Thank you for responding promptly to the range of comments made by the reviewers. The reviewers provided positive comments about the new data being produced here (new diol data, new temperature reconstructions being tested, important area to understand climate changes).

Both reviewers have commented on the manuscript discussion of previous work using long-chain diols (and the long-chain diol index) as a proxy for past temperature changes in lakes.
In the author comments a number of changes which could be made to the text are suggested; if undertaken as indicated these should provide both a more detailed and more nuanced discussion of what we know so far about these lipids. There was concern from the reviewers about the identification of likely producers, which the authors indicate will be removed in a subsequent draft. This should help to streamline the discussion.

There was also a concern from the reviewers about the statistical treatments of some of the data (in terms of its statistical analysis, and in terms of the strength of the statements about relationships being made). The authors have detailed several ways in which these concerns could also be handled in a revised draft. There was a particular concern from Reviewer 1 (point 11) about the analysis undertaken in Figure 3, particularly that there are four temperature data points for each single LDI value, but the regression is undertaken through all data points on the plot. Although the authors comment that the result is the same whether 4 regression lines or a single regression line is used, I share the reviewers concern that it would be better to show the 4 separate regression lines and their residuals (or the regressions from each of the two groups of data – lines 297-306, and Fig 3 caption). From the output of multiple different regressions, if a single preferred calibration emerged then that could be used for the reconstructions; if a preferred calibration line can’t be determined, then wouldn’t applying >1 calibration line to the LDI data be a better reflection of the uncertainty on the temperature reconstructions?

Finally, reviewer 2 commented on manuscript line 27 “what do you define as extreme responses”. I would recommend that to avoid confusion the sentence ought not to include the word ‘extreme’ – would ‘abrupt’ work better here instead?

I have carefully read your replies to the reviewers, and your revised manuscript. I thank you for responding thoughtfully to the comments made by the reviewers.

I have made some suggested minor edits to your manuscript in the pdf which is attached here, to improve the clarity of your arguments.

Two comments raised by Reviewer 1 which would still benefit from some clarification:
1) extrapolation. Rev1 notes that because your calibration data set are for the 20th century, the LIA data fall outside of that range. As a result, you are extrapolating the data from your calibration, to values for LDI which you did not record in your data set. (see reviewer comment: “Almost all of the fractional abundances of the 1,13-diols in the older sediments are higher, and almost all of the fractional abundances of the C32 1,15-diol in the older sediments are lower. As a result, the reconstructed temperatures before 1908 are a result of extrapolation of the dataset, and one might even argue that the LCD distribution was significantly different in the samples before 1908.” This is a common problem when a local calibration is generated, but then the down-core data move outside of the modern range for that site. The concern is that this could introduce error in down-core reconstruction because you don't know that your calibration definitely applied in the same way outside of the calibration range (maybe it is non-linear? maybe the errors are larger?). In your manuscript, you should try to add a line when you discuss the cold temperatures of the LIA, to remind readers that these data fall outside of the calibration data-set, and further study to extend the range of temperatures in the calibration could help to resolve any uncertainty here.

2) Figure 5. Rev1 noted that the linear / smoothing did not always seem to display an alignment with the original data. My understanding of what you have done with this samples, is take the age of each LdRS sample (e.g. 1950-1957) and found the same data in the other data-sets, and averaged those too. Is this not putting a lot of emphasis on your age model accuracy and precision? Would it not be better to show a running mean of your data, with the same value of the smoothing as the sediment data (e.g. a 7 yr smoothing)? That way, if the age of your sediments was actually 1951-1958, you would still see that in the smoothed data? This is how we handle longer time-series data, where the age models are more uncertain. For example, in reading your methods section I see that you made an extrapolation below 14 cm for the short core: so I’m not convinced that averaging individual, precise time windows on the instrumental data is appropriate.

Dear authors,

Many thanks for your replies to the minor comments I raised for your revised manuscript.

In relation to the comment on the smoothing for Figure 5: I am satisfied that you have explored the possibilities of the smoothing functions, and evaluated the impact on your data-sets. You have demonstrated that comparable relationships exist with the different functions, and that each have their own issues. I am therefore happy to accept the uploaded version of the manuscript without seeking further changes to figure 5.

Two final minor technical corrections:
1) in the final version of the manuscript, please ensure that the title of your Supplemental information matches that of the main manuscript (the latter has changed during the review process).
2) please carefully check over the grammar of the additional text which has been added through the review process, to ensure that there are no typos/errors.