Dear Rahul Raj and co-authors,

Thanks for the detailed responses to the two reviews your manuscript, “Bayesian integration of flux tower data into process-based simulator for quantifying uncertainty in simulated output”. Although i notice significant updates of the manuscript, I still believe that you have not fully answered few major concerns of both reviewers. I thus would like you to take care of them before

Reviewer #1 (and #2): About the use of time varying parameters: Although your experiment 2 is well defined, the request from reviewer #1 of having a simulation with time varying parameters sounds justified. You mention “This approach, however, is not possible in BIOME-BGC because it is a dynamic vegetation simulator”. Such answer is not obvious to understand. You could indeed increase the size of your “vector of parameters to be optimized” in order to have one parameter per month or proceed as proposed by reviewer #1 without significant model changes. This could indeed help to possibly identify what are the missing processes in the model controlling the temporal variation linked to the LFRT, FRC:LC, C:Nleaf, FLNR parameters at least. More generally the implication for model deficiencies that such experiment could highlight (either your current experiment 2 or a “true” time varying parameter experiment) is still not fully investigated in the new version.
About use NEE data to calibrate the model: Your answer “We argue that calibration of such a complex model to NEE instead of GPP may not be a good idea but calibration to NEE or respiration in addition to GPP is possible” sounds not fully justify and partially answer the reviewer’s request (given also that many studies have used NEE flux data to calibrate similar models)
Reviewer #1 ask your to discuss your finding in the context of other studies that already performed a Bayesian calibration of BGC-models against flux data. You have probably misunderstood such concern as indeed the reviewer does not refer to BIOME-BGC model but to BioGeoChemical models in general(BGC-models). He mention for example DALEC model but there are many other studies using FluxNet data based on model of similar complexity (JULES, ORCHIDEE, CABLE, LPJ, …). Please replace your study in the context of previous finding in order to bring specific conclusion that are relevant not only for the users of BIOME-BGC.
Reviewer #2: About his concern that a “thorough analysis of the results from Experiment 1 is missing”. You made a substantial effort to analyze more in depth the results but the findings are still fully put in perspective with previous studies (i.e. to draw general messages for modelers in terms of Data Assimilation or Forest C-Cycle modeling).

Finally as a personal note: your change of the colors for figures 4-5-6 does not make enough distinction between the two “grey zones” (for the model and the data). Your initial colors choice was more clear.

I am awaiting for your response before proceeding with the reviewers requests.
Best regards,
Philippe

Dear Authors,

Please find attached a document summarizing my last concerns about your paper before considering it for publication in GMD.

Best regards,
Philippe

Dear Authors,

Thank you for your revision and for taking into account my comments.
The article is now suitable for publication in GMD.
Best regards,
Philippe