The authors present results from an extensive parameter optimization study for a global scale ocean biogeochemistry model. They evaluate the impact of different optimisation set-ups, in terms of observation based data used, parameters to be optimized, and model spin-up length. Particularly, they focus on the benefit that assimilation of organic tracer information (phyto- and zooplankton, POP and DOP) might have for the optimization of ocean biogeochemical models. The authors generally find that the inclusion of organic tracer data has a strong impact on the representation of particle sinking in the model and improves model fidelity with respect to global oxygen biases and the represenation of oxygen minimum zones. Different spin-up times are shown to have a considerable influence on global oxygen and OMZ volume biase, even for optimized parameter sets that perform similarly with respect to surface nutrient and oxygen data.

Parameter otimization of global biogeochemical models is understudied due to its complexity and technical challenges, and this study is a very welcome addition to the field. The results are useful (beyond the technical aspects of optimization) for a wide community of ocean biogechemical and Earth system modellers since they tell us about model sensitivity in general. The paper is well written and the method is sound. I didn't find any problem with this manuscript, except a few rather minor points where the manuscript would benefit from clarifications. These points along with technical corrections are listed below.

General:

1) It is a bit confusing that the term J_RSME (equation 1) is used to denote the misfit (cost) function used in the optimization procedure, but also for the a-posteori (after optimization) quantification of model misfit. Although both misfit functions take the same form (eq. 1), they are different in which data and regions are considered. This is not explained in the methods section, rather there are only some hints scattered in the text. For example, the reader can guess what is shown in Fig. 1 based on the fact that the J_RMSE values are the same as in table 2, where there is a note in the caption. To make this more transparent, I would suggest to describe this two-fold use of the misfit function in the methods section. I also would prefer to use a (slighly) different notation for the cost function and the a-posteori misfit function (e.g. \hat{J} for the cost function).

2) The representativeness of the data products used for optimisation is discussed several places in the manuscript, but it would be helpful to gather a short description of this aspect in the methods section. The WOA climatologies for nutrients and oxygen should provide to first order a like-to-like comparison with the coarse model and the climatological ocean circulation. The same is probably true for the chlorophyl data? The discussions that are found later in the manuscript point towards the fact that the Martiny data is too sparse to be representative of the simulated model counterpart. It would then also be useful to frame the later discussions more consistently as a problem of representativeness, for example lines 262-266: Is this really a problem of the coarse resolution, or more the problem that the available data are not representative of a climatological average over a 1x1 degree gridcell? The same comment applies for lines 286-287, and also for lines 321-328 (where representativeness is finally mentioned). 

Also, is it really plausible that the lack of correlation (for large scale global patterns) can be explained by errors in the (data assimilated) ocean circulation? Doesn't this potentially also point towards a too low model complexity, i.e. only one phytoplankton and zooplankton type?

More specific comments:

-line 2: "...state of the ocean biogeochemistry..." I would find it justified to delete the word biogeochemistry here. The models tell us something about the ocean in general.

-line 15-16: "mainly located in surface layers" is a bit unclear, please consider rewording.

-line 32: Consider adding "combined with data assimilation techniques" or similar after "Global biogeochemical ocean models"

-line 49: "Far less than half of the studies...". Unclear which studies this refers to. Please consider rewording to clarify this.

-line 82: "one of the simulated compartments." It is unclear to me what "compartments" refers to (inorganic/organic? tracers/fluxes? nitrogen/phosphourous?)

-line 83: "basic optimisation procedure". Is "basic" a good word here? Maybe better "reference"?

-line 79-90: It is confusing that it reads "three further experiments" and "these five optimisations". The fact that the "basic" optimisation is actualy two different optimisations (one with range of b more constrained) is difficult to understand. Please consider explaining this better.

-line 101: consider changing "a circulation" to "a circulation field" 

-line 170: instead of saying "the DOP parameters", the two parameters could be spelled out for clarity.

-line 169-174: I don't understand the logic behind this experiment: If the objective is "to analyse whether the neglect of iron

limitation in MOPS yields a bias in parameter estimates" then why is the number of paramters to be optimized changed at the same time? This way it is not clear whether changes in the optimized model performance are due to the change in data coverage or due to different set of parameters to be optimized? Could the authors please comment on this?

-Table 1: instead of using italic font for fixed parameters, why not write "fixed at 0" and "fixed at 0.1848" for the two cases where this is relevant? Would be easier in my oppinion.

-line 204: "introduced" sounds odd to me. Maybe just "used"? 

-line 231: Why "a likely larger number"? L is given in the table, or does this refer to something else? Please clarify.

-line 252-253: "...in contrast to a reduction of JRMSE by about 25% for the consideration of DOP measurements." I don't understand this. In Fig 1a I don't see a very significant difference between S6-all and S6-DOP?

-Figure 4, caption: For panel b (grazing), I can't see any short or long horizontal bar? 

-Figure 7, caption: "Numbers on top of the panels..." I can't see any numbers on top of panels?

-Figure 9, caption: I don't think "sources of variability" is a good wording, please consider rewording. Also, "which is different for the individual model setups": It is actually only the L4-SO setup that is different, right? So maybe "which is different for L4-SO compared to the other setups"

Technical/typos:

-line 28: non -> not

-line 273: expose -> show (?)

-line 278: organics -> organic tracer data

-line 370: agrees with -> falls within

-line 373: too low -> below all other estimates