This study investigates and quantifies hazard, exposure, and vulnerability to heat and cold extremes in the Italian region Trentino Alto-Adige for 1980-2018 and calculates the resulting combined risk. The structure of the paper is generally clear, and the presented results are mostly convincing. My main comments concern 1) the language, 2) a more precise estimation of the contribution of hazard, exposure, and vulnerability to the overall risk, 3) extending the figure captions, and 4) adjusting the p-values of the statistical significance tests to control for the false discovery rate.

Main comments:

Although the manuscript is generally well comprehensible, the structure of some sentences and the some of the terms that are used make some parts difficult to understand. I would thus recommend to carefully check the whole text again with a special focus on rephrasing cumbersome sentences (some examples are listed under “specific comments”)

I think that it would be possible to calculate the contribution of changes in hazard, exposure, and vulnerability to the overall changes in risk ratio (e.g. by keeping exposure constant while changing the other parameters, and similarly for the other parameters). I think this could provide a valuable insight into the importance of climate change vs population and socioeconomic changes.

The captions of the figures are currently very short and contain insufficient information to fully understand the associated figures. A caption should be written such that it is possible to understand a figure and its main message only from watching the figure and reading its caption (i.e., without the need to read the main text). I would thus recommend extending the captions such that they explain the figures and the displayed features more comprehensibly.

Many of the figures contain estimates of statistical significance. As the multiple statistical tests (which I presume are conducted independently for each grid cell) may cause to overestimate the statistical significance (Wilks, 2016, https://doi.org/10.1175/BAMS-D-15-00267.1), I would suggest adjusting the p-values by controlling for the false discovery rate as proposed by Wilks (2016).

Specific comments:

• Lines 15-16 (and generally for the description of the Tweedie distribution): I think it would make sense to first mention that HWMId and CMWId are normalized to the interval (0, 1) to combine them with the exposure and vulnerability metrics, and only then writhe that the Tweedie distribution is used for this purpose.
• Line 17: Maybe better “which are used to derive vulnerability”
• Line 18 ff: I am wondering how the increased resilience is determined? Maybe the factors causing the increased resilience could be mentioned here (same for CW)
• Line 36 (and other occasions): I think that the text would be easier to read if an “s” would be added to the acronyms for the plural forms of “heat wave” and “cold wave” (i.e., HWs, CWs).
• Line 38: In which direction do they change? Increasing or decreasing?
• Line 42: How are heatwaves defined in this study? Based on percentiles? Or is it HWMId?
• Line 43-44: This part of the sentence about GCP losses is a bit difficult to understand. I would suggest rephrasing it.
• Line 71-73: Rephrase, as the last part reads rather cumbersome.
• Line 99: Maybe “are most exposed to” instead of “affect”
• Line 113-114: What does “normalized population” mean? Can this be shortly explained here?
• Line 134: Remove “for the”
• Lines 141-143: Something with the reference to Figure 1 is wrong
• Lines 145-146: I think it would be good to exactly state the population of Trento, Bolzano, Merano and Rovereto
• Lines 157-160: I think it would be good to shortly explain which variables are used for the extrapolation of the temperature dataset (e.g. height, land cover, something else?)
• Lines 164ff: What is the reference period for calculating HWMId? I would also explicitly mention that data are pooled from a window of 15 days before and after each day (currently this is not entirely clear).
• Line 175: I think rather “daily heat magnitude”
• Line 176: Are the percentiles calculated from the temperature distribution or from the yearly maximum temperatures? (the latter is done in the original publication by Russo et al.).
• Line 178: I would write “only consecutive days with HMd above 0”
• Line 189-190: I think it would be good to explicitly write that based on the definition used int this paper, CMd is always <0
• Lines 210-212: This is partly a repetition, maybe shorten it?
• Line 220ff: I would suggest writing more specific what the KS test has been used for in this paper (“statistical fit verification” sounds rather generic)
• Line 230: “population data”
• Lines 254-256: This sentence is not clear to me. Could it be explained a bit more in detail how this was done and why this approach was chosen?
• Lines 274-279: Another approach could be the temporal linear interpolation of the exposure and vulnerability variables
• Line 303: Here, does HW mean the yearly HWMId values or something else? Could that be specified?
• Line 305-306: If I understand correctly, there should only be 3-4 values for HW10Y in each pixel, given that a period of 39 years is used. I am not sure whether a trend can be deduced from such few data points.
• Line 312: Maybe “that was” instead of “and”?
• Lines 324-328: I would add “event” after HW and CW.
• Lines 329-331: But Figure 3 does not present a separation of both effects! It shows the combined effects of changes in HWs and of changes in population. I think that for disentangling both effects, one of them would need to be kept constant (see also main comment above)
• Line 350: Not sure that “extreme age” is the right term
• Line 360: I would delete “somehow”
• Lines 362-365: Does this refer to the study by Frigerio & De Amicis?
• Lines 368-372: These results cannot easily be seen in the figures. I would suggest to change the figures to make this better visible (see my comments to Figure 5 below)
• Lines 377-380: What are the main factors? Can they be identified, and can their contribution be quantified? (see also my main comment above)
• Line 407: Mainly “normalized” instead of “sized”
• Line 409: I do not really understand the meaning of this sentence.
• Line 428: Are there any proofs/studies showing that this is the case “likely also in many other regions”? Otherwise this statement should be deleted.

Figures:

• Figure 1: I think it would be good to have some more information in the caption, e.g. that Merano, Bolzano, Trento, and Rovereto are the main cities in the region and what the colours mean.
• Figure S-1: The abbreviations used in the legends and the titles should be explained in the caption.
• Table 1: What are the exact definitions of “population living in at risk housing” and “population with low income”? Could you be more specific? And does the diploma/degree for “population with low education” refer to school or university degrees?
  
  Why has the year 1960 been used for the category “people in old houses”? Is this an arbitrary choice or are there reasons to choose this year?
• Figure S-3: I wonder whether it would make sense to add the borders of the municipalities/districts shown in Figure 4 also to this figure. Moreover, I would suggest to use a linear color scale between 0.05 and 1 and add more ticks to the colormap (not just 0, 0.05, and 1)
• Figure 4: I think it would be good to add a figure (or a subplot) that shows the evolution for the four cities as it is impossible to identify them and to see their changes just from the maps. Also, what do the black borders in the figures depict? Is it municipalities or districts? This should be added in the caption.
• Figure 5: The trends are difficult to see due to the many hatching lines. I would suggest having only fewer hatching lines (like in Figure 2). Also, how is it possible to see that a trend is positive or negative? (the tau values are positive both for HWs and CWs). Like for Figure 4, I’d suggest adding a separate panel showing the results for the four cities.