|Review of the manuscript “Aerosol optical, microphysical and radiative properties at regional background insular sites in the western Mediterranean”, by Sicard et al. The paper deals with aerosol optical and microphysical properties derived from AERONET observations at three sites located in the western Mediterranean. It is well written and structured. However a major concern arises from the fact that many conclusions are derived from scarce data, especially the optical properties derived from inversions. I recommend the paper can be published after major revision. General and specific comments follow here.|
1. The paper extracts conclusions in a climatic perspective, indicating that they use a long-term database. In principle 8 years of data are available at Ersa and 5 years at Palma de Mallorca. However the number of level 2 inversion data is too low for a climatological analysis. The AOD440>0.4 threshold removes most of the inversion data. In this frame, the authors are no longer analyzing the aerosol properties of the sites, but only the high turbidity events (dust or pollution). This should be clarified and the discussion re-focused accordingly.
2. The analysis of AOD has in principle sufficient data for long-term analysis, even if some caution should be taken because the datasets do not comprise the standard 30-years to be considered climatological in a strict sense. However in the analysis of Gobbi plots (Angstrom exponent difference versus Angstrom exponent) the data with AOD675<0.15 are removed. The authors do not specify the percentage of data that are ignored because of this threshold, but from average values in Table 1, I can presume is clearly above 60% of observations. Therefore the analysis presented in the paper is actually ignoring the predominant aerosol conditions at both sites, which will typically consist of few polluted marine aerosol. IN the title you have the work “background”, but the background conditions are not investigated. It is critical that those data are part of the analysis and the conclusions. A simple graphical method (Angstrom exponent versus AOD, as can be seen in Holben et al., JGR 2001) allows aerosol typing without the need of removing low AOD cases. Once this is established, the Gobbi plot can complement the investigation with some further insight in fine mode fraction, etc. for the cases with AOD675>0.15.
3. The sampling of AOD and inversion data within AERONET strongly depend on cloudiness and solar declination. Therefore the aggregation into monthly, yearly or seasonal averages cannot be accomplished without taking this issue into account. The normal procedure (see for instance the AERONET website) is to produce daily averages, and from them compute monthly means, seasonal means or multi-annual monthly means. If this is not done in this way, the much larger number of observations in summer (longer day duration and less cloudiness) produces a bias in the dataset averages. See for instance how similar are the year and the summer mean size distributions in Fig 5. In the case of inversion products, it is critical to produce daily means and from them produce seasonal means. The yearly mean should be the mean of the 4 seasonal averages. And so on.
4. Following comment nr 3 above, the number of data indicated in all plots should be no longer the number of single observation points but the number of days. This is valid for most of the figures. In this sense, the analyzed inversion data in figure 6 would be even less, but that’s the real situation using level 2 inversions and that’s the reason of comment 1: if you restrict to AOD440>0.4, you are investigating just few cases, but ignoring the predominant aerosol conditions.
Page 5, line 6: >2 years is not that long-term. Please reformulate.
P5, L17: if you are asking for long-term analysis, why including the short term data at Alboran? It changes the focus of the paper: you cannot any longer extract “climatological” conclusions of the gradients, only about autumn 2011. How do you justify this?
P6, L15: reference to Dubovik et al., 2006 (the spheroid retrieval for dust) is missing
P13, L15: this is the reason why short periods of data should not be analyzed in a climatic perspective (see general comments)
P14, L22: note the contradiction here: you are talking about marine aerosol here, that according to the cited paper by Smirnov et al. (2002) is found for AOD440<0.15. However all those low AOD data are removed in the Gobbi-type analysis with the condition AOD675>0.15!
P15, L15: there are not enough level 2 inversion data to establish seasonal characteristics, which wouldn’t be in any case representative for the typical site conditions
P15, L28: you can’t compute such average for the dust events because you are only looking at the subset with AOD400>0.4. There exist weaker dust events (actually they are more frequent) and the mean AOD during dust events is not 0.47.
P17, L31: the Mie(spherical) computations are very unlikely to be the reason because AERONET version 2 inversion uses spheroids and normally the portion of spherical particles considered in dust events is very close to 0.
P24, L14: you are supposedly using satellite data to validate AERONET fluxes but in the end you use AERONET data to screen out satellite data contaminated by clouds or glint. So it’s a circular argument. If you select only data that are in agreement, then the agreement is good (Fig 9b). I don’t think this is a valid approach.
Fig 11d,e,f,etc.: replace “month” with “2011”.