Dear authors,
many thanks for your detailed responses. Please submit a suitably revised manuscript in which you fully address all concerns of the reviewers. I would like to ask you in particular to take the issues mentioned by reviewer #2 on board in more depth.
Best wishes,
Sönke

Dear authors,
my apologies for the delay in coming to a decision. I failed to find a second reviewer, therefore I reviewed the manuscript myself. I believe that the manuscript can be publishable in Biogeosciences, if you decided to further revise the manuscript according to the suggestions and comments by reviewer #1 and my comments below.
Best wishes,
Sönke


While I find that the manuscript has improved, I still see need for further improvements

Major comments:

My main worry with this manuscript is still that the authors state that they show their relationships to be robust against spatial upscaling, while they leave out the critical scale jump from the plot level to the 1km resolution level. I don’t think that this invalidates the results of the study per se, but I do believe that the design of this study is unsuitable to make claims about the scalability of the results, because for this one would need to address the scaling from the proximity of the forest plots to the 1km as well.

The use of the word detection is inappropriate. The paper demonstrates a sometimes significant log-linear correlation between these two variables, but does not attempt to disentangle the possible signal from canopy nitrogen from confounding factors, it does therefore not allow for a detection of canopy N trends.

The results of the manuscript are presented in a misleading fashion, because it states that (L221) that MTCI and canopy values are related via linear regression, whereas infact the authors use a log-linear regression. This needs to be made clear at every instance (for ease of writing possible by introducing two new symbols referrring to the log-transformed canopy values), since this affects the interpretation of the regression (i.e. the connection is not linear as written in the text) as well of the r2 value. I would also expect a reasoning as to why the authors believe that the use of a log-linear relationship between MTCI and canopy N is to be expected.

The reference list is incomplete (e.g. Bontemps 2011 and ESA 2010 are missing).

Minor comments:

please carefully check the usage of hyphen and punctuation.

L 10: I find this sentence confusing, because canopy concentration and content are nothing else then leaf concentration and content accumulated to canopy scale. Why then is it necessary to point to processes at “leaf and canopy scales”

L 17 specify whether the unit is per unit leaf area or per unit ground area

L20 here and thereafter (as noted above): note that this is a significant, log-linear relationship.

L25: starting with “, ultimately…” should be removed, because the paper, while showing that a log-linear correlation exists, does not provide the ground for such a mapping.

L31: “such as” before Zaehle & Friend.

L64:68: irrelevant here, can be removed.

L72ff: It seems to me that this paragraph would be better suited to follow L48, with the redundant sententences at the beginning of the paragraph currently beginning in L49 removed.

L96-100 can be removed

L154 how was fbiom determined?

L202: Rather: forest plots for which the dominant vegetation type of the gridcell in the MTCI data-set did not correspond to forest were excluded?

L203: Sentence beginning with “It was downloaded”. Remove sentence as add reference to previous sentence. Both references are missing in the reference list.

L211-215: This does not explain, why this aggregation study was performed. Of course you average across more forest plots when aggregating, but you also average across more MTCI grid-cells. I do not see why this would increase the confidence in the regression analysis. Please be more explicit as to why this should be of interest.

L290-294. Not necessary. Please remove,

L296-303. This partially repeats the results section. Isn’t the more relevant question to discuss uncertainties in the upscaling of 6m plots to 1km resolution?

L321 and all other places: is the R2 the adjusted r2 to account for the large variance in sample size?

L323: the use of may is not appropriate here. Either the correlation is mainly driven by the deciduous and mixed plots, or it is not. No need for speculation.

L362: Does this not suggest that the nitrogen signal is confounded by other factors, therefore limiting the ability to “detect” nitrogen with MTCI?

L380 I don’t understand what you want to say here that is new compared to the preceeding sentences.

L387: “…might reveal laborious”. Please be more explicit, do you mean to say that this is infeasible?

Dear auhors,

many thanks for the revisions. I think this is largely fine now, but I would like to ask you to include a short version of your response to the below comment into the manuscript. You can also revert your response to my comment on L154 - that seems to have been an oversight on my part.

Best wishes,
Sönke


Section 2.3.1 line 207. It is not clear why you need to resample the landcover map to the MTCI images resolution. If I properly understand you just want to identify and mask the field plots that changed from forest to other non forest covers. If so, you would just mask those field plots located in a landcover map pixel classified as those covers excluded from the analysis.

We decided to resample the Globcover landcover map (with an original resolution of 300 m) to MTCI original spatial resolution of 1 km before using it as a selection criterion for our forest plots to be on the safe side regarding plot selection and leave out plots located on heterogeneous MTCI pixels. For example, if one plot was located in the only 300 m natural vegetation area of the 1 km MTCI pixel, e.g. a small forest patch surrounded by agricultural crops, the MTCI pixel value would also be influenced by the non-forested area surrounding the plot. Resampling the Globcover landcover map was carried out using the majority option. That way, by using the resampled landcover map (1 km) as the selection criterion, we make sure that the plot located in the isolated vegetation patch is excluded. Also, the number of plots selected for both the analysis with and without resampling of MTCI pixels, i.e. the analysis at the original spatial resolution of 1 km and the analysis at the resampled spatial resolution, is equal. The difference in the results cannot be attributed to the difference in the plots selected.
However, when the analysis is conducted using the original landcover map instead of the resampled landcover map, the number of plot selected is almost the same (n = 866 instead of n = 846) and the relationships observed between the variables are almost not affected and we thus propose to leave out these results as they provide negligible additional information that could be relevant for the interpretation of our results.