the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Feb 2026
Channel Dynamics in an Experimental Alluvial Fan Under Constant Boundary Conditions: A Classification of Avulsion and Lateral Migration Events
Abstract. Alluvial fans exhibit complex channel dynamics, spanning gradual lateral migration to sudden avulsions. Although allogenic processes are recognized as key drivers of these behaviors, the autogenic mechanisms regulating channel change remain poorly understood. In this study, we present a quantitative analysis of the main channel kinematics on a widely graded, non-cohesive experimental alluvial fan, utilizing high-temporal-resolution RGB imagery and main channel centerline tracking. By employing two key metrics – displacement magnitude (normalized by channel width) and flow continuity, defined as the degree of overlap in the active channel footprint from one image to the next – we move beyond qualitative assessments, which are often subject to researcher bias, and establish a clear framework for distinguishing between migratory (continuous) and avulsive (discrete) channel behaviours. Our findings reveal that the fan alternates between supply-limited phases, when a small number of efficient channels route sediment to the toe with only localized reworking, and transport-limited phases, when a more complex, inefficient channel network traps sediment mid-fan, favoring abrupt reorganizations (i.e., avulsions). Contrary to the conventional assumption that systematic aggradation from toe to apex triggers large-scale channel abandonment, we demonstrate that lateral sediment redistribution often prevents fan-wide sediment buildup, thereby delaying or even preventing major avulsions. These results highlight the critical role of self-regulating autogenic processes, particularly the lateral reworking of coarse sediment, in controlling both the timing and scale of channel adjustments, emphasizing the need to incorporate localized feedback mechanisms into predictive models to improve our understanding of alluvial fan dynamics.
- Preprint
(7229 KB) - Metadata XML
-
Supplement
(1635 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-5934', Anonymous Referee #1, 07 Mar 2026
-
AC1: 'Reply on RC1', Nastaran Nematollahi, 14 May 2026
We thank the reviewer for the careful and constructive review. The comments were very helpful and led to substantial revisions, especially in the Discussion. We believe these changes have made the paper clearer and improved the overall flow of the manuscript. The main manuscript was prepared in R Markdown. For the reviewer’s convenience, I exported the manuscript to Word so that the revisions could be shown using Track Changes. The edits were applied in the Word version to make the changes easier to review.
Please note that this Word file is provided only for tracking and reviewing the revisions. It is not intended to reflect the final journal-formatted manuscript.
according to the journal requirements.
Response to Major Comment 1:
We agree that PIV, Dynamic Time Warping, and image-stacking approaches are important tools for quantifying channel mobility. In the revised manuscript, we clarify how our approach differs from these methods. Our framework addresses two related but separate questions: how channel displacement is measured, and how the mechanism of change is classified as continuous migration or discrete avulsion. The cited methods address these questions differently from our approach. PIV-based methods, such as Chadwick et al. (2022, 2023), estimate displacement by comparing small interrogation windows between successive images. The method takes a small patch of one image, searches for the most similar-looking patch in the next image, and uses the offset between the two patches to estimate motion. When channel boundaries or image textures move gradually enough that the same features remain recognizable between time steps, the method performs reliably; when channel relocation is abrupt, the original channel texture may move outside the search window or disappear entirely, and the algorithm either fails or returns a misleading match. PIV can therefore detect continuous migration but cannot classify cases where the main channel abruptly abandons one pathway and establishes another. Chadwick et al. acknowledge this limitation in both papers. DTW, as used by Sylvester et al. (2019) and Li and Limaye (2025), compares centerlines with different point spacing and finds an optimal alignment between two channel traces. DTW was not used here because our objective was not to optimize the geometric match between two curves but to quantify how far the main channel shifted across the fan at equivalent radial positions from the apex. Alluvial fans have a naturally radial geometry. Our method, therefore, resamples each centerline at fixed radial distances from the fan apex, computes the angular separation Δθ at each radius, and converts that angular change to arc-length displacement using L = r·Δθ, where r is the radial distance from the apex. This gives a direct measure of cross-fan displacement at a known radial position. DTW would improve curve alignment, but because it allows flexible matching between non-equivalent downstream positions, it obscures where along the fan the displacement occurred. Image-stacking methods, such as Lee et al. (2022), detect abrupt channel shifts from longer-term water-occurrence records. The approach is designed for remotely sensed river records where observations are less frequent and changes accumulate over time. In our experiment, the channel changed rapidly, and multiple adjustments could occur over short intervals. With one-minute active-flow maps, we used the high-frequency footprints directly rather than reducing the record to a long-term occurrence stack. This allowed us to classify each event using both displacement magnitude and whether active flow occupied the intervening pathway during the event window.
The mechanism of change in our method is identified by combining displacement magnitude with the fractional overlap of the active-flow footprint between successive time steps. The overlap criterion is feasible because the imagery has high temporal resolution, so successive channel positions are close enough in time for overlap to be meaningful. We do not present the overlap threshold alone as the main novelty. The contribution is the combination of dense optical-flow-derived active-flow maps, fan-based radial displacement measurements, and an overlap criterion to classify short-interval channel shifts in a controlled alluvial fan experiment.
Response to Major Comment 2:
Displacement magnitude alone does not distinguish migration from avulsion, which is why our framework combines it with flow overlap. Magnitude describes the scale of the event; overlap identifies the mechanism. The concern that overlaps would approach zero for any displacement greater than one channel width is valid for a snapshot-based footprint. Our active footprint, however, is a one-minute composite derived from 20 optical-flow frames collected at 3-second intervals, recording the full area occupied by active flow during that minute rather than only the start and end positions. When a channel migrates continuously across one channel width, all intermediate positions are captured in the footprint, producing substantial overlap with the next interval. When a channel avulses within a single 3-second frame, the footprint consists of two narrow, spatially separated active areas with inactive pixels between them, yielding low overlap even when the total displacement is comparable. To make this clearer in the manuscript, we added a sentence to the Methods section stating explicitly that the active-flow footprint represents the cumulative area occupied by flow during the full one-minute interval.
Response to Major Comment 3:
The main motivation for developing this framework was to reduce reliance on subjective visual interpretation in experimental alluvial fan studies. Experimental fans evolve rapidly, and channel networks can reorganize within minutes. However, channel behavior in laboratory experiments is often described qualitatively from video or imagery, or summarized using coarse metrics. We aimed to define channel behavior in more objective and reproducible terms. We agree with the reviewer that direct application to seasonal or annual satellite imagery is not straightforward, the temporal resolution gap between laboratory and field settings is narrowing rapidly. Daily satellite imagery (e.g. Planet Labs; Planet Team, 2017) and UAV-based surveys now make high-frequency monitoring of active fans increasingly feasible. The framework is not inherently limited to laboratory use. It requires high temporal resolution, which is becoming more accessible in field settings.
We have revised the Discussion to make this scope clearer. Specifically, we now state that the framework is most applicable to datasets with sufficient temporal resolution to resolve intermediate channel positions.
Response to Major Comment 4:
We agree that our original wording could sound like we were arguing against the role of aggradation and channel perching in avulsion. That was not our intention. We revised the Discussion and Conclusions to make this clearer. We now state that our experiment does not directly test the vertical perching mechanism because we did not track channel-bed elevation relative to the surrounding fan surface. We also clarified that aggradation-driven preconditioning and cumulative lateral migration are not mutually exclusive. Instead, our results suggest that repeated lateral reworking can redistribute sediment within the active corridor and may influence whether, when, or how abrupt channel abandonment occurs.
Response to Major Comment 5:
We agree that the Results needed to make the basis for the supply-limited and transport-limited interpretation clearer. Although discharge and sediment supply were held constant throughout the experiment, the local balance between sediment supply and transport capacity changed as the channel network reorganized. When the network expanded across the fan surface, flow was distributed among multiple threads. This likely reduced the transport capacity of individual channels and promoted sediment accumulation and reworking across the mid-fan, which we interpret as a relatively transport-limited state. When sediment accumulation was followed by channel consolidation, flow became concentrated in one or a few dominant threads and sediment appeared to be routed more efficiently toward the fan toe. We interpret this as a relatively supply-limited, or more efficiently connected, state. This interpretation is consistent with storage-and-release dynamics documented in previous experimental fan studies (Kim et al., 2006; Hamilton et al., 2013; Reitz et al., 2012).
Direct measurement of sediment flux or local transport capacity was not feasible in our small-scale physical model. We therefore inferred these states from channel-network behavior using three metrics: the number of active channel threads, the TMC ratio, and the angular spread of the active network. These metrics were combined into a single complexity index that describes whether the channel network was relatively simple and structured, with activity concentrated in one or a few dominant channels, or more complex and distributed across multiple active threads.
We added the full complexity-index methodology and the detailed quantitative procedure in the Supplementary Material so that the main text could remain focused on the displacement-classification framework. In the revised manuscript, we added a clearer explanation in the discussion describing how network complexity relates to the inferred sediment-routing state. We also added an example figure showing contrasting channel-network patterns: simpler, more concentrated networks are interpreted as relatively supply-limited or efficiently connected states, whereas more complex, distributed networks are interpreted as relatively transport-limited states.
Response to Comment 6
We have revised the manuscript to address each point. References to Fig. 4(d) and 4(e) have been added to the relevant paragraph in the Methods section, and Fig. 5 and Fig. 8 are now explicitly cited in the main text at the appropriate points in the narrative. A colour bar and description of what the colours represent have also been added to Fig. 3.
Minor commnet s:
We have addressed these points in the revised manuscript by adding the imagery spatial resolution, revising Fig. 3 to include a color bar and a clearer description of optical-flow motion magnitude, expanding the Fig. 4 caption to describe all panels, and adding missing citations to Fig. 4d–e, Fig. 5, and Fig. 8 in the main text.
We clarified that points were matched by equivalent radial distance from the fan apex, not by tracking individual pixels or material points. For each sampled radial distance, we compared the angular position of the main channel at time with the angular position at the same radial distance at time . Distal points without a corresponding radial position in both time steps were excluded from the angular displacement analysis. And Finally, we revised the sediment-connectivity section by softening the “first-order control” statement, adding supporting references, and removing the unsupported claim about preferred avulsion-node locations
-
AC1: 'Reply on RC1', Nastaran Nematollahi, 14 May 2026
-
CC1: 'Comment on egusphere-2025-5934', Safiya Alpheus, 14 Apr 2026
Publisher’s note: this comment is a copy of RC2 and its content was therefore removed on 21 April 2026.
Citation: https://doi.org/10.5194/egusphere-2025-5934-CC1 -
RC2: 'Comment on egusphere-2025-5934', Safiya Alpheus, 20 Apr 2026
In this manuscript, the authors present findings from their experimental study characterising different channel reorganisation regimes across an unperturbed, widely-graded alluvial fan. I believe their findings are both interesting and pertinent to this journal’s audience and the broader scientific community. The authors quantify—in extremely high resolution—channel migration and reworking across an experimental fan, and use these data to place quantitative constraints on different channel behaviours, differentiating between continuous events (lateral migration) and discrete events (avulsion). The framework that the authors propose classifies discrete vs. continuous events using displacement magnitude and continuity of channel overlap.
In their manuscript, the authors importantly show that cumulatively, lateral migration events can lead to as much channel displacement across fan surfaces as avulsion events, and can lead to comparable amounts of reworking. From the experimental observations, the authors also discuss how supply-limited and transport-limited flow regimes influence deposition and reworking across channel networks, where supply-limited regimes facilitate channel-coupling across the fan surface, efficiently transporting sediment through the network, while in transport-limited stages sediment accumulates more frequently in channels, triggering channel avulsions.Generally, I think this manuscript is well-written and the author’s have clearly documented the motivation, methodology and key findings of this study. I have highlighted 5 thoughts on how the authors could provide more clarity to the reader, particularly on contextualising this work and adding some clarity to the figures and results.
Major comments:
- The authors use high-resolution imagery of their experimental system to quantify migration and reworking across the alluvial fan, however I believe that the reader would have a greater appreciation for the utility of these methods and the boundaries that the authors propose if they were placed in context with existing methods for characterising channel reworking and migration from channel imagery (e.g., from Particle-Image Velocimetry e.g. Jarriel et al., 2021 and CHadwick et al., 2024; Centerline measurements e.g., Schwenk et al., 2015; Area-Based Metrics e.g., Wickert et al., 2013, Greenberg et al., 2024).
- Similarly, I think it would be helpful for the discussion if the authors touched on how these findings or the classification criteria could be translated to natural systems where, in the best case, repeat satellite imagery is available on daily to weekly timescales.
- The authors use supply-limited and transport-limited flow regimes to frame some of the discussion around avulsion, however they do not define these regimes for the reader. I think the main text would benefit from some discussion on how the authors are defining these different flow regimes in their model, especially as the system is under constant conditions.
- The figures, tables and captions could be made clearer for the reader. For example, there aren’t any details in the Figure 4 caption for panels 4e and 4f and it is unclear what these plots show from the legend. Also, Table 2 and Figure 7 are a bit difficult to follow—it is not clear from the caption in Table 2 that the numbers in the ‘Channel Behavior’ column map to the behaviours listed in Figure 7, this makes this section of the manuscript slightly difficult to follow. Furthermore, Line 220 says that the behaviours in Table 2 are listed in order of increasing mean displacement but in Table 2 they are increasing. This language should be made clearer.
Citation: https://doi.org/10.5194/egusphere-2025-5934-RC2 -
AC2: 'Reply on RC2', Nastaran Nematollahi, 14 May 2026
We thank the reviewer for the careful and constructive review. The comments were very helpful and led to substantial revisions, especially in the Discussion. We believe these changes have made the paper clearer and improved the overall flow of the manuscript. The main manuscript was prepared in R Markdown. For the reviewer’s convenience, I exported the manuscript to Word so that the revisions could be shown using Track Changes. The edits were applied in the Word version to make the changes easier to review.
Please note that this Word file is provided only for tracking and reviewing the revisions. It is not intended to reflect the final journal-formatted manuscript.
Response to Major Comment 1:
We agree with the reviewer that the utility of our approach is clearer when placed in the context of existing image-based methods for measuring channel migration and reworking. In the revised manuscript, we clarify how our approach differs from these methods. Our framework addresses two related but separate questions: how channel displacement is measured, and how the mechanism of change is classified as continuous migration or discrete avulsion. The cited methods address these questions differently from our approach. PIV-based methods, such as Chadwick et al. (2022, 2023), estimate displacement by comparing small interrogation windows between successive images. The method takes a small patch of one image, searches for the most similar-looking patch in the next image, and uses the offset between the two patches to estimate motion. When channel boundaries or image textures move gradually enough that the same features remain recognizable between time steps, the method performs reliably; when channel relocation is abrupt, the original channel texture may move outside the search window or disappear entirely, and the algorithm either fails or returns a misleading match. PIV can therefore detect continuous migration but cannot classify cases where the main channel abruptly abandons one pathway and establishes another. Chadwick et al. acknowledge this limitation in both papers. DTW, as used by Sylvester et al. (2019) and Li and Limaye (2025), compares centerlines with different point spacing and finds an optimal alignment between two channel traces. DTW was not used here because our objective was not to optimize the geometric match between two curves but to quantify how far the main channel shifted across the fan at equivalent radial positions from the apex. Alluvial fans have a naturally radial geometry. Our method, therefore, resamples each centerline at fixed radial distances from the fan apex, computes the angular separation Δθ at each radius, and converts that angular change to arc-length displacement using L = r·Δθ, where r is the radial distance from the apex. This gives a direct measure of cross-fan displacement at a known radial position. DTW would improve curve alignment, but because it allows flexible matching between non-equivalent downstream positions, it obscures where along the fan the displacement occurred. Image-stacking methods, such as Lee et al. (2022), detect abrupt channel shifts from longer-term water-occurrence records. The approach is designed for remotely sensed river records where observations are less frequent and changes accumulate over time. In our experiment, the channel changed rapidly, and multiple adjustments could occur over short intervals. With one-minute active-flow maps, we used the high-frequency footprints directly rather than reducing the record to a long-term occurrence stack. This allowed us to classify each event using both displacement magnitude and whether active flow occupied the intervening pathway during the event window.
The mechanism of change in our method is identified by combining displacement magnitude with the fractional overlap of the active-flow footprint between successive time steps. The overlap criterion is feasible because the imagery has high temporal resolution, so successive channel positions are close enough in time for overlap to be meaningful. We do not present the overlap threshold alone as the main novelty. The contribution is the combination of dense optical-flow-derived active-flow maps, fan-based radial displacement measurements, and an overlap criterion to classify short-interval channel shifts in a controlled alluvial fan experiment.
Response to Major Comment 2:
The main motivation for developing this framework was to reduce reliance on subjective visual interpretation in experimental alluvial fan studies. Experimental fans evolve rapidly, and channel networks can reorganize within minutes. However, channel behavior in laboratory experiments is often described qualitatively from video or imagery, or summarized using coarse metrics. We aimed to define channel behavior in more objective and reproducible terms. We agree with the reviewer that direct application to seasonal or annual satellite imagery is not straightforward, the temporal resolution gap between laboratory and field settings is narrowing rapidly. Daily satellite imagery (e.g. Planet Labs; Planet Team, 2017) and UAV-based surveys now make high-frequency monitoring of active fans increasingly feasible. The framework is not inherently limited to laboratory use. It requires high temporal resolution, which is becoming more accessible in field settings.
We have revised the Discussion to make this scope clearer. Specifically, we now state that the framework is most applicable to datasets with sufficient temporal resolution to resolve intermediate channel positions
Response to Major Comment 3:
We agree that the Results needed to make the basis for the supply-limited and transport-limited interpretation clearer. Although discharge and sediment supply were held constant throughout the experiment, the local balance between sediment supply and transport capacity changed as the channel network reorganized. When the network expanded across the fan surface, flow was distributed among multiple threads. This likely reduced the transport capacity of individual channels and promoted sediment accumulation and reworking across the mid-fan, which we interpret as a relatively transport-limited state. When sediment accumulation was followed by channel consolidation, flow became concentrated in one or a few dominant threads and sediment appeared to be routed more efficiently toward the fan toe. We interpret this as a relatively supply-limited, or more efficiently connected, state. This interpretation is consistent with storage-and-release dynamics documented in previous experimental fan studies (Kim et al., 2006; Hamilton et al., 2013; Reitz et al., 2012).
Direct measurement of sediment flux or local transport capacity was not feasible in our small-scale physical model. We therefore inferred these states from channel-network behavior using three metrics: the number of active channel threads, the TMC ratio, and the angular spread of the active network. These metrics were combined into a single complexity index that describes whether the channel network was relatively simple and structured, with activity concentrated in one or a few dominant channels, or more complex and distributed across multiple active threads.
We added the full complexity-index methodology and the detailed quantitative procedure in the Supplementary Material so that the main text could remain focused on the displacement-classification framework. In the revised manuscript, we added a clearer explanation in the discussion describing how network complexity relates to the inferred sediment-routing state. We also added an example figure showing contrasting channel-network patterns: simpler, more concentrated networks are interpreted as relatively supply-limited or efficiently connected states, whereas more complex, distributed networks are interpreted as relatively transport-limited states.
Response to Major Comment 4:
n the revised manuscript, we have expanded the Figure 4 caption to describe all panels, including panels 4e and 4f, and clarified the legends so that the channel-detection workflow is easier to follow. We also revised the caption and formatting of Table 2 and Figure 7 to make the connection between the numbered channel-behaviour classes in Table 2 and the behavioural categories shown in Figure 7 explicit. Finally, we corrected the wording around Line 220 to clarify the ordering of behaviours in Table 2 and avoid confusion about whether they are listed by increasing or decreasing mean displacement.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 890 | 488 | 87 | 1,465 | 216 | 167 | 188 |
- HTML: 890
- PDF: 488
- XML: 87
- Total: 1,465
- Supplement: 216
- BibTeX: 167
- EndNote: 188
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “Channel Dynamics in an Experimental Alluvial Fan Under Constant Boundary Conditions: A Classification of Avulsion and Lateral Migration Events”
This manuscript presents an experimental investigation of channel dynamics on a widely graded alluvial fan. The active flow area is identified using high-frequency imagery and optical flow analysis. Active channel centerlines were extracted from these areas, and a quantitative framework was developed to classify channel displacement into continuous lateral migration and discrete avulsion based on channel displacement magnitude and flow overlapping. The study demonstrates that cumulative lateral migration can produce net channel displacement comparable to avulsions and discusses the alternation of the fan between supply-limited and transport-limited states under constant boundary conditions. Overall, the manuscript is well written and addresses an important problem in fluvial geomorphology: how to distinguish between gradual channel migration and discrete avulsion. The use of high temporal resolution imagery and a classification scheme combining channel displacement magnitude and flow-path overlap is a valuable contribution. The results highlight the role of repeated lateral migration in redistributing sediment in alluvial fans.
However, the methodology and interpretation require further improvement and clarification before the manuscript can be considered for publication. In particular, the novelty of the proposed method relative to existing techniques needs to be discussed, the classification thresholds needs to be improved, and the implications and broader applicability of the method to natural systems require further discussion. Some figures need further refinement. Therefore, I recommend major revisions to the manuscript.
Major comments:
Minor comments:
Line 109: Suggest adding the spatial resolution of the imagery
Figure 3: It would be better if adding a colorbar and a description of “flow motion values” regarding whether it represents the magnitude or direction of the motion.
Figure 4: Panels e and f don’t have captions. Also, there seems to be no citation of Fig. 4d and e in the main text.
Line 155: “position of a point at two time steps” – How were points between the two time steps matched? For example, in Fig. 5a, there are several points at the end of the down fan in blue that do not have a corresponding point on the down fan that is colored in orange. Is there a reason why these points are removed from the angular displacement analysis?
Figure 5: There is no citation of Fig. 5 in the main text.
Line 212: “between consecutive images” – suggest reinforce the temporal resolution here
Figure 8: Section 3.4 is describing Fig. 8 but there is no citation of it in the main text.
Line 318-319: “Beyond distinguishing lateral migrations from avulsions, our data illustrate how sediment connectivity exerts a first-order control on fan dynamics” - Suggest rewrite this sentence as more data might be needed to support this statement.
Line 319-322: References might be needed for these sentences.
Line 329: “which, in turn, determine the preferred locations of avulsion nodes and ultimately drive shifts in alluvial style.” – The data here did not show anything about the location of avulsion nodes.