See attached

This manuscript provides an assessment of the performance of a new modelling platform applied across Africa. The study is well-aligned to GMD and provides new and interesting results for a domain that is often overlooked. I suggest that revisions are needed before the manuscript could be accepted.

I have provided comments on the pdf of the main document and supplementary material in the attached zip file. The main issue that I have is that the analysis did not use many available measurements that could have assisted with an improved analysis of the model and that there is still a lack of discussion of the results with local literature on the key issues that are discussed in the paper. I unpack both of these below and in the attached. I strongly recommend that these are addressed before the manuscript can be published.

Ground-based data are scarce, but they are not non-existent. This is a major failing of the paper. As it is a Pan-African study, it is understandable that the authors would not need to use all of the measurements from more local studies as some within the same city  would (for example) be within similar grid cells. However, it is not acceptable that they used so few. I have noted in the attached possible datasets that can be used and previous literature that can help in interpretations. For example, in Western Africa there are ground-based measurements from DACCIWA that overlap with at least some of 2017. In South Africa, there are freely available regulatory measurements across the country in 2017 from SAAQIS. My suggestions are not exhaustive, but are examples. I would expect that the authors would search for available data that could be used for model validation. In this manuscript it seems that the search was not yet complete nor comprehensive. I recommend that this gap is filled as this impacts upon the conclusions. For example, in southern Africa, the site used is Cape Point. This is a very unique site and not representative of southern Africa. I can understand why it was used as one of the sites (it is a well-characterized GAW site), but I cannot understand why it was the ONLY site used. There are regulatory grade monitors across the country. Using a handful of these could at least provide a more comprehensive assessment of the performance of the model across more of the region. I do understand that accessing data across Africa can be harder as there is not one central place such as in the US. However, such a manuscript should aim to represent the region it is modelling as well as it can, which includes trying to use as many measurements that are available.

In addition, there are gaps in the referenced literature that could support some of the analyses and provide more context for these (e.g. are these trends seen in other papers for some key aspects). Looking through, it seems that global analyses and references are used more than local studies. For example, the bias in Kampala could be because of the complex spatial distribution of PM2.5. The are two papers from Makerere University noted in attached that do not have data in 2017, but can support the spatial variability of PM2.5.

Also, for example, for the timing of fire events and impacts on pollution, there is a lot of literature across the regions. A starting point are papers such as,  

• Archibald, S.: Managing the human component of fire regimes: lessons from Africa, Philos. T. R. Soc. B., 371, 20150346, https://doi.org/10.1098/Rstb.2015.0346, 2016.
• Archibald, S., Scholes, R. J., Roy, D. P., Roberts, G., and Boschetti, L.: Southern African fire regimes as revealed by remote sensing, Int. J. Wildland Fire, 19, 861–878, https://doi.org/10.1071/WF10008, 2010.

Another example, ozone exchange between stratosphere and troposphere is an important component to the paper – but what is known about it in the domain (or even some regions)? This paper focuses on just one year and a few data sources, so other studies on this could help to support and provide context for this study.

I have tried to highlight some areas that I am familiar with where other studies can assist, however, I don’t know all the literature related to this topic. So I recommend that for the key points, that other literature in these regions is looked for to provide such context and support of the findings.

I have provided more detailed comments on the attached pdfs.