The manuscript provides a new global model for water quality assessment (DynQual). The model runs at a daily time step for the grid cell of approximately 10km by 10 km at the equator. The authors made a major effort to develop such a comprehensive dynamic model on a global scale. One of the model's strengths (compared to other existing large-scale water quality models) is the sector contributions (e.g., livestock, irrigation, manufacturing) to surface water pollution per grid and day while considering dynamics in pollutant routing through the river network. The model is process-based and largely uncalibrated. It makes the model more flexible to represent the processes based on characteristics (e.g., livestock number, people, runoff) and less dependent on observations (which are often scarce). In this way, there is an opportunity to apply the model for the future while considering the future characteristics of the areas. Validating the global model is not easy. The authors did a great job here and managed to validate in a good way. The manuscript is well-written, with beautiful maps and graphs. It is easy to follow the description of the model and results.

I suggest a minor revision. Below, I provide suggestions that can be helpful to improve the manuscript:

1. Abstract: It is a nice abstract, but does not provide any insights that we learn from the model application. Please add the main messages (2-3 sentences) that reflect the two main objectives of the model application:
   1. Pattern and trends
   2. Sector contributions
2. Methods: They are described rather concisely. Details are provided in the supplementary materials, which is nice. Nevertheless, I have four suggestions to elaborate on:
   1. Figure 1: please add the legend. The description of what colors and different arrows (dashed and solid) mean is not clear;
   2. Pollutant loadings to streams: this is well described in the supplementary information, but very concisely described in the main methods. Please add a few more sentences to tell the reader how pollutant loadings are calculated (e.g., the summary of the description from the SI). You can elaborate on the text where you mention the mass-balance approach. Here, you can briefly tell that pollutant loadings from livestock activities are simulated as a function of livestock number, excretion rates of pollutants per animal and day, and removal of pollutants during waste management practices while considering runoff from land to streams (see SI Section 1.4 for details). A similar description can be given for other sectors.
   3. Sector contribution: you do mention sectors in the methods, but briefly. Please elaborate more on what sectors exactly include and how. For example, the irrigation sector: what does it include? Which crops? rainfed irrigation? Livestock sector: which animals? I do know that some details are in the SI. But I do feel a need to give a bit more description of the main methods of the manuscript.  
   4. Downscaling: for example, some of the input data (e.g., livestock numbers) is regional, but DynQual requires the grid cell data. How did the authors go from regional to the grid cell, but also from annual to daily levels? Which model inputs require scaling (e.g., annual->daily; regional -> grid cell)? and which did not. This is not well elaborated. I suggest adding a few sentences on this in the main methods and giving more details in the SI. I suggest adding an overview table showing the list of model inputs and indicating which ones were aggregated from region to grid and from annual to daily.
3. Discussion: It is very concise and to the point. Some aspects can be expanded and a few aspects can be added:
   1. Comparison with other studies: the authors do this for the pollutants that they consider. I also think that the manuscript will benefit if the authors add comparisons in terms of modeling approaches, pollution hotspots, and sector contributions (what new aspects are added in this DynQual model and what new aspects we learn from the model application compared to other models). The authors may consider expanding the discussion (a few sentences) on comparing their pollution hotspots not only for TDS, BOD, and FC but also for other pollutants as well because pollution hotpots often match between pollutants.
   2. Implications of the limitations: any models have limitations. DynQual has as well Examples are livestock numbers in extensive and intensive production systems that do not vary among days, excretions rates of pollutants in manure, and human waste that are constant across the days and within the regions. I understand that sources such as open defecation, and direct discharges of manure to rivers are not considered. It is fine, but this needs to be discussed. It is important to give examples of the main limitations and reflect critically on their implications on the main conclusions of the manuscript.
   3. The usefulness of the model: DynQual has many useful applications (e.g., trends, patterns, future analyses, etc). The authors briefly mention this in paragraph 545. I think the authors can better emphasize how useful their model is compared to other models. For example: which scientific questions we can answer with this model that we could not answer with the previous models? The authors could add a few sentences on this in paragraph 545.
4. Supplementary information: It is well written. I have three suggestions:
   1. Units: they are missed in some equations. For example, units are not included for the following variables: R_dom,i,n and Pop_n (population per km²? Total population?) in equation [1], L_man,i,n and R_man,i,n in equation [2], L_urb,i,n and R_USR,i,n in equation [3]. Please also check the variables in equations [4] and [5] and add units for every variable. This will avoid misinterpretation.
   2. Livestock activities: is the number of livestock the same per day? Is this number per km² or ha? Did you consider soil processes and associated retentions of the pollutant in soil when you calculate loadings into the streams? All this was not very clear to me. Please clarify the description of equation 4.
   3. Scaling: please elaborate on which input data required to be scaled from region to grid and from annual to daily, and how this was done.

In addition to (or as an emphasis of) the community comment by Jason Ke and the comments from Referee #1 I have the following suggestions for improving the paper:

- Model description paper in GMD should focus on a detailed description of the scientific basis of a model and the technical/ numerical implementation. In the current version of the manuscript, there is a certain imbalance between the respective contents in the main paper and the supplemental material. Most of the description of model details in the supplemental material should be directly mentioned in the main paper.

- Another large part of model description papers should be dedicated to the model verification. This aspect is quite underrepresented in the current version of the manuscript. Model verification is only presented in line 304 to 321 and one figure in the main paper. Some additional results are included in the supplemental material but without a thorough presentation, analysis and discussion of the results. These aspects need to gain much more space in the main paper as this is a central part of model description papers in GMD. Along with a more detailed model verification, the authors also need to emphasize the discussion on possible model limitations. 

- A much larger portion of the manuscript (lines 323-491) is dedicated to the presentation of spatial patterns and trends of (long-year global) simulation results. Section 3 is named "model demonstration" and the model evaluation part (see above) is only a small part of this section. Showing possible applications of the model is, of course, interesting but model fidelity has do be demonstrated first and in much greater detail before spatial patterns and trends can be presented. Hence, there needs to be a better balance between the "model evaluation" and the "model demonstration" part. 

- The performance of DynQual should also be discussed in comparison to other water quality models, e.g. available catchment or regional scale models. The level of detail in terms of available input data and spatio-temporal resolution is, of course, different but the modelling community and potential users need to know what quality of output they could expect from this newly developed model compared to existing ones. Although the scope of the model might be a bit different, the interpretation of results needs to rely on the closeness to observation data (i.e. the prediction capability) and this one needs to be compared to already existing modelling approaches.

- It is unclear, why the normalized RMSE was chosen as the solely performance measure for model evaluation. Other performance measures such as the Nash-Sutcliffe efficiency (NSE) or the Kling-Gupta efficience (KGE) are much more common in model benchmarking and allow a clearer interpretation of the model performance, e.g. NSE lower or greater than zero. In addition, KGE even combines different aspects of model performance. 

- It is mentioned that DynQual can be run in two modes: (1) coupled to PCR-GLOBWB2 and (2) in offline-mode with any other hydrological model. Please describe in more detail the technical aspects of the coupling between DynQual and PCR-GLOBWB2 as well as the (technical) requirements for using DynQual with other hydrological models, e.g. what are the required input data from the hydrological model and in which form they need to be provided to DynQual (e.g. netCDF files and their format).

- Time-stepping: Some lower bounds for the sub-daily time steps are mentioned (line 184ff). Does the model ensure that numerical stability criteria (Courant number, Peclet number) are met for the reactive transport equations?

- Equation 9 for nRMSE: Please double-check this equation, the square is missing