Comments on: Comparing GPS RO Temperatures and Lapse Rates within Clouds with ECMWF and NCEP Analysis General Comments: This is a nice piece of work. To my knowledge, this is the first paper trying to retrieve temperature within clouds using a combination of GPS RO observation and global analysis. This is a very challenging task, and it is interesting to find and document systematic differences between temperatures inside the clouds from various different sources. Specific comments: 1. Comparison of different types of measurements: In this study, various different types of measurements are compared. It is important to recognize their differences in resolutions, and measurement characteristics: a. b. c. d. NCEP – This is a low resolution global reanalysis T62L28. ECMWF – This is EC’s operational analysis at T799L91 CloudSat – The resolution is on the order of 1.4 km x 2.5 km GPS RO – The averaging length is on the order of 300 km along the ray, and about 1 km cross the ray. Global analyses are providing “averaged” information within the grid point. The GPS RO measurements are affected by the azimuth angles of the measurements. Also in the vicinity of clouds, there will be significant horizontal refractivity gradients (due to moisture variations). Spherical symmetry assumptions may not work well. In other words, if the ray goes through clouds, the retrieved GPS refractivity does not necessarily a good representation of the “point” value. It would be an averaged value including both cloudy and clear-sky. It is clear that NCEP/NCAR reanalysis has the lowest resolution, and it also suffers a few disadvantages: old data assimilation system (3D-Var, SSI), do not use all available data (including COSMIC). The reviewer may ask: Why didn’t you use the FNL (NCEP GFS final analysis)? It would have been more compatible with the ECMWF operational analysis. Of course, EC uses 4D-Var, and has a much higher horizontal and vertical resolution. Another point is that the vertical interpolation (from the relatively low L28 resolution to the GPS vertical resolution) may induce some errors (or uncertainties). Again, using the NCEP FNL at native grids would have been better. When performing a comparison between GPS RO and the CloudSat measurements, it would be useful to keep in mind the azimuth angles of the GPS RO ray path. For example, I would expect different types of bias depending on whether the GPS RO ray path is parallel or perpendicular to the direction of the granule. In Fig. 16, you showed results for 10 positively biased sounding and 29 negatively biased soundings. Question is: Why some are positively biased, and some are negatively biased? Are there significant differences in their azimuth angle relative to the track of the CloudSat? Another question is that since the footprint of the CloudSat is so small, one may wonder “how representative” are these measurements? Can they really be used to represent area averages? 2. Retrieval of Cloudy temperature It is an interesting attempt to estimate the temperature within clouds, based on ECMWF global analyses and GPS RO measurements. I have been wondering, “What does this retrieved temperature represent?” “What is the applicable horizontal resolution for the retrieved ‘cloud temperature’ profile?” CloudSat is providing estimates of cloud top and cloud base, and other cloud properties based on a very small footprint. GPS RO is providing estimated “mean” refractivity along a ray path that has a scale of 300 km by 1 km. ECMWF is providing a box average value of about 25 km x 25 km. Would the new GPS RO Cloudy Retrieval algorithm produce significant changes in the GPS RO refractivity? Or, is the GPS RO refractivity kept constant? This point is not made clear. GPS RO derived refractivity is a quantity obtained following the retrieval procedures under various assumptions. All the real atmospheric structure along the ray path would influence the final derived refractivity. The “wet-retrieved” temperature assumes that the GPS RO refractivity is representative of a “point value”, and proceeds to derive the temperature under such assumption. In an inhomegeneous environment, the GPS refractivity can deviate significantly from the local refractivity. Therefore, the derived temperature would not be representative of the local value. On the other hand, the GPS RO refractivity can still be a robust “measurement”, we just need to recognize of “what it really represents”. Hope these comments are useful for your revision of the paper. Bill K.