|Review of the paper “Wet and dry spells in Senegal: Evaluation of satellite-based and model re-analysis rainfall estimates” by Cheikh Modou Noreyni Fall et al., submitted for publication in NHESS|
This paper analyses the occurrence of wet and dry spells in Senegal estimated from a set of precipitation products (in situ, satellite and reanalysis), aiming at assessing the performance of the analysed datasets.
This is my second review of the paper and, although I find it slightly improved compared to the first submission, I think that the overall quality of the analysis and presentation is still below an acceptable standard. The revision process has been particularly difficult due to inappropriate and sometimes confusing wordings and lack of explanations. I recommend the authors (in particular the senior scientists in the team) to take care of the quality of the manuscript for next submissions, to make the reviewer’s work easier and more effective. I highlight the importance of taking seriously this point, because lack of clarity makes more difficult (sometimes impossible) to evaluate the goodness of your results and the relevance of your findings.
Although the purpose of the study is valuable and results potentially significant, I believe that the paper still needs major and substantial revision before to be published.
In general, English language still needs fixings, many sentences need clarification, method and analysis need further explanations, and unnecessary repetitions should be eliminated.
Throughout the text, significance word is often used, whereas statistical significance of the results is never assessed, weakening the credibility of the conclusions.
The objective of the paper needs to be clearly declared. Specifically, while reading the manuscript, it is not clear whether observations are taken as reference. This is clarified in the Discussion Section, when it is stated that BK is considered as reference. This choice must be justified and clarified at the beginning of the manuscript.
There are large differences between OK and BK of the same dataset (see e.g. Fig. 4 and Fig. 8), this needs to be discussed.
Discussion and Conclusion Sections look as a summary of the results, with very succinct discussion. We understand that TRMM is generally better than other products. However, is this true for all the metrics? Which are the datasets outperforming on specific metrics? This needs to be clarified. Moreover, implications of the results needs to be highlighted. For instance, what do users of precipitation products in Senegal learn from your results?
Title: observations should also be mentioned, given that depending on the kriging method results are different. I suggest: “Wet and dry spells in Senegal: Evaluation of satellite-based, reanalysis and in-situ estimates”.
Abstract: A sentence on the implications of your results is needed.
L10: How the fact that dry spells are more frequent at the beginning and end of the rainy season indicates false start and early cessation? Please clarify.
L11: “wet strong rainfall events” is misleading. Do you work on wet sequences or sequences of very wet events? Please clarify.
L15: “hydrological climate”: I cannot find this expression in Giorgi et al. 2011. What do you mean exactly?
L9: “potentially high impact indicators”: what do you mean exactly?
L16: How many stations are used?
L25: “generally aggregated”: This is vague. Do you mean that satellite data have finite spatial resolution?
L29: Is the variogram of the region known? Where can we see it?
L2-3: This sentence is unclear, please explain how lambda coefficients are derived.
L19: “Very indirect”? Give a measure of indirectness, if any exists. Otherwise use only "indirect" or explain what you mean.
L23: which proxy?
L5: “Note that these wet and dry spells…” this sentence is not necessary here.
L23: I'd change "extreme long" with "very long". Extreme suggests this is a definition issued from a statistical analysis, which is not the case.
L5: “and because of the synoptic systems associated with the rainfall variability in Senegal” this sentence is not necessary here.
L10: What do you mean with "defined according to..."?
Section 2.1: In this section I have the impression that you use OK and BK as reference datasets, but this is never explicitly said in the manuscript.
L18: “seasonal precipitation” which season?
L20: why the 0.5 threshold is chosen?
L23-25: It's hard to assess similarities just by visual inspection of Fig. 2, an objective analysis (e.g. computing differences) would be helpful. Why do you say that reanalysis underestimate precipitation? Could it be the other way around, i.e. satellite and in-situ observations overestimating precipitation? What I see is that reanalysis seem to show less precipitation, but this is not underestimation, unless you are comparing with a reference dataset. Please clarify and rephrase.
L25-28: This statement is unclear. What is the effect of the morphing process on precipitation we see in Fig. 2c? Please clarify.
L1-2: Unclear. Which differences and areas are you discussing here? Where we can see the bias? Please clarify.
L7: Why BK should be more adapted to compare with satellite estimations?
L19-21: This sentence is too vague. Are you comparing datasets with a reference? Where we can see the overestimation of Heug/small rainfalls?
L22-23: what's the reference dataset?
L24-28: Here you discuss the occurrence of dry days during the rainy season, but there's no way to verify your claims. Are you referring to Fig. 3?
L12-13: In Fig. 4 I see clear differences between BK, TAMSAT and CHIRPS on one the one side and TRMM, CMORPH, CPC, NCEP, ERA5 and OK on the other side for DSl and DSxl. However, clear differences are not evident for DSC10 and DSC20. Any statistical tests are used to assess agreement among datasets?
L22: OK and BK do actually show large differences for DSl and DSxl. Any comments?
L24-25: All the DS* metrics focuses on specific dry spell duration. What do you mean with "DSl is more sensitive during transitional periods"? Please clarify this sentence.
L29: How can you infer a delay of the rainy season from increased occurrence of dry spells in May-July? I'd say that more dry spells increase the probability of late start, but whether this does really happen needs to be demonstrated.
L29: “ERA5 and NCEP reveal overestimation” in comparison to what?
L31-32: “The evolution of DSl etc.” this sentence is confusing, please clarify.
L17-18: “It is important to notice etc.” this sentence is confusing, please clarify.
L24-25: I actually see a slightly increasing trend in only ERA5 and CPC. You should assess trend significance statistically.
L32: By stating that TRMM is "better" than other products and closer to observations means that you consider observations as "truth".
L6: “differences are not significant”: Did you assess statistical significance?
L3: “Significant differences…”: Statistical significance of the results is not actually assessed. “The product resulting from the observations and kriged with the BK method is identified as the reference”: This should be motivated and stated at the beginning of the paper.
L4: “This is justified by the fact that kriged data are more likely to be compared with satellite observations or model data”: This looks as a-posteriori choice. Please clarify.
L11: “This explanation is therefore probably not sufficient”: I don't see any explanation here, you just highlight the outcome of your analysis.
L12: “trends” word is misleading, it looks as you refer to the interannual variability.
L21: what is a “unimodal region”?
L22: what is the difference between “seasonal cycle” and “seasonal evolution”?
Figure 5 and Figure 9: What do the plots actually show? Are data monthly aggregated? How frequencies are computed? Please explain.
Figure 10: Why do some datasets show gaps?
L2: “from several datasets”
L3: “two are based on reanalysis products”
L4-5: “three are based on raingauge observations: CPC Unified V1.0/RT and a 65 raingauge network that has been reggrided by using two kriging methods, namely Ordinary kriging (OK) and Block kriging (BK)”.
L6: delete “same”.
L6: “daily cumulated rainfall on a 0.25 degree regular grid”
L21: major floods in 2009 and heavy rains in 2012, references are needed.
L4: Program 2017: This reference is connected with a conflict database, please provide a reference on drought impact.
L5: ARC 2004: Impossible to find the link between funding and 2014 drought in this 53 page document. Please provide a more accurate reference.
L10: “to better understand multi-scale variability of the rainfall regime”
L12: “multi-scale variability”
L15: “illustrating a hydrological cycle intensification”
L28-31: please define TAMSAT, CMORPH algorithm and TRMM 3B42 data.
L5: “This paper aims to provide an inter-comparison between several datasets based on satellite data (TRMM-3B42 V7, TAMSAT V3, CMORPH V1.0, CHIRPS V2.0), reanalysis products (NCEP-CFSR, ERA5), and gauge observations (CPC Unified V1.0/RT, provided by ANACIM the National Agency of Civil Aviation and Meteorology). This Intercomparison focuses on the ability of these datasets to detect dry and wet spells”
L25: T62 resolution is around 1.8 degree, which is much coarser than the 0.31 resolution in Table 1.
L26: ERA5 data are provided at 0.25 degree resolution, why I see 0.1 in Table 1?
L28: 30’000 stations globally? Please define GTS and COOP.
L32: delete “<”.
L8: “a wet event”
L13: delete Maranan et al. 2018.
L15: please add units to standard deviation and RMSD.
L10: “distribution shows tipping points”
L10-11: “finest resolution”
L22: reanalysis are quite similar?
Figure 2: Please change the colour palette, light yellow may be confounded with white.
Figure 8: please check y-axis label.