Comments on: Evaluation of downward and upward solar irradiances simulated by the Integrated Forecasting System of ECMWF using airborne observations above Arctic low-level clouds

Authors: Hanno Müller, André Ehrlich, Evelyn Jäkel, Johannes Röttenbacher, Benjamin Kirbus, Michael Schäfer, Robin Hogan, and Manfred Wendisch

Summary:  This paper provides an overview of comparisons between aircraft-based radiation measurements and output from the IFS global model.  The paper provides some interesting perspectives, though the overall ability to draw firm conclusions is limited by the sample size connected to this comparison.  Therefore, it is challenging to say that there are any significant break throughs that come because of this work, though the conclusions provide interesting perspectives on potential avenues for model improvement.  I believe that this work is suitable for ACP and that it should be considered for acceptance after some minor items have been accounted for.  These are outlined below.

Major Comments:

• Lines 233-235: I agree that the low values in the measurements are due in part to the gradient between dark ocean and bright clouds, but isn’t it also simply due to the broader range of downwelling irradiances sampled during these times with higher solar zenith angles?
• Section 4.2: It seems likely that some of the differences between model output and aircraft measurements are the direct result of spatial variability in the 2D plane — particularly the case when CF ranges from 0-1 over open water. The aircraft is only taking a single transect, which could readily see differences between 0-100% when the grid-box average is around 60%.  Please consider adding some discussion or sensitivity studies related to the 2D sampling volume that is being captured by a single line from the aircraft data.
• Line 301-302: This conclusion incorporates some assumptions about sun angles and the relative albedos of clouds and underlying ice.  Some discussion is likely warranted.
• Section 4.3.1: I actually consider ice water path to be relatively “macro” physical in nature. What if the IWP is fine, but the crystal shapes and sizes are incorrect?  How would this impact the radiation?  Please consider adding some discussion.
• Paragraph starting at line 330: This paragraph can be clarified a bit. If I understand correctly, there are two issues at hand, namely: 1) the fact that the LWP observation is only available above open ocean; and 2) The fact that the IFS does not provide an LWP quantity that can be readily compared, due to the fact that some clouds may be above the aircraft.  I would expand upon what was done a little to make it clear.  At the moment, it sounds like the integration of IFS LWC is meant to solve the “over-ocean-only” problem in the observations.  It took me three read-throughs to fully understand that this is in fact two separate issues.
• Paragraph starting on line 335: I would like to see more information on how these LWP values were calculated. For example, is this only for “cloudy-sky” times?  Or is this the integrated average over the 60 s window?  What about in the model?  Is this the “all sky” LWP?  Or the cloud LWP?  Assuming that it’s a grid-box average for cloudy conditions only, do you have information on the variability within each grid box?  To what extent can some of this simply be due to sampling across a 2D cloud field?
• Line 360-361: “This indicates that the adjustments of the IFS need to be different above sea ice compared to open ocean”: This is speculative and based on a very limited dataset that was collected under a very limited number of seasons and synoptic regimes. Also, what would be the physical explanation for this need of different adjustments?  Please consider adding more discussion here. 
• Lines 361-362: The authors discuss “improving” DeltaF and then discuss “reducing” DeltaF. It’s not clear to the reader how to interpret “improve” versus “reduces”.  Please stick with one theme (e.g., improves and worsens, or increases and decreases).
• Line 374: It is not clear to the reader why or how droplet concentrations and wind speeds are linked.
• Line 375: Are the in-situ cloud drop distribution data averaged over a grid box as well (~60 s)? Or are these distributions of high-resolution data?
• Lines 392-393: Are the physical mechanisms supporting differences between sea ice and open ocean clouds described in a separate publication? If so, please include a reference.  If not, please provide some background on what drives these differences.
• Lines 424-426: Regarding a parameterization that accounts for different sea ice types, It would have been interesting to evaluate the output of such a parameterization in the context of these runs. It seems like all of the inputs would be available, and then the outputs could be inserted into the offline simulations to assess this specific claim.
• Line 429-430: Cloud macro- and microphysical properties? Also, again please keep in mind that you have only a small sample size from which to evaluate the performance of the IFS cloud parameterizations.

Minor Comments:

• The authors sometimes use “the” excessively. For example, in lines 28-34 all of the bolded “the” occurences can be removed, in my opinion: “One obvious issue in the Arctic results from the sparse observational coverage, which limits the data assimilation (Bauer et al., 2016; Jung and Matsueda, 2016; Lawrence et al., 2019; Ortega et al., 2022). Furthermore, the modeling of the sea ice cover is a major obstacle for correctly representing the Arctic surface energy budget but is still uncertain due to the complexity of sea ice dynamics (Day et al., 2022). The representation of low-level Arctic clouds and especially mixed-phase clouds has been identified as another major source of uncertainty (Forbes and Ahlgrimm, 2014). As shown by Morrison and Pinto (2006), especially the cloud microphysical schemes cause uncertainties in the cloud phase and precipitation.”  I recommend that the authors re-read the paper and evaluate their use of this short word.
• Line 157: “no prognostic variables” should be “not prognostic variables”.
• Line 228: Recommend changing this to read “Observations of upwelling irradiance above sea ice surfaces…”
• Figure 6: It could just be my eyes playing tricks on me, but when I do some quick counting on the dots in this figure and the associated “occurrences”, my intuition says this is more than the 210 cases over water that were mentioned above.  Can you confirm that the totals represented by these dots sum up to 210?

The authors use ACLOUD field campaign data to quantify the representation and uncertainty of Arctic low-level clouds in the ECMWF Integrated Forecasting System (IFS). The model’s horizontal resolution is 1.4 km to 7.8 km. It has 137 vertical levels. To focus on low-level clouds, the authors only use cases with clouds below the level of Polar 5 and no cloud above. The comparisons of IFS and observed downward and upward irradiance are presented statistically to avoid the error due to temporal and spatial mismatch. To quantify the sensitivity of cloud properties, cloud fraction, ice water path, liquid water path, and cloud particle number concentration, as well as surface albedo are perturbed. Then a off-line radiative transfer code is used to assess the impact on the upward irradiance. The authors find the sea ice spectral albedo is the largest contributor to the bias in the upward irradiance above sea ice.

The analysis demonstrates the utility of field campaign data to evaluate high resolution models. This paper can potentially become a high impact paper. However, the use of statistical analysis in identifying cloud properties that are responsible for the bias is limited. In the end, the authors only use mean differences to evaluate the contribution. In addition, how the sensitivity study was performed needs to be clarified.

Major issue.

• Equations 3 and 4 show relationship among cloud properties. When the liquid water path is perturbed by keeping the shape of vertical profile, the liquid water contend is scaled by the ratio of liquid water paths. The scaling liquid water content affects effective radius, according to Equation 4. It is not clear to me, therefore, when LWP is perturbed, whether this is a partial derivative or other cloud properties change according to Equations 3 and 4. This probably affect the result of the number concentration perturbation most. The authors need to describe more to clarify how the sensitivity study was performed.

• The authors show distributions of irradiances in Figures. Also, Hellinger distances were computed. However, the authors do not discuss distribution differences. I would like to see discussions of distribution differences and how the distribution differences are used to narrow the uncertainty to identify cloud properties contributing the bias of upward irradiances.

Minor issues.

Figure 2: dotted lines used for open ocean make the plot hard to see. Could them change to solid lines? Also tick labels of x-axis for the upper plots are missing. Are these the same as upward irradiance? Also, including mean irradiance values to the upper left helps.

Figure 3: Are IFS sea ice albedo climatologies indicated by the solid lines?