因此，一份审稿意见应该有方向性的指导，还要有具体的建议，如果能够对作者未来继续进行深入研究有启发性的作用就更好！

 

     在此展示的这份审稿意见，其实是作者根据审稿人的意见对文章进行了修改以后的二审意见。审稿意见洋洋洒洒好几页，非常具体中肯。对这样的评审专家，我们应该表示由衷的敬意！

 

----------------------------------------------------------------------------------

 

This manuscript is considerably improved over the initial submission, but there needs to be greater clarification of the analysis, more attention to important details, and improved organization.  Also, the use of the English language needs improvement in places, but I leave that to the copyeditors to assist the authors.  My comments below are organized by section of the manuscript.

 

Introduction.  I wouldn't say that interpolation techniques "resolve the shortage of observations", as the suitability of these methods is in part a function of the station density. The sentence that reanalysis "most closely estimate the state of real atmosphere" is also problematic as it is not clear what the reanalyses are being compared against. Also, in the introduction the authors state that their goal is to verify monthly 2 m air temperature in the ERA-Interim for the Tibetan Plateau, however, most of their analysis is at the annual and seasonal (rather than monthly) temporal scale.  

 

Section 1.2.  If a time series was not complete how was this handled? Were the missing data filled in with data from neighboring stations? Also, the 5 consecutive years criteria seems somewhat odd, especially as  it appears that only one station had less than 10 years of data.  Why not exclude that station and use a stricter criterion?  Also, the authors should provide the period of record in Table 1 for each of the stations to help readers interpret the results.  In addition, how did the authors deal with the 20 stations that didn't have data for the full period 1979-2010?  Were these stations included in the trend analysis and climatological maps, for example?

 

Section 2.1  Is the ratio of standard deviations for the annual standard deviation? I am assuming that the other parameters, bias, rmse, etc. also are for the annual means of Te and To.  Is that correct?  At this time, it would also been helpful to see a plot of bias and/or rmse against the difference in elevation between the observation station and the ERA-Interim gridpoint.  It is important for readers to see what the shape of this distribution looks like in order to better understand the impact of the lapse rate correction discussed later in the manuscript. This type of graph would also better support the authors' convention that bias is related to the elevation station, rather than providing examples in the text for only a few stations with small and large biases.  Also, are the differences shown in Table 3 statistically significant?  I doubt if they are.

 

Section 2.2 In my earlier review, I had suggested that the annual cycle be removed when correlating the time series of monthly data from observations and ERA-Interim, because the correlation was representing how well the annual cycles agreed between the two datasets rather than how well ERA-Interim was simulating the month-to-month variability seen in the observations.  However, the authors say on page 6 that they removed the annual cycle from the values of annual and seasonal mean temperatures rather than the monthly values.  I am perplexed on how and why they did this.  What method was used to remove the annual cycle?  Or when they say the removed the annual cycle are they just saying that they averaged the monthly values by year and season?  There really isn't any need to remove the annual cycle when calculating and comparing annual and seasonal means.  

 

I have some additional concerns regarding the Section 2.2 on temporal and spatial variability.  Rather than "annual variability" on line 7, I think the authors mean "inter-annual variability", or in other words, year-to-year variability in the annual mean. And by "seasonal variability" are the authors referring to the year-to-year fluctuations in the seasonal means? I suspect so, but they need to make this clear. Figure 3 for station No. 1 is not very useful in demonstrating how well the ERA-Interim is replicating inter-annual variability of the annual and seasonal means.  The bias is very large at this station and the vertical resolution of the graphs is coarse, thus the graphs for both observations and ERA-Interim appear very flat.  Based on the plot, only the winter averages of ERA-Interim show considerable inter-annual variability, but that is partly because the vertical resolution of the winter graph is finer than that for the plots of the others seasons and the annual mean.  Table 4 shows that at station No. 1 the correlation for winter is 0.295 but the annual value is only a little higher at 0.376.  But because of the scale of the graph, the curve for the inter-annual variations in the annual mean appears very flat for both the Te and To series when the correlation suggests that the inter-annual variability of the two series is rather dissimilar.  Also, perhaps a better station to use to show differences in inter-annual variability is station 14, which has low correlations for some seasons but the elevations of the observation station and the ERA-Interim gridpoint are more similar than for Station #1 and bias is much smaller.

 

In the second paragraph under 2.2 it appears that the authors are now looking at bias in the monthly values rather than the inter-annual variability of the monthly values.  This needs to be made clear.  The writing of this paragraph can also be tightened and the paragraph shortened to highlight the differences in bias between summer and spring, and between the eastern versus southern TP.  Again, I question the focus on Station No. 1 given the large difference in the To and Te elevations.  There are also some grammar errors in this paragraph.

 

Section 2.3. I don't understand what the authors mean when they say that the "The monthly lapse rates are obtained from Kunkel (1989) and Liston and Elder (2006)".  These studies were for other parts of the world -- weren't the lapse rates calculated specifically for the Tibetan Plateau?  Or do the authors mean that they used the procedures from Kunkel and Liston and Elder to calculate the lapse rates?  Whatever, it is not clear how the lapse rates were found. The authors need to explicitly describe how the correction was calculated. Table 5 doesn't need to include all the different parameters as the correlation, standard deviation ratios and standard deviation don't change much for the lapse-rate corrected temperatures.  Why not focus instead on the bias and the rmse and show the difference between the values for the corrected and uncorrected series or in other words show the difference between the values in Table 2 and Table 5.  Also, are the bias and rmse values given at the top of page 7 the average values across all the stations? Another question is why would the variation of Tc differ from that of Te? After all, for each month a constant is being added to Te to calculate Tc.  Even though the constant varies by month, I wouldn't expect much difference in the variation of Te and Tc.  Also, what is the "correlation model"?  And why isn't the bias reduced at all the stations, why just at 57 stations?  Does it have something to do with the shape of the relationship between bias and elevation difference? Here is where a graph of bias versus elevation difference would be very useful.  Note that at station #11 Te and To are at similar elevations and the bias was initially small.  What if you stratify the data by elevation?  Do you see that the correction was useful for higher elevation stations but introduced error for lower elevation stations?  And why didn't you also use the Gao et al. method in addition to the Kunkel method?

 

Section 2.4.  It is not clear from the climatology maps what new information the ERA-Interim provided.  What did the ERA-Interim tell you about the temperature climatology that was not previously known from observations only?  Also, the authors need to be more careful when they say that the "ERA-Interim captures the topographic features very well" since the climatology maps are not compared to similar maps prepared from observations.  I also would caution against speculating on the causes of the temporal trends, but rather focus on how similar the trends for the ERA-Interim are compared to the trends calculated from observations, especially given that this paper is an evaluation of ERA-Interim.  On what evidence are you basing  your statements  that the ERA-Interim does better than ERA-40 to capture temporal trends?  

 

Section 3. Rather than "whole time series", say instead "... complex TP at annual, monthly and seasonal temporal scales ..."  Also, what does "Overall the TP has great temporal and spatial variations" mean?

 

Figures and Tables. More attention needs to be paid to figure and table captions:

 

Figure 1. The caption needs to include what are the numbers in this figure.  Something like, "The numbers refer to the station number shown in Table 1".

 

Figure 2. Is this a plot of annual mean temperature? The variable that is plotted needs to be described in the figure caption.

 

Figure 3. The scales on these plots are not consistent -- note that the scale is finer for the winter plot and consequently there appears to be greater inter-annual variability.  Keep the scales consistent so that it is easier to compare the different plots.  Also, I question whether this plot is even necessary, as basically it shows that, with the exception of some years in winter, the Te is considerably warmer than To.

 

Figure 4.  Include units for RMSE (oC).

 

Figure 5.  A better caption would be "Average annual mean temperature (oC) across the Tibetan Plateau for 1979-2010 from the ERA-Interim reanalysis.

 

Figure 6.  Average temperature (oC) across the Tibetan Plateau for 1979-2010 from the ERA-Interim reanalysis for a) spring, b) summer, c) autumn, and d) winter.