Dear authors,
thank you for submitting your response to reviewers' comments, and for the revised manuscript. I communicated with both reviewers, and they feel that you adequately addressed their concerns. Meanwhile, a third reviewer sent a few further comments, which I paste below. For the final manuscript version, please take these into account to the extent possible at this stage.

With best regards,
Sebastian Schmidt

Comments by reviewer #3:

The study by Knobelspiesse and Nag addresses an interesting problem that can have important implication for the design of future aerosol observing missions. Specifically, the authors use an information content analysis to understand and quantify the value of different multi satellite configurations in terms of surface and aerosol properties retrieval. It is concluded that several single view instruments in constellation can actually perform equally well in comparison to a single instrument with multiangle capabilities and also that the distribution of observing geometries across the BRDF and BPDF does not seem to impact the conclusion, which the authors attribute to the smooth nature of TOA reflectance for the selected surfaces and aerosols.

The paper is well written, clearly organised and complete for comprehension. The conclusions are correctly supported by the material presented in the paper, but the context which they apply needs further discussion. In this regard, I have three main comments or concerns that the authors might want to address or discuss so that their conclusions are not misinterpreted in a more general context.

The first important assumption that is made concerns the configuration of satellite orbits which have been directly extracted from a previous study, in which observing configurations had been optimised for land surface BRDF retrievals. Although I would certainly agree that surface properties are an important and central assumption for aerosol remote sensing, it is not directly obvious that a configuration that has been optimised for surface BRDF retrieval will provide the maximum information content for aerosol retrievals. Could the authors comment on this? In particular do you have elements to support the fact that an optimisation of satellite configuration targeting aerosols retrieval would not result in a different selection of orbits? It is not necessarily clear that a set of geometries optimising BRDF carachterisation would work equally well for aerosols.

A second point concerns the limited selection of aerosol models used for the study. Given that multi-angle and polarisation information have been demonstrated to be particularly useful to discriminate spherical vs non spherical particles, I would have expected that at least one non-spherical aerosol model had been included in the analysis. Similarly, an additional aerosol model with higher absorption would help in consolidating the representativeness of the study. I wonder whether the conclusion that the distribution of observing geometries doesn’t seem to matter much, is related at all to the limited choice of aerosol models?

A third and last comment is more central to the application of the information content and concern the assumption made for radiometric noise of the various sensors. Given the multiple instruments implementation, wouldn’t it make sense to assume a higher noise for the instruments flying in constellation? The advantage in terms of resilience using multiple small satellite could be easily offset by difficulties in achieving the required accuracy from a set of different instruments under different conditions. Have the authors tried to perform the information content analysis for the groups of multiple satellite using a higher radiometric noise (ie 20% or 50% larger)?

Although these are fairly important questions in my opinion, they do not necessarily have to be fully addressed within this paper. However, the three points at least need some level of discussion before the paper could be published.

After completion of the manuscript's peer review process, a third reviewer submitted additional comments. The authors responded to those, but no changes to the manuscript were necessary. Since the exchange is relevant for this topic, it is published here alongside the other pertinent documents.

"AMT-2017-473
Response to reviewer 0000003
The study by Knobelspiesse and Nag addresses an interesting problem that can have important implication for the design of future aerosol observing missions. Specifically, the authors use an information content analysis to understand and quantify the value of different multi satellite configurations in terms of surface and aerosol properties retrieval. It is concluded that several single view instruments in constellation can actually perform equally well in comparison to a single instrument with multiangle capabilities and also that the distribution of observing geometries across the BRDF and BPDF does not seem to impact the conclusion, which the authors attribute to the smooth nature of TOA reflectance for the selected surfaces and aerosols.

The paper is well written, clearly organised and complete for comprehension. The conclusion are correctly supported by the material presented in the paper but the context in which they apply need further discussion. In this regards, I have three main comments or concerns that the authors might want to address or discuss so that their conclusions are not misinterpreted in a more general context.

We thank the reviewer for addressing our manuscript.

The first important assumption that is made concerns the configuration of satellites orbits which have been directly extracted from a previous study in which observing configurations had been optimised for land surface BRDF retrievals. Although I would certainly agree that surface properties are an important and central assumption for aerosol remote sensing, it is not directly obvious that a configuration that has been optimised for surface BRDF retrieval will provide the maximum information content for aerosol retrievals.

Could the authors comment on this ? In particular do you have elements to support the fact that an optimisation of satellite configuration targeting aerosols retrieval would not result in a different selection of orbits ? It is not necessarily clear that a set of geometries optimising BRDF caracterisation would work equally well for aerosols.

A key aspect of this work, and why we thought we could use the satellite configuration optimized for surface BRDFs, is that the surface BRDF largely defines the Top of Atmosphere observed BRDF (including aerosol effects). The impact of aerosols on the TOA BRDF is subtle, and does not impose artifacts with strong, angularly resolved signals. Thus, we can confidently expect that a satellite configuration optimized for surface BRDF retrievals would be the same as one optimized for aerosol retrievals, as surface BRDF is more variable in an angular sense.
A second point concerns the limited selection of aerosol models used for the study. Given that multiangle and polarisation information have been demonstrated to be particularly useful to discriminate spherical vs non spherical particles, I would have expected that at least one non-spherical aerosol model had been included in the analysis. Similarly, an additional aerosol model with higher absorption would help in consolidating the representativeness of the study. I wonder whether the conclusion that the distribution of observing geometries doesn't seem to matter much, is related at all with the limited choice of aerosol models ?
Multi-angle (and polarimetric) observations are indeed valuable for the retrieval of nonspherical and absorbing aerosols, but those aerosols do not impose a distinct, angularly resolved, BRDF change. What drives the need for multi-angle polarimetric observations is the additional information content that such measurements provide. In this sense, we have shown the information content is nearly identical regardless of the distribution of the observations within the observed TOA BRDF (ie within a plane or distributed). It is difficult to see how simulation of nonspherical aerosols would lead to another conclusion. Furthermore, while the number of aerosol models is limited, they were simulated for a wide range of optical thicknesses, which produces far greater variation in the TOA BRDF than changes to the aerosol models themselves. The relative (9 satellite vs 9 angle single satellite) information content remains the same throughout this range.

A third and last comment is more central to the application of the information content and concern the assumption made for radiometric noise of the various sensor.

Given the multiple instruments implementation, wouldn't it make sense to assume a higher noise for the instrument flying in constellation ? The advantage in terms of resilience using multiple small satellite could be easily offset by difficulties in achieving the required accuracy from a set of different instruments under different conditions. Have the author tried to perform the information content analysis for the groups of multiple satellite using a higher radiometric noise ? (ie 20% or 50% larger).

This is a matter, perhaps, of experiment design. We choose to assess one thing � the distribution of observations throughout the TOA BRDF � while maintaining uniformity for other aspects of the measurement systems. That way, we could isolate the impact of that variable alone. So, adding different amounts of measurement uncertainty to the different configurations would, to some extent, muddy the waters. We also don't have reasonable information about the degree to which measurement uncertainty would be different for the multi-satellite approach, and even if we did it would be specific to an individual implementation. A subsequent study could take this study as a baseline, and investigate the role of measurement uncertainty for a specific instrument configuration.
Although these are fairly important question in my opinion, they do not necessarily have to be fully addressed within this paper. However the three points at least need some level of discussion before the paper could be published.

Based on the above, and the past-deadline schedule of the review, we made no changes to the manuscript, although we appreciate the reviewer's comments. "

Thank you!