The paper presents an interesting approach of coupling hydrologic and hydrodynamic models to improve computational efficiency while maintaining numerical accuracy. However, to demonstrate the superiority of the proposed approach, it is essential to conduct a thorough comparison with state-of-the-art individual hydrology and hydrodynamic models. This will help highlight the advantages and necessity of the coupled modelling approach. It's crucial to show how the proposed method outperforms existing models in terms of both efficiency and accuracy. 

1, In the introduction, the authors should focus more on recent progress in coupled hydrology-hydrodynamic models, especially with respect to their proposed coupling method, which seems different from the common coupling methods. Additionally, a detailed explanation of the non-uniform grid generation should be provided to give readers a better understanding of its significance in the proposed approach.

2, The modelling performance is highly influenced by the underlying mesh generation. Even with advanced adaptive methods using meter-scale data, there can be uncertainties impacting the modelling performance. It is recommended that the authors perform an uncertainty analysis on the mesh generation process to understand its potential effects on the model results.

3, The paper mentions the use of a coarse grid. What is the resolution of a coarse grid? What is the ratio between find grid resolution and coarse grid resolution? Will the coarse grid resolution/ratio have a large impact on modelling performance? Understanding the impact of this coarse grid resolution/ratio on the modelling performance is crucial.

4, Providing a detailed description of the hydrology and hydrodynamic components, especially on their treatment of flow variables (e.g., discharge, depth), would greatly enhance readers' understanding of the coupling process at the interface. This information is vital to evaluate the robustness of the proposed coupling approach.

5, The paper uses small test cases to evaluate the modelling efficiency. However, it is important to validate the model's performance on larger scales, to ensure its practical applicability. Even the hydrodynamic models working on ~10m-30m can be configured for model run covering an area of several hundred kilometres and quite efficiently.

6, The choice of using Fortran for coupling the two modelling components, while the SWMM model is written in C++, raises questions about the rationale behind this decision. The authors should provide a clear explanation for this choice, considering factors like compatibility, performance, and ease of implementation.

The authors present an improvement to the multi-grid hydrological/hydrodynamic SWMM/IM-DBCM model which partitions the model domain into a coarse resolution part (away from rivers) and a fine resolution part (areas susceptible to inundation). The authors present the mesh-generation approach and test the model in 4 configurations with a variable number of grid cells (ranging between 59k-207k cells). Results of discharge are compared to observed values, inundation depth is presented without comparison to observations. Advances are said to stem from improved computational efficiency, the main reason for the multi-grid approach of the model, while retaining an acceptable model performance. I provide two

1) The state-of-the-art approach to the problem of variable resolution grids is a coarse-resolution hydrological model coupled with a 1D river routing model that activates a 2D model when channel capacity is exceeded. While the authors acknowledge this in the manuscript, they fail to compare their approach to results of such a model chain to demonstrate their advance. Admittedly, producing the same results with such a model chain to use as a baseline is a non-trivial task, but some comparison if not in the same catchment, should be considered mandatory.

2) The authors present model runtimes of the four grid configurations in Fig 13. It is apparent that runtimes scale linearly when comparing the uniform grid case to the variable-grid case (cf case00 to case12). This relationship does not hold, however, in coarser configurations of the multi-grid models (cases 15/10), i.e. the runtime/grid is significantly higher. E.g. going from 105k cells to 59k cells only brings moderate efficiency savings of 10-15%. This shows the limits of the approach, presumably because more time is spent on coupling the coarse and fine grids.

-l.74 -85 unclear, needs proofreading

- l.117 Ghost cells need to be defined before

- l.498 and l.501 "last moment" > in the last time step?

In conclusion, I feel the merit of this paper is confined to the user group of the presented model chain. In its current form, the improvements present limited advantages to other models. I thus leave it up to the editor to recommend major revisions or reject the manuscript in light of these shortcomings.