Envirotypes Based on Seed Yield Limiting Factors Allow to Tackle G × E Interactions

Round 1

Reviewer 1 Report

Dear Authors

I had the opportunity to review the paper title “Envirotyping highlights temperature and radiation stresses as the main abiotic constraints affecting seed yield in winter rapeseed and allows tackling the G×E interaction”

It is a very nice idea with envirotyping based on the PLS regression results and clustering. The manuscript has a good flow and is good to read / understand. There are some aspects in the manuscript need to be improved (minor revisions required). Only one aspect need major revision - the title of this manuscript should be amended.

Below are detailed comments for the manuscript:

Line 92. “Field experiments were conducted in 9 locations over 6 years”. That could be confusing for a Reader. See Table S1: Châteauroux only one 2013/14 crop season, Dijon 2012/13 and 2014/15, Liverdy only 2015/16., etc.
Line 108. replace “complete random block design” with “randomized complete block design”.
Line 111-112 and 199. The probes genotypes were scored over the 9 environments of the network, am I right?

-----------------------------------

Crop season  DS127   Probes

2010/11         -           2
2011/12         -           1
2012/13         -           2
2013/14         4           1
2014/15         7           -
2015/16         -           3

-----------------------------------

Total             11          9

-----------------------------------

Line 120. Replace “genotypeg” with “genotype g”.
Line 120 – 124. Many of your readers don't know how to interpret the value of ecovalence. This subparagraph is good place for a simple explanation, besides Fig. S1.
Line 134, 149, 152, 160, 232 and 49, 57, 71, 113. In the text, references should be harmonised.
Line 231-232. I’m certain that almost all of your readers don't know what is this “test value as defined by Husson et al”.
Line 234-239. I see many errors in equation 3:
i) replicate k nested in the j-th level of environment should be noted as Rk(j);
ii) lack of observational unit in y, e.g. yijkl;
iii) lack of nested observational unit in residual ε, e.g εl(ijk);
iv) what model it is, fixed, random or mixed?;
Line 240-247. Equation 4 – see above

Line 250. “±0.618” – it is standard deviation or standard error of mean?
Line 253. as above

Line 278-279. Replace “only 25% of the correlation were significant at a risk α of 5%” with “ 25% of the correlation coefficients were significant at α=0.05”
Line 296. The abbreviation BQ should be clarified in chapter 2.5.1.
Line 358. Here, incorrect model 3 is unnecessary.
Line 361-362. Should be “***, p<0.001; **, p<0.01, *, p<0.05; ns, not significant”
Line 364. Here, incorrect model 4 is unnecessary.
Line 368-369. Should be “***, p<0.001; **, p<0.01, *, p<0.05; ns, not significant”
Line 380-382. There is lack the envirotypes 1 and 3
Figure 1a. What indicate of 7 rows on heat maps? And, by the way, these are very nice figures.
Figure 2c. Seed yield was defined here as the mean or median, and why q·ha-1 not t·ha-1? There is lack information about box and whisker.
Figure 4. There is lack the envirotypes 1 and 3
Figure S2. Replace “Allen et al [xx] and Allen et al (1988)” with “Allen et al. [28]”
Figure S3. Seed yield was defined here as the mean or median, and why q·ha-1 not t·ha-1? There is lack information about box and whisker.
Table S1. Le Rheu have negative value of longitude
Table S3. The description under table must be improved: replace “correlation” with “correlation coefficients” and “***, p<0.001; **, p<0.01, *, p<0.05; -, not significant”

Author Response

Line 92. “Field experiments were conducted in 9 locations over 6 years”. That could be confusing for a Reader. See Table S1: Châteauroux only one 2013/14 crop season, Dijon 2012/13 and 2014/15, Liverdy only 2015/16., etc. The sentence was changed line 93 with “20 environments (combination of location and year) between 2011 and 2016” … Line 108. replace “complete random block design” with “randomized complete block design”. Ok Line 111-112 and 199. The probes genotypes were scored over the 9 environments of the network, am I right? No, the probes genotypes were scored over the 20 environments (they are included in DS127). The Table S1 was changed to be less confusing (population column). Line 120. Replace “genotypeg” with “genotype g”. ok Line 120 – 124. Many of your readers don't know how to interpret the value of ecovalence. This subparagraph is good place for a simple explanation, besides Fig. S1. A short note about the interpretation of the ecovalence has been added in the text (lines 121-123) Line 134, 149, 152, 160, 232 and 49, 57, 71, 113. In the text, references should be harmonised. Ok Line 231-232. I’m certain that almost all of your readers don't know what is this “test value as defined by Husson et al”. An explanation of the test value as been added in the text (lines 236-238) Line 234-239. I see many errors in equation 3:
i) replicate k nested in the j-th level of environment should be noted as Rk(j); ok
ii) lack of observational unit in y, e.g. yijkl;
iii) lack of nested observational unit in residual ε, e.g εl(ijk); we had a single observation for each genotype in each environment and in each replicate, therefore we did not think we need any supplementary index to define the residual
iv) what model it is, fixed, random or mixed?; Fixed linear model, term fixed was added in the manuscript Line 240-247. Equation 4 – see above Refer to the previous comment (equation 3) Line 250. “±0.618” – it is standard deviation or standard error of mean? Pearson standard deviation Line 253. as above as above Line 278-279. Replace “only 25% of the correlation were significant at a risk α of 5%” with “ 25% of the correlation coefficients were significant at α=0.05” ok Line 296. The abbreviation BQ should be clarified in chapter 2.5.1. ok Line 358. Here, incorrect model 3 is unnecessary. Deleted Line 361-362. Should be “***, p<0.001; **, p<0.01, *, p<0.05; ns, not significant” ok Line 364. Here, incorrect model 4 is unnecessary. Deleted Line 368-369. Should be “***, p<0.001; **, p<0.01, *, p<0.05; ns, not significant” ok Line 380-382. There is lack the envirotypes 1 and 3 Envirotypes 1 and 3 could not be described by an interaction plot because they are composed only by one DS127 environment. It has been precised in the text (line 388) Figure 1a. What indicate of 7 rows on heat maps? And, by the way, these are very nice figures. A mistake during edition has occurred and heatmap 1A has been corrected in the final version Figure 2c. Seed yield was defined here as the mean or median, and why q·ha-1 not t·ha-1? There is lack information about box and whisker. The mean was considered. We note that q.ha-1 is not the official unit but is often used to express seed yield in crops. In legend of figure 2c, we have added “boxplot of the distribution” Figure 4. There is lack the envirotypes 1 and 3 same as comment above line 380-382 Figure S2. Replace “Allen et al [xx] and Allen et al (1988)” with “Allen et al. [28]” ok Figure S3. Seed yield was defined here as the mean or median, and why q·ha-1 not t·ha-1? There is lack information about box and whisker. Same as figure 2c Table S1. Le Rheu have negative value of longitude latitude and longitude are given in decimal value. The sign of the longitude depends on the Greenwich Meridian, for this reason, Le Rheu has a negative value of longitude Table S3. The description under table must be improved: replace “correlation” with “correlation coefficients” and “***, p<0.001; **, p<0.01, *, p<0.05; -, not significant” Table S3 title has been slightly changed

Reviewer 2 Report

Formula in line 119 (and later) should use the common notation of putting a bar over the Y´s to indicate average (means).

In line 195 there should be a mention of the version of R used here, and the main libraries used. It is very important to to mention how the ANOVAs were done: have they have the function lm or lme?

In lines 357-369 (tables 3 and 4) the sum of squares mentioned are not only coming from the "fixed effects", but also from the random effect(s), i.e., the residual and may be (according to how they run the ANOVAS), the replicate effects.

I am attaching a few handwritten corrections in the documents.

Comments for author File: Comments.pdf

Author Response

Formula in line 119 (and later) should use the common notation of putting a bar over the Y´s to indicate average (means).ok In line 195 there should be a mention of the version of R used here, and the main libraries used. It is very important to to mention how the ANOVAs were done: have they have the function lm or lme? The version of R software was added (line 198) and the main libraries used were already quoted along the text. Line 239, we precise that we used the function lm to run linear models In lines 357-369 (tables 3 and 4) the sum of squares mentioned are not only coming from the "fixed effects", but also from the random effect(s), i.e., the residual and may be (according to how they run the ANOVAS), the replicate effects. We run only fixed linear models, we precise this in the title of tables 3 and 4, for this reason replicate effect was fixed for both models.