The manuscript compares three geophysical measurement systems in a triple collocation method, defined as 3CH, and in a pair-wise comparison for humidity variables. It remains very useful to stress the limitations and potential interpretation errors caused by pair-wise comparisons. The editor however also notes a few minor issues for consideration.

First, Stoffelen (1998) noted that three geophysical systems generally do not represent the same spatio-temporal average and not exactly the same variable. In order words, the true geophysical state may be sampled in more detail by two of the systems with respect to the third. This leads to a correlation of the representativeness error of the two systems, which he quantifies to affect the results for his wind analysis, later repeated by other authors for other geophysical comparisons. O'Caroll et al. (2008) recognize this potential problem, but state on representativeness for their SST analysis: "In this paper, we proceed tentatively on the assumption that these covariances are negligible, but we also make analyses to determine if this assumption is valid.". Hence, the TC equations of O'Caroll et al. simplify to what is now in this manuscript defined as 3CH. The current manuscript provides useful tests on another geophysical set and provides some analysis. However, on page 2, line 22, it is currently stated that "The major assumption in all of the above methods is that the errors of the three systems are uncorrelated" and "Correlation between any or all of the three measurement systems will reduce the accuracy of the error estimates". In fact, the in this statement implied TC method proposed by Stoffelen (1998) does take account of the correlation in the spatial representativeness error and corrects for its covariance. Hence, he does not assume uncorrelated error in all three systems as stated here. It further follows that in TC only UNKNOWN error correlation will reduce the accuracy of the error estimates, as in fact demonstrated in this manuscript too. Please revise these sentences to reflect these comments.

Second, on page 3, line 18 states: "The last three covariance terms in Eqs. (3)–(5) are unknown for real data sets". Given the above, the statement is not generally valid. Stoffelen (1998) and others found methods to estimate these covariances for real data sets. A revision of the sentence to make it less general seems appropriate.

The authors well considered the reviewer and editor suggestions.

A last minor comment is that Stoffelen (1998) elaborated on the necessity of error modeling before calibration and validation of geophysical data in his Appendix A and in section A2 particularly elaborates on the fundamental weakness of 2-way comparisons. This is closely related to your new section 4.3 and you may want to refer to it.