Rev_gmd-2023-147

Please excuse this hurriedly written review.  I need to clear this today and it has taken me most of day plus more to finish this review.  So, please excuse typos or non-sequiturs – if there are comments that make no sense, feel free to ignore them.

This manuscript presents a careful and conscientious evaluation of the process of updating the EMEP MSC-W chemistry-transport model (CTM) with a new photolysis code:  Cloud-J v7.3e.  It represents an incredible amount of work and analysis, providing new metrics for assessing the photolysis module in a CTM.  From the model-measurement metrics shown here, the addition of Cloud-J is a notable upgrade to the MSC-W CTM and the new combination shows excellent performance characteristics that other CTMs might hope to achieve.

The manuscript is ideal for GMD and is certainly publishable.  I include a number of editorial suggestions below that would improve the readability as well as some other bigger thought ideas that the authors’ may choose to include or not.  There is more than enough work here for the reader to absorb and so I would not recommend expanding this work.

EDITORIAL COMMENTS

L1 – ‘chemical transport model’  Sorry to be pedantic, but our community has had a long-standing problem with adjective-noun combinations.  ‘chemical’ is an adjective and as such must be modifying ‘transport’, the following noun.  Thus our model is a ‘transport model’ that transports chemicals, where in reality it is also a ‘chemistry‘ model.  Hence, the correct terminology for CTM should be ‘chemistry-transport model’.   I take part responsibility for this mistaken acronym since the term ‘CTM’ was first coined by the Theoretical Studies and Modeling Working Group of the 1986 Global Tropospheric Chemistry Report (UCAR OIES Report 3).  We needed an acronym in parallel with GCM and named it Chemical Transport Model, which has been hard to live down – but it did catch on.

L9-11 – the 20% daily variability makes sense but the hydroxyl budget increase of 26% then seems large.  Try to keep bias errors or shifts separate from variances.

L11: to say that the OH budget ‘increased by 26%’ is totally confusing.  Did the P and L terms increase by 26%?  Did the mean OH increase by 26%?  ‘budget’ is a vague term here. 

L10 and later -  ‘aerosol radiative effect’  This get very confusing with regard to the climate system for which these models are also used.  ‘radiative’ would normally mean Watts per meter squared, but here you mean the change in photolysis fields.  Maybe ‘aerosol photolytic effect’.  Later you heavily use ‘aerosol direct effect’.  This term is used heavily in the chemistry & climate community to describe climate forcing, and is contrasted with the aerosol indirect effect (clouds) and semi-direct effect (clouds disappear when the air is warmer).  Please find another term, maybe something with ‘photolytic’ in it.

L14 -  ‘The bias is worsened…’  Not sure what you mean here.  Do not hint at results in the abstract, just give the main ones.

L17:  ‘emissions, meteorology, and solar radiation.  Photolysis …..’   List the big 3 first, then go on to solar.

L22:  ‘reflections’ =>  reflectivity

L22: scattering and absorption applies to air molecules as well as clouds and aerosols.

 L31: ‘often referred to as’ sounds stilted, maybe just (J-values) or i.e., J-values,

L37:  ‘eight different approaches for averaging over overlapping, broken cloud fields.”

L54:  ‘As noted above,’ – delete, unnecessary

L63: ‘3-hourly IFS O3 concentrations are now specified at the top…’  You need full O3 profiles to get the columns correct.  Later you talk about monthly mean O3 columns.  Is this used to get STE folds? Where did the IFS O3 come from? is it assimilated?

L83:  drop ‘However’, it does not help the reading of the sentence

L84: Please use ‘S’ or ‘N’ on all latitude ranges:  30N-90N (with the degrees, OK)

L84-85: If you mirror the latitudes, if there not a seasonal shift?  Also, tropical atmospheres are very different from mid-latitude atmospheres, so using 30N-40N to do the tropics is not particularly good.  Since this is a just a description of what the lookup tables DID, it should be left as is.  Nevertheless, it may explain why there is such a large difference in the global budgets (e.g. OH) with Cloud-J.

L98:  This could be explained better.  Use of wbin=12 provides equally accurate J-values in the troposphere as the full wbin=18 versions of Cloud-J (Prather, 2015) and saves 33%.  The wavelengths important in the troposphere are:  200-220 nm (upper tropical troposphere only) and > 291 nm.

L105: Correct but could be simpler:  Rayleigh scattering (gases = air) and absorption by O2 and O3 only, not NO2 or other gases that do not affect the radiative transfer solution.

L106: surface albedo – not directly relevant to this paper, but Cloud-J v7.6 has added an ocean surface albedo that is wavelength and angle dependent.

L119: ‘aerosol direct effect’ – as noted above, find a new name for this.

L140: ‘photolysis of O3 for wavelengths below 320 nm is an important loss mechanism for tropospheric O3,’ ONLY when followed by O(1D) + H2O reaction (<10% of O(1D)). Also the quantum yield for O1D is <1 everywhere.  Please correct.

L143:  ‘overhead stratospheric O3’ – This is only partially correct, please be more precise: i.e., the overhead O3 column above 100 hPa…    You must calculate the lowermost stratospheric O3 column at mid latitudes (100-200 hPa) with EMEP.

L147:  Now what really are you doing in EMEP CTM?  You said you take the assimilated IFS O3 as upper boundary at model top = 100 hPa = ~16 km.  Here you say you use a monthly climatology down to 10 km ??  100 hPa is not the tropopause poleward of 30 degrees.

L154: do you mean: ‘a multi-year monthly zonal mean climatology’  Please be more precise.

L156: this level of variability due to monthly zonal means in column ozone is an interesting finding. 

L158-164:  Nice idea to control the control the number of degrees of freedom.  Why not test the different CJ cloud parameterizations with box Chem?

L195: Very nice sensitivity study and analysis, this is helpful in understanding what is driving the difference

L208: This is an interesting result.   The tabulated Js must have very bad reference calculations for low sun angles.  Added point – in Cloud-J v7.6 (beyond current version here), we have added refraction in addition to spherical atmospheres, which further adds to J-values at low sun and past sunset (in the stratosphere).

L209: Table 1.  This table is great.  It is clear and shows some interesting diagnostics.  There are some obvious messages here about the difference between Tab-J and Cloud-J that indicate that Tab-J has some very old cross sections and quantum yields.

• J-O1D fell -25% with Cloud-J, that should to first order greatly reduce OH, since O(1D)+H2O is the primary source of OH and H2O is presumably the same. I suspect that the quantum yields are old values.
• >>>>However, the plot of J-O1d (Fig. 1) shows that TB is much smaller than CJ ?? which is correct?
• J-NO2 increased 25%, that is large, not sure what happened there
• J-HNO3 appears to increase 4x, probably not a bid deal since OH+HNO3 is major sink.
• J-HNO4 increases 13x, probably missing the 1 micron photolysis found by Wennberg.
• “P-dep  from 177-999 hPa” is correct, but probably add a footnote that this range in P is tied to a tropical lapse rate, so it is both P and T interpolation.

L209 Table 2:  ‘16^th of August  2015’ should be simply 16 August 2015.

L210:  It is nice to see the extension of ATom analysis (Hall 2018) to EMEP, along with more detail. Nicely done.  The use of 2015 cloud fields in addition to the 2016 data is very useful.

L227ff:  This analysis of J-O1D and J-NO2 is excellent and quite interesting, definitely a high-point of the paper.  One case I found very thought provoking was the doubled-vertical-resolution case (CJLVL).  For this case, did you redo the EMEP model levels to match the IFS? Can you specific the native IFS grid (e.g., we use IFS cy38r1 at T159L60 for our CTM)?  It looks like the only big difference between CJ and CJLVL is for J-NO2 N.Pacific, is that true?  That would be the biggest cloud-affected J.  Also, you found out that WRF has very different cloud fields since they have different cloud physics than the ECMWF model.  It further demonstrates that clouds are hard to assimilate and not just determined by u,v,q,T. 

Fig.1: It looks like CAFS (ATom) is using fundamentally different NO2 cross sections and quantum yields.  These are tabulated separately in JPL.  It is not an interp/extrapolation problem with T because it is the same over all altitudes.

L256: Did you check the T difference and if it could change the O1D ?  You should do that before speculating here since it is under your control.

L259: I agree that this is what the Fig. 1 shows, but it is opposite direction and smaller than shown in Table 1.

L265: It is good to see CJW & CJWH match each other so well in both clear and all sky.  I cannot understand why CJW&CJWH separate for J-NO2 in Fig.1.  Can you explain it?

L270 Fig.3:  You want to use CJB because of the lower cost, so you may want to do a more thorough evaluation of the errors in CJB (blue) vs CJ (purple).  Comparing with ATom here is very good, but you must also recognize that given the cloud distributions (and a non-3D RT solution), CJ is much closer to the correct answer. 

• CJB has consistently lower Js above 850 hPa and higher Js in the marine BL, esp. in the Tropics.

 L275-279:  Some caution here.  ATom cannot fly in the MBL when there are thick clouds present. – there must always be clear enough patches to guarantee VFR when entering and leaving the MLB.  Unfortunately in Hall 2018, we could not /did not filter the model MBL cumulus layers to remove those conditions (did not know how to).  Thus ATom=CAFS would have higher than average 950-hPa Js.  You should probably revise this text.

L305-307:  Very nice result!

L308:  Yes, but you know the Briegleb model is in error.

L313:  The comparison with surface observations is a great addition.

L344:  Wow, overall this comparison is great.  It is good to see the difference between the real J (with albedo) and the observation with a black cloth.  As a fair warning, this is not the correct way to compare with the observations.  the light that is scattered up and tthen down, comes from a much wider region than can be blacked out from the black cloth under the instrument.  When we tried such comparison with Fast-J (a long time ago), we had to do the calculation with an albedo of say 0.30 and then go into the RT solution and remove the direct upwelling (after the multiple scattering).  Sorry, but that option is not currently available in Cloud-J.

Fig 5 shows that EMEP-Ao (blue, the best attempt to match the observations) is indeed the best fit to the obs (orange).  What is unusual here is that the EMEP-TB (which is clearly noted to have a surface albedo and thus overestimate J-O1D) is 30+% higher in all cases, and the EMEP-CJ (with the surface albedo) is only about 10% higher.  If we go back to Fig.1, the TB Js are not that much higher in the BL for J-O1D.  Can you explain?

L365: Very interesting, this daily variability with O3 column.  Presumably it is the 100-200 hPa stratospheric column which at this location will vary because it is near the sub-tropical jet.  That region keep shifting from stratospheric to tropospheric.

Figure 6: These results are very good, considering that MEGRIDOP O3 is monthly and that EMEP will be simulating the LMS (100-200 hPa) ozone via the boundary conditions at 100 hPa.  Well done.

L395:  Correct, it is seen in most of our Cloud-J tests.  J-NO3 is most sensitive to clouds.

Fig.7:  Looks like the J-H2O2 problem is with the corss sections being used?

L412 Fig.8:  Wow, this looks so different from the other surface comparisons.  Part of the problem here is that we are comparing new Js that we have not used before.  Another part of the problem is that (I suspect) that all three surface obs. are not using the same cross sections and quantum yields to convert the irradiances to J-values.

L417: Good point about the O3 column – is there anyway to check it during the observations?  However, the J-CH2O’s both of them look pretty good.

L426:  J-CH3COCHO (MGLY) is a very messy J-value, it has strong pressure dependent quantum yields and I doubt that TUV and the Chilbolton group are using the same formulation as Cloud-J. 

L427: CJ has large visible wavelength bins, but the cross sections are averaged correctly at very high (<0.1 nm) resolution.  So this is not the problem, look at the quantum yields.  The acetone Js are another big problem with EMEP-Ao and EMEP-CJ greatly over predicting the ‘observed’ Js.

L432:  What other optical properties?  Clouds must dominate unless there was an haboob.

Fig.9.  OK, this figure is excellent.  It shows that the corr. coeff. maxes out at ~0.90 for the best Js.  It show the Js where we have clear problems with the physics being uses in CJ vs obs-J. 

• J-O1d has consistent bias, the obs-J probably has old q-yields.
• J-H2O2and J-CH3CHO have similar problems
• J-CH2O (both) and J-NO2 look fine
• J-MGLY clearly is using different spectral data in the tow J calculations.

L442: This is a very important but surprising result.  I am surprised that the correl.coeff. was not affected.  Maybe it points out that the high-resolution cloud statistics have the same skill and RMSE as the low-resolution averages.  There is a basic limit as to how well we can do cloud variability with the IFS (~ 0.9 corr. coeff.)

L458:  I think you mean R1 here.  Also the Crutzen and Zimmerman paper is a classic, but terribly out of date if you want numbers.  For current models (including yours?) the CO sink is 40-45%, not 70%.

L459: Montzka 2011 only measure the variability in decay of CH3CCl3, the OH variability is inferred.

L470: Being part of the ‘reference climatology’, I can only say that it is not observed and is from an old model.  It is valuable as a published ‘reference’ OH set, but should not be mistaken for ground truth.

L472-3: OK, so CJ gives higher than current model mean OH, but the J values match J-O1D surface observations (?)  Fig 10, however, does single out the problem, the CH4 lifetime.  Most of the CH4 loss is from the lower tropical atmosphere and CJ greatly increase OH there.  Yet, we know that EMEP-TB is a poor estimate of J-O1D in the tropics, even though we may like the answers it gives in terms of OH.

L478:  you cannot use ‘overestimated’ here since we do not know the correct value, only ’larger’

L479-480:  It cannot be H2O since both EMEP models use the same H2O density- right? How much is the difference in trop O3 column between EMEP-CJ and EMEP-TB?  You can also look at the ATom sensitivity of the CH4 and CO budgets to T, q, O3, NOx, CO, CH4, HOOH in Prather et al.

2023, Table 2, https://doi.org/10.5194/essd-15-3299-2023 Earth Syst. Sci. Data, 15, 3299–3349.

L482: Nice work on the aerosol photolytic effect with EMEP.  I am surprised by the size of the change in the biomass burning regions.

L513:  I cannot argue with the logic, but cannot see how CJ vs TB would underestimate the the surface O3 diurnal variations. Maybe you could compare the large-scale surface diel variations in O3 derived from many more site by Schnell et al Figure 1a-h (Atmos. Chem. Phys., 15, 10581–10596, 2015).

L513: OK, I have some problems with the estimate of computational costs (which can be found in the Cloud-J 2015 paper).  So doing Briegleb every 60 min instead of 15 min we have 15% extra cost.  You compare this with doing AvQCA every 15 min (250% cost).  OK not a fair comparison.  We only update J;s every hour, that should be fine. So, the AcQCA cost is now 60%.  We know that the average extra calculations for AcQCA is  2.8x, which would have the Briegleb being 20%.  OK, closs enough.  So the choice is Briegleb at 15% or AvQCA at 45% (probably).  I do not know how much the Table lookup costs, but the 15% is differential I presume.  The 45% may be too much, but you should not that the vertical distribution of Js with Briegleb is the same as different cloud fields as you show.  I think there will be some other changes in your budgets, but either way is a better representation of the clouds.

L533 Fig 12:  I am underwhelmed by this figure.  Not sure what to make of the scattered results.

L544 Fig 13: Really fascinating daily comparison, actually quite impressive EMEP modeling with any J values!!  Something wrong with Trinidad Head in the model.Tthe EMEP model is doing an excellent job of following the maximum O3 levels.  The shift in r and NMB from TB to CJ is noticeable, but the bigger issue is the met fields and emissions.  It still looks nice, but this plot is not the reason why one would choose CL over TB, I agree.

Final note.  You have taken the Cloud-J v7.3e which is a solid version of Cloud-J and includes all the cloud overlap options. There has been further development of Cloud-J since 2015 and you and others should look at what has been fixed or updated (like physically based ocean surface albedos).  I include the notes of the successive changes below.  In general, I recommend you adopt Cloud-J v8.0c (which is being implemented in GEOS-Chem now) is you are only interested in J-values but v7.6 if you want the option of calculating solar heating rates in addition to J-values.

Michael Prather

Cloud-J notes following Cloud-J version 7.3e

FJ6.4 to FJ6.8, big change in J-NO2:  +11% @ 200K to +15% @ 300K

Cloud-J version v7.6c notes (Jul 2019, Prather)

The last published version of Cloud-J was v7.3c (see Prather, 2015 GMD)

Updates with minor fixes have been released: 7.3d, 7.3e, 7.4d, see

   ftp site:  ftp://128.200.14.8/public/prather/Fast-J_&_Cloud-J

Also see doi:

   Prather, Michael; Hsu, Juno (2019), Solar-J and Cloud-J models version 7.6c,

   Dryad, Dataset, https://doi.org/10.7280/D1096P

This new version 7.6c goes along with the publication of the paper:

    A round Earth for climate models, Prather & Hsu, PNAS, 2019.

Cloud-J and Solar-J have many overlapping data sets and subroutines,

    but Solar-J is more complex and has special needs.

Major updates, many described in the 2019 publication include:

    Spherical hydrostatic corrections for mass and geometry (new option)

    Corrected calc. of deposition of direct beam (FLXD) so that scattered

        flux is now conserved to 2/e6, and the TOTAL incident is correct

    A refractive ray-tracing code SPHERE1R is available, and a new cleaner

         algorithm for SPHERE1N (straight rays) is included.  Also a

         flat-Earth version SPHERE1F is available.

    Major re-coding of the Feautrier lower boundary conditions to allow for

         angle-dependent albedos, specifically the ALBEDO is now specified

         for each wavelength at the 4 quad angles AND for the incident SZA.

    New Ocean Surface Albedo (OSA) module depends on angle, wind, chlorophyl

         based on Seferian++ code; XWRC corrected for <400 nm from 0 to 0.2

    New simpler way of interpolating TAU and F for inserted cloud layers

    Dropped adding mid-layer odd-points for J's to cut cost.

    All the data tables needed to iniitalize are in a subdirectory /tables

Three sample standalone drivers show: a range of solar zenith angles for

cloudy vs clear, the effect of stratospheric (including GeoMIP) aerosols,

and the range of different cloud-overlap parameterizations.

Note:  the ACLIM_FJX module in fjx_sub_mod.f90 had to be corrected for Solar-J

to do a smooth interpolation with latitude.  This corrected version is included

here as fjx_sub_mod-corr.f90. It should replace the original version, but

the standalone reference cases have NOT been redone, and the corrected

version shifts the atmospheric profiles in T and O3.

Cloud-J version 7.7 notes (Feb 2020, Prather)

Small cleanup from version 7.6 (see notes below), some typos and non-std characters removed

Variables in calling sequence expanded to get diagnostics out.

'fjx_cmn_mod.f90' modified slightly to include variables for running SJ w/LLNL & CLIRAD

v7.7  (02/2020) Final synch with Solar-J v7.7

        Corrects problem with MAX-RAN that was caused by MAX-COR fixes

        New calling sequence of FPs, added OD18

'cld_sub_mod.f90' for Cloud-J v7.7 (02/2020) - last change = fixes for MAXRAN

Revised in v7.7 (02/2020) fixed MAX-RAN (#0 & #3) set CLDCOR=0 if need be

     SUBROUTINE ICA_DIRECT -- not used, finally removed in v7.7

Cloud-J 8.0 (April 7, 2023, Prather)

CJ-8.0 is designed for J-values, rather than for solar heating rates (Solar-J)

although it does calculate heating rates for uv-vis using standard 18 wide-band wavelength

bins of the traditional Fast-JX codes (~180-778 nm).

Most of the parameters describing various Cloud-J options are set in

  1) the cmn block via file FJX_CMN_MOD.F90 (most of the dimension variables)

  2) the subroutine INIT_FJX via the file FJX_INIT_MOD.F90

The spectral datasets are set for read-in length of SX_=27 super-bin used for the RRTMG Solar-J calculations. The current usage in terms of radiative transfer calculations is W_ = S_ = 18 (uv-vis, not near-ir)

A lot of the Solar-J features have been removed or commented out.

Version 8.0 has been cleaned up with a number of unused variables and parameters removed.

Also the FJX_osa_mod.f90 (ocean surface albedo) has be updated to real*8 and cleaned.

There is no Solar-J v8.0, any users wanting Solar-J+RRMTG will need to merge the CJv8.0 with SJv7.6.  If users want to run Solar-J based on the old LLNL or CLIRAD codes, those subroutines are still included but not activated.  Further, the wide-band spectral data sets would need to change to match the fewer number of near-ir super-bins, contact mprather@uci.edu.

The current default is for pseudo-spherical refractive atmosphere and cloud quadrature atmospheres (CLDFLAG=7). The clouds have 6 vertical groupings (max overlap) that are 33% correlated with the blocks above/below.

This v8.0 code has been tested for the "TROP-ONLY" option with a reduced number of wavelength bins and J-values.

  NWBIN (set in sub INIT_FJX) = 18 by default, but set to 8 or 12 it produces J-values good for

  tropospheric-only chemistry.  J-values with NWBIN=12 are within <1% of the standard code up to 20 km. (SZA=0 test)

        The NWBIN=8 is not quite so good, but the worst case is -1% error in J-O2 up to 17 km; and -3% by 20 km..

The major reason for CJv8.0 is to include X_H2O cross sections for absorption only in uv-vis (290-350 nm) from Pei et al., 2019

    Pei, L., Min, Q., Du, Y., Wang, Z., Yin, B., Yang, K., et al. (2019). Water vapor

    near-UV absorption: Laboratory spectrum, field evidence, and atmospheric impacts.

    Journal of Geophysical Research: Atmospheres, 124, 14,310–14,324.  https://doi.org/

    10.1029/2019JD030724

    Revised CJ code to pass WWW = molec H2O/cm2 for each layer through to PHOTO_J

IMPACT:  Reduces J(O1D) by 11% in tropics @surface (cloud-free, SZA=0), 1/2 @ 3 km, 1/4 @ 6 km.

New H2O cross sections:

        H2O   !H2O UV absorpt  ! Pei++ 2019   (2023-01-17 MJP)

         300a 0.000E+00 0.000E+00 0.000E+00 0.000E+00 0.000E+00 0.000E+00 xH2O

         300b 0.000E+00 0.000E+00 0.000E+00 0.000E+00 2.967E-25 4.392E-25 xH2O

         300c 2.675E-25 5.763E-25 3.613E-25 2.662E-25 3.553E-27 0.000E+00 xH2O

Cloud-J version 8.0c notes (May 30, 2023, Prather)

After publication of CJ v8.0 code and examples, it was noted at Harvard (E. Lundgren, thank you) that the standalone driver incorrectly calculated the in-cloud Ice Water Path as

        IWP(L) = 1000.d0*WIC(L)*PDEL*G100 / CLF(L)    ! g/m2

whereas it should be in parallel with the Liquid Water Path (as it was in Cloud-J v7.5 and earlier)

        IWP(L) = 1000.d0*WIC(L)*PDEL*G100    ! g/m2

This is corrected in the new standalone driver CJ80c.f90.  This code also turn on clouds as average (CLDFLAG=2)

Fortunately, all the test using CJ80.f90 were done with CLDFLAG=1 (clear sky) and thus the error did not matter.  A new set of output files with CLDFLAG=2 (average clouds) without uv H2O  absorption is now included.

The basic differences between with and without xH2O are shown in the original output files for v80.

Readme.MD for Cloud-J version 8.0c

The files are included along with a notes file in the UCI_CloudJ80.zip file

    Located with updates at ftp://128.200.14.8/public/prather/Fast-J_&_Cloud-J/

It is also published with a Dryad and Zenodo doi

   Prather, Michael (2023), "An updated cloud-overlap photolysis module for atmospheric

   chemistry models, UCI Cloud-J v8.0, with near-UV H2O absorption"

   Dryad, https://doi.org/10.7280/D1Q398.

GENERAL COMMENTS

################

This article describes the update of photolysis rates in the EMEP model.

The effects of the incorporation of these updated rates are thoroughly studied

through comparisons with a wide set of observations and a former version of model which uses

older rates.  Detailed references are provided for experiments and methods used.

I think this article deserves  publication in GMD after some minor corrections and clarifications.

SPECIFIC COMMENTS

#################

* Section 2:

Line 67: the reference for ECLIPSEv6b should be given

Line 76: what definition of all(cloudy)/clear sky do you use ? It should be precised, 

         since the distinction is used throughout the article.

Lines 78-79 Could you give the cloud optical depth (for instance at 550 nm ) ?

Line 95-96: does the use of more streams changes the results ?

            Since one of the differences between Cloud-J and EMEP-TB lies in

            the number of streams, it would be interesting to know.

          

Line 96: could you precise whether it uses discrete ordinates method or not ?

Lines 95-100: do you use methods such as delta-M in the presence of highly anisotropic

              phase functions ? If no, have you made trials to show that is does not

              bring significant improvements ?

Lines 95-100: do you use methods to correct the intensity, for instance the one of

              Nakajima and Tanaka (JQSRT 1988)  ? 

Lines 111-115 a basic description (without dwelling into details) on 

              MAX-COR AvQCA and the Briegleb method would be interesting 

Lines 153-154 could you precise what do you mean by a "climatology based on these years" ?

              Do you extrapolate after 2021 ? If so, how ?

Paragraph 2.3: you do not mention what TOA (Top Of Atmosphere) SSI (Solar Spectral Irradiance) 

             you are using. Have you tried more than one to asses their influence

             on the photolysis rates computed by Cloud-J ? The SSI seems to me an important

             part of a model which computes photolysis rates.

* Section 3.

Lines 225-226: was the conjugate dataset constructed from TUV alone or also from CAFS ?

Line 250 : Could you define what the WASM scheme is ?

Line 254: what is the tropopause altitude ? Please specify or plot a vertical temperature profile.

Line 261: difference with what quantity ?

Line 261-262: What do you mean by "comparatively largest differences" ? 

                 I can see some discrepancies between the text and the figures

                   -  for J-NO2

                      according to figure 2,  CJB is closest from CAFS than CJVL above 800 hPa for both zones, 

                      in the text you say that it presents the largest differences for both zones.

                   -  for J-O1D

                      accord to figure 2, for tropical pacific, CJ15 is closest from CAFS than other cloud-J simulations,

                      in the text you say that it presents the largest differences for both zones.

                  

                  You also say that, for J-NO2, the largest differences occur for CJB for both zones (line 261),

                  and for CJVL in the the Northern Pacific zone. This is not consistent. 

 

Line 271: the altitude of the tropopause would be useful.

Line 291: "diminished" => "diminished by more than 2.5 %"

Line 291: "enhanced" => "enhanced by more than 2.5 %"

Line 294: the crosses are not always located on the thick part of the lines. 

Lines 289-298: for figure 4, what determines the horizontal position of the bars ?  The legend of the

               horizontal axe should be precised. Their length is always 100 %, but they do not have the same abscissas.

               Detailing an example would help the reader.

Section 4

#########            

Line 320-329: could you give more details about the cloud cover ? 

Line 326: "The moderate aerosol mass... by marine aerosols". I think you meant that, in mass, the main aerosol constituent

 was marine aerosols. 

Line 339-343 is it possible for Cloud-J to distinguish between the downward actinic flux and the upward one ?

               it would allow to take into account the portion of the upwelling flux redirected downwards.

  What was the albedo of the surface (not in the model !), if close to zero, it would avoid the problem.

Lines 342-343 I understand  that the photolysis rates used in the chemistry were computed using a non-zero albedo.

              If I'm right, it should be explicitly stated.

p18: figure 5 could be separated in two figures, one with the photolysis rates, the other with the ratios model_rate/observed_rate.

p19, figure 4: could you add a second vertical scale on the right to plot the simulated column below 100 hPa ?

Line 395: "the impact of clouds ... for both model and observation" Where do we see it  ? Were some days cloudy ?

p21, Figure 7, plotting also the ratio simulated/observed would be interesting and allow a more easy comparison

     between the models. 

p32, figure 9: could you provide a parameter to assess the quality of the fit, e.g. mean squared residual ?

* Section 5

-----------

Line 463: Over which period  was the reference climatology from Spivakovsky computed ? The intercomparison of Naik done ?

          

Line 472-477. Two references are used to asses the models : the Spivakovsky  climatology, and the Naik intercomparison. 

              Which one is the best ?

p26, figure 10: could this figure be replaced or completed by a table with the values ?

p26, section 5.2 and p27, figure 11 It would be interesting to run simulations taking into account the radiative contribution

     of only one type of aerosol (dust, sea salt, biomass burning aerosols), and to plot the results in figures

     similar to figure 11.

Line 509-510: I think that "decrease" in "increase" mean decrease/increase from the TB values.  Could you precise it ?

Line 521: the decrease of  performance does not seem obvious for the United Kingdom.

Paragraph 5.3.1

A map showing the positions of the four sites used would be useful.

Fig 13:  It would be more clear, to use, for each site, one vertical scale for the plot obs = f(day of year), and another for the ratios EMEP-CJ/obs and EMEP-TB/obs. The information about the variation over a year would be conserved, and the ratio simulation/observation would be more easy to see.

In fig 13, we see that the absolute values of the NMB is higher for CJ than for TB. This could be understood as a decrease

of the performances of CJ relatively to TB. It should be mentioned in the text and explained. 

* Section 6

------------

Line 551-552: the "the efficacy ... from the CAFS instrument". This is true for clear sky, but less for cloud sky, according 

to figure 2.

 

 Line 557:  - The general improvement is not obvious for global sites, see remark for paragraph 5.3.1., it should be precised.

            - For Europe, the number of site for which CJ improves the performance is greater than the number

              of sites for which it is decreased. "General improvement" is not the best way to sum it up.

Code on Zenodo

--------------

The description of the role of the different files should be more detailed, to

help the reader understand the global structure of the model.

In the python scripts:

- it would be useful to have comment to link the plot commands to the figures of the article.

- some paths should be defined as parameters, for instance, /home/willemvc/Desktop does not exist everywhere,

  the user should be able to change once and for all these paths (in a python module, for instance).

TECHNICAL CORRECTIONS

######################

p12

  Figure 1, a and b "Tr" => "Trop" or "Tropical", it would be more clear and is enough

            space for it.

  Figure 1, c and d, "N." => "North", for the same reason.

Line 519 : "absolute annual mean" => "absolute value of the mean"

Lines 555 - 559 : "the Cloud-J based simulations show ... below 350 nm" 

  Could you recall the sections of your article were these assertions are detailed ?