Referee report on

"Geometric constraints on tributary fluvial network junction angles"

by

Pelletier et al.

The authors present an analysis of ~10 million junction angles

for which they extracted the corresponding stream network across the

Contiguous United States using a 10m Digital Elevation Model. They

find that the branching angles closely follow the  slope-angle relations

proposed by Horton in 1932 and later modified by

Howard in 1990, however with the twist, that they use the mean basin

slope of the two upstream basins rather than the valley slope of the

tributary channels at the confluence.

While the paper makes some interesting points on how topographic

features and geology may be expressed in a channel network's branching

geometry, several points need to be clarified and some

miss-perceptions of previous studies need to be resolved.

First, the fact that vertical features of topography like slope

correlate with horizontal geometric properties of landscapes such as

branching angles should surprising. Ultimately, Earth's surface

evolves as a result of erosion and uplift and valley networks are

simply characteristic features embedded in this three-dimensional

topography.

More important than correlations between different topographic

features is the understanding of how the underlying processes, namely

erosion, deposition, and tectonic uplift are expressed in the

network's branching structure. Quoting from Howard's 1990 paper on the

grometric model: , "[this geometric model] was only sketchily related

to the assumed channel processes ...".

This point actually prompted Howard in 1990, to explore additional

hydrological arguments to constrain the angle-slope relation foundon

more process-based descriptions of how the flow in a channel may

modify its bed, even if the particular expression may result in a

slightly lower r-square, than the rpedictions of the geometric model.

Also note, that much of the theoretical basis of Howard 1990 was

mostly done 20 years earlier in [Howard 1971b, 1971c] as referenced in

the 1990 paper. Similarly, Hooshyar 2017 and Seybold 2017 try to

relate branching angles to channel-forming processes, namely overland

flow v.s. debris flow in the first and seepage flow v.s. overland flow

in the latter, while the optimality models of Strong & Mudd e.g. use

hydrological conservation arguments at the junction, similar to the

arguments in Howard 1990. I would like to note here, that while the

authors promote the impression that the observed slope-angle relation

contradicts the above-mentioned interpretations, however, these

conceptual frameworks are more complementary as they simply provide

different aspects for explaining how landscapes evolve.

Consequently, the authors may consider a more "inclusive tone" in

their argumentation, particularly because the author's own

explanations remain rather vague on process understanding giving too

much focus on the trivial angle-slope relations of the geometric

model.

An interesting point in the manuscript which I was initially excited

to read about was the alleged finding that incisions into alluvial

piedmonts exhibit different structural controls on branching angles

than channel networks incised into bedrock. Here, the role of geology

is an important point for channel erosion and has not been addressed

adequately in previous work, including the recent papers cited by the

authors.

However,the more I was disappointed when reading through the text to

find out that the narrowing of branching angles in alluvial piedmonts

is fully explained by the fact that piedmonts are (usually)

preferentially sloped while bedrock valley networks do not display a

preferential slope.

Thus, what is thought to be a structural effect is simply the higher

prevalence of regional gradients that narrow the branching angle

statistics. 

This effect of regional slope and microtopography has been already

demonstrated by Casteltort in 2013 through numerical simulations and

also observed in multiple previous studies over the last decade.

Below I provide some minor comments/suggestions on specific parts of

the text, which may repeat some of the points raised above.

L.5: Ignoring some degenerated special cases, three planes, one for

each upstream tributary and one for the downstream channel can

intersect at any angle and slope and thus does not provide a geometric

constraint for branched valley networks. What actually constraints the

argument of Horton /Howard is the presence of fixed a general slope

(for all three basins) aligned with the downstream channel, while the

side valley is additionally inclined with respect to this dominant

direction.

L 9-10: Landscapes evolve through erosion/deposition and uplift

creating sloped valley networks. Thus I firmly object to the

implications that the geometric "model" actually contradicts the

process-based explanation by Hooshyar et al. or Seybold et al. as well

as the optimality principles proposed by Rinaldo, Strong & Mudd, and

many others. Earth's surface is shaped by processes and not geometric

features.

L. 11-12: "Junction angles are consistent with the geometric model..."

It would be good to have a 1:1 plot between prediction vs measurement

and a quantification of the spread and uncertainty of the model

prediction. Also, it would be favorable to see how the "classical

geometric model" compares to the "modified geometric model" in such a

comparison.

L.17-25: How does the type of geology (alluvial vs bedrock) relate to

the sloping of the topography? 

The tilting effect on branching angles and basin shapes observed in

the modeling procedures of Casteltort 2013) which the authors draw to

explain their findings makes absolutely no assumption on the ground

surface material. Neither piedmonts nor bedrock is part of

Casteltort's model. Consequently, I see nothing new in the

interpretation besides the finding that alluvial piedmonts at mountain

fronts often display a general sloping tilt and thus have narrower

branching angles on average. However, this has been known for over 10

years (Casteltort 2013) and also observed in multiple network studies.

The authors boldly use the word "demonstrate" multiple times in the

abstract but hardly prove anything in the main text. It would be

interesting if the authors could at least demonstrate that the

geological impact of alluvial v.s. bedrock is an actual effect and not

a spurious correlation because alluvial piedmonts often follow a

general slope.

L. 27: "demonstrates independent[ce] of climate and optimality

principles": How can the authors ensure that the generation of slopes

in landscape evolution is not the effect of climate-driven

erosion/deposition processes (rainfall leads to discharge and channel

incision as a consequence) which follow optimization rules of

landscape evolution?

While I agree with the arguments that the geometry of valley networks

are sensitive to the initial conditions such as regional tilt and

micro topography, it is hard to separate fluvial incision from the

formation of the initial landscape.

L 35: parallel and sub-parallel networks in piedmonts v.s. dentritic

and rectangular basins incised in bedrock: Is this a geological effect

(alluvium vs bedrock) or simply an effect that the piedmont has a

dominant regional slope?

L. 40: Piedmont deposition: I don't buy the claim that the deposition

cycles created initially unincised low-relief landforms because the

timescales for deposition are rather similar to the timescales of

valley incision. Thus one can hardly separate the two processes as

they occur simultaneously. Here tha authors need to provide some

additional arguments which support their claim.

L.56-58: If this is true, then the narrow branching angles in

piedmonts are not the result of the underlying geology but a spurious

effect of the gently sloping terrain. This effect has been very

clearly demonstrated by Casteltort using landscape evolution models

which are completely independent of the underlying geology.

L 63: The authors fail to demonstrate that tortuosity is indeed the

driver for wider junction angles compared to the arguments of Howard

1990. "may promote narrower junction angles ..." is simply not enough

for a central result in a scientific publication.

L 104: the importance of piedmont deposits: As the authors have

already explained in a previous paragraph, it is not the piedmont

deposit (geology) that narrows the branching angle but the gentle

regional slope that often comes with the depositional landscape at

mountain fronts. Consequently, any sloped geology e.g. volcanic

bedrock in Hawaii would also lead to the same effect.

L. 133: How do the authors define the basin averaged slope in the case

of n-th order junctions with n>1?  Do they follow the longest upstream

tributary up to its source, or do they average the slope of the whole

upstream network tree? This point needs to be clarified.

L 190: Same as above. Do the authors use the pixel-by-pixel slope of the

DEM and then average over all pixels of the upstream basin?   What is

the upslope along directions? These points need to be clarified.

L. 229-231: "... suggests that junction angles may be systematically

lower, on average, in fluvial networks incised into late-Cenozoic

alluvial piedmont deposits than those incised into adjacent areas of

bedrock/older deposits."  The authors need to show that the structural

control on junction angle is not a spurious slope bias in gently

sloped depositional landscapes such as piedmonts.

L. 316: comparison with NHD: Comparing USGS's medium-resolution blue

lines with a fixed flow accumulation threshold extracted channel

network is not a fair comparison, particularly as one can obtain any

drainage density by adjusting the drainage area threshold. NHDPlus (

and its higher resolution companion NHDPlusHR have been extensively

ground-checked. 

For example, I was unable to identify many of the fine channel

networks from Fig. 6b/d around 32.3N/110.9W on aerial images on Google

Maps.

 L. 560: variability in hillslope length: Minor point: While I agree

with the presented arguments it remains to be shown that hillslope

convergence is the dominant factor controlling drainage density.

L. 579: I agree with the AI~elevation argument which should be easy to

test. (correlation elevation~AI and correlation angle~elevation)

L. 595: higher or less infiltration in arid landscapes: I agree with

the authors that the question of groundwater recharge and its

dependency on environmental conditions is rather

controversial. See for example a recent Nature Climate Change which

argues that "Groundwater recharge is sensitive to changing

long-term aridity" which suggests higher recharge in humid landscapes

than in arid ones based on lysimeter measurements.

L. 611: "The result presented here could potentially be reconciled ..."

..." It should be easy to test and prove or disproof if the presented

results are consistent with Hooshyar 2017 or not.

L 615: I agree with the author's similarity arguments between minimum

power and MGM, but while MGM only relates different aspects of Earth's

tomography with each other, the optimal branching models seek to

explain WHY the slopes are as they are from a physical mechanistic

perspective of landscape evolution in general and channel erosion in

particular.  Here, the topographic slope itself is only a feedback

variable for flow accumulation in the erosional term of the landscape

evolution equation.

The authors present an analysis of approximately 10 million junction angles, for which they extracted the corresponding stream network across the Contiguous United States using a 10m Digital Elevation Model, suggesting that the topography of a landscape prior to fluvial incision exerts a key constraint on tributary fluvial network junction angles. This is due to the fact that different slope versus microtopography assemblages give rise to more or less sinuous patterns in the valley bed shape, as suggested by Lazarus and Constantine (2013). Indeed, when computed in the traditional way, junction angles are a function of slope ratios (as the Howard 1990 model predicts), but data deviate from the Howard model in a manner that the authors propose is the result of valley-bottom meandering/tortuosity.

I’ve read the paper with much interest and found it extremely well-written and fascinating. Only sometimes did I find myself a bit lost in the lengthy technical discussions, but this might be entirely my fault, as I am not an expert on the topic discussed here. Also, I appreciate that the literature and scientific debate on junction angles in tributary fluvial networks are extensive, so previous arguments need to be discussed in detail.

Being not an expert, I’ll refrain from delving into the debate regarding what drives the distribution of junction angles and how this study copes with or does not cope with earlier literature. The pros and cons of the approach are discussed in sufficient detail, and results are put into perspective with previous literature in a way that allows even a non-expert reader to navigate through the discussion, although it is very technical.

Still, I’d like to raise two possibly minor comments, one being purely terminological and the other perhaps more fundamental.

First, I’d suggest the authors avoid using the term “meandering” to describe sinuous valley patterns. Meandering refers to landforms that grow orderly rather than randomly. The fact that sinuous patterns can arise even from unordered growth, and that at low sinuosity random versus ordered shapes are not distinguishable (Limaye et al. 2021).

Second, and more importantly, the major limitation I see in the idea behind the paper is that valley sinuosity is assumed to remain fixed once their shape and sinuosity have been determined by the initial microtopography (in the way Lazarus & Constantine's idea illustrates).

Kwang et al. (2021) have demonstrated how lateral erosion/incision of rivers into bedrock is fundamental for the development of dendritic drainage networks, to the point that without lateral erosion, landscape evolution models cannot turn non-optimal, non-dendritic networks into optimal dendritic ones such as those observed in nature.

For the point being made in this paper (that valley tortuosity matters when it comes to junction angles) I think lateral erosion should be considered/discussed because it has the potential to:

• Alter river valley tortuosity over time, thereby impacting junction angles.
• Cause the system to lose memory of initial conditions (inducing persistent reorganization in dynamic steady-state network shape), thus critically diminishing (if not completely erasing) the effect of initial microtopography on which the authors anchor all their analyses and results.

In short, assuming that river networks evolve from initial conditions to a frozen state with no further modifications poses a strong limitations to the new perspective the authroes bring abotu regarding the drivers of junction angles in fluvial tributary networks.  Later network modification by lateral river erosion into valleys, which can induce major drainage reorganization and long transience in landscape shape, should be discussed more thoroughly because it’s potentially tied to how tributary channels intersect each others.

CITED REFERENCES:

Kwang, J. S., Langston, A. L., & Parker, G. (2021). The role of lateral erosion in the evolution of nondendritic drainage networks to dendricity and the persistence of dynamic networks. Proceedings of the National Academy of Sciences of the United States of America, 118(16), 1–6. https://doi.org/10.1073/pnas.2015770118

Lazarus, E. D., & Constantine, J. A. (2013). Generic theory for channel sinuosity. Proceedings of the National Academy of Sciences of the United States of America, 110(21), 8447–8452. https://doi.org/10.1073/pnas.1214074110

Limaye, A. B., Lazarus, E. D., Li, Y., & Schwenk, J. (2021). River sinuosity describes a continuum between randomness and ordered growth. Geology, 49(12), 1506–1510. https://doi.org/10.1130/G49153.1