The Scaling Relations of Galaxies with Different Morphology: Comparison Among WINGS, MANGA and Illustris Data Samples
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper presents a panoramic overview of the most commonly used scaling relations, derived from galaxies of different morphological types selected from two large surveys: WINGS (galaxies in nearby clusters) and MaNGA (Mapping Nearby Galaxies at APO).
- However MaNGA studies individual galaxies, while WINGS studies galaxies in a cluster. These are two distinct catalogues (what's more with different objects in each, only 58 galaxies in common), but that can be seen to some extent as complementary. The authors should explain in more details the value of comparing them.
- Following the authors the great interest of this panoramic overview lies in its use for building galaxy models through simulation. Unfortunately the data presented by the authors exclusively concerns the global aspect of the galaxies. When the authors say for instance:
“How many galaxies of elliptical type in the local Universe have a radius and surface brightness or mass or central velocity dispersion within a given range ?. Such information is very important when we try to model our Universe”.
This is not true for individual galaxies. In fact, each galaxy is a very special objet and implies a particular simulation where the statistical aspect (mean values of the parameters) eventually plays a minor role. This point must be specified.
3. The Log(n)-L figures are misplaced after the discussion on the L-sigma plane. Likewise Figs. 25, 26 have to be placed in paragraph 7. On the other hand rotational velocity fields versus dynamical mass for rotation-supported galaxies are not discussed. It is an important point for the galaxy model maker. Admittedly, the WINGS‑SPE component survey did not collect spectroscopic kinematic data (rotation velocities). I suggest adding (but the authors don't have to) a note on this item.
4. A paragraph on dark matter and modified gravity (for instance MOND) would be useful. What distinct role does dark matter (or modified gravity, e.g. EFE in MOND) play in an individual (or even isolated) galaxy or, on the contrary, located in a cluster environment? A short discussion on this point would be welcome.
Minor corrections:
- MANGA or MaNGA as shown in the conclusion ?
- In Figs 21-24 the log(n) is not clearly visible on the left-hand side.
- Missing journal names ?:
Faber, S.M.; Jackson, R.E. Velocity dispersions and mass-to-light ratios for elliptical galaxies. The Astrophysical Journal 1976, 204, 668–683. https: 776
//doi.org/10.1086/154215.
etc.
- Acknowledgments: We thank ...
Author Response
Referee: This paper presents a panoramic overview of the most commonly used scaling relations, derived from galaxies of different morphological types selected from two large surveys: WINGS (galaxies in nearby clusters) and MaNGA (Mapping Nearby Galaxies at APO).
However MaNGA studies individual galaxies, while WINGS studies galaxies in clusters. These are two distinct catalogues (what's more with different objects in each, only 58 galaxies in common), but that can be seen to some extent as complementary. The authors should explain in more details the value of comparing them.
Following the authors, the great interest of this panoramic overview, lies in its use for building galaxy models through simulation. Unfortunately the data presented by the authors exclusively concerns the global aspect of the galaxies. When the authors say for instance:
“How many galaxies of elliptical type in the local Universe have a radius and surface brightness or mass or central velocity dispersion within a given range?. Such information is very important when we try to model our Universe”.
This is not true for individual galaxies. In fact, each galaxy is a very special object and implies a particular simulation where the statistical aspect (mean values of the parameters) eventually plays a minor role. This point must be specified.
Reply: We have added a much clear explanation of the role of the WINGS, MANGA and Illustris data samples.
Certainly we agree that simulations should be able to reproduce single galaxies, but it is also very important that simulations
reproduce the statistical properties of galaxies, such as for example the distribution of morphological types, the luminosity function, the range of masses and velocities, etc. If these statistical properties are not reproduced by simulations, the ability of reproducing individual galaxies looses its significance and becomes a pure exercise not inserted in a cosmological context.
Indeed, the first comparison of simulations with observational surveys were performed on the statistical properties, such as the large scale structure, stressing the problem in reproducing the dwarf galaxies (e.g. the missing satellites problem), or on the mass and luminosity function, etc. These are the first things a numerical simulation should check.
Our analysis demonstrates that simulations can reproduce the scaling relations of galaxies, but the final answer on how well they do that is still far away. The analysis of the correlation among the parameters is only one aspect of the problem.
Our work has demonstrated that without a careful statistical analysis of the data samples that have to be compared, any conclusion on the ability of simulations in reproducing the 2D distribution of galaxies in the ScRs is not robust enough. Many problems should be solved to get the answer we are looking for.
==========================================
Referee: The Log(n)-L figures are misplaced after the discussion on the L-sigma plane. Likewise Figs. 25, 26 have to be placed in paragraph 7. On the other hand rotational velocity fields versus dynamical mass for rotation-supported galaxies are not discussed. It is an important point for the galaxy model maker. Admittedly, the WINGS‑SPE component survey did not collect spectroscopic kinematic data (rotation velocities). I suggest adding (but the authors don't have to) a note on this item.
Reply: The figures of the n-L relation are not in a misplaced position. The problem was that the Latex compilation puts the L-sigma chapter before the figures. We tried to solve this, but we did not succeed. Maybe during the production of the paper this problem can be solved by the editorial office.
Indeed, we do not have rotational velocities of both WINGS and MANGA galaxies. We agree with the referee that this is an important point.
Anyway, our analysis suggests that low luminous galaxies behave in a different ways with respect to bright galaxies in all ScRs.
In particular, in the log(n)-log(L) and log(sigma)-log(L) relations low luminous galaxies deviate significantly from bright galaxies.
The role of rotation should be one of the possible causes of such different behavior, but we cannot address it here.
==============================================
Referee: A paragraph on dark matter and modified gravity (for instance MOND) would be useful. What distinct role does dark matter (or modified gravity, e.g. EFE in MOND) play in an individual (or even isolated) galaxy or, on the contrary, located in a cluster environment? A short discussion on this point would be welcome.
R. Sorry, we cannot add any comment on the DM role. The reason is the we do not have observational measurement of the DM mass in both surveys. In addition the only ScR that can be affected by DM is the L-sigma relation for which WINGS has few data in particular for E galaxies. A short discussion of DM will make the paper even more difficult to follow. The paper is already too long.
==============================================
Minor corrections:
Referee: MANGA or MaNGA as shown in the conclusion ?
In Figs 21-24 the log(n) is not clearly visible on the left-hand side.
Reply: MANGA is now used consistently in the whole paper.
We tried to make the ylabel better visible without success. In the original figures they are well visible. This is again possibly a Latex problem. We will check this point during the production phase. If the proofs demonstrate that they are still not well visible we will attempt other solutions.
Referee: Missing journal names ?:
Faber, S.M.; Jackson, R.E. Velocity dispersions and mass-to-light ratios for elliptical galaxies. The Astrophysical Journal 1976, 204, 668–683. https: 776 //doi.org/10.1086/154215.
Reply: The references of the journals are correct in the .bib file. They come directly from the NASA/ADS system. The compilation of the Latex file do not put the Journal names in all references. We cannot solve this problem. We will ask the editor to solve it.
Referee: Acknowledgments: We thank ...
Reply: Acknowledgments corrected.
Reviewer 2 Report
Comments and Suggestions for AuthorsReferee Report universe-4067740
The main weakness of this manuscript is the use of two different surveys.
Both are SDSS derived but span different volumes and are not complete. The selection function is not discussed at all.
What are the volumes? Down to which luminosity or mass range are they complete. Why do the authors think they can be compared directly? There is NO discussion of the selection function for both surveys.
This makes the results quite problematic to interpret.
L84:
“in such a way that the same 2D plane obtained with different data-sets is often quite84 different from a statistical point of view. “
>> This is typically because of differences in survey volume. I find this the main issue of this paper that the volume argument for differences between scaling relations.
The authors mention this somewhat obliquely in L107
“iv) The galaxy samples under examination are not statistically similar. ”
>> But the observational design for extragalactic astronomy should strive to keep all systematics to a minimum before comparing different samples of galaxies and their properties.
v) The statistical analysis of111
the correlation changes.
>> this is not well substantiated
"How many galaxies of elliptical type in the local Universe117
have a radius and surface brightness or mass or central velocity dispersion within a given118
range?".
>> I think this has very much been answered by SDSS, S4G, 2MASS and other nearby galaxy surveys.
“”
>> There very much are since this was a single Sersic fit. There are differences for the size of the disk if one allows for a bulge for example.
L239
“Vega photometric system redshifted to the rest frame of239 each object. The k-correction was not considered and the B−V c
>> This is an issue since every other survey has most likely AB magnitudes.
L247
same range of redshift (0.04 ≤z ≤0.07).247
But do they span the same mass range? What is the difference or similarity in survey volume?
2. The range of galaxy parameters260
>> But is the photometry in the WINGS sample done with Vega or AB?
L323
“ …which qualitatively look identical but actually323
are not. “
Kolmogorov-Smirnoff tests would confirm.
There are enough statistical tests to check whether two distributions are similar or statistically significantly different from one another. Even for small samples.
L335
“ i) To trace the335
structural evolution of galaxies. By comparing slopes and intercepts at different redshifts,336
one can tests whether ETGs grow in physical size or simply fade due to passive stellar337
evolution. ii) To distinguish real size evolution vs luminosity fading; “
>> These two points read as the same point.
L378
“This is due to the special data analysis made by the WINGS team for the subtraction of the sky background (see [6]).”
>> and therefore likely also an environmental effect?
L392
The Λ-shape of the Ie−Re plane should392
be seriously considered when the KR is analyzed. T
>> for mass-limited, complete samples.
L426 and onwards
>> Alternatively a 2D K-S test or an Anderson-Dalrong test for multi-dimensional data. This would be perhaps a better way to identify similarity and significance.
Since this is not a very common statistical test, the null hypothesis should be stated specifically. Null hypothesis is that the two distributions are different?
8. Comparison with the Illustris simulation649
“log(M∗/M⊙)∼9.16 ÷12.4 for objects with redshift z < 0.05. T”
>> Not similar redshift range or mass range as the other surveys. If one wants to compare the distribution of properties there has to be a valid A/B test i.e. the sample is similar in other ways.
L695
“To compare observations with simulations we need to limit the mass interval of695
WINGS and MANGA to that of artificial galaxies. ”
This is something you should have done before comparing the two surveys!
“This lack of similarity indicates that we do not yet possess a standard, homogeneous748
sample of galaxies with robustly measured structural parameters suitable for direct com-749
parison with simulations.”
The manga and wings samples may not fit that bill but there are several other samples that may be a good z=0 epoch comparison. One can define volume and mass limited samples for the comparison. This statement could only be substantiated if one has checked all local samples with volumes and limits etc.
The relations754
may look qualitatively similar, but their full 2D distributions are significantly different.7
>> the word “significant” would have to be backed up where two —otherwise similar i.e. in redshift and mass range— samples of galaxies different in a given property to a degree that a statistical tests found significant (P<0.05).
I do not feel the presented results back this statement up.
Author Response
Referee: The main weakness of this manuscript is the use of two different surveys. Both are SDSS derived but span different volumes and are not complete. The selection function is not discussed at all. What are the volumes? Down to which luminosity or mass range are they complete. Why do the authors think they can be compared directly? There is NO discussion of the selection function for both surveys. This makes the results quite problematic to interpret. Reply: The referee is right. The comparison between the samples requires a careful preliminary analysis of the selection function. However, the aim of this paper is not that of comparing the ScRs when all the selection criteria are taken into account, but is exactly the opposite. We want to highlight what happens when two very different samples are compared and no preliminary analysis is done on the selection criteria. Indeed, the problems that enters in the comparison of the ScRs does not reside only in the volume samples of the surveys, but several factors play a role in producing different distributions. The paper wants to stress that among these factors we can list: 1. Different definitions of the scaling parameters (e.g. Radii, masses, brightness, etc.); 2. different data reduction (e.g. sky background subtraction, stellar mass derivation, etc.); 3. different volume and luminosity function for the different morphological types; 4. different environment. All these effects concur to produce different distribution of galaxies in the ScRs. We tried to explain better this point in our introduction. ======================================================== Referee: “in such a way that the same 2D plane obtained with different data-sets is often quite different from a statistical point of view. “ This is typically because of differences in survey volume. I find this the main issue of this paper that the volume argument for differences between scaling relations. The authors mention this somewhat obliquely in L107. Reply: As explained above the comparison of the ScRs is not only a matter of working with similar volumes. The statistical tests done in our study have demonstrated that when the interval of magnitude is very small, the two distributions under comparison become statistically consistent. This means that in addition to similar volumes, similar luminosity functions are necessary. Furthermore, the choice of similar environment and similar definition of the structural parameters, are also factors that might introduce systematic differences in the observed ScRs. The idea here, is to put in evidence all the effects that might introduce systematically different 2D distribution. ========================================================= Referee: “iv) The galaxy samples under examination are not statistically similar. ” But the observational design for extragalactic astronomy should strive to keep all systematics to a minimum before comparing different samples of galaxies and their properties. Reply: Right, but this is not our aim. Reduce to a minimum the factors that potentially influence the ScRs is not what we want to do. In this work we want to stress what is necessary to do when such a comparison is the goal of the analysis. The choice of the used samples has been better explained in our new introduction. ========================================================= Referee: v) The statistical analysis of the correlation changes. >> this is not well substantiated Reply: We tried to better explain the point. ========================================================= Referee: "How many galaxies of elliptical type in the local Universe have a radius and surface brightness or mass or central velocity dispersion within a given range?". >> I think this has very much been answered by SDSS, S4G, 2MASS and other nearby galaxy surveys. Reply: None of these surveys have quantified how many galaxies of a given morphology are present in each of the box built in the ScRs plane. The answer to this question requires an extra work, that of reducing to a minimum the possible differences in volume sampled, in luminosity function, in the working data analysis and in the definition of the scaling parameters. Indeed, any of these survey presented their own ScRs, but this not sufficient for a robust answer to the question posed in our paper. Qualitatively the ScRs are similar, but when the details matter, statistical differences are present among the data of different surveys. Up to now we are not aware of any work that attempted such complex comparison among surveys. ========================================================= Referee: There very much are since this was a single Sersic fit. There are differences for the size of the disk if one allows for a bulge for example. Reply: Yes. The way in which the Sersic parameter is measured is very important. 1D and 2D analysis might give different results, equally taking or not into account in the same way the role of seeing. We agree that this is an extra element of potential systematic difference in the ScRs. In our paper we tried to highlight this point. ========================================================= Referee: “Vega photometric system redshifted to the rest frame of each object. The k-correction was not considered and the B−V correction >> This is an issue since every other survey has most likely AB magnitudes. Reply: Right. Indeed MaNGA (SDSS) uses the SDSS photometric system, which is intended to be on the AB magnitude system (or at least very close to it), while WINGS photometry is traditionally on classical optical systems (e.g., Johnson/Cousins), which are usually Vega-based in calibration. While site calibrations can be tied to external systems (e.g., comparisons to SDSS for checks), the intrinsic calibration and zero-points differ from the AB system. The difference in magnitude depends on the filter used. On average in the V-band the expected difference is only 0.02 mag. Our figure 1 demonstrates that for the 58 galaxies in common there are indeed not strong systematic differences. The magnitude system used by MANGA and WINGS are very similar in the V-band. The small difference in zero-point does not introduce strong variations in the 2D planes. ========================================================= Referee: But do they span the same mass range? What is the difference or similarity in survey volume? Reply: Figure 1 shows that, despite the big difference in the distribution of L, the range of masses covered by the two survey is quite similar. ========================================================= Referee: 2. The range of galaxy parameters >> But is the photometry in the WINGS sample done with Vega or AB? Reply: See previous answer. ========================================================= Referee: “ …which qualitatively look identical but actually are not. “ Kolmogorov-Smirnoff tests would confirm. There are enough statistical tests to check whether two distributions are similar or statistically significantly different from one another. Even for small samples. Reply: In the paper we used the Energy Distance Test in addition to the Kolmogorov-Smirnov test. Both give the same answer. This is what we expect to happen when there is no an a priori work on the selection effects that might influence the ScRs. ========================================================= Referee: i) To trace the structural evolution of galaxies. By comparing slopes and intercepts at different redshifts, one can tests whether ETGs grow in physical size or simply fade due to passive stellar evolution. ii) To distinguish real size evolution vs luminosity fading; “ >> These two points read as the same point. Reply: Yes thanks, we correct it. ======================================================== Referee: “This is due to the special data analysis made by the WINGS team for the subtraction of the sky background (see [6]).” >> and therefore likely also an environmental effect? Reply: The sky subtraction depends critically on the depth of the survey. Environment do not play a big role in this. ======================================================== Referee: The Λ-shape of the Ie−Re plane should be seriously considered when the KR is analyzed. >> for mass-limited, complete samples. Reply: Yes, the analysis of slopes and intercept depends critically on this. ======================================================== Referee: L426 and onwards >> Alternatively a 2D K-S test or an Anderson-Dalrong test for multi-dimensional data. This would be perhaps a better way to identify similarity and significance. Since this is not a very common statistical test, the null hypothesis should be stated specifically. Null hypothesis is that the two distributions are different? R. In our work we used two tests: Kolmogorov-Smirnov and the Energy Distance test. We do not think the Anderson-Dalrong test would change the results. The Anderson–Darling (two-sample) test compares empirical cumulative distribution functions (ECDFs). It measures integrated squared differences between ECDFs, with extra weight in the tails. It is a rank-based, nonparametric test. The Energy Distance test compares distributions using pairwise distances between observations. It is based on expectations of Euclidean distances. The Anderson-Dalrong test and the Energy Distance test have different sensitivity. The following table summarizes them: Feature Anderson–Darling Energy Distance --------------------------------------------------------- Mean shifts Moderate High Variance changes Moderate High Tail differences Very high Moderate Multimodality Moderate High Multivariate data No Yes ---------------------------------------------------------- In conclusion: AD is especially powerful when distributions differ in tails. ED is more sensitive to global geometric differences, including shape and clustering. If applied to our data the AD tests would stress the big differences in the tails of the distributions that are generally quite big. ========================================================= Referee: Comparison with the Illustris simulation “log(M∗/M⊙)∼9.16 ÷12.4 for objects with redshift z < 0.05.” >> Not similar redshift range or mass range as the other surveys. If one wants to compare the distribution of properties there has to be a valid A/B test i.e. the sample is similar in other ways. Reply: The galaxies of the Rodriguez et al. sample (~1500 objects) span the range of solar mass 9.16-12.40 dex for objects with redshift $z<0.05$ from the TNG-50 simulation. The sample of Ferreira et al. contains 10121 galaxies at redshifts $z<0.1$ from the TNG100-1 simulation. Most of them (more than 9000) are ETGs. The mass range is 10.17-11.96 dex. We chose two very different sets of simulations with different range of redshifts and masses. As explained in the introduction this is explicitly done in order to amplify all possible differences in the observed scaling relations and to stress the importance of taking into account all possible effects that can modify the observed distributions. The lack of such analysis, when comparisons are made, is the source of wrong statements about the ability of simulations in reproducing the observational data. ========================================================= Referee: “To compare observations with simulations we need to limit the mass interval of WINGS and MANGA to that of artificial galaxies. ” This is something you should have done before comparing the two surveys! “This lack of similarity indicates that we do not yet possess a standard, homogeneous sample of galaxies with robustly measured structural parameters suitable for direct comparison with simulations.” The Manga and Wings samples may not fit that, bill but there are several other samples that may be a good z=0 epoch comparison. One can define volume and mass limited samples for the comparison. This statement could only be substantiated if one has checked all local samples with volumes and limits etc. Reply: As explained above the correct comparison of the data samples is not our aim. We explicitly did not consider the selection criteria to stress the effects on the ScRs. You are right, the comparison requires such preliminar analysis, but this is not want we want to show. The idea here is to show what effects (selection criteria, environment, data analysis, definition of parameters) can alter the observed ScRs. =========================================================== Referee: The relations may look qualitatively similar, but their full 2D distributions are significantly different. >> the word “significant” would have to be backed up where two - otherwise similar i.e. in redshift and mass range — samples of galaxies different in a given property to a degree that a statistical tests found significant (P<0.05). I do not feel the presented results back this statement up. Reply: We modify a little our statement that wants to stress that any comparison between data samples requires a big preliminary work to homogenize and make the differences as small as possible.
Reviewer 3 Report
Comments and Suggestions for AuthorsThis is a well-written paper describing the positions of observed and simulated galaxies on scaling relations and emphasizing the need for allowing for selection effects and requiring consistent datasets and definitions when performing such comparisons. It should be of great interest to those interested in comparing simulations and observations. I only have minor comments:
1) Abstract - 2nd sentence "ScR" should have a small "c" for consistency with the rest of the paper.
2) Section 1, Line 174 - it's unclear if the 20% error in z for the WINGS sample referred to here is an error in the photometric or spectroscopic redshifts (and what fraction of the sample used in the analysis came from the spectroscopic sample).
3) Section 1, L196 - it would be good to include the URL for the MANGA web catalog referred to here.
4) Figure 2, the symbols could be smaller as they overlap in some of the plots.
5) Section 3 , 3rd paragraph (L351) - the surface brightness dimming due to redshift is even bigger than (1+z)^4 for most galaxies due to the K-correction, which should probably be mentioned in this context.
6) Figs 9-32 - not essential, but the authors might consider using KDE plots to present these results more clearly. E.g. in python the Seaborn library contains routines to make these. e.g. https://seaborn.pydata.org/examples/layered_bivariate_plot.html
7) Section 3, L426: please give a citation for the Energy-distance statistical test.
Author Response
Referee: This is a well-written paper describing the positions of observed and simulated galaxies on scaling relations and emphasizing the need for allowing for selection effects and requiring consistent datasets and definitions when performing such comparisons. It should be of great interest to those interested in comparing simulations and observations. I only have minor comments:
1) Abstract - 2nd sentence "ScR" should have a small "c" for consistency with the rest of the paper.
Reply: done
Referee: 2) Section 1, Line 174 - it's unclear if the 20% error in z for the WINGS sample referred to here is an error in the photometric or spectroscopic redshifts (and what fraction of the sample used in the analysis came from the spectroscopic sample).
Reply: We better clarified this point.
Referee: 3) Section 1, L196 - it would be good to include the URL for the MANGA web catalog referred to here.
Reply: done
Referee: 4) Figure 2, the symbols could be smaller as they overlap in some of the plots.
Reply: done
Referee: 5) Section 3 , 3rd paragraph (L351) - the surface brightness dimming due to redshift is even bigger than (1+z)^4 for most galaxies due to the K-correction, which should probably be mentioned in this context.
Reply: Corrected.
Referee: 6) Figs 9-32 - not essential, but the authors might consider using KDE plots to present these results more clearly. E.g. in python the Seaborn library contains routines to make these. e.g. https://seaborn.pydata.org/examples/layered_bivariate_plot.html
Reply: Thanks for the suggestion, but the authors do not have a great expertise of python routines. We prefer to continue this work using SM. Using python for doing all figures will be a too long work that will not modify the conclusions.
Referee: 7) Section 3, L426: please give a citation for the Energy-distance statistical test.
Reply: done
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsDear authors,
I believe that you have addressed the questions appropriately; however, certain issues, particularly those relating to the formatting of the paper, remain outstanding.
Author Response
Thank you. We will try to solve this during the production phase.
Reviewer 2 Report
Comments and Suggestions for AuthorsDear Authors,
Thank you for taking my comments into account. I see no objections for publication. Congratulations!
the Referee
Author Response
Thanks a lot
