|I am reviewing this manuscript for the second time, and I recognize and appreciate that substantial effort went in this revision to address the reviewer comments on the first version of the manuscript. This has clearly improved the manuscript; in particular, I find the analysis at higher spatial resolution more convincing than the original one.|
There are a few remaining issues that need to be fixed before publication; except for one (see ‘major comments’), they mainly relate to readability, clarity, and logic – I felt that some statements were contradictory. In general, I think manuscript readability would benefit from some shortening, which could be achieved by more precise and concise language and by removing some repetition in the discussion.
Please note that line numbers in my comments refer to the revised manuscript without tracked changes.
1. In L20ff, the authors state that ‘Results show that […] increased the model prediction accuracy by more than 90 %.’, and the same statement reappears at the beginning of the conclusion. I find this statement a bit problematic for two reasons: Firstly, without context, it is not clear what this 90% improvement is compared to, nor where the number comes from. Moreover, it is presented as a major result of the study in the abstract and the conclusion, while there are multiple instances where the authors state that predictive accuracy was not the primary concern of the study. This is contradictory and should be handled more consistently, I believe highlighting the other main findings would be more appropriate
2. L271ff: This paragraph is a bit confusing. The authors first state that data was not split, but then argue that the general procedure IS to split the data and explain the in-bag / out-of-bag concept. It remained unclear to me what exactly was done for this study (i.e. whether data was splitted or not).
L44: ‘high-resolution micro variability’ -> I suggest dropping high-resolution to avoid repetition in the same sentence
L 54: ‘while a shorter (6 m) and longer (20 m) distance in non-vegetated areas are explained’ -> Should be changed to shorter distances (plural)
Section 2.2.4 I appreciate that the site variable is now better explained, but it took me a while to understand that the classification relied on the land cover map only; maybe it would be a good idea to refer back to Fig 1 here.
L234: The ‘DCE’ descriptor from Mazzotti et al. (2019) is also continuous, hence this statement is not strictly correct. I think the novelty is mainly that the edge metric used in this study takes wind direction into account. I suggest reframing this statement.
L290: ‘It is a risk-adjusted alternative to variable dependence.’. this statement (especially ‘risk-adjusted’) is unclear, consider rewording
L305ff: I am not convinced that listing the max. snow depth values at each site is very meaningful since the values all seem outliers at the upper end of the histogram. I would omit this for the sake of making this section more concise.
L420: ‘probably due to the influence of instrumentation’: It is unclear what instrumentation this refers to.
L430 ‘than their forest models’: consider rephrasing to 'than the forest model at the same site' (or ‘than the corresponding forest model'), I believe this would be more precise.
L445: ‘The overall amount of snow in the forest compared to field in the boreal forest of Montmorency could thus be underestimated due to poor lidar coverage under dense canopies’: It is unclear why a lower coverage implies an underestimation of snow depth
465: ‘or larger-scale effects': It is unclear what this refers to
472: ‘Shorter scale break distances in forested areas compared to open field areas (Fig. 5) is analogous to previous studies that studied the fractal distribution of snow depths with lidar data.’ : This statement doesn’t seem supported by the references listed in the following sentences, please revisit.
L514ff: For modelling, this is only true if the goal of the modelling is to fully resolve the variability, but not strictly necessary – a sub-grid parametrization is also an option
530ff: I suggest removing this introductory sentence to avoid redundancy
L581: ‘microtopography has a more restricted influence on deeper snowpack at this site compared to the shallower snowpack at the agro-forested sites’; vs L591: ‘Our findings however show that the microtopography, even under wind-sheltered conditions in the forest, still plays an important part of the spatial variability.’ – Note that while not contradictory, contrasting statements like these confound the key message of the paragraph. I would strongly advise to streamline such discussion sections.
Chapter 4.3.3 somewhat disrupts the story line and refers to results that are presented in the supplementary material only. I would recommend moving the entire section to the supplementary material.
Chapter 4.4.3: the table comparing RF and MLR performance seems better suited to the results section, I suggest reconsidering this
616: ‘grid-scale mean snow depth’: It is unclear what this means
Chapter 4.6: It would be nice to see statements on two additional aspects: 1. Amount and coverage of your data – how transferrable are insights obtained from so little data only? 2. The fact that statistical methods provide only limited understanding of the processes underlying the observed patterns.
679: ‘More often’: Than what? Consider omitting.
691: ‘would benefit from incorporating the meteorological conditions’: It is unclear what ‘meteorological conditions’ refers to and which result allows drawing this conclusion, please specify.