Dear Douglas Finch and co-authors,
I have struggled hard with this revised manuscript, but I’m afraid that it still falls down in one substantive way, and in a number of less substantive ways that need correcting nevertheless. I would like to see a further revision of the manuscript before making a final decision on whether the ms can go forward to ACP.
Please expand section 2.3, paying particular attention that the stylistic difficulties in other parts of the ms are not reintroduced here. Please start by defining whether L_{C,i} is constant or not in eqn 1. If it is a true constant in this ms, then I can’t see why it remains in eqn 1 and a simplified form should be given. This will mean that A_bar is independent of the chemical lifetime of CO and so there is no need to speculate on an appropriate value of OH.
Please then introduce the photochemical age expression, starting with equation 2 of Parrish et al. 2007 and making any changes to this explicit. In this way the reader can see immediately where OH concentration comes into both lifetimes and how and why the values might differ. It is essential to have this in order to understand the significance of the comparisons made later.
Please then take the reader through limiting-case examples of how A_bar and the photochemical age expression relate, so that we know under which conditions we can expect one to exceed the other. Is the difference always one-sided (eg A_bar > photochem age, always)? If A_bar is some combination of the mixing-ratio--weighted advective and diffusive transport time, and the photochemical age takes into account chemistry *and* diffusive mixing what does one deduce from a comparison? If A_bar is greater than the photochemical age, how do we decide whether this is because the advective transport in the model is not accurate, or the diffusive mixing is not accurate, or the OH is not what was assumed in the photochemical age calculation, or a mixture of all of the above? I’m sure you’ve thought of all this, but I’m afraid the ms does a poor job of taking the reader through your thought processes logically and without sudden jumps.
If you feel that all of this would break too much into the flow of the ms, then please supply the info as an appendix.

Further comments:
Abstract, line 5. Humour a cranky old editor: there are well-attested psychological reasons that make the use of parentheses to hold two contradictory meanings in a single sentence bad grammar. It really wouldn’t add much to the length of the ms to say (Spearman’s rank correlation, r_s = 0.65; the model has a positive bias for observations < 100 ppb, and a negative bias for observations > 300ppb). At the very least please correct to make sure that the positive bias clause doesn’t refer to r_s, as it does currently, and that the parentheses are closed properly (currently you are missing the final right close bracket).
Abstract, line 6: “NW North American”? I know what you mean, and perhaps that’s the main thing, but would “biomass burning from the far northwest of North America” not remove the clunkiness?
Abstract, lines 13-15. So much care is needed in the writing here. Does the sentence “We find that…” mean that the median size of delta_tau is 3 days, or that the difference in delta-tau varies with A_bar?
Equation 1 (see above): maybe I’ve gone maths-blind, but if you assume a constant chemical lifetime of 60 days for every tracer, does L_{C,i} not cancel out of this equation? I can see why you might want to keep it in just to show that you could have variable chemical lifetimes for the tracers (although then how would you compare to the photochemical age?), so perhaps just add a second, simplified, form or eqn 1 in-line? If what I’ve just said is right, I think you need to amend line 9 of the abstract, because A_bar does not depend on chemical decay.
Lines 129-130. I think you will frustrate readers with sentences such as that beginning “Mean and median…”. There is a huge difference between the mean and the median, and a difference in sign. Surely, some statistical preliminaries are in order? From the point of view of the skew distribution represented by the median, the model does *not* have a positive bias. Given that mean and median say something different, what then are we to make of the following sentence: “We find…”? Sending the reader to Figure 2 to work all this out is bad practice in my view.
Line 134: the mode is coincident with the 25th %ile? Gosh, I’d find a way of reassuring the reader that this is not a typo, for example: “The BORTAS-B observed CO has a highly skewed distribution with a mode of 87 ppb coincident with…etc”. I think it is important to be clearer too, about in what way this (the distribution? The mode? The interquartile range?) is consistent with ARCTAS-B.
Line 143ff: since you use quantitative measures of model skill prior to, and subsequent to, this statement, it might be prudent to say “We use Spearman’s rank correlation in this case…”
Line 168: I’m not sure you mean that the East Asian CO source is small per se, just that it makes a small contribution to the BORTAS-B measurements – is that right?
Line 201. Better I think to say “2010 and 2011 have median ages of five days”
Line 205ff: Reporting ranges of medians was one of the criticisms picked up by a referee. Could you not say “Considering each of the year and altitude distributions at the eastern boundary, medians of the distributions were similar (11-13 days) and interquartile ranges were…” ?
Line 207 “decreases with increasing altitude” – just to be absolutely clear?
Line 208: “The age…gets” – agreement of subject and verb. “Emissions clearly get older…” says the same thing and more succinctly.
Line 221. I’m afraid I can’t link the sentence beginning “The longitudinal…” to the figure. What is peaking? This sentence needs completely re-writing. Talking about peaks in Fig 7 will draw the reader’s attention to the blue areas – the red areas are local minima!
Line 236: I would avoid saying “generally consistent” and then discussing “substantial differences” 3 lines later. It is better to be precise about which distributions look alike and which don’t.
Line 238: I would avoid reporting, and especially comparing, ranges (I’m sure I’ve said this before). The percentiles are much more robust statistically. To discuss absolute maxima and minima properly would require a detour into extreme-value stats.
Line 253: “median difference”, surely? Why do you go straight to the fourth moment of the distribution (kurtosis) without mentioning the third, the skew, which explains the difference between mean and median?

Dear Douglas Finch, Paul Palmer, and Mark Parrington,
I’m sorry to say that I am going to ask you for some more clarifications before accepting the manuscript for publication in ACP. I’m sure this will be extremely frustrating for you, but please understand that I have at heart a desire to make the paper as persuasive and, hence, impactful as possible.
Please amend section 2.3 to make it absolutely clear how equation (1) is implemented in the model and to clarify the relationship between equations (1) and (2). My comments below will give you some steer as to the places where I lose the logic.
In equation 1, I presume the sum is over the number of tracers in a model run. So, for a run 30 days long, i = 30.
I presume that A_i is a real number calculated at every model time step (i.e., not an integer as could be inferred from the current text).
I presume that L_{m,i} is the value of the tracer, i, at a point, and that the chemical loss is multiplied in afterwards (online or offline) using L_{C,i} = L_{0,C,i}exp(-A_i/tau_i), where L_{0,C,i} is some (arbitrary) initial value and tau_i is the photochemical lifetime defined in the paper as 60 days. Tau_i presumably is given by 1/{}, where is the characteristic rate quasi-first-order loss rate for CO by reaction with OH. There is, therefore, a clear link between equations 1 and 2 relating to how is estimated by both approaches, and what assumptions allow to be separated from in equation 2.
The constant value of implied by the use of a constant CO lifetime must also imply a constant value for but this is not reported. Being clear about these values is important because of the relationship between equations (1) and (2).
The left-hand equality of Equation (2) defines t_a. as a concentration-time; the photochemical age is t_a only. This is not clear from the current text. is the time-and-space averaged OH concentration necessary to produce the observed linear correlation between the chemical tracers, and has been separated from the co-varying rate coefficient, , somehow.
is not OH-weighted so far as I can see. It is the space-and-time rate coefficient corresponding to the air parcel histories that determine and the tracer correlation, where the space-and-time averaging takes into account the temperature (and pressure, if relevant) dependence of the rate coefficient. If your L_{C,i} where space-and-time varying, it would give you a way of looking at but the manuscript as it stands uses two slightly different implicit constant values in (1) and (2), which probably makes understanding the differences between A_i and photochemical age difficult.
The manuscript seems to suggest that the photochemical age approach is highly uncertain because it can vary a lot with the choice of . This underplays the strength of tracer-tracer correlations, which contain all the real atmospheric processes in all their complexity. T_a. is very tightly constrained by the measured tracers; perhaps the model should be attempting to compare with this metric rather than t_a.

Dear Douglas, Paul and Mark,
Many thanks for your painstaking work to make sure the ms is as clear as can be. I am very pleased to accept the ms for ACP. It seems to me that there is more to be said about the exact relationship between the two measures of age that you use, but that is a topic for further research in the community at large now. I hope your paper is well-received and widely read,
Sincerely,
Rob