the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Apr 2025
Drivers and impacts of westerly moisture transport events in East Africa
Abstract. Equatorial East Africa (EEA) experiences strong intraseasonal precipitation variations; developing understanding of the processes that drive these variations can improve predictability and help local populations be better prepared for extremes. Previous research has highlighted anomalous westerly moisture transport from the Congo basin as an important driver of enhanced precipitation in EEA. Here, we have developed the first spatially unconstrained, objective framework to detect what we refer to as westerly moisture transport events (WMTEs) in ERA5 reanalysis data from 1980 to 2022, revealing new insights into the drivers of these westerlies and their impact on EEA precipitation. We show that over EEA, WMTEs were most common in January and February between about 5° S and 15° S, where there were typically 4–6 WMTE days per month, with each individual event persisting for around 2–4 days. During the March to May wet season in EEA, there were on average around 1–2 WMTE days per month. Using a precipitation attribution algorithm, we estimate that WMTEs were associated with up to 60 % of precipitation during January and February in Tanzania, and up to 20 % of precipitation during March–May to the East of Lake Victoria. Consistent with previous work, we found that WMTEs were more likely during phases 2–4 of the Madden-Julian oscillation (MJO). We expand on previous case-study based investigations by showing that the presence of a tropical cyclone anywhere in the south-west Indian Ocean makes WMTEs up to three times more likely, even during inactive or unfavourable phases of the MJO. This work builds on previous studies of the westerly wind feature by providing an objective framework to describe EEA westerlies and joins previous work in highlighting the complex nature of the interactions between different features of tropical meteorology that drive these short timescale variations.
- Preprint
(17612 KB) - Metadata XML
-
Supplement
(12117 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1694', Anonymous Referee #1, 14 May 2025
-
AC2: 'Reply on RC1', Robert Peal, 15 Jul 2025
Dear Reviewer,
Thank you very much for your thorough review of our work, for your positive assessment of our manuscript, and for your constructive feedback for improvement. Please find attached our responses to each of your comments.
Kind regards,
Robert Peal and Emily Collier
-
AC2: 'Reply on RC1', Robert Peal, 15 Jul 2025
-
RC2: 'Comment on egusphere-2025-1694', Anonymous Referee #2, 26 May 2025
This study investigates the prevalence and impact of westerly moisture transport events (WMTEs) on East Africa. The importance of moisture flowing from the Congo has been highlighted in previous studies, but this study applies a novel and more complex approach to identify the WMTEs. The method adapts previous work looking at atmospheric rivers for this tropical context. The method is applied to reanalysis and rainfall observation products. The impacts and wider features considered are rainfall, Madden Julian Oscillation (MJO) and tropical cyclones. For the methodological parameters and region chosen, the study finds that WMTEs are most prevalent during in January-February and are associated with a majority of rainfall in Tanzania during this period. Previously published relationships between WMTEs and the MJO are confirmed with this methodology. The results provide a more rigorous analysis than previous work of the connection between WMTEs and tropical cyclones, and finds there to be an association.
This manuscript is one of the best written and most thorough of any paper I’ve reviewed – thank you to the authors for making the review such an easy process. I consider the application of atmospheric river methodology to provide a useful step forward in rigour, and to have enabled the authors to provide some useful quantification of the phenomenon. I have no major objections to publication, but there’s a few aspects that I think the authors need to clarify – there were a few bits of text where it was difficult to understand how the conclusions related to the presented figures. I outline these points below.
Minor comments
L91 – 45 degrees. Is this plus/minus 45 or plus/minus 22.5 degrees. Please clarify.
L108 – The lat range used is quite a bit further south than in Finney et al. Whilst the sentence does not say anything incorrect, I think it is worth stating the difference clearly so that readers know to expect that there may be some fundamental methodological reasons for differences in results to Finney et al.
Sec2.2 - I’m would have liked supp text S1 to have been referenced in the methods section. I was pleased to stumble across it later in the paper, but it’s really at this point that I want to dig into sensitivities.
Sec2.2 - sensitivities to 500km distance to TC have not been discussed? Please share any findings you already have on this, and at least acknowledge this parameter choice when discussing the TC results in 3.2. In particular, the result in Finney et al. of suppressed occurrence of westerlies around Lake Victoria when TCs are in Mozambique channel would not be seen when using this constraint (I think). Please consider discussing this point more fully. Whilst, I think the authors have done a nice job analysing this, I think some further explaoration is needed to pick apart some of the regional details that might matter.
Sec2.3 - similarly to above comment, some sharing of any findings regarding sensitivity to precip attribution parameters, e.g. maxima separation distance.
Fig1 – It would be nice to see a range of different kinds of examples (including some that seem not to have worked quite as well) in order for readers to get a feel for this method. It’s nice to have an example showing it worked nicely, but it’s a new method so a few more in a supplementary figure would be appreciated.
Fig1 caption. “black dashed line” - I can’t see this well. I suggest maybe making it red, or something similar to make it stand out more.
L126 / Fig2 caption – I don’t think this figure and results are based on the constraint of WMTEs crossing the EEA line? Otherwise panels such as fig2d wouldn’t make sense because they ahve no events over that line. Can you it clear that these maps are independent of that constraint.
Fig2 – have you considered including a metric for interannual variability? I imagine some years there might be quite a few more instances, and in some years no instances. Perhaps an IAV metric could be added as another row to indicate where such variability is high? Possible metric that could indicate something useful related to this... 95th percentile of yearly WMTEs events.
L143– can you introduce to the reader why you are looking at the curl.
L152 – TC_WMTE looks to have more intense curl and vectors than noTC_WMTE. Is this true? And is it significant? If so, could you mention it. Possibly showing a 4 column with the difference of these two would help see that more clearly.
Fig3 – Whilst I think you have included the right main figure for your paper, there is part of the dynamics that is left out that you might like to add as supplement. Finney et al fig5 shows an intensification of low-level westerlies across the congo on westerly days. Adding a supp fig equivalent to fig3 but with 900 or 850hPa moisture flux or winds, might add a complimentary view of what goes on dynamically during WMTEs.
Fig 3 caption - “noTC-WMTE” is a bit confusing, I initially read it as “no(TC-WMTE)”. I don’t think you need to change it. But can you spell it out in the cpation instead of just saying a “noTC-WMTE crossing..”
Fig4 hatching – I feel like it would be more useful to show where there is a significant difference compared to inactive days.
L210 - “thoughout the year” - I’ve struggled to follow how yo ucan say this when your fig5 shows the full year and JF, but the full year is dominated by JF, I think? To say throughout the year, you’d need to look at MAM, JJAS and OND separately in your supplementary figures. This would be interesting to see, and would probably help compare to previous results, especially if you looked at MAM. I think you need adjust the text, if you dont analyse MAM on it’s own.
L215-224 - There’s also a slight other issue with your comparison to previous studies, because you have defined WMTE quite far south compared to the focus regions of previous studies. You may see a positive relation to WMTE because there is a WMTE over tanzania, but that WMTE is not spreading over Lake Victoria. I think you need to acknowledge this, and at least highlight the need for futrher research to pick it apart.
L246 – Finney et al showed that westerly day rainfall was upto 200% of daily average rainfall. I would call this up to 100% increase, as daily rainfall would be 100%. Check your sure of your phrasing.
Technical comments
L261 – typo – capital “S” on “WMTES”
Citation: https://doi.org/10.5194/egusphere-2025-1694-RC2 - AC1: 'Reply on RC2', Robert Peal, 15 Jul 2025
Data sets
Westerly Moisture Transport Events (WMTEs): A flexible framework for studying intraseasonal variability in East Africa Robert Peal and Emily Collier https://zenodo.org/records/15173985
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 673 | 42 | 21 | 736 | 35 | 24 | 57 |
- HTML: 673
- PDF: 42
- XML: 21
- Total: 736
- Supplement: 35
- BibTeX: 24
- EndNote: 57
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of « Drivers and impacts of westerly moisture transport events in East Africa » by Peal and Collier
This article analyzes events of westerly moisture transport across Africa (and nearby regions), and links those events to the phase of the Madden-Julian Oscillation (MJO), and the presence (or lack) of tropical cyclones (TCs) over the Indian Ocean. It finally provides an estimation of the precipitation that can be attributed to those westerly moisture transport events (WMTEs).
The authors developed a detection algorithm, quite similar to those used to monitor atmospheric rivers, to identify the WMTEs. The algorithm is unique in that ARs have a strong poleward component, while inter-tropical WMTEs are mostly zonal. The approach is sound and there was a clear need to have more in-depth understanding of those WMTEs.
The following analyses are less convincing. They remain very descriptive. The authors build samples consisting of different phases of the MJO, combined with the presence or absence of TCs, but the main object proposed in this work (the WMTEs) are not analyzed with more detail than just their presence or absence. Another possible issue comes from the attribution of rainfall to WMTEs: the justifications are not very clear but I understand (Fig. 1) that precipitation extending far beyond the contours of the detected WMTEs can be considered as linked to it. This is clearly not what is usually done when working with those detected moisture transports.
For those reasons, I believe that there is a very good study to be done with those WMTEs, but I think that the current version of the manuscript required major modifications and improvements before it can be accepted for publication. Even though I recommend major corrections, I would like to encourage the authors — there are really great ideas here, we’d just need more physical characterizations (and understanding) of the WMTEs, their properties, and their mechanisms (see detailed comments below).
Major comments
1. What are the WMTEs, physically speaking? ARs have been described as “A long, narrow, and transient corridor of strong horizontal water vapor transport that is typically associated with a low-level jet stream ahead of the cold front of an extratropical cyclone” according to the AMS Glossary. This is because ARs are a concept mostly used for extra-tropical climate and weather, where transient disturbances (atmospheric highs and lows) have fronts that separate air masses, where baroclinic instability develops. This definition (and even the schematics that are often provided in some studies) are useful to conceptualize the AR objects, from a physical point of view. The concept of WMTEs strongly differs, even if it is detected by algorithm that are not so different. This is not only because WMTEs are more zonal, but also, because they develop in a tropical climate, where vorticity is very low because of the proximity to the equator. In those regions, large-scale cellular circulations (that are, partly, conceptual objects) can develop (including zonal ones), as part of the MJO at the intraseasonal timescale, or ENSO at the interannual timescale. Are WMTEs linked, or even part, of those large-scale cells? Do they correspond to “bursts” or transient increases in their lower branch, when moisture convergence is located further east over the Indian Ocean sector (since you use a quite high threshold to define them, you only depict the most intense phases of those westerly circulations)? Do the WMTEs help refine (regionally) the conceptual schemes of the MJO as proposed by Madden and Julian (e.g., Fig. 3 of their 1994 paper, cited in this work), or propose more detailed vertical cross-sections? So, overall: what are WMTEs, physically speaking?
The same could be said about the interannual timescale, especially in OND when the EEA region experiences the Short Rains whose interannual variability is strongly tied to the Walker circulation (= are WMTEs partly driven by changes in the Walker-type circulation ?). See e.g. an (already old) paper by Hastenrath about the detection of those cellular circulations: Hastenrath, S (2000) Zonal circulations over the equatorial Indian Ocean : Journal of Climate 13, 2746-2756.
When reading the article I had a similar question, about their interpretation, when the dominant winds are easterly (usually the case where/when trade winds are well developed) vs. the regions where westerlies prevail (like the northwestern Indian Ocean in boreal summer, due to the monsoonal circulation). In other words, how can the threshold used differentiate transient from seasonal circulations. But the questions raised above are more general and encompass that particular case. All these questions are, in my opinion, super important, to know what are the objects we’re talking about, and that we consider so extensively in this work.
2. There are no metrics here to characterize the WMTEs, like their length, width, tilt / direction, duration, integrated moisture transport (and location of the maximum), … The AR community produced tens of articles showing that those descriptors are important to better analyze their impacts on rainfall, and help better understand the mechanisms responsible for rainfall, and its space-time variability (including daily amounts or even extremes). In addition, relating rainfall to the location of the outflow boundary of the WMTEs might give potentially interesting results; similarly, the inflow location might give insight into the moisture sources.
3. Attribution of precipitation to WMTEs. I understand you’ve considered the Boolean union of the WMTE and Precipitation > 1mm.day-1 contours and attributed all rainfall falling within that new contour to WMTEs. This would mean that precipitation occurring outside the WMTE contour is yet attributed to it. This would be clearly an issue, especially since WMTE can be linked to the MJO that promoted the development of large-scale convective clusters (some of which can reach 10,000km diameters). This would imply you could attribute MJO-caused precipitation to WMTEs. Considering the Boolean intersection instead of the union would certainly decrease the contribution of WMTE to rainfall totals, but the approach would be more conservative and more robust. This is the one traditionally used by the AR community. Yet, double (or triple!) counts are still possible (i.e., in the worst-case scenario, attributing rainfall to WMTEs, MJO and TCs). See e.g. Dacre’s papers about the interactions between atmospheric rivers, warm conveyor belts and cyclones: 10.1038/s41612-025-00942-z; 10.1029/2023jd040557; 10.1175/JHM-D-18-0175.1.
4. Relationship between WMTEs and TCs: is it a “chicken and egg” problem? Do the WMTEs feed TCs with moisture, or do WMTEs respond to the development of TCs (or convection, more generally) through ageostrophic circulations? While the causality itself might deserve dedicated studies, it would be quite straightforward to see whether the WMTEs develop before, or after the TC. This could be useful as a first clue to understand how both circulations behave and interact.
Minor points. There are not many of them because I mostly focused this first review on the main points listed above.
l. 92. How precisely is the 70th percentile of moisture transport calculated? By considering both signs (i.e. easterly and westerly), or just westerly transport occurrences?
Figure 2. If the main interest in this work is to assess westerly moisture transport across Africa and reaching the EEA region, then the choice of the domain is a bit strange — shifted eastwards, and giving more importance to the Indian Ocean region. Previous work (e.g. on the regional influence of the MJO) discussed zonal moisture convergence between the Congo basin and the Indian sector, so the WMTEs of interest, advecting moisture towards EEA, should be placed over inter-tropical or equatorial Africa. Moisture sources might be continental (Congo basin) or oceanic (Atlantic sector). Why such an eastward-shifted domain? WMTEs have lesser importance for EEA if they occur east of it?
Figure 3. Why show the curl of the moisture fluxes, rather than their convergence? I’m not saying this is wrong, but this needs to be explained. Moisture convergence may make more sense for rainfall analysis. Vorticity is certainly more meaningful when assessing the links with cyclogenesis.
l. 156. TCs themselves are not independent of the MJO. The authors did not discuss this point.
Bessafi, M. & Wheeler, M. C. (2006) Modulation of south Indian Ocean tropical cyclones by the Madden-Julian Oscillation and convectively coupled equatorial waves. Mon. Weather Rev. 134, 638–656
Klotzbach, P. J. (2014) The Madden-Julian Oscillation’s Impacts on Worldwide Tropical Cyclone Activity. J. Clim. 27, 2317–2330
Diamond, H. J. & Renwick, J. A. (2015) The climatological relationship between tropical cyclones in the southwest pacific and the Madden-Julian Oscillation. Int. J. Climatol. 35, 676–686
Figure 6. What is the use of defining an EEA region like in Finney et al. if it’s not used in this work, e.g. to compute regional indices?