The study “A climatology of atmospheric rivers over Scandinavia and related precipitation” by Holmgren and Chen provides an analysis of atmospheric rivers (ARs) in Scandinavia. Here, the authors use four detection algorithms (three based on relative thresholds and one using an absolute threshold), with most of the results showing the ensemble median of these algorithms. Additionally, they investigate AR-related precipitation using ERA5 reanalysis, define different AR pathways by using a k-means clustering approach, and analyze the relationship between AR activity and the North Atlantic Oscillation (NAO) index.

Overall, it is an informative study that gives an overview of ARs and their associated precipitation over Scandinavia, particularly highlighting their strong influence on the west coast of Norway. The manuscript is generally well written; however, certain parts – especially Methodology 2.2 – would benefit from better explanations. For example, including a schematic related to the Jaccard index would greatly enhance clarity.

I would recommend this manuscript for publication after addressing the following points:

• For the detection of ARs, you have used the Reid500 detection algorithm. For me, the use of this algorithm is a bit surprising, as the absolute threshold of 500 kg m-1 s-1 seems to be high for this region. Could you explain why you chose this algorithm, or why you did not consider lowering the threshold?
• Section 2.2: You have used the Jaccard index to analyse the temporal relationship between AR objects. In Line 121, you refer to the “similarity” of AR objects. Could you clarify what kind of similarity you are referring to? Is it limited to the IVT amount, or are other variables, such as moisture, wind, or temperature, also considered? Furthermore, in Lines 127-130, it seems that some AR objects may be excluded (I’m not sure if I understand this correctly). If this is the case, would this not result in an underestimation of AR occurrence?

• Section 2.3: Here, you use the k-means clustering to define different AR pathways. The terminology in this section is a bit confusing. First, you use the term “clustering”, but then you switch to “pathways”. I recommend using “pathway” consistently throughout the text. It would also be helpful to include a table summarizing the cluster/pathways.

Minor comments:

L6-7: Clarify which regions within Scandinavia are most active

L7: Specify that the 40% precipitation refers to local values

L14: Which clusters? Do you mean pathways?

L23-24: Add a reference supporting that ARs are important for Scandinavia

L31-47: The overview of ARs in other regions is useful but could be condensed, since these regions are not compared further in the manuscript.

L79-84: I would structure the manuscript in 3 main sections: 2 Methods, 3 Results, 4 Summary.

L93: dived  divided

L93: absolute IVT threshold or a relative IVT threshold  absolute or relative IVT threshold

L181: Clarify that the precipitation values refer to local contributions

L181: Here, you refer to Scandinavia as a region, but in the next sentence, you write: “specifically, the regions with the highest AR-related precipitation”. Here, I would suggest specifying subregions more clearly. For example, the southeast coast of Norway and the west coast of Denmark are mostly affected by AR-related precipitation.

Figure 4: Do you have any idea why the spread is higher for smaller precipitation rates?

Figure 5: I see why you have used different color scales, but you have to be very careful when comparing them with each other. Maybe you could use a gradient colormap with more colors

Figure 5: It is interesting to see the high precipitation fraction in Northern Sweden. I would suggest seeing it more in cluster 3

Line246-257: The described seasonal pattern does not apply to Cluster 4, where ARs are less active in winter and more active in summer

L265: Consider rephrasing: Following autumn, ARs are less frequent during winter and reach their maximum in spring.

L270: For Cluster/pathway 4: SON looks quite similar to MAM. It seems that there is already a decrease in autumn compared to summer

SUMMARY

This manuscript presents an analysis of the typical frequency and precipitation impacts of atmospheric rivers (ARs) in Scandinavia. The authors find that from 1980-2019, ARs contributed 40% of the total annual precipitation while only occurring 5% of the time – elevating their role as a primary driver for extreme precipitation in the region. There is large spatial variability in the precipitation impacts of ARs in the region, which is closely related to four key pathways of ARs and their proximity to topography upon landfall. They also identified peaks in AR activity during strong positive phases of the North Atlantic Oscillation.

The study is novel, robust, interesting, and organized in a logical manner. It is well-written and advances our understanding of how the latitudinal positioning and orientation of ARs affects their precipitation impacts. I have included a few major comments and some minor comments, organized by line number, below.

MAJOR COMMENTS

I noticed that the data analysis spans 1980-2019. I am curious if there is a specific reason why the study is not extended to a more recent year, such as 2024. Is this a limitation of the available AR detection tools? I see it as advantageous to include as many years as possible in the data analysis, particularly recent years, as these have been marked by events that are more extreme than we typically observe in the 1980s and 1990s.

Is there a reason why you do not separate the total precipitation into rainfall and snowfall explicitly, as these can have different impacts on the surface and long-term implications for the hydrology of the region? (for example, L90).

For the NAO-AR pathway results, it may help readers to include a map of the typical atmospheric circulation during the positive and negative phase of the NAO (or include it as a supplementary figure).

MINOR COMMENTS

Title – “associated” might be more precise than “related”

6 – “present during up to 5% of the 6-hourly timesteps in the most active areas” is a bit confusing. Does this mean they are present 5% of the time? Combined with “the region receives” in the following sentence, it is unclear which parts of Scandinavia are referred to, and whether the following sentence is talking about the hotspots for AR activity or all of Scandinavia.

8 - “probability density is greater for the highest precipitation rates during AR events” > can you rephrase and say “ARs are disproportionately associated with the highest precipitation rates”?

Rest of the introduction (25-73) – I would recommend reworking this section to hone into the Scandinavian region and what is already known about ARs and precipitation more quickly. Currently, the text covers ARs and their precipitation impacts in a lot of different regions across the world, and I think this is probably more information than is needed to contextualize the study. Furthermore, I think it would motivate the results better to dig into the climatology of the region of study and familiarize readers with what is and isn’t known about precipitation and specifically AR-driven precipitation here. It is currently addressed on lines 69-73, but this is brief compared to the rest of the introduction.

79 – consider explicitly stating the names of the ARDTs used

88 – recommend mentioning the 0.25-degree grid spacing of ERA5 here

89 – “to make the computation feasible” > “to improve computational efficiency”

96 – recommend explain the method for each AR detection tool in more detail. For example, the exact thresholds are listed in the table, but how might that lead to differences among the AR detection tools? Which one is more or less restrictive?  

106 – do all of the AR catalogs have a requirement for continuity of the AR? As in, in some tools there can be a timestep where there are little bits of AR that show up as noise or are likely part of a larger AR but are disjoint. How does the 4-connected component labelling method handle these, if they are present?

111 – regarding the buffering, I think this approach makes a lot of sense! Wondering if there is a precedent for the method that you can reference, or if this is the first time this has been applied in this part of the world?

Fig1 – I find this map very helpful. I would recommend adding the names of the different regions mentioned in the study to the map.

144 – why do you think the number of optimal clusters was different for the four AR detection tools? What characteristic exerts the greatest influence in the clustering, and what does that suggest about the differences among the catalogs?

153 – do you believe that non-AR associated precipitation occurs during an AR event, in the footprint of the AR? Or is this because you’re summing all precipitation in Scandinavia during AR landfall anywhere in the region?

177-179 – consider moving this to the methods

180-185 – what is the standard deviation in the frequency and precipitation?

191 – add “frequency” > “annual AR frequency and precipitation”

206 – how did you decide on this cutoff for high precipitation amounts?

209 – consider converting to hourly or daily rates, which may be more comparable across different studies, as opposed to mm per 6h

Fig5 – would it make sense to put 5a,c,e, and g on a common colorbar (value range for the colors)? And same for b,d,f, and h? Currently the total precipitation fraction for 5f looks much larger than any of the other panels, purely because of the dominance of dark blue colors, even though in reality it has smaller values than the other panels.

Fig6 – similar to the previous comment – you could normalize these so that the colorbar is the same and the percentage of ARs by season sums to 1.

264-267 – could add these place names to Fig1

319 – “are typically lower” add “due to lower air temperatures”