Review of: “A satellite view of the exceptionally warm summer of 2022 over Europe” submitted to NHESS by Martins et al.

In this study the authors perform a satellite-based analysis of the unusually high temperatures experienced during the summer of 2022 in Europe. They use a wide array of satellite products from MSG-SEVIRI, including the LSA-SAF All-Sky LST product. The manuscript is well-written, the methodology is clearly outlined, and the conclusions are well-supported by the results. The chosen topic also aligns with the scope of the “Natural Hazards and Earth System Sciences” journal. Although the study may not introduce particularly novel insights, remotely-sensed LST is not used so frequently for monitoring heatwaves. Consequently, this manuscript contributes valuable information to the existing body of knowledge.

However, prior to publication, I recommend that the authors make substantial improvements in two aspects of their study: (a) establishing better connection between their methodology and findings with previous works in the literature, and (b) enhancing the robustness of their methodology and conclusions when comparing LST with reanalysis variables.

I list below my concerns in more detail, along with some additional minor issues.

Major issues:

1) As the authors note in the abstract that the climate applications of LST are still poorly explored, it is crucial for them to offer a more in-depth analysis of previous studies, elucidating also points of agreement or disagreement in the findings. Therefore, additional discussion is necessary in the Introduction regarding past research that has utilized remotely sensed LST for studying heatwaves, and this discussion should also be incorporated when presenting the findings of the current study. While some relevant works have already been cited in the references, they would benefit from more comprehensive discussion. Furthermore, there is room to expand the list of prior works that contribute to the overall understanding of the topic.

2) Regarding the comparison with reanalysis I have the following remarks:

Why did the authors choose to use T2m from ERA5 instead of ERA5-Land? It appears to me that the latter would be more suitable for consistency, considering that SKT comes from the ERA5-Land product.

Lines 262, 302, 370: The authors in various parts of the text highlight the difference in physical meaning for T2m, however the patterns of T2m in Figure 4 (and also throughout the manuscript) appear to closely resemble those of SKT. Have the authors examined the difference between this pair of variables, similar to what was done with LST? It may be beneficial to include these figures as well.

Line 273: Please provide a quantitative measure of the differences in temperature anomalies over burned areas between LST and SKT.

Lines 275 – 288: Please present aggregated statistics about the magnitude of the difference between LST, SKT, and T2m anomalies.

Line 335: I may have missed it, but has SWI been defined? In addition, is SWI derived using inputs from MSG that are common with those used to derive the LST All-Sky product? If yes, this could explain part of the strong resemblance between the two, and it should be mentioned in the conclusions.

Minor issues / Technical comments

Lines 58 – 63: This discussion seems somewhat beyond the manuscript’s scope. Could you rephrase it to be more concise?

Line 84: Why is the significance of an all-sky product for heat extremes emphasized, considering that clear-sky conditions are typically the norm?

Line 114: It is unclear what the authors mean with “overall accuracy of 0.0 K”? Maybe bias is a more suited word?

Line 121: Have cases of active fire been included in the analysis? If yes, can the authors provide justification how this inclusion may not significantly impact the results about heatwaves?

Line 142: From this point onward, it appears that the section numbering is disrupted.

Lines 165 – 175: I recommend presenting the description of monthly and seasonal anomalies first, followed by the heatwave definition and metrics. This also aligns with the presentation in the results section.

Figure 2, 5: For clarity, I suggest avoiding the use of a dash (-) as a separator in subplot titles. This is particularly important to prevent confusion, as the minus sign is also used in titles to indicate differences.

Line 260: I believe you meant to write “ERA5-Land STK and ERA5 T2m”

Lines 351-352: Is there a reference to support this claim?

Line 352: Is it ERA5 or ERA5-Land? It seems the authors use these terms interchangeably also in other parts of the text.

Line 440: It is not clear to me what is meant by “for satellite observations these errors are mostly mitigated”. While satellite-based observations have their own sources of uncertainty, these do not typically involve “interpolation” or “model” errors, as suggested here.

## Overview

The manuscript from Martins and colleagues makes an exhaustive description of the 2022 heatwave in Europe with a specific focus on how land surface temperatures (LST) from geostationary satellite can provide valuable information to describe such event. The manuscript is well-written and well-documented. It may sometimes reads a bit too much like a weather report rather than a scientific paper, and it does not bring major novel scientific insights, but overall is informative enough to merit publication if some points are considered. 

The big question is actually 2023. It is a pity that the process of scientific analysis, writing and reviewing takes so long, and I fully understand that the authors had the intention to focus on 2022 before the summer conditions of 2023 came about. But now we are in December 2023 and arguably all the information to add 2023 to the current analysis could be done. It is true that this may somewhat complicate the message, as 2023 may show to be even more extreme than 2022 in some areas, but also maybe not. I would urge the authors to consider to add 2023 also to the analysis (at least in Figure 9 and maybe 10) to ensure their analysis does not become a bit obsolete before it is even published. Additionally, I would ensure something is said about 2023 in the discussion/conclusion.

While LST_max is seen to have differences with T2M_max and SKT_max, which is interesting and informative, I think this analysis could be pushed a bit more to give a bit more insight of these difference. Specifically these seems to scale with magnitude, and it would be wise to try to qualtify/visualize this to provide some info on the non-linearity of the relationship. I would thus strongly recommend some graph maybe showing the deltas of all pixels on the y-axis and the magnitude of each pixel on the x-axis (for both T2M and SKT separately)

On another take, I was quite interested in knowing more about the specific performance and appropriateness of the all-sky LST (versus the clear-sky LST). It seems that it is a bit taken from granted that it is assumed to be better (because more gap-filled). However, this paper could be the nice opportunity to evaluate better how it performs against the clear-sky in terms of relationship with other indices (SKT_max, T2M_max), and thereby giving an extra relevance for this paper. 

From the presentation point of view, figures would be clearer and more accurate using an appropriate geographical projection for Europe which respects the concept of equal area, thereby showing better the extent of the meteorological event described in this paper. I would recommend to use the Inspire LAEA for Europe.

## Specific comments... 

- L60 : "led to twice the yield", do you mean increase in yield?

- L72: yes, but this does not relate necessarily to NRT in general

- L86: perhaps it would be welcome here to add briefly a bit more on how this all-sky is produced, or rather, how come this is not the standard? what assumptions are being made to estimate what is below the clouds? I know this is explained later, but a word of two talking about energy balance would fit well

- L101: perhaps good to mention how many MSG satellites there are

- L114: what does overall accuracy of 0K mean here exactly? it reads as if there is no error whatsoever. is it not rather that you mean there is no bias?

- L117: which meteo variables from ERA5 are needed?

- L169: please make it explicit here that this is how you define "hot days" in this paper. It did not seem so clear to me later on that the definitin was here

- Figure 2: in panel a, please change the colourscale as this one is not colourblind friendly

- L269: I suppose the fact LAI is prescribed should also be mentioned here

- Figure 4: would be nice to relate this with absolute values of LST_max/T2M_max. Are the deviations larger particularly where the temperatures are larger? it would surely seem so.

- Figure 7: the colourscale on the right column of figures is als onot ideeal for colourblind people as it goes from green to red.

- Figure 8: graphically it is difficult to appreaciate how the corresponding areas are for non-2022 years, as these are not full. I wonder if there would not be a way to make things more comparable. Maybe shading also (with different colours maybe) the 3 next years with the largest cumulated max Area under Md(Td)>2.

- Figure 9: it would really be interesting to add 2023 on this graph