The paper is summarizing an important effort to simulate the impact of climate change on lakes using an ensemble of lake models and climate change scenarios. The paper is well-written and suitable for GMD. However, the presentation can still be improved following the minor revisions below.

Comments:

• In general, can you maybe also comment on the computational cost and the size of the datasets somewhere? If those simulations are cheap, you should also state this somewhere.
• It is a bit difficult to understand whether the paper is summarizing results for ISMIP 1, 2, 3 or all from the abstract. This should be stated explicitly, early in the paper.
• P8L5: “with the aim of conducting innovative science” should be rephrased.
• P8L14: “were simulated for existing lakes in the local domain and “representative” lakes” is rather cryptic and should be rephrased.
• P8L20: “for this first phase of lake sector simulations” I thought this would be about ISIMIP 2 and 3?
• P9L18: “is expanded”, can you be more specific?
• P9L24: 0.5/0.5: You are using different ways to write ½ degree resolution throughout the paper. Maybe you can state once that you are working on a lon/lat grid with ½ degree grid-spacing and then remove the information from the rest of the paper?
• P10L22: “for which the original data was remapped from 30’’ to match” This needs more information.
• P13L13: “the lake layers can be defined according to water volume” No sure what this means.
• P13L20: “suite of Swiss lakes” Can you be more specific? What kind of predictions, how many lakes?
• P14L15: Simstrat (v2.1.2) was described above, not Simstrat v1.4, or am I missing something?
• P20L20: “^-0424” is not a very nice format.
• P24L17: “robust fit for all eight local lake models was found” What does this mean?
• P25L16: “calibrated parameters c” This should be rephrased.
• P29L24: “the first stage of simulations undertaken by the ISIMIP Lake Sector” But this was about ISIMIP 2 and not 1?
• P29L27: “it will clearly not be the case for” should be rephrased
• P30L2: “Within the ISIMIP framework, the simulated climate change impacts are (inter-) comparable with results from 13 other sectors, supporting cross-sectoral aggregation of impacts (Vanderkelen et al., 2020).” I do not understand this sentence.
• Table 1: “L” and “G” should be defined in the caption.
• Table 2: What does the dot in brackets stand for? Maybe?
• Table 3: “Calibrated parameter name” should be rephrased. Maybe “Names of model parameters that need to be calibrated”
• Table 3: “Cross-site calibrated parameters summary statistics” Can you explain exactly in the caption what the two numbers are standing for and what the numbers in the brackets are?
• Table 4: “Highlighted” I guess this refers to the grey colour?
• Figure 1 A: Why is there more than one colour for the dots?
• Figure 2: The panels are too small, and A, B and C are not visible.
• Figure 3: I do not understand the sub-plot on the bottom left.

The paper submitted by Golub et al. is a presentation of the lake sector of the ISIMIP initiative. ISIMIP is definitely a key project to structure our community and the work done is absolutely amazing. The lake sector in itself is also very important, in particular because most of the Earth system models don’t include lakes despite their importance in the land/atmosphere exchanges of heat, water and greenhouse gases. The paper is well written and clear. It fits well with the scope of the journal. I proposed few modifications to help improving the reading. Two may need a bit of restructuration and all the others are minors.

• First, I found a bit misleading to mix results from ISIMIP2 and the new protocol of ISIMIP3. To improve this, I propose first to change the title of the paper into something like “The ISIMIP Lake Sector: an ensemble modelling of climate change impacts on lakes worldwide. Results from the last simulations and framework for the next step”. I also propose to better separate in particular in section 3 what is coming from ISIMIP2 and what will be done in ISIMIP3. Maybe a dedicated section for ISIMIP 3 will help the reader to make the difference between what is done what will be done. Another option could be to split the paper into 2 parts to separate both but I let the decision to the editor.
• You compared model metrics in table 3 (RMSE and R2) but the models can be adjusted on a different number of parameters. As a consequence, the model with more parameters to adjust can better explore the space and at the end have better RMSE or R2 because of an overparameterization. I suggest to used metrics that take the number of adjusted parameters into account such as the BIC.

Minor comments:

P4 line 16-17 “…lakes are among the most anthropogenically altered ecosystems on Earth…” This kind of statement is a bit weird I would prefer to read specific examples such as eutrophication, change in the water regimes etc.

P8 line 14: Maybe worth to better define what is a representative lake.

Section 3.2: In general when the ISIMIP3 protocol is presented it would be nice to have few lines to explain what is the rational behind the modifications from ISIMIP2 to ISIMIP3.

P10 line 22-23: So at the end you may have situations with more than one lake within a grid cell. In this case how do you manage, do you "merge" the 2 lakes to have a single water body or do you have a more complex description of the sub grid heterogeneity.

Section 3.3: It would be interesting to know what are the criteria to be included in the group of models. For instance, should a model pass a couple of benchmarks before being incorporated? Is it based on the representation of some key mechanisms?

Section 3.5: I am not really a specialist of lake model parametrization so I assume that it was done following state of the art methods.

P24 line 10: Maybe it worth reminding here that all information to download the data are in the code and data availability section.

Table S2 and S3 for some models (GOTM, CLM Mylake there is no answer to the question "Was a spinup scenario used?". I guess that when it is not answered yes it means no but it should be clearly written to avoid confusion.

Fig2: The letters are visible in the plots.

General

This paper by Golub et al. presents the simulation protocol by the Lake Sector of the ISIMIP to model the impact of climate change using an ensemble of lake models, a large and important effort. The paper is well written and structured and the topic is interesting  for the journal audience. Please find below some questions and comments that require mostly minor modifications for further impovement of the paper:

• While there is some overlap between lake models (ALBM, GOTM and Simstrat) for local and global simulations, it is not clear to me why not all 6 global lake impact models were also applied to the local domain and calibrated in this setting. This should be clarified in the paper. In fact, you could use the same model sets for both applications. This approach can also give useful information about the value of default/a-priori model parameters at the global scale when compared to local domain calibrated parameters and simulation results. In the Conclusions lines 9-12 states that reasonable parameter and coefficient values from the local domain were used in the global domain. How was this exactly done (it seems this is only described for GOTM)?
• The water balance is not considered as part of the lake modelling efforts (paragraph 3.5.3). In lines 20-23 of paragraph 3.5.3 the explanation is that one should use caution for (only) seven lakes or reservoirs with large water level fluctuations (Table 1). If think the wrong Table is referenced here?  Seven lakes or reservoirs are part of the local domain, what about the global domain? And what are large water level fluctuations (definition)? But even without large water level fluctuations, large input-output changes (inflow, outflow, precipitation, evaporation) because of climate change can have a significant impact on lake temperature (e.g. changes in residence time)? This part in the paper requires more explanation, either including appropriate references  that confirm that the omission of water balance components has only a significant impact on lakes with large water level fluctuations, or rewrite this to a more cautious statement (this omission can have a broader impact).
• It is stated that the global lake models were not calibrated because of lack of a global-scale data set of measured lake water temperatures. What about using other datasets like satellite based datasets for example for surface water temperature? Or are there any (planned) efforts to setup a central data repository to collect measured lake water temperature (and other variables) data, for example similar to the Global Runoff Data Centre (discharge data for hydrological applications)? Also, when the water balance is considered, additional data like water level, surface area dynamics etc. (for example from satellite data) could be considered for calibration/validation purposes. Would be good to add a section/alinea to the paper that considers some of the solutions/ideas from the authors for a of lack of a global-scale data set.
• Generally a calibration strategy also includes the validation of the calibrated model. It is not clear if this was done, please describe this in more detail or explain why a validation was not carried out.
• An extra section to describe future work (ongoing ISIMIP 3, possible solutions lack global dataset?) would be useful (see also the last specific comment below).
• And, finally, briefly some differences between models (P26, lines 24-27) are mentioned. Are there more examples from his study? I think it could be useful to include this kind of information more extensively in the paper.

Specfic comments:

• P22 line 6-7: different objective funtions were used by the different models. Why was not the same objective function used for each model? I think using the same objective function is an important aspect of an ensemble modelling protocol.
• Table S2 and S3 seem to have missing information in some table cells (empty)
• P22 line 25: spinup periods were different. Bit similar to objective function, please state clearly why not the same spinup period was used (if applicable) for each model.
• P25 line 17: Gao reference is missing.
• Paragraph 4.1.2 Better move this part to an “Outlook or further work” section, this is not really a result, but part of possible further study.