Summary

This paper investigates the effect of urban areas on supercells using a semi-idealized approach. The main findings are that supercells weaken when passing over an urban area. Overall, the paper is interesting but I believe that the authors’ are overselling their results and are misrepresenting previous studies in an attempt to validate their current findings. The authors state that previous studies of interactions between supercells and urban areas ‘lack generality’ and rely on ‘simplified urban representations’.  This seems to imply that this current study ‘fixes’ those issues. However, this current study suffers from the same limitations. The authors attribute their findings primarily to the presence of an urban dry island which reduces CAPE over the city. This is not a universal truth. There are many studies that have found that urban areas can increase CAPE. The driving factor for this difference is the overall synoptic pattern—dry conditions tend to lead to UDIs and a reduction in CAPE, while moist seasons can produce increased CAPE over urban areas. Additionally, I believe that the results of this current study are largely a result of the environmental wind profile chosen. The wind profile is producing southeasterly flow near the surface, which is advecting the urban air to the northwest of the city. While this type of wind profile can be observed in the Great Plains of the US, it is not particularly common in other areas of the US. While studies in Europe are less common, there are several examples of severe weather environments that do not contain easterly wind at the surface. Since the right-moving supercell is not even passing over this ‘urban plume’ that produces positive vertical velocity and initiation of a new cell, I believe that a different wind profile would produce drastically different results.

In addition, I question some of the methodology, particularly the grid spacing. It has been known for over 20 years that sub-grid turbulence is not well represented at grid spacing of 1 km. The authors claim that they are addressing the physical processes that result in storm modification—something that they claim is lacking in previous studies—yet they employ a model configuration that is known to be deficient at resolving storm-scale processes. I do not understand the purpose of simulations with 1 km horizontal grid spacing. This is too coarse to study storm dynamics yet to fine to mimic operational convective-allowing models.

Specific Comments

1. I think the results are only valid for this specific environmental scenario. For example, it is easily seen in figures 13 & 14 that the largest perturbations to CAPE and upward motion are found ‘downwind’ of the city at 950 mb and the primary supercell in your simulations is not passing over that area. In an environment without a near surface easterly wind, this area of upward motion (and strongest SRH anomalies) would be to the east of the urban area. A storm passing over this may be amplified. A similar argument can be made for SRH. Perturbations in SRH appear to be due to obstacle flow around the urban area. These perturbations would be different with a different wind profile.
2. UHI magnitude of 1.5 K, which the authors’ state is in-line with a previous study (Hidalgo et al. 2010). However, that study is based on theoretical scaling arguments. The first line of that paper says that nighttime UHI magnitude can be as high as 5-8 K in European cities. The stated UHI value of 1.5 K seems to be what they calculated the UHI to be for a specific urban breeze—they then stated that this breeze is close to that observed in a specific city on a specific day. My main point is that this value seems small. Naylor and Mulholland (2023) and Naylor et al. (2024) found that a stronger UHI was necessary to induce substantial modification.
3. How deep is the UHI in your simulations? Does the depth match observations?
4. There’s a lot of emphasis on urban dry island. I would like to see more justification of  the dry island assumption as well as the magnitude of the UDI, especially since several studies have noted that urban areas can increase CAPE. The authors are justifying the UDI magnitude with results from a single study (Meili et al. 2022), yet another study from the same journal shows that UDI and UMIs both occur with regularity (Huang and Song 2023). Other previous studies have noted (e.g., Huff and Changnon 1973) noted that urban modification to convection is most prominent during wetter summers. In addition, the reduction in CAPE noted by the authors is much greater than that observed by Rozoff et al. 2003 and  Reames and Stensrud (2018). Reames and Strensrud also note that the reduction in CAPE is ‘generally neutralized as the storm approaches’, while Rozoff et al. noted an increase in CAPE near the urban center.  How can you be confident that your observed reduction in CAPE is a reasonable representation of reality?
5. How is your methodology any more sophisticated than that of Naylor et al. 2024? You are performing idealized simulations similar to Naylor et al. 2024 with a simple circular city (I believe with horizontally homogeneous characteristics) but at a reduced horizontal and vertical resolution. I do not understand how the use of BEP makes the simulation more ‘realistic’. Changing the BEP parameters such as urban fraction, building height, etc are simply modifying surface fields. For example, changing the building plan area fraction modifies the surface drag coefficient. These experiments seem conceptually similar to that of Naylor and Mulholland 2023 in which the surface roughness and skin temperature were altered.
6. Why is 1 km horizontal grid spacing appropriate for this study? It has been known for over 20 years that turbulence is not well resolved at this grid spacing and that that convection needs grid spacing of 250 m or less (e.g. Bryan et al. 2003).
7. Your input sounding has a rather sharp low level inversion. Is it possible that this inversion is preventing UHI air from entering the storm updraft? Have you ensured (via trajectory analysis) that urban area is being ingested by the storm updraft?
8. What is the integration depth for your UH calculation? In addition, have you looked at 0-1 km SRH in addition to 0-3 km SRH? How are you calculating CAPE? Are you using a surface-based parcel or a mixed layer? Are you calculating CAPE with temperature or are you using virtual temperature?
9. Periodic boundary conditions are an interesting choice. Why not use open boundaries like other studies? The initial warm bubble creates a fast moving wave that propagates across the domain

References

Bryan, George H., John C. Wyngaard, and J. Michael Fritsch. “Resolution Requirements for the Simulation of Deep Moist Convection.” Monthly Weather Review 131, no. 10 (2003): 2394–416. https://doi.org/10.1175/1520-0493(2003)131%253C2394:RRFTSO%253E2.0.CO;2.

Huang, Xinjie, and Jiyun Song. “Urban Moisture and Dry Islands: Spatiotemporal Variation Patterns and Mechanisms of Urban Air Humidity Changes across the Globe.” Environmental Research Letters 18, no. 10 (2023): 103003. https://doi.org/10.1088/1748-9326/acf7d7.

Huff, F. A., and S. A. Changnon. “Precipitation Modification By Major Urban Areas.” Bulletin of the American Meteorological Society 54 (1973): 1220–32. https://doi.org/10.1175/1520-0477(1973)054%253C1220:PMBMUA%253E2.0.CO;2.

This manuscript presents results from 1 km WRF model simulations of supercell storms moving across an urban area.  Various experiments are conducted for different urban sizes and characteristics.  Eleven member ensembles are used for each experiment and results compared with an ensemble that has no urban area.  Convection initiation is produced using a warm bubble in each ensemble member.  Results indicate that the urban area leads to a mean weakening the supercell, with larger urban areas producing more weakening for most variables.  The results are interesting, but I have a number of significant concerns that merit attention and should be answered.   

1) The plot of normalized mean UH (Figure 5) shows that the number of ensemble members changes across the different urban sizes. For Figure 5a there are 8 members for an urban diameter of 15 km and 7 members for an urban diameter of 60 km.  Some urban diameters have as few as 5 members displayed on the plot. 

It is earlier stated that every experiment has eleven ensemble members, indicating that only some of the simulated supercells are actually reaching the urban area before they dissipate. This large variability in the number of the members that are used in the analysis is a significant concern.  This variability influences the calculations of ensemble mean UH from the urban runs (Figures 6b-f) which show large fluctuations from the beginning of the storm track and could be due in part to having fewer members.  It also will influence the calculations of mean slope shown in Table 9.  The no_urban ensemble has the most consistent mean UH track of the ensembles; other tracks show greater variation in mean UH along the track.  If roughly half the supercells are dissipating before reaching the city within such a favorable environment, then this behavior raises questions about the experimental design.  For example, Figure 7 shows that the developing supercell dissipates entirely before reaching the city (and doesn’t appear to produce a left-moving storm) while Figure 10 shows only small differences in environmental conditions in this area of storm demise that is well outside the zone of perturbations associated with the urban areas shown in Figures 12-15. It is important to understand what is causing so many storms to dissipate within such a favorable environment prior to reaching the city, as this behavior clearly influences (and indeed may dominate) your results.  

2) Your results show that many supercell characteristics decrease as city size increases, but the results also have large spread and some members intensify as city size increases. A deeper dive is needed to understand this behavior. 

Results shown in the supplemental materials indicate that the supercell tracks from the ensemble members deviate from each other sufficiently to pass over different parts of the urban area in the no_urban run.  Similar tracks are expected for the urban runs, such that some storms may pass along the southern urban edge, while others may pass along the northern urban edge.  Results from Reeves and Stensrud (2018) indicate that the impact of the urban area on the supercell is influenced by the location of the urban area relative to the storm. 

Since the figures of normalized UH as a function of urban area size show both enhanced UH and reduced UH, the question of where the urban area is located relative to the storms for these simulations is important to address.  One way to explore this question is to look at a few of the members with the largest normalized UH and compare with the same number of members with the lowest normalized UH for a fixed urban size.  Plot the tracks, explore the supercell characteristics, and determine if these storms are passing consistently across different parts of the urban area. 

3) The control no_urban ensemble plot of cumulative ensemble mean UH (Figure 6a) shows a lowering/break of mean UH over the city location that is specified. This suggests that the simulated supercells are undergoing some transition at this point.  Thus, all the supercell storms likely all pass over the urban area at a time when the storm behavior in the control ensemble is arguably most sensitive to the environment, as the storm is in transition.  This needs to be mentioned and discussed, as it could influence the results.

4) The simulated sounding at storm initiation (4 pm LST) shows that the surface winds are quite weak at this time (Figure 1b). This will greatly reduce the impact of friction owing to buildings on the flow.  This deserves greater attention in the discussion. 

5) It is also important to mention that the no_urban run shows SBCAPE, 0-3 SRH, and 0-1 SRH all decreasing by storm initiation time. Thus, the supercells are moving into a less favorable environment as they approach the city, which could influence your results.

6) Details are missing and need to be added for completeness and reproducibility. These include:

Soil moisture in urban and cropland regions

Depth used in UH calculation (2-5 km AGL?)

How mean UH is calculated (over what area? How determined?)

Multiple figures show comparisons without a time stamp (Figures 11-14).

7) You state on lines 100-101 “the most accurate parameterization of urban processes in mesoscale modeling.” This is an opinion, unless you have a reference to a study that clearly shows this.  Please delete.