The work of Qiao et al., focuses on the PWV and AOD retrievals of EKO MS711 and MS712. The method presented is based on methods of other instruments , but it is novel and since the spectral measuring instruments are becoming more popular, will become valuable in the future and fits the scope of AMT. However, the manuscript is poorly written, a lot crucial information for the reproducibility of the methodology are missing and the validation of the retrievals is very shallow. Hence, I suggest to be considered for publication after major changes. More specifically:

The two instruments are considered as one for most of the manuscript. I think it should be separated and make clear what is the performance of each one. Since the area around 940nm is overlapped by both them, the comparison of the measurements should be presented. Also, the different spectral steps and FWHM will result to very statistics in the validation process. It is crucial to present that, since the instruments are usually sold and installed separately and also in case of parallel operation, a decision should be made for the overlapping region.  Finally, in section 2 more details should be mentioned such as the calibration of the instruments, the reported uncertainty and their measuring schedules.  Specially, the calibration of the spectral bands is very important and could lead to high deviations for the algorithm. Is there any wavelength shift? How are the spectral channels characterized?

L75-80 More details should be provided on the cloud screening procedure. How effective was it? Give a figure showing the cloud screen data and discuss the results.

It is implied that the radiative transfer model used is MODTRAN. Please, add a section in 2, about the model, the setup , the selection of variables and the bibliographical accuracy.

In general the 1370 absorbing window is more sensitive to PWV changes, but the Direct Irradiance signal at this spectral range is a lot lower. Hence, before using it, signal to noise ratio for the instrument should be discussed and the expected uncertainty should be estimated.

L103/figure 4. This approach should be discussed thoroughly and the results need to be evaluated.

L117 Results showed in figure 5 are not enough to prove that one band is more efficient than the other. We don’t know what is the testing sample, how representative is and all other effects on the measurements are already eliminated. A discussion leading to figure 5 is clearly missing.

3.2 It is not clear at which wavelengths this inversion will be used. It is a odd to name this aod inversion in general, since it is not valid for the most wavelengths (where other gases absorb). I suggest to focus in water vapor bands and close bandwidths and just calculate aod for those and keep the full aod inversion for future work that will include more trace gases.

L139 this uncertainties should be discussed and estimated in a separate section.  Also, the fact that is compared with CIMEL retrievals, which was found in other studies to drift above 70º sza.

L142 “here we say”, I don t understand this phrase.

Figure 7 discussion. It is clear that band 2 is underestimating PWV constantly. It is more like a constant bias of 0.02 between the two bands. So this seems more a calibration issue  (between the model and the instrument) than a systematic error of the method.

Figure 2.L82  This figure does not show water vapor absorption windows. It is just two random measurements. Do we know that there was different PWV at these days? Figure 3 Clearly shows the windows, but the figure 2 has no use at this version of the manuscript.

Figure 3. I don’t understand the purpose of visualizing cimel filters. Also, the aerosol line, corresponds to a specific AOD (which will  change the transmittance) . Please change the legend to the actual AOD value. Also, move the legend to a position that does not hide the drop at 1300-1500nm.

Figure 4. Describe better at the caption. Information on how these spectras were retrieved.

Figure 5.  What are the “real values”? If it is CIMEL retrievals, keep in mind that previous studies showed that CIMEL was the most erroneous from all the methods (GPS,  radiosondes, microwave radiometer).

Figure 8. It is not wise to provide spectral AOD, when all the trace gases but the water vapor are ignored.

The authors provide us a physical algorithm based on radiative transfer to derive the Precipitable Water Vapor (PWV) and Aerosols Optical Depth (AOD) from a pair of ground-based spectroradiometers, EKO MS711 and MS712. In their algorithm the water vapor band near water vapor band near 1370 nm as well as near 940 were also used to derive the PWV, this is very important for dry atmosphere, e.g. for very cold area such as Tibeat  or other high altitude plateau.

All my questions and concerns in my first review has been answered, and a new revised manuscript was submitted, I have no more concerns for this revised version except a few typo mistakes and language sentences. I agree it publishing in AMT after the revision.

Minor concerns:

1. Line 31-33, “MWPS measures the radiation emitted from the atmosphere by microwaves, yields a vertical profile of water vapor, which can then be integrated to give PWV, where aerosols have little effect, but this measurement is very expensive (Güldner and Spänkuch, 2001; J. and Güldner, 2013).” should be rewritten as: “MWPS measures the radiation emitted from the atmosphere by microwaves, yields a vertical profile of water vapor, which can then be integrated to give PWV (Güldner and Spänkuch, 2001; J. and Güldner, 2013). The advantages of using microwave for PWV is that aerosols have little effect, but the disadvantage is that this kind of instruments is generally is very expensive”
2. Line 39, the short name “PHOTOS” should be clearly mentioned the first time.
3. Line 89, it’s better to replace “based on this” as “therefore”
4. Line 114-116 “When simulating spectral curves, the US standard atmospheric model was selected, regardless of clouds and aerosols, randomly inputted PWV of 0-0.5 cm and solar zenith angle of 10°-45°, and superimposed -1 %-+1 % noise on the simulated curves.” Should be rewritten to make it clear, maybe better written as: “In spectral simulations, the US standard atmospheric model was used with random PWV between 0-0.5 cm, and solar zenith angle of 10°-45°, regardless of clouds and aerosols. The simulated spectral plus +/-1% was used for retrieval.”