The authors have presented a useful dataset with clear explanations of the processing. The dataset allows for the fair comparison of extreme heat and cold events across the globe, subject to the geographical biases inherent in the available data. The dataset clearly provides added value to the EM-DAT database.

I have a suggestion and also a question. Firstly, in section 2.2.4 the authors could state the use of a moving average window earlier in the first paragraph to avoid confusion. Secondly, I'd like to know why MSWX was chosen over ERA5-LAND? It isn't clear to me from the text what advantage MSWX offers over the hourly 0.1 degree scale reanalysis provided by ERA5-LAND.

This data paper introduces SHEDIS-Temperature, a curated dataset linking EM-DAT national disaster records for heat and cold waves to subnational geometries, meteorological data, and population exposure. The dataset is useful, well-structured, and has clear potential for cross-national hazard–exposure analysis, model benchmarking, and policy applications. Overall, the manuscript is strong, but several clarifications and additional details would improve transparency, reproducibility, and usability.

Major comments

1. Abstract: The abstract should include some key evaluation results (e.g., mean absolute error between EM-DAT and MSWX extremes) to convey dataset reliability at first glance.
2. Detrending procedure: The manuscript should explicitly clarify what “detrending” means—whether it refers to removing the long-term climatological trend or the seasonal cycle. Equations or a concise methodological description would help. Please also discuss the sensitivity of detrending results to the choice of reference period.
3. Advances over EM-DAT: Although Figure 1 touches on this, the global advances in spatial coverage and finer geometries relative to EM-DAT are not clearly visualized. A global map showing EM-DAT vs. SHEDIS coverage would highlight the added value.
4. Choice of MSWX: The manuscript should justify why MSWX was selected as the meteorological input, rather than ERA5-Land, which has the same spatial resolution. A short rationale (e.g., bias corrections, variable availability) is needed.
5. Cross-comparison with independent datasets: The study compares EM-DAT records with MSWX-derived extremes, but this is not fully independent from SHEDIS. A cross-check with another dataset (e.g., E-OBS, GHCN, Berkeley Earth, or reanalyses) would provide an independent validation.
6. Apparent temperature: Provide more detail on how apparent temperature was calculated (equations, inputs). This is important for reproducibility and comparability with alternative indices such as UTCI or WBGT.
7. Area vs. geo-projection: Tables define variables in km², but the gridded input is in WGS84. Clarify whether the data were reprojected or area-corrected to ensure comparable cell areas across latitudes.
8. Percentile thresholds: Provide references for the use of a 31-day window for percentile determination.
9. Minimum duration: Provide references or justification for the choice of a three-day minimum duration for events.
10. Uncertainty guidance: There is no quantified uncertainty guidance, required by ESSD.

Minor comments

1. Correct minor typos:

• “logaritmic” → “logarithmic”
• “recrods” → “records”
• “percieved” → “perceived”
• “Jammu and Kasmir” → “Jammu and Kashmir”
• “the the” duplication
• “Files within in these subfolders” → “Files within these subfolders”