This paper describe the Determining Instrument Specifications and Analysing Methods for Atmospheric Retrieval (DISAMAR) computer code, which performs both forward radiative transfer model and inversion simulations for the Earth’s atmosphere with different components, for instance, trace gases, aerosols, and clouds, and properties of the ground surface from passive remote sensing observations of the Earth. The main novelty is that DISAMAR flexibly uses a variety of radiative transfer methods when solving multiple scattering of light in the atmosphere, including the layer-based orders of scattering method, adding and doubling, integration of source function, etc. For weakly gas absorbing spectral region, the DISMAS (DIfferential and SMooth Absorption Separated) method is developed to significantly expedite the simulation time while keeping the accuracy. Jacobian (differential of radiance field with respect to retrieval parameters) can be calculated semi-analytically, which is a great advantage in comparison with the finite difference method. The manuscript is clearly written and organization is logical. I suggest the publication of this paper at Geoscientific Model Development (GMD) with some minor revisions. Specifically:

1. Figure 1 and 2 are referenced out of order. I suggest the authors revise the manuscript to reference them in sequence. Minimally, they could simply rename Fig. 1 as Fig. 2 and vice versa.
2. Figure 2, the selection of wavelength grids is quite vague to me, especially when absorbing lines are involved in the channel. I strongly recommend the author revise the description the wavelength selection scheme.
3. Figure 3, I thought it would be more natural to use optical depth as a vertical coordinate in radiative transfer. Thus to me using pressure as the vertical coordinates is a bit unusual. This is just a comment and I won’t force the authors to make any changes, as this would be most likely an overhaul of the computer code.
4. Line 150-151, the paper discussed different Gaussian quadrature points for different optical depth situation. Again it would much natural to use optical depth as the vertical grid, so that you would easily built a universal criterion of how many discrete layers are needed in terms of optical depth. By the way, how large is the optical depth considered as “thick”?
5. Line 408, please give a list of “strong” absorbers and their associated wavelength ranges to which DISMAS should not be applied. How strong of a gas absorption line is considered strong?
6. For the spherical shell correction, there are some new developments recently. Specifically:
7. Korkin, E.-S. Yang, R. Spurr, C. Emde, P. Zhai, N. Krotkov, A. Vasilkov, A. Lyapustin,

Numerical results for polarized light scattering in a spherical atmosphere, Journal of Quantitative Spectroscopy and Radiative Transfer, Volume 287, 2022, 108194, ISSN 0022-4073, https://doi.org/10.1016/j.jqsrt.2022.108194.

(https://www.sciencedirect.com/science/article/pii/S0022407322001297)

Peng-Wang Zhai, Yongxiang Hu, An improved pseudo spherical shell algorithm for vector radiative transfer, Journal of Quantitative Spectroscopy and Radiative Transfer,

Volume 282, 2022, 108132, ISSN 0022-4073, https://doi.org/10.1016/j.jqsrt.2022.108132.

(https://www.sciencedirect.com/science/article/pii/S0022407322000693)

7. For Raman scattering and other inelastic scattering in the ocean waters, there are some new development as well:

Peng-Wang Zhai, Yongxiang Hu, David M. Winker, Bryan A. Franz, and Emmanuel Boss, "Contribution of Raman scattering to polarized radiation field in ocean waters," Opt. Express 23, 23582-23596 (2015)

Peng-Wang Zhai, Yongxiang Hu, David M. Winker, Bryan A. Franz, Jeremy Werdell, and Emmanuel Boss, "Vector radiative transfer model for coupled atmosphere and ocean systems including inelastic sources in ocean waters," Opt. Express 25, A223-A239 (2017)

A forward modelling is essential for understanding the physics behind radiation measurements, it is also the base of building accurate algorithm for remote sensing, DISAMAR is of a such important tool for retrieving atmospheric gases and particles, and it has been used for TROPOMI and SP4/SP5 observations. This model introduces several numerical techniques such as layer-based SOS method, semi-analytical derivatives etc. to improve its computation efficiency as well as its computational accuracy. I agree its publication in GMD after minor revision, and I’m looking forward its application for more space-born instruments in future.

1. Around line 280, it is not clearly to see how can we get EQ. 25 from EQ 24, please add more description.
2. For the convolution of the ISRF around Line 109. “If there is a strong absorption line in the interval, the interval is reduced until the position of the strong line is reached, so this interval is smaller than the FWHM”, this is easy to do for O2-A band which is regular spaced, for irregular spaced absorption band, say, water vapor band, is this method easy to apply this algorithm?
3. Line 114,”Typically the number of Gaussian division points is between 3 and 30 per wavelength interval in the O2 A-band for a FWHM of 0.5 nm.” Is there a rule for the reader to know how to choose the number between 3 and 30.
4. “only the adding of different layers and the subsequent calculation of the internal field is replaced by the successive orders of scattering method.” Please add several sentences to make clear how to calculate the internal field?

Journal: Geophysical Model Development

Year: 2022

Title: Introduction of the DISAMAR radiative transfer model: Determining Instrument Specifications and Analysing Methods for Atmospheric Retrieval (version 4.1.5)

Comments:

In the manuscript, the authors describe the so-called DISAMAR one-dimensional radiative transfer model.

The model is described by the authors as a polarized all-sky radiative transfer model, i.e. it is suitable for purely absorbing clear-sky atmospheres and atmospheres with scattering clouds.

The application focus of said model are satellite radiance retrievals, in particular for the TROPOMI instrument on board the european Sentinel-5p satellite. 

Nevertheless, the authors list a suite of different available solvers and a range of additional model features that are not required for retrievals, such as irradiance computations.

It is emphasized by the authors that the primary advantage of their model is the seamless combination of all necessary features for satellite remote sensing in their model. 

Comments on the Introduction:

- line 44: The assumption that the atmospheric input profile of the radiative transfer model is hydrostatic is of some importance. How does this approximation impact the DISAMAR retrieval results?

- line 47: Is a Lambertian reflectance the only surface reflectance type available? Does this limit the accuracy of the model over ocean surfaces where the Cox-Munk model is typically applied?

- It would be advantageous to provide a list of other relevant radiative transfer models with similar purpose and complexity in comparison to DISAMAR.

Examples include the CRTM [1-3] and RTTOV [4].

Comments on Section 2:

- line 65: Please provide a short explanation on the purpose of the wavelength grid.

- lines 71 and 72: Please provide references for the application of the derivatives for optimal estimation and the application to the error covariance matrix, gain vectors, and averaging kernel.

- line 73: The formal theory of evaluating the derivatives of a computer program is quite well developed [5]. 

Could you please elaborate whether your semi-analytical approach computes the forward-mode (tangent-linear) or reverse-mode (adjoint) derivative of your code output/ the radiance spectrum?

Comments on Section 3:

- line 92: Are there any restrictions when using a tabulated ISRF? It is stated in Section 2 that the radiance wavelength grid is given on a set of Gaussian quadrature points. Are the tabulated values automatically interpolated onto the grid?

- line 106: How does the line-by-line absorption model impact the calculation time of DISAMAR?

Does DISAMAR include faster absorption models when calculation time is a constraint?

- line 140: Please explain how the pressure levels are translated into altitude levels. Are you using the hypsometric equation?

- line 185: There are different adding-doubling initialization schemes [6] and this is known as the infinitesimal generator initialization.

- I have not checked equations (19) to (26) for correctness.

- line 335: If the Layer-Based Orders of Scattering method is fast for optically thin clouds, wouldn't it provide some advantages to initialize the Adding-Doubling solver with a LABOS solution, since it spends a lot of computation time doubling a small initial layer?

- Is the LABOS method related to the Successive Order of Scattering approximation? If so, what are the characteristic differences?

References:

[1] C. H. Lu, Q. Liu, S. Wei, B. T. Johnson, C. Dang et al. (2021): The Aerosol Module in the Community Radiative Transfer Model (v2.2 and v2.3): accounting for aerosol transmittance effects on the radiance observation operator. Geosci. Model Dev., 15, 1317–1329.

[2] B. M. Karpowicz, P. G. Stegmann, B. T. Johnson, H. W. Christopherson et al. (2022): pyCRTM: At python interface for the community radiative transfer model. J. Quant. Spec. Rad. Trans. 288.

[3] P. G. Stegmann, B. T. Johnson, I. Moradi, B. Karpowicz, W. McCarty (2022): A deep learning approach to fast radiative transfer. J. Quant. Spec. Rad. Trans. 280.

[4] R. Saunders, J. Hocking, E. Turner, P. Rayer, D. Rundle, P. Brunel, et al. (2018): An update on the RTTOV fast radiative transfer model (currently at version 12). Geosci. Model Dev., 11, 2717–2737

[5] A. Griewank, A. Walther: Evaluating Derivatives, Principles and Techniques of Algorithmic Differentiation. Society for Industrial and Applied Mathematics; 2nd edition (November 6, 2008)

[6] Wiscombe, W. J., 1976. On initialization, error and flux conservation in the doubling method. J. Quant. Spectrosc. Radiat. Transfer 16, 637-658.