Review of  “Angular Sampling of a Monochromatic, Wide-Field-of-View Camera to Augment Next-Generation Earth Radiation Budget Satellite Observations” by Jake J. Gristey et al.

This paper provides an initial description of the methodology that will be used to generate ADMs for the split-SW radiometer onboard the future Libera mission. The paper is very well written, with the motivation, concept and details well explained, and the presentation and quality of the figures is excellent. I don't have any significant comments, I believe this paper is ready to be published as it is.

I've only spotted one typo on L310: CIRES- > CERES.

The paper describes an alternative approach to rapidly derive empirical ADMs for the broadband VIS channel planned for the Libera mission.  These ADMs are fundamental to the derivation of the primary user products, the irradiances, from radiance observations.  The proposed approach provides a clear advantage in terms of the speed of building up the ADMs during the mission and avoids the need for the broad band instrument to spend a large amount of time in the non-standard scanning modes that are required to provide the ADM observations.  The subject matter is clearly within the remit of the journal and of interest to future users of the products as well as to the flux retrieval community.

I find the overall approach interesting, and the material presented relatively convincing but have some queries about the details of the application that I outline below. I have separated these by the section of the paper that they are relevant to.

Section 1

Is the conversion to broadband that is proposed (fig 1) required for some other purpose such as scene identification and is the accuracy of this step important, it seems superfluous in the ADM generation given the later averaging and normalisation that occurs for each scene? 

How is scene identification of the monochromatic observations performed and how will it be matched to the VIS scene classification, is this likely to be a significant error contribution?

Section 3

It is not entirely clear to me if the intention of the ADM sampling with the camera pixels is to match the nadir footprint of the Libera radiometer at all angles or to match the view angle variation in footprint size that would occur if the radiometer itself was obtaining the samples.

Section 4

I think I need a bit more convincing that the correlation across scenes optimizes within scene angular fidelity and more information to assess the likely accuracy and the derived ADMs and hence the flux determined using them.

As the ADMs only need to provide information on the angular distribution of energy within each scene class and solar zenith angle, it is not obvious to me that correlation between the monochromatic and broadband radiances across these variables is important.  Rather ADM accuracy would only seem to require good correspondence across angles within each scene and solar zenith class as the ADM data is averaged and normalised on a scene by scene basis.  I wonder if trying to have good correlation across these additional variables to ensure you can perform a monochromatic to broadband VIS (or NIR) conversion with high fidelity is overly restrictive.  In feel that the expanded dynamic range that results from combing scenes and/or solar zenith angles would likely dominate the correlation and hinder the search for the optimum wavelength for ADM derivation in favour of the optimum wavelength for overall broadband fidelity.  For the restricted angular coverage data primarily used in this section (the nadir OSSE data and the near nadir SCIAMACHY observations and simulations) this across scene and solar zenith variation becomes the only variable of interest.  Whilst some across scene variation effects would mimic angular variations others such as spectral variation in surface reflection are I think not so relevant. Thus, I think using the across scene correlation at nadir and near nadir wavelength as the only basis on which to justify the optimum wavelength for capturing the broadband angular variation within scene classes needs more support.  The CERES unfiltering database scene specific correllations shown in figure 8 would need to be further statrified by solar zenith to provide this support I think. As currenlty presented the variation in incoming shortwave with solar zenith which isn’t relevant for the ADMs  becomes a factor dominating the correlations shown.

It is also not obvious to me how the relationships shown in section 4 can be easily translated into an error in the ADM and hence the derived flux.  Some exploration of this would be helpful and may identify some scenes or angles that are particularly problematic. Figure 8 (d) comes the closest to addressing this but as it shows reflectance and only explores a single viewing zenith can’t be understood in terms of the likely flux error. What I would like to understand is the difference in the inferred anisotropy between a broadband and monochromatic derived ADM.

Section 5

I’m not sure that the results in this section really answer the question of how long will be needed to build the required ADMs. Angular coverage that isn’t stratified by solar zenith angle (figures 10 and 11) whilst likely a minimum requirement doesn’t provide reassurance that the ADMs will be filled at the solar zenith angles of the scenes observed by the VIS channel.  In fact when the effect of of the sun synchronous orbit is considered on the pixel array angular sampling the mis-match between solar zenith for particular angular bins would seem inevitable.  Perhaps I misunderstand but the reassurance that these figures give about the ability build up the ADMs is unclear to me.  The solar zenith angle stratified results shown in figure 12 and that could be derived for other scenes, including those stratified by optical depth, also don't directly translate into incomplete ADMs for scenes that are actually required, as the orbital characteristics and illumination conditions for any particular time of year will restrict the solar zenith angle - scene combinations observed by the VIS channel.  I feel that to understand how quickly the camera can build the required ADMs, the angular coverage needs to be stratified by solar zenith and scene (including cloud optical depth if this will be used for the ADMs ), but the resulting coverage needs to be considered in the context of the actual ADMs required at that point in the record.  So, the question of how long it will take in practice to build the required ADMs with this technique is a question about how quickly good angular coverage can be obtained for the scenes viewed by the broad band instrument (in the centre of the pixel mask I assume), by the camera pixel array.  I expect this can be answered with the data used here, and I wonder if this analysis would change the scenes considered difficult to acquire, the length of time needed or point to a need for a few days every month or season of monochromatic observations to keep up with the evolving make up of the observed scenes.  Given the sun synchronous orbit and the effect this will have on the solar zenith sampling of the different camera pixels I also wonder if this analysis might impact the choice of optimum subsampling of the camera pixels.  For the case of the random pixel mask, I also wonder if this study shows if random coverage of the ADM angular bins is achieved once the scene variation and orbit effects are superimposed.