|I have read through the revised version of this paper. I thank the authors for taking note and responding to comments made in my original review. I recommend this version for publication, but do have the following specific comments on this version of the paper.|
(1) The Abstract and Conclusions identify welcome developmental steps involving incorporation of aerosol information and use of ensemble techniques. As pointed out in the body of the paper, the representations of Ta in ERA-Interim and JRA-55 are much better than those in other reanalyses, almost certainly in significant part due to their analysis of synoptic observations of the variable. This could be mentioned in the Abstract and Conclusions as another way the near-surface products of other reanalyses might be improved.
(2) Table 1 has been tidied up following a comment in my original review, but the entry for ERA-20CM remains incorrect. The entry under ”Assimilation System” for ERA-20CM has been changed from 4D-Var to 3D-Var, but this is still not right. There should be no entry under “Assimilation System” for ERA-20CM as it did not directly assimilate any atmospheric observations. Observations influence ERA-20CM only indirectly, through the forcing provided by CMIP5, and the prescribed SST, sea-ice and other model fields. ERA-20CM had no assimilation system of its own.
(3) Regarding point (6) of my original review, I can accept the terminology “NWP-like reanalysis” and “climate reanalysis”, even though I still do not particularly like it. In Dee et al. (2014) – of which I am a co-author – we use the phrases “NWP-like reanalysis” and “extended climate reanalysis” which is subtly different, in that we (or at least I) regard the NWP-like reanalysis as one type of climate reanalysis and the extended climate reanalysis as another (longer) type. However, this slightly different nomenclature will not work in the present authors’ case as CFSR, which they classify as a climate analysis, is not “extended” in the sense used by Dee et al..
(4) Linee 229. It would probably be better to write “incorporate many observations” rather than “incorporate some observations”, even though the reanalyses have access to under 10% of the Chinese observations available to the authors. Globally, the number of observations used by ERA-Interim is some thirty thousand per day for the 1980s, and that number has risen considerably in recent years.
(5) Lines 310 and 311. The sentence that spans these lines needs rephrasing as the correlation coefficient does not assess the absolute value.
(6) Line 359. “A homogeneous adjustment” should, I believe, be changed to “a homogenizing adjustment”.
(7) Line 398. Delete “in ERA-Interim and JRA-55”
(8) Line 431. It would be better to replace “jointly determined” by “determined in part”, “determined to a significant extent” or some such phrasing. This is needed because surface air temperature is sensitive to a number of factors – such as circulation patterns and snow cover – that are not considered in the authors’ analysis, or only partially considered – a circulation anomaly may change Ta through differences in precipitation frequency that are taken into account, but may also change temperature due to advection. Similar qualifications are needed in several similar statements made in this and the following section.
(9) Line 512. The trend in precipitation frequency is given in units of days/decade, implying that the precipitation frequency has a unit of days. How is precipitation frequency defined? Is it the number of days per month with appreciable rainfall? Maybe this is stated somewhere in the paper, but I do not recall seeing it.
(10) Page 527 and 528. Elsewhere the biases in temperature trends are related to biases in precipitation frequency, Ld etc.. But in this sentence it is the other way round. Is this a mistake?
(11) Section 3.4 is a long one, and a more succinct synthesis of the results would probably help many readers.
(12) Line 578. Incorporating observations of Ta as done by ERA-Interim and JRA-55 is expected generally to reduce biases in the trends of analyses of Ta, not to introduce biases, as the reanalyses tend generally to reproduce the trends present in the observations, even in the presence of biases in the assimilating models. An exception can occur for regions where observations are not available for a substantial part of the time, particularly early or late in the period covered by the reanalysis. In this case, trends will be in error if the assimilating model has significant biases that are corrected at times when there are observations, but uncorrected when observations are absent.
(13) Lines 617 and 618. Elevation differences can occur over time if stations are relocated, but this should be largely taken care of by the homogenization.
(14) Lines 647 and 648. ERA-20CM should be referred to as ensemble simulation not ensemble forecasting.
(15) Lines 663-670. It is stated that the comparison is for the reanalyses that do not incorporate observations, but ERA-Interim and JRA-55 are listed in the comparison. This needs amending.
(16) Lines 691 and 691. This reads rather as if MERRA2 included time-vary aerosol loading in order to make a pioneering attempt to improve regional warming over the North China Plain. One suspects that the motivation for including time-varying aerosols in MERRA2 was much more general than this. The text could be amended slightly to read something like “MERRA2’s pioneering incorporation of time-varying aerosol loadings provides a way of improving the representation of regional temperature changes over regions such as the North China Plain where the impacts of aerosols on surface temperatures are significant.”