Xi et al. add an expanded tiling approach to ORCHIDEE-MICT. This new approach, termed MICT-teb allows for per-PFT tiling of the energy budget with effects cascading to the hydrology and carbon cycle. The paper reads OK but has issues with clarity in some sections, which I note below. The actual approach is relatively straight forward and present a comprehensive overview of how MICT changed with tiling. Many groups have done this in their models and these results are similar to others, but it is interesting to see how it was implemented in ORCHIDEE and the impacts. One thing that I missed in the paper was a consideration of what actually makes the most sense to tile. Here they tiled by PFT, as others have done before, but the decision seemed to just be a default option rather than one carefully considered. Based upon my experience, I am not sure if tiling by PFT makes the most sense. In some landscapes the PFTs are truly distinct (e.g. peatlands vs. nearby uplands), but in others they likely shouldn't be separated (e.g. savannah landscapes). I would like to see some more discussion about their choice to tile based upon PFT and then at the end a discussion of whether that choice was a good one. Within our group it has been a learning process of when to tile and when it doesn't seem worth the complexity (e.g. we have tried by PFT - Shrestha et al. 2016, soil texture - Melton et al. 2017, and by disturbance - Curasi et al. 2023). I personally seem to have landed on peatland vs. upland and by disturbance requiring tiling, but I would be quite interested to hear of the authors' finding here. Otherwise, I have many small comments but don't forsee a problem with them being addressed so can recommend publication after acceptable revisions.

Smaller comments:

Line 48: Cooling effect already implies it is a reduction so should just be 2.5C, not -2.5C.

L 55: The sentence should be rewritten for clarity. 

Fig 1. I found this to be relatively hard to understand. First, the caption includes a duplicate jibberish caption label. For the figure itself, I am not sure what the red rectangles mean. The MICT-teb has only one but the MICT has two(?). I think this could better be redrawn for clarity. There are other model schematics in the literature that the authors could look to for inspiration.

Line 111: Reference for the <5% number?

L 123: specific heat 'capacity' of dry air

Table S1 - are your layer thicknesses really so thin? In our model we find stability issues if the layer thickness goes below about 10 cm. The first layer here seems to be 1 mm thick(?!). Also to be clear - this shows depth to the bottom of the layer or layer thickness?

line 144: missing 'taken'

eqn 5: why heat capacity now lower case, was upper case earlier.

Table S2: Minimum snow albedo after aging can be incredibly low (0.14). Even the oldest, melting snow should have albedo around 0.5 as far as I have ever seen. Can these low values be justified?

Fig S1 - where is the soil albedo from MODIS coming from? This is not straight-forward to produce so assumedly this is from some other source?

Fig S2 - this is hard to compare between b and c. The different PFTs are not labelled in a manner that would make comparison simple. As it is now, I can't really tell what the differences are except that the model's SOC drops too quickly with depth. It would be nice if this figure was improved to allow a proper comparison.

Fig S3 - the model and ref dataset appear to be on different grids. What are we supposed to make of the inland white cells? Does the model blow up in those locations or ? It would be good to see a difference plot to make it clear where the biases are, against at least this one ref dataset.

L 265: provide ref for CRU-JRA

L 266: Provide proper ref for land cover. Yes it is used by TRENDY but it comes from somewhere else. It is a lot of work to create these datasets so the least we can do is properly cite them.

I suggest you choose either mteb or teb. The model name uses teb but the text uses mteb. It will be more clear if you just use one.

L301 - didn't you say earlier in the paper that albedo was insensitive to soil moisture?

Fig 3: Is the peatland parameterization also a tropical one in addition to a boreal representation?

Line 306: How does this work 'We note that the cover fraction of a PFT in the model includes both valid vegetation and bare soil.' - wouldn't you want the PFT cover fraction to be only the PFT cover and bare soil fraction be accounted for separately? Or is this meant to be talking about the soil below the canopy? If so, I suggest you use different, more clear, terminology.

L 312: So roughness height is not static? Table S4 makes it seem like it is static per PFT. This is confusing.

Fig 4 + 5: I think I know what this shows for the MICT-teb simulation, e.g. bare soil is the values over the bare soil tiles, but I am less certain what that is compared to for the MICT simulation. Looking at Fig 1 makes me think that all panels of this figure will be comparing a MICT-teb tile against the same values from MICT. Is that right? If so, then naturally the differences will be large. The most interesting changes then are the grid-cell mean row. I would be tempted to put the rest in the supplement since the comparison is not quite clear as it is for the grid-cell mean. And then add a second row showing the relative difference (or a second scale as done in Fig 3). Have you also looked at the NH totals for these quantities (where applicable)? This might get at the problem with tiling, in my experience at least, where is it most valuable/needed? It adds a lot of complexity so should add some real benefit. I am wondering about something like Fig S14 but for the whole NH, not just a few cells.

Table 2: Have these effect sizes been checked statistically? It would be good to know which are more significant. Esp if you are assigning an effect size, when you are comparing two things of differnt units (e.g. albedo change and SWout change). I see this is done in Table S5, why not here?

Fig 6 has a lot of info but similarly to my comment about Fig 4 + 5, an important change is the grid-cell mean, which is not shown here.

L 414: Assumedly the quick equilibration is due to defined vegetation heights? If the vegetation could grow according to conditions, that should take longer than 2-3 years to find a new equilibrium.

L 424: But surely the subtle effects are when looking at grid cell mean values and not, e.g. the peat tile specifically? I would expect the peat tile to be much different (as shown in Fig 8). Perhaps tighten up the language here to be more specific.

Fig 9 - what is going on with the Tsoil between S0 and S1? For example, in panel e it looks like there is some sort of restart bug that prevents a smooth transition? Actually I am finding this whole figure confusing. What is grey vs. yellow? What is the x-axis? Is the S1 plotted in every figure, but just overplotted in some? Same questions about fig 10, with addition of wondering why the scale break for mean annual snow and why the values start so low?

Fig 12 - Here it seems that SOC for all three simulations is higher than for the 'OBS' (side comment- this is not observed since it is a gridded product of a fundamentally point-scale phenomenon, perhaps 'observation-based' would be a more accurate name), but it looked from Fig S2 that MICT tends to not have the soil C be deep enough. Does that mean that the SOC is now deeper with MICT-teb or is it just higher amounts in shallower soils?

Line 483 - same complaint as earlier about citing the proper refs for the model inputs.

Fig 14: The threshold fractions are the minimum fractional coverage of that land cover type in the gridcell? Unclear what is meant here... Also instead of 'SIM', please keep consistent and put what model was used to produce the values, i.e. MICT-teb.

line 537: I am not really sure how Fig S20 relaties to this discussion of a weak correlation between biases. Perhaps this could be made more clear?

Papers mentioned:

Melton, Joe R., Reinel Sospedra-Alfonso, and Kelly E. McCusker. 2017. “Tiling Soil Textures for Terrestrial Ecosystem Modelling via Clustering Analysis: A Case Study with CLASS-CTEM (version 2.1).” Geoscientific Model Development 10 (7): 2761–83. https://doi.org/10.5194/gmd-10-2761-2017.

Shrestha, R. K., V. K. Arora, and J. R. Melton. 2016. “The Sensitivity of Simulated Competition between Different Plant Functional Types to Subgrid-Scale Representation of Vegetation in a Land Surface Model: Competition Between PFTs.” Journal of Geophysical Research: Biogeosciences 121 (3): 809–28. https://doi.org/10.1002/2015JG003234.

Curasi, S. R., Melton, J. R., Humphreys, E. R., Hermosilla, T., and Wulder, M. A.: Implementing a dynamic representation of fire and harvest including subgrid-scale heterogeneity in the tile-based land surface model CLASSIC v1.45, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2023-2003, 2023.

Surface heterogeneity has large impacts on surface energy, water and carbon cycles. This study developed and evaluated a new multi-tile surface energy budget scheme in the ORCHIDEE-MICT model. Expectedly, the improved model indeed shows better performance in some regions, especially the permafrost areas. However, some issues are needed to be resolved first: 1) the research gap is not clear apart from using a different model; 2) many details on the methods are missing, e.g., the vegetation albedo calculation, the unreasonable assumption of 1 for the emissivity, the definition and estimations of surface temperature, the snow cover fraction parameterization, etc. 3) The statistical tests on the significance of the model-model and model-observation differences are needed; 4) The authors just compared the surface temperature between model and observations, and more comprehensive comparisons are needed using the benchmark datasets for all the surface energy balance variables, also the carbon and water cycles-related variables. Please see below for my specific comments.

Major concerns:

1. In the abstract, I suggest the authors provide some quantitative metrics for the performance of the improved and original models.
2. In the second paragraph of the introduction section, the authors only used the surface temperature as one example to introduce the background. However, surface temperature is just one import factor in the surface energy budgets. Actually, there are already many existing studies analyzing the impacts of surface heterogeneity on surface energy balance and water cycles, as well as land-atmosphere interaction. I suggest the author reorganize this paragraph to better introduce the existing studies and background.
3. Considering that the multi-tiling scheme has been used in other land surface models, please clarify the research gaps apart from the specific model used in the study.
4. In section 2, Figure 1: It is unclear how many soil/snow columns are included for each grid cell for the improved and original versions. Whether do different PFTs have different snow cover and soil characteristics or not? These (especially the snow cover) may have big impacts on surface energy balance. Besides, the authors used the standard rectangle grids to represent different PFTs, which may mislead the authors, because the same PFTs may distribute in different sub-grids.
5. In equation 2, the emissivity is assumed to 1, which is not reasonable. For vegetation, the emissivity depends on LAI. Different land types also show very different emissivity.
6. In equation 2, I am also confused about how did the authors define the surface temperature here, because the surface temperature can change within the vegetation canopy and understory background. Please clarify how the model calculated surface temperature here as well as the relationship between surface temperature, canopy temperature and ground temperature.
7. Line 117: How did the authors retrieve the soil albedo for the vegetated regions from remote sensing data?
8. Section 2.3: How did the authors set the empirical values for different snow-related parameters?
9. Section 2.1: It is unclear how the model calculates the surface albedo.
10. 11: what is the reference of this equation? Are the impacts of topography on snow cover fraction considered in this equation?
11. Section 3.3: There are already high resolution (e.g., 250m) SOC datasets at the global scale, e.g., soilgridv2.
12. Section 4: I am curious about why did the authors run the simulations cycling the forcing from 1901-1920 rather than the present-day simulations? The present-day simulations can be more suitable for the comparisons with the available remote sensing data.
13. Figure 4-5 and 7-8: Please show whether the differences are significant or not for each grid cell. Table 2: Please also clearly define the magnitude of one arrow and two arrows. The statistical tests on the case differences are also needed in section 5.2 for all variables.
14. Section 6.1: When comparing with MODIS data, did the authors extract the model values in the MODIS overpass time? Besides, also show whether the differences between model estimates and MODIS are significant or not. Did the improved version show better performance than the original version?
15. The authors just compare the surface temperature. I suggest the authors compare all the surface energy balance components with the available remote sensing data or reanalysis data, e.g., surface albedo, net radiation, etc. Comparing all of them can give more hints on the model improvement and potential drawbacks.

Minor concerns:

1. Line 53: In some ESMs, e.g., E3SM, the topography-based tiling scheme has also been used.
2. Line 100: Modify the captions of Figure 1.
3. Line 264: what is the meaning of “offline” here?
4. Line 290: why did the author select the three grids?
5. Line 306: Looks like such calculation is not accurate because of neglecting the light interaction between canopy and soil.