General Comments

This paper presents 5 months of particles fluxes measured by eddy covariance at Ny Alesund, Svalbard. Deposition velocities as a function of particle size are derived for snow-covered, melting and snow-free conditions. In addition to the ultra-fine particle numbers obtained by a condensation particle counter, the 16 size channels of a optical particle counter are summed into 2 broader size classes in order to obtain workable statistics. An analysis is performed first on all fluxes regardless of direction, and downward fluxes are then analyzed separately to construct deposition velocity parameterizations for the three periods. Results are compared to previously published numbers and models and agree reasonably well.

There are some issues with this analysis that cannot be improved upon after the fact, such as the measurement site on top of a building that is likely to cause flow conditions less than ideal for eddy covariance, or the limited spectral response of the particle instrumentation. These should be discussed and justified in more detail than is currently the case. There are also a few issues that may warrant some additional thought and perhaps some further analysis. Simply concentrating on only positive (deposition) fluxes without accounting for natural scatter across the zero line is somewhat arbitrary and is likely to introduce artificial biases towards larger numbers; therefore, this should be justified in more detail. Some additional information, such as an ogive of particle fluxes and some statistics on the upward (emission) particle fluxes, either in the manuscript or a supplement, would also be helpful to convince the reader of the validity of the results.

Given the paucity of published direct measurements of particle fluxes, and especially in the Arctic, this paper can represent a useful addition to the field if these issues are addressed.

Specific Comments

Line 104: Please include a photo of the building and mast, and state the dimensions of the building (height, footprint) and the mast. Doing flux measurements on a rooftop can cause all kinds of problems due to flow distortion, and for some directions more than others, but a bit more on that later.

L150: If T comes from the sonic (T_s), it’s not the actual temperature, but rather a virtual temperature (or close to virtual, for all intents and purposes) which also depends on moisture. Therefore, what you are calculating is not strictly the sensible heat flux H, but a virtual sensible heat flux that is probably close enough to H for all practical purposes. But it’s best to be exact at the definition phase.

L155 and after: make the “i” a subscript

L157: here you define V_di, but later you only use V_ex. Make this consistent throughout. V_ex is more transparent and therefore a better choice.

Eq. (2): should include T_s rather than T

Section 2.4 could use some rereading and grammatical clean-up

L192: rephrase/clarify. There was no correction for large angles, but the small ones were corrected?

L194: please clarify.

L202: dependent on (not according to)

Fig. 2:

• those slopes in the inertial subrange are not even close to -7/3, except for the highest 3 points perhaps! I would temper that assertion in L257.
• In a), the particle flux signal drops by 5 orders of magnitude over the frequency range shown whereas the heat flux signal drops by only 3.5 orders of magnitude. Please discuss what this means in terms of underestimating fluxes associated with smaller eddies, and how that may affect the integrated total flux.
• Showing heat fluxes in b) is fine, but it would be more insightful to show the ogives for the particle fluxes as well. Obviously these will be significantly noisier, but showing that these look reasonable will go a long way towards convincing the reader that they behave properly.
• You could easily clean up the neutral ogive by widening the z/L range definition for “neutral”.

L288: “below 0C” would be more accurate than “about -10C”.

Fig. 3: wind directions look better when plotted using dots rather than lines

Fig. 5: averaging together the whole measurement period, consisting of vastly varying climatic and heat flux regimes, may not be the best way to do this. I would suggest splitting into at least 2 panels, one for snow-covered and one for snow-free, plus perhaps one for the transition period.

L357: according to the flux footprint estimate in Fig. 1, and the stated maximum footprint in L356, the ocean is well outside the footprint, and therefore it seems very unlikely that swell would be the cause of these upwards momentum fluxes. Could it be katabatic winds coming down Zeppelin Mountain? Or a very shallow sea breeze? Would need to look at wind directions carefully.

L349: these are very large roughness lengths for snow or tundra surfaces, an order of magnitude higher than expected, suggesting that the building is introducing a significant roughness element to the flow observed at the sonic.

L403: rephrase. The average value of F_ACC for the whole period was ….

L405: Finally, the mean of F_CRS …

L420: overlap with

L421: the village and shipping are all well outside the flux footprint (see comment to L357)

L439: simply throwing out all negative Vd’s and continuing the analysis with the rest will bias the readings (unless the histogram is strongly bimodal and there are hardly any numbers around zero), and including these would bring the averages down. This is an important enough point that it should be discussed in some detail, and the potential biases evaluated. They may be part of the reason for your numbers being higher than measurements by others and the Zhang and Pleim & Ran models for much of the range (Fig. 8).

Table 4: did you also try doing the statistics for V_d^–  ? Could be interesting, and lend some support to the validity of separating out all the positives into V_d ⁺  and simply ignoring negative values.

 Fig. 8d) snow free

L490: Seems unlikely that water surfaces have anything to do with this since a snow surface, with all its fractal roughness elements, would be a good surface for deposition. Seems more likely that the increasing z₀ as the snow melts has an effect, as you reasonably state in the conclusions [L 574].

L534 and conclusions: state clearly that the polynomial only applies to the size range 0.25 – 3 µm and does not cover UFP.

Donateo et al. present fluxes of particles in ultrafine, accumulation and quasi-course mode over an Arctic site. They use these data to evaluate models of size-dependent particle deposition models. This is a region that is in great need of measurements, and this manuscript offers useful data to the community. 

Major Comments

I do have concerns over the placement of the flux measurements. The manuscript indicates that the measurements were on a top of a building on a mast: photos would be useful, but more importantly, we need to undrstand the potential for the building to impact the turbulence. We need a good diagram of the inlet setup.

The particle losses are unclear: the authors suggest that because the D50 cutoff is 5 nm, they consider their system to detect particles between 5-1000nm. That is a strange assumption: the particle losses should have substantial size dependences, meaning that some particles in the size range of a given measurement will be far more effectively sampled - and as the paper shows, the size dependence of the fluxes can be substantial even within a single mode. A figure showing the size-dependent losses, and a quantitative discussion of how those losses might introduce biases to the analysis are essential.

Further, an important piece of QA/QC for eddy covariance fluxes is the cospectrum of the sonic anemometer data. The data shown in Figure 2 are disconcerting, as the suggest that the sensible heat flux data do not follow the expected decay. The authors state similarity to a -7/3 slope, but the plot is on a log-log curve. What actual slope is the data (i.e. -x/3?). This suggests a substantial loss of flux or disruption of the turbulence regime. You are most definitely not in the inertial subrange. I think the authors must add a substantial paragraph acknowledging these problems and considering how they would bias the data. 

I generally disagree with the approach of fitting the Vdep/u* versus size range to a polynomial as this is driven purely by the data, and not at all by the underlying process. I think this needs to be made exceptionally clear that this is purely a fit, not a parameterization - and that it only applies to the accumulation / quasi-coarse mode (that should be noted in the Abstract as well).

Overall, this manuscript includes unique data and is well-written. I do hold concerns over hte measurement site, and request that the authors provide substantial additional details and consider the implications of not measuring fluxes in the inertial subrange. Further, the authors need to be more careful over their discussion of polynomial fits. However, the data are intriguing, and warrant publication after revision in this journal.

Minor Comments

- The Introduction is well-written, but quite difficult to read as the bulk of it is in a single paragraph. I encourage the authors to consider breaking the main paragraph into about three paragraphs, each addressing the different aspects of the motivation and background