|General Comments: |
I would like to thank the authors for addressing the key points from my first review of this paper to a high standard. The PPE design now covers the parameter space considered much more evenly, leading to a more robust emulator and hence resulting analysis. This is a novel use of the emulation and sensitivity analysis approach to assess model behaviour under uncertainty. I have a few further minor comments, but once these are addressed I would recommend the publication of the manuscript in ACP.
- Line 139-144: Here, the sections of the paper are introduced, but there is no mention of the sections relating to the seasonal and spatially resolved analysis, which is mentioned in the abstract? Perhaps add a sentence here to also point to these for easy reference?
- Figure 2 and Section 2.2. I like the new Figure 2 that has been added to the manuscript to help with the interpretation on how the effects of the perturbations point to potential for process simplification. However, there is no link or reference to the new Figure 2 in the text of this section, or any text description of this interpretation, so it’s not clear. And, the caption just says ‘Sketch of the envisioned interpretation…’, but interpretation of what? Please add some informative description as to what Figure 2 shows in this section and make the caption clearer as to the interpretation it corresponds to.
- Line 297-298 (and elsewhere: L289, L469): ‘…only 47 PPE members were used as with the 48th member the computational constraint was too tight for the emulator’. This is very ambiguous, and I don’t think a general reader will understand what is meant by this and so will find this statement confusing. If you need to state that there were issues with the emulator construction, then it needs to be phrased more directly as that… something like: ‘… only 47 PPE members were used due to instabilities in the computations when constructing the emulator. ’
I am a little surprised that this happens, with <50 data points over a 4-d parameter space. For my interest - Is it a specific run that causes this? i.e. the same run each time, so some sort of outlier in your PPE?
- Line 327 (and throughout Section 3.1, and in Section 3.3, 3.6): ‘…that it’s inhibition leads to…’ I mentioned this in my previous review, and I still don’t fully understand the meaning of the word ‘inhibition’ when describing the parameter effects. What is a parameters’ inhibition? Do you mean it has very little effect? Or switching it off? – it’s not clear. [When I google the meaning of this word, I don’t find a relevant meaning for this context.] If you must use this word, then please define what it means before you first use it. Or alternatively, re-phrase the sentences to be clear in meaning and take it out.
- Figure 7 caption: The projections of the sampling here are 2-d, not 3-d. And the response surface is a 4-d response surface, not 5-d. The response variable (here, IWP) is not a dimension of the sampling – You only have 4 input parameters that the sampling is over, and so the dimension of the response surface has to be 4-d, as it is showing how the response variable changes over those 4-dimensions of parameter space. Each individual plot here considers 2 of those input dimensions, and therefore it is showing the response over a 2-d projection on the space. Please amend the caption.
- Figure 8 caption: Please remove the term ‘correlation panel’ from this caption – see previous review.
- Line 445 - 446: Why do you need to use the difference to the control simulation for this analysis? I don’t understand what that achieves… Could it affect (reduce) the amount of signal that you see?
- Figures 11 and 12: Following a comment in my previous review – I think it would be more informative to show the first-order effects in these figures and have the total effect figures in the appendix. If the first order effect plots look similar to the total effect plots then they are more informative, as with the total effect, the reader is left to ponder/guess whether some of the effect is in fact interaction, when it’s probably not? The first order effects correspond to individual parameter effects alone, and so are surely more informative and conclusive here?
- Line 481. I’m not sure where the value of 0.2% comes from – please clarify.
- Line 289: Change ‘…As these were only few cases …’ to ‘… As these were only a few cases…’
- Line 326: I know IAV is defined on page 10, but I got to the acronym here and couldn’t remember what it meant… maybe give the full wording here as a reminder? In fact, do you really need to use an acronym for this, given it only appears 3 times?
- Line 372: Change ‘Only for LWP Lohmann and Ferrachat (2010) find…’ to ‘Only for LWP do Lohmann and Ferrachat (2010) find… ’
- Figure 12: The plots in the figure need to be labelled to indicate which is plot a), b), c) and d).