|Hawker et al, 2021 Review|
In this study, the authors test the sensitivity of a mix of Eastern Atlantic Tropical cloud types to INP parametrizations with different temperature dependencies as well as the Hallett-Mossop process. They show that the slope (temperature dependence) of INP parametrizations has a significant impact on the TOA outgoing radiation for the clouds studied. Additionally, even in the presence of the Hallett-Mossop process, the INP parametrizations still produce significant results, indicating that correctly parameterizing the temperature-dependence of INPs is important to accurately simulate the radiative properties of clouds.
The manuscript is well written and the explanations are well thought through and explained nicely. The authors have also done a nice job responding to the previous reviewer comments. Nevertheless, I still think some points could be better explained as well as have a few additional comments listed below. The line and figure numbers correspond to the line numbers in the track changes version of the manuscript with the reviewer comments and author responses.
As previously mentioned, it is not immediately clear which hydrometeor classes are able to gain mass through riming and therefore produce secondary ice through the Hallett-Mossop process (HMP). Please clarify this as in several places, ice crystals are mentioned specifically but no other ice phase hydrometeor classes (snow and graupel) are mentioned. In response to a previous reviewer comment it is stated that ice crystals are meant to represent all cloud ice but for clarity it might be easier to replace ice crystals with ice phase hydrometeors or something more inclusive in the text when referring to the HMP. See specific line numbers below.
As previously mentioned by a reviewer, it is stressed that INPs are not scavenged in the text. But what does this actually mean? In the response, the authors state that as long as a threshold number of droplets is present and based on the ICNC already present and T, a freezing rate is calculated. This makes me wonder if the ICNC in a grid box is already above the predicted number of ice crystals from a parametrization due to settling (lifting) of ice from above (below), is no new ice formed? Please clarify this and also discuss what impacts this might have on the influence of the INP parametrizations used and the resulting TOA radiation if this is the case.
Similarly, a threshold number of cloud droplets is used. But what is this number? Is it large enough so that there are always enough cloud droplets such that the number of formed ice crystals does not exceed the number of cloud droplets? Please clarify this and discuss the impacts on the results if this is not the case.
“Hallett Mossop” should be changed to “Hallett-Mossop” throughout the text.
It is not always clear if the results are referring to the TOA changes in radiation over the entire daylight period (10 -17 UTC) or the entire simulation (10-24). Please make that clearer in the text and the figure captions rather than just mentioning it in the Methods.
I find the equations nice to explain the different calculations that have been conducted. However, there is no return to the acronyms used in the equations and therefore it is not immediately clear which values are calculated with which equation. Consider integrating the equation acronyms into the text and figures. Although, in general I find the text explanations quite clear, but also mentioning the acronyms, especially in the figures might make things clearer as how things were calculated.
It is stated in the manuscript that there was little sensitivity to the time frame (day versus day and night) considered for the impacts on TOA outgoing radiation. However, I find this quite surprising especially in light of the results in the convective anvils (high clouds). As at night when no shortwave radiation is reflected and the longwave is the only outgoing radiation, the cirrus anvil extent will likely have a large impact on the TOA outgoing radiation. Was this investigated and see comment below?
As I am unfamiliar with the SOCRATES radiation scheme, do the clouds ever become saturated in the amount of radiation that they can emit? I would think this would occur quiet quickly in the cirrus anvils.
The authors do a nice job of showing that the slope of the INP parameterization used influences the strength of the changes in TOA outgoing radiation. However, as the A13 parameterization has the steepest slope at temperatures above ~-26 C but also no slope at colder temperatures, how does that impact the argument that the slope of the parameterization is critical? Perhaps it is better to state that the number of INPs at cold temperatures is more important than at warmer temperatures in the MPC cloud regime?
It is mentioned clearly that more studies are needed to investigate what the impacts of adding INPs has over a longer period. However, is it possible to hypothesize on what impacts the increase in TOA outgoing radiation has on subsequent cloud development on subsequent days? More specifically would the reduction in surface temperature reduce the ability of convective clouds to form and ultimately over a long period offset any changes to the TOA outgoing radiation?
Line 170: Is the rime mass calculated for the snow hydrometeor class or only for the ice crystal class? Please clarify here as well as in the following comments on this.
Line 190: Here it is stated that the radiative cloud properties are not affected by changes in ice or snow number or any changes to rain and graupel. However, based on the following lines it sounds like the cloud radiative properties are sensitive to the mass of these hydrometeor classes. If that is the case, please specify that in this sentence as the mass is sensitive to the number/size and therefore this sentence is potentially misleading.
Line 221: Please clarify why the change in radiation from clear sky areas is only multiplied by the clear sky fraction of the sensitivity run (s) and not the change in the clear sky fraction from the sensitivity run and reference run (s-r), as is done for the cloudy sky fraction.
Line 246: Was cloud base always at temperatures above freezing i.e. was the melting layer always within cloud? If not then omitting rain alone may not be enough to ignore falling precipitation. Please clarify this here.
Line 257: please remove the additional “our” before “one of our”
Line 274: define DMT ie. Droplet Measurement Technologies and please provide the specifics on the CDP and CIP e.g. size range of measurements.
Line 281: should SODA2 also include a reference?
Line 290-291: Please clarify here that all ice-phase hydrometeors contribute to HMP.
Line 311-312: Same here.
Line 334: please add a “to” in “due a reduction”
Line 342-345: Here it is stated that the increase in snow and graupel production due to heterogeneous freezing increases sub-cloud evaporation of rain. However, when looking at Figure A5, the snow mass and graupel mass suggest that the A13 parametrization would lead to the largest amount of available melted mass to increase below cloud humidity. Yet, when looking at Figure A6, A13 has one of the lowest increases in sub-cloud humidity. How is this justified?
Line 352-360: In line with my general comment about the slope of the A13 parameterization, I generally agree that the A13 will produce less ice and therefore remove fewer cloud droplets at low levels in the cloud (below 5 km). However, the number of ice crystal produced at this low-level is quite insignificant to the number of cloud droplets. This can be clearly seen in Fig. 2 where at -6 C (Approx 5 km, assuming dry atmospheric lapse rate) the shallower sloped INP parameterizations (with significantly higher INP concentrations than A13 (~3 orders of magnitude)) still only predict < 1 INP per liter while in Fig. 6c, the concentration of cloud droplets is between 3 and 5 per cm3. Thus, the low number of cloud droplet activation due to ice nucleation, and following loss due to riming and depositional growth is likely insignificant to the number of cloud droplets observed at this height and that can be advected to higher regions of the cloud. Rather, is it more likely, that the increase in relative humidity of the low-level air mass due to evaporation of rain that is increased due to enhanced ice nucleation due to the higher INP concentrations of the steep parameterizations at higher levels of the MPC responsible for the observed changes in cloud droplet number? Additionally as HMP is rather unimportant on the CDNC for the steeper sloped parameterizations (N12 and A13) could it be argued that the higher INP concentrations predicted at colder temperatures are much more important as discussed below? If yes, then again, perhaps the slope of the INP parameterization is not as important as presented, especially as A13 is flat starting at -26 C.
Line 367-371: Again here it is mentioned that adding evaporating precipitation acts to increase humidity. But again A13 which has the highest snow and graupel mass, has lower humidities outside of cloud. Please clarify this. Additionally, when discussing the invigoration of updraft velocity due to enhanced latent heat release due to riming, it is clear that A13, which has the largest snow mass also has the highest updrafts. However, as the updraft increase is occurring primarily in a region where A13 predicts fewer INPs than the other parameterizations, is it really the slope of the INP parameterization at these temperatures that is important, or the concentration of INPs predicted by the parameterization at colder temperatures that settle to these altitudes?
Line 374: The idea of reducing homogeneous freezing by removing condensate via heterogeneous freezing has been well studied before in several geoengineering papers. Perhaps it is worthwhile citing some of these papers here (e.g Gasparini et al., 2020 and references within).
Line 390: There is no discussion about how anvils are classified in section 2.1.4, just how clouds are defined as high/mid or low. Also, it seems a bit misrepresentative to constrain anvil extent to only areas where no cloud is found below. I agree that the anvil (high cloud) is linked to the mid-level and low-level cloud below. However, the anvil should absolutely be considered when it extends over low-level clouds only. When looking at the example of the cloud classifications in Fig. 8, it looks like this does not really occur. Nevertheless, it might be more important at other times.
The previous point also raises the question as to why 20 UTC was selected as the example of the cloud classifications to the cloud field in Fig. 8, rather than 1330 or 1030 when the comparison between the satellites and the model were conducted? Especially since the 20 UTC does not fall within the daytime calculations.
Line 394-396: Such a drastic decrease in cirrus anvils between the noINP and INP simulations must have large impacts on the domain averaged outgoing longwave radiation at nighttime. Perhaps it is worthwhile showing this even if it is not the same amount of time that is compared, as mentioned in a response to one of the reviewers. That being said, with such a reduction in cirrus cloud fraction, does the addition of INPs makes the nighttime outgoing longwave radiation significantly higher than reported?
Line 396-399: Again it might be worthwhile to cite previous studies here that have investigated this as well e.g references in Gasparini et al, (2020).
Line 419: The review paper by Korolev and Leisner, (2020) could be included here.
Line 439-440: What effect does the formation of the “anvil” occurring at warmer temperatures have on the outgoing longwave radiation? This may be important as outgoing radiation is related to the temperature to the 4th power.
Figure 2: It is mentioned in the caption that the INP parametrizations shown are for an aerosol concentration of 8 cm-3. As three of these parameterizations are surface area dependent or at least size dependent (D10 e.g. aerosols > 500 nm) what sized particles are assumed here? Also, a concentration of 8 cm-3 seems to be an unfair comparison to the observations from the Welti et al, (2018) study as the ambient surface aerosol concentration were likely significantly higher when these INP measurements were conducted. Indeed the modelled and measured surface aerosol are approximately 2 orders of magnitude higher than at the 4 km level. This would put the parameterizations, with the exception of A13, significantly higher than the observations by Welti et al (2018). Or are you relying on the values of the modeled insoluble aerosol concentration to justify the comparison with the Welti et al, (2018) data? This should be expanded upon or at least mentioned. Additionally, if the values are normalized to surface area then this should also be mentioned. Furthermore, how was the data from Price et al, (2018) presented? Was it scaled to the surface area of the modelled insoluble aerosol or are these just the absolute concentrations per liter of air reported from the airborne samples? Also please fix the Welti et al, (2017) to (2018) in the figure legend. Lastly, I am not sure it makes sense to include the information that D10 was linearized for the correlation analysis in the figure caption here. Perhaps this is better suited in the text or in the first slope correlation plot.
Gasparini, B., McGraw, Z., Storelvmo, T. and Lohmann, U.: To what extent can cirrus cloud seeding counteract global warming?, Environ. Res. Lett., 15(5), 054002, https://doi.org/10.1088/1748-9326/ab71a3, 2020.
Korolev, A. and Leisner, T.: Review of experimental studies on secondary ice production, Atmospheric Chem. Phys., 1–42, https://doi.org/10.5194/acp-2020-537, 2020.