Aljoscha Rheinwalt

Quantifying the structure and heterogeneity of complex spatial patterns remains a key challenge in spatial data analysis. Recurrence plots (RPs) offer a powerful method for visualizing the recurrent spatial patterns in such data. To quantify the structure of spatial recurrences, we extend the normalized recurrence lacunarity (nRL), a measure of the homogeneity of an RP, to the analysis of spatial RPs. Even though RPs can be derived from data with a topological dimension larger than one, their nRL has yet to be evaluated. To address this gap, we evaluate the nRL of spatial RPs, validate the method using synthetic test patterns and model data, and then apply it to analyze the hillslope gradients of 47 river catchments near the Mendocino Triple Junction. The results suggest that the nRL effectively detects and quantifies subtle differences in the spatial structure of the catchments. We show indications that these relate to local uplift rates, a previously identified possible driver of the structural differences. The nRL provides a robust and scale-sensitive tool for comparing diverse spatial data sets and for detecting and quantifying how their spatial structure may relate to external parameters.

Urban vegetation provides essential ecosystem services, including temperature regulation, air purification, noise reduction, and carbon storage. However, urban densification and climate-induced stresses increasingly threaten these ecosystems. Accurate classification of urban vegetation is critical for sustainable urban planning, yet remains challenging due to the structural complexity and high data density of urban LiDAR (Light Detection and Ranging) point clouds. To address current research gaps, including insufficient model interpretability, high computational demands, and limited generalization capabilities, we introduce transmissivity, a novel feature that combines echo-based LiDAR properties with spatial context to more effectively characterize urban vegetation structures. This feature enhances vegetation classification by estimating whether laser beams tend to traverse or terminate within a local neighborhood, independent of the specific return order of individual beams, thereby characterizing vegetation’s volumetric permeability. This makes transmissivity highly interpretable, unlike other echo-based statistical features. Transmissivity was evaluated alongside 52 conventional features using three different feature-importance measures across two distinct urban LiDAR datasets from Berlin and Hessigheim 3D (both datasets are from Germany). Transmissivity consistently ranked among the most influential features in both datasets across multiple scales and achieved the highest average gain (Berlin: 0.465; Hessigheim: 0.286) and the second highest mean absolute Shapley value (Berlin: 2.066; Hessigheim: 0.824). Permutation importance confirmed that transmissivity has the strongest impact on the mean decrease in F1-score (Berlin: 0.53; Hessigheim: 0.34) if used in uncorrelated feature subsets. Our findings support that echo-enriched point clouds allow efficient and accurate monitoring of urban vegetation. The feature is simple to compute and does not require additional data beyond standard multi-return LiDAR attributes.

Sediment characteristics and grain-size distribution are crucial for understanding natural hazards, hydrologic conditions, and ecosystems. However, traditional methods for collecting this information are costly, labor-intensive, and time-consuming. To address this, we present OrthoSAM, a workflow leveraging the Segment Anything Model (SAM) for automated delineation of densely packed pebbles in high-resolution orthomosaics. Our framework consists of a tiling scheme, improved seed (input) point generation, and a multi-scale resampling scheme. Validation using synthetic images shows high precision close to 1, a recall above 0.9, with a mean IoU above 0.9. Using a large synthetic dataset, the two-sample Kolmogorov-Smirnov test confirms that there is no significant difference between the predicted and the ground-truth grain size distributions. We identified a size detection limit of 30 pixels; pebbles with a diameter below this limit are not reliably detected. Applying OrthoSAM to orthomosaics from the Ravi River in India, we delineated 6087 pebbles with high precision (0.93) and recall (0.94), based on manual verification of each predicted mask. The resulting grain statistics include area, axis lengths, perimeter, RGB statistics, and smoothness measurements, providing valuable insights for further analysis in geomorphology and ecosystem studies.

Urban vegetation provides essential ecosystem services, including temperature regulation, air purification, noise reduction, and carbon storage. However, urban densification and climate-induced stresses increasingly threaten these ecosystems. Accurate classification of urban vegetation is critical for sustainable urban planning, yet remains challenging due to the structural complexity and high data density of urban LiDAR point clouds.We introduce a novel feature, transmissivity, which combines echo-based LiDAR properties with spatial context to capture the urban vegetation structures more accurately. This feature enhances vegetation classification by estimating the proportion of laser beams that pass through a point's neighborhood. It is a local measure of how many laser beams pass through the neighborhood rather than terminating within it, explicitly targeting vegetation’s volumetric permeability.Transmissivity was evaluated alongside 52 conventional features using three different feature-importance measures across two distinct urban LiDAR datasets from Berlin and Hessigheim 3D. Results show that transmissivity is highly informative for vegetation detection across multiple scales and improves model interpretability and classification performance. Our findings support that echo-enriched point clouds allow efficient and accurate monitoring of urban vegetation. The feature is simple to compute and does not require additional data beyond standard multi-return LiDAR attributes.

The Central Andes are one of the most tectonically and geomorphologically active regions on Earth. The orographic barrier between the eastern foreland and the Central Andean plateau induces a strong East-West climatic gradient, with peak rainfall occurring on the steep eastward-facing slopes. Frequent rainstorms during the South American summer monsoon, coupled with fault-weakened lithologies, drive mass-movement processes including landslides and high-altitude periglacial creep. However, monitoring these instabilities over large areas and long time periods remains computationally expensive using traditional CPU-based methods.We implemented a high-performance GPU-based workflow using sub-pixel optical image correlation to process 20-year time series using the Landsat 7, 8, and 9 data archive spanning from 22° to 27° S (12 Path/Row combinations) . The workflow consists of two main steps: (1) identification of areas of interest by oversampling the scenes before correlation at a coarse, but sufficiently dense step size, and (2) sub-pixel matching for refined displacement derivation within the detected regions. To ensure data integrity, we employ a pair selection process based on sun-elevation geometry and filters for cloud cover, snow, vegetation changes, and topographic shadows. Additionally, we applied a kinematic filter to exclude displacements inconsistent with hillslope aspect. By stacking multi-decadal imagery, we improved the signal-to-noise ratio and successfully detected velocities ranging from less than 0.5 m/yr to several meters per year. Our results highlight the extent of permafrost processes and the influence of the East-West climatic gradient on hillslope dynamics by capturing the transition from the humid foreland to the arid high-elevation plateaus. The stacking method effectively removed outlier signals caused by transient snow cover at higher elevations.This 20-year record provides a vital baseline for understanding how Andean hillslope processes respond to a changing climate and how they depend on pre-existing, weakened lithologic conditions related to tectonic stresses. The GPU-accelerated framework enables a transition from localized monitoring to large-scale kinematic analysis in high-relief terrains.

Quantifying pebble size, shape, and roundness is fundamental tounderstanding sediment transport and abrasion in fluvial systems, yetremains challenging in natural, densely packed settings. Most existingapproaches rely on 2D imagery and therefore fail to capture truethree-dimensional morphology. Here, we present a curvature-based instancesegmentation framework for reconstructed surface meshes and demonstrate itsrole as a key step enabling 3D roundness and orientation analysis.In our approach, individual pebbles are detected directly from 3D surfacereconstructions using curvature features, without prior shape assumptions.Validation against high-resolution reference models yields a high detectionprecision of 0.98, with remaining errors mainly due to under-segmentationin overly smooth reconstructions. Estimates of 3D pebble orientation arestrongly controlled by the represented surface area, highlighting both thepotential and current limitations of orientation retrieval from incompletesurface segments.We illustrate how reliable segmentation allow downstream 3D shape androundness analyses that are not accessible in 2D, including curvature-basedsurface metrics and volumetric descriptors. Example fluvial scenesdemonstrate that segmentation quality directly controls the stability ofroundness estimates and their geomorphic interpretation. Our resultsestablish curvature-based 3D pebble segmentation as a methodologicalfoundation for reproducible analyses of pebble shape, roundness, andorientation in natural river systems.

Quantifying the structure and heterogeneity of complex spatial patterns is a key challenge in the analysis of spatial data across many scientific disciplines, including geoscience and geomorphology. The complexity of spatial patterns can be analysed by considering the recurrence of specific properties. While recurrence plot based methods are well established for analysing dynamical systems, their application to spatial patterns has received less attention. Here, we propose a novel approach that combines spatial recurrence analysis with a measure from fractal geometry, lacunarity, which originally quantifies homogeneity in spatial patterns. Applied to recurrence plots, it is referred to as recurrence lacunarity (RL) and quantifies the homogeneity of recurrences. Although recurrence plots can be generated from higher-dimensional data, RL has not yet been calculated for higher-dimensional (spatial) recurrence plots. To address this gap, we evaluate the RL of spatial data by validating the method using synthetic test patterns and then applying it to analyse hillslope gradients of river catchments near the Mendocino Triple Junction. The results demonstrate that the RL effectively detects and quantifies differences in the spatial structure of the catchments, which can be related to local uplift rates and the geological setting. RL provides a robust measure for comparing diverse spatial data sets and for quantifying how their spatial structure relates to external parameters, and may also be used as features in machine-learning models, complementing existing descriptors of spatial structure.

Sediment characteristics and grain-size distribution are crucial for understanding natural hazards, hydrologic conditions, and ecosystems. However, traditional methods for collecting this information are costly, labor-intensive, and time-consuming. To address this, we present OrthoSAM, a workflow leveraging the Segment Anything Model (SAM) for automated delineation of densely packed pebbles in high-resolution orthomosaics. Our framework consists of a tiling scheme, improved seed (input) point generation, and a multi-scale resampling scheme. Validation using synthetic images shows high precision close to 1, a recall above 0.9, with a mean IoU above 0.9. Using a large synthetic dataset, we show that the two-sample Kolmogorov-Smirnov test confirms the accuracy of the grain size distribution. We identified a size detection limit of 30 pixels; pebbles with a diameter below this limit are not reliably detected. Applying OrthoSAM to orthomosaics from the Ravi River in India, we delineated 6087 pebbles with high precision (0.93) and recall (0.94). The resulting grain statistics include area, axis lengths, perimeter, RGB statistics, and smoothness measurements, providing valuable insights for further analysis in geomorphology and ecosystem studies.

Sediment characteristics and grain-size distribution are crucial for understanding natural hazards, hydrologic conditions, and ecosystems. However, traditional methods for collecting this information are costly, labor-intensive, and time-consuming. To address this, we present OrthoSAM, a workflow leveraging the Segment Anything Model (SAM) for automated delineation of densely packed pebbles in high-resolution orthomosaics. Our framework consists of a tiling scheme, improved seed (input) point generation, and a multi-scale resampling scheme. Validation using synthetic images shows high precision close to 1, a recall above 0.9, with a mean IoU above 0.9. Using a large synthetic dataset, we show that the two-sample Kolmogorov-Smirnov test confirms the accuracy of the grain size distribution. We identified a size detection limit of 30 pixels; pebbles with a diameter below this limit are not reliably detected. Applying OrthoSAM to orthomosaics from the Ravi River in India, we delineated 6087 pebbles with high precision (0.93) and recall (0.94). The resulting grain statistics include area, axis lengths, perimeter, RGB statistics, and smoothness measurements, providing valuable insights for further analysis in geomorphology and ecosystem studies.

Sediment characteristics and grain-size distribution are crucial for understanding natural hazards, hydrologic conditions, and ecosystems. However, traditional methods for collecting this information are costly, labor-intensive, and time-consuming. To address this, we present OrthoSAM, a workflow leveraging the Segment Anything Model (SAM) for automated delineation of densely packed pebbles in high-resolution orthomosaics. Our framework consists of a tiling scheme, improved seed (input) point generation, and a multi-scale resampling scheme. Validation using synthetic images shows high precision close to 1, a recall above 0.9, with a mean IoU above 0.9. Using a large synthetic dataset, we show that the two-sample Kolmogorov-Smirnov test confirms the accuracy of the grain size distribution. We identified a size detection limit of 30 pixels; pebbles with a diameter below this limit are not reliably detected. Applying OrthoSAM to orthomosaics from the Ravi River in India, we delineated 6087 pebbles with high precision (0.93) and recall (0.94). The resulting grain statistics include area, axis lengths, perimeter, RGB statistics, and smoothness measurements, providing valuable insights for further analysis in geomorphology and ecosystem studies.