The research focuses on supercritical flow, the regime that develops when inertial forces overwhelm gravity, expressed technically as a Froude number greater than one. In this state, water slamming into a cylinder behaves in ways that have no counterpart in slower, subcritical rivers. The flow climbs the upstream face as an upward wall jet, periodically collapses backward in a phenomenon called reverse spillage, and generates surface shock waves analogous to sonic booms in aerodynamics. Previous work had identified these structures, including the horseshoe vortex system that wraps around an obstacle’s base and drives scour, but almost all of it assumed a smooth bed. Natural mountain rivers, however, are floored with coarse gravel where the ratio of obstacle size to grain size can be as small as two to ten, meaning individual stones are large enough to matter.

To capture this reality, the team combined laboratory experiments with extraordinarily detailed computer simulations. In a flume at the Universidad de Concepción, they placed a four-centimeter cylinder at the centerline of a mobile bed of agate stones between six and ten millimeters in diameter, tilted the channel to a four percent slope, and ran water at a discharge of 4.92 liters per second. The resulting flow was just under two centimeters deep, moving at 0.70 meters per second, giving a Froude number of 1.68 and a Reynolds number near 49,000. Such shallow, fast, air-entraining flow defeated conventional measurement techniques like particle image velocimetry, so the experiments pinned down bulk quantities while the simulations resolved the full three-dimensional flow field.

The digital side of the study is where the technical ambition becomes remarkable. The researchers photographed the bed with a smartphone camera and used Structure-from-Motion photogrammetry to build digital elevation models with an average resolution of 0.2 millimeters, fine enough to capture every individual grain. These surfaces were converted into computational meshes of roughly 13.5 million cells using OpenFOAM, with refinement down to 0.1 millimeters near the bed. The simulations employed detached eddy simulation, a hybrid turbulence approach that resolves large eddies directly in the main flow while modeling near-wall regions, coupled with a volume-of-fluid method that tracks the churning air-water interface. Time steps averaged around 2.5 microseconds, dictated by the demands of tracking the free surface through violent aeration.

The team examined three morphodynamic states that mirror natural channel evolution: a perfectly smooth bed analogous to bedrock channels, a flat rough bed representing gravel deposited after a flood, and an equilibrium scoured bed produced by running the flow at sediment-moving velocities for eight hours until a symmetric scour hole 3.5 centimeters deep had developed around the cylinder. The model validated well against the experiments, reproducing the wall jet climbing the obstacle, the reverse spillage rolling backward at the base at frequencies between roughly five and seven hertz, the aerated lateral jets, and the measured upstream water depth of 1.76 centimeters.

The headline finding concerns what roughness does to flow organization. Over the smooth bed, the time-averaged flow displayed textbook coherence: a well-defined horseshoe vortex system with a primary necklace vortex located about one obstacle radius upstream, secondary vortices further out, and a nearly symmetric bed shear stress distribution whose mean and median differed by only 1.5 percent. The coefficient of variation of stress, a measure of spatial variability, was 37 percent. Over the flat rough bed, that organized picture collapsed. Individual grains acted as micro-obstacles, each generating its own wake and separation zone, fragmenting the horseshoe vortex until the main necklace vortex was barely recognizable and secondary structures vanished entirely. The stress distribution became skewed, its coefficient of variation jumping to 86 percent.

The scoured bed produced the most complex fields of all, because large-scale topographic modification and grain-scale disruption operated hierarchically, with the scour hole reshaping the approach flow within which individual grains continued to generate their own turbulence. Across the three configurations, the coefficient of variation of bed stress climbed from 37 percent to 86 percent to 115 percent, while the mean stress actually fell, from 0.00183 to 0.00133 to 0.00115 square meters per second squared. The paradox is the point: smoother statistics hide wilder extremes. Localized stress concentrations around individual grain crests exceeded smooth-bed conditions by factors of two to three, and the stress distributions grew progressively right-skewed, from a skewness of 0.41 over the smooth bed to 2.60 over the scoured bed.

These results carry immediate consequences for engineering. Bridge piers and bank protection in steep channels are routinely designed using smooth-bed assumptions inherited from subcritical flow research, yet the study shows that natural rough beds subject structures and sediments to frequent low-stress periods punctuated by intense local peaks, a loading pattern that standard methods do not capture. The findings also reshape understanding of sediment transport, because entrainment of any given grain depends on the force balance on that single particle, not the reach-average stress. Meanwhile, the patchwork of high- and low-stress zones may sustain the microhabitat diversity that benthic organisms depend on, and the pressure gradients generated around individual grains may help drive the exchange of water between the river and its subsurface.

The authors are careful about the limits of their approach. The bed was fixed and impermeable, so the extreme stress heterogeneity they observe represents an instantaneous upper bound that grain mobility would erode over time as mobile stones respond to the flow. Validation relied on bulk parameters and visual comparison rather than spatially resolved velocity measurements, and grain-scale stress values at individual particle crests should be read as indicative patterns rather than precise predictions. Even so, the demonstration that supercritical flow over rough beds abandons the organized physics of subcritical flows, and instead becomes grain-dominated chaos, marks a turning point. Predicting how mountain rivers carve their channels, and how the structures we build within them will fare, may now require models that resolve, or at least honestly parameterize, every last stone on the bed.

Subject of Research: Grain-scale bed roughness effects on velocity and bed stress fields around obstacles in supercritical mountain river flow

Article Title: Grain roughness controls on velocity and bed stress fields around a fully protruding obstacle in supercritical flow

Article References: Monsalve, A., & Link, O. (2026). Grain roughness controls on velocity and bed stress fields around a fully protruding obstacle in supercritical flow. Earth Surface Dynamics, 14(3), 469-491. https://doi.org/10.5194/esurf-14-469-2026

Image Credits: AI Generated

DOI: 10.5194/esurf-14-469-2026

Keywords: supercritical flow, grain roughness, horseshoe vortex, bed shear stress, scour, mountain rivers, detached eddy simulation, Structure-from-Motion, sediment transport, bridge piers, turbulence, geomorphology

To understand why this matters, it helps to recall how geometric model fitting has evolved. Since Martin Fischler and Robert Bolles introduced the Random Sample Consensus paradigm, or RANSAC, in 1981, the standard approach has been to repeatedly sample small subsets of points, hypothesize a model such as a line, a plane, or a homography, and count how many points agree with it. When a scene contains only one structure and a moderate number of outliers, this strategy works remarkably well. But real scenes are rarely so cooperative. An image pair of a building contains dozens of planes at different depths; a point cloud of a street contains ground surfaces, vehicle bodies, and facades simultaneously. Multi-structure fitting requires the algorithm not only to find models but to segment the data, deciding which points belong to which model instance and which are simply noise.

In recent years, researchers have increasingly turned to density-based clustering algorithms to perform this segmentation. Methods such as DBSCAN, introduced by Ester and colleagues in 1996, group points based on how densely they pack together, and their robustness to noise made them attractive for fitting pipelines. However, as the authors of the new study point out in their abstract, these algorithms carry a critical limitation: they depend on a fixed threshold parameter, conventionally called MinPts, which specifies the minimum number of neighbors required for a point to be considered part of a dense region. This threshold often lacks adaptability to varying data characteristics. When input data is severely contaminated with outliers, determining suitable threshold values becomes genuinely difficult, and a poorly chosen value leads directly to suboptimal fitting results, with inlier structures fragmented or outliers absorbed into spurious clusters.

The core innovation of IACF is to let information theory decide the threshold automatically. Entropy, in the sense introduced by Claude Shannon, quantifies the disorder or uncertainty of a distribution. The researchers propose an entropy-based adaptive strategy that measures the disorder of local neighborhoods in the data and uses this measurement to dynamically determine an optimal clustering threshold, denoted MinPts*, for each specific dataset without any prior knowledge. In other words, instead of an engineer guessing how many neighbors should define a dense region, the algorithm inspects the statistical structure of the data itself and derives the answer. This is a meaningful shift: the parameter that most strongly controls density-based clustering behavior is no longer a fixed constant but a data-driven quantity that adapts to the contamination level and geometric complexity of each problem instance.

The method unfolds in three coordinated stages. First, the authors construct a reliable neighborhood graph by integrating motion and preference information. Preference analysis, a technique well established in the multi-model fitting literature through works such as T-linkage by Magri and Fusiello and accelerated hypothesis generation by Chin, Yu, and Suter, represents each data point by its pattern of agreement with a large set of sampled hypotheses. Points belonging to the same structure share similar preference vectors, so combining this preference signature with motion consistency information allows the graph-building step to prune invalid edges while preserving the potential local inlier structures that matter for fitting. This graph pruning is the first line of defense against outliers, removing connections that would otherwise let mismatched points contaminate each other’s neighborhoods.

Second, the entropy-based adaptive strategy operates on this graph, quantifying the disorder of local neighborhoods to set the optimal MinPts* threshold for the dataset at hand. Third, building on both components, the authors develop an Entropy-Adaptive Density-based Dominant Set Clustering algorithm, or EADSC. Dominant set clustering, formalized by Pavan and Pelillo in their influential work on pairwise clustering, is a graph-theoretic framework in which clusters emerge as dominant sets, subsets of vertices that are internally coherent with respect to their external environment. By making the dominant set search entropy-adaptive, EADSC can autonomously detect model-related subgraphs corresponding to individual model instances and accurately partition inliers from outliers, without the analyst supplying the number of structures in advance.

The experimental evaluation reported in the paper covers several challenging datasets, and the authors state that the proposed IACF method outperforms several state-of-the-art fitting methods in terms of both fitting accuracy and computational efficiency. That dual improvement is notable because robustness and speed are usually in tension in this field. Methods that achieve high accuracy on heavily contaminated data, such as hypergraph optimization approaches, energy minimization and mode-seeking frameworks, and residual-based consensus techniques, often pay a steep computational price, while faster methods sacrifice accuracy when outlier ratios climb. An adaptive threshold that correctly identifies inlier structure on the first pass reduces wasted hypothesis sampling, which is where much of the computational cost of fitting pipelines accumulates.

The context in which this work appears is a field under growing pressure from real-world deployment. The reference list of the paper traces the trajectory of the problem: from RANSAC and its many guided-sampling descendants, through preference-analysis methods and hypergraph formulations, to recent deterministic fitting approaches such as latent semantic consensus and spatial clustering guided two-view fitting, and graph-based techniques for mismatch removal. Applications cited in the surrounding literature range from multimodal image matching and feature matching with large numbers of outliers, to instantaneous frequency estimation of multi-component signals with multi-sensor consensus, to 3D point cloud map merging for large-scale environments and defect detection in printed-circuit boards. Each of these tasks shares the same mathematical skeleton: data generated by multiple geometric models plus noise, and an algorithm that must recover the models without knowing in advance how many there are or how badly the data is corrupted.

What makes the entropy angle particularly compelling is its generality. The authors connect their contribution to a broader line of research on autonomous clustering, including recent work by Yang and Lin on clustering by fast find of mass and distance peaks and toward autonomous distributed clustering. The shared ambition across these efforts is to remove human tuning from the clustering loop, letting statistical measures of structure determine algorithmic parameters. In the IACF framework, entropy serves as a universal yardstick of neighborhood disorder: a well-organized inlier neighborhood has low entropy, while a neighborhood polluted by outliers exhibits high disorder. By searching for the threshold that best separates these regimes, the algorithm effectively asks the data how much contamination it contains and adjusts its density criteria accordingly. This is precisely the kind of self-configuration that matters for systems operating in uncontrolled environments, where the outlier ratio can swing dramatically between frames or between scenes.

The practical implications extend across the computer vision stack. Robust multi-structure fitting sits upstream of fundamental estimation tasks: fundamental matrix and homography estimation in two-view geometry, motion segmentation in video, plane detection in point clouds, and registration of 3D scans. Improving the accuracy and speed of the fitting stage propagates improvements into every downstream application, from augmented reality anchoring to industrial inspection. The authors have made the datasets used in the study publicly available through the Adelaide RMF benchmark repository and a project page, and state that the custom code is available from the corresponding author upon reasonable request, which should make the method straightforward for other groups to evaluate against their own pipelines.

Funded by the Natural Science Fund of Fujian Province, the Natural Science Foundation of Xiamen, the National Natural Science Foundation of China, and several Xiamen municipal science and technology programs, the research reflects a sustained investment in the geometric foundations of machine perception. Whether entropy-adaptive thresholds will become a standard component of future fitting pipelines remains to be seen, but the study makes a clear case that the era of fixed MinPts values in density-based fitting may be drawing to a close. As vision systems are asked to operate in ever messier environments, algorithms that measure the disorder of their own inputs and adapt on the fly offer a persuasive path toward fitting that is both more accurate and more efficient, no matter how many structures the world throws at them.

Subject of Research: Information-theoretic adaptive density clustering for robust multi-structure geometric model fitting in computer vision

Article Title: Information-theoretic adaptive clustering for robust multi-structure model fitting

Article References: Information-theoretic adaptive clustering for robust multi-structure model fitting. (n.d.). https://doi.org/10.1007/s10586-026-06563-2

Image Credits: AI Generated

DOI: 10.1007/s10586-026-06563-2

Keywords: model fitting, density-based clustering, information entropy, adaptive threshold, dominant sets, RANSAC, computer vision, outlier removal, preference analysis, multi-model fitting, neighborhood graph, Cluster Computing

The review, authored by Bernardo Cândido of the University of Missouri, synthesised peer-reviewed studies published since 2005, tracing the methodological lineage from close-range digital photogrammetry through to today’s UAV-based workflows. Its central conclusion is deceptively simple: structure-from-motion photogrammetry directly measures soil-surface geometry and, in repeated surveys evaluated against uncertainty, detectable net surface change. It does not directly measure detachment, soil loss, sediment flux, sediment delivery or conservation success. Those are interpretive claims, and they become defensible only when detectability, mechanism, storage context and scale are explicitly constrained.

The technical reasons for this caution lie in the measurement chain itself. Structure from motion estimates camera geometry and sparse scene structure from overlapping images, and multi-view stereo then densifies that solution into a point cloud from which elevation models are derived. Each stage introduces uncertainty. Weak image networks, particularly largely parallel acquisition geometries, can introduce systematic deformation even when models look visually plausible, as James and Robson demonstrated in foundational work. Bare agricultural surfaces compound the problem: they are often planar, weakly textured and repetitive, so they constrain camera self-calibration less effectively than rougher natural terrain. Bright illumination and homogeneous texture can further degrade keypoint matching and introduce broad-scale vertical noise.

Georeferencing decisions matter just as much. Ground-control-point-based orientation, direct georeferencing from RTK or PPK camera positions, and stable local coordinate frames each carry distinct strengths and weaknesses, and camera-center coordinates used in bundle adjustment are observations rather than independent validation data. The review stresses that RTK or PPK positioning does not eliminate the need for spatially distributed checkpoints, and that a single global RMSE can coexist with poorer local precision. For multitemporal analysis, the decisive requirement is a stable date-to-date reference frame accompanied by independent, spatially explicit uncertainty assessment, because detectability varies across the scene rather than behaving as one uniform number.

The evidential strength of drone photogrammetry also varies systematically across erosion processes, and the review organises this variation into a four-level inference ladder. At the base sits surface reconstruction: recovering microtopography, roughness, rill form and many exposed gully shapes. Here the method is already robust. Studies have shown that structure from motion can recover bare-soil microtopography with useful accuracy, that rill geometry can be measured rapidly and non-destructively at plot scale, and that automated rill mapping across agricultural fields can achieve accuracy statistics above eighty percent, although performance drops for shallow incisions and where vegetation or linear surface features resemble rills.

Gullies, with their large relief signals, are particularly well suited to the technique, and early studies demonstrated volumetric analysis of headcut erosion in exposed sections. But strong relief does not eliminate geometric blind spots. UAV-only surveys become restricted on steep gully walls above roughly fifty to sixty degrees and on overhangs, and additional terrestrial imagery substantially improves the resulting three-dimensional models. Vegetation imposes a harder limit still: agreement between photogrammetric surfaces and terrestrial laser scanning degrades as ground cover increases and is significantly affected beyond about fifty-three percent cover. When soil visibility changes, the review notes, the target of the measurement changes too, and elevation differences recorded over vegetation or standing water cannot automatically be read as soil loss.

Diffuse, sheet and interrill erosion is the most demanding domain, because the expected topographic signal is small, spatially dispersed and often close to the magnitude of non-erosional change. Controlled experiments have shown that compaction and settlement can exceed the volume attributed to diffuse erosion under some conditions, and that part of the mobilised material may remain stored within convergent features rather than leaving the plot. On cultivated hillslopes, repeated surveys captured rainfall-related surface lowering, yet inferred mass loss decreased once soil settlement was accounted for through bulk-density measurements. In other words, soils can lower without exporting sediment, and a negative elevation difference that exceeds local uncertainty should first be reported as surface lowering, not erosion.

The strongest studies are those that close the inferential loop with independent evidence. One benchmark combined UAV monitoring of bare erosion plots under natural rainfall with direct sediment collection and spatially variable levels of detection, reporting strong correspondence between photogrammetric soil-loss estimates and measured sediment yield. Another paired drone surveys with RFID soil-particle tracking, runoff and sediment measurements at the plot outlet, and a sediment-connectivity index, revealing that short particle travel distances could coexist with increasing connectivity and higher sediment discharge. A post-fire Mediterranean catchment study combined photogrammetry with terrestrial LiDAR, GNSS and georadar to quantify erosion and sedimentation around a retention dam over several years. In each case, topography gained explanatory power because it was embedded in a broader process-observation framework.

Management claims sit at the top of the ladder and are the sparsest of all. A comparative study using post-storm drone imagery found that cover-cropped plots showed lower soil loss, shallower rills and smaller rill surface area than bare plots, a genuinely comparative design that moves beyond description. But even that study acknowledges that its reconstructed-baseline approach is constrained by vegetation conditions and requires stronger ground validation for quantitative use. Demonstrating that an intervention reduced erosion requires showing a change relative to a credible counterfactual, and the review finds that validated conservation-effectiveness claims remain much rarer than claims of detectable net surface change.

The practical implications reach beyond academic terminology. For soil conservation, drone surveys excel at spatial diagnosis, identifying runoff pathways, erosional source areas and depositional zones at the scale where interventions are actually placed. For erosion modelling, repeated UAV surveys provide spatially distributed time series that field-scale models have long lacked; one study using nine surveys found that WEPP reproduced observed erosion more closely than USLE, which systematically overestimated soil loss at the site. But surface lowering alone should not be assumed to represent sediment yield. The review’s agenda is clear: shared benchmark datasets across surface states, explicit separation of erosional from non-erosional change, standardised validation protocols for diffuse erosion, nested event-based monitoring, and reporting standards that keep every claim proportional to the evidence that supports it. Detailed surfaces are no longer the hard part; proving what they mean is.

Subject of Research: UAV structure-from-motion photogrammetry for measuring soil surfaces and inferring soil erosion

Article Title: A critical review of UAV structure from motion photogrammetry for measuring soil surfaces and inferring soil loss

Article References: Cândido, B. (2026). A critical review of UAV structure from motion photogrammetry for measuring soil surfaces and inferring soil loss. Discover Soil, 3(1), Article 140. https://doi.org/10.1007/s44378-026-00298-7

Image Credits: AI Generated

DOI: 10.1007/s44378-026-00298-7

Keywords: UAV photogrammetry, structure from motion, soil erosion, soil microtopography, rill erosion, gully erosion, diffuse erosion, DEM of difference, sediment redistribution, soil conservation, uncertainty, remote sensing