Tarroso et al. Spatial climate dynamics in the Iberian Peninsula since 15 000 Yr BP
Revised manuscript
The authors appear to have taken on board many of the comments made in the first round of review. However, there are still substantive problems remaining. Three of the most important are:
1. The methodology still remains opaque. From the methods presented it would still not be possible to independently reproduce the results of the study from reading the paper and accessing the resources it refers to. As a minimum, the paper should include enough information so that the study could be independently reproduced. This is particularly important in this case because the method appears to differ significantly from previous pdf methodologies that are otherwise well documented. Even if we take on trust that the R scripts can be obtained from the author (although it would be better to upload these to a recognized R script archiving site), the authors still do not provide information about the taxa list and how they resolved the not insignificant problem of equating modern distribution data available at species level with pollen taxa which are of variable taxonomic resolution. Until this information is made available, it is not possible to say that this has been done satisfactorily.
2. Critically, there still remains no proper evaluation of the reliability of the technique, and no evaluation of potential error in the reconstruction through time. This is such a fundamental part of transfer-function development, let alone the scientific method, that I am somewhat shocked to find the authors scrabbling around to put together some kind of evaluation at this stage. It is great to find so much effort has been put into developing a new method, but ultimately you still need to ask the question is it any good? how do we know it is any good? can we quantify how good it is, and can we make that assessment back in time? The evaluation provided by the authors in this revised manuscript is not sufficiently rigorous to provide this information, and would not be acceptable in comparison with previous studies where such evaluations have been provided. The authors have only to look at the literature associated with other pollen-based climate reconstruction techniques to see what is required, and particularly those based on the pdf method.
3. The English has deteriorated considerably from the first draft, to the point that almost every other sentence (including figure captions) has some issue or other. These are far too numerous to correct, but absolutely need to be corrected before the paper could be considered for publication. I believe the co-authors should be asked for help here, or a professional proof reader employed. At times the entire meaning of the sentence is not clear, and this has made the task of reviewing the paper more difficult.
Main comments:
p3, Section 2.1: Still needs taxa list, see 1) above
p3, Section 2.1: The acronyms for the temperature parameters Tjan and Tjul are not self descriptive, something like Tjan_min and Tjul_max would be better. However, Aprc should definitely not be used and should be replaced with the standard Pann, which is also consistent with the chosen temperature acronyms.
p4: lines 198-232: I still find the description of the method difficult to follow, and as a result quite opaque. Perhaps this may have something to do with the English terminology used, which may mean something different to the average reader than the author. I would suggest that the author asks someone unfamiliar with the technique to read the passage and then recount (without prompting) the method to test the effectiveness of the description.
p4, lines 233 onwards: The authors now include an evaluation of the method. This is described, but the results of the evaluation are not discussed. What is the average error for each climate variable, and what is the standard deviation? Are there any systematic errors? (for instance, reconstructed summer temperatures look systematically cooler than expected). The method used to assess these errors is very poor. For instance, the comparison of samples within the last 500 years against a modern baseline does not take into account changes in climate over the last 500 years that would undermine the comparison even if the transfer function worked perfectly. Critically, no uncertainties are provided for the palaeoclimate reconstruction back in time. It seems to be a serious failing in the development of the method that no consideration was given to assessing transfer function performance.
p4, line 238: do you not mean ‘all samples younger than 500 years BP’ rather than ‘inferior’?
p5, line 243-251: How was the 100 BP baseline calculated? Is this based on eg historical climate data, the top sample of each core or a gridded reconstruction? This is not described clearly.
p5, line 252-254: How are the 1000 year timeslices defined? For example, using a time window (+/-500 years), or exactly every 1000 years using the interpolation between samples? Please explain.
p6, line 297-300: The authors apply an unusual evaluation of uncertainty. The description (Appendix A) is written in poor English ‘A set of 250 replicates were performed by sample randomly one reconstructed age within 500 years for each site (gray lines) and than averaged.’ but I presume this means that they iteratively refitted a linear regression line 250 times after randomly removing one of the samples in the training set on each iteration. This is not a normal cross-validation method, which involves the ability of the model to predict an individual missing sample and not some kind of global regression relationship. The evaluation shown by the authors only demonstrates that the sample values are symmetrically distributed about the mean, and that with enough samples in the dataset the removal of one sample does not unduly disturb the regression line. The ability of the transfer function model is not based on a regression line, but the ability of the individual samples to reproduce the present climate, which should be a one-on-one relationship. It is the deviation from this one-on-one relationship (residuals), which is the measure of uncertainty. If the present day January Tmin climate at the sample site is -2 but the reconstruction gives a value of -8.2, then the reconstruction is in error by -6.2. It is common to report some kind of global average error based on all of the samples in the evaluation training set, but this is not the same as reported by the authors. Looking at the figures, the ability of the reconstructed values to reproduce the modern climate on a sample by sample basis indicates substantial errors. These are probably no worse than other methods that have large uncertainties (eg inverse modeling), but they should be properly evaluated and reported using a commonly used methodology. This evaluation should also be extended to the fossil samples, so that each fossil sample also has an uncertainty. See Norbert Kuhl’s work to see how this could be applied to the pdf method.
p6 Section 3 Results: There is a long description of the climate reconstruction results based on absolute values. As I mentioned in my first review, and indeed this was also mentioned by the second reviewer, the use of absolute climate values rather than anomalies greatly reduces the relevance of this section to the average reader. I respect the decision of the author to insist on presenting their results in this form, and can understand the relevance to bioclimatic parameters, but I would again strongly recommend presenting the anomalies first, and absolute values second. Even just showing modern day values for the different regions in figure 4 would help.
p7, line 366-374: Comments about the evaluation of the method would be better placed altogether in the methods or results section rather than spread about the text. What is the ‘historical’ climate referred to on line 368? is this the modern climatology? if so, the word historical is very confusing, if not then its use needs to be explained. What is meant by a ‘significant linear trend’? on line 367, as discussed previously the importance is not that the line is linear, or that there is a regression line at all, only the deviation from the expected fit and the spread of the residuals.
p7, line 383-385: As was mentioned in the first round of reviews, why does winter temperature rise in response to a falling trend in summer insolation? the authors need to explain their logic.
p7, line 407-410: Again, this was already mentioned in the first round of reviews, the authors need to explain their logic better here. The comment “Summer temperatures, on the other hand, provide enough energy to plant growth (Sykes et al., 1996), and are likely resulting in less responsive July temperature.” does not make sense, neither grammatically or scientifically. You might have a good point, something about plants becoming insensitive to progressively higher temperatures because their energy needs are non-linear with respect to temperature, but you need to explain and justify it better.
p7, section 4.1: ‘OD’, ‘BA’ what are these undefined acronyms? I take it that they are probably something like ‘Older Dryas’ etc but such things need to be defined for the reader. If it is the Older Dryas, then the date provide on line 416 is wrong (~8 to 14.7 ka). Also section 4.2 etc ‘YD’ etc
p7, line 425: ‘extreme January temperatures’? ‘extreme’ depends on your perspective, an anomaly of -5 doesn’t look very extreme compared to other parts of Europe.
p8, line 445: I think you mean ‘depending on the location in Northern Europe’ since as you say later in the sentence, Southern Europe experienced a cooling at this time.
p8, line 449-451: “Our results point to a decrease on Tjul temperature but a stable minimum temperature, indicating mild summers.” doesn’t make sense, even if you meant ‘in’ not ‘on’. Are you saying that Tjul temperatures were lower than present? and what does a ‘mild summer’ mean? colder or warmer than today? when talking about minimum temperature, do you mean your Tjan reconstruction? Please be more precise in your use of language.
p8, line 473: Seppa and Birks only refer to Northern Europe, the ‘warm period’ experienced in Northern Europe is not typical of Europe as a whole and particularly Southern Europe which experienced a cooling.
p9, 532-534: Appendix E is not well explained, but I presume this is a plot of the maximum temperature change for each grid point arranged by altitude. This suggests that each grid point represents an independent observation, but in fact this is based on interpolation from just a few sites often distant in both horizontal and vertical space. I think you could make some general inference at an aggregate level, but to make a plot of the grid points like this without any consideration of both the interpolation and reconstruction uncertainties is very misleading.
Data archiving: are the reconstructions and associated uncertainties to be made public through a data repository such as NOAA paleoclimate or PANGAEA? Also it would be very helpful if the pollen data that is included in the study but is not publicly available is made available through the European Pollen Database.
Finally, the English needs extensive correction, far too numerous to list here since almost every other sentence needs attention. This has deteriorated significantly from the first draft. Perhaps some of the co-authors can help here, since their English is excellent, else the manuscript needs to be professionally proof read. Here are just a few examples on page 2 alone:
p2, line 57: “with a strong relationship with climatic” with a strong relationship with climate
p2, line 71: “understanding this climate dynamics” understand climate dynamics
p2, line 73: “this intimate relation” this intimate relationship
p2, line 75/76: “and is, thus” and are therefore
p2, line 81-83: “the potential location of suitable climate favouring long species persistence and serving as refugia” grammar
p2, line 94: “milder climate than the northern” milder climate than northern
p2, line 127: “taxon are created” taxon were created
p2, line 127: “distributions in the cl |