Well-aligned point cloud time series data generated with Unmanned Aerial Vehicles (UAVs) can be a significant asset to geoscientists.
Practitioners benefit from multi-temporal point clouds with high comparative accuracy, e.g. to evaluate landscape changes after landslides and quantify mass wasting.
Two approaches are usually applied to achieve the accurate alignment of point clouds: indirect and direct georeferencing.
Indirect georeferencing uses well distributed Ground Control Points (GCPs) in the study area.
While this method significantly enhances the precision and accuracy of time series point clouds, the placement and measurement of GCPs are time-intensive and may even be impossible in difficult terrain.
Direct georeferencing depends on highly precise and accurate location information embedded in images, which is often viable only with expensive real-time kinematic (RTK) positioning equipment or post-processed kinematic (PPK) services.
Beyond the extra cost, this approach faces the same challenges as indirect georeferencing, particularly in the placement of equipment and scalability for large areas.
Recent research has introduced an alternative method called Co-Alignment, which enables the alignment of point clouds with high local precision without GCPs and RTK data. Moreover, when GCPs or RTK are used, co-alignment can further enhance accuracy of the point cloud alignment.
This method aligns multiple point clouds with good local precision without requiring GCPs or RTK equipment, though it lacks global accuracy.
The workflow uses common, unchanged features in the study area, such as anthropogenic structures or boulders, to establish spatial references across multiple epochs using computer vision algorithms.
We developed FACA - Fully Automated Co-Alignment to implement the Co-Alignment workflow.
With FACA, we aim to offer easy access to a scalable point cloud alignment method.
FACA is automatable from the command line and user-friendly through a custom graphical user interface, making it adaptable to common point cloud generation workflows.
Released as open-source software under the GNU General Public License v3, FACA is freely accessible and modifiable to meet diverse user requirements.
By integrating with Agisoft Metashape Professional, FACA leverages advanced photogrammetric features to enhance performance and output quality.
We present the FACA workflow, emphasizing its ease of use, scalability, performance, supported by results from data acquired at Germany's Baltic Sea coast and in Svalbard.
Furthermore, we discuss the potential for custom software solutions to further improve and expand the workflow’s capabilities.
How to cite: Schüßler, N., Torizin, J., Gunkel, C., Fuchs, M., Schütze, K., Tiepolt, L., and Kuhn, D.: An Introduction to Fully Automated Co-Alignment - FACA, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4331, https://doi.org/10.5194/egusphere-egu25-4331, 2025.
The assemblage of multiple sensors on Unmanned Aerial Systems (UAS) to collect high resolution geospatial data represents one of the most significant advances in remote sensing, including oceanographic applications. Coastal inundation of pelagic Sargassum has been thoroughly documented as a natural hazard that jeopardizes the ecological integrity of coastal ecosystems, unbalancing several livelihoods and local economies. Sargassum patches (rafts) are drifted offshore by surface ocean currents, with distinct drivers at different geographic and time scales. UAVs have revolutionized the immediate local remote sensing of Sargassum as they can identify rafts that are expected to reach the coast in terms of hours, becoming a strategic tool for rapidly management actions, bridging the on-site actions with high and medium resolution satellite detections. To obtain primary data on the extent, frequency, and magnitude of floating and beached Sargassum in the Mexican Caribbean, a rapid assessment protocol based on aerial photogrammetric techniques was implemented in the Yucatan Peninsula. We documented the arrival of sargassum rafts in the nearshore environment used to perform statistical comparisons with other remote sensing products. High resolution orthomosaics, DSMs, and 3D reality models were created to document the extent and quantity of beached Sargassum and the contiguous “brown tide” areas. Floating sargassum rafts were also identified in real time using long-range telemetry UAVs between 2 and 20 km offshore, that were consistent with field-based observations. Ocean circulation model outputs are also presented, which demonstrate that including UAV-mounted multi-sensors data acquisition is fundamental towards a comprehensive description and monitoring of the Sargassum coastal dynamics. These results strongly suggest that UAV-derived cartographic products represent an efficient tool for Sargassum-management actions, downscaling satellite detections and linking them with local observations, a strategy that needs to keep addressing as the future research agenda in Operational oceanography.
How to cite: Gudino-Elizondo, N., Cuevas, E., Uribe-Martinez, A., Garcia-Nava, H., Flores-Vidal, X., and Avendaño-Gastelum, O.: Multiple airborne sensors to monitor rafts and beached Sargassum in the Mexican Caribbean: Documenting different UAVs applications for management actions., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15301, https://doi.org/10.5194/egusphere-egu25-15301, 2025.
The stability of rock masses is crucial for the safety of hydraulic engineering, as the integrity of the rock mass directly influences the stability of structures such as dams, reservoirs, and tunnels. Accurate extraction and orientation of rock mass discontinuities plays a key role in stability analysis, providing essential geometric data for assessing rock mass behavior. However, traditional manual measurement methods used to extract these orientations are not only time-consuming and labor-intensive but also fraught with safety risks, especially when working on large and steep slopes. These limitations hinder the efficiency and accuracy of rock mass stability assessments.
To address these challenges, this paper proposes a novel approach for acquiring 3D rock mass scenes using unmanned aerial vehicles (UAVs), coupled with oblique photogrammetry technology for 3D scene reconstruction. With UAVs equipped with high-resolution cameras to capture image sequences from various angles, the Structure from Motion (SfM) algorithm is then applied to reconstruct the 3D scene. This method allows for the generation of high-precision point cloud data through geometric uniform sampling, ensuring accurate representation of rock mass. Once the 3D scene is reconstructed, local geometric features (including surface curvature, planarity, scattering, and verticality) are calculated based on neighborhood search. Combined with RGB texture information, machine learning method is employed to analyze the importance of these features, and further identify and differentiate rock mass features from vegetation and outliers within the large-scale slope scene, followed by a region-growing and merging algorithm for the segmentation of rock mass patches. For each individual patch, a local planar coordinate system is established to generate a grayscale image, which is then used for edge detection to identify structural boundaries. Following this, line extraction is carried out using an energy-optimization-based graph cut algorithm, and the closed contours of the structural patches are delineated through vectorization, ensuring an accurate and detailed mapping of the rock mass structure.
The effectiveness of the proposed method was validated through experiments conducted on a large-scale rock mass slope scene. The results demonstrate that the method can accurately extract the rock mass structural regions, identify the fracture network, and provide crucial geometric features, such as dip, strike, and trace information for each structural plane. The extracted features significantly contribute to evaluating the structural integrity and stability of large-scale slopes, offering a more efficient, accurate, and safer alternative to traditional manual measurement methods. Moreover, this method can be applied to a wide range of geological environments, providing a valuable tool for real-time monitoring and assessment in engineering projects.
How to cite: Han, X., Zhai, R., Huang, Y., and Ding, B.: Extraction and Orientation Analysis of Rock Mass Discontinuities Using UAV-Assisted Photogrammetry, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2736, https://doi.org/10.5194/egusphere-egu25-2736, 2025.
The project's goal is to create a software tool for detecting and predicting a higher form of (rill) erosion on agricultural land. The planned tool's innovative potential is the use of neural networks on the joint remote sensing and erosion-hydrological modelling data. Morphological parameters and erosion-hydrological causal event response thus enhance common inputs for the neural network-driven semantic segmentation.
By combining morphological parameters, event-based hydrological responses, and a calculated critical water layer thickness (hcrit) from physical SMODERP model - the threshold at which rill erosion begins - the tool enhances the precision of high-risk area delineation, supporting smart agriculture and climate adaptation.
The project utilizes a unique dataset of manually digitized erosion rills from over 20 years of aerial orthophotos, enabling comprehensive training of neural networks. Multi-resolution data, including satellite imagery, aerial orthophotos, and UAV images, are combined to identify and refine morphological properties critical for rill erosion detection. Several types of neural networks were tested, notably FCN, U-Net, SegNet, DeepLabv3+, to evaluate their effectiveness in handling diverse input data and optimizing predictive accuracy. Automated workflows for dataset expansion and retraining ensure adaptability to new data.
Validation of the model will be performed using the original dataset of manually digitized erosion rills as a benchmark for accuracy. By comparing the predicted rill locations with this dataset, the model’s performance can be rigorously evaluated and adjusted. Real-time erosion event mapping, supported by the Agricultural Land Erosion Monitoring system, will complement this process by incorporating contemporary data to further enhance model reliability. This innovative tool addresses gaps in existing methods by combining predictive capabilities with detailed spatial data, improving erosion detection accuracy for sustainable land management under changing climatic conditions.
The research is funded by the Technological Agency of the Czech Republic research project (TQ03000408)- Detection of Increased Erosion Damage Using Neural Networks on a Combination of Remote Sensing Imagery and Erosion-Hydrological Modeling and an internal student CTU grant (SGS23/155/OHK1/3T/11).
How to cite: Tejkl, A., Kavka, P., Pesek, O., and Landa, M.: Detection of Increased Erosion Damage Using Neural Networks on a Combination of Remote Sensing Imagery and Erosion-Hydrological Modeling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5875, https://doi.org/10.5194/egusphere-egu25-5875, 2025.
Raw terrain data acquired by sensing techniques such as SfM or LiDAR typically contain non-terrain components that require filtering, such as vegetation occlusion and other non-terrain features. While filtering helps remove non-terrain data, it can introduce discontinuities and local voids in the dataset. These data gaps can affect both the completeness of the terrain representation and subsequent analysis tasks. Therefore, it is crucial to develop effective terrain data completion methods for reliable terrain analysis.
Traditional terrain data completion methods, such as interpolation-based algorithms and Poisson surface reconstruction, typically model and optimize data continuity from a mathematical perspective. Although these methods address local voids to some extent, they generally fail to exploit terrain features and semantic information, limiting their effectiveness in completing complex terrain scenarios.
To address these issues, we propose a deep learning-based framework for terrain data completion. Our methodology explores different neural network designs with supervised and unsupervised learning, incorporating geomorphological constraints to improve terrain feature representation and semantic understanding. The framework leverages the representational capabilities of deep learning to improve the robustness of terrain data completion, contributing to a more consistent and reliable basis for subsequent terrain analysis and applications.
How to cite: Lu, T. and Jaboyedoff, M.: Deep Learning-based Terrain Data Completion with Geomorphological Constraints, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12156, https://doi.org/10.5194/egusphere-egu25-12156, 2025.
Different people perceive landscapes in various ways depending on their cultural and social background as well as their own values. However, characteristics inherent in the landscapes also have an impact on their perceived beauty. Accordingly, it remains unclear to what extent personal assessments and landscape properties influence how much people appreciate landscapes. In this study, we had 50 test subjects evaluate alpine landscapes represented by 30 historic and recent rendered pictures each. Since the recent pictures should display the exact same part of the landscape as the historic ones, digital elevation models (DEMs) and orthophotos were used to render the current scene in the same greyscale range as in the historic photographs. Additionally, DEMs and landcover maps for the captured images were analysed. These results were used to explain the test subjects’ values of the appreciation of and desire to travel to the landscapes using linear mixed models.
The key finding is that perceived landscape attractiveness depends more on the people assessing the landscapes than the landscape characteristics themselves. The number of distance zones (surrounding, near, middle and far zone) present in the viewshed has a significant impact on the appreciation of the landscape. The maximum slope affects the desire to travel to the landscapes, and the relief energy, the viewshed size and the ratio of the recently glaciated area influence both the appreciation of and the desire to travel to the landscape. Furthermore, the historic photographs are perceived as more beautiful than the recent rendered ones. Taking into account the ratio of the glaciated area, this difference is even more pronounced for the desire to travel to the landscape. The bigger the difference in the glaciated area between the historic and recent image is – hence the more glacier has melted – the more the test subjects desire to travel to the scene shown in the historic picture than in the recent one.
How to cite: Koschmieder, M., Pfeuffer, C., Mikolka-Flöry, S., and Heckmann, T.: Landscape attractiveness – It depends on the observer, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11270, https://doi.org/10.5194/egusphere-egu25-11270, 2025.
Along many coastlines worldwide, a variety of direct and indirect anthropogenic influences are influencing natural processes of coastal erosion and deposition. Both traditional change reconstruction and monitoring techniques (e.g. repeat surveys) and increasingly sophisticated approaches (e.g. photogrammetry, LiDAR, drone imagery) require specialist knowledge and equipment, can be time consuming to apply, and may be restricted to assessing relatively recent changes over short timeframes (e.g. typically years to a few decades). Here, we evaluate the potential for archival visual sources - maps, paintings, geological sketches, and historical photographs – to help document changes in the coastal environment of Ceredigion County, west Wales, over the past 100-150 years. Two extant sites of geoscientist interest, both located within 20 km of Aberystwyth, were investigated: Harp Rock (Craig y Delyn), which represents the westward-dipping limb of a synclinal fold, and Monk’s Cave (Twll Twrw), which has essentially now developed into a coastal arch. Egg Rock (Tŵr Gweno), a coastal stack which was previously located near to Monk’s Cave but has since disappeared, was also investigated. All three sites were well-known tourist attractions in the late 1800s and early 1900s, and various maps, paintings, sketches and photographs help to provide both qualitative and quantifiable insights into the nature of coastal change, including the sequencing, rates, and timing of key changes, as well as volumes of mass loss. For example, Harp Rock is retreating landward as sandstone strata of ~37 cm thickness are removed by wave action and mass movement; for every 1 m2 of stratal loss, a mass of 858.4 kg is removed. For Monk’s Cave, the average vertical erosion rate of the cave entrance is estimated to ~0.65 cm/yr over a timespan of 139 years. Based on the last known photograph of Egg Rock (early 1900s), the total mass loss is approximated to be 197.70 t. Collectively, the findings from these three sites provide insights into rates of Holocene shore platform development along this dynamic coastline.
Wider use of archival visual sources clearly has potential for complementing more technically sophisticated short-term change reconstruction and monitoring approaches. Key challenges include sourcing well-dated, high-quality archival visual sources to enable establishment of robust timelines of change and the generation of quantitative data, and safely accessing potentially hazardous locations to enable new paintings, sketches, or photography. If these challenges are surmounted, opportunities include enhanced potential for: i) providing quantified landscape change case studies for inclusion in school/university geoscience syllabi; ii) demonstrating the relevance of geoscience for local/regional natural and cultural heritage; and iii) enhancing public engagement with coastal geoscience (e.g. through citizen science projects or science-art collaborations).
How to cite: Goode, C. and Tooth, S.: Can archival visual sources be used to quantify coastal change?: insights from the dynamic coastline of Ceredigion, west Wales, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18399, https://doi.org/10.5194/egusphere-egu25-18399, 2025.
Historical imagery captured from aeroplanes since the early 1900s and from spy satellites from the 1960s onwards have long been used in natural sciences for military, civil, and research purposes. These images have the unequalled potential for documenting and quantifying past environmental changes caused by natural and anthropogenic factors. Especially when acquired in stereo mode, these images enable the generation of point clouds and digital elevation models (DEMs), allowing us to quantify surface elevation changes over the past century.
Recent advancements in digital photogrammetry and the increasing availability of historical photographs as digitised/scanned images have heightened the interest in these data for reconstructing long-term surface evolution from local to regional scale. However, despite the large archive of historical images, their full potential is not yet widely exploited. Key challenges include accessibility, lack of metadata, image degradation, limited resolution and accuracy and lack of standardised workflows for generating DEMs and orthophotos.
We reviewed 198 journal articles published between 2001 and 2023 that processed historical aerial and spy satellite imagery. Our review spans methodological advancements in photogrammetric reconstruction and applied research analysing past 2D and 3D environmental changes across geoscience fields, such as geomorphology, cryosphere, volcanology, forestry, etc. We provide a comprehensive overview of these studies, summarise the image archives, applications, and products, and compare the methods used to process historical aerial and spy satellite imagery. Furthermore, we highlight emerging workflows and offer recommendations for image processing and accuracy assessment for future research and applications.
How to cite: Piermattei, L., McNabb, R., Elias, M., Ressl, C., Dehecq, A., Girod, L., Dewez, T., and Eltner, A.: Unlocking the potential of historical aerial and spy satellite stereo-imagery in geosciences: access, processing, and applications, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5168, https://doi.org/10.5194/egusphere-egu25-5168, 2025.
Climate change is resulting in rapidly increasing temperatures in the European Alps, rising twice as fast compared to the global average, and leading to unprecedented glacier retreat. Deglaciating alpine landscapes are considered extremely dynamic, evolving rapidly over space and time. The use of DEMs (Digital Elevation Models) of Difference (DoDs) to study changes occurring in these environments has significantly increased in the last years and has been used for a wide range of disciplines. This approach builds on the growing availability of datasets (e.g. historical imagery), accessibility of drones and their sensors (e.g. LiDAR) and facilitated use of digital photogrammetry through commercial and open-source Structure-from-Motion software. However, DoDs of deglaciating landscapes tend to disregard the diversity and complexity of processes in these environments.
In this research, DEMs were obtained using aerial archival photogrammetry (1977) for the Turtmann basin, a rapidly deglaciating Alpine valley in the Canton of Valais (southwestern Switzerland. A 2021 DEM was used as a reference to create a DoD of the basin (28km2), in order to determine net sediment erosion and deposition during this 44-year time period.
Most changes identified in the DoD could not be attributed to sediment displacement, but rather to various ecological (e.g. tree growth), glacial (e.g. glacier ice melt) and periglacial (e.g. rock glacier and buried ice melt) processes, as well as error in the photogrammetry. The latter is amplified by the inherently steep topography of alpine basins, which means that small georeferencing errors can cause significant apparent vertical change. A series of post-processing steps were required to obtain precise sediment volumes from the DoD.
DoDs are extremely valuable for assessing changes in rapidly deglaciating environments. However, challenges exist when applying them to such topographically complex and dynamic landscapes. These challenges must be identified and thoroughly dealt with through DoD post-processing in order to exploit DoDs to their full potential and obtain precise volumes of change. The specific post-processing steps will depend on (1) the research objective, which determines the desired precision as compared to the limits of detection, and (2) the spatial and temporal scales of the DoD, which influence the detectability of changes. In this research, the large temporal (decades) and spatial (basin-wide) scales exposed the challenges and opportunities of using DoDs in rapidly deglaciating environments. The workflow developed to overcome these challenges can be applied to other alpine basins for more precise change detection and thus allow for a better quantitative understanding of processes in deglaciating environments.
How to cite: Repnik, L., Breillad, A., Giovanardi, A., Comiti, F., Gianini, M., Argentin, A.-L., Pitscheider, F., and Lane, S. N.: Historical photogrammetry for DoDs in deglaciating environments: challenges and opportunities, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6774, https://doi.org/10.5194/egusphere-egu25-6774, 2025.
The digitisation of historical aerial photograph archives offers a unique opportunity to analyse long-term environmental changes. One such valuable resource is the archive of 1935-1941 aerial photographs of Ethiopia, one of the largest and oldest collections in Africa, comprising 34,000 images. While a portion of these images has been localized, many remain without known coordinates. To address this, we developed a computer vision approach that combines scale invariant feature transform (SIFT) keypoint matching and nearest-neighbour search, achieving 99% accuracy and 80% recall in matching images. This method increased the localization rate from 40% to approximately 70%, though manual verification and coordinate determination remain necessary. A proof-of-concept further demonstrated the potential of utilizing depth information to localize photographs: by leveraging the spatial proximity of images within the quite erratic flight lines, we significantly reduced the search area. Additionally, we show that 3D scene reconstruction from consecutive images, matched to a digital elevation model using the ICP algorithm, is feasible.
We demonstrate the potential of historical aerial archives for studying long-term environmental change through a case study on river geomorphology. At 70 locations where aerial photographs intersect major unconfined rivers, we analysed key hydrogeomorphological variables to assess river dynamics. By comparing river morphology in 1935-1941 with that on the most recent Google Earth imagery, our results reveal significant morphological changes, including channel widening, gullying, bank erosion, and in-stream sediment accumulation. These findings highlight how a detailed understanding of local river dynamics, derived from historical and modern imagery, can enhance the broader understanding of environmental changes and their impacts on catchment behaviour.
Key words: Aerial Photographs, Environmental Change, Hydrogeomorphology, Environmental Change, River
How to cite: De Baets, M., Lamote, L., Sterckx, J., Annys, S., Nyssen, J., Luong, H., Gebreyohannes, T., and Frankl, A.: The use of computer vision to relocate historical aerial photographs that enhance the understanding of hydrogeomophic changes in Ethiopia, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6611, https://doi.org/10.5194/egusphere-egu25-6611, 2025.
Alpine alluvial fans are important storages in the alpine sediment cascade. It is to be expected that climate change and the resulting changes in precipitation will have a massive impact on the dynamics of alpine alluvial fans. In order to differentiate between short-term and long-term dynamics, we compile a dataset quantifying sediment redistribution of a small mountain river and its alluvial fan on centennial, decadal and sub-annual scales. Our dataset comprises historical topographic maps from 1826 to 1912, 25 sets of historical aerial images collected between 1945 and 2024, and 17 high-resolution UAV-campaigns collected between September 2018 and October 2024. We identify the spatial changes in the sediment body, quantify the sediment redistribution and relate both to precipitation.
On centennial timescales, our data show a shift from presumably low geomorphic activity that persisted for at least 100 years (1820s-1930s) in the eastern sector of the fan, to high geomorphic activity with rapid channel migration across the central fan within the past 60 years. The onset of intense geomorphic activity may be contemporaneous to the increase in debris flow activity at nearby lake Plansee in the 1920s (Kiefer, Oswald et al., 2021). Decadal changes to the active area are largely explained by median precipitation (r2 = 0.66, p < 0.002) measured at a weather station c. 10 km east. Since the 1960s, incision at the apex and deposition at the toe of the fan can be observed. Sub-annual change detections show that for most epochs, erosion and deposition balance out within the uncertainty margin and the main channel gradually shifts its position by bank erosion and gravel bar construction. However, following an extreme deposition event between August and September 2019 with a net deposition of 8000 ± 3500 m3, the course of the main channel abruptly shifted. Our preliminary results show that while historical maps and aerial images are useful to reconstruct long-term trends, repeat topographic surveys with a close temporal spacing are needed to understand the processes behind these trends.
Kewords: alpine alluvial fan, sediment redistribution, geomorphic change detection, multiscale investigation
Kiefer, C., Oswald, P. Moernaut, J., Fabbri, S.C., Mayr, C., Strasser, M. & Krautblatter, M. (2021): A 4000-year debris flow record based on amphibious investigations of fan delta activity in Plansee (Austria, Eastern Alps). – Earth Surface Dynamics, 9: 1481–1503. DOI: 10.5194/esurf-9-1481-2021
How to cite: Gewalt, P., Wagner, T. C., and Krautblatter, M.: Constraining centennial to sub-annual sediment dynamics on alpine alluvial fans – first insights from the Friedergries (Germany), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15820, https://doi.org/10.5194/egusphere-egu25-15820, 2025.
Trend determination for earth surface processes requires long and continuous and certain measurements, but long-term records of landscape change are often limited in temporal and spatial extent. Scanned historical aerial imagery serve as a valuable resource to derive data products like digital elevation models (DEMs) to document the historical state of the Earth's surface and to calculate trends for different processes e.g. glacier dynamics.
Classic Structure-from-Motion (SfM) photogrammetry workflows have demonstrated the capability to automatically generate DEMs and orthoimage mosaics from such historical images, as highlighted in a few studies. These workflows typically consist of the following steps: (a) pre-processing, (b) tie-point extraction, (c) matching, (d) bundle adjustment, (e) dense reconstruction, (f) co-registration, and (g) orthoimage mosaic generation. However, classic methods struggle with the challenges historical imagery coming with. For example: inconsistent image quality, limited metadata documentation, image distortions and distinct viewpoint geometries.
Recently, advances in robotics and computer vision have introduced learned models for tasks such as tie-point identification, matching, dense reconstruction as well as part of the co-registration stage (e.g. SuperPoint, ALIKE, SuperGlue, LoFTR and more). These networks have shown promising results in different stereo-matching scenarios by outperforming classic SfM methods. However, since they were primarily developed for modern robotics and computer vision tasks, their performance on scanned historical aerial imagery remains uncertain. As historical imagery exhibits the properties described above, these networks were not optimised with them during training.
We boost existing pipelines in tie-point extraction and matching with these models and compare the quality of resulting DEMs from different model combinations together. We also highlight issues encountered when applying these learned models to historical aerial imagery and proposes solutions to address them. We demonstrate our findings using scanned historical images from the Southern Patagonian Ice Field (Chile) recorded in 1980, particularly for the Grey & Dickson Glacier area, as well the south-west flank of Cordon Mariano Moreno Mountain and adjacent fjords. These two sites providing different acquisition geometries and overlaps. The results evaluate the average RMS reprojection error following the bundle adjustment, to determine the quality of different extractors and matchers as well as the median distance between closest points to evaluate the co-registration.
How to cite: Kugler, L., Ioli, F., Wegner, J. D., Dussaillant, I., Rada, C., and Piermattei, L.: Advances in Historical Aerial Image Analysis: Boosting SfM Pipelines with Learned Models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17688, https://doi.org/10.5194/egusphere-egu25-17688, 2025.
During the last decades the high Arctic has undergone substantial changes as a result of global warming and arctic amplification. Melt seasons are expanding rapidly, and landscape and ecosystems are shifting into new states. To quantify these changes from the historical baseline requires datasets on pre-warming states, which can be extremely rare in the high Arctic. Prior to the satellite era, starting in the 1990s, a commonly used data source for baselines in geosciences is aerial photographs, which if one is lucky can reach back to the 1930s. These aerial images are most often recorded at high elevation and perhaps also obliquely which results in spatial resolutions of 2-10 meters, limiting the level of detail that can be resolved on the ground.
With this presentation we reveal a new exciting dataset of aerial images from East Greenland recorded in the 1950s and ‘60s. Contrary to other aerial campaigns, these images were recorded at very low elevation in order to conduct geological mapping, ultimately yielding spatial resolutions surpassing those of the newest high resolution satellites.
The images were recorded by geologist John Haller during the Lauge Koch expeditions to central East Greenland in the 1950s and 1960s, and comprise a dataset of c. 3600 high resolution oblique images recorded at low elevation from plane and helicopter. The images are recorded in stereo, which allows us to recreate the terrain surface in 3D and construct orthorectified imagery that allows a direct comparison with modern satellite images, for use in all aspects of landscape- and ecosystem evolution.
How to cite: Bjork, A., Deichmann, A., and Socher, T.: A new high resolution historical aerial image dataset from East Greenland, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16381, https://doi.org/10.5194/egusphere-egu25-16381, 2025.
We tested whether public high-resolution airborne LiDAR data could be suitable for structural geology applications by comparing fracture orientation measurements on Virtual Outcrop Models (VOMs) to field measurements from the same outcrops. We found that the fundamental requirement for taking full advantage of such data is good bedrock exposure, which is also dependent on lithology. Whenever this requirement is satisfied, VOM measurements are comparable to field measurements. VOMs can help considerably in both reducing the time it takes to collect measurements, and in expanding the area in which measurements can be collected without adding significantly to the time budget. They are also especially useful in remote regions and at high elevations, where access is more difficult and yet good exposures are more likely to be found, and they should always be used when planning field work. At present the main limitations, apart from LiDAR coverage not yet existing in places, are due to the hardware and software capabilities needed to create and especially to analyze VOMs.
How to cite: Carena, S. and Friedrich, A.: Mapping fractures in 3D from airborne LiDAR: comparison with field mapping, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2144, https://doi.org/10.5194/egusphere-egu25-2144, 2025.
