Dear referee and editor,

We appreciate your feedback and comments, which definitely helped us to
improve the manuscript.  It took us almost 6 months to reply because we had
to invest a lot of time and effort into the web-site re-design and
improvement of the web services.  Following the collegial agreement between
all co-authors, we invited our undergraduate student Kirill Grishin as a
co-author of the paper, because of his valuable contribution to the data
access services and improvement of the project presentation.

We also deeply investigated and validated our non-parametric emission line
fitting technique, which was the main point of criticism of the reviewer. 
It was done in connection to our project of AGN studies based on RCSED.  We
did not include all technical results in this manuscript because it is
already very long, however an executive summary was added to the following
section of the paper.

We have answered most of the reviewer's comments and questions. All the
changes in the manuscript are highlighted in blue.

> Reviewer's Comments:
> This is a referee report for the paper "RCSED - A value-added
> reference catalog of spectral energy distributions of 800,299 galaxies
> in 11 ultraviolet, optical, and near-infrared bands: morphologies,
> colors, ionized gas and stellar populations properties" submitted for
> publication on ApJ by I. Chilingarian and colleagues.
> 
> This paper presents a new value-added catalogue for the Sloan Digital
> Sky Survey which brings new gas-phase metallicity estimates for
> star-forming galaxies, provides new measurements for the
> stellar-population age and metallicity and - most important - adds an
> impressive suite of total and central broad-band measurements from the
> UV to the NIR domain to the existing SDSS optical measurements. This
> catalogue has also the advantage of being easily accessible and
> Virtual-Observatory complaint. All these aspects are going to be very
> useful to the astronomical community and thus make this paper
> certainly worth publishing.
> 
> There are a number of issues with this catalogue and associated paper
> that need to be addressed before this can happen, however, and in
> particular as regards their non-parametric approach for dealing with
> nebular emission. These are described below in their order of
> appearance in the paper.
> 
> Please accept my apologies for the delay in writing this initial report.
> 
> 
> - Abstract
> 
> Towards the end of the abstract, the authors indicate that the paper
> will include a discuss existing and future scientific application of
> their catalogue. I could not find a section specifically dedicated to
> this, except for some sparse reference to past work by the authors in
> the second-last paragraphs of the introduction (where some reference
> are provided) and the summary (where none are). Please either
> explicitly add a section (perhaps before the summary) that delivers
> what promised in the abstract, or remove that sentence from it.

We have added a couple of paragraphs in the Summary section  (# 6)
describing the existing results obtained with RCSED.

> 
> - Introduction
> 
> Here the authors must explicitly acknowledge the existence of the
> MPA-JHU and the Oh et al. (2011, also known as OSSY) value-added
> catalogues. Indeed, although based only on SDSS data only, these have
> provided for several years most of the spectroscopy measurements
> presented here, including the stellar and gas kinematics,
> emission-line fluxes (MPA-JHU, OSSY) and the gas-phase metallicity
> (MPA-JHU), in addition to other quantities not provided here, such as
> absorption-line strength corrected for emission-infill (OSSY) or
> stellar mass and star-formation rates (MPA-JHU). In fact, the
> statement "We aim to provide ... (iii) the first consistent analysis of
> absorption and emission line in SDSS spectra including parametric and
> non-parametric emission-line fitting" is not entirely correct in this
> respect, since both the MPA-JHU and OSSY catalogue analyse the stellar
> and emission-line spectrum in a similar way as done here, with the
> exception of the non-parametric emission-line fitting. Please amend.
>
fixed

> - Sec. 2.1
> 
> Please provide the number/fraction of objects with robust Galaxy Zoo
> classification in your catalogue and the redshift range of the latter,
> since I do not think that these would have covered all 800,000 of
> them.

added (661k galaxies with nvote>10)

> - Sec. 2.2.1
> 
> Regarding the mismatch between the GALEX and SDSS or UKIDSS
> resolution, I actually don't think that the corresponding
> central-magnitude offset for compact sources can be so easily
> dismissed. A simple calculation assuming typical seeing conditions
> suggests that for a compact Gaussian source with an intrinsic
> extension of just 1" the GALEX fluxes within a 3"-wide aperture would
> be around 75% of that collected with SDSS images, or 0.3
> magnitudes. Although this may be comparable to the GALEX uncertainties
> in the NUV magnitudes, the effect is systematic, and would apply to
> many compact or distant objects. I suggest that the authors estimates
> the impact of such a small offset on their stellar ages, in particular
> for old objects.
> 

Added and explained: the number of compact galaxies is tiny (~800, i.e. 0.1% of the sample). However, this does not affect our stellar age estimates because we use spectra alone. We plan to include photometric information in future updates of RCSED, then we will have to deal with this issue more seriously.



> - Sec. 2.2.2
> 
> This section should be re-written to make it clearer, in particular
> the 2nd paragraph. The reasons for anticipating an offset between SDSS
> and UKIDSS should be anticipated at the beginning of the section. A
> figure could also help to better understand the process.
> 
The reason is given in the first sentence: an observed large spread of optical-NIR colours (e.g. g-J). We have restructured the entire section and included some explanations. we have added a phrase that we compared UKIDSS GCS Z-band photometry with the SDSS DR7 z-band photometry "for exactly the same objects". The figure demonstrating the z-Y inconsistency is presented in Chilingarian et al. 2010 (K-corrections paper) and we prefer not to duplicate it here because this paper is already too long.

> 
> - Sec. 2.3
> 
> The NBURST procedure adopted here would seem rather similar to the
> pPXF code of Cappellari & Emsellem (2004), so perhaps the authors may
> possibly want to start by stating any advantages of their procedure.
> 
> The authors may also want to provide some motivation for adopting an
> exponentially declining star-formation history that assume an initial
> burst of star-formation nearly 13 Gyr ago. For instance they could
> quote here the work of Chiligarian & Zolotukhin (2012), and specify
> the limitation of this approach for objects dominated by young stellar
> populations.
> 
> More important, it would be important to specify what entails a
> detection of nebular emission, which where excluded in a second
> iteration of NBURST. Do the authors stick to a threshold based on
> the measured line-peak amplitude to residual-noise level ratio? And if
> so, what is that threshold? I would presume a Gaussian emission-line
> profile would have been adopted in this first assessment.
> 
> Similarly, please specify exactly what it is provided in the catalogue
> from the SSP approach. I would imagine these are luminosity-weighted
> ages and [Z/H] values.

We have added some explanations and clarifications. NBursts has important differences from the ppxf code regarding the stellar population fitting and we briefly described them. We did not use the emission line flux as a detection criterion but rather the reduced chi2 values of the fit. We also refer to the paper by Serra & Trager (2007) regarding the analysis of composite stellar populations using SSP models

> 
> - Sec. 2.4
> 
> This is the most problematic section of this paper, for several
> reasons. First, this is not the first catalogue that carefully
> subtracts the stellar spectrum before measuring emission lines, and
> simply stating that this approach is superior to the standard SDSS
> pipeline is not sufficient.
> 
> Second, if on one hand the non-parametric profile fitting could
> capture gas outflows and peculiar motions on the other hand it could
> lead to misleading results in object with central Type 1 AGNs. Indeed
> in the latter case it would make no sense to provide a single flux,
> velocity and velocity dispersion for recombination lines when these
> originates from two physically distinct emission-line region
> (i.e. narrow- and broad-line regions). In this respect, the choice of
> a Seyfert nuclei to illustrate the advantages of their non-parametric
> approach is most unfortunate, since this is exactly the kind of
> objects where the addition of a BLR could help just as well in
> improving the fit to the observed profile. In fact, authors should at
> least caution the users of their catalogue when using their
> non-parametric results for objects with central AGN. The recent paper
> by Oh et al. (2015) vividly illustrates how relatively common these
> objects are in the SDSS, and provides a catalogue of such objects.

We are sorry -- apparently, it was unclear from the description that we performed the non-parametric profile recovery independently for allowed and forbidden emission lines (in fact, the approach was targeted especially on Seyfert galaxies when we were developing it). We have added some explanations in the text.

At present, we do not provide physical interpretation of non-parametric fluxes in case of Seyfert galaxies. However, we are now working on a project on the line profile decomposition which allows us to separate broad and narrow line components and make precise flux measurements for each of them.

> 
> Third, even putting broad-line regions aside, the use of a more
> sophisticated non-parametric fitting of the emission-lines contrasts
> with the relatively cruder interactive approach whereby emission-lines
> are measured from the residuals of a stellar-continuum fit performed
> while masking regions affected by emission, which could include many
> age-sensitive features, in particular in heavily star-forming
> galaxies. Comparing this methodology to the approach used in the OSSY
> catalogue, whereby emission-lines are fitting with Gaussian profiles
> but where the nebular and stellar component of the spectrum are
> simultaneously matched, it would not be clear to a careful reader
> which catalogue would provide the most reliable emission-line
> measurements. As a sanity check, the authors should start by
> comparing their results based on Gaussian profiles with the ones
> provided by the OSSY catalogue, in particular for star-forming objects
> where the recovery of the underlying stellar continuum may suffer from
> the fact that most Balmer absorption lines would have been masked
> during the NBURST fits.
> 

It was demonstrated in several papers (Chilingarian et al. 2007, Chilingarian 2009 and others) that masking Balmer line regions biases neither age nor metallicity determinations. We have added a passage to the beginning of the section 2.4. Therefore, we do not expect any problems with the non-parametric fitting related to the template mismatch. We have added the comparison of our Gaussian fluxes with OSSY in Fig12.


> In this respect, I note that the fraction of objects with Hb emission
> with a S/N > 3 (48%, with Gaussian fit) is nearly the same as the
> fraction of objects with Ha emission above 2.86 times this threshold,
> that is S/N > 8.6 (estimated to be 55% from Tab. 2), whereas in the
> presence of dust absorption one would expect that the Hb fluxes
> corresponding to these Ha emitting objects would be generally smaller
> than just Ha/2.86. Perhaps this means that in general there is little
> reddening across the whole population of emission-line galaxies, but
> it would be worth checking that the Hb fluxes are not systematically
> biased as an outcome of the NBURST fitting procedure.

The dust would lower the H-beta fraction even more. Same for AGNs where the Balmer decrement is higher (3.1 or so). The explanation is, however, very simple: already at z=0.1 H-alpha shifts to the region affected by telluric absorptions and bright OH emissions, therefore, it can be detected and measured only if it happens to land between those lines. That, given relatively low spectral resolution of SDSS, substantially reduces the fraction of objects with detected H-alpha and [NII] lines. Also to be noticed, that the H-alpha detection is affected by the presence of [NII] lines around it.

> 
> Fourth, coming back to the non-parametric fit, I suspect this would be
> prone to biases in the case of weak emission either in objects with
> noisy data or in the presence of substantial template-mismatch. How
> robust is the non-parametric fit in these cases? And what defines a
> detection? Test are needed to reassure the reader against these
> effects. For instance, the author could produce artificial spectra
> with Gaussian emission lines and see how well these profiles are
> recovered non-parametrically as the lines becomes progressively
> immersed in the noise level associated to the stellar continuum, for
> lines with different equivalent widths. Similarly, they should compare
> their Gaussian and non-parametric results for massive early-type
> galaxies, where the Gaussian fits suggest the presence of weak but yet
> detected emission (i.e. most of the objects with LINER-like
> emission). For instance, the authors could check for systematic
> shifts or broadening in the emission-line position (a tell-tale
> behaviour of emission-line fits when adjusting to template-mismatch)
> as a function of the mismatch between the data and models in the Na D
> (e.g. Fig 3) or Mgb region. This would assume that weak LINER-like
> emission has intrinsically Gaussian line profiles, which should be the
> case considering that this emission most-likely originates from
> diffuse ionised-gas emission and that gas in early-type galaxies is
> generally relaxed.
> 
We performed a lot of simulations related to the non-parametric line fitting technique in the framework of our project of measuring BH virial masses in AGNs and we leave the discussion on that subject for the forthcoming paper (Katkov et al. in prep.) In fact, we forgot to mention in the text here about the regularization technique which we apply to the non-parametric inversion and which helps a lot in cases of noisy data. The regularization is applied only in the regions of low-S/N of a line profile (e.g. the "wings") and by design does not affect narrow lines / cores of broad lines. The RCSED data release will contain both regularized and original recovered profiles of emission lines.

Perhaps we misunderstand the referee's point, but the regions of NaD and MgB do not contain any strong emission lines, only [NI] and HeI (and the He line is 14A apart from NaI 5890). Because we do fix relative positions of emission lines while performing both Gaussian and non-parametric fitting (and as a result obtain two velocities: v_allowed and v_forbidden in both cases) if one weak line falls into a region of strong template mismatch, it will not affect the measurement of radial velocity and intrinsic broadening dominated by strong lines locate all over the spectral range.

We also ran MC simulations to get realistic flux uncertainties from the non-parametric fitting technique on a small sample of galaxies (2000 objects).


> 
> Finally, but this is a minor quibble, in Fig. 4 you may want to use
> different colors in the central panel, for the stellar-continuum and
> nebular fits
> 

We keep the version of Fig.4 in agreement with the spectrum preview service in our web-site. We will possibly change the color schema later before the public release of the web-site.

> 
> - Sec. 4.1
> 
> I was a bit puzzled by Fig. 8 here, and in particular by the absence
> of bright objects among the color-coded symbols in the four panels
> corresponding to z < 0.25. I can see why at higher redshift faint
> objects would be absent (either due to the SDSS magnitude cut but also
> to the difficulty in visually classifying distant objects), but I
> don't understand why bright objects are absent at lower redshift.
> Please clarify.
> 

SDSS rejects bright large galaxies -- it's a well known problem. Therefore, there are often groups without central galaxies and clusters without a cD -- those have to be added from other sources (e.g. 2MRS, 6dF).

We have double checked the figure. All points shown in individual panels (except dwarf galaxies from Chilingarian et al. 2008) make up the contours.

> - Sec. 4.2
> 
> This section is a quit confusing. The authors want to prove here when
> comparing the results from SSP or exp-SFH fits?
> 
> The authors are trying blame the fragmented nature of the SSP results
> on the way the SSP models are built, but I frankly do not fully
> understand their arguments for this, in particular when they point to
> the - obviously - smoother results from the exp-SFH fits. To be
> honest, I think most likely the SSP fits are based on solutions that
> are non-unique, similar to the case of orbit-superposition models for
> the stellar kinematics of galaxies. Also, it is not clear what it is
> that is plotted in the top panel of Fig. 9. Does each point represent
> the luminosity-weighted age and metalliticy? Maybe here the author
> should instead convey the weight given to each of the template used in
> their NBURST fit. For instance instead of plotting one point per
> object, they could plot 100 points, e.g. for a NBURST fit returning
> 70% weight to one template and 30% to another, they could plot 70
> times the point corresponding to the age and [Z/H] of the most
> important template, and 30 times the age and [Z/H] of the least
> important one. Maybe this would remove the "spotty" appearance?
> 
> As regards the velocity-dispersion degeneracy, does the 15% systematic
> errors on sigma depend on the velocity dispersion itself? Does this
> hold also for large velocity dispersions? This part would be worth
> expanding and explaining better in my opinion.

There is no relative template weighting involved in NBursts, therefore this argument cannot be applied there. We provide an explanation why we think we have this problem (see Section 4.2, par.4). Exponentially declining models are combinations of large number of SSPs therefore the effects are smeared our.

It has been demonstrated in Fabricant et al. (2013) that this problem (sigma-Z degeneracy) affects the high velocity dispersion end because the degeneracy effects become worse at very high velocity dispersions.


> 
> - Sec. 4.3.1
> 
> Given my previous comments, Fig. 13 should really be purged of those
> objects potentially containing a broad-line region. Care should also
> be take to plot here only objects where a non-parametric fit would
> make sense and would not be biased in the presence of weak lines in
> noisy spectra or by template mismatch. After this is taken care of,
> histograms would help in both axis to really appreciated the fraction
> of objects where a non-parametric fit really makes a difference in the
> Ha flux estimates. On the Chi^2 axis, it is more difficult to form a
> judgment without knowing the number of free involved in the
> non-parametric fit. I presume these correspond to the velocity
> sampling of the gas line-of-sight velocity distribution, so it's
> probably a considerable number (this should in fact be specified above).
> 

As we explained earlier, our non-parametric technique is a really unique approach to the interpretation of AGN spectra. Indeed, even though we rejected objects classified as "QSO" by the SDSS pipeline, there is a substantial number of AGNs as one would conclude from the galaxy distribution in BPT diagrams.


> - Sec. 5 and catalogue
> 
> Maybe it would be clearer to split table 4 in two tables, thus
> isolating the stellar-population fitting results.

No, we cannot do this, because those tables reflect the database schema and there is absolutely no reason from the database design point of view to isolate stellar population fitting results in a separate table.

> 
> - Summary
> 
> See my previous comments on the lack of a comprehensive discussion of
> present and future applications for this catalogue.

We have added a couple of paragraphs in the Summary section.


With best regards,
the authors