Reserve and beyond. Improving forest conditions for bats will contribute to the long-term conservation of these mammals.

How to cite: Dammert, L., Arav, R., and Suarez-Rubio, M.: 3D Habitat characterisation for targeted bat conservation in the Vienna Woods Biosphere Reserve , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19140, https://doi.org/10.5194/egusphere-egu26-19140, 2026.

EGU26-19140

Download Close

Demand is growing for cost-effective, current, and spatially detailed data on forest attributes— including species composition, growing stock, disturbances, and mortality—driven by management requirements, the multifunctional roles of forests, and their sensitivity to climate change. Advances in high-resolution remote sensing, deep learning, and rapid data processing now enable reliable, reproducible, wall-to-wall forest products that complement traditional inventories with regularly updated, spatially explicit information essential for sustainable, multifunctional, climate-adapted forest management.

Despite four decades of developing remote sensing–based forest products, their adoption by forestry practitioners remains slow and often incorrect or limited (e.g., Barrett et al., 2016; Waser and Ginzler, 2023; Fassnacht et al., 2024; Waser et al., 2025). In some cases, products fail to meet user expectations for accuracy or update frequency, revealing a mismatch between development and practical needs. This gap stems largely from poor knowledge exchange between researchers and practitioners, leading to differing expectations and misunderstandings of product content. Misalignment arises from differing expectations, limited understanding of practical needs, and technical challenges. While datasets like canopy height models are widely and effectively used, more complex products such as tree species or disturbance maps remain challenging and prone to misinterpretation. Adoption is further hindered by technical terminology, the need to integrate products into existing workflows, and the time, cost, and complexity of adapting decision-making processes.

In this study we show how to bridge the gap between remote sensing research and stakeholders, including forest industries, service providers, practitioners, and forest owners. We identify core challenges limiting the adoption, accuracy, and utility of forest products and propose a collaborative framework emphasizing cooperation between researchers and practitioners. We present examples of active user involvement to further improve the quality of remote sensing–based forest products by incorporating additional training data, adjusting model settings, and retraining iteratively based on new feedback. Active user involvement benefits both sides: it helps develop user-friendly products and provides supplementary reference data essential for machine learning, thereby advancing remote sensing research.

We tackle the key challenges and opportunities for integrating remote sensing research into forestry practice and propose strategies to improve utilization and acceptance of these products. We focus on five critical components:

• Enhancing collaboration between researchers and forestry stakeholders to ensure product development matches user requirements and fosters technological progress.
• Engaging applied research initiatives, engineering firms, and start-ups to translate discoveries into practical products.
• Tailoring methods and products to practical, real-world applications, while maintaining relevance in informational content, accuracy, spatial resolution, and alignment with existing datasets.
• Integrating user feedback through quality checks, validation, and iterative improvements.
• Promoting clear communication and documentation, including intended use, interpretation guidance, and transparency regarding accuracy and uncertainty.

In summary, we show that addressing these issues requires active engagement of stakeholders in product development, iterative quality assessments, and alignment of methods with real-world use cases.

How to cite: Waser, L. T., Schwenke-Beloiu, M., Stereńczak, K., Adler, P., Havryliuk, S., and Rehush, N.: Strengthening forest remote sensing by linking research and practice: a collaborative framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5343, https://doi.org/10.5194/egusphere-egu26-5343, 2026.

EGU26-5343

Download Close

Bangladesh’s forests are vital for maintaining biodiversity, supporting rural livelihoods, and contributing to climate resilience. However, they face growing challenges from agricultural expansion, population pressure, infrastructure development, and climate-induced hazards. While several reforestation and community afforestation programs have been initiated, the overall trend, spatial variability, and underlying drivers of forest cover change remain insufficiently understood. This study integrates remote sensing, machine learning, and socioeconomic analysis to examine the dynamics and determinants of forest land change across Bangladesh over a 19-year period (2000–2018).

The study employs Landsat time-series imagery processed in Google Earth Engine. Random Forest classification generates annual land cover maps, followed by least-square trend modeling to detect forest growth rates. Hotspot and zonal statistical analyses will identify regions of significant change. Expert interviews, literature review, and secondary datasets will be used to examine drivers such as population pressure, shifting cultivation, community forestry programs, governance challenges, and climatic influences.

Preliminary results show that the total forest area in Bangladesh increased at an average annual rate of 0.78% (approximately 25,932 hectares per year) between 2000 and 2018. However, this growth is spatially uneven. The study will provide a comprehensive understanding of how forests in Bangladesh are changing across different forest types and vital project areas for forest development, such as community afforestation and coastal greenbelt projects, several reserved forests, including the Chittagong Hill Tracts and parts of the Sundarbans; and what factors are driving these dynamics. Findings will inform national forest management, policy development, and sustainable land-use planning.

How to cite: Ullah, K. and Tiwari, V.: Spatiotemporal Dynamics and Drivers of Forest Cover Changes in Bangladesh, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-50, https://doi.org/10.5194/egusphere-egu26-50, 2026.

EGU26-50

Download Close

Canopy height growth is a key determinant of the state and functioning of forest ecosystems. As traditional ground-based inventories can not exhaustively capture growth and ensure hotspots detection, mixed-source canopy height time series from multiple remote sensing platforms now enable extensive characterization of these dynamics, provided that measurement biases between sources are addressed. We proposed a transferable workflow to map spatially explicit patterns of vertical growth across forested landscapes. By leveraging recent aerial imagery and lidar data regularly acquired across Belgian temperate forests over 2006-2021, standardized against ground-based inventories at ~1000m² spatial resolution, we estimated plot-level vertical growth and modeled species-specific reference trajectories from which we quantified plot-level deviations, providing both absolute and contextualized assessments. Across acquisitions, the standardization approach reduced the top-of-canopy height bias from 2.64±2.01 m to 0±1.77 m (RMSE = 1.77 m, R² = 0.92). Canopy structure, rather than acquisition parameters, was the main source of bias when estimating forest height from aerial imagery. Plot-level growth exhibited decreasing trends as initial height increased. Importantly, deviations from reference vertical growth displayed significant spatial clustering (Moran's I = 0.36, p < 0.001), suggesting systematic variations indicative of potentially declining or over-performing stands. Our workflow offers transferability, reproducibility, and multi-scale applicability for spatially exhaustive characterization of forest growth dynamics, providing actionable insights to support adaptive management and conservation planning.

How to cite: de LAME, H., Bastin, J.-F., and Messier, C.: Spatial patterns of forest growth dynamics with mixed-source time series of canopy height, a novel approach using belgian temperate forests as case study., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4839, https://doi.org/10.5194/egusphere-egu26-4839, 2026.

EGU26-4839

Download Close

Remote sensing plays a central role in providing qualitative and quantitative information on forest ecosystems for sustainable management and forest carbon inventory. The ESA BIOMASS satellite mission launched in 2025 introduces the first global P-band Synthetic Aperture Radar (SAR) system dedicated to forest structure and biomass, but its potential for refining tropical forest ecosystem maps remains largely unexplored. Here we compare P-band (~70cm wavelength) BIOMASS data with C-band Sentinel-1 SAR to assess their respective ability to discriminate structurally and hydrologically different forest ecosystems in Brazilian and Bolivian Amazonia, including terra firme forest, flooded forest and forest–wetland mosaics, in the presence of sparse solid ground-truth data.

We use MapBiomas Amazonia and MapBiomas Bolivia as primary land-cover references, taking advantage of their annual time series from 1985 to 2024 to address label noise. Because these maps are not error-free at the pixel level, we develop a noise-labelling pre-processing workflow to derive high-confidence forest samples at the epoch of the first BIOMASS acquisitions (2025). The workflow combines (i) spatial homogeneity constraints (distance to class boundaries, neighbourhood purity, minimum patch size), (ii) temporal stability of the MapBiomas class history (to identify pixels with persistent forest or flooded forest trajectories), and (iii) physical plausibility checks using auxiliary optical and terrain indicators. Pixels that satisfy these criteria are retained as reliable proxies for different forest ecosystem types.

For these filtered samples we extract BIOMASS Detected Ground-range Multi-looked (DGM) backscatter and Sentinel-1 Ground-range Detected (GRD) backscatter, derive polarisation ratios and simple texture metrics, and quantify within-class variability and between-class separability for both frequencies. We pay particular attention to forest–non-forest transitions and to distinctions among terra firme forest, flooded forest and adjacent forested wetlands that are relevant for high-carbon stock and peat-forming systems. Preliminary results from the Brazilian test site indicate that P-band reduces within-class variance in forested classes and enhances separability between terra firme and flooded forests compared to C-band alone, while C-band performs comparably or better for some open and anthropogenic land covers. By extending the analysis to Bolivian Amazonia and to a richer legend of forest and forest-wetland classes, and by testing a similar workflow for peatland-prone flooded forests, this study provides a first evaluation of the potentials and limitations of BIOMASS P-band SAR for tropical forest applications under sparse ground-truth conditions.

How to cite: Dong, Y., Avoiani, E., Dabiri, Z., and Blaschke, T.: Evaluating P- and C-band spaceborne SAR for refined tropical ecosystem mapping in Amazonia under sparse ground-truth conditions, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20155, https://doi.org/10.5194/egusphere-egu26-20155, 2026.

EGU26-20155

Download Close

How to cite: Koukos, A., Kondylatos, S., Grogan, K., and Nord-Larsen, T.: National Tree Species Mapping in Denmark with Machine Learning Using Multi-Temporal Sentinel Data , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19343, https://doi.org/10.5194/egusphere-egu26-19343, 2026.

EGU26-19343

Download Close

Monitoring phenology over multiple decades is crucial for understanding how forest productivity responds to climate change. In this regard, satellite remote sensing is indispensable to capture land surface phenology (LSP) at regional to continental scales. Sensors with high spatial or temporal resolution, such as combined Landsat/Sentinel-2 or MODIS time series, have been used to estimate annual LSP. However, these time series either cover relatively short periods (<10 years) or aggregate signals across multiple land cover types, limiting our understanding of long-term phenology change. The full Landsat archive spans over 40 years, offering long-term coverage, but sparse observations before the 2000s have limited its use for annual LSP estimation. Here, we explore the potential of the Landsat archive for estimating phenological parameters across European forests. We processed all available Landsat Level 1 images from 1984-2024 (>300,000) using the Framework for Radiometric Correction for Environmental monitoring (FORCE), including radiometric and topographic corrections as well as cloud and cloud shadow masking. To isolate phenological changes, we excluded forest pixels affected by disturbances including windthrow, fire, bark beetle outbreaks, or harvest. For the remaining undisturbed pixels, long-term phenological parameter distributions were first estimated from the full 40-year time series using a double-logistic Bayesian model. These parameter distributions were subsequently used as informative priors in a Bayesian hierarchical framework to estimate start (SOS), peak (POS), and end (EOS) of season from sparse annual observations, while accounting for regional tree species composition. Our two-stage modelling approach enables robust annual phenology estimation across the full Landsat era, including data-sparse early decades, and provides a basis for analyzing long-term forest phenology dynamics at continental scales.  

How to cite: Kowalski, K., Viana-Soto, A., and Senf, C.: Towards long-term monitoring of forest phenology using Landsat time series, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10061, https://doi.org/10.5194/egusphere-egu26-10061, 2026.

EGU26-10061

Download Close

Robust estimation of aboveground biomass (AGB) plays a pivotal role in forest resource management and carbon accounting. These estimates are especially relevant within the framework of climate mitigation strategies such as REDD+, yet direct tree-level estimates over large areas are still challenging to obtain. AGB predictions commonly rely on allometric models calibrated from destructive sampled trees. While diameter-based allometric models dominate, the high costs related to measuring diameters at tree-level have recently driven interest in alternative allometries. In this context, advances in remote-sensing technologies enable direct and spatially explicit characterization of three-dimensional forest structure, including tree height and crown attributes. Crown width continues to expand even when tree height growth slows, offering valuable information for AGB prediction. Together, these developments support diameter-independent, remotely sensed AGB models, though challenges remain in data availability, segmentation accuracy, and cross-site generalization.

Thus, the objective of this study was to develop an alternative methodological framework for assessing single-tree and stand AGB, along with its associated uncertainty, by upscaling predictions from field-calibrated terrestrial laser scanning (TLS) data to airborne laser scanning (ALS) data.
To achieve our objective, we conducted a case study at the Remningstorp study area in southern Sweden (58.5° N, 13.6° E), where the forest is dominated by Norway spruce (Picea abies), Scots pine (Pinus sylvestris), and birch (Betula spp.). In 2014, the site was surveyed collecting a combination of field measurements (diameter at breast height and tree height), TLS, and ALS data. In addition, destructively collected single-tree measurements from Marklund (1998) are being used to define diameter-independent models for AGB prediction, using crown-related features, such as tree height and crown diameter, as explanatory variables.

Our assumption is that crown related features can be reliably characterized from medium-density ALS data (approximately 10–100 points/m²). Thus, we will use the diameter-independent model to predict tree-level AGB from ALS and perform a rigorous assessment of associated uncertainties. This approach relies on accurate single-tree segmentation, the matching of field-measured trees with remotely sensed trees, and the extraction of crown and height metrics from TLS and ALS data. The accuracy of our method will be further tested comparing the proposed survey approach with the traditional Swedish AGB models, based on the tree height and DBH as predictors (Marklund 1998).

The expected results of this study are twofold: (i) the development of an AGB allometric model based on tree height and crown diameter, applicable to both field-measured and remotely sensed data, (ii) a comprehensive evaluation of the uncertainties inherent in upscaling this model from TLS to ALS data, and (iii) wall-to-wall AGB mapping with associated uncertainty analysis across the study area.

How to cite: Papucci, E., Pires, R. D. P., Yrttimaa, T., Valbuena, R., Persson, H., Mensah, A. A., Roberge, C., and Ståhl, G.: Remote sensing-based forest aboveground tree biomass and uncertainty assessment through upscaling from single tree to stand level, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6916, https://doi.org/10.5194/egusphere-egu26-6916, 2026.

EGU26-6916

Download Close

Forests, which cover around one-third of the Earth’s land surface, play a crucial role in climate regulation, the global carbon cycle and biodiversity conservation. Forest-related datasets derived from Earth observation (EO) data often serve as baseline layers in various applications, ranging from biosphere monitoring to policy- and decision-making. However, their accuracy and temporal availability vary across regions and climatic zones, and trees used for agricultural purposes are frequently misclassified as forests. Therefore, there is a need for an accurate, up-to-date, and globally consistent forest cover layer that clearly distinguishes forests from tree crops and is available across multiple years. The current advent of big EO data with a combination of advances in Artificial Intelligence lead to the development of geospatial / EO embeddings as ready-to-use products for local to global applications, including forest monitoring. In this study, we develop a highly accurate global forest/non-forest (F/nF) classification at 10 m spatial resolution, while explicitly classifying tree crops as a sub-class of the non-forest category. Our approach implements Google Alpha Earth Foundation’s Satellite embedding dataset in an automatic training process through simple machine learning approaches, including linear Support Vector Machine (SVM), k-nearest neighbors (kNN) and random forest (RF). Automation is achieved through the generation of training data by intersecting multiple forest-related, land cover, plantation and agroforestry datasets across more than 200 training areas, proportionally representing all global biomes. Classification accuracy is assessed through ~21,000 global F/nF reference samples for the year 2020, complemented by several open-access tree crop and plantation validation datasets for 2019-2021. Our F/nF map for the year 2020 achieves an overall accuracy (OA) of 92% and macro F1-score of 0.91, with balanced omission and commission errors for the forest class of 14% and 13%, respectively. Validation of the tree crop sub-class showed high accuracies with OAs exceeding 90% for oil palm, while additional tree crop classes are still being assessed. Among the evaluated classifiers, both the linear SVM and kNN outperform more complex models, including non-linear SVM variants and fine-tuned RFs. In comparison to other global F/nF layers and widely-used land cover datasets, our F/nF dataset’s performance is better or comparable to these alternatives, while it additionally provides information on tree crops. Moreover, the initial transferability tests demonstrate that the trained models produce accurate and spatially consistent results for the period 2017-2024, showing their strong potential for global multi-year change detection analysis at 10 m spatial resolution. These results can support decision-making for policies and regulations, including the European Union Deforestation Regulation (EUDR). The open-access availability of both the resulting dataset and trained models enables global applicability and encourages further testing, adaptation and development by the EO and forest monitoring communities.

How to cite: Paluba, D., Hastie, A. T., Puerta Quintana, Y. T., Marsocci, V., and Onačillová, K.: Global forest, non-forest and tree crop mapping at 10 m resolution using satellite embeddings, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21065, https://doi.org/10.5194/egusphere-egu26-21065, 2026.

EGU26-21065

Download Close

There is an ever-growing toolshed of processing solutions for close-range LiDAR scans of forests. But these tools often cover only a fraction of the workflow from a point cloud to a full forest inventory, which includes tree position, diameter at breast height (DBH), tree height, and species. Many tools either perform single steps, such as segmenting individual trees, or extract only geometric information, such as DBH and tree height, but not species information, while others do just this. Even though many of these tools are validated, and first benchmarks exist for individual tasks, it remains unclear whether a pipeline can be conducted across multiple tools to generate a full inventory and how errors propagate through such pipelines. Therefore, we validate a pipeline that includes single-tree segmentation (SegmentAnyTree), species classification (DetailView), and geometric parameter extraction (CspStandSegmentation) against manual full inventories of two contrasting forests in south-west Germany. The first forest is a flat, mature mixed forest dominated by Fagus sylvatica (approx. 1500 trees), while the second forest is on steep terrain with a diverse age structure, mostly dominated by coniferous species like Picea abies and Abies alba (approx. 750 trees). Therefore, these forests depict a strong gradient in structural complexity. We illustrate how reproducible, easily usable and scalable pipelines can be implemented across programming languages using the Galaxy platform. We clearly depict how errors propagate from the segmentation to the subsequent processes and how this influences the overall performance of forest inventory tasks.

How to cite: Frey, J., Kröner, K., Weidenfeller, M., Wardius, Y., Larysch, E., Gerberding, K., Kattenborn, T., and Seifert, T.: From Trees to Forest Inventories: Do End‑to‑End LiDAR Pipelines Really Work?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19228, https://doi.org/10.5194/egusphere-egu26-19228, 2026.

EGU26-19228

Download Close

Mobile laser scanning can be used to efficiently measure tree stems with high precision. However, the laser beam divergence will affect the accuracy if a curved surface, such as a tree stem, is to be measured. In this work we present a method for correction of the bias introduced by the physical properties of the emitted laser pulses. The aim of the work was to estimate tree stem centres and stem diameters for different heights above the ground level (i.e., stem vector). We used two different laser scanners (Ouster OS0; Ouster OS1) mounted on a forest harvester operating in northern Sweden (64.3° N, long. 19.8° E). The beam divergences were 0.35 and 0.18 degrees, respectively. For validation, the trees were measured with stationary laser scanning (Leica RTC360). The mobile laser scanning data were combined with data from an inertial navigation system (INS) and point clouds were derived using a simultaneously localization and mapping (SLAM) algorithm. To avoid influence of errors remaining after the SLAM computations, laser data were in a first step processed scan-wise to estimate circle centre and circle radius based on laser returns from the tree stems. The correction of the stem diameter bias caused by beam divergence was in this step also performed using a new algorithm using solely data from the same scan rotation. In addition, laser returns from the ground were extracted for each scan rotation. In a second step, circle estimated and ground returns from all scans were merged for further processing. The circle locations were in this step clustered to build up tree stems, and a ground elevation model was created using an active contour surface to normalize height values. Stem diameter profiles (stem vectors) were estimated for each tree stem using all circles associated to an individual tree. A priori information about tree stem allometry was used for the final interpolation of stem diameter vectors. The vectors of stem diameters and tree stem centre locations were validated using data from the stationary laser scanning. The results show that stem diameter estimation bias could be corrected using the new scan-wise bias correction method. Furthermore, stem shape could be estimated with sufficiently high accuracy to make the method useful for practical applications. The method could therefore be used in the future for real-time bucking optimization to improve utilization efficiency of wood resources.

How to cite: Holmgren, J., Förster, N., Fransson, J., and Lindgren, N.: Estimation of tree stem diameter with mobile laser scanning using beam divergence bias correction and allometric models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9310, https://doi.org/10.5194/egusphere-egu26-9310, 2026.

EGU26-9310

Download Close

Forests are critical ecosystems that sustain biodiversity conservation, carbon cycling and climate regulation. Recent advances in laser scanning technology have provided unprecedented opportunities for detailed forest inventory and monitoring. Airborne, unmanned aerial, mobile, and terrestrial laser scanning systems produce complementary 3D point clouds that capture forest structural attributes across multiple scales and viewing geometries. However, the inherent heterogeneity in data characteristics across platforms severely limits the generalizability of conventional data-driven models. Additionally, naive multi-platform data mixed-training strategies that simply combine multi-platform data often lead to negative transfer, degrading segmentation performance and hindering consistent results across different acquisition systems. To address these challenges, we propose a Multi-platform Synergistic Training (MST) paradigm, a data- and model-driven representation learning framework, which can be seamlessly integrated into both semantic (tree components segmentation) and instance (individual tree segmentation) segmentation deep learning architectures. MST explicitly captures shared structural representations of forest environments through Cross Platform Aware Tokens (CPATs) and a Context Integration Module (CIM), which together enhance transferability and stability across heterogeneous forest point clouds. Furthermore, MST employs a two-stage training strategy in which platform-invariant features are first learned from pre-training on virtual synthetic multi-platform forest datasets, followed by fine-tuning on real-world data. This design lays the foundation for robust, platform-agnostic forest scene understanding while substantially reducing reliance on large volumes of manually annotated real-world data for training. The code for the proposed representation learning framework is available at: https://github.com/jdjiang312/MST.

The effectiveness of the proposed method is evaluated on nine benchmark forest point cloud datasets covering airborne, unmanned aerial, mobile, and terrestrial acquisitions, for both semantic and instance segmentation. Cross-dataset generalization experiments demonstrate that our framework achieves robust performance across all platform datasets and consistently outperforms models trained on single-platform data. Furthermore, by pre-training MST on a virtual synthetic forest point cloud dataset and subsequently fine-tuning on real-world data, the framework attains accuracy comparable to fully supervised training for both single-tree segmentation and tree-component segmentation, while relying on only 20% of the real annotations (semantic - mIoU: fully supervised 69.42% vs. MST 69.71%, instance - F1 Score: fully supervised 88.69% vs. MST 86.96%). These results highlight MST as a promising paradigm for cross-platform forest point cloud analysis, significantly reducing labeling costs while improving robustness and scalability. The framework thus offers a practical tool to enhance forest monitoring, inventory, and ecosystem assessment.

How to cite: Jiang, J., Shen, Y., Wang, J., Hollaus, M., Kissling, W. D., Ferreira, V., and Pfeifer, N.: Toward Platform-Invariant Forest 3D Perception: A Multi-platform Synergistic Training for Forest Point Cloud Segmentation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4814, https://doi.org/10.5194/egusphere-egu26-4814, 2026.

EGU26-4814

Download Close

Terrestrial laser scanning (TLS) enables the detailed three-dimensional characterization of forest stands, capturing structural elements from stems to individual branches in an objective and reproducible way. This high-resolution structural information is valuable for a wide range of applications, including precision forestry, forest management, and ecological and biodiversity monitoring. In addition, TLS-derived forest structure can serve as reference data for the calibration of area-wide remote sensing products, such as airborne laser scanning (ALS) point clouds, and for identifying the structural contributions to synthetic aperture radar (SAR) backscatter signals.

Recent technological developments, particularly devices becoming lighter, easier to operate, and capable of functioning in both static and kinematic modes, have considerably broadened the applicability of TLS in forest environments. While multi-scan static TLS acquisitions still represent the gold standard in terms of geometric accuracy, kinematic laser scanning setups are increasingly able to provide point clouds suitable for many forest-related applications and offer advantages with respect to acquisition time and field logistics. In addition, improved usability and increasingly automated processing workflows have expanded the user base of TLS beyond remote sensing and surveying experts. As a result, TLS is now frequently integrated into applied forestry as well as inter- and transdisciplinary forestry research and academic education and training.

Within the Earth Sensing Summer School 2025 in San Vito di Cadore (Italy), a student project group conducted forest point cloud acquisitions using multiple terrestrial laser scanning systems operated in both static and kinematic modes, complemented by UAV-based laser scanning (ULS) data. In the presented study, we show results derived from the data of this campaign, focusing on data acquired with a RIEGL VZ-600i terrestrial laser scanner operated in both static and kinematic acquisition setups. Both data acquisitions are performed with long-baseline RTK GNSS to provide absolute georeferencing, although GNSS accuracy is severely degraded within the forest. The analysis is based on a representative forest plot of approximately 2500 m², including around 150 trees. The plot is dominated by coniferous species, primarily Picea abies (Norway spruce), and is located on sloped terrain with sparse understory vegetation.

We systematically compare the static and kinematic TLS acquisitions and the resulting point clouds with respect to acquisition time, data processing, point cloud completeness and occlusion effects. Furthermore, the point clouds are analyzed at the individual tree level, including semantic segmentation of individual trees and the derivation of key tree metrics. ULS data are used as a reference for the assessment of tree heights and the representation of upper canopy elements.

The data acquisition was performed by students unexperienced with TLS after giving a 30 min introduction to the device and TLS in forest environments. Comparing data acquisitions, the two kinematic acquisitions took ~10 min each, the static acquisition resulted in 22 scan positions and an acquisition time of 1 h 35 min. Preliminary results of the initial data inspection indicate that the kinematic point clouds provide a more complete representation of tree tops than static point clouds.

How to cite: Iglseder, A., Pöppl, F., Groiss, B., Bocaux, L., Brandolese, A., Brunetto, N., Li, F., Maes, L., Naito, C., Perry, N., Reato, I., Van Sebroeck Martins, B., Cabo, C., and Balestra, M.: Static and kinematic point clouds using a single terrestrial laser scanning system for forest structure characterization, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10738, https://doi.org/10.5194/egusphere-egu26-10738, 2026.

EGU26-10738

Download Close

How to cite: Kaharlytskyi, R., Robinson, D., and Guglielmi, R.: Radiometric Features Enable Cross-Platform (TLS-to-ULS) Generalization for Forest Structure Assessment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15759, https://doi.org/10.5194/egusphere-egu26-15759, 2026.

EGU26-15759

Download Close

Accurate quantification of single-tree structural attributes is essential for improving estimates of terrestrial carbon stocks and for supporting sustainable forest and urban tree management. While traditional forest inventory methods and advanced technologies, such as terrestrial laser scanning (TLS) provide high-quality measurements, their spatial and temporal coverage remains limited due to cost and logistical constraints. Citizen science offers an underexploited opportunity to complement expert-based data collection and enhance data availability at large scales.

We present an overview of recent advances in integrating citizen science with digital tools and remote sensing for single-tree assessment, with a particular focus on urban environments. Our contribution specifically explores the use of mobile applications, low-cost sensors, and participatory approaches to support crowdsourced identification of tree species diversity and mapping of vegetation carbon stocks in urban environments.

To this end, we developed Tree-Quest (TQ), a free citizen-science mobile application, designed to measure single-tree attributes, such as tree species (ID), tree height (TH) and stem diameter at breast height (DBH). We compiled a dataset comprising 700 measurements acquired from 30 volunteers across peri-urban landscapes located in the vicinity of Vienna. Volunteers achieved a mean absolute error (MAE) of 3 cm for DBH (R² = 0.97; rMAE = 6%) and 1.5 m for TH (R² = 0.91; rMAE = 11%), thus demonstrating comparable measurement accuracy with traditional forest inventory.

Our findings indicate the potential of citizen science to complement remote sensing estimates and forest inventory measurements, thus supporting climate adaptation strategies, and improving our understanding of tree-level carbon dynamics in urban environments, beyond traditional estimates derived from natural forest ecosystems.

How to cite: Hofhansl, F., Milenković, M., Weinacker, R., Sturn, T., Karanam, S., Georgieva, I., Wild, B., Pfeifer, N., Hollaus, M., Zappa, L., Bruckman, V. J., Mccallum, I., and Fritz, S.: Tree-Quest: A Citizen Science App for Collecting Single-Tree Attributes, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3906, https://doi.org/10.5194/egusphere-egu26-3906, 2026.

EGU26-3906

Download Close

Forests play a central role in biodiversity conservation, nutrient cycles, and climate regulation. At the same time, forests are increasingly affected by biodiversity loss, disruptions of nutrient cycles, and global warming. Forests respond to these pressures through a range of dynamics, including increased tree mortality, shifts in tree allometry, and changes in species composition. Accurate and detailed forest monitoring, which can track changes in these parameters, is therefore essential.

Mobile and terrestrial laser scanning (MLS and TLS) have proven to be among the most precise tools for assessing key forest characteristics such as forest structure, tree architecture, and aboveground biomass. However, these technologies are typically ground-based, and the resulting point clouds are strongly affected by sparse point density in the canopy and occlusion, i.e., when main branches block laser pulses. This leads to systematically vertically biased point densities and data gaps in the canopy.

Here, we present a novel methodological framework that integrates mobile and terrestrial laser scanning with canopy access methods to reduce occlusion and improve point cloud quality. We use modern arborist techniques to access tree crowns using ropes and harnesses. Once in the canopy, we abseil from two sides of the tree while carrying an MLS unit by hand. In addition, we distribute targets on branches and mount the TLS on custom-built platforms installed on lateral branches and along the main stem. This workflow is repeated several times per year to quantify changes in tree morphology, such as tree growth at the millimeter scale, and to derive accurate estimates of forest characteristics.

The resulting point clouds show a more homogeneous point density and substantially reduced occlusion. Consequently, estimates of tree allometry, volume, and biomass are significantly improved. These highly accurate digital forest representations can be extrapolated to larger spatial and temporal scales and used to calibrate and validate airborne and satellite remote sensing products.

This work demonstrates that accurately characterizing the structural complexity of trees and forests requires innovative measurement approaches to enable improved monitoring and sustainable forest management in a changing climate.

How to cite: Hepner, S., Wingate, V., and Ifejika Speranza, C.: Abseiling for science: integrating mobile and terrestrial laser scanning with arborist methods to improve point clouds, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16958, https://doi.org/10.5194/egusphere-egu26-16958, 2026.

EGU26-16958

Download Close

Accurate estimation of aboveground biomass (AGB) is essential for quantifying forest carbon stocks. Missions such as NASA’s Global Ecosystem Dynamics Investigation (GEDI) provide valuable forest structure data that can be converted into AGB estimates. The most robust approach relies on calibrating these metrics against National Forest Inventory (NFI) plots. However, GEDI’s sampling remains sparse in both space and time, limiting opportunities for local calibration and validation of biomass models [1]. To address these limitations, the forest and remote sensing community increasingly uses simulators to generate GEDI-like measurements at NFI locations. Among the available tools for emulating GEDI waveforms, the simulator developed by Steven Hancock [2] has been widely adopted. Yet, its accuracy and biases have not been systematically assessed beyond the initial test areas or against real GEDI observations. 

By evaluating the Hancock simulator across diverse French forests using high-density national airborne LiDAR data (LiDAR HD), this work investigates the validity of a globally developed tool when applied at the local scale. We quantify discrepancies between simulated and actual GEDI data with a focus on bias metric due to its potential propagation into downstream biomass models. Such errors may lead to significant over- or underestimation of carbon stocks. 

Our approach focuses on a bottom-up, empirical evaluation of GEDI-simulated metrics to diagnose local biases and their drivers. It does not provide a comprehensive review of the simulator's theoretical framework. Results reveal systematic structural biases of up to 1 m in RH metrics. We investigate these errors in relation to pulse shape, algorithms and beam energy differences, canopy cover, forest type, seasonal effects and topography. Finally, we propose correction strategies in which a multi-layer perceptron (MLP) is trained to adjust simulated RH metrics to better match real GEDI observations. Our findings provide practical recommendations for simulator users, implications for improving GEDI-based biomass estimation and insights for the design of future LiDAR missions.

[1] N. Besic, et al., “Using structural class pairing to address the spatial mismatch between GEDI measurements and NFI plots,” IEEE JSTARS, vol. 17, pp. 12854–12867, 2024. DOI: 10.1109/JSTARS.2024.3425431  

[2] S. Hancock, et al., “The GEDI simulator: A large-footprint waveform lidar simulator for calibration and validation of spaceborne missions,” Earth Space Sci., vol. 6, no. 2, pp. 294–310, 2019. DOI: 10.1029/2018EA000506

How to cite: Behloul, S., Besic, N., Hancock, S., Fayad, I., Vega, C., Durrieu, S., Renaud, J.-P., and Ciais, P.: Towards using GEDI in the French NFI-based AGB estimations: Systematic Assessment of Plot-level Simulations Using Nationwide Airborne LiDAR HD, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9762, https://doi.org/10.5194/egusphere-egu26-9762, 2026.

EGU26-9762

Download Close

Forests are undergoing rapid structural changes driven by droughts, storms, pests, and long-term climatic stress. Quantifying these dynamics requires detailed three-dimensional information on forest structure, biomass, and diversity. Drone-based, mobile, and terrestrial LiDAR have become essential for acquiring such data, yet their broader use remains constrained by fragmented processing workflows, heterogeneous data formats, limited cross-platform integration, and restricted access to scalable computing resources.

3Dtrees.earth is an open, cloud-based platform designed to overcome these barriers through integrated, scalable, and reproducible extraction of forest information from multi-platform LiDAR data. The platform supports standardized processing of LiDAR point clouds terrestrial (TLS), uncrewed aerial (ULS), and mobile laser scanning (MLS), applying modular pipelines for data harmonization, instance and semantic segmentation, species prediction, and structural trait extraction. Building on the recent advances in 3D deep learning, 3Dtrees.earth integrates state-of-the-art models for single-tree detection, species classification, and tree-level inventory generation. 

All processing workflows are containerized and deployed via Galaxy Europe, enabling users to analyze large LiDAR datasets without local software or dedicated computing resources. A core design principle is accessibility combined with transparency: users interact through web-based workflows and shared histories that fully document tool versions, parameters, and data provenance, ensuring reproducibility across regions, sensors, and applications. Derived products - including canopy height models, tree-level inventories, biomass estimates, and structural diversity indices - are curated according to FAIR principles with persistent storage, rich metadata, and standardized access interfaces. 

Co-developed with forest managers, researchers, public agencies, and AI developers, 3Dtrees.earth serves multiple communities. Practitioners gain access to operational products such as tree density and height maps, gap and deadwood indicators, and structural diversity metrics that can be directly integrated into management planning. Scientists benefit from harmonized benchmark datasets and reproducible workflows that facilitate method comparison across regions, forest types, and sensor platforms. AI developers are provided with large-scale, well-labeled 3D forest datasets for training and evaluating generalizable forest analytics models.

By lowering technical barriers and standardizing 3D forest analytics, 3Dtrees.earth aims to accelerate the integration of LiDAR-derived structural information into forest research, monitoring, and management at a global scale.

How to cite: Gerberding, K., Vajna-Jehle, J., Kattenborn, T., Scharinger, C., Lusk, D., Brede, B., Sperlich, M., Seifert, T., Grüning, B., Puliti, S., and Frey, J.: 3Dtrees.earth - A platform for accessing, analyzing, and visualizing LiDAR data in forest environments, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20434, https://doi.org/10.5194/egusphere-egu26-20434, 2026.

EGU26-20434

Download Close

Canopy height is one of the most important forest structural variable, generated by remote sensing for applications such as forest inventory, sustainable management, carbon assessments and disturbance monitoring. But generating accurate and frequent canopy height maps over large areas remains a challenge. Airborne laser scanning (ALS) provides highly reliable and detailed canopy height models, repeated acquisitions are often limited by cost and availability. With the spaceborne LiDAR from NASA’s GEDI (Global Ecosystem Dynamics Investigation) there is globally distributed relative canopy height observations but these are affected by noise and terrain-related uncertainties. This has created a gap for generating consistent, wall-to-wall canopy height products at annual timescales. With the growing availability of high temporal multispectral imagery from satellite missions such as Sentinel 2 raises the question to what extent dense optical time series can be used to support operational canopy height mapping when combined with LiDAR observations. In this work, we investigate the potential and limitations of using dense Sentinel-2 time series in fusion with LiDAR data for generating CHMs at 10m resolution across Bavaria. We downloaded and processed all available Sentinel-2 imagery for Bavaria from 2019 to 2024 (~9 TB) by correcting it radiometrically and geometrically and regridding it into a non-overlapping datacube structure. From this datacube, we generated multi-seasonal composites and interpolated time series to capture forest phenology at the pixel level. Using the Sentinel-2 time series products created, we trained a CNN based model (UNet) with (i) high-resolution ALS derived CHMs and (ii) GEDI waveform relative height metrics as reference data. Preliminary results demonstrate that integrating multi-seasonal Sentinel 2 information substantially improves model performance at generating annual CHMs at 10m reoslution. At the same time, we also highlight limitations related to the choice of training supervision data and that models trained with higher quality ALS based CHMs yield the most reliable canopy estimates whereas GEDI based supervision can introduce increased uncertainty in heterogeneous terrain and areas with limited footprint samples. We thus provide a technically workable, scalable and semi-automatic forest canopy monitoring approach which - once trained for a region - uses only open-scource data, making it highly reproducible.

How to cite: Nagarajan, S., Marty, M., Ginzler, C., and Senf, C.: Mapping canopy heights from space using deep learning with Sentinel-2 time series and LiDAR data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16942, https://doi.org/10.5194/egusphere-egu26-16942, 2026.

EGU26-16942

Download Close

Europe’s forests play a critical role as carbon sinks, yet their capacity for climate change mitigation is increasingly threatened by rising disturbances and increasing demand for wood. Reliable data on disturbance rates and trends are thus needed. Several Earth observation-based products have been released in recent years, providing an outstanding source of information on forest change. However, many of these products lack proper quantification of accuracies, rendering rates and trends derived from them uncertain. We address this problem by developing a new database of forest disturbances for Europe, based on consistent manual interpretation of satellite imagery. Using this database, we derive robust annual disturbance rates at both national and regional scales, as well as trends over time. We further compare our sample-based estimates with state-of-the art map-based products, showing significant differences in map accuracies and thus area and trend estimates. We finally provide a framework for incorporating different map products into an ensemble estimate with well quantified uncertainties. Our results underscore the need for consistent, transparent, and independent reference data, and highlight that relying on a single map product might lead to biased conclusions about forest change in Europe.

How to cite: Senf, C., Wieland-Glasmann, F., Kowalski, K., and Viana-Soto, A.: All maps are wrong, but some are useful: Benchmarking European forest disturbance products using a consistent reference database, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8189, https://doi.org/10.5194/egusphere-egu26-8189, 2026.

EGU26-8189

Download Close

Windthrow events, characterized by tree uprooting or breakage by strong winds, can result in substantial structural change and biomass loss, altering biodiversity and carbon dynamics. They have led to severe disturbance of native forests in temperate Australia in recent years, and yet their extent and impacts remain largely unquantified. This study aims to advance the monitoring and understanding of windthrow dynamics by integrating remote sensing technologies, machine learning, and field-based data to quantify windthrow severity and associated impacts on the structure of native eucalypt-dominated forests. As a case-study example, it focuses on a major storm event on 9 June 2021 that affected an estimated 40,000 ha of the Wombat State Forest in Victoria, southeastern Australia. The study employs high to very-high resolution satellite and aerial imagery [PlanetScope (3m), NearMap (7.5cm)] and derived indices (Normalized Difference Vegetation Index, NDVI; Blue Normalized Difference Vegetation Index, BNDVI) to nominally map None, Low, Medium, and High severity windthrow zones. These zones were used in stratified random sampling to select 650 (30m×30m) plots in the NearMap imagery, which were analyzed for change in canopy cover using a machine learning workflow involving a Random Forest model. The workflow provided canopy cover reduction estimates from pre- to post-event scenario with high accuracy (96.9%), precision (92.5%), recall (92.8%,) and F1-score (92.68%) across plots in high windthrow severity locations (260) initially and significantly reduced the amount of time and labour for this task. Building on these canopy-level estimates, the final stage will upscale damage quantification across the entire Wombat State Forest by training PlanetScope imagery with very-high resolution canopy cover estimates data from NearMap while employing machine learning models integrating spectral predictors (including dNDVI, dBNDVI, and key multispectral bands). This will produce a high-resolution windthrow severity map, enabling an accurate assessment of windthrow severity across the large and heterogeneous landscape. These outputs will enable biomass-loss estimation from canopy and tree-fall metrics, and will support risk models that integrate remote sensing, biophysical variables, and climate data to predict windthrow susceptibility across the landscape of Australian temperate forests.

How to cite: Singh, A., Bennett, L., Hinko-Najera, N., and Fairman, T.: Optimizing remote sensing workflows using machine learning techniques to quantify windthrow severity across Australian temperate forest., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-289, https://doi.org/10.5194/egusphere-egu26-289, 2026.

EGU26-289

Download Close

Integration of Digital Cover Photography and Multi-Source Remote Sensing Approaches for Forest Canopy Cover Estimation in Southwest Ethiopia

Mohammed S Ozigis¹, Byongjun Hwang¹, Thierno Bachir Sy², Matthew Snell² and Desyalew Fantaye³, Adrian Wood²

¹Department of Biological and Geographical Sciences, School of Applied Sciences, University of Huddersfield, Queensgate, Huddersfield HD1 3HD, UK

²Department of Management, Huddersfield Business School, University of Huddersfield, Queensgate, Huddersfield HD1 3HD, UK

³Ethio-Wetlands and Natural Resource Association, Ethiopia.

Abstract

Forest loss through deforestation and degradation is an important factor shaping the global climate change and its attendant short- and long-term impacts. Forest canopy cover (FCC) estimation has evolved to become an important and essential parameter for establishing degraded forest. Recent advances in Earth Observation (EO) satellite sensors have opened a new frontier in estimating, mapping and monitoring forest cover using high-resolution imagery and machine learning (ML). These have typically relied on canopy cover extracted from aerial or satellite images to establish baseline reference data. While several studies have alluded to the suitability of field-based Digital Cover Photography (DCP) for forest canopy characterization, none have explored their potential in predicting forest canopy cover through its integration with EO satellite data in-combination with ML methods. This study explores the integration of multi-sensor EO data from Sentinel-1 and Sentinel-2, along with topographic information (Digital Surface Model, DSM) and field-based DCP canopy cover measurements, to enhance the accuracy of EO-derived forest canopy cover estimates in Southwest Ethiopia. Over 1,000 DCP measurements were obtained during a field campaign conducted from January to February 2025 in southwest Ethiopia. The DCP data were then used to train both simple linear regression and advanced ML regression models to predict and map canopy cover. Initial results suggest that the integration of Sentinel-2 raw spectral bands with DSM produced the most accurate canopy cover estimates, with Random Forest (RF) model achieving the highest R² (0.63) and lowest RMSE (7.4%). In addition, the XGBoost model achieved R² of 0.59 and an RMSE of 7.9%, while the Generalized Additive Model (GAM) outperformed the other linear models tested, producing a higher R² (0.52) and a lower RMSE (8.63%). This study demonstrates that integrating field-based DCP measurements with EO data provides a more accurate approach for estimating baseline forest canopy cover, thereby advancing existing knowledge and methodologies for EO-based canopy cover mapping.

 Keywords: Forest Canopy Cover, Deforestation, Forest Degradation, XGBoost, Random Forest, Digital Cover Photography, Sentinel-1, Sentinel-2

How to cite: Ozigis, M., Hwang, B., Bachir, T., Snell, M., Fantaye, D., and Wood, A.: Integration of Digital Cover Photography and Multi-Source Remote Sensing Approaches for Forest Canopy Cover Estimation in Southwest Ethiopia, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10987, https://doi.org/10.5194/egusphere-egu26-10987, 2026.

EGU26-10987

Download Close

We recently implemented Sentinel-1 radar-based weekly, high-resolution (10m pixel spacing) forest disturbance alerts for Europe and processed it for 2020-2025 (Reiche et al., 2021; van der Woude et al., in revision), providing the basis to analyze disturbance seasonality. However, identifying and interpreting seasonal disturbance patterns requires disturbance type information, as seasonal patterns can vary greatly between and within disturbance types (Wohlgemuth et al., 2022). The European Forest Disturbance Atlas (EFDA) is based on Landsat imagery and provides thematically detailed annual maps of forest disturbance, distinguishing between three disturbance types: fire, wind/bark beetle and harvest (Viana-Soto and Senf, 2025).

We analyzed seasonal patterns of forest disturbance across Europe by combining radar-based RADD Europe forest disturbance alerts with optical-based EFDA forest disturbance type information. Disturbance alerts were overlaid with disturbance type maps, aggregated to an ~20 km hexagonal grid, and summarized as mean disturbed area per day of year over a 4.5-year period from January 2020 to June 2024. We characterized disturbance seasonality using three complementary indicators: magnitude, timing, and modality. Seasonal magnitude was quantified using a seasonality index that measures the temporal concentration of disturbed area relative to a uniform distribution. Timing was described by deriving the mean day of year of disturbance occurrence. Modality was defined as the number of seasonal disturbance peaks, distinguishing between uni-, bi-, and multi-peaked patterns.

Our results showed strong contrasts in seasonal disturbance regimes across disturbance types. Fire exhibited the greatest seasonal magnitude, with disturbances primarily occurring during summer months. Wind and bark beetle disturbances were most concentrated in spring, while harvest-related disturbances were more evenly spread throughout the year. Substantial within-type variability was also observed, particularly for harvest, where differences in management practices between countries and regions lead to pronounced spatial variation in timing and a higher prevalence of bi- and multi-peaked seasonal patterns.

We emphasize the benefits of combining radar- and optical-based disturbance products for improved disturbance characterization, allowing for a better understanding of disturbance seasonality, as well as interactions between disturbance types and disturbance sequences. The launch of Sentinel-1C and 1D and the continued availability of optical satellite missions such as Sentinel-2 and Landsat will be crucial in reducing uncertainties in the analysis of forest disturbance seasonality.

Reiche, J., Mullissa, A., Slagter, B., Gou, Y., Tsendbazar, N.E., Odongo-Braun, C., Vollrath, A., Weisse, M.J., Stolle, F., Pickens, A., Donchyts, G., Clinton, N., Gorelick, N., Herold, M., 2021. Forest disturbance alerts for the Congo Basin using Sentinel-1. Environmental Research Letters 16. https://doi.org/10.1088/1748-9326/abd0a8

Senf, C., Seidl, R., 2021. Mapping the forest disturbance regimes of Europe. Nature Sustainability 4, 63–70. https://doi.org/10.1038/s41893-020-00609-y

Van der Woude, S., J. Reiche, J. Balling, G.-J. Nabuurs, F. Sterck, A.-J. Welsink, B. Slagter, and M. Herold  (2025). “Near real-time European forest disturbance alerts using Sentinel-1”. In revision.

Viana-Soto, A., Senf, C., 2025. The European Forest Disturbance Atlas: a forest disturbance monitoring system using the Landsat archive. https://doi.org/10.5194/essd-17-2373-2025

Wohlgemuth, T., Jentsch, A., Seidl, R. (Eds.), 2022. Disturbance Ecology, Landscape Series. Springer International Publishing, Cham. https://doi.org/10.1007/978-3-030-98756-5

How to cite: van der Woude, S., Viana-Soto, A., Reiche, J., Senf, C., Nabuurs, G.-J., Sterck, F., and Herold, M.: Uncovering seasonal patterns in European forest disturbance regimes, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-22624, https://doi.org/10.5194/egusphere-egu26-22624, 2026.

EGU26-22624

Download Close

Europe’s forests are under increasing pressure from natural disturbances, while there is growing demand for wood. With disturbances expected to further intensify under climate change, quantifying the impact of disturbances on forest resources has thus become a key challenge for Earth Observation. In particular, there is a strong need for spatially explicit information on how forests change over time, driven by disturbance and post-disturbance recovery. Most existing large-scale assessments derive this information using spectral indices. However, spectral indices tend to saturate in dense canopies while being limited in their ability to capture changes under mixed land cover conditions. In that sense, using estimates of tree cover might provide a clearer signal of forest canopy changes. Building on previous applications of spectral unmixing for mapping forest cover in European forest ecosystems (Mandl et al. 2024, Viana-Soto et al. 2022, Senf et al. 2019), we here present a novel framework for mapping tree cover across all of Europe’s forests. Specifically, we (i) estimate annual tree cover fractions from 1985 to 2024 at 30 m spatial resolution using spectral unmixing of Landsat data, (ii) assess the temporal consistency and accuracy of these estimates across Europe’s forests, and (iii) characterise tree cover loss from disturbance and post-disturbance tree cover gain, thereby distinguishing it from land use changes. As a data basis, we built a consistent Landsat data cube of atmospherically and topographically corrected Landsat surface reflectance data, including cloud and shadow masking, totalling to 363,088 images. Annual gap-free best available pixel composites were generated by selecting high-quality observations closest to 1^st of August, minimizing phenological effects and ensure intra-annual consistency. Based on these composites, we developed a multi-year endmember library consisting of pure and temporally stable pixels representing treed and non-treed land cover types (herbaceous, shrubs, bare ground, and shadow). We collected endmember spectra by randomly sampling pure pixels from LUCAS database, providing in-situ land cover information across Europe, and by cross-checking their spectral–temporal stability and cover proportions using high-resolution imagery. To simulate the full range of possible spectral mixtures, we generated synthetic training datasets by linearly combining endmember spectra in known proportions. Lastly, these mixtures and their associated ratios were used to train regression models predicting annual tree cover fractions. Preliminary results indicate that the spectral unmixing framework enables consistent mapping of annual tree cover fractions across Europe, capturing losses associated with disturbance or land use conversion and gradual gains reflecting post-disturbance recovery. By delivering harmonized annual maps of tree cover fractions for Europe, this work advances continental-scale forest monitoring efforts and supports policy frameworks for forest adaptation to climate change.

How to cite: Viana Soto, A., Kowalski, K., Mandl, L., and Senf, C.: Mapping multi-decadal tree cover change from disturbances across Europe using spectral unmixing of Landsat time series, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9660, https://doi.org/10.5194/egusphere-egu26-9660, 2026.

EGU26-9660

Download Close

Non-stand replacing disturbances (NSRs) are events that do not result in complete removal of forest stands and generally occur at a low intensity over an extended period (e.g., insect infestation), or at spatially variable intensities over shorter periods (e.g., windthrow). Forest structural change associated with NSRs can impact both timber supply and ecosystem services, necessitating the need for both detection of NSRs and characterization of their impact.  The increased accessibility of high frequency revisit, medium spatial resolution satellite imagery, has led to a subsequent increase in algorithms designed to detect sub-annual change across broad spatial scales. The Bayesian Estimator of Abrupt change, Seasonal change, and Trend (BEAST) has shown promise in both detecting NSRs on a sub-annual basis and estimating forest structure changes indicating the potential for continuous characterization of NSRs. This study assesses the impact of NSRs on forest structure across a dry interior forest in Western Canada with a specific case study on aspen leafminer (Phyllocnistis populiella) and two-year budworm (Choristoneura biennis). To do so, the BEAST algorithm was applied to a time-series of medium resolution optical satellite imagery for six disturbance-sensitive indices for the time period 2013-2021 to generate predictor variables capturing annual phenological variation (i.e., amplitude, slope, and trend). Three LiDAR derived forest attributes were modeled (i.e., canopy cover, height and height variability) using predictors variables as inputs (R² values between 0.5 - 0.7).  These models were then applied across the study areas, and changes in structure estimated over NSR impacted stands. Results showed changes in forest structure over the period of continued NSR events, including an 11% decline in canopy cover. This approach enables the structural change caused by NSRs to be more rapidly identified, providing forest practitioners with approaches to better identify areas in need of intervention.

How to cite: Brown, M., Coops, N., Mulverhill, C., and Achim, A.: Characterizing the impact of non-stand replacing disturbances on LiDAR based forest structure using a Harmonized Landsat-Sentinel-2 time-series, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-283, https://doi.org/10.5194/egusphere-egu26-283, 2026.

EGU26-283

Download Close

Extreme windthrow events have increasingly affected mountain forest ecosystems, highlighting the need for robust monitoring approaches to assess post-disturbance recovery and support adaptive forest management. In October 2018, Storm Vaia damaged more than 42,500 ha of forest across northern Italy, causing an estimated 16.5 million m³ of windthrown timber and providing a regional-scale case study to evaluate forest recovery dynamics. This study aimed to investigate post-windthrow vegetation trajectories, with a particular focus on the effects of edge forest stand characteristics and salvage logging strategies on vegetation recovery. Vegetation dynamics were reconstructed using Sentinel-2 Normalized Difference Vegetation Index (NDVI) time series (2016–2025) for 148 permanent plots established and surveyed in the field within an extensive monitoring programme. Temporal trajectories were interpolated and classified using a temporal similarity clustering approach. Seasonal behaviour was characterised by deriving phenological metrics (start, peak, and end of the growing season) for individual plots and cluster-level confidence intervals. Statistically significant differences were tested both among clusters and across successive years. We used a Multivariate Factor Analysis (MFA) to integrate topographic variables, forest stand characteristics, and salvage logging methods to assess their influence on the identified trajectories. Our analysis identified four distinct vegetation recovery trajectories. One trajectory, representing the most severely impacted areas and associated with herbaceous-dominated stages, exhibited a pronounced post-disturbance reduction in NDVI (approximately 45%) while maintaining a high seasonal amplitude in the later years. A contrasting trajectory showed progressively dampened seasonal oscillations, with a 2024 amplitude of about 0.45 and a mean NDVI recovering to approximately 0.77, reflecting a more stable and less seasonally variable recovery pattern. The timing of the peak growing season was significantly altered across the study period, with post-hoc comparisons showing that the immediate post-disturbance years (2019 and 2021) differed markedly from both the pre-storm baseline (2018) and subsequent years. The MFA showed that edge forest stand characteristics explained 33% of the observed variance in vegetation trajectories, while salvage logging strategies exhibited limited explanatory power. Overall, our results demonstrate the potential of dense optical time series to reconstruct complex post-disturbance vegetation dynamics and highlight the value of integrating satellite observations with ground-based surveys to improve the interpretation of recovery trajectories in mountain forest ecosystems.

How to cite: Grande, C., Candotti, A., Stein, M., Alberti, G., Lingua, E., and Tomelleri, E.: Integrating Sentinel-2 time series with in-situ monitoring to evaluate post-windthrow vegetation recovery trajectories and management impacts, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19352, https://doi.org/10.5194/egusphere-egu26-19352, 2026.

EGU26-19352

Download Close

Understanding when and where satellite disturbance products miss forest change is critical for reliable large-scale monitoring. Landsat-based products such as the National Land Cover Database (NLCD) provide long-term coverage, but their disturbance signals are primarily driven by spectral change and can fail to capture canopy structural losses, especially when changes are subtle, fragmented, or occur in short/transitional vegetation. Here we use airborne LiDAR canopy height models (CHMs) from 32 NEON sites across the continent of United States as an independent structural benchmark to quantify Landsat detection limits.

We compared multi-year LiDAR-derived canopy height change with temporally matched NLCD disturbance layers. From CHMs we derived pre-disturbance canopy height, canopy-height loss (severity), and patch size. We quantified Landsat recall at pixel scale, and evaluated how recall varies with height, severity, forest type, and disturbance patch size.

LiDAR revealed systematic detection biases in Landsat disturbance detection. Recall increased with canopy height and remained low for low-to-moderate structural losses, rising sharply only for the most severe canopy-height reductions. At the patch scale, detection fraction increased with disturbance size: small patches were rarely detected, whereas larger patches showed substantially higher detection. Detection agreement varied across forest types and was weakest in open-canopy woodlands and transitional vegetation. In conclusion, Landsat disturbance products preferentially capture large, high-severity canopy-loss events while frequently omitting smaller and lower-severity structural changes evident in LiDAR.

How to cite: Yang, X. and Jucker, T.: Structural Forest Disturbances Revealed by LiDAR: Limits of Landsat Detection, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2083, https://doi.org/10.5194/egusphere-egu26-2083, 2026.

EGU26-2083

Download Close

How to cite: Müller, F., Bastos, A., and Camps-Valls, G.: Balancing Spatial, Spectral, and Temporal Information: Which Dimension Drives Deep Learning Performance in Forest Disturbance Classification?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1823, https://doi.org/10.5194/egusphere-egu26-1823, 2026.

EGU26-1823

Download Close