The manuscript “Convection-permitting climate model representation of severe

convective wind gusts and future changes in southeastern Australia” by Brown et al. investigates how future convective wind gusts might change over Southeastern Australia. Previous research suggests that surface hazards from thunderstorms, like severe convective winds (SCWs), may alter with climate change, yet global climate models struggle to resolve these due to their small scale, leading to uncertain projections. The authors find that SCW events using a convection-permitting climate model (BARPAC-M) over southeastern Australia for December–February, and comparing with a regional parent model (BARPAR) improved representation of extreme wind gusts in BARPAC-M but overestimation of SCW frequency, particularly in certain environments. Projected changes in SCW frequency for 2050 show uncertainties, with potential decreases under certain conditions and increases under others, highlighting the complexity of future SCW trends in the region. The study is very well structured, and written, and the images and text are of high quality. The differentiation of wind gusts into different categories adds a lot of novel insights about model biases and future climate change impacts on these extremes. This is one of the best-written and interesting papers that I have read in a while. I have only a couple of minor suggestions for changes and recommend publishing this manuscript after those are addressed.

General comment

Adding more discussion on the importance of climate internal variability on your results would be beneficial. You mention the high degree of spatial and temporal variability of convective gusts already but making the role of internal climate variability more explicit would be important (see e.g., Deser et al. 2012). Internal climate variability could easily be the dominant source of uncertainty in your future climate projections.



Deser, C., Phillips, A., Bourdette, V. and Teng, H., 2012. Uncertainty in climate change projections: the role of internal variability. Climate dynamics, 38, pp.527-546.

Specific Coments:

1. Is BARPAC-M online or offline nested into BARPAR? The online nesting would have the benefit of providing higher-temporal resolution at the lateral boundaries which generally reduces that spatial spinup in the high-resolution domain.
2. How did you account for boundary effects in BARPAC-M? Do you use a sponge zone and did you exclude boundary grid cells from the analysis?
3. L100: It is optimistic to assume that the resolved gust in BARPAR is 10-min if you have a 5-min time step. I would assume that your resolved temporal scales are at least 4Dt based on numerical considerations and model diffusivity.
4. In the analysis of Fig. 2 you directly compare the grid cell wind gust from the models with observed gusts at point locations. Should the model be able to capture point-scale wind gusts? I would assume that the model wind speed should be lower than that observed at point locations since the model is representing a spatial (e.g., grid cell) average wind gust, which in case of ERA5 and BARPAR is a quite large area.
5. 3: Maybe using a log y-axis would make this figure easier to read.
6. L223: Why are you using the 6 km speed here? Are you assuming that this is the source height of downdrafts?
7. 6: Please add a legend that describes the circle sizes.

The manuscript “Convection-permitting climate model representation of severe convective wind gusts and future changes in southeastern Australia” by Brown et al. investigates how future convective wind gusts might change over Southeastern Australia. Previous research suggests that surface hazards from thunderstorms, like severe convective winds (SCWs), may alter with climate change, yet global climate models struggle to resolve these due to their small scale, leading to uncertain projections. The authors find that SCW events using a convection-permitting climate model (BARPAC-M) over southeastern Australia for December–February, and comparing with a regional parent model (BARPAR) improved representation of extreme wind gusts in BARPAC-M but overestimation of SCW frequency, particularly in certain environments. Projected changes in SCW frequency for 2050 show uncertainties, with potential decreases under certain conditions and increases under others, highlighting the complexity of future SCW trends in the region. The study is very well structured, written, and the images and text are of high quality. The differentiation of wind gusts into different categories adds a lot of novel insights about model biases and future climate change impacts on these extremes. This is one of the best written and interesting papers that I have read in a while. I have only a couple of minor suggestions for changes and recommend publishing this manuscript after those are addressed.

General comment:

1. Adding more discussion on the importance of climate internal variability on your results would be beneficial. You mention the high degree of spatial and temporal variability of convective gusts already but making the role of internal climate variability more explicit would be important (see e.g., Deser et al. 2012). Internal climate variability could easily be the dominant source of uncertainty in your future climate projections.
   
   
   
   Deser, C., Phillips, A., Bourdette, V. and Teng, H., 2012. Uncertainty in climate change projections: the role of internal variability. Climate dynamics, 38, pp.527-546.
   
   
   
   

Specific comments:

1. Is BARPAC-M online or offline nested into BARPAR? The online nesting would have the benefit of providing higher-temporal resolution at the lateral boundaries which generally reduces that spatial spinup in the high-resolution domain.
2. How did you account for boundary effects in BARPAC-M? Do you use a sponge zone and did you exclude boundary grid cells from the analysis?
3. L100: It is optimistic to assume that the resolved gust in BARPAR is 10-min if you have a 5-min time step. I would assume that your resolved temporal scales are at least 4Dt based on numerical considerations and model diffusivity.
4. In the analysis of Fig. 2 you directly compare the grid cell wind gust from the models with observed gusts at point locations. Should the model be able to capture point scale wind gusts? I would assume that the model wind speed should be lower than that observed at point locations since the model is representing a spatial (e.g., grid cell) average wind gust, which in case of ERA5 and BARPAR is a quite large area.
5. 3: Maybe using a log y-axis would make this figure easier to read.
6. L223: Why are you using the 6 km speed here? Are you assuming that this is the source height of downdrafts?
7. 6: Please add a legend that describes the circle sizes.

The authors investigated the severe convective wind (SCW) events based on the climate simulations over eastern Australia using the regional climate model at convective permitting resolution (BARPAC-M) to resolve the deep convections and coarse resolution (BARPA-R) to represent the environmental conditions. The work demonstrated the improved simulation of SCW events with convection-permitting resolution and investigated the SCW in different categories and their possible changes by the middle of the 21^st century under the SSP585 scenario. The manuscript is very well written. The design of the experiments and methodology is sound at the bulk part, and the clustering-based analysis is very interesting! I would recommend minor changes before it is accepted for publication.

First, there is a concern about the direct comparison of a given cluster between observation and simulation. The clusters from the k-mean method represent the relative differences in the same dataset. Although the partitioning of the events (e.g. percentage of each category) is comparable between the observation and simulation, the direct comparison of each cluster between observation and simulation may not be reasonable since a given cluster may not be physically similar enough between observation and simulation.

Another concern is that the 20-year simulation may not cover long enough climate variability, thus the future changes are sensitive to the resampling. It might be beneficial to show the phases of the major climate variability are similar between the historical and future runs.

The specific comments are as follows,

Line 30-31: The changes in large-scale environment affecting deep convections may not be limited to temperature and moisture. Other factors may be worth noticing/mentioning, such as changes in tropospheric lapse rate, low-level wind shear, and relative humidity.

Line 54-64: The discussion on model uncertainties touches only on the influence of model resolution on dynamics. It might be worth mentioning other sources of uncertainty although resolution is the focus here. For example, the uncertainties in parameterizations, especially the microphysical (MP) parametrization may be worth mentioning since the deep convective systems and growth of hydrometers are strongly affected by MP processes, which further affect the evaporative cooling and downdraft.

Line 90: Climatology based on the “20-year” simulation may be sensitive to climate variability depending on when the period starts. It may be beneficial to show the historical and future periods cover similar phases of the major climate modes that affect deep convection over eastern Australia.

Line 132-136: It might be easier for the readers to follow if the introduction of the terms is in the same sequence as their occurrence in Eq. 2.

Line 243-246: Since k-mean clustering captures the relative difference within the same data, is it possible that the “severe high moisture events” from the observation and simulation are not physically similar enough to make a direct comparison? It would be better to show some evidence of how similar the two clusters are in terms of moisture and related properties before we conclude here. Same suggestion for other clusters if compared directly between observation and simulation.

Figure 7: Since both BARPAC-M and BARPA-R are presented in each panel, It is better to use a common y-axis title to describe both simulations.

Line 258-259: I don’t see a causality relation between the bias in intensity distribution and an overestimation of event numbers here

Line 261-262: Based on Figure S7, the overestimated amount of steep lapse rate SCW events does show consistent bias to the large-scale environment from BARPA-R that the model produces more frequent “steep lapse rate” conditions than observed.

Line 271-272: This uncertainty may be due to uncertainty in climate variability.

Line 300: My guess is the BDSD and criteria used to identify the F_ENV were developed based on a different dataset other than the model simulations. It is worth considering 1) the uncertainty in this parameterized metric when applied to a different dataset, and 2) whether the model can well capture the physical relations between the parameters considered in BDSD and SCW. It would be beneficial if some physical explanations could be provided for the opposite changes in SCW and F-ENV.