Review of „Sensitivity of atmospheric rivers to aerosol treatment in regional climate simulations: Insights from the AIRA identification algorithm“ by Raluy-Lopez et al.

This is an interesting study which addresses the question of how the interactive coupling of aerosoles models influences the representation of dynamics and precipitation of atmospheric rivers (ARs) in a climate simulations. In the first part a new AR detection algorithm is proposed for the specific application in limited area regional models. This is certainly of great interest for coordinated modelling frameworks as e.g. the CORDEX initiative. The second part analyzes the effects of direct and semi-direct effects of an interactively coupled aerosole model on the trajectory of ARs and AR related precipitation patterns. Both aspects are highly relevant in current research on ARs and the manuscript certainly can provide new insights in these topics. However, the authors should consider the below listed suggestions and comments to improve the manuscript substantially before publication can be recommended.

General comments:

This study is not the first that aim at detecting ARs in regional models. This should be mentioned. Some remarks are given in the special comments.

The description of the algorithm should be improved and some deviation from existing ones should be explained. Unlike others, the AIRA algorithm detects ARs on two longitudes at 10 and 12°E and infers additional length and direction criteria by employing trigonometric functions. Though this is described briefly in the text, a figure sketch with a zoom on L1 L2 to draw the trigonometric elements used to derive the relevant parameters would be helpful, e.g. the direction and length scales. As far as I understand the IVT threshold was calculated using all time stamps and not only those at 12:00 UTC (when moisture is at the higher end) as in for e.g. Lavers et al. (2013). Likely this may result in a lower threshold which should be discussed.

The BASE, ARI and ARC experiments should be better described for those readers who are not specialists in aerosol modelling and those who are not familiar with the WRF-Chem model. What precisely is meant by semi-direct and direct effects on a physical basis? The interaction of aerosoles with radiation beyond the optical depth in ARI should be physically explained. The same would help for the interaction of cloud (micro-)physics. Are condensation-nuclei reduced due to precipitation for example?). If so, in which of the BASE, ARI, and ARCI experiments is this the case? In the current version only references to literature about the WRF model and it’s coupling is given. A brief summary about coupling prognostic variables, input, and output etc would be helpful. This knowledge is essential for the understanding of the results.

In the results sections the physical processes that lead to differences in the three experiments should be better and more verbosely explained to meet a broader readership which are not only atmospheric or aerosole researchers. For example, often a heating or cooling is proposed but as no corresponding temperature anomaly is shown this is hard to see. Also the clustering procedure which is based on leading EOFs of salt and aerosoles is not sufficiently described. All this makes it difficult follow the results and final conclusions. More examples are given below.

Special Comments

line 7: “The analysis of common AR events showed that the differences between simulations

were minimal in the most intense cases, and a negative correlation was found between mean direction and mean latitude differences.

please rephrase: what is meant? you have three sensitivity simulations. When the ARs are located more to the North in e.g. BASE, then the direction is more south in ARI and ARCI? Perhaps it’s better to remove the second part of the sentence.

line 11 deviations from what? What precisely is meant by reinforcement and attenuation? is it the moisture transport (in most studies taken as a proxy for intensity) or precipitation?

Introduction

line 33: what is meant by “anomalous”? Heavy precipitation above a certain threshold?

Climate change is indeed assumed to impact on ARs. However, besides the important studies of Payne and Algarra, there also relevant studies with more focus on Europe and in particular the Iberian Peninsula. Please consider these to mention, like e.g.

Gröger, M., Dieterich, C., Dutheil, C., Meier, H. E. M., and Sein, D. V.: Atmospheric rivers in CMIP5 climate ensembles downscaled with a high-resolution regional climate model, Earth Syst. Dynam., 13, 613–631, https://doi.org/10.5194/esd-13-613-2022, 2022

Ramos, A. M., Tomé, R., Trigo, R. M., Liberato, M. L. R., and Pinto, J. G. (2016), Projected changes in atmospheric rivers affecting Europe in CMIP5 models, Geophys. Res. Lett., 43,  9315–9323, doi:10.1002/2016GL070634.

Lavers, D. A., Allan, R. P., Villarini, G., Lloyd-Hughes, B., Brayshaw, D. J., and Wade, A. J.: Future changes in atmospheric rivers and their implications for winter flooding in Britain, Environ. Res. Lett., 8, 034010, https://doi.org/10.1088/1748-9326/8/3/034010, 2013.

line 46: “tracking its long 2D structure”. Do you mean tracking its elongated 2D structure?

line 51: That’s true. The effect of resolved spatial orography on the representation of AR over land was found most evident over the Iberian Peninsula (see e.g. aforementioned study by Gröger et al.).

line 57-61: The interactive online coupling between aerosole modules and other climate compartments will represent feedbacks by aerosoles in a much more realistic way. May be this could be explained a bit more in the Introduction. Can you mention some feedbacks we neglect if we use only prescribed fields of aerosoles instead of simulated ones?

Methods

line 75: “The WRF-Chem model (v.3.6.1) was used for the simulations, both in a decoupled configuration (WRF alone (Skamarock et al., 2008)) and in a fully coupled configuration with atmospheric chemistry and pollutant transport to account for aerosol-radiation and aerosol-cloud interactions (Grell et al., 2005)”

What does fully coupled mean and how is the coupling precisely done in the three experiments? This is essential to understand the results in this study. The section could benefit from a brief description of the WRF-Chem and how aerosoles have direct and semi direct effects on the models physics.

section 2.1 Data

line 81: “... encompass major dust emission areas”. Which are these areas? The Sahara desert?

line 87: what is CCN and how does it interact with model physics?

aerosol-radiation interactions: does radiation then alter the optical properties of Aerosoles and or the number of condensation nuclei?

line 131: “First, the magnitude and direction of the IVT are bi-linearly interpolated to the detection lines, L1 and L2, enabling the computation of the required variables. “

What variables are meant here? The sentence implies that IVT is calculated from specific moisture, u, and v as a first step and thereafter IVT is interpolated from the models grid to L1 and L2. What variables do you mean here in addition to u,v, and q and for what are they necessary?

Line 134: How is the threshold value determined? Is this threshold latitude dependent? Is it determined from climatological values like e.g. the 85^th percentile as in other algorithms? L1 extents over a wide range of latitudes ranging from semi-arid climates to more wet conditions. Are the northern latitudes more represented in the threshold than those from the south?

Line 139. “...direction of AR...”. If I interpret equation 4 right, wouldn’t the term orientation not better than the term direction? Direction might be related more to the movement of the AR over time.

Line 165: How is “s” determined? Do ARs not move over time so that changes in their axis latitudinal position are not unusual? Please explain why this is necessary.

line 169: “.. estimation of the AR length..”. Do you mean AR duration here? The length scale isn’t determined so far, is it?

185 ff:

Please explain how the value of the mean 90^th percentile is calculated. Is it determined over all latitudinal points (i.e. m=22) at L1 and over the whole time period? Or do you calculate 22 90 percentiles an at the end average over the 22 points? Also contrary to other algorithms you take into account all day times while others include only time steps of 12:00 UTC time stamps (when moisture content is high due to solar heating). This is likely the cause why your value of 260 kg m^-1s^- seems a bit lower than in other studies (see e.g. Lavers and Villarini, 2013: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/grl.50636

Is there any empirical evidence to support the limits for w (150 – 800 km)?

The spatial / temporal criteria listed in Table 1 seem to be more or less reasonable from theoretical/geometrical considerations, but ultimately lack empirical evidence. So it would supportive if sensitivity tests could be made to estimate the sensitivity of the thresholds on the AR frequency, duration and intensity. If this is too much effort, this should be at least discussed in terms of uncertainties associated with the algorithm.

section 3.2.1s

Figure 5 fits very well with result from Gao et al. (Fig. 8) and Gröger et al., 2022 (Fig. 5d). The could be mentioned to support the validity of the new developed AIRA algorithm.

Gao et al.: https://journals.ametsoc.org/view/journals/clim/29/18/jcli-d-16-0088.1.xml

section 3.3 Common events

you may consider renaming the section, e.g. coherence of events or so

I would speculate that the different treatment of aerosoles will alter not only the precipitation pattern of AR related precipitation events but also alter systematically the mean precipitation rates. Could it be that the alteration seen in AR related P are similar to those in mean P? Implying that aerosoles impact similar mean and AR precipitation events.

3.3.1. Analysis of differences

What is the idea of eliminating non coherent AR intervals to elaborate the effect of aerosoles? I think here a more profound explanation for the strategy should be added. From a methodological point of view I would guess ARs penetrate into the EuroCordex model domain roughly at the same time and at the same position. Then, differences in precipitation, IVT intensity and frequency etc. would be attributed to the different treatment of aerosoles. Can you confirm this? Consequently, the non coherent AR time steps would be the result of the aerosole treatment which would neglected in this approach. Would it be wrong to calculate Fig. 6 without the eliminating step?

line 245: “...The maximum IVT was obtained by averaging the maximum IVT magnitude of each AR event in the three experiments…”. You mean intensity here?.

line 246: What is meant by spatial deviation. Does it refer to the deviations in latitude (Fig. 6 middle)? Please be consistent with the terms throughout the manuscript.

line 247: what is meant with “the three magnitudes”. A distinction between three magnitude categories was not done before.

line 249: Can you summarize Fig. 6 to explain what you aim to analyze with the EOF analysis. Are there systematic differences in the deviations to BASE in Fig. 6? At a first glance, it seems like noise (with the exception that most intense ARs seems to be consistent in the experiments). Also, it would be interesting to show at least the first or three leading EOFs for sea salt to get an impression where most variance is concentrated.

section 3.3.2

Please explain more verbose how the EOF analysis was performed, e.g. how were salt and aerosole anomalies calculated, what clustering algorithm was applied etc. Moreover, is there a seasonality in the aerosol fields (as you showed for AR incidents)? Could the clustering also explained by over-representations of certain seasons? Did the classification procedure require to choose the number of different clusters? If so, why were 8 classes chosen?

line 259/Figure 7: “ ...it was observed that an AR weakening occurs in clusters 2 and 3.” Not clear at first reading what is meant. Figure 7 top (which I think this statement is related to) shows the IVT difference at the y-axis, in the caption it reads as “magnitude”, and in the text it is termed weakening.

Figure 7 shows red points which are not explained.

line 262: what kind of frontal surface? that of a storm? An explanation at this first place what is meant by a thickness to non-specialists is lacking.

line 267: “In cluster 3, a wider cooling effect is present, but the more pronounced cooling in the south (over the north of Africa) leads to the observed weakening”. Where is this cooling derived from? Figure 9 show thickness [m]. There is no information about temperature differences at this place.

Accordingly I have to go back to Figure 7 where an elevated dust concentration in the region is visible. Shall I interpret this as proxy for cooling in this region (in the sense of dimming?). So far no explanation for the assumed cooling is given at the place of line 267. I get lost here...

line 279 ff: The ARCI-BASE comparison reads much better than the previous ARI-BASE comparison because physical explanations that appear plausible are given to the reader. This should be likewise provided for the ARI-BASE comparison. Saying this, most of the explanation is based on the interpreted aerosole effect on temperature which is not shown itself, though often it is argued with “cooling” or “warming”. Therefore it would help to show additional plots for temperature either instead of thickness or as supplementary material.

3.3.3 Case studies

line 309: “...-70.32 and 58.01 kg m−1 s−1...” over which area has this been averaged. Over the AR area? Model domain? Iberian Peninsula?

lines 309 to 333:

This paragraph reads very well as it provides a process-based discussion about the aerosol effects on ARs, involving a chain of interactions between temperature, clouds, droplets etc. The role of heating/cooling and temperature gradients is again highlighted and the reader may wonder if it would be possible to support this statement by a figure showing e.g. temperature anomalies.

In particular, the paragraphs (and already the previous ones) emphasize the cooling effect by aerosoles as well as a heating effect from more abundant droplets, prolonged cloud presence, and latent heat gain are discussed. However, it becomes not quite clear why the individual effects (cooling or heating) dominate in the respective cases. This could be more explained.

line 323-324. Isn’ it rather a southwestward shift seen in Figure 15 ARCI-BASE?

Conclusions

Including an atmospheric chemistry and trajectory model yields likely the most realistic and physically consistent treatment of aersoles. But it is likely also the most expensive? If so can we derive from the experiments a statement wether or not the additional online coupling of an expensive chemistry/aerosole model is worth and/or in which cases? Can we expect systematic shifts in AR related precipitation and or moisture convergence which may be of importance on climate related time scales? Would the conclusions also hold for e.g. the U.K. which is further away from major dust aerosole sources?

Review for EGUsphere (GMD, MS type: Methods for assessment of models)

Title:  Sensitivity of atmospheric rivers to aerosol treatment in regional climate simulations: Insights from the AIRA identification algorithm

Authors: Eloisa Raluy-López, Juan Pedro Montávez, and Pedro Jiménez-Guerrero

Summary and overall comments:

Raluy-Lopez et al. present a nicely designed set of experiments to test the sensitivity of ARs to different aerosol treatments with WRF-Chem. It is written well and clearly organized, and their results show that aerosols impact both the character and placement of ARs, thus highlighting the importance of aerosols in predicting and projecting ARs in weather and climate simulations. However, I have three main concerns that need to be addressed before this manuscript is ready for publication. 

1. The relationship between the dust, sea salt and the AR mechanisms needs to be more clearly and directly shown. The thickness diagnostics (and differences) for the composite “common” ARs are hard to interpret. I would recommend keeping these diagnostics for the case studies, but illustrating the connection between the ARs and aerosols cluster groups more directly, or in a more focused way. Ideas include using AR variables themselves,(IVT, IWV, or low level winds) for the “strengthening/weakening” component with the aerosols in lat/lon space, rather than box/whisker, and only showing the clusters that are significant. Or perhaps AR-spine centric averages vs aerosols cluster (highest density areas?)(and/or perhaps thickness) in scatter plots, to show this relationship. I think it is there, but at present it is a little unfocused. Also, significance needs to be shown in any difference plots. 

2. AR and ARDT uncertainty needs to be addressed. AIRA needs to be put into context of published ARDTs, and specifically, regional-specific algorithms that cover the Iberian Peninsula (e.g. IDL/Ramos, Lavers, Brands). Given the IDL code uses transects and also a Lagrarangian framework, this is the most similar type of code). ARTMIP (https://www.cgd.ucar.edu/projects/artmip/algorithms will have the reference list for the above mentioned ARDTs) has robustly shown that threshold choice is the largest source of AR metrics variability across ARDTs with dramatic differences in frequency, for example, depending on how this is chosen. See specific comments for details on suggestions on how to address this issue.  

3. Referencing needs to be improved and representative of the recent AR literature. 

Specific comments:

Line 21: In the midlatitudes, this is indeed the case, but not necessarily for high latitude ARs.  I recommend amending this statement with “in the midlatitudes”. 

Lines 24 and 25: There are many many references that could fit this statement, I recommend adding an “e.g.,” to your citation list, or add a few more references. 

Lines 28,32,34: Again, there are quite a few that could be listed here, so “e.g.” should be used. I am surprised not to see any Lavers references as this group was among the first to discuss North Atlantic ARs. 

Paragraph Line 36: I appreciate the author's discussion here, but there are some major gaps in the literature review. ARTMIP has had a number of workshops, plus 5 major group/overview papers, and many contributed papers. All discuss the issues of defining and detecting ARs, and the philosophy of using an ARDT (AR detection tool) that is appropriate for the science question asked. In addition to referencing the workshop report (or instead of), please read and cite the following papers. (Note: the climate change papers, O’Brien and Shields/Payne, would be good additions to the climate change literature review sentences, with the Rutz and Collow papers for reanalysis). 

Shields, C. A., Rutz, J. J., Leung, L.-Y., Ralph, F. M., Wehner, M., Kawzenuk, B., Lora, J. M., McClenny, E., Osborne, T., Payne, A. E., Ullrich, P., Gershunov, A., Goldenson, N., Guan, B., Qian, Y., Ramos, A. M., Sarangi, C., Sellars, S., Gorodetskaya, I., Kashinath, K., Kurlin, V., Mahoney, K., Muszynski, G., Pierce, R., Subramanian, A. C., Tome, R., Waliser, D., Walton, D., Wick, G., Wilson, A., Lavers, D., Prabhat, Collow, A., Krishnan, H., Magnusdottir, G., and Nguyen, P.: Atmospheric River Tracking Method Intercomparison Project (ARTMIP): project goals and experimental design, Geosci. Model Dev., 11, 2455-2474, https://doi.org/10.5194/gmd-11-2455-2018, 2018.

Rutz, J.J, Shields, C.A., Lora, J.M, Payne, A.E., Guan, B., Ullrich, P., O'Brien, T., Leung, L.-Y., Ralph, F.M., Wehner, M., Brands, S., Collow, A., Goldenson, N., Gorodetskaya, I., Griffith, H., Hagos, S., Kashinath, K., Kawzenuk, B., Krishnan, H., Kurlin, V., Lavers, D., Magnusdottir, G., Mahoney, K., McClenny, E., Muszynski, G., Nguyen, P.D., Prabhat, Qian, Y., Ramos, A.M., Sarangi, C., Sellars, S., Shulgina, T., Tome, R., Waliser, D., Walton, D., Wick, G., Wilson, A., Viale, M.: The Atmospheric River Tracking Method Intercomparison Project (ARTMIP): Quantifying Uncertainties in Atmospheric River Climatology, Journal of Geophysical Research-Atmospheres , https://doi.org/10.1029/2019JD030936, 2019.

O’Brien, Travis Allen and Wehner, Michael F and Payne, Ashley E. and Shields, Christine A and Rutz, Jonathan J. and Leung, L. Ruby and Ralph, F. Martin and Marquardt Collow, Allison B. and Guan, Bin and Lora, Juan Manuel and et al., (2022) Increases in Future AR Count and Size: Overview of the ARTMIP Tier 2 CMIP5/6 Experiment. JGR-A https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021JD036013.

Collow, A.B., Shields, C.A., Guan, B., Kim, S., Lora, J.M., McClenny, E.E., Nardi, K., Payne, A., Reid, K., Shearer, E. J. , Tome, R., Wille, J.D., Ramos, A.M., Gorodetskaya, I.V., Leung, L.R., O’Brien, T.A., Ralph, F.M., Rutz, J. Ullirich, P.A., Wehner, M., (2022) An Overview of ARTMIP’s Tier 2 Reanalysis Intercomparison: Uncertainty in the Detection of Atmospheric Rivers and their Associated Precipitation, Journal of Geophysical Research, Atmospheres, https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021JD036155.

Shields, C. A., Payne, A. E., Shearer, E. J., Wehner, M. F., O’Brien, T. A., Rutz, J. J., Leung, L.R., Ralph, F. M.,  Collow, A. B. M.,  Ullrich, P. A. Ullrich,  Dong, Q.,  Gershunov, A.,  Griffith, H.,  Guan, B.,  Lora, J. M., Lu, M.,  McClenny, E.,  Nardi, K. M.,  Pan, M.,  Qian, Y.,  Ramos, A. M. Ramos,  Shulgina, T.,  Viale, M.,  Sarangi, C., Tomé, R., Zarzycki, C. (2023). Future atmospheric rivers and impacts on precipitation: Overview of the ARTMIP Tier 2 high-resolution global warming experiment. Geophysical Research Letters, 50, e2022GL102091. https://doi.org/10.1029/2022GL102091

More details on ARTMIP here: https://www.cgd.ucar.edu/projects/artmip

Line 48: The statement that GCM’s “may not accurately represent their (AR) behavior” is a bit misleading. Most GCMs (and ESMs) are able to simulate the synoptics, bulk numbers, duration, etc. realistically. I recommend amending this statement specifically to AR-precipitation, given it is the precipitation piece that does better with high resolution (citations are needed here, there are quite a few out there now for high resolution global/earth system models, and ARs). 

Line 51: I am not sure I understand why a timeslice approach doesn’t work for limited area models?  Many timeslice ARDTs work well within a limited area domain (see the ARDT list on the ARTMIP webpage, some of these are both timeslice and regional). I agree with the authors that regional ARDTs tend to do a better job because localized considerations are made for regional-specific that would not otherwise be considered in globals (for example, for IP, the complex topography and the North Atlantic storm track climatology). If this is the intent of the authors, I recommend using this as motivation for the newly developed ARDT for the IP, rather than timeslice vs lagrangian approach. If I misunderstood, please make this statement more clear. 

Introduction general comment:  I am surprised there is no mention of the Calwater experiment.  Although this was focused on the western U.S., it was an important and groundbreaking study to look at aerosols with observations and AR. Here is a citation from CalWater that uses the same model as this study, i.e. WRF-Chem.

Naeger, A. R. (2018). Impact of dust aerosols on precipitation associated with atmospheric rivers using WRF-Chem simulations. Results in Physics, 10, 217-221, https://www.sciencedirect.com/science/article/pii/S2211379717318223

Paragraph at line 74: It might be useful to readers familiar with climate models, but not WRF forecast systems, to add a sentence or two explaining how lateral boundary conditions nudge the model back to the “observations”. This is important for when you describe your common ARs periods later, it makes sense to use common periods given each simulation is reproducing the same forecast period, but just with different aerosol treatments. If I am misunderstanding the design, please clarify. 

Line 88: Just checking how “online” is meant here, as an active coupled component and not stand-alone simulation? 

Line 108: I think this a Lagrangian approach, i.e. tracking rather than timeslice, given Figure 2?  I am not sure I understand why a regional ARDT can’t track an AR? This approach is similar to the IDL ARDT (an ARTMIP contributor, Ramos et al., 2016).  I think it would be helpful to add what aspects of AR science that AIRA addresses that the IDL does not. Or, how it compares to IDL, especially given both of these ARDT look at Iberian ARs. 

Ramos, A. M., Nieto, R., Tomé, R., Gimeno, L., Trigo, R. M., Liberato, M. L. R., and Lavers, D. A.: Atmospheric rivers moisture sources from a Lagrangian perspective, Earth Syst. Dynam., 7, 371–384, https://doi.org/10.5194/esd-7-371-2016, 2016

Line 134 and Paragraph at Line 185: From Table 1 and paragraph at Line 185, I think this is an absolute threshold, used for all simulations and does not change with the respective simulated climatologies? If so, please state that an absolute threshold is used for all simulations in the initial description, and point to the application for further explanation. 

Line 196:  Which ARDT catalogues/datasets were compared? The Brands ARDT contributions to ARTMIP are regional algorithms.  

Line 203: This is consistent with ARTMIP findings as October being the month with the maximum frequency for these latitudes (Rutz et al. 2019, Fig 13). 

Line 207: The mean intensity values are somewhat “baked in” to the values given the application of an absolute threshold.

Figure 5:  I noticed is that the AR metrics presented in this paper do not agree with other published results that look at aerosols, ARs, and climate, (Baek et al., 2021) where the Baek shows very little change over the Iberian Peninsula in the thermodynamic/precipitation and more of a change with the dynamics. There could be many reasons, including model resolution, aerosol treatment, ARDT, but this should be discussed or addressed in some way. 

Baek, S.H., Lora, J.M. Counterbalancing influences of aerosols and greenhouse gases on atmospheric rivers. Nat. Clim. Chang. 11, 958–965 (2021). https://doi-org.cuucar.idm.oclc.org/10.1038/s41558-021-01166-8

Line 238: I am not convinced that 80 AR clusters is enough to overcome natural variability, could you add some discussion on the robustness of only using 80? Have you considered playing with your threshold to increase your sample size?  Would the results be the same if you used a fixed-relative threshold, based on the “base” climatology? And/or a simple relative climatology unique to each of your experiments (base, ari, aric?) This would increase your sample size and also test uncertainty in your AR definition. (One thing that ARTMIP has shown is that the moisture threshold value is by far the biggest influence on AR frequency, and quite significantly so). 

Figure 6: I am not sure if this figure adds much to the manuscript as currently described. Their differences don’t seem significant by eye (?) How are they important? If they are not, then maybe omit this figure. 

Figure 8: Add an explanation for the box and whisker styled plots: mean, median, quantiles? What is the color scheme showing?  As clusters 2 and 3 are primarily discussed, perhaps only show these instead of all the clusters? It will be more focused. 

Figure 11: Same comment as Figure 8, as well as only showing the significant clusters.

Thickness field diagnostic :  Have you considered showing low level winds and/or IWV instead of the frontal boundaries via thickness field for these composite plots? I would think that IWV might be a better diagnostic to show ARs, given it is the moisture stream that makes the AR unique, and not all ARs are associated with the warm conveyor belt? To show strengthening/weakening of the thickness fields, the gradient value (i.e., anomalies ahead - behind the front might be more intuitive than the difference plots which are hard to interpret. I like the thickness plots for the case studies, which help to highlight the relationship between the AR and the strengthening/weakening of the frontal boundaries, but for the composites, they are hard to interpret. If difference plots are continued to be used, then significance should be added. 

Figures 8,11: There is a lot of information packed into these figures, but not alot of explanation in the text. Consider adding more description and inference with these figures to make your points. 

Figure 13: Contour labels need to be a bit bigger, it is hard to see them even after zooming in. 

Case studies: I really like the figures with the dust and IVT overlays as this shows the displacements of the ARs. I would recommend trying to do something similar with the composites to help illustrate your conclusions that the aerosol locations and magnitudes impact intensity and location of the ARs. The case studies show this, but the current figures  8-12 aren’t as convincing. 

Line 342: This was not explained or motivated convincingly and AR uncertainty (that is, the uncertainty in AR metrics due to ARDT alone) is not addressed in the manuscript. This should be done given that AR frequency is highly sensitive to thresholding values. Suggested ways to address this: (1) Uncertainty can be discussed in the text addressing the limitations of using one ARDT,  (2) For extra robustness and my recommendation, repeat the AR analysis by running the AIRA ARDT using different threshold values to both increase the sample size and attempt to bound ARDT uncertainty, (3) More work, but useful could be to compare AIRA with other ARTMIP ARDTs. Other ARDT catalogues for MERRA2 and ERA5 available, in addition to source data so AIRA could be run for a sample period for direct comparison. Data available at https://www.earthsystemgrid.org/dataset/ucar.cgd.artmip.html.  Comparing to other regional ARDTs such as the IDL (Ramos), or the Brands ARDTs are highly recommended, especially if there are plans to use AIRA for other applications, including climate change where more than one ARDT is typically needed (O’Brien et al., 2022). 

Review of the paper: Sensitivity of atmospheric rivers to aerosol treatment in regional climate simulations: Insights from the AIRA identification algorithm by Eloisa Raluy-López, Juan Pedro Montávez, and Pedro Jiménez-Guerrero 

This investigation aims to develop a robust approach to identify ARs (called AIRA algorithm) under the constraint of the regional domain provided by the area-limited simulations. By developing this approach, the authors are able to inspect the impact of online aerosols on this meteorological phenomenon. The key finding of this work is how aerosols are treated in simulations may influence both in ARs intensity and trajectory by radiative changes cooling or heating the air. 

General comments

The paper is well structured, and the algorithm adopted is easy to understand thanks to the illustrative figures. Although several approaches have been developed to identify ARs, their innovation relies on overcoming the RCMs' limitations where most of the runs are focused over land, and this precludes capturing the long way over the ocean. The success of this approach will allow the use of RCMs to provide more accurate precipitation amounts than GCMs and to perform less computationally costly simulations such as online aerosol runs to understand ARs mechanisms. Then, I found this work a valuable advance to analyze the impacts of the AR’s landfalling.

Under these arguments, I recommend accepting this work after addressing a minor revision detailed below. 

Introduction

In line 55, the authors mention the lack of research about the impact of aerosols on ARs but they did not discuss the challenges nor mention previous works such as Counterbalancing influences of aerosols and Greenhouse gases on atmospheric Rivers by Baek and Lora

Methods

How can the AIRA be sure that is detecting an AR and not the branch of a low system with a bigger enough radius? Does Δϴ < 25 guarantees this fact? Maybe introducing SLP values will avoid this concern. 

In Table 1 the authors show the imposed parameters. To demonstrate the robustness of the approach some discussions about the sensitivity of these parameters are needed. For instance, how many percentages of ARs increase/decrease if the IVT threshold is modified?

To better contextualize your methodology, I missed a discussion comparing the AIRA approach with other methodologies of other tracking approaches, For instance, a review can be found in: Atmospheric River Tracking Method Intercomparison Project (ARTMIP): Project Goals and Experimental Design by Ruth et al.

Results

Following the previous comment, some validation against observations (e.g. satellite images) and/or using the ARs inventory/catalogs is needed to be the coherence of your approach with the ARs already identified along the bibliography. 

In line 196 the authors mention.  “It was found that most of the ARs identified by AIRA also matched those identified by global-scale algorithms, as reported by Brands et al. (2017).”  How many coincidences did you find? Did you find more ‘real’ ARs in BASE or in ARCI?  Do you think that some discrepancies may be due to a different approach or the use of an RCM instead of a GCM?

In Line 224 the authors assert that the ARs explain the 30% of the precipitation, it is not clear what area did you use to obtain this value, and the Fig. 5 shows strong spatial variability to perform a spatial average. Furthermore, how accurate is the precipitation during these events? Is ARCI or BASE more representative of the observed precipitation?

In Line 232. Only 37 % of the coincidence of ARs between ARCI and BASE looks like a few percent. When the simulations are described there isn't any mention of nudging or re-initialization of initial conditions has been mentioned. What percentage of these discrepancies could be due to different treatments of aerosols or due to internal variability of the simulations?  

When sea salt and dust clusters are analyzed (Fig. 7 and 10) It will be interesting to see mean ARs trajectories for each cluster (for instance superimposed with dotted lines).

In the analysis of the differences to better understand the thermodynamics and dynamics changes, it will be illustrative to analyze whether the IVT changes are more due to  IWV or winds.

Throughout the work, I missed more analysis about the impacts of ARs on precipitation. I understand that may be the scope of future work. For the case studies will be interesting to show the spatial distributions of the precipitation (accumulated during the whole event and/or hourly) for the three simulations; BASE, ARI, and ARCI. These will provide some insights about how the intensity and trajectory of ARs impact on the precipitation distributions. Furthermore, the authors found around 30% of ARs impact precipitation but this percentage will have spatial and temporal variability. For instance, as ARs have an interannual variability also their impact on precipitation will be significant. Finally, it will be interesting a further understand the low impact on precipitation of the ARs over Galicia, Is it less frequency of ARs, more precipitation due to cold fronts, or orographic arguments?