Causal effects of dams and land cover changes  on flood changes in mainland China

Yang, Wencong; Yang, Hanbo; Yang, Dawen; Hou, Aizhong

doi:https://doi.org/10.5194/hess-25-2705-2021

Articles | Volume 25, issue 5

https://doi.org/10.5194/hess-25-2705-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-25-2705-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 25, issue 5

Research article

|

20 May 2021

Research article |

| 20 May 2021

Causal effects of dams and land cover changes on flood changes in mainland China

Wencong Yang, Hanbo Yang, Dawen Yang, and Aizhong Hou

Download

Final revised paper (published on 20 May 2021)
Preprint (discussion started on 26 Nov 2020)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Review of "Causal effects of dams and land cover changes on flood changes in mainland China"', Anonymous Referee #1, 13 Jan 2021
- AC1: 'Reply on RC1', Wencong Yang, 16 Feb 2021
RC2: 'Review', Marc F. Muller, 21 Jan 2021
- AC2: 'Reply on RC2', Wencong Yang, 16 Feb 2021

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (further review by editor and referees) (23 Feb 2021) by Louise Slater

AR by Wencong Yang on behalf of the Authors (05 Mar 2021) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (09 Mar 2021) by Louise Slater

RR by Marc F. Muller (06 Apr 2021)

RR by Anonymous Referee #1 (07 Apr 2021)

Suggestions for revision or reasons for rejection

The authors have made many improvements to the paper and have addressed most of my concerns. My remaining concern relates to the amount of underlying variability that is used to estimate the panel regression coefficients. The authors should communicate the underlying variability (or lack of) more clearly, so that the reader is able to properly evaluate how the results may or may not apply to other regions or catchments. Below, I have explained my concern in detail and how I believe it could be addressed in a straightforward manner.

Panel regressions estimate the effect of a within-catchment change by taking advantage of panel datasets, which have variation in both space and time. This requires that there are temporal variations in urban area, RI, and crop area within catchments. As I suspected, and as shown in Table 1, a relatively small portion of catchments experience changes in RI over 1992-2017. For example, it appears that less than half of 207 watersheds (or <103 catchments) had changes in RI. Thus, the estimated effect of RI on Q will be almost entirely based on the effects within this small subset of catchments. My concern is that this subset of catchments may not be a representative subset of the overall 1,625 catchments (in terms of size, geographic distribution, topography, climate, etc).

Similarly, from Table 1, it appears that more than 75% of the 757 catchments had changes in urban area of less than 1%, and more than 50% of catchments had changes in crop area of less than +/– 1%. The few watersheds that had larger changes in urban area or crop area would have a larger influence on the estimated coefficients. Overall, this adds uncertainty for drawing conclusions about effects of RI, crop area, and urban area across the full region, if the regression coefficients are heavily influenced by a few catchments.

I agree with the authors’ choice to use bootstrapping to estimate regression uncertainty. I think there are two additional aspects needed to properly convey the underlying variability and the implications for the analysis:

1. Clarify the number of catchments with variability in each of the causal factors during the 1992-2017 period, as well as the distribution and characteristics of these catchments compared to the overall region. This could be accomplished with an additional table (or a revised version of Table 1), as well as some clarifications in the text. Fig. 3 is helpful, and it would also be helpful to see the spatial distribution of ∆RI, ∆%Urban, and ∆%Crop. If there are large differences between the subset of catchments and the full set of 1,625 (or 2,739) catchments, the authors should add a caveat about applying results from the subset to the full set.

2. In the case of RI (where different time periods are used to estimate the regression vs. cumulative effects), the authors should compare the amount of within-catchment variability during the period used for fitting the regression (1992-2017) to the RI changes during the period used to estimate cumulative causal effects (1960-2017). This could be accomplished in a table, or through clarification in the text. It is possible that changes in RI over this longer period are much larger than over the shorter period, and thus there may be higher uncertainty when extrapolating to these larger changes. If this is the case, I think it deserves a statement or caveat in the text.

Additional Comments:

L164: “1. There are no other important time-varying sub-regional variables that significantly affect both human factors and floods”. I think the assumption would more accurately be stated as: There are no other time-varying sub-regional variables that are correlated with human factors and affect floods.

Figure 8: I think this analysis should be restricted to the common period when both dam and land type data are available (1991-2017). Otherwise, potential effects of ∆%Urban in the early part of the period are unknown. Additionally, I am confused by the authors’ choice to only examine basins where RI=0 and %Urban=0 at the beginning of the time period. Any catchments that have no change in RI and %Urban would be free of effects from these factors over the time period used to calculate trends in Q (regardless of initial RI or %Urban values). I don’t think that the current analysis is wrong, per se, but a more comprehensive version of Fig. 8 would include all catchments where ∆RI and ∆%Urban yield expected changes in Q of less than 10% (regardless of initial values), and is something the authors should consider.

Hide

ED: Publish subject to minor revisions (review by editor) (08 Apr 2021) by Louise Slater

AR by Wencong Yang on behalf of the Authors (14 Apr 2021) Author's response Author's tracked changes Manuscript

ED: Publish as is (16 Apr 2021) by Louise Slater

Short summary

This study quantified the causal effects of land cover changes and dams on the changes in annual maximum discharges (Q) in 757 catchments of China using panel regressions. We found that a 1 % point increase in urban areas causes a 3.9 % increase in Q, and a 1 unit increase in reservoir index causes a 21.4 % decrease in Q for catchments with no dam before. This study takes the first step to explain the human-caused flood changes on a national scale in China.