Ayoub Fatihi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Ayoub Fatihi has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Anindita Samsu
Institute of Earth Sciences, University of Lausanne, 1015 Lausanne, Switzerland
5 shared publications - Jefter Caldeira
Institute of Earth Sciences, University of Lausanne, 1015 Lausanne, Switzerland
5 shared publications Samuel T. Thiele
Helmholtz-Zentrum Dresden-Rossendorf, Helmholtz Institute Freiberg, Chemnitzer Str. 40, 09599 Freiberg, Germany
4 shared publications- Tom Beucler
Institute of Earth Surface Dynamics, University of Lausanne, 1015 Lausanne, Switzerland
4 shared publications - Ahmed Jamal Afifi
Helmholtz Institute Freiberg for Resource Technology, Freiberg, Germany
2 shared publications - Lihong Zheng
Charles Sturt University
2 shared publications
Latest Publications
- Open Access
- 8 October 2026
- Published by Springer Nature in Natural Hazards
- Vol. 122 (21) , 665
Glacial lake outburst flood hazard and downstream impacts: lessons from Afghanistan examples
Glacial lake outburst floods (GLOFs) pose a growing risk in high mountain regions as climate-driven glacier retreat accelerates lake expansion and new lake formation. In the Hindu Kush Himalaya, particularly in Afghanistan, triggering processes and GLOF propagation downstream are largely unstudied. This study evaluates two Afghanistan GLOFs (Peshghor 2018; Bam-Tanab 2021) using field observations and satellite imagery, and applies two-dimensional hydrodynamic simulation (BASEMENT-MD with BASEbreach) to the Kunj-Peshghor lake event to assess flood propagation, dam-breach dynamics, and downstream hazards under different scenarios. Both lakes formed and expanded over 33–42 days before catastrophic drainage, triggered by extreme temperatures, rapid-melt, ice-cliff collapse, and englacial tunnel failure. Both impacted Peshghor village downstream. The Kunj-Peshghor lake flood wave produced estimated peak velocities of 9 ms −1 and depths >10 m in confined valleys, associated with peak discharges >170 m 3 s −1 , high turbidity >800 NTU, and channel widening of up to 250 m. Secondary surges caused by temporary river blockages prolonged flooding and amplified downstream impacts. Hazard mapping indicates that ~46% of inundated zones in the Peshghor village fell within high to very high hazard levels, threatening settlements, roads, markets, and agricultural land. These findings provide the first quantitative GLOF propagation modeling for Afghanistan and highlight the need for site-specific monitoring, early warning systems, and mitigation measures. Both GLOFs were caused by the rapid formation and expansion of new lakes. This raises the question about whether current approaches to identifying dangerous lakes, based on remote-sensing of existing lakes are sufficient to detect all dangerous lakes and the associated hazards.
- 8 October 2026
- Published by Springer Nature in Natural Hazards
- Vol. 122 (21) , 665
Open AccessGlacial lake outburst flood hazard and downstream impacts: lessons from Afghanistan examples
Glacial lake outburst floods (GLOFs) pose a growing risk in high mountain regions as climate-driven glacier retreat accelerates lake expansion and new lake formation. In the Hindu Kush Himalaya, particularly in Afghanistan, triggering processes and GLOF propagation downstream are largely unstudied. This study evaluates two Afghanistan GLOFs (Peshghor 2018; Bam-Tanab 2021) using field observations and satellite imagery, and applies two-dimensional hydrodynamic simulation (BASEMENT-MD with BASEbreach) to the Kunj-Peshghor lake event to assess flood propagation, dam-breach dynamics, and downstream hazards under different scenarios. Both lakes formed and expanded over 33–42 days before catastrophic drainage, triggered by extreme temperatures, rapid-melt, ice-cliff collapse, and englacial tunnel failure. Both impacted Peshghor village downstream. The Kunj-Peshghor lake flood wave produced estimated peak velocities of 9 ms −1 and depths >10 m in confined valleys, associated with peak discharges >170 m 3 s −1 , high turbidity >800 NTU, and channel widening of up to 250 m. Secondary surges caused by temporary river blockages prolonged flooding and amplified downstream impacts. Hazard mapping indicates that ~46% of inundated zones in the Peshghor village fell within high to very high hazard levels, threatening settlements, roads, markets, and agricultural land. These findings provide the first quantitative GLOF propagation modeling for Afghanistan and highlight the need for site-specific monitoring, early warning systems, and mitigation measures. Both GLOFs were caused by the rapid formation and expansion of new lakes. This raises the question about whether current approaches to identifying dangerous lakes, based on remote-sensing of existing lakes are sufficient to detect all dangerous lakes and the associated hazards.
- Open Access
- 29 September 2026
- Published by Copernicus GmbH in Solid Earth
- Vol. 17 (9) , 1087-1115
Towards robust fracture mapping: benchmarking automatic fracture mapping in 2D outcrop imagery
Analysis of fracture geometries, orientation, distribution, and connectivity is commonly conducted to characterise the mechanical and hydraulic properties of a fractured rock mass and to interpret its deformation history. These studies increasingly leverage high-resolution drone imagery (e.g., orthomosaics or orthophotographs). However, consistently and accurately extracting fracture traces from these large datasets remains a persistent challenge. In this contribution, we present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation of combined high-resolution RGB orthophotographs and digital elevation models (DEMs) (ground sampling distance ranging from approximately 0.5 to 32 mm). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB + DEM patches spanning diverse lithologies and imaging conditions (ground sampling distance, illumination, etc.) to systematically assess the effectiveness of conventional image-processing approaches for fracture extraction (Canny, Sobel, Gabor, Sato, and phase congruency) and two deep-learning (DL) models (U-Net and SegFormer). Quantitative comparisons using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU), and new task-specific error metrics show that the deep models substantially outperform classical filters (F1 ≈ 0.3–0.5 vs. ± 0.29) and produce smoother, more continuous fracture traces. Training on the combined dataset (M_all) improves cross-site generalisation compared to models trained on individual sub-datasets used in this study. Probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work establishes a unified benchmark, curated dataset, and reproducible baseline for developing robust automated fracture-detection tools.
- 29 September 2026
- Published by Copernicus GmbH in Solid Earth
- Vol. 17 (9) , 1087-1115
Open AccessTowards robust fracture mapping: benchmarking automatic fracture mapping in 2D outcrop imagery
Analysis of fracture geometries, orientation, distribution, and connectivity is commonly conducted to characterise the mechanical and hydraulic properties of a fractured rock mass and to interpret its deformation history. These studies increasingly leverage high-resolution drone imagery (e.g., orthomosaics or orthophotographs). However, consistently and accurately extracting fracture traces from these large datasets remains a persistent challenge. In this contribution, we present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation of combined high-resolution RGB orthophotographs and digital elevation models (DEMs) (ground sampling distance ranging from approximately 0.5 to 32 mm). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB + DEM patches spanning diverse lithologies and imaging conditions (ground sampling distance, illumination, etc.) to systematically assess the effectiveness of conventional image-processing approaches for fracture extraction (Canny, Sobel, Gabor, Sato, and phase congruency) and two deep-learning (DL) models (U-Net and SegFormer). Quantitative comparisons using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU), and new task-specific error metrics show that the deep models substantially outperform classical filters (F1 ≈ 0.3–0.5 vs. ± 0.29) and produce smoother, more continuous fracture traces. Training on the combined dataset (M_all) improves cross-site generalisation compared to models trained on individual sub-datasets used in this study. Probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work establishes a unified benchmark, curated dataset, and reproducible baseline for developing robust automated fracture-detection tools.
- Open Access
- 31 July 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-35
AC1
Extracting consistent and accurate fracture traces from large volumes of high-resolution imagery remains a persistent challenge in structural analysis. We present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation in high-resolution RGB orthophotos and digital elevation models (DEMs). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB+DEM patches spanning diverse lithologies and imaging conditions. We use this dataset to systematically assess traditional image-processing filters (Canny, Sobel, Gabor, Sato, phase congruency) and two deep-learning (DL) models, U-Net and SegFormer, for per-pixel fracture detection. Quantitative comparison using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU) and proposed similarity FracSim metrics suggest that the deep models substantially outperform classical filters (F1 ≈ 03 −0.5 vs ≤ 0.29), giving smoother, more continuous fracture traces with reduced noise. Training on the combined dataset (M_all) improves cross-site generalisation relative to models trained on the individual sub-datasets. Challenges remain in handling annotation misalignments, illumination artifacts, and thin traces. More importantly, probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work thus establishes a unified benchmark, curated dataset, and reproducible baseline to support further development of robust automated tools for fracture detection.
- 31 July 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-35
Open AccessAC1
Extracting consistent and accurate fracture traces from large volumes of high-resolution imagery remains a persistent challenge in structural analysis. We present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation in high-resolution RGB orthophotos and digital elevation models (DEMs). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB+DEM patches spanning diverse lithologies and imaging conditions. We use this dataset to systematically assess traditional image-processing filters (Canny, Sobel, Gabor, Sato, phase congruency) and two deep-learning (DL) models, U-Net and SegFormer, for per-pixel fracture detection. Quantitative comparison using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU) and proposed similarity FracSim metrics suggest that the deep models substantially outperform classical filters (F1 ≈ 03 −0.5 vs ≤ 0.29), giving smoother, more continuous fracture traces with reduced noise. Training on the combined dataset (M_all) improves cross-site generalisation relative to models trained on the individual sub-datasets. Challenges remain in handling annotation misalignments, illumination artifacts, and thin traces. More importantly, probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work thus establishes a unified benchmark, curated dataset, and reproducible baseline to support further development of robust automated tools for fracture detection.
- Open Access
- 22 May 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-35
Comment on egusphere-2026-1097
Extracting consistent and accurate fracture traces from large volumes of high-resolution imagery remains a persistent challenge in structural analysis. We present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation in high-resolution RGB orthophotos and digital elevation models (DEMs). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB+DEM patches spanning diverse lithologies and imaging conditions. We use this dataset to systematically assess traditional image-processing filters (Canny, Sobel, Gabor, Sato, phase congruency) and two deep-learning (DL) models, U-Net and SegFormer, for per-pixel fracture detection. Quantitative comparison using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU) and proposed similarity FracSim metrics suggest that the deep models substantially outperform classical filters (F1 ≈ 03 −0.5 vs ≤ 0.29), giving smoother, more continuous fracture traces with reduced noise. Training on the combined dataset (M_all) improves cross-site generalisation relative to models trained on the individual sub-datasets. Challenges remain in handling annotation misalignments, illumination artifacts, and thin traces. More importantly, probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work thus establishes a unified benchmark, curated dataset, and reproducible baseline to support further development of robust automated tools for fracture detection.
- 22 May 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-35
Open AccessComment on egusphere-2026-1097
Extracting consistent and accurate fracture traces from large volumes of high-resolution imagery remains a persistent challenge in structural analysis. We present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation in high-resolution RGB orthophotos and digital elevation models (DEMs). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB+DEM patches spanning diverse lithologies and imaging conditions. We use this dataset to systematically assess traditional image-processing filters (Canny, Sobel, Gabor, Sato, phase congruency) and two deep-learning (DL) models, U-Net and SegFormer, for per-pixel fracture detection. Quantitative comparison using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU) and proposed similarity FracSim metrics suggest that the deep models substantially outperform classical filters (F1 ≈ 03 −0.5 vs ≤ 0.29), giving smoother, more continuous fracture traces with reduced noise. Training on the combined dataset (M_all) improves cross-site generalisation relative to models trained on the individual sub-datasets. Challenges remain in handling annotation misalignments, illumination artifacts, and thin traces. More importantly, probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work thus establishes a unified benchmark, curated dataset, and reproducible baseline to support further development of robust automated tools for fracture detection.
- Open Access
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Computer Society
- Vol. 7, 1038-1050
Boosting Brain Tumor Detection Accuracy in MRI Using Transfer Learning and Fine-Tuned DeepLabv3+
Detecting and diagnosing brain tumors in adults is a complex task, given their varied appearance and serious health risks. Interpreting magnetic resonance imaging (MRI) scans manually can be time-consuming and error-prone, making the development of reliable automated solutions highly valuable. In this study, we introduce a deep convolutional neural network (CNN) framework built on a refined DeepLabv3+ architecture for segmenting glioma, meningioma, and pituitary tumors in contrast-enhanced magnetic resonance imaging (CE-MRI). The design employs a ResNet34 backbone, atrous spatial pyramid pooling (ASPP), and a boundary-aware decoder, specifically tuned to capture the structural details of CEMRI tumor images. To address the pronounced imbalance between tumor and background pixels, we incorporate a Weighted Binary Cross-Entropy (WBCE) loss, which improves the detection of smaller or irregular tumor areas. We also conduct a step-by-step ablation study to assess the contributions of WBCE, ASPP, and ImageNet pretraining to the model's performance. The framework is evaluated on an internal CE-MRI dataset, an external CE-MRI dataset without additional fine-tuning, and multiple BraTS benchmark editions (2020, 2021, 2023), demonstrating strong performance across varied imaging protocols. By achieving a Dice similarity coefficient of 98.0%, an accuracy of 99.1%, and a sensitivity of 99.3%, the proposed framework-comprising architectural choices, loss-function optimization, and thorough validation outperforms other existing techniques.
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Computer Society
- Vol. 7, 1038-1050
Open AccessBoosting Brain Tumor Detection Accuracy in MRI Using Transfer Learning and Fine-Tuned DeepLabv3+
Detecting and diagnosing brain tumors in adults is a complex task, given their varied appearance and serious health risks. Interpreting magnetic resonance imaging (MRI) scans manually can be time-consuming and error-prone, making the development of reliable automated solutions highly valuable. In this study, we introduce a deep convolutional neural network (CNN) framework built on a refined DeepLabv3+ architecture for segmenting glioma, meningioma, and pituitary tumors in contrast-enhanced magnetic resonance imaging (CE-MRI). The design employs a ResNet34 backbone, atrous spatial pyramid pooling (ASPP), and a boundary-aware decoder, specifically tuned to capture the structural details of CEMRI tumor images. To address the pronounced imbalance between tumor and background pixels, we incorporate a Weighted Binary Cross-Entropy (WBCE) loss, which improves the detection of smaller or irregular tumor areas. We also conduct a step-by-step ablation study to assess the contributions of WBCE, ASPP, and ImageNet pretraining to the model's performance. The framework is evaluated on an internal CE-MRI dataset, an external CE-MRI dataset without additional fine-tuning, and multiple BraTS benchmark editions (2020, 2021, 2023), demonstrating strong performance across varied imaging protocols. By achieving a Dice similarity coefficient of 98.0%, an accuracy of 99.1%, and a sensitivity of 99.3%, the proposed framework-comprising architectural choices, loss-function optimization, and thorough validation outperforms other existing techniques.
- Open Access
- 13 March 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-35
Towards robust fracture mapping: benchmarking automatic fracture mapping in 2D outcrop imagery
Extracting consistent and accurate fracture traces from large volumes of high-resolution imagery remains a persistent challenge in structural analysis. We present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation in high-resolution RGB orthophotos and digital elevation models (DEMs). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB+DEM patches spanning diverse lithologies and imaging conditions. We use this dataset to systematically assess traditional image-processing filters (Canny, Sobel, Gabor, Sato, phase congruency) and two deep-learning (DL) models, U-Net and SegFormer, for per-pixel fracture detection. Quantitative comparison using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU) and proposed similarity FracSim metrics suggest that the deep models substantially outperform classical filters (F1 ≈ 03 −0.5 vs ≤ 0.29), giving smoother, more continuous fracture traces with reduced noise. Training on the combined dataset (M_all) improves cross-site generalisation relative to models trained on the individual sub-datasets. Challenges remain in handling annotation misalignments, illumination artifacts, and thin traces. More importantly, probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work thus establishes a unified benchmark, curated dataset, and reproducible baseline to support further development of robust automated tools for fracture detection.
- 13 March 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-35
Open AccessTowards robust fracture mapping: benchmarking automatic fracture mapping in 2D outcrop imagery
Extracting consistent and accurate fracture traces from large volumes of high-resolution imagery remains a persistent challenge in structural analysis. We present a harmonised benchmarking dataset, FraXet, for pixel-wise fracture segmentation in high-resolution RGB orthophotos and digital elevation models (DEMs). FraXet curates images from three publicly available datasets, totalling 8953 256 × 256 RGB+DEM patches spanning diverse lithologies and imaging conditions. We use this dataset to systematically assess traditional image-processing filters (Canny, Sobel, Gabor, Sato, phase congruency) and two deep-learning (DL) models, U-Net and SegFormer, for per-pixel fracture detection. Quantitative comparison using image-quality (e.g., MSE, PSNR), segmentation (e.g., Precision, Recall, F1, IoU) and proposed similarity FracSim metrics suggest that the deep models substantially outperform classical filters (F1 ≈ 03 −0.5 vs ≤ 0.29), giving smoother, more continuous fracture traces with reduced noise. Training on the combined dataset (M_all) improves cross-site generalisation relative to models trained on the individual sub-datasets. Challenges remain in handling annotation misalignments, illumination artifacts, and thin traces. More importantly, probability maps derived from the DL approaches enable confidence-based triage and visualisation of model uncertainty. This work thus establishes a unified benchmark, curated dataset, and reproducible baseline to support further development of robust automated tools for fracture detection.
- Open Access
- 9 February 2026
- Published by Elsevier in Journal of Structural Geology
- Vol. 205
Multi-scale structural analysis of fracture networks in folded limestones: A case study from the Central Internal Jura, Switzerland
- 9 February 2026
- Published by Elsevier in Journal of Structural Geology
- Vol. 205
Open AccessMulti-scale structural analysis of fracture networks in folded limestones: A case study from the Central Internal Jura, Switzerland
- Open Access
- 14 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Computing for Healthcare
- Vol. 7 (1) , 1-32
An Empirical Analysis of Loss Functions for Deep Retinal Vessel Segmentation
Retinal vessel segmentation underpins computer-assisted screening and monitoring of ocular and systemic disease. While encoder–decoder networks such as U-Net are widely used, their behavior is strongly shaped by the training objective. This work presents a controlled empirical study of loss functions for vessel segmentation using a U-Net architecture that employs strided convolutions in the encoder, together with a consistent pre-processing pipeline based on morphological enhancement and principal component analysis. We compare cross-entropy, weighted cross-entropy, and Dice losses on the DRIVE and STARE datasets under identical settings, reporting pixel-wise and overlap-based measures to reflect both detection and spatial agreement. The configuration with weighted cross-entropy provides a balanced outcome, achieving sensitivity and accuracy of 0.873 and 0.969 on DRIVE, and 0.821 and 0.961 on STARE. Rather than proposing architectural novelty, the contribution of this study is a reproducible data-driven comparison that clarifies the tradeoffs each loss imposes on recall, specificity, and boundary fidelity, offering practical guidance for selecting objectives in retinal vessel segmentation.
- 14 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Computing for Healthcare
- Vol. 7 (1) , 1-32
Open AccessAn Empirical Analysis of Loss Functions for Deep Retinal Vessel Segmentation
Retinal vessel segmentation underpins computer-assisted screening and monitoring of ocular and systemic disease. While encoder–decoder networks such as U-Net are widely used, their behavior is strongly shaped by the training objective. This work presents a controlled empirical study of loss functions for vessel segmentation using a U-Net architecture that employs strided convolutions in the encoder, together with a consistent pre-processing pipeline based on morphological enhancement and principal component analysis. We compare cross-entropy, weighted cross-entropy, and Dice losses on the DRIVE and STARE datasets under identical settings, reporting pixel-wise and overlap-based measures to reflect both detection and spatial agreement. The configuration with weighted cross-entropy provides a balanced outcome, achieving sensitivity and accuracy of 0.873 and 0.969 on DRIVE, and 0.821 and 0.961 on STARE. Rather than proposing architectural novelty, the contribution of this study is a reproducible data-driven comparison that clarifies the tradeoffs each loss imposes on recall, specificity, and boundary fidelity, offering practical guidance for selecting objectives in retinal vessel segmentation.
- Open Access
- 28 July 2025
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-G-2, 359-365
Permanent Laser Scanning and 3D Time Series Analysis for Geomorphic Monitoring using Low-Cost Sensors and Open-Source Software
This study presents a first assessment of the performances of a low-cost permanent laser scanning (PLS) system for geomorphic monitoring applications. The goal is to evaluate the applicability and accuracy of these accessible technologies in comparison with high-end, commercial laser scanning systems. The assessment focuses on accuracy estimations and reliability in detecting and quantifying geomorphic changes over time in a target area in the Rotmoos valley, located in the Ötztal (Tyrol, Austria), featuring sediment movement and riverbed changes that are manually induced in an experimental setup. In this study, we use a Livox Avia scanner, controllable via an open SDK and Raspberry Pi, as low-cost monitoring setup in comparison to a high-end RIEGL VZ-2000i TLS. We acquired 14 epochs of point clouds from both systems simultaneously while inducing changes to the scene in-between acquisitions. Changes are quantified via direct point cloud comparison using the M3C2 algorithm and assessed both spatially per epoch as well as regarding the time series information at selected locations. Our results show consistent change values and patterns obtained with both Livox and RIEGL scans, demonstrating that, despite minor differences in time series trends, the low-cost Livox scanner effectively captures geomorphic changes comparable to those measured by the RIEGL. Our presented approach, by leveraging affordable hardware and open-source software tools, could provide a cost-effective solution for long-term environmental monitoring. By comparing the results obtained from both systems, this research highlights the potential of low-cost alternatives for continuous geomorphic monitoring, offering valuable insights for cost-effective environmental management and research.
- 28 July 2025
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-G-2, 359-365
Open AccessPermanent Laser Scanning and 3D Time Series Analysis for Geomorphic Monitoring using Low-Cost Sensors and Open-Source Software
This study presents a first assessment of the performances of a low-cost permanent laser scanning (PLS) system for geomorphic monitoring applications. The goal is to evaluate the applicability and accuracy of these accessible technologies in comparison with high-end, commercial laser scanning systems. The assessment focuses on accuracy estimations and reliability in detecting and quantifying geomorphic changes over time in a target area in the Rotmoos valley, located in the Ötztal (Tyrol, Austria), featuring sediment movement and riverbed changes that are manually induced in an experimental setup. In this study, we use a Livox Avia scanner, controllable via an open SDK and Raspberry Pi, as low-cost monitoring setup in comparison to a high-end RIEGL VZ-2000i TLS. We acquired 14 epochs of point clouds from both systems simultaneously while inducing changes to the scene in-between acquisitions. Changes are quantified via direct point cloud comparison using the M3C2 algorithm and assessed both spatially per epoch as well as regarding the time series information at selected locations. Our results show consistent change values and patterns obtained with both Livox and RIEGL scans, demonstrating that, despite minor differences in time series trends, the low-cost Livox scanner effectively captures geomorphic changes comparable to those measured by the RIEGL. Our presented approach, by leveraging affordable hardware and open-source software tools, could provide a cost-effective solution for long-term environmental monitoring. By comparing the results obtained from both systems, this research highlights the potential of low-cost alternatives for continuous geomorphic monitoring, offering valuable insights for cost-effective environmental management and research.
- Open Access
- 18 March 2025
- Published by Copernicus GmbH
Challenges and perspectives of climate data compression in times of kilometre-scale models and generative machine learning
Climate data compression urgently needs new standards. The continuously growing exascale mountain of data requires compressors that are widely used and supported, essentially hiding the compression details from many users. With the advent of AI revolutionising scientific computing, we have to set the rules of this game. Minimizing information loss, maximising compression factors, at any resolution, grid and dataset size, for all variables, with chunks and random access, while preserving all statistics and derivatives, at a reasonable speed — are squaring the compression circle. Many promising compressors are hardly used as trust among domain scientists is hard to gain: The large spectrum of research questions and applications using climate data is very difficult to satisfy simultaneously.Here, we illustrate the motivation behind the newly defined climate data compression benchmark ClimateBenchPress, designed as a quality check in all those dimensions of the problem. Any benchmark will inevitably undersample this space, but we define datasets from atmosphere, ocean, and land as well as evaluation metrics to pass. Results are presented as score cards, highlighting strengths and weaknesses for every compressor.The bitwise real information content shows a systematic way in case no error bounds are known. In the case of the ERA5 reanalysis, errors are estimated and allow us to categorize many variables into linear, log and beta distributions with values bounded from zero, one or both sides, respectively. This allows us to define error thresholds arising from observation and model errors directly, providing another alternative to the still predominant subjective choices. Most error-bounded compressors come with parameters that can be automatically chosen following this analysis.Also new data formats are on the horizon: Chunking and hierarchical data structures allow and force us to adapt compressors to spatially or length-scale dependent information densities. Extreme events, maybe counterintuitively, often increase the compressibility through higher uncertainties, but lie on the edge or outside of the training data for machine learned-compressors. This again increases the need for well-tested compressors. Benchmarks like ClimateBenchPress are required to encourage new standards for safe lossy climate data compression.
- 18 March 2025
- Published by Copernicus GmbH
Open AccessChallenges and perspectives of climate data compression in times of kilometre-scale models and generative machine learning
Climate data compression urgently needs new standards. The continuously growing exascale mountain of data requires compressors that are widely used and supported, essentially hiding the compression details from many users. With the advent of AI revolutionising scientific computing, we have to set the rules of this game. Minimizing information loss, maximising compression factors, at any resolution, grid and dataset size, for all variables, with chunks and random access, while preserving all statistics and derivatives, at a reasonable speed — are squaring the compression circle. Many promising compressors are hardly used as trust among domain scientists is hard to gain: The large spectrum of research questions and applications using climate data is very difficult to satisfy simultaneously.Here, we illustrate the motivation behind the newly defined climate data compression benchmark ClimateBenchPress, designed as a quality check in all those dimensions of the problem. Any benchmark will inevitably undersample this space, but we define datasets from atmosphere, ocean, and land as well as evaluation metrics to pass. Results are presented as score cards, highlighting strengths and weaknesses for every compressor.The bitwise real information content shows a systematic way in case no error bounds are known. In the case of the ERA5 reanalysis, errors are estimated and allow us to categorize many variables into linear, log and beta distributions with values bounded from zero, one or both sides, respectively. This allows us to define error thresholds arising from observation and model errors directly, providing another alternative to the still predominant subjective choices. Most error-bounded compressors come with parameters that can be automatically chosen following this analysis.Also new data formats are on the horizon: Chunking and hierarchical data structures allow and force us to adapt compressors to spatially or length-scale dependent information densities. Extreme events, maybe counterintuitively, often increase the compressibility through higher uncertainties, but lie on the edge or outside of the training data for machine learned-compressors. This again increases the need for well-tested compressors. Benchmarks like ClimateBenchPress are required to encourage new standards for safe lossy climate data compression.