The two plots shows the relation between 70/100 flux density ratios as functions of the 250/500 and 250/350 flux density ratios. The galaxies are metallicity-coded. The lines are PACS/SPIRE colors deduced from single modified blackbodies of different temperatures for different emissivity index. Those plots indicate that the galaxies of Kingfish present a wide range a behaviour in the submm regime with a clear trend with metallicity. It also indictaes that introducing the 500 um data tend to flatten the slope (the galaxies move to lower emissivities). Feel free to use them if you think that relevant for your analysis. =================================================== 1- Introduction: End of the 1st paragraph: You can also cite the Meixner et al. 2010 Science Demonstration paper on the LMC and perhaps some papers of the Planck Collaboration: Results on nearby galaxies where they find evidence for colder dust than has previously been found (Planck Early Results: The Planck View of Nearby Galaxies), or in the Magellanic Clouds (Planck Early Results: Origin of the submm excess dust emission in the Magellanic Clouds). *** I already had a reference to the paper by Bernard et al. on the SMC, but it's hard to know I did since the first author is Ade ... I added the other "Ade" reference you suggested on nearby galaxies, and the one by Meixner. Thanks. 3rd paragraph: "do not directly take advantage of one of the key features to Herschel data" or "only indirectly profits". I find the tone of the two sentences a bit too negative compared to the huge amount of work you produced for this paper. After all, the paper takes advantage of the other (and perhaps main) advantage of Herschel: the submm coverage of the SEDs. *** I removed one of the two instances of "indirectly". Section 4.1: You mention in the 3rd paragraph that your applied color corrections to compare MIPS and PACS data. Did you apply those factors as well when performing the SED modelling? Sorry, I don't remember if it is mentionned in the paper. *** The color corrections are built into the SED modelling. They are taken into account when the IRAC-like and Herschel-like model fluxes are extracted for comparison to observations. Section 4.2: So SCUBA fluxes are not used in any of the fits or only in the galaxies mentionned in the paragraph? *** Like the end of Section 4.3 suggests, I didn't use them in any of the fits. There are relatively few KINGFISH galaxies with SCUBA data, and I find them less reliable than Herschel submm data. Section 4.3: - It would be interesting to test what happens to the parameter Umin if not fixed in the modelling. SPIRE bands should normally provide sufficient constraints to avoid the model to require too much cold dust. *** I took Bruce's and your advice and extended U_min down to a possible low of 0.01. Turns out none of the fits yielded U_min below 0.7. In short, no real change here. I added a note concerning this to Section 4.4. - Should we give a clue on how to translate the Umin parameter in term of temperature to make this parameter more understandable to a reader who is not familiar with the modelling? *** It depends on the grain size, as exemplified by Figure 4 of Draine & Li (2007). I added some text suggesting they can go to this figure to gain some understanding. - It would perhaps be nice to add a few words about how Bruce (Draine & Li 2007, Li & Draine 2001) has already modified his model to account for the FIR/submm emission observed for high latitude Milky Way emission (standard astronomical silicate emissivity modified for lambda>250um). Especially since it is the wavelength range at which we are looking at with SPIRE. *** Done Section 4.4: -If qPAH is dependent of the dust mass estimated, would it be better to plot instead: qPAH(Dale2011)*Mdust(Dale2011) / qPAH(Draine07)*Mdust(Draine07) ? *** That's an interesting suggestion. I made such a plot, but it looks very similar to the Mdust panel since most of the qPAH ratios are so close to unity. So I'll stick with the original version. - Is there anything to learn about plotting the fraction of dust heated by the quiescent ISRF to that heated in PDR regions as a function of metallicity? *** Another interesting idea. I made the plot and the distribution is flat as a function of oxygen abundance. One small, technical point: the plot I made is slightly different than you requested: I plotted (1-gamma)/gamma and not f(heated-by-Umin)/f(heated-by-PDR), i.e., I plotted the dust mass heated by Umin divided by the dust mass heated by Umin < U < Umax. But you asked for the dust mass heated by Umin divided by the dust mass heated by U>100. In other words, the discussion surrounding Eqn 29 of Draine & Li 2007 suggests that PDR heating is where U>100. - The fact that you are not finding a huge decrease of the dust mass when using Herschel data compared to what we showed in Galametz et al. 2011 is not surprising. It is hard to compare both approaches because in the model I used in the paper, we did not fix the Umin parameter, so the use of submm data was even more crucial for us and drastically decreases our dust masses. Bruce (in Draine et al 2007) fixed the Umin parameter to values higher than 0.7 especially to avoid an overestimation. Your result clearly shows that the assumption he made was the right one to adopt for most of the kingfish galaxies. *** Yes. In fact, even though I have allowed Umin to be as low as 0.01, the smallest fitted value using Spitzer+Herschel data is 0.7. Section 4.5: Magrini et al. 2011 found, for Virgo spirals, that their dust masses using the Draine et Li 2007 procedure is higher by only 10% compared to that found with a single modified blackbody using a power law with beta = 2, an emissivity ko = 0.192 m^2kg^-1 at lambda_o = 350um, so within their error bars. Could you check that the results you find are not strongly dependent on the assumptions you made for the single modified blackbody parameters? Especially when you let the emissivity index beta vary. Allowing beta to reach low values will clearly decrease the amount of dust mass required to account for the same emission and thus systematically lead to smaller dust masses. An effect that may perhaps be more significant than the lack of warm dust in the total dust mass budget. But that may be hard to quantify. *** Thanks to your prodding, I've taken a more careful look at this, including carrying out my DL07-BB comparisons using uncertainty-weighted values. Here's what I find for Mdust(BB)/Mdust(DL07): Variable beta, 100-500um: 0.59 Variable beta, 70-500um: 0.60 beta=1.5 , 100-500um: 0.59 beta=1.5 , 70-500um: 0.56 beta=2.0 , 100-500um: 0.52 beta=2.0 , 70-500um: 0.80 In short, the BB dust masses are 50-60% of DL07 dust masses, except for the case of beta=2 and using data as short as 70um. I briefly mention these results in the manuscript. I'm not sure what to say about the Magrini et al. result (I do reference them now). In their comparison, they don't fit DL07 to anything below 100um, and part of my point here is that when you fit data with DL07 shortward of 100um, you pick up contributions to the overall dust mass. Not most of the dust mass, of course, but an appreciable amount. Typo: "only Herschel photometric bands from 110 \um through 500 \um" --> 100 ? "are restricted to 110 \um" --> 100 ? *** Thanks Section 5: Typo: "These subtle differences in the fits presumably reflects the unprecedented" --> reflect? "these excesses are attributabe to increased" --> attributable *** Thanks