Dear friends and colleagues,

I have read through the manuscript several times, as well as the comments by the two referees (Kump and Sexton) and your responses.

First off, this is a nice piece of work. I stress this because, at times, my commentary comes across as pretty critical. My goal as an editor and reviewer is always to make the end paper as good as possible.

Lee makes a very insightful comment concerning the d13C vs d18O gradients – these may be telling us something about the carbon source and its dynamics. Unfortunately, both Lee and Phil, at least my opinion, miss the larger issue by making a bad suggestion of comparing the supposed temperature response (d18O) to the supposed carbon input (d13C), a problem that is confounded by the writing and presentation in the manuscript (and many other manuscripts). I elaborate on this point below.

I am also not a great fan of currently labeling and lumping all identified CIEs by polarity chrons as suggested by Phil. This is because it is not obvious, at least to me, that all such events are real, globally significant or similar. Moreover, I am not so sure that the paleomag is correct at Site 1258. I would very much agree with him that the situation is somewhat confusing right now, especially to those who do not know the details, but until we have really clean records at multiple sites, I see no point in potentially making things even more confusing in the long-run.

Most of the other comments are reasonable and can be addressed.

I will accept the manuscript pending modest revisions, and sincerely hope that you change certain aspects.

Gerald (Jerry) Dickens

****
Major Problem:

The overarching problem behind this manuscript (and the referee comments and unfortunately many published papers) lies with a premise stated on page 1797, namely that the short-lived warming events were DRIVEN by the release of 13C depleted carbon. There remains ZERO evidence that this is the case, at least to my knowledge (although I am open to edification). Clearly, the signals for warming and carbon input are somehow coupled (as nicely shown in this manuscript). However, all information to date, admittedly still sparse and debatable, suggests that warming DROVE in part the carbon input, at least that recorded by the negative d13C excursions. Until someone provides solid evidence that the carbon inputs CAUSED the associated warming, this sort of writing and discussion of data absolutely should be avoided. Indeed, because of this, the discussion (e.g., Page 1803) becomes inconsistent and awkward. Think: how does one invoke orbital changes, presumably in solar insolation, that cause carbon release to drive all temperature change? It almost has to be a coupled system where the carbon release was a POSITIVE FEEDBACK TO WARMING BUT NOT A FORCING OF WARMING.

This is crucial, because it leads to two very important concepts. First, the change in temperature should always exceed that expected from mass balance calculations for the carbon input (as emphasized 18 years ago for the PETM, Dickens et al., 1997); second, and more interesting to the current manuscript, the gradient between d18O and d13C should increase if the time between successive events is significantly less than the recharge rate of the carbon source from which the carbon derived. Both these effects appear in the data of the manuscript, but because the background through presentation sets things opposite, all becomes messy, as one would expect with a wrong framework.

The very first things to change are the axes of Figure 5. It should be d18O on the x-axis (driver) and d13C on the y-axis (response).

The next things to change are to provide a proper set-up to the problem and to the discussion.

Additional Editor Comments

(A) The manuscript has EECO in the title, and briefly defines this interval as 52-50 Ma. As pointed out by Slotnick et al., 2012, the use of EECO, especially with these dates, causes confusion. In fact, the manuscript, with this definition, discusses an interval that occurred 2.4 to 0.4 Myr before EECO so the title makes no sense!

The root problem lies in how EECO first came to widespread usage (Zachos et al., 2001), where it was given an age but on an outdated time scale and where some key information was not aligned correctly in the depth domain. However, as one can see in the records of Zachos et al. (2001), the relative age for the start of EECO is clear – it begins approximately at the low in d13C, which we now know is the K/X event and at nominally 52.8 Ma.

In records from the Clarence Valley, New Zealand, there is a marked change in lithology that begins at the J-event, presumably when background temperatures finally passed some threshold. Slotnick et al. (2012) thus suggested that the easiest way to think about EECO is that is began after a “step” coinciding with the J-event, but is characterized by variance (and many CIEs and temperature variations) throughout the next 3 or so million years. Indeed, the data in this paper is really good for discussing EECO – the long-term rise in deep-ocean temperature appears to stop at the J-event. This would be absolutely consistent with the way Slotnick et al. presented things.

Anyway, the presentation of EECO should be fixed and explained, so it does not perpetuate confusion and the title makes sense.

(B) The writing in Section 5.1 is not good for multiple reasons, all related to my primary criticism. First, in the original modeling, the “carbon capacitor” does not have to go to zero, if the long-term temperature rise stops; rather, the answer depends on the starting carbon mass, and the “recharge rate” (i.e., the fluxes in and out of the capacitor, Dickens, EPSL 2003). However, the EECO world would be in a situation where large CIEs could not happen.

Second, it is by no means clear that any of the post K/X CIEs were hyperthermals in the current, albeit poorly defined, sense. In fact, if one looks carefully at the Site 1258 record, one will see that, unlike the H-1, H-2, I-1, I-2 and K/X, many of the later CIEs are not associated with a MS spike as expected for dissolution. Of course, there is also the problem that the K/X event is not really recorded in the Site 1258 data.

Third, a change in the d18O vs d13C “gradient” should be what happens if one accepts that temperature is driving carbon input (at least in part) and there is a carbon capacitor. In fact, the change in gradient should be related, albeit in a complicated manner, to the recharge rate. Basically, for a given temperature change, and over a time frame shorter than the recharge rate, the amount of carbon released has to be less the second time.

This section really should be rewritten and rethought.

-- Page 1796 --

Lines 24-3: This sentence is awkward in writing and concept. The warming trend is not only in deep marine records, but also in high latitude surface water records (e.g., the cited Bijl et al., 2009 paper); it also culminated in the EECO. Should be something like “… trend that culminated in an extended period of warmth …”

-- Page 1797 --
Line 8: Should be something like “… conditions and rates of change (REF).”

Line 12: Should be “rose”.

-- Page 1800 --
Line 6: Probably should be “… the benthic foraminifera values (d18Oc) … “ to be crystal clear.

Line 10: It would be good to briefly introduce two usually omitted issues (and perhaps even discuss them later). These are the assumption on d18Osw, and that whatever fractionation that causes an offset from equilibrium is constant in the time domain.

Figure 1: This can be improved, mostly because it conveys limited information relevant to the text. Why not add the locations of other sites that have relevant information across this time interval and are discussed in the text.

Figure 3: The Stap data is not obvious in some of the records.

Line 1819 – should be “at” not “in”

There is an important figure missing: the data in the depth domain. There are far too many records in far too many papers that jump to the time domain without clearly showing how things look in the depth domain first. For example, the record at Site 1258 is either spliced wrong or sampled wrong, as there is no significant change in d13C across K/X.

There are several typos and grammatical oddities, but the present manuscript is significantly modified according to referee commentary, and at a level to proceed to publication. I encourage all the authors to carefully read at least one more time and fix the silly errors.

Gera-L-d

Will be a nice paper.