First, we want to thank you for the positive comments on our manuscript!

              We agree that the main outcome is that the models exhibit significant variability across N pools and fluxes. We attempted to examine the mechanisms underlying this spread and uncertainty by analyzing the representations of four central N cycling processes: N limitation of vegetation growth, biological N fixation, vegetation response to N limitation, and N limitation of decomposition. However, no significant differences between models with different representations of these processes emerged. This is likely due to other structural differences between models that confound their effects (an issue in all model intercomparisons). Further analysis is clearly warranted in future work. Similarly, while the implications of this spread and uncertainty are not within the scope of this work (given that the TRENDY simulations focus on the historical period), we conjecture that they are likely to influence future projections (see below). In our revised manuscript, we will reframe the text to highlight our main findings more clearly in the abstract and in the discussion section as well as the focus and constraints of our study. Furthermore, we agree that the title of the manuscript could be more descriptive and we will brainstorm a better title (perhaps, “Evaluating Nitrogen Cycling in Terrestrial Biosphere Models: Spread Across Models Suggests Uncertainty in Projections of the Future Terrestrial Carbon Sink”).

              You are completely correct that, although our results find that the ability of this suite of TBMs in reproducing the terrestrial C sink appears to be disconnected from the N cycle, the central purpose of the N cycle is to constrain CO2 fertilization which is not directly evaluated in our study. Evaluating the ability of a TBM to simulate present-day N cycling processes is one way to assess its ability to simulate N limitation of CO2 fertilization in the future albeit an indirect one. While not a direct test of the sensitivity of the C cycle to the N cycle, the significant variability across present-day N pools and fluxes that we assessed suggests that there could be uncertainty in future projections of the terrestrial C sink. We agree with you that a robust test of the response to CO2 fertilization and N fertilization across this ensemble of TBMs would be ideal for evaluation of the N cycle. In our revised manuscript, we will clarify this distinction and the limitations of our study. We will also emphasize the importance of experimental manipulations during future model evaluations and intercomparisons.

Furthermore, as per your suggestion, we will add an additional assessment of the correlations between the bias in simulated BNF, vegetation C:N, and soil C:N and the bias in simulated NBP (as well as other C cycle variables). If this analysis yields interesting results, we will add it to the revised manuscript. In our original analysis, we assessed correlations between the overall scores for BNF, vegetation C:N, and soil C:N and the overall score for NBP (as well as other C cycle variables) in Figure A3. However, no significant correlations emerged which is likely due to the many structural differences between models that confound these relationships. In the revised manuscript, we will highlight this analysis more in the main text.

              With respect to the minor comments, we will update our explanation of BNF in CLM5. While Cawse-Nicholson et al. 2021 points to many useful products, they are only at the site level or at a regional scale. While Fisher et al. 2012 provides a global product, it is for nutrient limitation which, as discussed above, would require the simulation of nutrient fertilisation experiments in the TBMs. We considered a remote sensing product for leaf N at the global scale (Moreno-Martinez et al. 2018; doi:10.1016/j.rse.2018.09.006) but TRENDY model outputs are given as vegetation N rather than as leaf N. This necessitates an analysis of root and wood C:N and relative tissue allocation as we conducted in our study. We will include references to these remote sensing studies in the revised manuscript as they are useful resources for future work. We will also include reference to the energy exascale Earth system (E3SM) land model (ELM) (Braghiere et al. 2022). While not part of the TRENDY ensemble, ELM is an important example of a TBM with both N and P cycling.

First, we want to thank you for the positive comments on our manuscript!

              We agree that the main outcome is that the models exhibit significant variability across N pools and fluxes. We attempted to examine the mechanisms underlying this spread and uncertainty by analyzing the representations of four central N cycling processes: N limitation of vegetation growth, biological N fixation, vegetation response to N limitation, and N limitation of decomposition. However, no significant differences between models with different representations of these processes emerged. This is likely due to other structural differences between models that confound their effects (an issue in all model intercomparisons). Further analysis is clearly warranted in future work. Similarly, while the implications of this spread and uncertainty are not within the scope of this work (given that the TRENDY simulations focus on the historical period), we conjecture that they are likely to influence future projections (see below). In our revised manuscript, we will reframe the text to highlight our main findings more clearly in the abstract and in the discussion section as well as the focus and constraints of our study. Furthermore, we agree that the title of the manuscript could be more descriptive and we will brainstorm a better title (perhaps, “Evaluating Nitrogen Cycling in Terrestrial Biosphere Models: Spread Across Models Suggests Uncertainty in Projections of the Future Terrestrial Carbon Sink”).

              You are completely correct that, although our results find that the ability of this suite of TBMs in reproducing the terrestrial C sink appears to be disconnected from the N cycle, the central purpose of the N cycle is to constrain CO2 fertilization which is not directly evaluated in our study. Evaluating the ability of a TBM to simulate present-day N cycling processes is one way to assess its ability to simulate N limitation of CO2 fertilization in the future albeit an indirect one. While not a direct test of the sensitivity of the C cycle to the N cycle, the significant variability across present-day N pools and fluxes that we assessed suggests that there could be uncertainty in future projections of the terrestrial C sink. We agree with you that a robust test of the response to CO2 fertilization and N fertilization across this ensemble of TBMs would be ideal for evaluation of the N cycle. In our revised manuscript, we will clarify this distinction and the limitations of our study. We will also emphasize the importance of experimental manipulations during future model evaluations and intercomparisons.

Furthermore, as per your suggestion, we will add an additional assessment of the correlations between the bias in simulated BNF, vegetation C:N, and soil C:N and the bias in simulated NBP (as well as other C cycle variables). If this analysis yields interesting results, we will add it to the revised manuscript. In our original analysis, we assessed correlations between the overall scores for BNF, vegetation C:N, and soil C:N and the overall score for NBP (as well as other C cycle variables) in Figure A3. However, no significant correlations emerged which is likely due to the many structural differences between models that confound these relationships. In the revised manuscript, we will highlight this analysis more in the main text.

              With respect to the minor comments, we will update our explanation of BNF in CLM5. While Cawse-Nicholson et al. 2021 points to many useful products, they are only at the site level or at a regional scale. While Fisher et al. 2012 provides a global product, it is for nutrient limitation which, as discussed above, would require the simulation of nutrient fertilisation experiments in the TBMs. We considered a remote sensing product for leaf N at the global scale (Moreno-Martinez et al. 2018; doi:10.1016/j.rse.2018.09.006) but TRENDY model outputs are given as vegetation N rather than as leaf N. This necessitates an analysis of root and wood C:N and relative tissue allocation as we conducted in our study. We will include references to these remote sensing studies in the revised manuscript as they are useful resources for future work. We will also include reference to the energy exascale Earth system (E3SM) land model (ELM) (Braghiere et al. 2022). While not part of the TRENDY ensemble, ELM is an important example of a TBM with both N and P cycling.

Thank you for your positive review of our manuscript!

We agree that the major finding of our paper is that, despite the empirically established importance of N limitation of vegetation growth and CO2 fertilisation as well as its implementation in these terrestrial biosphere models (see Table 1), the ability of these models to reproduce the terrestrial C sink appears to be disconnected from the N cycle. Essentially, all models are calibrated to reproduce the behaviour of the C cycle and the terrestrial C sink over the historical period regardless of whether they represent N cycling or not. As such, although N cycling is implemented in these models, it does not have a consistent influence on C cycling across models. While this does not have a significant influence on historical simulations, this has important implications for future simulations of the terrestrial C sink (see Kou-Giesbrecht and Arora Geophysical Research Letters 2023). Our study’s main goal is to identify this important disconnect between C and N cycling and to encourage further investigation. As we discussed in our response to reviewer #1, our study does not directly evaluate how N constrains CO2 fertilisation because we only evaluated historical simulations that were performed for the Global Carbon Project (GCP). A robust test of the response to CO2 fertilisation (and N fertilisation) across these models would be ideal for evaluation of the N cycle but such simulations are not performed within the GCP.

In the revised version of the manuscript we will first outline the empirically established importance of N limitation. Then, we will explain how, despite N limitation being implemented across models, a disconnect between C and N cycling occurs because models are calibrated to reproduce C cycling over the historical period. We will also emphasize the importance of experimental manipulations (such as CO2 and N fertilisation) for comprehensive model evaluations and intercomparisons.

In our study, one of our aims was to identify how differences between representations of important N cycle processes are crucial to implementing N cycling. We identified four central N processes: N limitation of vegetation growth, biological N fixation, vegetation response to N limitation, and N limitation of decomposition. We explained how different models represent these central N processes and attempted to distinguish significant differences between models in their abilities to reproduce both C and N cycling datasets. However, no significant differences between models with different representations of these processes emerged. This is likely due to other structural differences between models that confound their effects (an issue in all model intercomparisons). Further analysis is clearly warranted in future work. We will emphasise this in the discussion section of our revised manuscript.

Minor points:

• Thank you for pointing out that this was unclear. We will add an explanation of the differences between datasets.
• We will add a figure showing the variability of the geographical distributions of N cycle processes across models to the supplement.
• We will add more concrete future research directions to the discussion section (as described in our response above).