Annual flood damage influenced by El Niño  in the Kan River basin, Iran

Hooshyaripor, Farhad; Faraji-Ashkavar, Sanaz; Koohyian, Farshad; Tang, Qiuhong; Noori, Roohollah

doi:https://doi.org/10.5194/nhess-20-2739-2020

Articles | Volume 20, issue 10

https://doi.org/10.5194/nhess-20-2739-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/nhess-20-2739-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 20, issue 10

Research article

|

17 Oct 2020

Research article |

| 17 Oct 2020

Annual flood damage influenced by El Niño in the Kan River basin, Iran

Farhad Hooshyaripor, Sanaz Faraji-Ashkavar, Farshad Koohyian, Qiuhong Tang, and Roohollah Noori

Download

Final revised paper (published on 17 Oct 2020)
Preprint (discussion started on 01 Aug 2019)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review nhess-2019-166', Anonymous Referee #1, 23 Aug 2019
- AC1: 'Response Letter to Reviewer #1', Roohollah Noori, 13 Dec 2019
RC2: 'Referee comments', Anonymous Referee #2, 26 Sep 2019
- AC4: 'Response Letter to Reviewer #2', Roohollah Noori, 13 Dec 2019
RC3: 'Comments on “Evaluation of the probable annual flood damage influenced by El-Niño in the Kan River Basin, Iran”', Anonymous Referee #3, 01 Oct 2019
- AC3: 'Response Letter to Reviewer #3', Roohollah Noori, 13 Dec 2019
AC2: 'Response Letter to Reviewer #2', Roohollah Noori, 13 Dec 2019

Peer-review completion

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

ED: Reconsider after major revisions (further review by editor and referees) (13 Dec 2019) by Bruno Merz

AR by Roohollah Noori on behalf of the Authors (19 Jan 2020) Author's response

ED: Referee Nomination & Report Request started (14 Feb 2020) by Bruno Merz

RR by Xun Sun (25 Feb 2020)

RR by Anonymous Referee #2 (09 Mar 2020)

Suggestions for revision or reasons for rejection

The paper has much improved. Fig. 5 and 6 in the author response now also quite clearly illustrate why the authors want to construct the connection between El Nino and flood risk.

However, a major criticism remains in that the presented changes in flood risk due to El Nino are entirely arbitrary. Fig. 4 in the paper clearly illustrates that the median % change of precipitation during El Nino years is equal to 0. Taking the 60 and 90% quantiles of this distribution and declaring these as representative for flood risk changes due to El Nino is simply not a valid approach, because it completely ignores the lower tail of the distribution where rainfall during El Nino years is actually decreased. I included some possible options for fixing this below, but the author's may have different ideas. In any case the risk analysis probably needs to be redone, so a major revision will be required.

The above assumes that it is a valid approach to apply annual precipitation change factors directly to extreme rainfall on short time scales. This is another major limitation of the paper which needs to be stated clearly in the discussion and conclusions. The author's quite nicely illustrate this in their reply to reviewer 1 (comment 1) in that average change of monthly precipitation in El Nino years is +36% in January, but -28% in April. The former is likely to be linked to snowfall, while the latter actually suggests a decrease of rainfall during the flooding season.

Main changes required:
-quantification of rainfall changes - I could imagine 2 ways of doing this:
-1. Compute flood risk for 10 or so quantiles of the distribution in Figure 4 (considering positive AND negative changes) and then numerically integrate over the distribution.
-2. Compute flood risk for each of the 9 El Nino years and compare average against non El-Nino years.
-merge Fig. 5 and/or 6 from the author's response into Fig. 3 in the paper. These are much more more illustrative, while the data in Fig.3 currently may or may not support the assumption of a trend.

Minor comments:
-p.4 l.13: all your result figures suggest annual rain depths around 250-300mm, so I'm surprised to see 640mm here?
-Eq.2: it is still unclear how you compute the probabilities. Please specify, possibly in an appendix
-p.8 l.25: This paragraph describes methodology which should be merged into the description of assessment steps (probably step 3)
-the motiviation of the paper is still not quite clear. I would say we are trying to identify the potential variability due to El Nino to be able to separate from other (e.g., climate) effects?
-the paper does require language revisions if accepted

Hide

ED: Reconsider after major revisions (further review by editor and referees) (15 Mar 2020) by Bruno Merz

AR by Roohollah Noori on behalf of the Authors (06 May 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (11 May 2020) by Bruno Merz

RR by Anonymous Referee #2 (24 May 2020)

Suggestions for revision or reasons for rejection

I'm sorry to send this back one more time, but the statistical analysis is still not valid. It becomes very clear when we look at Fig. 3. The current approach in the paper is the following:
1. Take all rainfall values with SOI below -1
2. Compute % change in annual rainfall for each of these years compared to mean annual rainfall for the entire series
3. Fit a distribution to these values and use the 60 and 90% quantiles for evaluating the impact of El Nino.

Step 3 implies that the authors find the "60 and 90% biggest rainfalls" during El Nino years and evaluate how much bigger these are than the average annual rainfall. However, these quantiles are simply an expression for random variation in the data (we can see similar levels of variation in non-El-Nino years. In fact, the 3 years with highest annual precipitation are outside the El Nino season).

The correct question to ask would be:
How much bigger is the mean annual precipitation in El-Nino years compared to the mean annual precipitation in non-El-Nino years?

If have done a quick bootstrap and permutation test on the Mehrabad dataset (R script inserted below). The answer is that when we consider a SOI threshold of -0.8, there is no difference between the mean annual precipitation in El-Nino and non-El-Nino years. Considering a threshold of -1, there is a mean difference in the order of 11% (which is in line with the median difference mentioned in the paper). The statistical evidence for the difference in mean annual precipitation is weak - in the permutation test it is significant at a confidence level of 63%.

I suggest the following:
-Remove any mentioning of confidence levels in the paper (abstract, page 7, discussion, conclusions)
-Consider the 12% median change in annual precipitation as an expression for the change inflicted by El Nino and consider only these results.

In addition to the above, please mention clearly in the abstract and the outlook that
-Annual change factors cannot necessarily be transfered to extreme values.
-In fact, your own data suggest that when considering monthly rainfall the effect of El Nino might negative when considering periods where extreme rainfall occurs. These are badly understood effects that require more research.

##################################################
#R script for bootstrap and permutation test on Mehrabad data
##################################################
#data pairs from Figure 3a (Mehrabad station)
valuepairs=c(1.5,200,1.4,260,1.3,175,1.1,180,1,160,0.9,270,0.8,135,0.75,150,0.75,155,0.6,220,0.5,230,0.45,280,0.45,360,0.4,150,0.35,120,0.35,175,0.35,185,0.34,290,0.33,330,0.25,130,0.23,190,0.23,210,0.2,300,0.19,330,0.18,140,0.17,215,0.16,170,0.1,130,0,240,0,335,-0.05,155,-0.1,215,-0.1,225,-0.1,235,-0.1,280,-0.18,125,-0.19,290,-0.2,200,-0.2,300,-0.26,175,-0.3,280,-0.35,275,-0.4,290,-0.45,180,-0.48,225,-0.5,245,-0.51,245,-0.6,265,-0.6,275,-0.65,400,-0.7,220,-0.71,340,-0.72,160,-0.72,205,-0.73,290,-0.75,275,-0.75,250,-0.79,205,-0.8,180,-0.81,260,-0.85,90,-1.1,150,-1.2,240,-1.25,245,-1.3,280,-1.65,255,-1.7,340)
SOI=valuepairs[seq(1,133,2)]
Pannual=valuepairs[seq(2,134,2)]
plot(SOI,Pannual)

#bootstrapping to compute confidence interval for difference between the two groups
#sample values from x and y and check the difference in mean between the groups (how much are the means affected by removing single values from the distribution)
#https://02402.compute.dtu.dk/enotes/chapter4-SimulationBasedStatistics
for (threshold in c(-0.8,-0.9,-1)){
x=Pannual[SOI>threshold] #rainfall in normal years
y=Pannual[SOI<=threshold] #rainfall in el nino years
k=10000
xsamples = replicate(k, sample (x, replace = TRUE))
ysamples = replicate(k, sample (y, replace = TRUE))
mymeandifs = apply(ysamples, 2,mean)-apply(xsamples, 2, mean)
hist(mymeandifs,xlab='annual precipitation difference [mm]',main='') #histogram of differences between the bootstrapped samples
quantile(mymeandifs,c(0.025,0.5,0.975))
#mean annual rainfall change / mean annual rainfall in non-El-Nino years
print(c(threshold,mean(mymeandifs)/mean(x)))
}

#permutation test - compare the difference in means between x and y against the difference in means when we create two groups that are composed by randomly drawn values from both x and y
#if we perform 100000 replications and in most of these the differences between the randomly composed groups are similar to the difference in mean between x and y, we have no evidence that the mean values of x and y actually are different
for (threshold in c(-0.8,-0.9,-1)){
x=Pannual[SOI>threshold] #rainfall in normal years
y=Pannual[SOI<=threshold] #rainfall in el nino years
k=100000
perms = replicate(k,sample(c(x,y)))
mymeandifs = apply(perms[(length(x)+1):nrow(perms),],2,mean)-apply(perms[1:length(x),],2,mean)
#what ratio of the random samples has a smaller difference between the two sets than the difference between our actual datasets for El-Nino/non-El-Nino years?
print(sum(abs(mymeandifs)<=abs(mean(y)-mean(x)))/k)
}
##################################################

Hide

ED: Reconsider after major revisions (further review by editor and referees) (02 Jun 2020) by Bruno Merz

AR by Roohollah Noori on behalf of the Authors (14 Jul 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (28 Jul 2020) by Bruno Merz

RR by Anonymous Referee #2 (19 Aug 2020)

ED: Publish subject to technical corrections (19 Aug 2020) by Bruno Merz

AR by Roohollah Noori on behalf of the Authors (27 Aug 2020) Author's response Manuscript

Short summary

The effect of El Niño on flood damage was investigated. The methodology was based on the calculation of increasing rainfall amount during El Niño events compared to normal conditions. With the southern oscillation index equal to −1.0 as the threshold of El Niño, the annual percentage of increased rainfall is 12.2 %. The annual change factor may not necessarily be transferred to extreme values. Nonetheless, the change factor was applied for generating simulated storms of different return periods.