Review of “Evaluating the ability of the operational High Resolution Rapid Refresh model version 3 (HRRRv3) and version 4 (HRRRv4) to forecast wind ramp events in the US Great Plains” by Bianco et al.

The manuscript evaluates the ability of the High Resolution Rapid Refresh (HRRR) numerical weather prediction model in forecasting wind ramp events in the U.S. Great Plains. The study focuses on two operational versions: HRRRv3 and HRRRv4. Utilizing the Ramp Tool and Metric (RT&M), it demonstrates that HRRRv4 outperforms HRRRv3 in skill, particularly in detecting up-ramp events during summer, which is vital for wind energy integration into the electric grid. The methodology includes using 10-m observational data from METAR stations and model outputs, with statistical analyses carried out for annual and seasonal variations.

The work is timely, addressing the critical need for reliable wind energy forecasting in the context of renewable energy integration. However, some areas require additional clarity or justification. The manuscript is well-written and accessible to a broad audience in meteorology and renewable energy fields.

Major Comments

1. While the manuscript provides a broad description of RT&M, it would benefit from a clearer explanation of its mathematical implementation. How is the skill score derived for different ramp characteristics (e.g., timing, amplitude, and duration)? Reference to specific equations (e.g., Bianco et al., 2016) is helpful but insufficient for readers unfamiliar with RT&M.
2. The decision to use 10 m wind speeds is justified by correlations with 80-m data and data availability. However, as acknowledged by the authors, this introduces uncertainty, particularly when converting the wind speed into power generation – small changes in wind speed can result in large changes in wind power and associated ramps. This limitation should be discussed in more detail. Since the focus of this study is the ramps, I would suggest also evaluating the ramp statistics (with wind power instead of wind speed) of those two levels to address the potential biases.
3. Both HRRRv3 and HRRRv4 have longer periods of data than what is used in this study. Why was only one year of V3 data used? Additionally, although 2021 and 2022 were both simulated by HRRRv4, the large differences observed between these two years (Figure 5,6,7, and 9) indicate that the inter-annual variation in skill may not be fully explained by the model improvements alone. This raises concerns about the representativeness of the dataset. For instance, the conclusion that there is a 50% increase in skill for summer up-ramps; how much of this improvement can be attributed to the model improvement vs inter-annual variability? Can this conclusion apply to other years? Expanding the analysis to include multiple years and evaluate the interannual variability for both versions would strengthen the conclusions.
4. For many figures, the captions are repeated in the main text. I suggest removing this redundancy to save space and instead expanding on the discussion of the figure contents.
5. The geospatial distribution of results has not been sufficiently addressed. Most statistics are averaged over all sites, and there is little discussion of the spatial variability. This aspect could be tied to the physical developments in HRRRv4. Analyzing and discussing the spatial distribution would provide additional depth to the analysis.  

Specific Comments

1. Lines 37-39, please add a reference for this statement.
2. Lines 85-88, this paragraph seems out of place and may connect better to the paragraph starting at Line 70.
3. Lines 108-109: consider moving this sentence to figure caption.
4. Line 111, please also note on the rated speed.
5. Lines, please see my major comment.
6. Line 180, since METAR data is assimilated, should good performance at those locations be expected? How this result applicable to the area without METAR station available should be discussed.
7. Lines 186-188, many down ramps occur around 00 UTC (Figure 7) when artificial “ramp” are also expected. Why not use simulation start at other hours (e.g., 6 UTC) when ramp is less frequent?
8. Figure 5, my understanding is this figure based on model results at site locations. How about the observations? Meanwhile, please indicate that the how the size of the circles was normalized. How does this normalization influence the results?
9. Line 230, larger difference between 2021 and 2022 are observed compared to their difference from 2020, suggest that the interannual variability is more important than difference in model versions?
10. Figure 6, this figure suggests a good consistency between the years. However, the blue color spreads over a wide range of data from 0 to 100%, potentially masking large differences. Consider using more colors within the 0 to 100% range.
11. Lines 238-240, please move this to figure caption.
12. Lines 245-248, Redundant with the figure caption; remove this repetition.
13. Figure 7, please change the title to “Diurnal variability in ramps and wind at 10 m”. Visionally a better wind speed simulation in HRRRv3 compared to V4.
14. Line 257, could you elaborate how the statistics are calculated?
15. Lines 258-259, this has already been included in figure legend.
16. Lines 276-278, could you discussion the spatial distribution? we do see a larger improvement in the region with less ramps.
17. Lines 281-284, already in figure caption.

The paper aims to show the modeling improvements of wind ramp events of the HRRRv4 model relative to the previous version HRRRv3. The study is well-written, highly timely, and relevant. However, the conclusions of the study, that HRRRv4 is demonstrated to generally outperform HRRRv3 for ramp events is not satisfactorily supported in the evidence provided. 

I recommend major revisions to the paper before it is accepted.

General comments 

• The study does not significantly quantify the influence of its assumptions. The first assumption is that model performance at 10 m is a good proxy for performance at hub height. Showing a decent overall correlation between the two model levels for the whole period is insufficient. You should at least focus on showing it for ramp events and actual observations, even if you only have a few sites available. The second assumption is that because the spatial distribution of ramp occurrence is somewhat similar in 2020 for HRRRv3 and 2021 and 2022 for HRRRv4, the comparison in performance between these different years and model versions is warranted (implied in L220-221). You should do more to show the influence of interannual variability on the results and perhaps present the results in a way that makes it easier for the reader to convince themselves of the improvements (the dots in Fig. 5 and 6 are difficult to compare).  
• Alternative hypotheses that could explain the results (the performance improvements seen), such as natural variability, are not discussed or tested, weakening the results and conclusions drawn. Given that HRRRv3 and HRRRv4 are compared across different periods, at the least, some effort must be made to rule out natural variability as the driver of differences. L230 of the paper could indicate that interannual variability is not negligible. 
• Please sure you are following the guidelines of the journal regarding notation, dates, math symbols, etc.: https://www.wind-energy-science.net/submission.html#math, e.g., “1700 UTC” -> “17:00 UTC”, avoid hyphens with abbreviated units (e.g. “10-m wind”), and many more cases. 

Specific comments

• L100-101: The RT&M method is so central to this study that you should spell out the details here, not simply refer to another paper
• Figure 2: I suggest indicating the study area on the map(s). In general, Fig. 2 is presented but not discussed much. Perhaps you can relate the mean and standard deviation to the number of ramp events experienced. Perhaps you could even make a ramp occurrence map. 
• L169-170: Please state how the temporal interpolation was done
• L189-190: Please explain the 3-point smoother in more detail. Is it simply the average between the two? Or something else?
• Figure 4: please add the runs for the 2021-04-07 00Z and 2021-04-11 00Z initializations to the figure to allow the reader to follow the source for the red line throughout the period
• Figure 5: How much of the spatial variation in ramp events is explained by the variation in mean wind speed?
• Figures 5 and 6: it would be helpful to show the frequency distributions of all the samples. This would also help the reader see more clearly the improvements you mention
• Figure 7: It would be valuable if you reflected on how these diurnal cycles may look different for typical hub height. For example, you mention the importance of low-level jets in the text. How would they change the picture? One could, perhaps, expect a reverse cycle at higher altitudes. Also, I would suggest using local time-of-day values or indicating the typical ranges corresponding to day- and nighttime. 
• “Newmann” -> “Newman” in three places on page 6 
• Small suggestion: your author contributions are very short and general/vague. You can take a look at https://publications.copernicus.org/services/contributor_roles_taxonomy.html and perhaps make it more specific