General comments

The paper addresses the relevant scientific questions related to the evaluation of model simulations and observed data, and within the scope of CP. Only one citation Hegerl and Zwiers, 2011 is not enough and even was discussed earlier in IPCC 2001 Detection of Climate Change and Attribution of Causes — IPCC. Reports TAR, Climate Change 2001, 2013 AR5 Climate Change: The Physical Science Basis Detection and Attribution of Climate Change: from Global to Regional.

Some parts of the methodological approaches are mixed up with results. The authors often avoid clarifications and explanations, which can be helpful. There is no clear conclusion, only assumptions. It is not clear how TRW chronologies were pre-selected, which method of standardization was applied. Moreover, all chronologies have different age trends, periods, site-specific, and species-specific differences. All uncertainties can be related to the methodological approaches and pre-selection procedure.

Specific comments

1. L. 29 Starting from the first sentence - brings confusion between different forcing and factors. The main research question/hypothesis in the article can be better formulated. It is not clear if the main focus is irradiative forcing before /after volcanic eruptions or in general forcing or any other forcing factors or mechanisms that will be taken into consideration. It should be clearly formulated.
2. L. 80 unclear which time period (past 600 years from xx to xx?). Which grid net was used (lat, alt)? TRW observations – common period? Which chronologies, citations, how many chronologies n=? Pre-selection high latitudes, mid-latitudes? Please provide citations and refer to Fig 1.
3. L.89 no citation .” to prior studies”, which one? Please cite. …” of reconstructed surface temperature”? Summer temperature? Annual temperature? Please specify.
4. L.101 – historical temperature. Is it reconstructed temperature? If yes, please provide the period. If not, please clarify.
5. Subsection 2.1 It will be good to provide more details about the TRW database used for analysis, e.g., time period, regions, species.
6. Subsection 2.2 Please explain what T and M mean. “Parameters T1, T2, M1and M2.
7. L. 160 please clarify why a 71-year high-pass LOESS filter was applied.
8. L. 194 this description should be provided earlier in Figure 1 legend
9. L. 210-220 info about TRW chronologies, length, sites should appear earlier in section 2, subsection 2.1
10. L. 262 it is unclear based on which criteria the 12 largest volcanic events were pre-selected (VEI?) and which one (names). Please clarify.
11. Figure 5. It is unclear why annual temperature and annual precip. are considered? In legend VOLC – volcanic forcing, in Table 1 – V. Please select one abbreviation through the whole manuscript.

Technical corrections        

1. L.16 Abstract: tree ring width replace with tree-ring width
2. L.42 – references are not in the correct order. Please correct.
3. L.43 instrumental period of observations, please specify the period. For many stations outside Europe the instrumental period of observations for precipitation is rather short (ca. 50 years).
4. Figure 1. Abbreviations should be clarified in the Figure legend. E.g., optimize S/N ratio. Please clarify numbers (B14)? Please check abbreviations and provided an explanation in scheme precipitation or precip.,) in the text L. 101 (PREC). Please be consistent.
5. L. 105 – Eq. 1 is not in Section 2.4. Firstly, it was mentioned p.2. It should be Eq. 2. Please correct the numbers.
6. L. 110 consider revision .. “is constructed is illustrated”
7. Polson et al, (2013), replace with Polson et al. (2013),
8. L. 176, 178 – please check (is/are)
9. L. 203 GT, GM – please clarify what is what.
10. Fig 3, x-axis please write Year (CE)
11. Fig. 3 in plot – edf and citation edf – please clarify
12. L. 232 replace to "..a 11-year.."
13. L. 341 AOD – please clarify.
14. L. 342 please add a citation.

This is a thorough and very interesting study combining proxy systems modeling with the detection and attribution framework applied directly to tree-ring width proxy chronologies.  I especially appreciate all the work that authors have invested in dealing with the many challenges of the data, model(s), and simulation output (these are considerable).  In particular, the attention to evaluating VSL in the 'real world' before moving into the simulation and D&A framework, observations about the nature of parameter sets (e.g. information around Line 195 is really interesting to think about the implications and potential interpretations of this), attention to temperature and precipitation bias issues, and various other aspects.  This will be a useful touchstone paper and I suspect also motivate further work, since tree-ring proxy systems models are both valuable but then again challenging to use in frameworks such as the one here because of model bias, parameter uncertainty, and often mixed or weak climate signals in large tree-ring datasets (particularly for temperature) compared to the deterministic climate signals that emerge from VSL.  My major comments below are primarily around the ability of VSL to simulate the chronology set here and how this propogates into the differences between observed and simulated series and how this then impacts particularly the moisture-sensitive D&A:

1. Patterns of successful simulations (Line 194 and elsewhere): I think it would be desirable to get a better idea of where and for what chronologies the VSL simulations are successful - I get the sense from the manuscript and the Tolwinski-Ward papers show that (in general and not surprisingly), VSL will do better when the chronology in question has a strong climate signal itself (because VSL is driven by climate filtered through some possibly nonlinear simulated processes).  In any case, it would be helpful to visualize the success of VSL here - where (which chronologies, that is) can VSL successfully simulate and how many of these are moisture vs. temperature - my guess would be that the majority or plurality of the Breitenmoser chronologies are moisture-sensitive or mixed sensitivity based on e.g. St. George 2014 and the original a quick look at the Breitenmoser paper - so, does VSL do really well with more (% wise) T or M limited sites?  Are mixed sites generally not as well simulated?  Some of this is likely already part of the original Breitenmoser paper, but this is useful information when evaluating where the observations and simulation (e.g. Figure 4) agree or disagree and what might be the potential reasons behind this. 

Particularly for moisture, there is the question of the seasonality of the climate response vs. the seasonality of tree growth.  For instance, in western North America and the Mediterranean, winter/spring moisture will be important for growth, while in Northern Europe and other parts of North American, annual or summer moisture will control moisture-sensitive tree growth.  The extent to which VSL can do this adequately would seem to be key to making the connection from climate forcing (e.g. volcanism) to local climate to tree growth with as much confidence as possible.  I was also surprised (e.g. Figure 4) by the lack of chronologies further to the west (the Great Basin, Sierra, California, etc) - these are some of the most moisture sensitive sites in the world - why are they not represented here? Is this a VSL problem? A model simulation data/bias limitation?   

2. Regarding Figure 4 and results shown there: Are all the locations shown in these maps really places where (1) VSL successfully simulates the chronology/ies at the location and, (2) where there is a true T or M limited site?  I ask because I find myself surprised, for instance, to see apparently T sensitive sites in mid-latitude or arid North America and parts of the Mediterranean, and note in particular that several of these T-sensitive sites show increased growth post eruption, suggesting perhaps these are not simple temperature sensitive sites in the real world (observations)?  Whereas the simulation shows (as expected) a growth reduction everywhere.  I wonder if the difference in observations and simulations for T sensitive locations can be explained by the strength of the confidence that some of these are really temperature sensitive?  Again, I look at North America and find myself wondering if many of those mid-continent sites are sufficiently temperature sensitive to be confident they can be compared to VSL limited by temperature alone.  Or, put another way, VSL (driven by climate) will have a strong temperature-mediated growth response if the parameters and local climate make the simulation at that location temperature sensitive (and, this also leaving aside landscape-scale changes in sensitivity, e.g. differential tree growth response in the same grid point - Bunn et al. (2018). Spatiotemporal variability in the climate growth response of high elevation bristlecone pine in the White Mountains of California. Geophysical Research Letters, 45(24), 13-312.).  

As well in Figure 4, there seems to be several important and interesting mismatches for moisture sensitive sites as well - for instance, for Crowley et al. eruptions (left and right columns) in North America the simulations show drying/reduced growth in the northeastern United States and a negligible response on the central and western part of the continent, while the observations show the opposite - e.g. a negligle signal in the eastern/northeastern part of the country, and a wet anomaly in the central/west.  The authors do note some of these features (Lines 280 to 286), but what stands out to me for the purpose of this manuscript is the differences between simulations and observations even when the same forcing dataset is used in North America in particular.  Perhaps though the most consistent signal is indeed the European dipole (wet/more growth in the Mediterranean, drier/reduced growth in Northern Europe) - this latter feature somewhat consistent with Fischer et al. 2007 (10.1029/2006GL027992) and more so I think with Rao et al. 2017 (10.1002/2017GL073057) who look at PDSI.  

3. Figure 5 - given the inconsistencies in simulated vs. observed patterns particularly for moisture in North America, how much of the detection for moisture is being driven by the largely successful observed vs. simulation Mediterranean vs. northern European pattern?  The caption says that the moisture D&A refers to 'aggregate mean response grouped by the two regions of homogenous response indicated in Fig 4', but nothing is indicated (should there be a box or the region otherwise outlined?), and it isn't clear from the text alone (e.g. around Line 280) - given the mismatch in North America I note above and evident in Figure 4, I think the statement about detection and attribution in Line 310 and onward should probably be caveated - I suspect (and would ask the authors to establish if this is the case with some regional tests) the signal and successful Moisture D&A is being drive by the Mediterranean/European pattern - the authors can also consult Fischer et al. 2007 and Rao et al. 2017. 

Minor comments:

Line 113: Just to verify: these are all tree-ring width data, and no density data correct, in Breitenmoser?  

Line 119: suggest changing to 'As input to VSL we use the ...'

Line 159: suggest also citing the first paper on this, Cook, E. R., Briffa, K. R., Meko, D. M., Graybill, D. A., & Funkhouser, G. (1995). The 'segment length curse' in long tree-ring chronology development for palaeoclimatic studies. The Holocene, 5(2), 229-237.

Line 340: should probable add a citation near here to Stevenson, S., Fasullo, J. T., Otto-Bliesner, B. L., Tomas, R. A., & Gao, C. (2017). Role of eruption season in reconciling model and proxy responses to tropical volcanism. Proceedings of the National Academy of Sciences, 114(8), 1822-1826.