General consideration:

This manuscript shows the impacts of climate change in the European cropland and grassland production systems towards year 2100. Specifically, by using two biogeochemical models, the work focuses at assessing changes in plant productivity and biogenic GHG (N2O, CH4, CO2) balance. The paper is well detailed and structured, the language is appropriate and is fluently flows. However, to my opinion, some very important weaknesses are presents:

1. The novelty of the paper is quite low. Further modelling studies have been carried out at larger scale to assess changes in productivity and GHG emissions. The 2100 threshold is quite far and, even in the view of the EU policies mostly focused at nearest GHGs thresholds, the provided results risk to be very speculative and not very close-to-reality to provide suitable information for policymakers as suggested in line 74-75.
2. Modelling works do need appropriate model parametrization such as climate variables, soil properties, vegetation parameters and management. Whereas climate and soil properties are clearly reported, several information in the other components are missing or shows very low confidence. For instance, for the management characteristics, I appreciate the effort of the authors to obtain all these information. However, some parameters need to be better explained and discussed. For instance, sowing dates and fertilization were imposed but, for sure, these cannot reflect all the possible variability observed in the whole EU. I understand the need to impose fixed parameters for running the model, however it is also necessary to indicate the results can be affected by uncertainties due to the application of these fixed parameters. My main concern is about crop parametrization. In line 175-176 authors indicate a crops parametrization based on those applied in previous/other works. However, no reference has been reported. Also, I was wondering how authors were able to retrieve and summarize all these crop/grassland information since modelling study are often carried out at single point/area, with different parametrization for the same crop over different areas, and do not report these data. This is also true especially for grasslands where, as the same authors says, very low information are presents. There is no explanation about these parametrization (i.e. water efficiency, radiation use efficiency, maximum and minimum productivity, etc.) were found, from which studies, and how these parameters were summarized to find the most representative/suitable for each of the applied crops.
3. My main concern is the point 3.1.1., model validation. Model validation should provide a confirmation about the model capability to represent the crop growth, development, and production in different environment. However, looking at picture 1a, does not seem that models are able to reproduce the correct behavior of each single crop. Putting all crops together may create an inthrinsic error and does not indicate if each single crop is well reproduced. For instance, looking at sugar beet or potato, I’m not able to see a proper correlation between modelled and observed crops. In order to provide a robust validation, each crop should be singularly validated and then reported, so as readers are able to see the model confidence and robustness at reproducing the crops. This is needed also to provide robust information about the expected changes due to climate change. Therefore, I suggest authors to provide single validation for each crop and then summarize them in scatterplot or table with the relative statistics. Concerning grasslands, results are very poor for all areas. For instance, for the sole Mediterranean data (L272) the overestimation of 55% do not allow to indicate these data as robust enough to be accepted. This overestimation led to unplausible results under future conditions, resulting not useful and misunderstandable for readers and policymakers. I understand the lack of data, however more recent and affordable information on productivity could be taken by remote sensing and new analysis may be done to provide more robust results. I understand the effort of the authors, but grasslands systems are very complex and sensitive to climate change, especially considering the dynamics involving water reduction and species changes. These results do not take in account changes in composition that, in turn, also affect productivity, neither provide robust statistics. Finally, as indicated for crops, at least model calibration/validation for single areas (if not for pasture composition) should be carried out, to make results more robust.

I appreciate the wide analysis done by authors; however, the above-described issues make the level of confidence of these results very low. Discussion section do not address this high level of uncertainty and only report the agreement in the impacts with other studies. This information is quite negligible if not accompanied by a strong model performance in the magnitude of the results. Simulated dynamics are found expected since driven by common algorithms (i.e., GDD for crop growth), but robust information about the expected change need to be provided. Whilst crop data may be improved, I’m quite concerned about pasture dynamics which need to be adequately addressed through a more consistent approach since, at this time they cannot be accepted as here proposed.

The manuscript describes the set-up for spatial runs of crop productivity and emissions from croplands and grasslands in Europe. This is analysed on simulation model results provided by two model simulation approaches for a historic (1985-2004) and a future period (2005-2999). The future scenario simulations include two scenarios, representing a moderate and an extreme climate development.

Spatial data about emissions of croplands and grasslands are very important and model approaches are most suitable to provide these spatial data. The data demand for models is high and the errors and uncertainties of used data as well as of the model results is high. This combination makes the judgement of the quality of spatial model studies on large scales, like this one here, very difficult. There is no doubt that the technical set up of the model is up-to-date and the used data represent the state of the art quality of available data. However, there are still a couple of genera aspects that let this manuscript down:

• I mentioned the uncertainty and error affected data. I appreciate the open communication about the used data and the limitations in the approach. I see the extrapolation of the data for fertilizer application rates and crop rotation difficult. This extension introduces a different error and shifts the uncertainty compared to the period these data sets were originally aggregated for. Changed climate will affect the rotations and the fertilizer application rates and it is difficult to predict these changes. I do not mind that these data are used, but I would expect a more critical discussion on these points and a more careful conclusion.
• I highly appreciate the detailed comparison of the different results with literature, but I see two problems with this. One is that some of the studies that are used for comparison are model approaches that used similar data and use similar model approaches (still worth to compare, but not necessarily completely independent). More relevant, I do not see the new or innovative contribution of the here presented data. There is a lot going on in the manuscript and it is not all linked with each other. Even though, croplands and grasslands are analysed, both are analysed parallel and not integrated. The most obvious connection (using manure from livestock production) is not included (I understand why and I agree to this point) and spatial emission budgets are not provided (as land use data are not used). Overall, the data are nicely presented, but more moderated than discussed. This is also reflected by the conclusion, which is more a summary of the manuscript and the conclusive sentences are not really convincing (lines 773-774: I do not think that the study proves this, even though it adds a strong indication and to conclude that C stock changes are variable over time does not require a complex study). I think that the errors and uncertainty and their impacts on the results should discussed stronger.
• The presentation of the results for the historic period sounds partly like observed data or robust baseline. Considering the uncertainty of the use data (the crop rotation and fertilizer data are also simulation results, soil maps are developed over larger time periods and climate realisation are partly weak in the representation of extreme events) this is not a robust baseline. As mentioned above, the data that are used for the historic period are aggregated in a different way than the data for the future scenarios. While the trends, productivity and emissions during the historic period are driven by management changes (the farmers action is considered, as these are statistical data sets) the future scenario results are driven by climate impacts. Both can be presented and in the end they can be compared by a direct comparison including conclusions is difficult. Figure 5 shows the spatial distribution of the fluxes, but it would be great to see the emissions separated for croplands and grasslands. There are a couple of places where the N emissions surprise me. I would expect higher N emissions around Paris, less low emission spots (and more extreme emissions) in the Netherlands and I cannot follow the logic that North-West Germany shows lower emissions than the East of Germany (historically simulations). I am also critical about the results for Eastern Europe and, again, using these data for future extrapolation. Eastern Europe make these data doubtful as baseline for the future simulations. While management in historic data changed due to political changes or to adapt changing conditions, the management for the future simulations remains constant. I think simulations on the historic data are a good point to prove the functionality of the models, but I would only compare the future simulation scenarios with each other. Obviously, you can show general changes and trends, but I wouldn’t use the historic simulations or observations as baseline or reference. Errors and uncertainties are different in both data sets, which affects the results and the differences.
• The spin-up for such a model approach is fairly short. Was the model tested for equilibrium in all grid cells? What were the criteria for equilibrium (no change over 10 years? Only 1 % change over a fixed period?). If the spin-up was only tested for an average, I wonder, if the authors think this is good enough? Where did the authors got the initial SOC values from or did the model used spin-up SOC?
• I have some concerns about the management and the potential impact of the management options on the results. I find the amount of 80 % residues removed from the field very high, compared to other comparable model approaches. Do the authors think that this is a realistic number for Europe? I didn’t find any support for this number and would like to understand how the authors came to this result. Changing this number will change the NEP significantly. I also wonder, if the authors thought about introducing a second growing period. The shorter growing periods are only presented as a problem, but this can be used for double cropping seasons. Again, I do not mind limitation, by they should be discussed and in a best case the relevance (quantitative impact if possible) estimated.
•  

In summary, I think this is an interesting study, but I do not see the clear objective. For me all the different points does not come together. There are plenty of options to improve the manuscript (focussing on single aspects; defining key/new aspects of the presented study; include a more critical discussion about the limitations, including estimates about their relevance; etc.). In the actual form I am not convinced by the manuscript, but there is definitely enough potential for improving it,

Line 51: become neutral

Line 179: species

Line 205-209: Was cutting and grazing on all grasslands?

Line 218: Do you mean 1901-1977?

Line 309: decreased (check writing).

Line 309: What is driving the yield reductions for maize: drought, heat, water or nutrient limitation? A shorter growing period or other weather impacts have a stronger impact for spring crops, as the growing season is short and impacts show a stringer effect.

Line 360: Higher productions

Line 654: lower case 2

Line 686-689: is this a realistic effect or a model artefact? I would expect increased emissions and leaching, as there is increased soil water content and water transport in the soil. Even though the nutrient up-take by plants is increased, I would still expect an increase of leaching and emissions. Only optimised systems would show a decrease.

Lines 707-710: Does the decline of NBP means stringer sink? Is this not affected by increased production due to rising temperature in an area with sufficient water availability?

Line 779: agro-ecosystem

Figure 1: please mention the spatial unit that is used here. Does each point represent a NUTS2 unit or a country?

Figure 5: Please indicate in the legend delta N2O.