This paper presents enough materials to describe the upgraded key features from  uDALES v1.0 to uDALES v2.0 (LES), as well as the crucial technique details.  Four cases were tested and validated against the data in the literature,  showing the success of the upgrading.  I am very pleased to see that the authors have published the uDALES v2.0 codebase and user manual for open access.   

The MS is concise and is well prepared.  I would recommend ‘accept’ for publication, subject to addressing some minor comments below:

1. Lns 8-9. Please rephrase this sentence ‘Good agreement …’
2. 2, in the convection term u_i to change to u_j. Give a bit more details about K_h (it should be SGS eddy diffusivity).
3. Ln 112, change ‘transported’ to ‘convected’.
4. 5, define RH, q_{sat}
5. Lns 142-144. “Note that the surface energy balance is generally evolved using a larger timestep than the LES, therefore these fluxes are time averaged”. This is not clear enough. These fluxes are constant during the LES, or vary slowly in a quasi-steady manner?
6. Section 3 title ‘Improvements’ to change to ‘Upgraded key features from uDALES v1.0 to v2.0
7. 3.2. As in the inlet-outlet direction (streamwise), only the Neumann BC can be used.  Therefore, the constant pressure boundary condition cannot be used.  Given all boundary conditions for pressure are Neumann, any specific treatment is given to settle down a unique solution?
8. Ln 228. Non-slip boundary condition means non-zero shear, which is wall-normal momentum flux.  Please comment.
9. Lns 245-246. Please describe a bit more of the interpolation here, is it a 2^nd-order accuracy linear interpolation?
10. Lns 335-338. The ‘real’ turbulence generated by the driver simulations using periodic lateral boundary conditions is governed by the upper and lower boundaries, which are just a simplified condition (but not 100% ‘real’). Please comment.
11. Ln 374. ‘These differences are most likely caused by differences in the inflow conditions above z/hm = 5 (Fig. 8c)’. The upper boundary conditions in the LES might be one more reason to cause the discrepancy?  Please comment.
12. Section 4.2. Please give the dimensions of the LES domain in the text, although they are shown in Fig. 9.
13. Section 4.3. The Ri number is kept the same as in Richards et al (2006) and Boppana et al (2013), whereas the Reynolds number is increased by 100 times (or kept the same as Richards et al (2006))? If not, please comment on this.
14. Section 4.4. Please give the Ri number in the current problem, with a discussion on the capability of uDALES for larger Ri flows. 

I found the presentation to be clear, and the topic is definitely appropriate.  I am not qualified to review the technical aspects of the manuscript.

see attached report

Review of " A conservative immersed boundary method for the multi-physics urban large-eddy simulation model uDALES v2.0 "

This study shows result from uDALES v2.0. model which is an upgraded version uDALES v1.0, and compared results with the previous version of uDALES v2.0. and existing literatures. The upgraded version (i.e., uDALES v2.0.) results are consistent with the previous version (i.e., uDALES v1.0) and literatures, and have flexibility to study larger urban areas and run longer simulations. I found this manuscript well written and explain, although, I am not much expert on the technical aspects of the manuscript. I believe this manuscript is suitable for publication in GMD but I have few minor comments that I think would improve the manuscript:

I noticed in the tile that author refers to uDALES v2.0. as a model, whereas in the abstract and rest of the manuscript, author refers to uDALES v2.0 as a framework, which is a bit confusing to me. Therefore, I would suggest stick with one of them to be consistent. 

In line 13, it is not clear to me what trend author is referring there.

Line 36-39, I would suggest splitting the sentence, as it is a very long one to follow what information author wants to convey.

General Comments

In the present study , the authors have taken advantage of the existing  knowledge to produce a new version (v2.0)  of the uDALES code, demonstrating in addition the positive impact of the changes on prediction capabilities.

The effort here is to present the improvements on uDALES 1,0 to become uDALES 2.0. Therefore common descriptions are better to be limited and give more emphasis in the differences.

It is assumed that each application selected serves a specific objective. It is good this specific objective to be described and at the end of the specific exercise, conclusions need to be drawn in relation tothis objective. A generic objective is to compare results of  uDALES v2.0 and uDALES v1.0 and commenting on the differences and their causes. Please make sure that those principles are kept to a large degree, for all applications presented. 

It is most probable that some limitations still persist in the v2.0 version. It would be helpful,  in a separate chapter, such limitations are discussed including the way forward.

Specific Comments

Lines 41- 43 “RANS only resolves the mean flow, and entirely relies on turbulence modelling to incorporate the effect of fluctuating components. As a result, RANS often fails to accurately reproduce transient flow features and quantities such as turbulent kinetic energy (van Hooff et al., 2017).” . It is not evident that the RANS characteristics prevent them to reproduce transient flow features. Do you mean unsteady flow features ? . Concerning turbulent kinetic energy predictions with RANS Models probab;y there are models that claim they can do this. See for example : Bartzis, J,G , Boundary-Layer Meteorology (2005) 116: 445–459 ,DOI 10.1007/s10546-004-7404-y

Lines 45 -46 :  “… LES is inherently more accurate than RANS….” . It would be better to say  “… LES is expected to be more accurate than RANS”.   We should not forget that we have to apply LES correctly and this not always easy for open flows such as the atmospheric boundary layer.

Lines 105-107 : Concerning horizontal grid, it seems that there is a different treatment for x- and y -directions. In real problems this might impose limited degree of freedom in selecting x- and y- directions. In other words is this selection   application dependent ?

Lines 110 -113 ; The inlet flow conditions are briefly discussed. The key problem here is how the upwind large eddies are introduced into  the computation domain, keeping in mind that any introduced error persists downwind.  Please expand.