Schmitt et al. first compare climatological data from the ATTO experimental site to two gridded climatological datasets. They identify and discuss biases of the datasets in relation to the experimental site. Then, they adjust one of the datasets and use it to force the land surface model JSBACH. Finally, they highlight the differences that arise (mainly in the soil ) when using the model with standard gridded forcing data compared to the adjusted forcing data.

I like the approach of the authors by first comparing and then adjusting the forcing datasets before using them in a vegetation model. I also think that comparing and testing multiple forcing datasets is great and should be the standard for any modelling study in locations where the meteorological data is quite uncertain. å

However, I have some comments mainly regarding the structure, flow and presentation of results and discussion that I think should be addressed.

Major points

I found it difficult to follow the results and discussion. While it is well written from a language perspective it was hard to understand which part of the text belonged to the result and the discussion. The authors refer to point 3 as results and discussion, but I am unsure why the authors did go this way. Associated with the alternating part of results and discussion there is also a lot of jumping around with the figures. For example, the first figure mentioned in the results is Figure 3c. What about 3a and b? Figure 12 is only referred to in the text after Figure 13. I can see that the authors wanted to address each climate variable consecutively, but when the variables are spread across figures it is quite hard to follow.

The authors find that soil temperatures and soil water content are the key differences that arise when using different forcing datasets to drive JSBACH at the ATTO site. I am wondering about all the other variables and states of JSBACH, such as GPP, NPP, stomatal conductance and carbon pools. How are they affected by the different forcings? I understand that not all model output variables should be addressed in this study, but I think it would be very interesting for the modelling community to see if some of them are also affected by the choice of forcing dataset.

The authors highlight and describe the key important results well. However, at some points, I am missing that these results should be discussed in greater detail and I am especially missing some  I will outline these parts at the minor points.

Minor points

Lines 3-5: Introducing the abbreviations JSBACH, ERA5 and MERRA-2. Given the length of the abstract, I think it would also be nice to explain the abbreviations more. For example ERA5 reanalysis data? I would add one more sentence about what ERA and MERRA are.

Lines 13: Great!

Line 18: How would a separate canopy layer improve that? The way it is written that it appears to be quite speculative/=.

Line 32-35: I think this part should be at the end of the discussion.

Line 44: Is JSBACH a land model or a land surface model? Choose one and use it consistently.

Line 74: there are two dots at the end of the sentence.

Line 74-75: Something is missing in this sentence.

Lines 75-76: There is no mention of the discussion in this paragraph.

Line 82: I would suggest using the word ‘aspects’ instead of ‘features’.

Figure 1: While I like this figure I think it is needed to understand the key points of the manuscript and would suggest moving it to the appendix.

Table 1: I appreciate the effort of the authors putting together this list of measuring devices, but similar to Figure 1 I think it can be moved into the appendix.

Lines 102 (equation 1): These two equations are standard in atmospheric sciences and I would also suggest moving them and the explanation to the appendix.

Line 122 Section 2.2: Why did the authors choose ERA5 and MERRA-2? Why not any of the other datasets? There should be 1-2 sentences about why such datasets are preferred over other available ones.

Lines 165-169: This part is rather a model setup than a model description and I suggest moving it to the end of this subsection.

Line 174: What is a T63 grid?

Line 199: What does ‘almost perfectly agree’ mean? Such statements should also be supported by some statistics like RMSE.

Line 206: See major points. Why start by explaining Fig. 3c?

Line 230: The authors describe the bias in windspeed that they find. I am missing the implications for LSMs if the windspeed is much lower or higher. What should that do in theory to LH and SH and canopy temperatures? Or is that hard to say at all?

Line 235: This statement about generality needs a reference.

Line 252: Can you test if that is true or not?

Figure 5: What kind of measurement uncertainty is used? How can it be so small given we are looking at rainrates over 5 years?

Line 258: I think the specific humidity (Fig. 3b) deserves more attention. Why are both models going down while the observations are going up in August-October?

Line 270: ATTO does not have that radiosonde instrument, right? I would add why we can’t do that in that study.

Line 274-277: This describes what is happening in the models used for reanalysis data, right? Why does this not apply to the ‘reality’ or the ATTO tower?

Line 293: Typo: biases instead of biased

Line 298: How has the data been optimized? Should be described in the method section.

Line 300: Section 3.2.1: Are there any other impacts on e.g. GPP or NPP or transpiration or stomatal conductance in general?

Line 310: Why is that expected?

Figure 7: The description of this figure is confusing. I recommend adding ERA5 and MERRA to the legends. I also recommend explaining each subfigure in alphabetical order.

Figure 9a) Missing legend for JSB.

Line 443: Figure 13a,b does not support that statement before!

Line 495: I like the summary of the results. However, here I am missing the implications for the vegetation. Why should we care about these biases? If there are almost no roots in the deepest soil layers, is a 20% difference important? Again: are there no other biases for other variables?

Review of:

Title: Modeling atmosphere-land interactions at a rainforest site – a case study using Amazon Tall Tower Observatory (ATTO) measurements and reanalyis data

Author(s): Amelie U. Schmitt et al.

MS No.: egusphere-2023-679

MS type: Research article

The manuscript addresses the very relevant process of the land surface-atmosphere interactions in the presence of a (high) forest. In many atmospheric models for numerical weather prediction or climate simulations simplifying or conceptional approaches are used to describe this issue. It needs to be understood for which applications this is good enough, and where it needs more complex (multi-layer) canopy models. A novelty of this manuscript is to use tower measurements in the Amazon rain forest in order to very specifically evaluate the land surface model JSBACH in this respect.

Major remarks:

- Sect. 2.3.2: Do I understand it correctly that JSBACH is run in offline mode at one (grid) point, using an atmospheric forcing from observation or reanalyses? Then, why is a spin-up of ten years needed? For instance, Chen et al. (1997) have shown that in PILPS Phase 2a the vertical profiles of soil temperature and water content converged to an eqilibrium after two to three years. This was in Cabauw (The Netherlands), not in the rainforest. Anyway, since there is an extensive rain period, you could start the simulation there, initialize the soil water content at saturation, and run the model into equlibrium. You should show that this is not already reached after two or three years, but that you really need ten years. This would make more sense in order to understand the model behaviour of JSBACH. As you say, the soil is not even particularly deep. Maybe, with a shorter spin-up, you find periods of the ATTO measurements with less data gaps?

- Sect. 3.3.1: The wind looks difficult. Likely, there is a problem with the representativity, since the reanalyses can not describe the high forest canopy, but in reality it exists. It is not clear to me how a bias correction should work here. Maybe, it is mainly a matter of finding the correct heights above ground (or canopy top) to make the quantities comparable? Could you discuss this a bit more?

- L. 472-474: Instead of "temperature damping effect" I would call it rather "shading effect". Without a vegetation canopy the model has no chance to get any of the following temperatures right: Soil temperature (different depths), surface temperature, and 2-m temperature. Instead of a complex canopy scheme, a simpler way to represent this mechanism is e.g. a conceptional "skin temperature" scheme, see e.g. Viterbo and Beljaars (1995), Heidkamp et al. (2018), or Schulz and Vogel (2020). The work of Heidkamp et al. (2018) is available in JSBACH. It may be advisable to apply it in your study, in order to represent the shading effect due to the vegetation. This would reduce the amplitudes of the diurnal cycles of the soil temperatures, and increase the amplitude of the surface and 2-m temperature. It would be good if you could demonstrate this in your manuscript, because the observations you have available.

L. 475-476: The evaporation of water from the interception reservoir is usually less relevant for the simulated soil temperature. It may play a role after dew fall in the morning (or after rain fall) for the 2-m temperature. Please rephrase.

Minor remarks:

- L. 49: ... errors in the ...

- L. 54: ... based on two ...

- L. 58: ... (Yang et al. 1995), ...

- L. 66: ... of the forcing data ...

- L. 71: ... turbulent heat fluxes ...

- L. 72: shortcomings

- L. 74: Section numbers are missing

- L. 90: less or equal, or larger or equal 36 m?

- Fig. 9: (a) and (b) are mixed

- L. 398: "field" capacity. Anyway, field capacity is not saturation, this would be pore volume. Please rephrase.

- L. 404: soil types -> soil textures

- L. 407: an exponential root profile would be typical

- L. 468 and around: soil types -> soil textures