General comments

This is a welcome paper addressing the thorny issue of how best to integrate historical and meteorological evidence to reconstruct climates of the past. It is clearly structured, generally very well written and makes persuasive arguments. However, as you will see from my specific comments below, it includes some statements that are problematic. These mainly relate to the way that the text falls into the “documentary evidence unreliable, instrumental data reliable” trope common to many climatological studies.

There are also some methodological issues that need addressing. These include the need for a clearer description of the way in which documentary index classes are derived. Explanation is also needed as to why – given that the goal of the study is to produce a time series that is “interoperable with Nicholson et al. (2012)” (line 167) – the method used to convert modelled rainfall levels from 20CR and GCMs is not the same as that used by Nicholson et al. (2012).

I’m very surprised that the authors do not cite the recent papers by Nash et al. (2018) and Mutua & Runguma (2020), which present 19^th century documentary climate series for Malawi and Kenya respectively. I would have thought that these are essential for comparison with the results presented for Tanzania.

Mutua, T.M. and Runguma, S.N. (2020) Documentary driven chronologies of rainfall variability for Kenya, 1845–1976, Journal of Climatology and Weather Forecasting, 8, 255, available at: https://www.longdom.org/open-access/documentary-drivenchronologies-of-rainfall-variability-for-kenya–18451976.pdf

Nash, D.J. et al. (2018) Rainfall variability over Malawi during the late 19^th century, International Journal of Climatology, 38 (Suppl. 1), e629–e642.

Specific comments

Lines 16-17, 60 and 507 – I’m not sure the wording “…a more scientifically grounded interpretation of documentary materials…” in the abstract and main text is ideal. This makes a value judgement about the validity of documentary evidence. Maybe ‘climatologically grounded’ rather than ‘more scientifically’ grounded would be better.

Line 38 (and throughout) – best to avoid the use of terms derived from ‘East Africa’ as this is a colonial construct. Academics that I have worked with from the region tend to prefer the term ‘eastern Africa’.

Line 45 – I don’t think Endfield & Nash (2002) used a term as strong as ‘distorted’ to describe European perceptions (and hence descriptions) of African climate. Rather, the descriptions made by Europeans were often framed relative to their ‘home’ climate (particularly during their early years of residence in Africa), so tend to over-emphasise drier conditions. As noted in section 2, they may also be shaped by imperial knowledge-making.

Lines 46-48 – this is a very strong statement – are you sure that absolutely no records made by Europeans describing climate in Tanzania exist between 1861 and 1868? I find that very hard to believe.

Lines 114-127 – this paragraph makes some valuable points. However, it paints the rather sweeping picture that all explorer and missionary descriptions of weather and climate were shaped by imperial agendas and are therefore unreliable. Some descriptions might well be ‘highly subjective’ – especially the broad overviews of climatic conditions in the more general explorer monographs – but other accounts of specific weather events and related phenomena (e.g. delays to the start of the rainy season, counts of rainy days, descriptions of flood events, descriptions of pasture conditions etc) will likely be reliable. I suggest that this paragraph be tweaked to provide greater nuance.

Lines 128-129 – the emphasis on describing extreme conditions is not unique to African documentary evidence and is well documented in historical climatology studies around the world – have a look at some of the excellent reviews by Christian Pfister or Rudolf Brazdil for further details and cite relevant methodological sources.

Lines 130-139 – these kinds of uncertainty surrounding ‘climatically indirect’ indicators of climate variability are routinely dealt with in historical climatology studies and there is a wide literature on this. Again, have a look at some of the reviews by Christian Pfister or Rudolf Brazdil for details. These include explicit guidance on how to handle an ‘absence of discussion’.

Lines 1444-148 – this is a very long sentence – suggest you fragment.

Lines 155-157 – on a more pragmatic note, it is also very likely that they were interested in weather conditions as they relied upon them to grow their own food.

Lines 169-178 – I’m slightly unclear over the methodology used here. Are you following directly the methodology used by Nicholson et al. (2012) whereby individual pieces of narrative evidence (i.e. individual quotes) are read and graded from 1-7 (and then averaged to give an annual picture), or the approach used in most other documentary-based climate reconstructions around the world where collections of quotes from specific months or seasons are read together and given a collective grade? This is important because, as Nash et al. (2021 – section 8.3) have discussed, the Nicholson method tends to lead to an over-representation of drier conditions in the resulting reconstruction. Figs 2-4 seem to suggest some sort of hybrid, which could be problematic if the authors are aiming to replicate Nicholson’s approach as they suggest.

Lines 199-203 – I would appreciate a little more explanation over the way in which individual diary entries are incorporated into Figs 2-4, particularly where they are merged with results from quotations in letters that could refer to conditions over periods of longer than a single day. In effect, you appear to be giving equal weight to (for example) a single daily diary entry describing drought and a letter documenting dry conditions that could span weeks or months. If you are following the Nicholson method described above this could lead to an over-representation of particular conditions, especially if these are isolated quotes from a personal diary rather than a ‘weather diary’ with daily weather-related entries.

Line 225 – there are many reasons for famine, not simply climatic. Do you have any contextual data from missionary sources that might explain the causes?

Lines 319-321 – this sounds like the approach used in the majority of historical climatology studies based on the European tradition, where monthly indices are summed and averaged. There is nothing new here methodologically, so you would do well to cite related sources – see section 8.1 in Nash et al. (2021) for more detail.

Lines 337-347 – I’m intrigued to know why you have adopted this approach when Nicholson has published her method for converting rain gauge data into 7-point index values based on standard deviations from the long-term mean (see section 8.3 in Nash et al. 2021 for a summary). If, as suggested earlier by the authors, they are trying to make results that are interoperable with those of Nicholson, then surely the same method needs to be used in this study?

Lines 513-517 – these sentences again oversimplify the apparent subjectivity of European observers. If you are going to make statements that European observers “regularly misunderstood the climatic and environmental contexts they reported on”, then you need supporting evidence. I would suggest softening of these two sentences. There is as much evidence in the literature supporting the idea that European observers provided reliable eye-witness testimonies of climatic conditions in Africa as there is that their observations were unreliable.

Lines 522-525 – in light of the work by Mutua and Runguma (2020), this sentence requires revision.

Many thanks for this paper. Whilst not an 'expert' in the field of documentary analysis, I approach this paper as a potential user of the outputs and the application of the methodology as a way for testing / comparing records derived from lake sediments against an independent archive. The approach using documentary evidence from Tanzania fill an important gap in knowledge in part of the world where historical (and palaeo!) records are still few and far between. I would also, before listing my corrections, like to apologise profusely to the authors for the delay in submitting this review. I have been in and out of the office for work, and as with everyone, workloads have simply got the better of me this year.

Overall, I think this is a good paper. I have some clarifications to elements of the text. And just being a little pedantic, the authors have over-used the number of figures; many of these should be panel figures or stacked figures. As they are all on the same timeframe, it would be so much easier to compare the figures generated from different datasets if they were stacked and became (for example) Figure 1 (a,b,c). I will also point out where this could be improved. Obviously, this is not a critical element to the robustness of the science presented, but to make it accessible and easier for the reader of the journal, I would strongly suggest this change to the presentation of figures is made.

Just an aside, I wondered whether you had worked with / engaged any Tanzanian scientists in this work, given this is the country of focus? Again, not critical to whether or not this paper is accepted, but the landscape is changing, and it would be wonderful to see Climate of the Past publishing work that is delivered in partnership with scientists from the country that is having science done to them.

The remainder of clarifications are given by line number. Given I am not a technical expert, my comments are really directed to help users of your work (i.e., not historical / documentary analysts) to understand what you are presenting, and to provide them with data / figures that are easily comparable to other archives.

Ln25: I’m not sure I fully understand how the approach is complementary to / different from existing published methods? What issues will this cause of users of the approach used by Nicholson (for example)?

Ln175: What is the cut-off used for high uncertainty vs low uncertainty? How did you determine this (i.e., had to score 3 out of 4 criteria)? How subjective is it? How transferable is this method from one researcher to another if it were developed, for example, in another country within Africa?

Ln187(ish): Figures 2-4. This really isn’t a set of 3 separate figures. Even the combined heading that has been used suggest this. Rather, this is a stacked / panel diagram on a single page, on a single timescale labelled Figure 2a, b, c. Further, the text on this diagram needs to be larger / readable. It’s a big strain to see what is there. Also, on this (and all diagrams with the timescale on the x-axis). What is the notation you use? Is 01 January, 02 February etc or is it seasonal? You need to make this clear somewhere (in figure caption), so we are all on the same page. Originally, I thought it was record 1 from 1855, record 02 etc. The way the notation is at the moment is just too hard to understand and read. Try to align all x-axis scales (you can do this if you change your chart-type – it will also save the pain of the axis labels being fixed).

Ln187(ish): A follow up on the figure caption. No matter how many times I read it, I can’t quite work out what you mean in the last line of the caption (‘Values of +0.05…’) you are going to need to spell this out in clearer language. I can’t even work out where +0.05 would come from if your scale is -3 to +3 (or have I missed the point?). Can’t you use a symbol or colour on the x-axis label to identify where there are no documentary data? There has to be a better way than arbitrarily adding a score. You may also want to increase the resolution of the image, my printer isn’t wonderful, but the copy here is quite low res.

Ln201: Somewhere here you start referring to ‘documentary references for each line’ – what’s a line? Is this a bar on your graph? In which case you would refer to it by month and year (or season and year (see comment above). If this isn’t what is meant here, this whole paragraph needs a re-write to clarify exactly what you are referring to, because I can’t quite follow t. Can you provide a supplementary data table that shows a; ‘lines’ and all data a source used in the paper. At the moment there is no transparency, just your interpretation here. Providing the underpinning data would help the reader better understand the approach you are taking, especially when the text 9as written) is quite complex to understand.

Ln210: How do you define the ‘quality’ of the data? What were the parameters that you set? They would need to be presented here (or in a supplementary document) also.

Ln280: As previously, these are not 6 different figures. It is a single figure with multiple panels. They all need to be stacked and appear on a single page to enable the reader to make an assessment of the data you are providing. They also need to be labelled sequentially (i.e., not refer to 5, 7,9, but to 5,6,7 – or my preference would be a,b,c). Perhaps consider a different way to display these – there are lots of examples in the literature that would help you to better present these. At the minute it’s all very confusing and I don’t think will land with the audience of CoP. At a push this could be 2 figures 5a,b,c and 6a,b,c where you would have a column for each model. They are related (a-c), not independent (5-10) figures. Follow the layout for your Figure 11 – you have not labelled these as 25 different figures!

Ln312: I know it’s going to be hard, but the text size is quite small in your figure 11, is there an outside chance you can make it slightly bigger?

Ln327: Another figure. You could approach it that in any given panel figure ‘a’ is always Mpwapwa, ‘b’ is always Tabora, and ‘c’ is Ujiji. That will also make it easier for the reader to follow the story. You need to take the reader along with you, not make it overly complex for them to engage.

Ln360: Again, think of the labelling. If you do want to keep these as separate figures, they would all need separate captions (i.e. one for Fig 15, a new one for Fig 16 and another for Fig 17; they can’t be all as one. In doing so, you are suggesting to me they are related and should be presented slightly differently.