The authors have responded to all comments, and most of the concerns raised by the referees were accordingly addressed. On one hand, I consider that this manuscript presents a cutting edge case study, which is very straightforward and potentially beneficial to other studies in this region. On the other hand, I still found the manuscript weak in some parts. The main aspect, as referee #2 pointed out, is that the discussion is missing content. To that end, I have some comments and suggestions:
1. The manuscript is not convincing – and probably redundant – in dealing with the effects of open water bodies on the performance of CMORPH rainfall. To support open water bodies’ criterion, the manuscript only uses a relatively vague statement with two supporting references (L119-121). In fact, one of these references (Rientjes et al., 2013, DOI: 10.1016/j.jag.2012.07.009) is not actually a study that assesses the effects of water bodies on rainfall patterns and, therefore, not a suitable reference to support the demand for such analysis.
The study concludes (L842-846) that the CMORPH estimate has the largest mismatches in two situations: at low elevation and in close proximity to open water bodies. Considering that the water bodies selected in this study are actually in the relatively low elevation zones of this watershed, I wonder how redundant this analysis is by considering the terrestrial elevation and the distance to the water bodies separately.
The effect of large water surfaces has consequences to the climate, the time response and scale follow very complex dynamics that probably could not be addressed solely by the effect of one parameter (distance to the water body). To that end, I recommend either removing the bias analysis that considers the distance to the water bodies or, in case the authors have an original argument, adding some solid analysis about the relation of distance to open water bodies and elevation in the discussion in order to exclude a possibility of redundancy;
2. I wonder whether the authors are familiar with the study by Xie et al. (2017, DOI: 10.1175/JHM-D-16-0168.1), which was published shortly before your submission. This seems to be a very relevant study for this type of research. I understand that the timing was not the best, but I suggest considering refining your discussion by using some insights from this global scale analysis of the CMORPH datasets;
3. I am not surprised (not that I should be) that the STB was the best bias correction method. This method forces the estimates to behave as observations. I was intrigued to know what the implications of this result in the precipitation dynamics in this watershed are. In other words, how would the authors explain the good efficiency of the linear method for the Zambezi River basin?
By the way, I suspect that Eq 1 is wrong (L287). Shouldn’t the gauge rainfall (G) be divided by the rainfall estimate (S)? And isn’t the first S also spatiotemporal dependent?
4. One of the advantages of this study would be its application to other hydrological studies since rainfall time series are scarce and difficult to obtain in this region. Bearing this in mind, is the current manuscript providing enough information to be used in other studies? I am not convinced about this. The current version of the manuscript does not provide a data statement (required by HESS), and the authors could fill this gap in a very positive and supplemental manner if they also provide useful information and datasets (e.g. add the best bias correction factors to Appendix 1 and the raster of Fig. 2 as supplemental material/external dataset). I believe it will cost you no substantial time, it will significantly improve the usability of your research and be very appreciated by the community.
Some observations and suggestions I found while reading the manuscript:
L36: Here is a very appropriate place to name the correction schemes.
L40: Why “whereas”? I did not understand the contrast.
L45: “Gauge estimates”. If the study considers the gauge data as the “real” rainfall, I suggest to avoid use the term “estimate” to referring to it throughout the text.
L47-51: If I understood it right, the bias is most overestimated for the very light rainfall (<2.5 mm/d), which is also the range that shows the best bias reduction, which in turn is most effective during the wet season. Would be possible to rewrite these statements to improve its concision?
L68-69: What water resources applications? I suggest the authors be more specific because I cannot see that this is the focus of this study (as it is declared in the text).
L83: This is the place where CMORPH should be written out (not in L110-111).
L105-107: Rewrite this sentence to connect it better to the rest of the paragraph.
L121-122: Please, explain how SRE may be affected.
L125: I do not think applications are limited. Do the authors mean past studies?
L134: I cannot see how these studies highlight the need to correct SREs, especially Cohen Liechti et al. (2012: DOI: 10.5194/hess-16-489-2012), who reports that only 7% of the rainfall in the Zambesi River basin contributes to runoff.
L149: I wonder what in science/hydrology has been fully investigated. Better rewrite this science.
L150: It is very obvious and general any improved data set or method in water resources is important for IWRM. I suggest cutting through the jargon by removing this sentence.
L157: Too many “applications”. Better rewrite it.
L178: What are the common types of precipitation in each of the regions in this river basin? Is the orographic type significant? This information would probably support some results about the elevation bias analysis.
L197: What is the original source of this data? Does this data repository belongs to this software or to the NOAA?
L206: Since the criteria to filter the stations is not shown, I suggest to skip this vague statement and go straight to a statement that defines the number of stations used.
L208: Here is a good place to mention the Appendix 1 (you do not need to remove it from L240).
L215: I find it odd that nothing about the assumption that the rainfall within the 8 x 8 km pixels is taken as homogenous. This could be used to discuss, later on, some bias found in this study.
L246: How this 700 km2 threshold was defined? It needs to be explained in the manuscript. Large lakes are often considered as water bodies with surface areas over 50 km2.
As other reviewers pointed out, the figures need some major work:
Figure 1: a) the map is polluted with unnecessary information: African country names (in the continental map – then this map can be reduced to give some more space to the main map), rainfall gauge station names; b) the elevation palette is not helping its visualization (suggestion: leave it as monochromatic); c) Please improve the river streamline; d) since the results are sectioned in lower, middle and upper Zambezi, it would be very useful show these regions in this map.
Figure 2: what bias is shown on this map? The manuscript presents many bias schemes and on this map it is only written as “bias”;
Figures 3 and 10: a) the axes of these images are not at the same scale; b) the line colours, patterns and thickness do not help the visualization of results. Please keep in mind that this is a graphic that is summarizing many results, therefore it should be very well made.
Figures 5 and 9: Please, do not use 3D-like effects (or similar) in this type of graphics.
Figure 4: It is strange that according to the gauge data in this graphic the mean annual precipitation to all Zambezi regions is lower than 1000 mm/year. It contradicts the information in L178.
For all figures (except Fig 4 and 9): improve resolution, captions and sharpness of each element in the figures.
In the text: please avoid to give any role to the figures but visualization (e.g. do not use the term “Figure reveals” L557).
I am not convinced that the results and discussion as one section was a good choice for this manuscript. As it was mentioned, the discussion is often forgotten and this section remains with a structure that is primarily made to present results (e.g. “Standard statistics”, “Significance testing”, “Taylor diagrams”, “q-q plots” subsections). This section also has a considerable amount of methodological information, which should have remained in the methodology.
I cannot see the relevance of the conclusion “3” in this section (surely it is important in the discussion, but not in the conclusions). This section is somehow addressing the three main objectives of the study (L152-159), but it needs to be more concise.
Last but not least: the manuscript still shows some typos, doubtful word choices and vague statements. It would be very positive if the authors read the text carefully and feel free to do any type of improvements in the text. It will surely be positive for the review process as well as for the reading experience of this manuscript. |