General comments:

The authors address with their research an important aspect of offshore wind - modelling of operation and maintenance. While many models exist, especially for fixed bottom wind turbines, validation and verification of said models is lacking to date. The authors compare two models and present the modeling results, and most importantly driving factors for differences between the two models.

The methodology of the study is mostly well explained and fits the purpose of the study. 

In some areas, the research article can benefit from revisions, mainly consisting of how the ideas and findings are presented.

Specific comments:

In section 2.1 it is not quite clear which modelling principles are general and common for all kinds of DES models, and which are concrete examples for the WOMBAT and UWiSE models. 

In line 147f you write: "... to how DES-based O&M models in general represent operational states and process flows. The shared modelling principles of the DES models are summarized below:" However, there exist DES models (in the context of (fixed bottom) offshore wind O&M models) which use slightly diverging methods than the ones listed below. Some examples include:

Failure modeling: While, to my knowledge, most models use Weibull distributions for stochastic failure modelling, some models use Poisson or Bernoulli processes. Linear fatigue accumulation or exponentially increasing failure probabilities are also seen in the literature.

Weather: Recorded historical data is most commonly used, however not all of the datasets have an hourly resolution. The FINO data, which is commonly used by European researchers, includes 10min resolution for wind speeds. 30min wave measurements are also common. Additionally, sometimes simulated weather based on the historical data is used.

I believe therefore that section 2.1. will benefit from rephrasing and making the distinction between the two models investigated in this paper and the general body of DES models clearer.  

Section 2.2.: 

185: I am not familiar with the use of "harmonized" in the context of creating a baseline scenario - can you explain the difference of creating a baseline scenario to be used in the analysis and creating a harmonized baseline scenario? Are you hamonizing between multiple existing baseline scenarios? If there is no difference between the concepts, you might consider removing the use of harmonized in this context.

187: What is the order of magnitude of the "multiple random seeds"? If possible, provide the reader with a concrete number (i.e. "using 17 random seeds").

Figure 1 and 190ff: From the written explanantion, I am understanding that you are iteratively modifying the assumptions in the UWiSE model and then comparing those results to the baseline scenario. In Figure 1, you write "Adjust one modeling assumption in UWiSE to align with WOMBAT's assumption". This second part of aligning with WOMBAT's assumption is not mentioned in the description. Could you please explain this further?

220 and Figure 3: It would be beneficial to mark the BoP components in the overview, to make the distinction between components used in both fixed bottom and floating wind farms and those specific to floating even easier. 

Section 3.1:

While reading about the simulation results, the reader would greatly benefit from having deeper knowledge of the two models - if at all possible, could you include all of the assumptions and boundary conditions already in Section 2? (e.g. the implementation of minimum charter period, first mentioned in 267ff), Table 2 already summarized some of the differences, this might be a good place to include this kind of information.

Section 3.2:

I do not quite understand the value of scenario 5 at this point in the paper. Scenario 1 is limiting the number of crews being transported per vessel, the hypothesis will be that this increases the time it takes to complete maintenance tasks, hence increasing the length of downtime events. Scenario 2, with uninterupted shifts, has the potential to reduce downtime, by removing the mid-day crew change and handover. These two scenarios Applied at the same time will most likely, cancel each other out to a certain extend? Similar arguments can be made for scenarios 3 and 4, one increasing (wind turbine) downtime and one having the potential to decrease (wind farm) downtime. Applying all four scenarios at once (while most likely very easy in the verification setup you have built), does not offer an obvious benefit. Later, we implicitely learn that scenario 5 makes UWiSE most like WOMBAT, could you please explain this reasoning behind scenario 5 already at the start of section 3.2. to avoid confusion?

324ff: It would be more logical to present the results for the scenarios in "chronological" order, i.e. starting with scenario 1 and ending with 4. 

Section 3.3: The unit of k€ /MW /y as presented in Figure 15 (and formatted wrong in the text) is a concept that is not readily understood. Cost per MW or even cost per MW*time (e.g. MWh) is more readily understood. Consider refomulating, in order to avoid the double fraction, which creates ambiguity. 

Technical corrections:

Caption Figure 1: Iteratre -> Iterate

229: suggestion to add an additional comma: towing as well as reconnection -> towing, as well as reconnection

238: replace "," with "."

323: observed variation is maintenance cost -> observed variation in maintenance cost

353: On opposite -> on the opposite

Figure 13: The labeling inside the figure is too small. Consider including just one legend (as it is the same across all scenarios, labelling the scenarios next to the graph, reducing the number of scenarios shown in the figure, a combination of the above suggestions, or any other measures to improve readability.

Section 3.3: the units for costs are formatted wrong.

394: workin -> working

General Comments

Overall, the paper addresses an import topic of the economic viability of floating offshore wind turbines and the challenge associated with modelling these future operating scenarios under a high degree of operational uncertainty. The paper presents a useful contribution to the science of Wind Energy and is relevant to the topic of "Wind energy economics and markets with high shares of renewables" and overall should be recommended for acceptance if some specific scientific quality areas are addressed.

Specifically, the choice to iterate UWiSE simulations to approach the WOMBAT model output has not been justified. This is an issue since the output of the WOMBAT model shows a continually degrading availability which would not be economically viable, practically desirable for an owner and is not representative of typically observed wind farm availability trends. This suggests there is a fundamental issue with the inputs of this model and this should have been updated to provide a more representative availability profile before the model alignment exercise was performed. It is recognised that a stated objective is to provide “generalized insights on how assumption choices shape O&M simulation outcomes, thereby improving the transparency, robustness, and reproducibility of logistics modelling practices in the wind community” and the representativeness of the simulation does not affect this, however this somewhat unrealistic starting point reduces the value of verification and evaluation of operational models which the paper also aims to achieve.

The scientific quality and presentation are of a good standard with clear explanation of the investigation approach and sufficient detail about the modelling tools and inputs to allow a full and fair assessment of the work. Given the open-source nature of the WOMBAT tool, it is recommended that a full input data collection for this model is provided so that this model can be independently scrutinized and ultimately other researchers can verify their own models and establish this work as an important reference study.

Specific comments

Introduction is well written and clearly states the objectives.

Section 2.1 – O&M Models

Stochastic failure modelling – A single weibull distribution is typically not able to capture the behaviour of complex electro-mechanical machines with different shape factors observed for different operational periods, especially representing early-life and late life-failures of mechanical components such as gearbox, pitch and yaw systems. While a reasonable approach to use a single failure rate per subsystem / failure type for this exercise, some commentary on the impact and sensitivity of models to this assumption should be provided. This is recognised in ‘Future Work’ but suggest it should be moved to Section 2 when describing the modelling approach and other limitations in Future Work remain where they are.

Section 2 - Strategy and dispatch logic and Section 3 ‘O&M Strategy; – The dispatch logic is not sufficiently detailed but rather described as ‘a coordinated logistical framework’ which is partly driven by overlapping maintenance demands. It is evident how the constraints would influence dispatch feasibility and simulation implementation but not how the prioritisation logic is implemented. Specifically, how do the models prioritise the utilisation of resource when there are constraints to decide which actions are done and in what order? Table 2 describes the modelling assumption differences once an action has been allocated but not the action prioritisation decision logic. To use a practical example, is unscheduled maintenance always prioritised over scheduled maintenance (or is this a model input or time variable), are minor failures that can return turbines to operation sooner prioritised over longer failures, is it a shared technician resource pool and if yes how is it prioritised?

Verification process, see comment in ‘General Comments’

In model inputs, duration of some activities are < 1hr but weather models are noted to be hourly intervals, the simulation resolution should be clarified and whether maintenance actions are rounded up/down or weather model is sub-sampled.

Results

20 runs may or may not be sufficient, there are no error bars presented on the main results for costs and losses. A presentation of model convergence and uncertainty of these key results is required to assess whether sufficient simulations have been performed and provide better context overall (such as when presented in Figure 6 and Figure 7).

Scheduled inspection – this level of incomplete actions would be in breach of any turbine warranty or servicing contract suggesting the models in both baseline models have insufficient resource to be fully representative of real-world operations. Only scenarios 3 and to some extent 4 using UWiSE presented later show broadly acceptable levels of scheduled maintenance (>70%).

The comparison of models generally provides satisfactory explanations for the variation between models but should be presented in order consistently.

Technical corrections

Some figures (particularly Figure 13 could benefit from increased font size)

1. Page 8, Figure 1 flowchart – "Iteratre with next assumption" should be "Iterate with next assumption."
2. Page 3, line 50 – "to authors' best knowledge" should be "to the authors' best knowledge".
3. Page 3, line 56 – "tow-out/in" should be "tow-out/tow-in" for clarity.
4. Page 12, line 257 – "in WOMBAT then in UWiSE" should be "in WOMBAT than in UWiSE"
5. Page 12, line 258 – "between two models" should be "between the two models"

1. Page 15, line 282 – "In UWiSE, time-based availability occasionally drops to around 80% in certain period" should be ""in certain periods".
2. Page 20, line 352 – "On opposite" should be "On the opposite", "Conversely" or "In contrast."
3. Page 24, line 395 – "workin vessel" should be "working vessel."