Reviewer's comments are attached in the supplement file.

I want to thank the authors for an interesting manuscript. I agree the topic is worth of investigation and publication. The study is well performed and the manuscript mostly well written. Still, I would like to ask authors to adress some comments before the publication.

Specific comments

L17: authors mention various applications of GNSS data usage. It would be worthy for some readers to speficically name at least some of the applications.

L18: "GPS receivers" are mentioned - is it really only GPS receivers? Would it not be possible to use the term GNSS receiver here?

L22: the manuscript is focusing on GNSS derived horizontal tropospheric gradients. Their estimation and utilizitaion has been the subject of scientific studies in the last ten years or so. It would therefore be useful to extend the Introduction section in this regard and provide a better prepared summary of the current state of the art in this area (on top of several already cited works).

L65: I miss some information about the GNSS data processing:

- definitely you should provide elevation cut-off angle as it can impact estimation of gradients

- what about the interval of input observations? Was it 30s or different?

- did you process 24h data or different time periods? If 24h, have you somehow dealt with the problem of decreased quality of tropospheric parameters estimated at the boundary of the day?

- what about PCV/PCO corrections?

L71: what exactly do you mean with sentence "Equal weighting of the observations was applied"? Do you mean observations from individual GNSS systems? And/or that you applied no elevation dependent weighting?

L81: you mention usage of hydrostatic tropospheric gradients from VMF data server. Please, consider adding some information on what these gradients looked like, for example in the form of basic statistics (min, max, mean, sdev).

L91: I wonder why the elevation cut-off for WVR measurements was 25 degrees. Please, can you explain in the manuscript? I guess that for GNSS data processing you applied a much lower elevation cut-off (probably something between 3 and 10 degrees). Can't the different cut off angle cause part of the difference between GNSS and WVR results? Please, discuss in the manuscript.

Figure 3: on the right figure it seems to me that on azimuth of 45°, the measurement on the lowest elevation is missing. What is the reason? Some blockage of the sky view around the WVR?

L151: why do you expect "the true values of the mean gradients to be close to zero over a time period of almost two years"? What is your reasoning behind this statement? I would expect that mean wet gradients would point to the south on the northern hemisphere (and north on the southern hemisphere) due to global distribution of water vapour (higher values around the equator compared with higher latitudes).

Table 3: you present RMS differences for gradients from individual GNSS stations. It can be hard for the reader to interpret them without knowing the typical range of values of the gradient themselves. I therefore suggest to provide information about the standard values of Ge and Gn over the studied period. It can be just basic statistics for Gn/Ge for the single GNSS station (e.g. ONSA). I know you provide such an information by Figure 5, but for the so called gradient amplitudes, not Ge/Gn values themselves.

Do you have any explanation for Ge (east gradient) having a slightly better aggrement than Gn (north gradient)?

Table 4: correlation coefficients for GNSS vs. WVR comparison are not that high (typically 0.6 - 0.7). Although you compare the correlation coefficients between the two constraining options, LWC values, etc. In your writing, you do not specifically mention or comment the values themselves. Please, consider it.  

L189: I find the manuscript lacking a deeper discussion, which is inevitable in any scientific writing. Please include it in a separate added section named Discussion, or expand the Conclusion section with it. Some of the topics which can be discussed are mentioned in my comments above.