** We thank the referee for another careful reading and for once again providing useful, articulate advice.  We believe we have finally addressed the major concern over the uncertainties, which should alleviate the related concerns over the CIGALE-based modeling.  However, other papers (cited in the text) have studied the utility and accuracy of CIGALE, and so we feel that any detailed analysis of the model itself (besides the standard Monte Carl-type analysis we perform) is beyond the scope of this paper.


I appreciate the authors adding more information about the equivalent surface brightness and limiting depth of the data. But the authors say they achieve a limiting surface brightness of "~ 28 mag/arcsec2". This should be band-specific -- just as the quoted point source depth is different in different bands, the surface brightness limits should be as well. There needs to be a discussion of depth/uncertainty at all wavelengths, IR and UV included (particularly since the exposure time on the UV data varies greatly).

** This good suggestion has been incorporated, with these relevant excerpts:

UV: For these longer integrations the limiting surface brightnesses are ~32.5 and 32.0 mag arcsec^{-2} AB, respectively for the far-ultraviolet and near-ultraviolet channels; for the small set of shorter integration ultraviolet images, the surface brightness limits are about 1--1.5 mag arcsec^{-2} shallower.

ugr: The limiting ugr surface brightnesses are ~28.2, 28.4, and 28.0 mag arcsec^{-2} AB

3.6um: fainter than 29 mag arcsec^{-2} AB

IR: The surface brightness sensitivity of the infrared data is approximately 26.7 mag arcsec^{-2} at 12um, 25.0 mag arcsec^{-2} at 22um, 26.0 mag arcsec^{-2} at 24um, and 19.5 mag arcsec^{-2} at 70um.


Regarding my request to see demonstrate the photometry by showing data with photometric masks, the authors have chosen to only show the Spitzer data. Given that the new dataset here is the optical imaging, and that it's the depth/analysis of this data that really needs to be better described, the authors need to show the optical imaging. In fact, nowhere in this paper are the galaxies shown; it was only after looking these systems up on NED that I realized the sample mixed face-on and edge-on galaxies. Face-on and edge-on systems will have very different behavior in their light profiles (and optical depths), particularly if one is doing measuring profiles in annuli that follow the isophotes.

So the authors really need to show the optical images for all galaxies, displayed in a way that the reader can see the flux at least out to the radii that are being analyzed (better yet, to the limiting depth of the data). That's really the best way to demonstrate the depth and radial extent of the data.

** We have added a figure that displays the r' imaging for the entire sample (and have indicated with an ellipse the outermost regions sampled in this work).


And the authors claim that they don't need to do additional masking on anything but the Spitzer data. That's simply not true if you are doing deep optical surface photometry. If the authors really are claiming to go to 28 mag/arc sec2, at these surface brightnesses and for galaxies this big, there will be significant contamination from discrete sources (background objects, largely). Now it may be that the authors aren't interested in going that deep with the data -- in looking hard at Fig 3, it looks like both the g- and r- profiles generally stop at around mu=25-26 mag/arcsec2, 2-3 magnitudes brighter than the quoted 28 mag/arcsec2 limit. This is presumably because the IR and UV data do not go as far out. This means the analysis isn't truly probing the extended outer disk, and the authors need to pare back their language about how deep their study goes -- their profiles don't go very deep by modern optical imaging standards.

** In fact we did carry out additional masking at all wavelengths, but as we stated in our previous response to the referee, the additional masking at non-3.6um wavelengths was less onerous than what was done for the deeper 3.6um mosaics.  The referee is correct that we are limited by the wavelength-dependent extent of the emission and surface brightness sensitivity.  For these reasons we only analyze out to 1.5R25.  It can be a matter of semantics and left to the individual reader whether or not probing out to 1.5R25 reaches the "outer disk", but nonetheless we have taken the referee's suggestion and pared back some of the language, e.g., "outermost" --> "outer" in the Summary (twice); emphasized we probe only out 1.5R25 in the Summary; and qualified in the Introduction that "outskirts" in this paper refers to 1.5R25.


But ultimately I'm still confused by the photometric uncertainty and errorbars. Section 4.1 gives formulas for calculating errorbars, which as I said in my earlier report needed to be explained more thoroughly (and for which the authors gave no further explanation of in their revisions). But then the errorbars on the colors shown in Fig 4 refer to simply changing the sky background by 1-sigma, so they don't seem to be related to the description of photometric uncertainty in Sec 4.1. Particularly since sky uncertainty at low surface brightness can be very non-gaussian -- the dispersion of individual pixel intensities can often be very different from the dispersion between the average intensity of different sky regions, and arguably it's the latter number which is most important at low surface brightness.

In looking at the color profiles shown in Fig 4, the errorbars seem much too large. The outermost data points are at a surface brightness of mu=25-26, but the colors here have errorbars of +/- 0.5-1 magnitudes! For data as deep as claimed here (mu_lim ~ 28), the color uncertainty at these high surface brightnesses should be *much* smaller. So either the errorbars/uncertainties are being calculated wrong, or the limiting depth is not as deep as claimed.

So given the inconsistency in how the errorbars are being described, and the surprisingly large errorbars shown in Fig 4, I simply don't know what to make of the photometric accuracy claims, or the claims of limiting depth.

This then presents concern for the SED modeling. I find it hard to believe that with that level of uncertainty in the colors beyond a_25=1, the results of the SED fitting can be as accurate as they are. I can only presume that the other bandpasses are similarly uncertain (there is no discussion/demonstration of uncertainties for the UV/IR data). The inferred e-folding times of the inferred star formation histories (shown in Fig 8), all have errorbars of ~ +/- 1 Gyr even while the uncertainty on the optical colors ranges from < 0.1 mag to > 0.5 mag. I simply don't find this credible.

The SED fitting seems not to factor the observational uncertainty into the analysis, particularly as it pertains to the increased uncertainty at large radius in the profiles. So the errors shown in Fig 8 must only be due to systematics in the models. The authors try to show the effect of observational uncertainty in Fig 10, but this is neither sufficient nor explained thoroughly enough. The authors only say they "inject uncertainty" into the fluxes, but don't say what that uncertainty is, or if it is typical of the inner high surface brightness regions or the outer low surface brightness regions.

Given that the primary result (the switch in the age trend between the inner and outer disk; Fig 5) is a weak trend, and much smaller than the scatter in the data, it's particularly important to demonstrate that the data can constrain the SF histories to this level of accuracy. Without a proper consideration of how the observational uncertainties factor in to the analysis (as a systematic function of radius), this result is not at all believable.

** The referee is correct to be persistent on this issue.  We have revised our approach to the photometric uncertainties and now consider both small-scale and large-scale (e.g., flat-fielding) fluctuations in the sky levels.  We also have incorporated into the writing the calibration uncertainties for each photometric band (UV 15%; ugr 3%; 3.6um 5%; WISE 12 & 22um 10%; MIPS 24um 7%; PACS 70um 5%).  Note that in the previous version of the paper, the +/- 1 sky sigma approach for the uncertainty was only applied to color plots, not to the actual flux errors (and hence the 0.1-0.5 color uncertainties displayed in the figure were not carried over into the SED modeling).  

** As for the uncertainties incorporated into the Monte Carlo simulations, we have tried to clarify the writing:  "In each simulation a random flux offset was added to each flux, with the flux offset derived from a Gaussian distribution with $\sigma$ scaled according to the measured uncertainty."


Finally, the authors seem to have missed my point about the confusion with the different star formation models. If the two models being considered are the delayed model and the single exponential model -- if double exponentials are not being considered (and I agree they shouldn't be) -- then why have a long discussion of double exponential models, along with an explicit equation for them (Eq 3), and why show a double exponential model (Fig 2) rather than a single exponential model? Just describe/show the models being considered: delayed and single exponential.

** Agreed.  We have removed the references and graphical display of the double exponential SFH.


Another minor point: Table 4 has a footnote "c", but nothing is marked with this footnote.

** Thanks; fixed.