Maria Val Martin et al., presented an interesting study that developed new capability for CLM5 N cycle and used this new model to investigate the impact of ERW-induced pH changes on agriculture soil N20/NO/NH4 emissions. This is an exciting and timely work that greatly advances existing modeling tool and will benefit the research community a lot. Below are my comments and suggestions.

The model development description is very clear; however, the model simulation plan needs to be improved. Important information is missing, and the choice of forcing data is potentially problematic. For example, section 2.3, which forcings data are used for spin-up simulation? How many years of simulation were conducted for accelerated spinup versus regular spinup? What are the criteria of soil N variables for a fully-spunup condition? Why use GSWP3 forcings for 1850-2014 simulation but NLDAS forcings for 2015-2019? When transition from 2014 to 2015, the difference between NLDAS and GSWP3 forcings will result in sharp change of vegetation and soil dynamics. For global simulations, again, how many years of accelerated spinup versus regular spinup were performend? How to determine if global land has reached a fully-spunup condition.

New development introduced a lot of constant parameters (e.g., fN2Onit = 721.86*e^-2.387*pH). fN2Onit has two constant parameters: 721.86, 2.387. These parameters were derived from previous studies, however, also had large uncertainties. And their incorporation into a global model like CLM, will have corresponding parametric uncertainty. How do we confidently know the constant parameters work well at global scale? A common strategy in previous CLM development is to define parameter ranges and conduct model tuning for newly introduced parameters. In such way, we will not only improve the model performance, but also gain knowledge about model parametric uncertainties.

How CLM5 estimate the dynamics of soil pH is missing? A brief description of the weathering module and how that affect soil pH is needed in the methodology section. Also, a comparison and discussion of observed versus modeled pH variability is needed to demonstrate that CLM5 was able to capture reasonable pH dyanmics after e.g., applying the crushed silicate rock.

Site level simulation is convincing, but the global comparison needs to be improved. Based on the model performance shown in Figure 6, Table 2, the simulated changes in N2O/NO/NH3 fluxes (figure 7) were not convincing, therefore, the major conclusion of ~30% reduction in N2O/NO emission due to ERW was not strongly supported. Overall, I think this study is timely and important, but the model seems not ready for a global application.

The paper is written very clearly and well presented overall. As my only major comment, I am under the impression that the discussion on the global results is too optimistic. I suggest that the authors should keep the discussion in Section 3.2 more in line with what is actually shown in the results. I have further minor comments below.

Line 41: The literature can be made updated. The scenario community is exploring scenarios without much relying on CDRs (e.g. Riahi et al. 2021), a shift in paradigm from the time of Fuss et al. (2014).

Line 46: The authors may include the most recent State of CDR report. https://www.stateofcdr.org/

Lines 48-55: The authors could also point out the release of phosphorus as another biogeochemical consequence of ERW, as discussed in Goll et al. (2021).

Line 135: What happened to other factors such as soil temperature and water content, which were just mentioned a few lines above?

Line 141: Is the soil pH kept at the nominal value throughout the simulation period?

Line 171: Usually “taken up”, not “uptaken”

Line 262: “qualitative” may be replaced with a more appropriate word. I understand what the authors try to say, but numerical comparisons are always quantitative.

Line 290: “increases”, not “increased”

Lines 293-295: Please describe what kind of calibrations are needed to get a better agreement.

Line 297: it should be “with respect to”.

Line 305: Why is the range from the 2019 soy simulations so large?

Line 344: Maybe “simpler validation”

Line 378: The numbers are “r”, not “r2”. I would rather see it as a relatively poor correlation.

Line 379: I think that the statement here is also too optimistic. NMB from EDGAR is 147%.

Line 379: NMBs for NO are 21%, 71%, 142%, and -11%. What are the reasons for the very high bias?

Line 382: What is the basis for good correlation?

Line 383: Table 2 indicates 142%.

Line 402: I think that the discussion should continue with a comparison between CLM5 and inventories. For example, there is a rather large difference in the estimate of NH3 emissions in China between CLM5 and inventories.

Lines 437-439: Here again I think that the statement is too optimistic and unsubstantiated. I would rather see it as a mixed outcome.

Line 481: Table 3 indicates that CLM5 generally gives a lower estimate of NH3 emissions than emission inventories do. Does this imply that the effect of ERW on NH3 emissions can be larger than what is indicated from CLM5?

References

Goll et al. (2021) Potential CO2 removal from enhanced weathering by ecosystem responses to powdered rock. Nature Geoscience 14 (8):545-549. doi:10.1038/s41561-021-00798-x

Riahi et al. (2021) Cost and attainability of meeting stringent climate targets without overshoot. Nature Climate Change 11 (12):1063-1069. doi:10.1038/s41558-021-01215-2