Journal: Hydrology and Earth System Sciences (HESS)

Manuscript ID: hess-2022-315

Article Type: Research Paper

Summary

Increased Nonstationarity of Stormflow Threshold Behaviors in a Forested Watershed Due to Abrupt Earthquake Disturbance assessed changes in hydrologic response of a forested experimental watershed in the eastern Tibet Plateau following an earthquake. The authors characterized longer-term changes in threshold behavior in the watershed and introduced a new metric to quantitatively express thresholds for watersheds with areas of disparate land use, ecology, and physiography. The authors found that lower threshold values were observed in disturbed landslide regions and that non-stationarity in thresholds was mainly controlled by changes to the dominant runoff generation mechanisms of subsurface stormflow and the variable source area.

Significance

This work is significant in several ways:

• It contributes to our growing understanding of threshold-mediated hydrologic response.
• It contributes to the further advancement of a unified threshold-based hydrologic theory.
• It assesses longer-term trends in threshold behavior following an environmental disturbance.
• It introduced a new metric to quantify and compare thresholds.

General Suggestions

I found the abstract difficult to digest. Multiple results are communicated, but there is little context for the reader, making it difficult to understand the methodology or the jargon used in the abstract. Consider revising the abstract to be more general to start and highlighting only key results.

Starting at the end of Line 50 the authors suggest that most threshold behavior in rainfall-runoff relationships reported in the literature has been of the hockey-stick diagnostic shape. I think it is notable that most of the listed studies had an identification procedure only compatible with this shape of a threshold. Otherwise, the wording is somewhat ambiguous and may lead to readers assuming that the dominance of the hockey-stick shape is process-driven or a reflection of some common element in watershed behavior.

In L58-59 I see that the authors have referenced Wei et al. (2020) and the proposed three-linear hydrologic behaviors. I find this wording hard to follow, which I also address in comments about the abstract. I think it might make more sense to describe this form of rainfall-runoff relationship as having multiple inflections/thresholds with intervening linear segments.

The paragraph spanning L62-89 was very clear and informative. It contrasted with the writing style of earlier paragraphs. I hope that a revised version of the manuscript more broadly applies the tone and writing quality of this section.

L109-112: I am unsure if this information is a study area description or is an early interpretation of results. Perhaps, it is just the wording, particularly “let to an unstable trend of the disturbance-response-recovery trajectory….” that is confusing me.

In L123-125, the authors mention the disturbance recovery process of vegetation and how analyzing this might help better understand runoff generation. I think that this information is critical, and a more detailed process-based description of these relationships would be a welcome addition to the introduction.

In section 2.4, I was hoping for more details rationalizing the proposed integrated watershed average index for the thresholds. In the discussion, I think that a section should be added to further elaborate on the efficacy of this metric and some introspection about how this metric may or may not be well suited for other environments/conditions where the control factors on the threshold behaviors differ.

In the discussion, I think some attention should be given to uncertainty in the actual threshold values. I understand that the use of PRA in this context is to characterize the relationship shape rather than to be used in prediction. Still, how robust was the PRA, and are there any concerns about the unequal distribution of events and leverage from particularly large events?

In Section 3.2, especially in later parts, interpretation and discussion begin to creep in a bit.

I think that a stronger definition of threshold is needed to maintain clarity throughout the manuscript. On the first introduction of the tipping points, I also feel that a clear distinction should be made so that the reader can more readily determine that different patterns are being assessed.

L237-230: The authors describe bedrock depression storage and soil moisture deficit as the main factors controlling a runoff initiation threshold. How do the environments of the referenced studies compare to that of the area in the current study? Are there common characteristics that make this process-based interpretation transferable to this stud environment?

Figures 6 and 7: I like figures 6 and 7! They were a nice conceptual addition to the manuscript.

For Section 4.1, controls on threshold behaviors, I found that the author's rationalization of the controls was detailed. With that said, it did read as a mere explanation of different runoff generation mechanisms, and I found there to be a lack of synthesis connecting the experimental observations and analysis results to these more processed-based interpretations. It would be nice if the authors could add some checkpoints in the theoretical explanations to better articulate how their interpretations are supported by their data and how these observations differ from or parallel other studies.

For Section 4.3 point 3. I think that this is an interesting recommendation. Can the authors provide an example of how this could be done? It is a little ambiguous, but I think that this could be a potentially appealing avenue for future work.

Specific Suggestions

Abstract L12: Consider “former hydrologic functioning” rather than “original hydrologic functioning”.

Abstract L16: I am confused by “three-linear stormflow threshold behaviors are examined”, as graphical representations of threshold behaviors are nonlinear.  Also, the following segment refers to “both thresholds”, which I also find confusing.

L38-40: This sentence was confusing to me. I interpret these thresholds as emergent patterns or hydrologic signatures that are an integrated representation of processes spanning spatiotemporal scales. If I have correctly interpreted what the authors were aiming for, I do not believe that this is conveyed in their writing.

L40-44: I understand the intent of this sentence, but I found the wording unusual. Consider revising for clarity.

L44-46: Ambiguous wording. I suggest providing a concise definition of the threshold behavior in runoff response being referred to. The Ali et al., 2013 reference provided, offers one such definition.

L46-48: It is unclear to me why threshold is plural in this sentence – I also think that this information can be incorporated into the former sentence where I have suggested clearly defining the author’s operational definition of threshold behavior.

L50: “They might indicate…” is vague. Are the authors referring to the different diagnostic shapes or the transition from pre-threshold to post-threshold behavior?

L50: “The runoff behaviors….”. Are the authors referring to the thresholds in the cited literature or in the current study?

L78-L79: I found this hard to follow and after reading it multiple times am not sure about the intended messaging.

L119-L120: Please clarify.          

L129: why is volumetric soil moisture content italicized?

L129-131: this sentence is unclear/ hard to follow.

L132-134: tense changes from rest of paragraph.

L139-140: please include in-text which method of baseflow-stormflow separation was used.

L193: “non-stationary” rather than “non-stationarity”.

L203-205: I found the first half of this sentence difficult to understand/follow.

L271: “severally” rather than “several”?

L274-277: I am not sure what the authors are saying in this sentence.

L305: “from2011” missing space.

L306: “It triggered…” what are the authors referring to as “it” in this context?

L310: I do not understand the messaging of the Section 4.3 title.

L320: The spatial patchiness of which characteristics?

Review of “Increased Nonstationarity of Stormflow Threshold Behaviors in a

Forested Watershed Due to Abrupt Earthquake Disturbance”

Author(s): Zhang et al.

MS No.: hess-2021-228

1. Band, University of Virginia

General comment:

Zhang et al. present an interesting study of stormflow runoff threshold non-stationarity over a time-line before and following a major earthquake in the eastern periphery of the Tibetan Plateau.  The earthquake resulted in a massive disturbance of the dominant forest cover due to extensive landsliding which subsequently expanded with monsoon-initiated landslide growth, then slowly began to recover with revegetation, and presumably, renewed colluvial infilling of scars. 

The paper provides a good illustration of specific controls of non-stationary threshold behavior in response to geomorphic disturbance and a chronology of ecosystem recovery.  This adds to our knowledge of storm event-based threshold behavior with good evidence of the watershed system dynamics and time scales of adjustment.  It may be argued that this is an end-member in terms of magnitude of disturbance, but may be increasingly applicable to other cases of large, sudden land use change and slow recovery due to devastating storms, fires, or other disasters.

Specific comments (scientific questions/issues):

The documentation of the threshold stormflow behavior is interesting, but there are a set of areas in the text that are unclear.  Specifically, the methods need to be clarified.  Otherwise, some of the interpretation and conclusions may appear to be more qualitative and speculative, and not specifically supported by the data.  

Figure 3 is a major result and contributes prominently to the conclusions.   However, there do not appear to be sufficient observations to separate out the highest thresholds and trend with statistical significance as it appears this is determined by a single, large event.  It is also not clear from the methods whether discharge was separately measured or determined for grass shrub, forest, and landslide areas. The position of the gauges suggests each drainage area is a mixture of all three land covers, and more information is required to see how each land cover contribution is deconvolved.  Add more detail to this discussion. If separate measurements were not made it is not clear how these piecewise regressions were made.  If this is done by modeling using curve numbers or HEC-HMS this should be clear. 

Finally Figure 3 is difficult to interpret as all data points have the same symbol and color, over all land uses.  Either color code or use a different symbol so the reader can assess the degree of separation between the trends.  Clarifying the statistics provided would also help. There is a composite r² provided in table 2 for each of three distinct land uses, including two thresholds and three slopes.  While the overallthe correlation is significant What is the confidence in each of these parameters?  Is it possible to provide SEE for each?  I presume this may not be possible for the highest flow slope values if they were support by a single large storm observation. 

The authors cite the scarcity of measurements pre-earthquake, and the logistical difficulty of accessing areas post-earthquake as limiting the information available to assess stormflow threshold behavior through this time.  Some information is derived from simulation modeling, developed in a previous paper.  More information on the number of actual measurements, the information provided by the HEC-HMS model, and its reliability should be provided.  The authors point to a specific “tipping point,” after which the stormflow thresholds begin to increase again.  Are these based on land cover change derived curve numbers within the model, and are there discharge measurements sufficient to verify these changes?  In figure 4b we see peak discharge for a set of events first increase and then decrease as the forest ecosystem begins to recover.  How are these peak discharges adjusted for the size of the storm, or are they averaged from a larger number of events?

The analysis of threshold behavior shown in figure 5 is presented in the discussion section.  I think this should be in the results section, then discussed/interpreted in the discussion section.

Technical corrections, clarifications:

1. Reword so it is clear that it was estimates there were roughly 2 x 10⁵ landslides initiated (which is amazing)
2. Line 107, the term indeciduous is not clear. Remove and simply call the canopy conifer.
3. Line 168-170, sentence may be better placed either in the introduction or discussion. It is not a specific result of the analysis done here.
4. Line 261 – I presume you mean the time to peak decreased not increased following the earthquake?