ACP-2021-597: Convection-Aerosol Interactions in the United Arab Emirates: A Sensitivity Study

This manuscript ‘Convection-Aerosol Interactions in the United Arab Emirates: A Sensitivity Study’ mainly investigate the impacts of aerosol loading and properties on the atmospheric circulation, convective activity, surface/air temperature, and local precipitation by Weather Research and Forecasting (model) in UAE on 14 August 2013. The authors carried out ten different scenarios for WRF simulations and compared the different results of circulation, radiative effect, convective, and rainfall. In general, the paper presents in a logical way, but the English writing need to be greatly improved. Some interesting results of this manuscript will be helpful to understanding the interactions between the convection and aerosol. I therefore recommend publication of this paper in Atmospheric Chemistry and Physics after major revisions. My comments are listed as follows. for his/her valuable comments/suggestions, which helped to improve the quality of the manuscript. Following his/her feedback, we have rephrased poorly written and potentially confusing sentences and put more emphasis on the main findings of the study. Below we address the reviewer’s queries one by one, highlighting in the text where changes, if any, were made.

Many conclusions of this manuscript are consistent with the previous publications. For instance, -(Page 1, Abstract, lines 13-15) 'The convection on 14 August 2013 was triggered by the low-level convergence of the circulation associated with the Arabian Heat Low (AHL) and the daytime seabreeze circulation.' This conclusion is the same as the previous publications in (Page 3, 1. Introduction, Lines 113-116.) 'As discussed in Schwitalla et al. (2020) and Branch et al. (2020), it is normally triggered by the convergence of the low-level circulation associated with the Arabian Heat Low (AHL; Fonseca et al., 2021), the sea-breeze circulation from the Arabian Gulf and Sea of Oman, and the upslope flows on the mountains.' -(Page 6, 1. Introduction, Lines 123-124.) 'Here, they are commonly triggered by the low-level convergence of the AHL and sea-breeze circulations (Steinhoff et al., 2018)'. -(Page 2, Abstract, lines 31-32 and the Conclusions) 'The surface downward and upward shortwave and upward longwave radiation fluxes are found to scale linearly with the aerosol loading, ….' This conclusion is consistent with (Page 4, 1. Introduction, Lines 80-84.) 'Liu et al. (2020) used the WRF model with Chemistry (WRF-Chem;Grell et al., 2005) to investigate the effects of biomass burning aerosol on radiation, clouds and precipitation in the Amazon basin. The authors found that ACI effects prevail at lower emission rates and low values of aerosol optical depth (AOD), while the ARI plays the largest role at high emission rates and high AODs.'

REPLY:
We thank the reviewer for his/her comment.
The statements in the first two bullet points above are related to the triggering mechanisms of the convective event considered in this study. As noted in the Introduction (lines 112-134), there are two main types of summertime convective events in the United Arab Emirates: on the eastern side around the Al Hajar mountains, for which the cloud development is aided by the topographic circulation (Branch et al., 2020, Francis et al., 2020, and on the western side, where clouds typically arise from the low-level convergence between the Arabian Heat Low circulation  and the sea-breeze from the Arabian Gulf (Steinhoff et al., 2018). The 14 August 2013 event falls into the latter category, and therefore the triggering mechanism is expected to be consistent with that of previous studies of similar convective events. We would also like to stress that the sentences the reviewer refers to here are mostly in the Introduction, where a literature review is normally given, while in the remaining of the paper, and in particular in the Discussion and Conclusions section, the focus is on the novel results of the work.
Regarding the sentences in the third bullet point, Liu et al. (2020) did not reach the conclusion that the surface downward/upward shortwave and upward longwave radiation fluxes scale linearly with the aerosol loading. Instead, the authors focused on the aerosol-cloud and aerosol-radiation interaction effects on precipitation, which we also discuss in our study. As a result, that particular conclusion is novel and worth being reported in the Abstract.
Given the reviewer's comment, and in particular in the Discussion and Conclusions section, we have put greater emphasis on the novel findings of the study (lines 790-835) and now clearly highlight the take-home messages for potential future readers of this work (lines 837-853).

(Page 11, 2.2 WRF Experiments and the whole context):
The authors implemented 10 different scenarios for WRF simulations based on two aerosol distributions (an idealized aerosol distribution profile and a climatological profile) and compared the different impacts of aerosol loading and optical properties on the atmospheric circulation, radiative effect, convective, and rainfall. The authors carried out a lot of simulations for sensitivity experiments and acquired many conclusions, but it is not clear for the readers, which conclusion is important and which one is close to the observed results for this manuscript.

REPLY:
We fully agree with the reviewer we have to emphasize the main findings of the study. As stated in the reply to his/her previous comment, we have now made it abundantly clear in the Discussion and Conclusions section our main findings and take-home messages for readers of this work (lines 790-853).
Regarding a comparison with the observational measurements, no simulation clearly outperformed another.
In fact, Table S1 shows that, by and large, WRF's cold and dry biases are present in all model runs, and readers interested in running WRF over hyper arid regions need to be aware of this bias. Having said that, we are in a position to issue recommendations for users interested in running the WRF model for such convective events in hyper-arid regions located next to major dust sources like the United Arab Emirates (UAE): ➢ When accounting for the observed aerosol loading, using a climatology-based distribution is preferable to an idealized distribution as it can improve the representation of deep convection, as evidenced by the increased precipitation generated by the model and the colder cloud tops, in particular when the aerosolradiation interaction (ARI) effects are switched on. The vertical profiles of variables like temperature are also better simulated; ➢ Even in short-term (2-day) simulations, the fields in the interior of the WRF nests can be substantially different from those in the input (in this case reanalysis) dataset. Employing nudging in the outer nests (in this case in the first two model grids) is preferable to only applying it in the outer nest or not doing it altogether, as it helps to at least partially correct some of the WRF biases; ➢ It is vital to accurately represent the properties of the observed aerosols in the model, more so than the amount, provided the order of magnitude is in line with that observed. If the aerosols are more absorbing, the heating in the aerosol layer will peak closer to its top instead of in the bottom half, which has implications for the dynamics and convection in particular if the aerosol layer is deep and/or multiple layers are present.
We have stated this in the text (lines 837-853) and would like to thank the reviewer again for raising this issue.
Whether the changes of aerosol loading and optical properties in the WRF sensitivity simulations could reflect the true observations or not?

REPLY:
We thank the reviewer for raising this point. In Fig. 6c we compare the model-predicted aerosol optical depth (AOD) with that given by the MERRA-2 reanalysis dataset. This particular dataset explicitly accounts for aerosols and their interactions with the climate system, and is found to perform well in this region (Roshan et al., 2019;Ukhov et al., 2020). While we can get the correct order of magnitude when scaling the climatological aerosol loading by a factor of five, the diurnal trend in the reanalysis dataset is not simulated by the Weather Research and Forecasting (WRF) model. We speculate on why this may be the case (lines 504-518). Due to the extensive cloud cover on 14 August 2013 (Figs. 2a-c), AOD estimates from ground and satellite assets exhibit gaps and missing data and hence cannot be used to directly evaluate the WRF predictions. What is more, we do not have aircraft measurements of aerosol loading at different heights to assess the vertical distribution nor information regarding its optical properties. We understand this is a limitation of the study and have noted it in the text (lines 504-510). We believe a comprehensive assessment of the simulated aerosol loading and properties would require additional observational data that is not available for us. This has also been highlighted in the Discussion and Conclusions section (lines 858-863).
In this manuscript, the authors indicated that 'The 14 August 201 was also a rather dusty day in the UAE, with Aerosol Optical Depths (AODs) in excess of two', and I suggest the authors should implement the sensitivity of the potential effects of dust aerosols' loadings and optical properties on the circulation, convection, radiative forcing, and precipitation.

REPLY:
We thank the reviewer for his/her comment. This is precisely what we do in our study, and the 14 August 2013 event is selected as it features dusty and convective conditions in the country on a day for which observational data is available for model evaluation, as noted in lines 394-397. In this work, and through sensitivity experiments with the Weather Research and Forecasting (WRF) model, we explore the changes in atmospheric circulation, convection, radiation and precipitation to different aerosol loadings and properties, considering both an idealized and scaled version of a climatological distribution of aerosols. Examples of this are listed below: ➢ Circulation & Convection: Figs. 6a-b for ERA-5 and the WRF-3 simulation; Fig. A1 for runs WRF-4 to 5; Figs. 9c-d for simulations WRF-5 and 9; and Figs. 10c-d for runs WRF-6 to 8; ➢ Precipitation & Convection: Fig. 8 and S2 for all WRF simulations; ➢ Radiation: Figs. 9a-b for simulations WRF-5, 6 and 9; Fig. 10a-b for simulations WRF-6 to 8.
Besides, we compare the WRF predictions with in-situ measurements at the location of 35 weather stations spread out over the UAE (Fig. 1c). We present the results for the diurnal cycle for runs WRF-1 to WRF-4 in Fig. 7, and for all simulations we give the skill scores for the full day in Table S1.
However, and as stressed by the reviewer's previous comments, we agree that in the previous version of the manuscript we have not clearly highlighted our findings in the text, which is a cause for confusion. In the revised version of the paper we have done so in the Discussion and Conclusions section (lines 790-853) and in the Abstract (lines 20-34).