cc: katie.farnsworth@iup.edu, lsmith@geog.ucla.edu, P.Jones@uea.ac.uk, kxu@vims.edu date: Mon, 10 Sep 2007 22:02:58 -0400 from: John Milliman subject: Nature manuscript 2007-07-07125 to: j.mossinger@nature.com Dear Dr. Mossinger: After several days of mulling over the two reviews of our paper, "Climatic and anthropogenic factors affecting river discharge to the global ocean, 1951-2000", it becomes increasingly clear that neither you nor my co-authors and I were well-served by one of your reviewers, and that, combined with some points of confusion by the other reviewer, resulted in your decision to reject the paper. We did submit an earlier version of manuscript to Science, which was rejected in part based on the comments of Reviewer #1. Parenthetically, we saw no reason by inform Nature of the manuscript's prior history nor changes we had made to the manuscript based on those reviews. We did, however, take the Science reviews into account as we revised manuscript for Nature. As a single example, Reviewer #1 requested that we compare our trends with those obtained by Milly et al. (Nature, 438, p. 347) for a 100-yr period, something that a second reviewer also suggested. In fact, there were several valid reasons for not initiallyl comparing our results with either Milly et al. or those of Labat et al. (Adv. Water Res., 27, -p. 631), the most obvious being that their "100-yr" records were strongly biased towards North American and northern European rivers, compelling the authors to rely on questionable models to extrapolate long-term records for rivers from the less-developed countries (e.g., Asia, South America, Siberia and Africa). Realizing, however, that other readers might also wonder why we did not refer to the Milly or Labat papers, we added both new words and footnotes to p. 2 of the Nature manuscript; note particularly footnote (8): Although 50- to 100-yr discharge records are desirable, access to globally distributed long-term data remains problematic. Of the ~650 rivers listed in the Global Runoff Data Center (GRDC) database that discharge directly to the ocean, for instance, only 20 records extend back to 1900, only two of which (Karun and Nile, both of whose accessible records cease in the mid-1980s) lie outside northern Europe or the USA. Attempts to compensate for the lack of long-term empirical data include wavelet-based runoff reconstruction (6) and climate models to simulate regional discharge (7). The lack of data (8), however, calls into question such reconstructions (9, 10). 6) Labat, D., Godderis, Y., Probst, J.L. & Guyot, J.L. Evidence for global runoff increase related to climate warming. Adv. Water Res. 27, 631-642 (2004). 7) Milly, P.C.D., Dunne, K.A. & Veccchia, A.V. Global pattern of trends in streamflow and water availability in a changing climate. Nature 438, 347-350 (2005). 8) The 100-yr trends for the 165 rivers in (7), for instance, are based on a median of 59 years of observational data; the median record for Asian, African and South American rivers in that dataset may be closer to 50 years. 9) Legates, D.R., Lins, H.F. & McCabe, G.J. Comments on "Evidence for global runoff increase related to climate warming" by Labat et al. Adv. Water Res. 28, 1310-1315 (2005). 10) Peel, M.C. & McMahon, T.A. A quality-controlled global runoff data set. Nature 444, E14 (2006). That Reviewer #1 did not note these rather significant changes (which appeared in the second paragraph of the Nature manuscript) suggests that he did not read it ("Therefore, my review is the same, and if the authors in fact have made changes, they should indicate specifically what and why to the Nature editor."). In my 40 years of publishing and 23 years editing (Deep-Sea Research) I have never come across a reviewer who submits a review without reading the paper in question. It may be a first for you also. Many points raised by Reviewer #2 seem valid; his insights are greatly appreciated and will be addressed fully in a revised manuscript - no matter where it is eventually published. However, several of his points, we feel, are off-base, four of which I cite below four (in blue), together with our responses in brackets: 1) The reviewer has some problem to understand why the authors discuss the increase and decrease of trends for rivers without statistical significance. In Table S2, only half of the river basins show statistically significant increases or decreases. Therefore the reviewer is puzzled to see a sentence like "local and regional changes were significant" at line 3, page 5. The reviewer recommends discussing the long term trends only for statistically significant river basins. [With the exception of the Yana River all deficit and excess rivers showed a statistically significant change in runoff and/or precipitation. In contrast, 21(out of the 31 normal rivers show no statistically significant change in either runoff or precipitation. But this should not be surprising: if normal rivers are defined as rivers in which DR reflects DP, is it not logical to assume that temporal trends for either both or neither should be statistically significant? To discuss long-term trends based for statistically significant river basins , as suggested by the reviewer, necessarily negates all those rivers (including, for example, the Amazon, Yangtze, Rhine, etc.) for which there was no statistically significant trend in either DP or DR. To that extent we may have added to the reviewer's confusion by using the word "significant" in both a statistical sense and a non-statistical sense; substituting, where appropriate, a word like "relevant" might help eliminate this confusion.] 2) It is also strange to compare the linear relationship between DP and DQ by percentage, as Fig.2 or the text at the middle of page 5. As far as the reviewer knows, there is no geophysical theory that DP and DQ are equal in long term trend under natural condition. Recalling the Budyko's empirical curve of the water balance considering the available radiative energy, the relationship between DP and DQ is non-linear. Of course, any functional relationship may be able to be approximated as linear, the changes discussed here is more than a few tens of percent and applying perturbation concept would not be appropriate. [Nowhere did we infer that DP and DQ are equal (the DP and DQ axes in Fig. 2 are not equal) or that they correlate linearly; the purpose of Fig. 2 is simply to show global and latitudinal similarities in the two trends, the notable exceptions being the deficit and excess rivers identified and discussed in the paper. (Parenthetically, this figure may not be as essential to the paper as the other figures; if space is a problem and if the editor were insistent, we might be "persuaded" to eliminate it.)] 3) Since global river discharge from land to ocean is estimated to be approximately 40,000km3/y or more, the article covers less than half of it. Moreover, most of the findings and discussions in the article are changes in individual river basins, and the title "To The Global Ocean" seems not appropriate. [As Fig. 1a clearly shows, our 135 rivers represent a global distribution of temporal trends. Moreover, global basin areas are more or less equally apportioned, cumulative African, Arctic, Australasian, North American and South American rivers each ranging between 8-11 x 10^6 km^2. With the data at hand, this represents just about the best that one can hope for in terms of a global picture - and the task grows almost exponentially more difficult as one attempts to increase the coverage. To achieve 75% global coverage, for instance, would require long-term temporal trends for >250 rivers, including from at least 25 Indonesia and the Philippines, for which there may be no long-term data.] 4) Secondly, there should be longer river discharge record for Arctic rivers, and the authors will be able to show more robust long term trend and would be able to discuss the reason of negative DP and positive DR. [Extending our record for Arctic rivers further back in time, which Peterson, McCllelland and others have done with Siberian rivers (whose records date back to the 1930s), would compromise our synthetic global approach, as the records for many rivers (e.g., Chinese, Brahmaputra, etc.) only extend back to the early 1950s - the reason that we chose the 1951-2000 interval for this study.] We are somewhat perplexed as to our next step with Nature - if, in fact, there is a next step. Your rejection letter seems to have left the door opened for your re-evaluation of a revised manuscript. Any revision obviously would address the reviewers' points (the manuscript that we submitted to Nature, in fact, already had addressed many of the points raised by Reviewer #1). We must wonder, however, whether Reviewer #1 would/could give us an impartial review; I suspect not. On the other hand, we think that we can respond to all of Reviewer #2's comments to his satisfaction. Assuming that you will need a new reviewer, Des Walling (Exeter) read an early draft of this paper, but would serve as an excellent reviewer in terms of his global perspective on temporal trends in fluvial discharge. I might add that I/we never have nor never would submit a manuscript simultaneously to more than one journal. We hope that this manuscript will published in Nature. Sincerely, John Milliman