Evaluation of the First Negative Ion-Based Cloud Seeding and Rain Enhancement Trial in China

Round 1

Reviewer 1 Report

The paper attempts to estimate the difference in precipitation due to cloud seeding for two regions in China.  They attempt this with these methods:

• Predicting precipitation from historical records for  one (or 3?) weather near one of the sites
• Comparing 2020 precipitation to historical averages for both sites (to one weather station)
• Investigating differences in precipitation for multiple sites in the region over 1 season. From multiple purpose built stations.

Overall, I am not convinced by the arguments in the paper.  The paper’s own statistics generally indicating you can’t accept their conclusions aside,  I have multiple concerns on the test methodologies

I don’t think Section 2 faithfully describes the actual experimental design, there is so little information regarding the historical stations, that I can’t evaluate section 4.1, and even so, I don’t think their methodology for test and validation data (3.1) makes sense and the model clearly doesn’t work for the time frame that interests them.

Section 4.2 seems to include control data for Wushalong for times when the seeding was going on, and again, there is no information about the stations so I can’t determine if this is a causal effect, and using two different seasons for comparison makes no sense w/out 

In addition, wind direction which is referenced as so important has no bibliographic references to the previous studies cited. 

Sections 3.3 indicates a randomized testing with these new stations, but only a limited selection for control was performed in Wushalong, and none in Liupan.

Section 4.4 could only be interested in predicting a total effect of increased rainfall, and other studies have better methods of this type of interpolation, so Figure 8 is not too believable.  Also, why are you using lat-lon in these images? You need an equal area projection.

I expected sections 4.5  and 4.6  to be interesting investigations of the randomized sampling of the regions.  (Ala the descriptions around lines 220-240) but there was no statistical comparisons of these components.

Section 4.7 Could basically show that 2020 was going to be wetter.

More Detailed Notes

 English pretty good until about line 80, then it goes downhill.  

Line 88, not a hypothesis

Figure 1 not required

Add to Table 1:

• Date of installations
• Times of operation / in-operation
• Date(s) precipitation gages added, range of precip data

Line 136 says Wushaoling precipitation has increased, because of cloud seeding? When did operation start? Doesn’t this affect all your met station based comparisons? 

Line 169 - 171 must be missing a bunch of text.

Figures 2 and 3. 

• The legends need to be in a box w/ a white background
• Scale required
• Location of the long Term Weather Station
• Prevailing winds

In addition there should be an overview map.

Line 180: Need a bibliographic reference to wind direction, OR a much better description.  Was is surface only? Weather ballons? Cloud direction mapping? 

Line 190 Where are these stations?  You describe (Line 235-240) The need to know wind direction, and the distance of the effect.  Where are these stations wr.t to these questions?

Line 267: Very confused on AI training.  Presumably you are training on daily data, why aren’t you just pulling a random set of days for validation? Why focus on one year? 

On Line(s) 289, you say you performed a randomized comparative experiment, w/ random days for both sites, but then

355: Need the exact parameters included, not ‘such as’

Lines: 355-364: Again you say, you use three weather stations.  Where are they? Why only precipitation from Longgd and Guyuen? What is the difference of using these three stations in your predictions?   Why not just randomize the days you use? 

Table 4: The bias in your prediction is the worst in the very timeframe you are comparing? Why use this prediction then? Why not predict precipitation for just that time frame?  Later you just subtract this bias from your measured values, This is not right.  

Table 5 and 6, the difference should be Real-Predicted so  a positive % means more rainfall.

Table 5 and 6 are only really helpful if you can show that 2020 wasn’t just a more rainy year.  You have three stations, presumably, some of them are not in the ION shadow.  Can you plot the correlations of rainfall from station to station for these years?  A 12% increase in precip over an average is not too meaningful.

Line 380.  Just make a better model, don’t assume a bias where you are trying to measure.

Table 6, If you want to find a seasonal correlation of rainfall (that is that a high in Jan-Jul implies higher in Aug->November) then you should just be able to show that correlation for the previous years.  Otherwise, you can’t assume that means anything (not natural as stated in Line 391)

4.2.1

On line 421 you state you compare 2008-2019 w/ 2020 For Wushalong, but on line 136, you state precipitation has increased in recent years due to the ionic cloud seeding. 

Line 431: Unless you can show historically that dry yearly beginnings statistically predict dry late years, these numbers don’t mean anything.  Where is the justification? 

4.3.2: Not even worth reporting.

4.4 Spatial Interpolation:

Better examples of these interpolations,  https://popups.uliege.be/1780-4507/index.php?id=10003

 https://doi.org/10.1080/02693799508902045

FIgures 6,7,8,9 should all be in an equal area projection.

4.5 - This section suffers from the small number of control days actually performed in the study, and doesn’t use them really at all to try and add wind direction, ions, and cloud direction as variables of study.

Lines 567-572 indicate that you didn’t use the randomized off dates as control, I don’t understand this?, then in Figures 11,12,13,14 you compare these randomized dates.  There is no reason to predict the 2020 precip, you could just do the individual station by station comparisons from control and experimental days.  

Lines 580, you just remove items that would decrease the predicted results.

Lines 591 indicate that 2020 was a wet year, making us question the results in the previous sections.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This is an interesting paper describing full-scale experimental work on cloud seeding for increasing precipitation. There is no discussion of any potential problems outside the project area if rainfall is displaced by seeding activities, but that is not what this is about. I see two ways (other than a mention of the possible downsides) that it could be improved, however. First, there is very little information about the negative ion generating devices and how they differ in terms of input of energy and output of negatively charged ions so that the between site comparisons can be better understood. Second the use of the t-test is not well explained or reported. No discussion of degrees of freedom or p value. Minor editing of English could be helpful.                              

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I think the second draft is better, but didn't address all my concerns, and I still had trouble following the paper becuse of the experimental description,

I also didn't think some results were analyzed the best way.

Since I made example picutres, I added my comments to the doc below.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

OK