The manuscript is well written, with great efforts in the absent of sufficient data.

The following points would be better considered for discussion and future research.

Regarding the lower calculated concentrations in river, the authors have indicated that river discharge may be higher than observed and veterinary and industrial are not considered.

I agree on this point. But on the other hand, it should be mentioned that there are factors that further decrease the river concentration.

Within this paper, there is no mention of load reduction before entering the wastewater treatment plant.

If the direct discharge coefficient is considered as the inflow from the conduit to the river in urban untreated area, I believe that something similar may be happen in the sewer pipes.

Taking this into account will lead to a decrease of river concentration.

In addition to this, advanced wastewater treatment plants could lead to further load reductions.

It would be desirable to mention these points in the discussion or as future research topics.

Minor comments:

It would be good to indicate the 10 km on line 147, if there is any reasoning behind it.

Isn't ds,r in line 321 a mistake for dl,r?

Check line 332 for a reference error.

There is a spelling error (individual) in 533 in Fig. 5.

Dear authors,

Thank you for the well-written manuscript. The efforts of combining hydrologic and population models and validating the model results at the global scale are fully acknowledged. Although I am not an expert in this field, I can follow the arguments and the model assumptions. In the following, I have provided some comments that may improve the manuscript, ranging from minor to critical.

Critical:

According to GMD code and data availability policy: "Where the authors cannot, for reasons beyond their control, publicly archive part or all of the code and data associated with a paper, they must clearly state the restrictions" (https://doi.org/10.5194/gmd-12-2215-2019). In the supplement python file, HydroFATE_v09.py, the imported library arcpy is not publicly available. Please state the restrictions in the code and data availability section. For example: "A license of the software provided by ... is required to run the provided scripts."

The method presented in Section S.1 of the supplementary material is crucial to reproducing the modeling result of HydroFATE v1.0. In essence, Section S.1 details the generation algorithm of the WWTP service area. Section S.1 should be moved to the main manuscript, perhaps in the Appendix.

Follow up on the previous comment on the generation of the WWTP service area. Line 242, which is in Section 3.1, reported a "successive trial-and-error approach in which intermediate results were mapped, visually inspected for plausibility, and statistically tested to verify whether they led to further improvements." I understand the difficulty of the service area generation, and it is acceptable to inspect the model visually. However, the realization of the WWTP service area is a critical part of the HydroFATE model. Hence, it is within the scope of the GMD code and data availability policy. I urge the authors to provide the scripts you use to generate the service area since I cannot find them in HydroFATE_v09.py. (Please review Section 3.2 of the GMD editorial: https://doi.org/10.5194/gmd-12-2215-2019.) Furthermore, the sensitivity analysis in Section S.3 does not provide the sensitivity of different realizations of the WWTP service area. Therefore, it is necessary to make the service area generation process more transparent by providing the script.

Suggestions:

Figure 2 provides an example of the generated WWTP service area used in this study. It would be helpful to show examples of intermediate iteration results and explain why this iteration is rejected or accepted so the readers can have a gist of the visually inspected acceptance criteria.

In equation (1), the removal efficiencies of wastewater treatment plants (WWTP), e_{WWTP, j}, contain three treatment levels, which are primary, secondary, and advanced. However, the parameters of the case study do not include such complexity. Only one removal efficiency is used in each scenario. During the review, I have always considered WWTP treatment levels spatially varying, and they can be sensitive to the modeling result. I find the presentation of Figure 2 a bit misleading. Therefore, I suggest adding another figure in Section 4, which is based on Figure 2, but the WWTP areas are only color-coded once to represent the model better.

Minor:

The unit in Figure 4 should be presented as m³ s^-1.

Fix the unit in Figure S-3: ng L^-1.

Fix the reference error in Table S-5 of the supplement.

Dear Authors,

I read with great interest this manuscript. I think it is an important work that enhance previous large scale modelling approaches on contaminant fate in rivers.

Specifically, the HydroFATE (v1) model, is built upon previous smaller scale models such as the one of Grill et al. (2018) (https://doi.org/10.1016/j.watres.2018.08.053 ). HydroFATE (v1) includes higher spatial resolution respect to models previously published in GMD such as GLOBAL-FATE (version 1.0.0) (Font et al. 2019, https://doi.org/10.5194/gmd-12-5213-2019 ) and additional physic-chemical processes.

In my opinion, the MS is well written and it is suitable to be published in GMD. However, I have particularly focussed on the scientific matters of the modelling approach, and I have found some weaknesses in the model methods description, validation and discussion that need to be addressed. I list here below the general comments and I attach a pdf with detailed comments. I hope this revision will be of help to improve the manuscript.

General comments

• Methodology
  • The mechanism of water/soil partitioning (or absorption to soil) is complex. Here, the choice of their relative parameters is not process-based. It appears that are from back calculations or calibrations for the specific case of China (Grill et al. 2018) and may be not valid for other areas of the world. I understand that this is difficult for such global scale. But this needs to be better discussed.
  • On-soil and groundwater contaminant degradation are here accounted with a degradation parameter based on the Euclidean distance from hypothetical source point and closer stream. But, groundwater flow does not necessary follow straight lines, but also follows flow directions trough positive gradients related to the local terrain geomorphology. Also, aquifers increase residence time, and therefore degradation. Because you state that the model is not very sensitive to this parameter, why did you add it on your model? This way, it seems that you add unnecessary complexity.
• SMX case study and validation
  • SMX has both human and veterinary use. Therefore, it is complex to account only for human uses when validating the model. In the validation, you attempted to focus only on catchments were human use dominates. However, at global scale and with such big basins, this is very difficult.
  • You defined several scenarios by changing relevant parameters. Hoever, all parameters choice should be justified. For instance, you discussed the WWTPs removal variability in literature being 2% (min), 49% (ave), 73 (max) based on literature values. What about the other parameters in the different scenarios?
  • High flow conditions, although favourable for higher dilution, is not considered as extreme low-end scenario.
  • The choice of using or not MECs below detection limit for validation is contradicted a couple of time in the MS. I would describe it better and univocally in the methodology section. In my opinion, MECs below detection limit are still important for model validation.
  • You accounted the uncertainty in model prediction due to discharge condition into your PECs. If you include it again in MECs, it means that you are considering this uncertainty twice, which is not correct.
• Discussion
  • I would appreciate a deeper discussion on the quality prediction (i.e. on NRMSE, NSE, PBIAS, KGE parameters). Are they expressing a good or bad prediction performance of your model? How do they compare with other similar large scale models performance? Is it an acceptable performance for predicting contaminants at global scale?