Review of Harel et al.: “Drainage reorganization induces deviations in width-area-slope scaling of valleys and channels”

Please contact me at charles.shobe@mail.wvu.edu if anything is unclear.

In this contribution, the authors seek to test the hypothesis that drainage reorganization changes the scaling among drainage area, slope, and valley (and channel; more on that later) width. They use topographic data and field observations from an area undergoing drainage divide migration to show that beheaded channels have large, positive area-valley width exponents (width is large for a given area) and reversed channels have negative area exponents (width decreases with increasing drainage area) relative to channels they deem to be unaffected by piracy. Slope results are less clear, but the authors do find that reversed channels show different slope exponents between the channel, which has likely adjusted to the reversal, and the valley, which has not. Results suggest that area-width scaling might be useful for finding channels affected by divide migration, and also point to the fact that assumptions of instantaneous adjustment of width to area variations (i.e. those included in effectively all modern landscape evolution models) are to some extent wrong.

This paper speaks to a couple of topics that are of great relevance to geomorphology right now, notably 1) the importance of drainage divide migration in shaping landscapes and 2) the need to improve our understanding of channel width dynamics if we want to be able to simulate landscape evolution. I found this paper a joy to read, and it is overall in good shape. It is mostly easy to follow, asks and answers an interesting question, and the results are largely well supported by the data. I do have a couple of conceptual/methodological concerns that I would like to see addressed before publication. I hope the authors find my suggestions helpful in improving what is already a very interesting piece of work.

Overall points:

1. Writing and organization of the introduction: We as readers do not find out the knowledge gap that the authors intend to fill (aside from the abstract) until line ~115 or so, after a reasonably thorough discussion of channel width theory and drainage reorganization dynamics. I strongly suggest that before section 1.1, the authors make a statement similar to lines 116-119 stating that they are going to investigate width patterns in basins undergoing capture events. This could follow reasonably well from the existing lines 61-63. That way, readers will know the point of the paper before they slog through the width theory.
2. Channel width versus valley width: I know the authors are aware of this distinction, but it is a tricky issue. The classical width scaling ideas apply to river channels, but the width measurements that ultimately get made are of valley width (Figure 3). While I appreciate that Equation 2 is in some sense “general” (line 218), it does ultimately rest on the idea of open channel flow operating according to Manning’s equation and having a constant bankfull width/depth ratio, whereas valley width is often conceptualized as being dependent on some measure of lateral channel mobility (e.g. Hancock and Anderson, 2002; Limaye and Lamb, 2014) which may or may not be correlated with channel width. Given the massive size of these beautiful valleys (Figure 2), surely their cross-sections are not analogous to those of a channel in terms of being self-formed by some formative discharge. I do not suggest that the authors re-do their analysis to try to only use channel width, because that would be a lot of work and probably not possible in this intriguing landscape of broad-valleyed ephemeral channels. I would like to see however two things: 1) a dedicated, well-referenced subsection in the discussion clarifying for readers how the focus on valley width (as opposed to channel width) might influence the results of the study (for example, incorporating—or dismissing if you disagree—my comment above that valley width can in some cases be a function of channel mobility rather than width) and 2) more careful word choice when describing the quantities measured in the study. Often just the term “width” is used; I know it might seem tedious but I do think it would be helpful to make sure it is always prefaced by either “channel” or “valley.”
3. Slope data: Slope data extracted from DEMs is notoriously noisy (the classic discussion is Wobus et al., 2006, already cited by the authors). In this case the authors blame that noise for the strange near-complete lack of slope influence on width in all but the reversed valleys (line 429). I suspect that they are right, but this leads me to wonder about the slope measurement methodology. Of course slopes in these channels can be low as the authors note, and that makes things hard. But I have a few other ideas about methodological causes for this result.
   1. Figure S5 brings up an interesting point, which is that the D8 flow path along which slope is measured may be substantially more sinuous (and therefore lower sloping) than the actual path “felt” by the water during a formative flow event. Is the flow path in Figure S5 actually showing a thalweg that is substantially enough incised into the valley that it would guide formative flows? Or would flow basically be straight down-valley (and therefore experience a much higher bed slope)? I could easily be wrong, but supposing that the latter case was true it might make more sense to simply use the valley centerline as the line along which slope is approximated.
   2. On a separate note, Wobus et al 2006 discuss log-bin averaging schemes for slope data to reduce the noise. Did the authors try log-bin averaging? If so and it didn’t work, it would be good to say that in the paper. If not, it would be worth a shot to see if that changes the relationships between slope and width.
   3. One last thought: calculating the slope between any two DEM points is always an error-prone thing given the vertical error in DEMs. What about a linear regression along the five pixels where the best-fit slope then gets assigned to the middle pixel? This must be less sensitive to the error in any individual pixel.
   4. My concrete suggestion here is that the authors make sure that their somewhat surprising results with respect to slope are not simply a function of methodology. This could be done by testing other slope calculation and averaging methods, or by explaining very clearly why other methods failed or weren’t tried.

Line comments:

34: “control river dynamics”

37: “and hydrologic modeling”

64: As detailed in main comment 1, this is a good place to say how your study contributes to understanding of width under transient conditions

65: Consider being a bit more specific in the header: “channel and valley width scaling” or similar

70 and other equations: the multiplication symbol is not needed

83: Consider including a brief half-sentence explanation for why landslides influence width scaling. It will not be obvious to everyone I think

86: specify rock uplift?

97: see line 70 comment

112: I think there is more certainty here than your language implies. I suggest changing to “width adjustment does not occur…”

113: Can you be more specific about what constitutes a “long” timescale?

120: This is one example of an issue that crops up repeatedly. Which type of width, channel or valley? If both, state that explicitly.

127-131: This is an excellent statement of the work’s contribution!

152: can delete “had”

Figure 1: The level of zoom in (a) is awkward: not close enough to see distinct geomorphic features, and not far enough out for unfamiliar readers to know where in the world they are. I suggest adding another panel for a true “location figure,” which could then allow (a) to be zoomed in farther if you want.

Figure 1: (b) is a beautiful illustration; very nice work.  

Figure 1: looking at (c) brings up a question for me: how undisturbed are the “undisturbed” basins? Surely they must be experiencing some drainage area loss to escarpment migration. I agree with the authors that the data show diagnostic differences between those streams and the beheaded and reversed ones, but it might be worth adding a few sentences assessing the extent to which any stream draining this divide can be considered “undisturbed” since likely it just means losing drainage area at some much slower rate than the beheaded streams.

172: can delete “a”

178: there is some vague terminology here: “relatively rapid” and “slower.” Can these be made any more specific?

183: maybe “indicated” instead of “recognized?”

215-219: see major comment 2. This is where I think you need a much more detailed justification for extrapolating a relation based on open-channel flow mechanics from the channel scale to the valley scale.

235-244: This procedure necessarily contains some steps that could be considered subjective (e.g., manual polygon editing, etc). That is understandable. But when paired with the authors’ statement of “data available upon request,” it makes it hard to view this study as easily reproducible. I don’t think the journal requires it (though they should!), but I would strongly encourage the authors to archive all the GIS files relevant to their valley width measurements in a Figshare repository or similar. Then I would be less concerned about the fact that polygons had, for example, to be manually edited.

245: I have always written “thalweg,” but the EGU typesetting staff will sort it out one way or the other.

255: See my major comment 3 on slope extraction. In short, I am not sure that this is the best procedure (endpoint differencing along the flow line that is not binned/averaged in any way).

270: Again the lack of any binning worries me given the recommendations of Wobus et al (2006). There could be, at a minimum, some justification of why none was done.

300: I also recommend reporting the DEM versus dGPS measurement differences in % in addition to m.

330-334: In line with major comment 3, I worry that this result could be partially due to methodology. I could certainly be wrong, but it is at least worth a bit more of a detailed treatment in the discussion.

Figure 5: An easy way to make this more readable would be to label the exponent x-axis labels with the quantity they correspond to, e.g. “b exponent (area)”

351: “Farther,” not “further”

354: “for an additional”

Figure 6: The valley bottom outline does not look very convincing in the context of the air photo because it is hard to visualize the breaks in slope that should delineate it. Consider replacing the air photo with a DEM as the background on this image, or add another panel with the DEM.

373: you undersell your results! Consider replacing “could be” with “are”

374: “could be” -> “can”

453: can we get any more specific than “relatively rapidly?”

473: Again related to my point about channel versus valley width: if there are places where the two are especially similar, it seems like we should be prepared to understand differences that might occur in the places where the two widths are NOT similar. This is one of the points I’d like to see added in a discussion of this issue.

535: You could mention specifically here that this simplified approach is what is used in effectively all large-scale LEMs. One of the key exciting results of your study to me is a path forward for breaking down some of these oversimplistic assumptions regarding concordance between area and width.

Thank you for the chance to read this interesting paper!

Review of "Drainage reorganization induces deviations in width-area-slope

scaling of valleys and channels" by Harel et al.

Summary:

This contribution uses field observations, TanDEM-X-derived DEMs, and high-resolution SfM (and GNSS) field surveys to study how valley and channel width vary with drainage area for watersheds experiencing drainage reorganization.  To do this, they use a series of 12 watersheds, some of which have been previously documented, in the Negev desert.  Here, a rift-like structure has produced a vertical offset on the order of 500 m, which cuts across the heads of a series of (formerly) west-flowing drainages.  This offset creates a knickpoint, which redirects flow within a portion of the formerly west-flowing channels to the east.  As the knickpoints expand, the divides migrate.  However, incision has not been sufficient to erode a valley-depth of the landscape, and so the valley geometries of the former flow direction are preserved within this landscape.  The authors exploit this fortuitous circumstance to measure how the headward migration of the drainage divide produces channel width - area relations at odds with the valley width - area relations encoded in the former regime.  This, with some scaling arguments used to calculate stream power in the channels, suggests that divides may be more mobile than might appear due to a simple stream-power rule because of the narrowing of channel width that accompanies incision of the east-flowing channels.

General Comments:

1) This is a very nicely executed study in a fascinating natural system.  The fact that the old valley network can be clearly delineated and measured is fortuitous for addressing this problem.

2) The study is predicated on the idea that valley width scales as a function of watershed area in a way that is similar to channel width.  While this may empirically be the case, the valley width (in fluvial environments) may be set by a different set of factors (e.g., erodibility of the bank materials that scales migration rate, human modification of channel systems such as mill ponds, landslide damming of rivers) than the channel width.  I'm fine with an empirical measurement of valley width to demonstrate that scaling (for instance, the Beeson work), and these different factors are acknowledged.  But, it could be valuable to dedicate a bit more description to this distinction – that one can be related, somewhat directly to the flows that traverse the channel, while the other will be related to the evolution and migration of the channels over time. 

3) One contribution of this work is to present a novel method of extracting valley width, which is useful for a number of types of studies as the citations of the authors indicate.  It seems to work particularly well for a class of valley morphologies in which the valley floor can clearly be distinguished from the valley side-slopes by assigning a (calibrated) slope threshold.  It's worth noting that this is not generally the case, and so this method will fail when there is a gradation between valley floor and side-slopes.  Also, it's probably worth noting that the valley width in this case is the width of the valley floor.  Finally, I would suspect that the method works best when the valley and channel sinuosity are similar.  Otherwise, might a sinuous channel lead to aberrant projections normal to the trend of a sinuous inner channel to the VBET?  These issues might be discussed a bit more in the methods section.

4) The three classes of watersheds, "Undisturbed", "Beheaded", and "Reversed", seem to imply a conception of how water is routed through this landscape.  In particular, if one were to simply consider the DEM, probably all of the watersheds would be "Beheaded" or "Reversed", since the escarpment cuts across the apparent heads of the former drainages.  But, it seems from Figure 1b that the network is incised into a low-relief surface.  Figure 1b thus implies that the distinction between a beheaded and undisturbed channel rests on whether or not the incised portion of the drainage has been cross-cut by the escarpment / other channels.  Maybe I am confused about this, but if this is indeed the case, then a channel is undisturbed only if water is sourced exclusively from the incised portion of the landscape.  If this is the case, I think the watershed areas must be calculated based on those incised areas of the landscape, rather than the entire DEM.

5) The idea that channel widths adjust more quickly than valley widths seems intuitive.  Thus, in a landscape with such clear divide migration, one would expect channels widths to increase with area, while the valley widths might behave in an opposite manner, which is I think the main point of the paper.  A couple of thoughts on this:

A) I think the reversed valley width scaling can persist for no longer than the valley width divided by the incision rate * average hillslope angle.  The signature could be much shorter, but this would be the maximum amount of time that would be required to remove the former valley morphology.

B) I was not completely on board with the association of the valley and channel widths in the eastward-flowing channels for the stream-power calculation (and the inferences that flow from this).  The authors assert that in the westward-flowing channels, there are a series of anastomosing low relief-channels that occupy virtually the entire valley width.  I think this is being shown in Figure 7, where the width measurements for the stream power calculations are the yellow "width measurements"?  Yet, the photograph of the valley seems to indicate that only a fraction of those valleys are occupied by a channel.  If this is the case, the width effect on the stream power calculation might be somewhat overstated.

C) Another way to have a larger contrast in stream power between the reversed and beheaded channels might be to have increased infiltration and transient storage of precipitation in the valley alluvium on the low-sloped, eastward flowing channels relative to their west-flowing counterparts.  I don't know much about the flashiness of the Negev desert, but some alluviated valleys in California produce surface runoff only in very large storms because of the effect of infiltration into the low-sloped valleys.  This might stand in contrast to the steeper-sloped segments that form as the drainage divide migrates.  Perhaps this is worth some discussion as well?

Thanks very much for the opportunity to be part of this work.  The field site is a real gem, which has been exploited well by the proponents.  I am support of its acceptance to ESurf, pending consideration of some of the issues I mentioned above.

George Hilley.

This manuscript explores the relationships among drainage area, slope, and valley width in three types of drainage basins in various states of basin reorganization: undisturbed basins, beheaded basins, and reversed basins. The authors find that the values of the exponent on drainage area distinguish among the three basin types, and a significant, negative exponent on slope is present in reversed basins. I thought this manuscript was well written and nicely organized. I think this manuscript needs minor revisions prior to publication. The main issue I think the authors need to address is increasing the impact or value of the findings from this study, largely in the discussion section. Below I describe several sections in the manuscript that I think should be expanded and/or discussed in more detail. At the end of this review, I have some minor line-by-line comments.

The authors find unique values of the drainage area exponent b in the three valley types and mention that the different valley types could be characterized by the value of the exponent b (lines 319 – 323 and 373 - 374). But the authors later point out (lines 407 - 411) that there are several other reasons one might find distinct values of b, even in neighboring basins (e.g. Schanz, et al., 2016). What distinguishes the characteristic values for the b exponent in these reorganized basins from different b values due to differences in lithology (Schanz et al., 2016; Langston and Temme, 2019; Brocard and van der Beek, 2006)? What I would like to see here is a more thorough description of when or under what conditions reorganized drainage basins can be identified by the value of the b exponent. This is mentioned but needs more discussion. Further discussion of when variation in the b exponent may point to drainage basin reorganization will help to guide future readers to using the results presented here properly and increase the impact of the paper.

Also related to impact of the paper, would it be worthwhile for future researchers to extract valley width, drainage areas, and slopes from basins with an unknown reorganization history to attempt to identify new locations with past drainage basin reorganization? Perhaps not given the difficulty of extracting valley bottom width; therefore, it seems like the key takeaways of this paper are the differences in unit stream power across drainage divides and evidence for different time scales of channel and valley width adjustment.

Figure 6 that shows increasing channel width going downstream but decreasing valley width going downstream is especially interesting. Does this pattern exist in the other reversed drainage basins? Can the authors give us any idea on the timescales of adjustment for channel width vs. valley width? The authors start to discuss this on lines 458 – 461, but don’t say anything more on timescales besides “longer” vs. short timescales for valleys and channels respectively. I understand the authors might not have the data to give a hard number for the timescale of adjustment, but a more in depth discussion of what factors play into the timescale of adjustment would also be helpful.

I would also like to see a discussion of how the hyper-arid climate plays into these findings. How would these findings be different in a humid climate? Discussion on this point would add to the impact of the manuscript and findings.

Line by Line comments:

Line 50 – 51: is this section discussing channel width or valley width? I found that in several places in this manuscript, I was not sure which one the authors were referring to. See also lines 120 – 121; 471

Line 179 – 181: Nice definition of valley vs. channel, thank you.                        

Line 232 – 234 and Fig. 6a: outlining a polygon representing the valley bottom can be difficult. In figure 6a, the area inside the polygon doesn’t look like a valley bottom to me, with many tributaries carved into the defined valley bottom and no visible break in slope that distinguishes the valley bottom from the sides of the valley. Perhaps this polygon was created based on a slope map of the area, with a threshold slope dividing valley bottom and valley wall. But I can’t see this in figure 6a. Would be helpful to see a figure that shows how this valley bottom polygon was drawn.

Figure 3: Is Figure 3d showing data from valley 10 and the fit from valley 11? Or is this a typo? Additionally, this figure only shows W-A plots for three of the 12 valleys in this study. I suggest adding similar figures of all valleys to the manuscript or in the supplemental material. This could be 1 figure with 3 panels showing all data for normal, beheaded, and reversed valleys.

Line 322 – 323: valley width decreases going downstream. Does this necessarily mean a reversed valley? What else is needed to make that determination?

Line 324 – 325: Clarify that slope exponent values are non-unique only for normal and beheaded valleys. This is not entirely clear until readers reach lines 332 – 334.

Figure 5: In my opinion, not a great figure. Can't see what the values of b and c are. Can only see that they are between 0 and 1 and either close to or far from 0 or 1.

Line 388: Clarify. This makes me think that the authors are talking about channel slope, not slope of the W vs. Area line.

Line 404 – 405: yes, scaling differences between adjacent drainages is potentially evidence for drainage basin reorganization, but under what conditions? Scaling differences can also be lithology dependent could represent changes in uplift rate across catchments

Line 433; 435 – 436: Does “slope exponents” refer to channel slopes or valley slopes?

Line 448 – 449: A reference to a figure that shows this exception for valley 10 would be helpful.

Line 459 – 461, Figure 6 caption: Channel width increasing downstream while valley width decreases downstream is a cool finding. Do you see this in any other of the reversed valleys? What more can you say about this? What else can we get out of this finding?

Line 493: here's explanation of blue dots on figure 7. This explanation should be in fig 7 caption.

Figure 7: Why show channel bottom polygon in reversed and valley bottom polygon in beheaded? Why calculate three different ways west of wind gap? which one is right/best?

Line 506, fig 7 caption: should be red rhombuses, not blue?

Line 518 – 519: “exploited with caution”. What does that mean? Say more about how this can help identify and categorize drainage basin reorganization.

Line 536 – 537: This is a very interesting thought. I would like to see the authors take a stab at answering these questions, or at least share their thoughts on how climate and lithology might affect timescales of deviant scaling.