Review of "FLEXPART version 11: Improved accuracy, efficiency, and flexibility"

GENERAL COMMENTS

This manuscript presents a new version of the Lagrangian particle dispersion model Flexpart. The main changes for this new version are described and relevant examples are given for several aspects of the model. The new version is compared with the previous version in terms of computational performance and accuracy by using observations obtained from two tracer experiments. As the Flexpart model is widely used by many researchers for several applications, the manuscript is of high interest to the community. The manuscript is well written and figures and tables are neat.  The methods and assumptions are clearly described or properly referenced.

SPECIFIC COMMENTS

l 90: Could the authors add instructions/hints to install Flexpart-11 and its dependencies (or refer to the online documentation)?

l 95: “FLEXPART calculates particle trajectories using interpolated meteorological fields”: the interpolation is performed in time and space?

l 101: “All units are in International System of Units (SI) units, unless otherwise specified.”: the context for this sentence is not very clear. Does it relate to the input meteorological data?

l 149: “In addition, to avoid regions with low Coriolis force…”: could the authors provide a motivation for avoiding regions with low Coriolis force?

l 157: “N is the total number of particles in the sample”: sample refers here to all particles in the domain between [-80, -40] and [40, 80] latitude?

Figure 1: can the authors provide an explanation why the improvement in the stratosphere is so large for potential temperature compared to the improvement for potential vorticity and specific humidity?

Figure 2: the authors provide a number of possible causes for deviations from the well-mixed criterion. For |lat|>66° and between the surface up to +-1.5 km, the eta coordinates seem to result in a worse performance in terms of well-mixedness than the z coordinates. Do the authors have any possible explanation for this specific behavior?

Figure 3:

- there is only one sentence of discussion for Figure 3, so I suggest to elaborate the discussion a bit or consider omitting the figure.

- the results shown in Figure 3 represent total deposition, that is, the combined effect of dry and wet deposition. Wet deposition will diminish the relative difference between the total deposition of spheres versus fibers? Therefore, it would be useful for the reader to have an idea of the amount of precipitation in January 2018 (that is, was it particularly dry or wet in that period?).

l 251: “Experiments show that non-spherical particles experience a larger drag in the atmosphere and therefore have lower settling velocities than spheres (Tatsii et al., 2024).”: this is assuming identical particle volume and particle mass?

l 273: “PLA=2 PIA=2 PSA”: do the authors mean “PLA=2, PIA=2 * PSA”?

Figure 4: Comparison of panel (a) and (c) seems to suggest that in panel (c) wet deposition occurs even if there would be no rain in the meteorological data. Could the authors comment on this?

Subsection 5.2: it is not clear to me whether the underlying problem with the interpolation relates to time, space or both. In the text, the focus seems to be on the temporal interpolation, while I thought the problem was more related to the fact that precipitation represents a grid box average value rather than a point value? In addition, could the authors briefly mention the two meteorological fields (large scale precipitation and convective precipitation?) that are used when numpf = 3?

Subsection 5.3: the fix for the presence of clouds in case of convective precipitation seems quite arbitrary? Could the authors give an idea of the sensitivity of the choices and whether they think this is a large source of model uncertainty or not? Lastly, is there a particular reason for starting the cloud at ground for convective precipitation above 0.1 mm/h?

Figure 5: what is “sum” in the figure?

Subsection 5.4: could the authors provide motivation for aiming for an aerosol decay time of 9.3 d, close to the model median, rather than that based on the observations, which was 14.3 d?

l 368: “The mixing ratios for each species are calculated using the ratio of the mass of the species over the mass of air, where the mass of air is always carried by species number 1.”: does this mean that the mass of air attributed to a particle changes during the simulation? If so, what processes in Flexpart modify the mass of air attributed to a particle?

l 713: “There also exists a FLEXPART version for very-high-resolution simulations (1 km), where turbulence parameterisations have been adapted to account for the fact that turbulence is partly already resolved by the meteorological input data (Katharopoulos et al., 2022).”: this seems quite similar to Flexpart-AROME?

TECHNICAL CORRECTIONS

l 110: “The ECMWF’s IFS employs a hybrid pressure-base vertical coordinates”: omit “a”

l 193: “byCassiani”: add space

Figure 3: in the caption: “top” should be “bottom”

Figure 4: title in panel (c): consider writing numpf3 rather than npf3 to make it consistent with the text.

Table 1: “Number gives the total…”: should be “n gives the total”.

l 413: “FA5” should be “FMS”.

Figure 8: the figure size should be smaller.

Table 2: In the caption, move the sentences “Weak scaling is defined by…” and “The serial comparison is done by …” to the text.

l 511: omit “solid black lines” (should be in the figure caption or legend).

l 512: omit “dashed black lines” (should be in the figure caption or legend).

l 581: I suggest to omit “will”.

Figure 9: labels (a), (b) and (c) are missing.

l 641: Should be “Appendix A”.

l 793: omit the second “of”.

l 858: omit “a of”.

Review of “FLEXPART version 11: Improved accuracy, efficiency, and flexibility”

The authors highlighted some recent changes in the FLEXPART model since the last similar publication for FLEXPART version 10.4 in 2019.  The detailed descriptions of the changes are quite useful not only for the model users but also for other model developers.  While the presentation is mostly clear, minor revisions are still needed.  General and specific comments, as well as some editorial corrections are listed below.

General comments:

The improved accuracy of the new version is not convincingly demonstrated from the examples in the manuscript.  Although the semi-conserved property tests in section 3.2 show clear improvement of the new version, the tested properties are not exactly conserved as the authors also pointed out.  The statistic results listed in Section 7 with tracer experiments show very marginal differences between the new and old versions.  While Figure 6 compares the new model results with the ETEX measurements, it will be helpful to show the concentration fields predicted with the old version (or the difference between v 11 and v10.4).  It will be beneficial to compare the vertical profiles of CAPTEX aircraft measurements and the predictions with both FLEXPART versions.

There is no doubt that the OpenMP parallelization implementation is important for the FLEXPART users, but many details of the technical aspects are probably not needed for a scientific paper.  This reviewer suggest moving some of the contents in Section 8 to a supplementary material.

Specific:

Lines 24-25, "... they can take into account all processes occurring during transport including nonlinear atmospheric chemistry":  This gives an impression that Lagrangian models can not account for such processes.  However, it is not true.  Although it is convenient to use the Eulerian methods for such processes, the statement could be misleading.

Line 48, “FLEXPART combines a unique set of capabilities no other model can offer, …”:  Other Lagrangian 3D particle models have most if not all the capabilities listed here.  Thus it is not accurate.  Please remove “no other model can offer”.

Lines 149-150, “In addition, to avoid regions with low Coriolis force, we only used particles at latitudes north of 40◦N and south of -40◦S”:  Why?  Can this be elaborated?

Lines 411-412, “In fact, with the exception of FA5 and FOEX, all statistical values are slightly better for FLEXPART 11 than for FLEXPART 10”: FLEXPART 11 using the z coordinate system actually has a better FA5 than FLEXPART 10.4.  In addition, it is better to differentiate FLEXPART 10 and FLEXPART 10.4.

Line 425-426, “We also see no systematic large differences between FLEXPART 10.4 and FLEXPART 11, except for the NMSE values which again are better for FLEXPART 11”:  The second part of the statement is not true. 

Line 421, “In table 1 we list the average and medians of …”: What are actually listed in Table 1?

Line 445: Please clarify what “convection computations” mean here.  Does that include the horizontal transport?

Line 559, “For the largest, this …”:  What does the largest refer to?

Line 610, “part_i.nc  in the output directory”: Should it be the input directory?

Line 706, “…, whereas NCEP-based input comprises only pressure-level fields”: Current GFS model has a hybrid sigma-pressure vertical coordinate.

Line 776, 0.002 km^{-1}: Should the unit be K km^{-1}?

Table A1, “E-ward & N-ward turbulent surface stress row”: The unit of turbulent surface stress should be “N m^(-2) s” instead of “N m^2 s”.

Table A2: Please explain what α and β are.

Table A3: Units are needed for some of the parameters such as T1/2.  In addition, it is better to have “1/2” as a subscript.

Editorial:

List of affiliations are not in order.

Line 193: “byCassiani”-> “by Cassiani”

Line 471,” Replace “printed” with “written” or “recorded”.

Line 735: Remove “to” after “making use of”.

Line 762, “turbulent modtion”: Please correct the typo.

Table A1: What is “etadot” in the Vertical velocity unit for IFS?

Please see comments in attached pdf file