Flood risk reduction and flow buffering as ecosystem services &ndash; Part 1: Theory on flow persistence, flashiness and base flow

van Noordwijk, Meine; Tanika, Lisa; Lusiana, Betha

doi:https://doi.org/10.5194/hess-21-2321-2017

Articles | Volume 21, issue 5

https://doi.org/10.5194/hess-21-2321-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/hess-21-2321-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 21, issue 5

Research article

|

05 May 2017

Research article |

| 05 May 2017

Flood risk reduction and flow buffering as ecosystem services – Part 1: Theory on flow persistence, flashiness and base flow

Meine van Noordwijk, Lisa Tanika, and Betha Lusiana

Download

Final revised paper (published on 05 May 2017)
Preprint (discussion started on 19 Jan 2016)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review comments HESS-2015-538', Anonymous Referee #1, 29 Feb 2016
RC2: 'Flood Risk Reduction and Flow Buffering as Ecosystem Services', Anonymous Referee #2, 25 Apr 2016
AC1: 'Author response to reviews and proposal to split the manuscript into two parts', M. van Noordwijk, 24 May 2016

Peer-review completion

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

ED: Reconsider after major revisions (30 May 2016) by Jan Seibert

AR by Meine van Noordwijk on behalf of the Authors (19 Jul 2016) Author's response Manuscript

ED: Referee Nomination & Report Request started (21 Jul 2016) by Jan Seibert

RR by Anonymous Referee #1 (17 Aug 2016)

Suggestions for revision or reasons for rejection

I think the restructuring and responses to the reviewer comments on the previous submission have materially improved the MS relative to the previous one. I have only reviewed the first paper of the two though because I have encountered issues that need to be resolved.
My overall comment is that much of the lengthy introduction on the various possible interpretations of flow persistence (Fp or the slope of the recession curve) adds little value and could even confuse readers. My preference would be reduce the lengthy detail and digressions, and focus on why and how catchments respond temporally to rainfall, what shifts in those responses may mean, and how an understanding of response mechanisms can lead to actions aimed at recovering catchment function.
It seems that the authors have failed to grasp a major comment I had on the previous version. In my opinion this MUST be addressed before this version can be taken any further. I thought my comment was straightforward and easily understood, but it seems the authors have failed to grasp the issue so I will try again. The whole study is predicated on finding a simple index of flow persistence which can be measured over time to detect whether land-use changes are altering the responsiveness of a catchment to rainfall input. Fine and good. An examination of flow hydrographs will show that the flows out of a catchment catchment initially rise after a rainfall event and then decrease again, with the decrease following what is known as a recession curve. The rise is typically more rapid than the recession or fall. With this as the context, I now take their equation (1) which is as follows:
Qt =Fp Qt-1 + Qa,t
Assuming that Qa,t = 0 for now, then Qt =FpQt-1 (i.e. the flow at time t is related to the flow at time t-1 by Fp). For this equation to be true for every rise or increase in flow (i.e. where Qt>Qt-1) Fp must be >1. Yet they only deal with values of Fp in the range from 0 to 1. So the Fp they are describing must only calculated on the falling flows. Yet this is not mentioned or described anywhere in the paper even though they explicitly note that Fp is equivalent to a recession constant (line 222).
The authors provide data that show flow responses in a catchment (Figure 2 and assoc text) where the variability of flow (i.e. the magnitude of both the rises and falls) increases, yet there are no values of Fp>1. This needs explanation.
At some points they discuss the term Qa,t as though it were stochastic, at other points it seems that Qa,t is used to account for all flows greater than some level of base flow (implying that Qa,t is ≥0). (BTW My understanding is that if Qa,t represents a proportion of the observed flow, it is not actually stochastic although there may be factors that give it a degree of stochasticity?) If Qa,t represents all non-base-flows, then all the values where Fp would be >1 are used to estimate Qa,t (which is what I infer from lines 243-246)? Did they effectively vary Qa,t to get the results they present in Figure 1? If so, this needs to be explained. I also question then why they do not simply describe this approach as flow separation technique with the aim of quantifying the catchment responsiveness rather than the flow persistence. To my mind, they have not adequately explained these key points, so until I get an understandable explanation I cannot accept the paper.
Some other points:
Lines 119-138 provides a discussion of whether changes in land cover lead to changes in flows. They note that this has been shown in small catchments but not in large catchments. Yet they do not discuss the simple issue of detectability given the accuracy of the flow recording system. The design of most large weirs is simply not suitable for accurately measuring base-flows or relatively small changes in flows, which means that changes have to affect most of the catchment to be detectable and accurately quantifiable.
In line 247-249 they describe Qa,t as being interpretable as effective rainfall. My understanding effective rainfall is that is this is simple a measure of how much rainfall is taken up by the soil. Only a daily time step (which is what they argue is logical to use) much of this water will not appear as flow over 24 hours unless the catchment is already saturated or has some rapid flow response mechanisms (i.e. rapid interflow). So it will not be adequately measured by Qa,t which needs to be made clear.
There are a number of places where they abruptly introduce new symbols and fail to explain them e.g. line 237 QT and Qo are not explained.
There are still several basic typographical errors that would be found by a spell checker and should not be in a submitted MS.

Hide

RR by Anonymous Referee #2 (22 Aug 2016)

ED: Reconsider after major revisions (22 Aug 2016) by Jan Seibert

AR by Meine van Noordwijk on behalf of the Authors (28 Sep 2016) Manuscript

ED: Referee Nomination & Report Request started (02 Oct 2016) by Jan Seibert

RR by David Le Maitre (28 Oct 2016)

ED: Publish subject to revisions (further review by Editor and Referees) (29 Oct 2016) by Jan Seibert

Thanks for your efforts with the revisions, which clarified a number of points. At the same time, reading your manuscript again, I also realized that there are still a number of issues:

1) The runoff values you present in figures 2-4 seem unrealistically high. Could you please comment on these values. As it is not just one figure, I am afraid there is something fundamentally wrong - or I misunderstand something.

2) Your approach basically is similar what people previously have used as flow separation techniques. While I can see your point that you are using these techniques here with a different goal, I still think that it for the sake of scientific clarity is mandatory to clearly link to the previous work and to better describe what is similar/different with your approach.

3) Fp is affected by both rainfall time series and catchment characteristics. It is therefore not clear to me whther Fp really is a good measure of catchment status since Fp also could change without any catchment change if rainfall distributions change.

4) Human impacts can both increase and decrease Fp, again, I am wondering whether Fp really is a good measure of 'watershed health'

5) You claim that Fp is related to flood risks. can this be shown on real data? I am not so sure I would agree. Your approach basically assumes that catchment changes result in linear changes in flood risks, but this is obviously often not the case. For instance, if a medium-size reservoir is build, this would increase Fp, but the largest floods would hardly be influenced (because then the reservoir is filled anyway). I guess, another issue here is that the Fp values are based on the continuous, average catchment behavior, whereas floods are single extreme events.

6) Your approach is not at all applicable to catchments with snow. This should be stated even clearer.

Hide

AR by Meine van Noordwijk on behalf of the Authors (05 Nov 2016) Author's response Manuscript

ED: Publish subject to minor revisions (further review by Editor) (12 Nov 2016) by Jan Seibert

Dear authors,
thanks for these clarifications. However, I am afraid some of the replies left me rather more confused than before. Below I reiterate and further explain my concerns. As these concerns are fundamental to the approach put forward in you manuscript it is crucial to resolve these.

Best regards,
Jan Seibert

1) The runoff values you present in figures 2-4 seem unrealistically high. Could you please comment on these values. As it is not just one figure, I am afraid there is something fundamentally wrong - or I misunderstand something.

You now changed two of the figures, although the catchments remain to be on the very wet side. For one figure you kept the very high runoff (I guess there are not too many catchments with 6000mm/y worldwide). This makes me wonder whether you would suggest that this approach is suitable mainly for very wet catchments and to be less suitable to ‘normal’ or more arid catchments. If so this should be clarified.

2) Ok

3) Fp is affected by both rainfall time series and catchment characteristics. It is therefore not clear to me whther Fp really is a good measure of catchment status since Fp also could change without any catchment change if rainfall distributions change.

In your reply you agree with my concern, but say that FP ..” is reflecting the “health” of the way a watershed interacts with its climate, rather than the land cover as such.”. I am not sure I understand the term ‘health’ here. Basically you are saying a catchment could change from healthy to unhealthy if the climate changes. And it is not really the ‘interacting’ with the climate which changes. The interacting processes could still be the same, just the forcing changes. Let’s assume that you for some reason suddenly have a constant rainfall, and thus runoff if we ignore E here, for the half of the year, where usually no floods occur. This would largely increase Fp (more healthy) but the risk for flooding would stay the same in reality, wouldn’t it?

4) Human impacts can both increase and decrease Fp, again, I am wondering whether Fp really is a good measure of 'watershed health'

Your reply “We try to provide empirical evidence to help readers answer this question for themselves.” honestly did not convince me. I am still struggling to understand why Fp should be a good measure of catchment health, if human activities can shift it in both directions and this measure is heavily influenced by rainfall patterns. Leaving it to the reader to answer this fundamental question, does not seem to be appropriate.

5) You claim that Fp is related to flood risks. can this be shown on real data? I am not so sure I would agree. Your approach basically assumes that catchment changes result in linear changes in flood risks, but this is obviously often not the case. For instance, if a medium-size reservoir is build, this would increase Fp, but the largest floods would hardly be influenced (because then the reservoir is filled anyway). I guess, another issue here is that the Fp values are based on the continuous, average catchment behavior, whereas floods are single extreme events.

I am afraid I am not convinced by your reply. Here I want to emphasize on the second part of my question: the way you define and compute Fp means that this value reflects the average behavior of a catchment over long periods. In other words, Fp is also influenced by low flow periods, which should not influence the likelyhood for floods.

6) ok

Hide

AR by Meine van Noordwijk on behalf of the Authors (13 Nov 2016) Author's response Manuscript

ED: Publish subject to revisions (further review by Editor and Referees) (14 Nov 2016) by Jan Seibert

AR by Meine van Noordwijk on behalf of the Authors (14 Nov 2016) Author's response

ED: Referee Nomination & Report Request started (18 Nov 2016) by Jan Seibert

RR by Anonymous Referee #3 (20 Dec 2016)

Suggestions for revision or reasons for rejection

I find the concept of this paper to be really interesting, but feel that the papers are still poorly organized, figures and tables need better labeling and updating, and that the authors still need to make a better case for what Fp adds to a given analysis (both in the text and through benchmarking in Part II).

Major points are summarized as follows:
• The connection from health, to floods, to Fp I think gets lost amongst both papers. I would almost change all references from health to refer to Fp as an indicator of alteration. The threat of flooding differs based on your location within a catchment (upstream or downstream) as well as many different watershed characteristics, including human impact. Also, as stated in a comment below, floods are healthy, so attributing their increase solely to human impact is perhaps missing some points. I think reframing this as an indicator of alteration could be useful.
• I find it very striking that two large bodies of literature are missing from the paper – the work of Leroy Poff (Natural Flow Regime, etc), and reference to many of the (small) benchmark catchment studies in the US that linked forest harvest to streamflow responses – these would be interesting candidates for further testing of Fp. I understand that you are focusing on floods at the larger scale, but given the connections you draw between forest harvest/recovery and watershed response, it would be remiss to not reference these landmark papers.
• Floods are a natural part of the flow regime – while Part II demonstrates that the deviation from these natural flow regimes through time is really what you are looking for, I think this theoretical approach would be worth stating and referencing upfront throughout Part I. Just like most of the indicators of hydrologic alteration literature, you are interested in a deviation from average.
• I am left to think that the Fp metric may be worthwhile for comparing within a catchment through time, but may not be effective at comparing across catchments, due to their heterogeneity in all of the effects summarized in Figure 1. It would be nice to include some discussion of the possible limitations of Fp.
• From Figure 2 (Part I), I am really left thinking that this indicator is a measure of flashiness
• Part of the utility of this indicator over flashiness could be your ability to partition it between wet and dry seasons and different flow pathways. At first, the description in Part I was lost on me, but I understood it after reading Part 2. I would better organize this description to frame Fp as a flexible indicator that spans the empirical and modeling realms. Currently, this is somewhat described in Part I, but highlighting the different ways it could be used by a Figure, or by organizing the text better would be really useful.
• As a new reviewer coming to this paper with fresh eyes, I found that many of the comments from previous reviewers were not addressed, especially points of ambiguity in the text .e.g, sufficiently long period (line 224), and the wording throughout section 2, which I found difficult to follow. It is unclear if changes were made to Figure 1, despite comments from reviewers to this effect. I also find this figure difficult to dissect.

Minor points:
-I found interpreting the figures in part I of the paper based on their captions alone to be very difficult, especially Figures 2, 3 and 8 (and their formatting). Please revise these captions.
-All figures would benefit from some revising, especially using subscripts for the p in Fp, sizing font and axes text to the same size, labeling x and y axes, better labeling of Figures (Figure 4, Part II especially), and construction of figures (Figure 4, Part II – lines should not be used to connect different values – this implies a continuity, but these are different catchments)
-I found the results/discussion of part I to be haphazard – if data is presented, whether constructed or real, it should be introduced and discussed prior to being included in the discussion. Furthermore, while I like the comparison of Fp to the flashiness index, this was introduced so late in the paper, and not touched on in the methods, and then not truly analyzed. The authors missed an opportunity to give some thought to how this flashiness index compares to Fp – if the flashiness index describes Fp, then why do we need Fp? Does it say the same thing as Fp? Constructive analysis would certainly make the case for Fp here.

•
• Line 45: 30-50% of what?
• Line 48: This is true at large scales, but differences in catchments of similar size would break this relationship down
• Line 52: this is not always true – there is a difference between upstream and downstream floods, and their drivers
• Line 80: this is likely due to the size of the watershed being investigated
• Line 109: there are several historical papers out of the US that were the first to perform the paired catchment study, highlighting the effect of land cover change on streamflow across longer time periods – work at Coweeta, HJ Andrews, Hubbard Brook, and Fernow Forestes would be relevant
• Line 127: I’m not sure I fully agree – what is the alternative? Mechanistic models?
• Line 130 on: also soils
• Line 136: the rational method and curve number approach were developed to do just this
• Line 150: I would recommend including more citations here – there are many papers that have demonstrated this
• Line 310: I think there is an incorrectly placed word on this line!
• Line 469: wording
• Line 491: For Figures 6 and 7 – If the data is used in this paper, you should describe where it came from – I think these implications may be better explored in paper 2, or the data source should be described in paper 1.
• Line 508: This paper made several assumptions that also should be acknowledged – doing hydrology backward only works well under certain cases
• I find Figure 1 to be too busy to follow! Consider reducing color, size, changing font, and adding arrows in such a way to better show the “flow”

Part II:
• Given that you wish to relate health to Fp, and its change through time to land cover, it seems as though Part II would almost benefit more with a comparison to flashiness indicators. If flashiness tells the same story, then what is the value of Fp? I still think the argument could be made that it allows some simple process-based analysis. The contribution of Fp in Part II needs to be more clear, and should be benchmarked against another indicator of hydrologic alteration, to show the clear value of using Fp over other indicators.

• Lines 960 – 64 – these are relatively relaxed targets, and may miss peaks. Given that the emphasis of this paper is on high flows, you should include a hydrograph of your model to demonstrate that the model adequately matches high flows. As this would affect your Fp values, it may be why you end up with wide scatter in Figure 3.
• Lines 1011 on: please use words for scenarios instead of abbreviations
• Line 1097: wording is confusing
• Table 5, please use more descriptive titles or cite your abbreviations – I cannot tell what these titles mean

Hide

ED: Reconsider after major revisions (further review by Editor and Referees) (20 Dec 2016) by Jan Seibert

AR by Meine van Noordwijk on behalf of the Authors (31 Jan 2017) Author's response Manuscript

ED: Referee Nomination & Report Request started (31 Jan 2017) by Jan Seibert

RR by Anonymous Referee #3 (01 Mar 2017)

ED: Publish subject to minor revisions (further review by Editor) (14 Mar 2017) by Jan Seibert

AR by Meine van Noordwijk on behalf of the Authors (24 Mar 2017) Author's response Manuscript

ED: Publish as is (04 Apr 2017) by Jan Seibert

AR by Meine van Noordwijk on behalf of the Authors (07 Apr 2017) Author's response

Short summary

Deforestation is commonly understood to increase and reforestation to reduce flood risk, but scientific evidence at the relevant landscape scale is scarce and contested. A measure of day-to-day flow persistence is defined here. It is proposed as simple performance indicator for watershed health that can respond to changes in climate and land cover quality, quantity and spatial pattern. Data for four watersheds show decrease or increase in flow persistence in degradation and restoration phases.