Li et al. developed a multi-stage model to predict the reactive nitrogen concentrations during 2000-2100. Overall, the topic is very interesting and the manuscript fits the scope of ACP. However, the manuscript still suffers from some major flaws especially some sections lacks of necessary examination and discussion. I recommend the manuscript for publication on ACP after the following comments have been well addressed.

1. “Ambient” should be added between “of” and “reactive” because the authors simulated the ambient reactive nitrogen concentrations.
2. Line 60-69: The introduction about reactive nitrogen estimates in future scenarios is too simple. I suggest the authors must point out the importance and novelty of the simulation in the future scenario.
3. Line 94-102: More detailed GEOS-Chem scheme should be introduced.
4. Why do you choose XGBoost and LightGBM among all of the decision tree models? Please add the introduction of hyperparameters of these models.
5. Section 3.1: How about the spatiotemporal transferability of this multi-stage model? The authors must add some tests about it.
6. Line 223-225: I cannot agree with this view because many croplands were also distributed on Southeast China.
7. The historical trends of reactive nitrogen concentrations in some other regions such as Africa and South America could be also introduced simply because the authors only focused on China, Europe, and the United States.
8. Section 3.5: This part is just result rather than discussion. The author must add some discussion about the future trends of reactive nitrogen species.
9. Section 3.6: The GEOS-Chem output also shows some uncertainties.
10. Section 3.6: The authors should add the conclusions rather than the implications alone.

general comments:

The manuscript of egusphere-2024-69 entitled “Global estimates of reactive nitrogen components during 2000-2100 based on the multi-stage model” presented the evaluation of reactive nitrogen at the current status and projected to the future using data fusion with CMIP6 results. This kind of topic should be clarified and I can fully understand the importance of this study. However, many parts need to be amended. Especially, as a major comment to this manuscript, I do not fully follow why the main discussion has been focused on only the U.S., Europe, and China. Because the validations were based on NNDMN over China, CASTNET in the U.S., EMEP over Europe, and EANET over Asia, these areas need to be fully discussed, but I do not find the detailed discussion over Asia, and the status over only China seems to be focused. Moreover, other areas could be precisely estimated based on the fusion of satellite data and can be discussed, but why the status of India was sometimes discussed? Toward the dramatic change in the 2100s, the targeted areas regarding reactive nitrogen can be discussed through this study, and I expect such perspective discussions. In addition, the discussion for each region and each species will be discussed in longer sentences in general, so how about summarizing as one or two table(s)? Taking into consideration the importance of the findings through this study, I would like to request a clear presentation. Please also see and address the following specific comments before the possible publication from the journal of ACP.

specific comments:

• Title and abstract: I do not fully follow the intention of “stage” in the manuscript title itself. This point can be followed from Section 2.3, so it will be better to clarify this methodology in the abstract.
• Page 2, Line 31-32: What are the direct emissions for HNO3? The oxidation pathway from NOx will be the source of HNO3. Please clarify.
• Page 3, Line 60-62: I am confused reactive nitrogen themselves affect the land carbon cycle and greenhouse gas emissions, or vice versa.
• Page 3, Line 70-73: Why only four components of NO3-, HNO3, NH3, and NH4+ is regarded as important species? How about NOx (NO and/or NO2) and PAN? The appropriate references are required to support this target.
• Page 5, Line 139 and Line 143: Were the input datasets unified or different at the first and second stages? The wording “meteorological factors” and “meteorological parameters”, “land use type” and “land use data” impressed us with the different datasets. So, please clarify the difference. If the same dataset were used at these two stages, the expression should be unified.
• Page 6, Line 147-149: At the final stage, did the authors use the CMIP6 dataset from 16 earth system model as an ensemble average? Please clarify.
• Page 7, Line 178: I do not fully follow how the Geos-Chem 2.5 degrees dataset was resampled into 0.5 degrees. Please add more explanations for “appropriate algorithms” explained here.
• Page 7, Line 180-183: I can understand this approach, but what kind of variables were removed during the model development process? Such additional information would be helpful for readers, so please consider including them in the supplemental information.
• Page 7, Line 190 and 191: Even though such statistical metrics have been well used in the model evaluation papers, the abbreviations for RMSE and MAE should be introduced, and defined within the manuscript (or, refer to appropriate references).
• Figure 1: “NO3-N” and “NH4-N” should be unified into the main manuscript. From section 2.3, I feel LightGBM method was applied at the first stage whereas this Fig. 1 seems to be using XGBoost. Please confirm this figure.
• Figures 3, 5, 6, 7, and 8: The size of these figures is too small to see the details of structure over the world. Please revise.

technical corrections:

• Page 4, Line 88: The monitoring site numbers are needed for EANET as is the same manner for EMEP and CASTNET.