|Review of the paper by Panagiotopoulou et al.: "Comparison of PMCAMx aerosol optical depth predictions over Europe with AERONET and MODIS measurements" |
The paper reports the comparison of PMCAMx AOD predictions with AERONET and MODIS observations. The comparison is made for limited time (1 month) and presented only for monthly average values. Periods with high dust contributions have been screened out, so that the evaluation is expected to reflect mainly the model's ability to reproduce the organic and secondary inorganic aerosols.
In addition to the PM concentration, AOD depends also on the PM composition, size distribution, mixing state and interaction with atmospheric humidity. Thus, there are many possible reasons for the model measurement discrepancies and additional information is necessary to apportion the error. The paper heavily relies on a previous publication by Fountoukis et al. (2011), who compared the PMCAMx predictions of aerosol composition for the same period with ground based and airborne AMS observations. However, that comparison was fairly limited, including the data from only 4 stations for OA and secondary inorganic aerosols in PM1.
The paper is clearly written and the methodology is presented in an understandable way. The authors have adequately answered majority of the previous reviewer requests with a few exceptions:
I agree with the previous reviewers, that more extensive comparison with the observations of e.g. the EMEP network would give higher confidence in the models ability to reproduce the near-surface concentrations of the aerosol components in Europe. Currently no evaluation is presented for instance for the modelled sea salt and the paper does not discuss how it could influence the AOD predictions over the sea.
The temporal and spatial correlations of the model with the MODIS and AERONET observations should also be reported, as was requested by referee #2 - the current set of model scores does not provide the information on how well the model can reproduce the AOD patterns and variations.
The authors have presented a variety of model scores for a number of regions, however, they have not adequately discussed the discrepancies in the modelled and observed AOD patterns over Europe. What is the reason for underestimating the high AODs, while the low AODs are overestimated? How does the AOD underestimation in sea areas compare with PMCAMx ability to model sea salt?
Page 4, l 17 - change EUCAARI intensive to EUCAARI intensive campaign
Page 5, lines 9-12 "The major new methodological improvement in this effort is the screening of the satellite retrievals for periods with high dust (or coarse particles in general) concentrations as well as the combination of the MODIS and AERONET datasets so that the conclusions can be more robust." - Unclear, please restate
Page 6, lines 22-24. The authors say that they use gas phase anthropogenic emissions form GEMS inventory and carbonaceous aerosols from EUCAARI. However, the anthropogenic emissions also have a non-carbonaceous component, and based on Kuenen et al. (2014) this is not negligible for PM2.5 in Europe. Was the non-carbonaceous fraction of primary anthropogenic PM ignored in the model computations?
Page 7, line 6. IS4FIRES provides PM10 emissions. How was this speciated to PMCAMx species and sizes?
Page 8, lines 16-18. How is the aerosol size used in the Mie computations? Is the mean bin diameter used or is there an integration over the size bin?
Page 11, line 2 - What is the explanation, why only 0.4% of the MODIS AOD-s get discarded over water due to dust influence, while for AERONET this fraction was much higher and very similar to land?
Page 11, line 23 - Page 12, line 4. Replace the number of data points with the number of stations and flights.
How does the model bias at ground stations compare with the bias at higher altitudes and how this is expected to influence the modelled AOD?
The authors should discuss the discrepancies in the modelled and observed AOD patterns. Currently major differences are visible between the model and MODIS maps on Figure 3, which are not adequately addressed in the paper.
What is the reason for underestimating the high AODs in England, Northern Italy, Balkans region and southern part of the Eastern Europe region, while the low AODs are overestimated in Scandinavia, Northern Russia and Central Europe?
Does the AOD underestimation in Atlantic, Mediterranean and Black Sea reflect underestimations of sea salt?
The temporal and spatial correlations of the model with the MODIS and AERONET observations should also be reported, as was already requested by referee #2.
Figure 4 - if possible, redraw with colour scale which would clearly distinguish between the over- and underestimations and accurate predictions (e.g. white around zero)
Page 14, lines 16-17, Figure 5 - are the error envelopes computed for the monthly average values?
Page 16, lines 22-24. Fountoukis et al. (2011) reports about as large negative absolute bias for organic aerosol in Finokalia as for sulphates. That should also contribute to the AOD underestimation.
Page 18, line 14. As AOD at all other seas is also underestimated, could sea salt be underestimated in PMCAMx?
Page 18, lines 16-17. Model absolute bias is a better statistic than the fractional bias when explaining the AOD underestimation on monthly level.
Page 19, line 12. Please change to "Sulfates were the major fine PM components predicted in the Black Sea region during the simulation period."
Page 20, lines 12-14. When increasing the particle size, was the particle number kept constant? I would expect, that when mass concentrations would be kept constant, AOD would decrease noticeably.
Page 20, lines 15-18. Usually when discussing the BC mixing state, the BC core is assumed to get coated with a scattering layer, which scatters the solar radiation towards the absorbing core, thus increasing both the particle absorption and scattering. Neither of the cases presented here (internal volume averaged mixture and external mixture) take into account that option.
Section 7, Conclusions
Page 21, line 10. I would strongly discourage the authors from using the word "excellent" when describing their results, while discrepancies between the model predictions and observations are clearly visible on Figure 3 for majority of the regions they mention.
Kuenen, J.J.P., Visschedijk, a. J.H., Jozwicka, M., Denier van der Gon, H. a. C., 2014. TNO-MACC_II emission inventory: a multi-year (2003-2009) consistent high-resolution European emission inventory for air quality modelling. Atmos. Chem. Phys. Discuss. 14, 5837–5869. doi:10.5194/acpd-14-5837-2014