the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 02 Mar 2026
Changes in rainfall extremes over Southern Africa over the 20th and 21st centuries simulated by a high-resolution regional climate model and their connection to the Agulhas Current System
Abstract. In Southern Africa, precipitation is a crucial variable that is closely linked to agriculture and water supply. Additionally, extreme precipitation leads to devastating flooding, and heavy rainfall events pose a significant threat to the population in this region. Here, we analyse historical and scenario-based climate simulations, focusing on the spatial patterns of extreme precipitation and its projected future changes. We also investigate whether the Agulhas Current, a major regional oceanic current system, influences the frequency or intensity of extreme precipitation.
For this purpose, we analyse high-resolution simulations with the regional atmospheric model CCLM, conducted with a higher resolution (16 km x 16 km) than that normally used in the CORDEX project. One simulation is driven by meteorological reanalysis, whereas other simulations are driven by global coupled simulations that regionally resolve the Agulhas Current, its leakage and retroflection. The simulations cover the last few decades and the 21st century.
During the present period, the regional simulations indicate the strongest precipitation over Madagascar, the Mozambique Channel, and the adjacent mainland. Extreme rainfall events are most intense in Madagascar's mountainous regions, the Drakensberg, and the African Great Lakes. The extremes are generally stronger in the Summer Rainfall Zone than in the Winter Rainfall Zone. This climatological pattern agrees with available observations.
In the scenario simulation, extreme events are projected to intensify along the South African coast. In KwaZulu-Natal province, the heaviest future rainfall event is twice as strong as the strongest extreme simulated in the historical period and the recently observed disastrous extreme event in April 2022.
The simulations do not reveal a discernible impact of variations in the Agulhas Current System on strong rainfall events along the South African coast.
- Preprint
(43634 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-806', Izidine Pinto, 09 Apr 2026
-
AC1: 'Reply on RC1', Nele Tim, 15 Jun 2026
Thank you very much, Mr Pinto, for revising our manuscript. In the following, we respond to your comments. Your initial comments are in bold face, followed by our response
L37: I’m unclear on what the authors intend by the phrase 'models are burdened…”
- We would change it to: „In addition, models are affected by uncertainties in the representation of processes that influence precipitation extremes.”
L39: The intent of figure 1 is not clear in the introduction and how it fits within the general background.
- Figure 1 shows the estimated precipitation trends for Southern Africa in CMIP6 and CORDEX scenario simulations. Our scenario simulations fit in this framework, although, as we explain in the text, they go beyond the IPCC setup because the spatial resolution is higher, and we also include, regionally, a very high-resolution ocean that resolves the Agulhas current. Figure 1 summarises the findings of the CMIP6 simulations for this region. It shows the projected changes, how well the ensemble members agree on them, and the differences between global (CMIP6) and regional (CORDEX) simulations. With this figure, we introduce the current state of research, summarised by the IPCC. We would make these points slightly clearer in a revised version.
L42-46: The sentences are not clear, and more context is needed, such as time periods, scenarios and seasons.
- We would add the corresponding time periods, scenarios and seasons in a revised version.
L83: Suggest replacing “negative trends’ by “downward trends’
- We would change it as suggested.
L85: This manuscript (https://doi.org/10.1029/2020EA001466) could be interesting here.
- We would cite and discuss this paper in a revised version.
L106-108: I recommend having a look at this paper https://doi.org/10.1002/wcc.70025
- This is a very good suggestion, which we missed as this paper was published shortly before our submission. We would cite and discuss this paper in the introduction.
L146-149: I recommend adding a box of these regions with labels in one of the figures.
- We would add boxes of the regions in Figure 4.
L150: Is this value representing the 99th percentile of wet days?
- It represents the 99th percentile of all days. Days without any precipitation are included in the calculation.
L51-153: Could you clarify or provide justification for why the order of operations varies depending on the size of the regions?
- In the larger regions (SRZ, WRZ, and the coastline), precipitation varies substantially across space. If we calculate the spatial mean before calculating the 99th percentile, we would falsely lower the extreme events. The spatial variability is much smaller for the Cape Town and KwaZulu-Natal region, as they cover only a fraction of the number of grid boxes compared to the other regions. Therefore, as the small regions are more homogeneous, this effect is less important. We will include a brief comment on this point.
L163: For the GEV fit, is it assumed that the time series is stationary or a scale fit is included as covariate (e.g. section 4.3.2 https://ascmo.copernicus.org/articles/6/177/2020/)
- Yes, we calculated them for the historical CCLM simulation and the observational datasets, assuming the distributions are stationary. Here, we are evaluating the CCLM simulations against observations for the magnitude of extremes, not for possible historical trends.
L170: The authors highlight challenges related to observational data for model evaluation in lines 84-86. Are ERA5, TAMSAT, and CHIRPS the most suitable datasets for this purpose? Could alternatives such as ERA5-Land or Multi-Source Weather (MSWX) offer improvements? For reference, Tim et al. (2023; doi:10.5194/wcd-4-381-2023) utilized different datasets in their analysis.
- Based on our experiences and learnings from the previous data analysis of Tim et al. 2023, we are sceptical that further data sets would improve the present study. The MSWX data set is geared towards operational forecasts, and not towards long-term climate studies, and ´thus its stationarity is not assured. It is a combination of several datasets, for which the authors conduct a suitability analysis based on the timescale. Thus, it is probably not a temporally homogeneous data set. ERA5-land is a subproduct of ERA5, for which a higher resolution model has been driven by the ERA5 reanalysis with the inclusion of a land model. Quite probably, while the individual precipitation extremes might be better captured, the long-term evolution of the extremes will probably be very similar. The main problem, in our view, is that the available datasets, as shown by this study and the cited studies, do not fully align. There are clear discrepancies depending on the period considered. Thus, adding more would widen the spread still further and only confirm that precipitation datasets do not fully agree, especially in this data-sparse region, and that uncertainties are large.
L174: Would incorporating the seasonal rainfall cycle for the regions identified in Section 3.2 provide additional value here (and section 4.2)? Additionally, I would assume that section 4.1 would be already covered in Tim et al. 2023.
- Validation of the seasonal mean precipitation was already included in Tim et al. (2023) using ERA5, CRU, GPCP, GPCC, and JRA-55. The literature indicates that TAMSAT and CHIRPS are reasonable choices for validating extreme precipitation; thus, we include a brief validation of the annual mean precipitation using these two datasets here. We refer interested readers to Tim et al. (2023) for seasonal mean precipitation.
L196: “This study…” It is not clear to me which study is referenced here.
- It refers to Treblanche et al. (2022). We would make this clear in a revised version.
L205: Please clarify what you do with the resolution mismatch of the datasets. This should be clarified in the methods section.
- We interpolated CHIRPS to the CLM grid using nearest-neighbour weighting. We would add a description in the method section.
L169: Should section 4 be under Results section?
- We think this is a matter of writing style. We consider that the validation is not a result but a necessary precondition to test the model’s skill, prior to analysing extremes in the results section. The interested reader can skip the validation section and go directly to the result section or vice versa.
L206: Figure4. Should the authors also include the regions identified in Section 3.2?
- Yes, as we mentioned in the response to a previous comment. We would include the regions as boxes in this figure.
L209: “In our coastal domain…” Please provide boxes identifying the regions in the map.
- Yes, we would include the regions as boxes in this figure.
L216-221: In here the authors should not compare return periods of two datasets with different lengths and periods! Furthermore, it is unclear whether this analysis and its findings contribute meaningful insights.
- We do not agree with the reviewer on this point. The return periods estimated from series of different lengths can indeed be compared. The uncertainty estimates for these two estimates will differ, the ones derived from the shorter time series being wider, but the return period and its uncertainty (a function of the GEV parameters) can be estimated from the available data and then compared. This comparison should take into account the different uncertainty ranges.
- The comment regarding the different periods (not just the length) is indeed a good point, which we will explain better in the revised manuscript. The underlying distributions can be different, but this is precisely the point of our test: to test whether that parameter significantly differs in the two periods.
L222: Is this because of the short sample?
- That observational estimates differ may indeed be due to the small sample size, particularly for the satellite-derived data, since a small sample sizes causes a smaller precision in the estimate. The monitoring methodologies, however, surely also have an impact. Without a specific comparative study of observational datasets, it is difficult to ascertain the cause. We include a sentence in this paragraph.
L229-230: I suggest adding this information in the methods section and the reasons. Which Niño index was used and from which dataset was it calculated, what are the regions with high correlation between El Nino and rainfall (see for e.g. 10.1088/1748-9326/ade60e)? As it stands, the analysis appears arbitrary and lacks context within the current narrative. Additionally, I suggest including a paragraph with a literature review on the relationship between El Niño and extreme precipitation to provide necessary context in the introduction.
- We would expand the ENSO topic in the section Region to provide more background on the climatology of this region and its connections to ENSO. There, we would specify the methods and data sets used in the methods section.
L237: I recommend merging Sections 4 and 5 to create a more cohesive and accessible narrative. In its current form, the structure is somewhat disjointed and challenging for readers to follow. Additionally, I recommend including a detailed description of the analysis and methods in the methodology section to ensure the study is easily replicable.
- As per the response to a previous comment, we prefer to separate the validation from the results. Validation is a prerequisite to answering the research question itself. Some readers, perhaps users of the regional climate model, may be more interested in the validation of the model, so they can go directly to this section. Other readers may be more interested in the results. In practice, this makes very little difference in the manuscript, but keeping the two sections helps the reader navigate the paper more easily
- As explained in previous comments, we would add the missing technical information for the methodology in the methods section.
L242-247: I am not sure how relevant it is to provide dates of extreme rainfall from climate models that have been run in a climatological setup as the model is not initialized to reproduce the real-world sequence of weather.
- Our aim is not to compare extremes that are temporally aligned, but to provide the maximum range of daily precipitation. This provides information on when, during the simulation period, the most extreme event is projected to occur. For instance, is it towards the end, when the climate has changed the most, or at the beginning of the simulation, which corresponds more to the recent state of the climate? We think this is relevant information, although we, of course, are aware that the model dates cannot be directly compared to real dates. We will add a sentence to make this point clear
L247-248: “Thus, the heaviest rainfall occurs over the Mozambique Channel…”. Mozambique channel is over the ocean and the figure are not showing it!
- We thank the reviewer for picking up this oversight. The reviewer is right. We will include a warning in the text that this is not included in the Figure. We did not want to expand the figure including ocean regions, as this would reduce the visibility of land areas, since our study focuses on extremes on land.
L250: Would this information be better presented as timeseries for the regions of interest introduced in section 3.2? The current map is ‘too noisy’ and I’m not sure what is the main take away of this information.
- The reviewer is right that the maps are noisy, but they do show a discernible structure that we think is interesting to discuss. The text is admittedly too brief in this regard, and we will expand the discussion of Figure 7.
- Figure 7 shows the trend of the annual maxima in the historical and scenario CLM simulations. The choice of the annual maxima as a target is dictated by the focus on the strongest extremes, but the drawback is that the trend estimation is noisy, as the annual maxima are determined by just one single event. Nevertheless, the spatial structures shown in Figure 7 do have similarities between the two simulations, indicating that they represent the signal of greenhouse gas forcing: a slight decrease in the south-west and north-west and an increase in the north and north-east. The spatial pattern is also somewhat stronger in the scenario simulation, thus supporting the interpretation as greenhouse gas forcing.
- Replacing this figure with a few time series would fail to convey this spatial agreement between the different simulations. Thus, the point of figure 7 is not only the magnitude of the changes per se, but also their spatial distribution.
L284: Why were this region chosen? It is unclear whether this region experiences the greatest impact from the Agulhas Current.
- It seems plausible that ocean impacts would be stronger in coastal regions, and previous studies have indicated that the Agulhas Current affects mean coastal precipitation. This coastal region is also more densely populated.
- Before conducting the study, it was not known where the impacts on precipitation extremes would be, so it seems to us to be a plausible choice.
L293: I’m surprised by the fact that tropical cyclones contribute to extreme rainfall in this region! Are there any references supporting this?
- It’s true that most tropical cyclones make landfall further north. But several have caused devastating flooding on the northeast coast of South Africa: https://en.wikipedia.org/wiki/Tropical_cyclones_in_Southern_Africa, Chikoore, H., Vermeulen, J.H. & Jury, M.R. Tropical cyclones in the Mozambique Channel: January–March 2012. Nat Hazards 77, 2081–2095 (2015). https://doi.org/10.1007/s11069-015-1691-0)
L295: I suggest including these details in the methods. As mentioned before, the analysis appears arbitrary and lacks context within the current narrative.
- We would include the methodological details in the methods section in a revised version.
- The characterisation of a depression as a Tropical Cyclone from gridded data is, however, subject to some extent. As per the comment by reviewer #2, we will include a more objective identification of tropical cyclones or subtropical cut-off lows using geopotential and air temperature data aloft. So, this section will be expanded.
L296-299: I suggest aligning the definition of tropical cyclones and cut-off-lows with the literature. Cut-off-low do also form north of 30°S as it is the case of Durban floods in April 2022 and others.
- See, previous comment, we would adjust our definition of tropical cyclones and cut-off lows by the temperature in 300-500hPa for a better distinction between both causes of extreme precipitation.
L330-389: I am not sure what is the validity of studying specific events with a climate model as they’re not designed to reproduce exact individual historical events exactly as they happened and they have internal variability.
- We conducted these analyses to assess how these types of extreme events, such as the drought in Cape Town and the flood in KwaZulu-Natal are projected to occur in the future. We agree that climate models do not reproduce individual historical extremes and that the climate simulations are not temporally aligned with observations. However, they produce extremes of a similar type that correspond to the climate states during the simulation periods. Therefore, our study puts the extremeness into perspective in the warming climate.
L391-392: This is not the correct.
- We disagree with the reviewer on this point. The IPCC ‚Regional fact sheet - Africa’, states: „The frequency and intensity of heavy precipitation events are projected to increase almost everywhere in Africa with additional global warming (high confidence).“ https://www.ipcc.ch/report/ar6/wg1/downloads/factsheets/IPCC_AR6_WGI_Regional_Fact_Sheet_Africa.pdf
L392-397: Sentence is unclear.
- We would rewrite this sentence in a revised version.
L400-402: I’m not sure how is this information relevant here. Should this be in the introduction. Furthermore, “Tropical cyclones are projected to intensify” is not clear what aspect of tropical cyclones are projected to intensify and for which ocean basin. And the correct reference should be Seneviratne et al 2021 (https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_Chapter11.pdf)
- As tropical cyclones are among the causes of extreme events in southern Africa, their projected changes in frequency and intensity are linked to the occurrence and intensity of extreme precipitation in our research area. This is why it was mentioned in the discussion to place our results in context with previous studies on extremes and their causes. We will place the main part of this information in the introduction, and just summarize it in the discussion to help with readability.
L390: The conclusion needs to be revised considering the above recommendations and suggestions with clearer acknowledgment that the results are based on a single model and simulation, and with the associated uncertainties and limitations explicitly discussed and contextualized (see for e.g. L48-L51). Additionally, do these results improve or are they better/worse than the ones from CORDEX models? What can we learn from these higher resolution simulations for the region that we didn't know before?
- We would restate in the discussion that our study is based on a single model and mention the associated uncertainties. Certainly, our simulation adds value due to its higher horizontal resolution compared to CORDEX. Precipitation is a variable that is highly dependent on high-resolution data due to its spatial heterogeneity. We would also discuss this in the revised version.
And lastly, I suggest improving the colours in the maps. Here is a visual guide that may be helpful
https://www.ipcc.ch/site/assets/uploads/2022/09/IPCC_AR6_WGI_VisualStyleGuide_2022.pdf
- Thanks, we would do so in a revised version.
Citation: https://doi.org/10.5194/egusphere-2026-806-AC1
-
AC1: 'Reply on RC1', Nele Tim, 15 Jun 2026
-
RC2: 'Comment on egusphere-2026-806', Anonymous Referee #2, 07 May 2026
This manuscript addresses an important relevant topic by investigating extreme precipitation over southern Africa using high-resolution (~16 km) regional climate model (CCLM) simulations. The study has potential value for publication in the journal Weather and Climate Dynamics; however, the manuscript in its current form requires substantial revision.
The main concerns relate to the robustness and interpretation of the extreme precipitation analyses, the physical realism of some simulated extremes, the simplified synoptic classification methodology, and the overall structure and clarity of the manuscript. Several methodological choices and interpretations require stronger justification and supporting references, while some sections and figures would benefit from substantial tightening and clearer presentation, respectively.
Please find attached the referee comments PDF file.
-
AC2: 'Reply on RC2', Nele Tim, 15 Jun 2026
We thank the reviewer for their time and effort in reviewing our manuscript. In the following, we respond to the reviewer’s comments. The original comments are in bold face, followed by our response
Major comments
Comment 1
Lines 12–13: Please explicitly state the latitudinal and longitudinal extent of the Summer and Winter Rainfall Zones in the Abstract. In addition, Figure 1 should clearly delineate all regional boundaries and key locations described in the manuscript (e.g., SRZ, WRZ, coastal strip, Cape Town region, Good Hope Line, Cape Basin, Great Lakes, Drakensberg). This would substantially improve readability and spatial interpretation throughout the manuscript.
- We will add the latitudinal and longitudinal extent of the Rainfall Zones and the mentioned locations in brackets. Additionally, we will add boxes of the analysed regions in Figure 4.
Comment 2
The manuscript highlights the important role of synoptic disturbances in driving intense precipitation; however, the Introduction lacks sufficient discussion of the relevant synoptic-scale systems. For instance, studies along the Australian east coast have shown that extreme precipitation events are often linked to East Coast Lows forming over the East Australian Current system and nearby regions.
These systems can include embedded thunderstorms activity with spatial scales smaller than the ~16km resolution used here (see https://doi.org/10.1007/s00703-015-0382-4). In this context, it would be useful to clarify whether mesoscale convective systems play a role in the study region, and if so, to provide some justification for the chosen 16 × 16 km resolution, especially given that convection is likely parameterised.- We discuss the drivers of precipitation in the Region section. We would extend this description of the synoptic drivers of precipitation for this region. An extensive description can be found in the cited book chapter (Rouault et al., 2024), and we will summarise it here.
- The resolution of the regional model is already at its finest limit due to computational limitations. One simulation with this resolution at this long climate time scale takes over 6 months on a High-Performance Computer at the German Climate Computing Centre. Also, convection is parametrised, as this version of the model at the 16 km resolution is not convection-permitting. We acknowledge that this is a limitation, but it is so far technically not feasible to increase the resolution and still conduct several century-scale simulations.
Comment 3
Please justify the choice of JRA-55 as the driving reanalysis for the CCLM hindcast, particularly given the widespread use and higher spatial resolution of ERA5, which is also used later in the manuscript for the KwaZulu-Natal event analysis.
The manuscript also states that the driving global model FOCI has a horizontal resolution of ~0.1° around Southern Africa (Line 137), which appears comparable to or potentially higher than the regional model CCLM (~16 km). If so, the added value of the regional downscaling should be clearly discussed and justified.- The hindcast simulation using JRA-55 as the driving reanalysis started running in April 2019. Unfortunately, ERA5 was not available at the time. Surely, it would have been a good choice. ERA-Interim covers a shorter time period and has a coarser resolution. Therefore, we chose JRA-55.
- As we mentioned in the previous comment, these simulations are costly and were conducted over a few years of real time.
- The quoted resolution of FOCI refers to the ocean resolution, not to the atmosphere resolution. This fine ocean resolution is needed to resolve the Agulhas Current System. An atmosphere resolution that fine is certainly impossible to achieve at multiyear timescales. Even Higher-resolution simulations with, let’s say, 0.1° as FOCI, aren’t feasible for such long simulation periods (1958-2019, 1951-2013, and 2014-2099) with daily outputs and for such a large domain. The added value of that extremely high atmospheric resolution is also not so clear.
Comment 4
The analysis of extremes during El Niño and La Niña years is interesting; however, the motivation for focusing specifically on El Niño–Southern Oscillation is not clearly introduced. The Introduction should provide background on the role of large-scale climate modes in regional rainfall variability and extremes. In particular, the potential influence of the Southern Annular Mode on southern African rainfall and storm tracks should at least be acknowledged and discussed.
- Thanks for the hint. The motivation is the well-known connection between seasonal-scale precipitation and ENSO in Southern Africa. As this manuscript is focused on extremes, it seems logical to extend the analysis of that seasonal link to extreme events. We would add a paragraph about the role of ENSO and other climate modes in the section ‘Region’ for precipitation and explain our motivation in this manuscript.
Comment 5
The reported daily precipitation maxima (e.g., 2225 mm on 24 February 2001; Line 242) appear exceptionally high and require further validation. It is unclear whether these values are physically realistic, particularly given the model resolution and likely use of parameterised convection. I strongly recommend evaluating these extremes against observations (e.g., CHIRPS or station data) and/or relevant literature to assess their plausibility and potential model biases. The future analysis also appears to use the full simulation period (2014–2099) rather than fixed climatological windows. Using consistent fixed-length periods (e.g., 30-year windows) would provide a more robust comparison between historical and future climates. Figure 6 would benefit from either showing climatological annual maxima or using a non-linear colour scale, and the statistical significance of trends in Figure 7 should be explicitly indicated.
- We would calculate the daily maxima of CHIRPS to discuss these simulated values.
- We will compare fixed-length periods for a more robust comparison between the historical and future climates in a revised verison.
- Showing the climatological annual maxima wouldn’t add value to the plot of the 99th percentile threshold. Whereas showing the absolute maximum underscores the extremeness of the simulated extremes.
- We would add the significance in Figure 7.
Comment 6
Figure 8: The historical and future simulations appear broadly similar, with only limited local differences. I strongly recommend showing difference fields (future minus historical) in addition to the absolute fields to more clearly highlight climate change signals. A similar approach should also be considered for other historical-future comparisons throughout the manuscript.
- We agree with the reviewer, and we will add a third panel for figures 8 and 9 to show the differences between the future and historical fields.
Comment 7
The colour scale in Figure 9 is dominated by a very small number of extreme grid points and does not adequately represent the broader spatial variability. A non-linear colour scale is strongly recommended. Also, the isolated daily precipitation maxima (>200 mm) appear inconsistent with the broader distribution, where the 99th percentile remains below ~40–45 mm. This again raises concerns regarding potential model artefacts or overestimation associated with the model resolution and parameterised convection. These extremes should be carefully verified and discussed.
- Yes, the colour scale emphasises that most daily maxima are below 200 mm and that these very high extremes are locally restricted to the east coast, where the more intense extreme events occur. Here, we do not agree with the reviewer’s comment and prefer it like that. Regarding the realism of the very high precipitation extremes, we would like to point out that extremes of up to 600 mm/day, even with a 99th percentile of 40 mm, don’t necessarily have to be a model artefact; they can occur in the model, as extreme events in the past have also shown. For instance, in the extreme event in April 2022, between 200 and 400 mm of rain fell within 24 hours(https://wmo.int/media/news/south-africa-declares-state-of-emergency-after-deadly-rains). A further increase in the future is possible due to rising temperatures in the air and sea.
Comment 8
Lines 293–299: The attribution of precipitation extremes to tropical cyclones and cut-off lows appears overly simplified and insufficiently justified. The classification relies solely on the location of minimum sea-level pressure, without adequate methodological detail or supporting references. This approach may misclassify dynamically distinct systems (e.g., tropical lows, hybrid systems, extratropical cyclones, cold fronts, or mesoscale convective systems), thereby limiting the physical interpretability of the results. In addition, the absence of supporting diagnostics or visualisation makes the validity of these attributions difficult to assess. A clearer and better-referenced methodology, including supporting diagnostics and discussion of limitations, is required.
- We agree with the reviewer. We would augment our definition of tropical cyclones and cut-off lows, including the temperature in 300- 500 hPa, for a better distinction between cold and warm core cyclones. We will include a figure showing two examples of a warm-core and a cold-core cyclone extracted from the model simulation.
Comment 9
Section 5.6: If I understood it correctly – this section mixes a single observed event (08–16th of April 2022), ERA5 climatological extremes, and model-simulated historical and future extremes in a way that is not fully physically or statistically comparable. As a result, several interpretations appear overstated and require stronger justification. In particular, the claimed “good agreement” (Line 344) between ERA5 and the simulations is not convincingly demonstrated, as it compares a single event with multi-decadal maxima rather than event-based simulation and further validation. Furthermore, the hindcast simulation does not include the April 2022 event, limiting direct model evaluation.
The discussion of future extremes (e.g., approximately doubling of maximum rainfall) also relies heavily on single extreme values rather than a formal extreme value framework and should therefore be interpreted with greater caution.
Lines 364–389: The circulation-based interpretations (e.g., cyclone influence – also not clearly mentioned tropical or extratropical cyclones, SLP intensification as key driver) are visually inferred but not supported by objective diagnostics or a cyclone identification methods. A clearer separation between event evaluation, climatological analysis, and future projections is strongly recommended.- We agree with the reviewer that a comparison of April 2022 would be very informative. However, we ran the simulation in 2019, and the analysis later evolved to the investigation of this extreme. Therefore, the comparison cannot be conducted on a one-event basis. Alternatively, we compared CLM not only to that specific event but also to the daily maximum of the period 1970-2020. We believe that the simulated precipitation of 383 mm/day is indeed in „good agreement“ with 335 mm/day from ERA5. We are comparing here the climatological orders of magnitude: the ERA5 event and the simulated event are not the same, but both are extremes in the respective data set.
- We would add a statement that the future doubling of rainfall is based on the absolute maximum here.
- Here, we compare the drivers of the most intense rainfall events in our simulations and ERA5. They are event evaluations in all cases.
Comment 10
It is strongly recommended to revise the figure titles and in-panel labels for clarity and consistency. In many cases, the titles require multiple readings to understand (e.g., Figure 10). Figures should clearly and explicitly describe what is being shown. Figures should also be cited more precisely within the text. For example, references such as “Fig. 10” do not clearly indicate which panel(s) the reader should examine. Several figures also could potentially be merged to improve conciseness (e.g., Figures 2 and 3; Figures 6 and 7).
- We would do so in a revised version.
Comment 11
The Discussion and Conclusion sections should be clearly separated. At present, a substantial amount of interpretation is embedded within the Results section, which weakens the manuscript structure. The Discussion should focus on interpreting the main findings (after considering all the referees’ comments) in the context of previous studies, supported by appropriate references, rather than introducing extensive discussion throughout the Results section.
- We thank the reviewer for this advice, which will help to strengthen the manuscript structure. We will change the Results and Discussion section accordingly, including references, after consideration of all referees' comments.
Minor comments
Comment 1
Lines 25–27: This sentence appears to summarise the work undertaken in the manuscript and would be more appropriate near the end of the Introduction, where the study objectives are typically presented. A similar point applies to Lines 62–64.
- We would move the sentences to the end of the introduction
Comment 2
Several statements and/or findings in the Introduction are not adequately supported by references (e.g., Lines 28–29, 36–37, 56–58, 106–108). It is also unclear whether some results originate from the present study (e.g., Lines 80–83). If so, they should be moved to the Results section; otherwise, appropriate references should be provided.
- We would add references where necessary.
Comment 3
The Introduction is generally well written; however, the overall flow could be improved. In particular, the progression from previous work to research gaps, motivation, and study objectives is not always clear, and some research gaps are introduced too late in the section (e.g., Lines 106–115).
- In lines 106-115, we provide an overview of recent publications in a Springer book and outline this paper's objectives. As the reviewer commented in Comment 1, study objectives are typically presented at the end of the introduction.
Comment 4
The use of the 99th percentile of daily precipitation is reasonable and widely used; however, sub-daily extremes are often more sensitive to climate change and may show stronger intensification signals. While I recognise that long-term hourly observations may be limited, the authors should briefly justify the use of daily-scale extremes and discuss how this choice may influence the interpretation of the results, particularly if mesoscale convective processes contribute to regional extremes.
- We would add a sentence about that. The choice of daily averages is motivated by both the availability of observational data and their impacts. An hourly-scale extreme is certainly interesting from the dynamical point of view, but the major impacts are given by more sustained extremes.
Comment 5
Lines 216–221: The use of 15-year samples (for CHIRPS) to estimate return levels potentially associated with longer return periods raises concerns regarding robustness, as this requires extrapolation beyond the available data range. The associated uncertainties should therefore be quantified and discussed more explicitly.
- The choice of this relatively short window to estimate return values is motivated by the objective of estimating trends in the return period during the observational period. Too wide a window leaves very few degrees of freedom to estimate changes in the return periods; too narrow a window is associated with larger uncertainties in the estimation of the return periods, as the reviewer indicates. The choice is thus a compromise between the two constraints.
- We will quantify and discuss the uncertainties derived from our choice.
Comment 6
Several sections, particularly within the validation and results, would benefit from substantial tightening. Many paragraphs contain excessive reporting of individual values, which obscures the main scientific message. For example, Section 5.2 reads more like a listing of values than a synthesis of key findings. In many cases, the historical and future values are also very similar, which further reduces clarity.
- We think that quoting the values of individual extremes can be more informative, but we acknowledge that this style requires an additional guiding line for the reader. We will complement this section with a tightened summary to increase readability.
Comment 7
Line 276: Please clarify what is meant by the “timing of extreme events”. It is currently unclear whether this refers to intraseasonal, seasonal, or interannual variability. If such analyses were performed, they deserve clearer explanation and interpretation.
- We refer here to the exact dates of extremes. We would clarify this in a revised version.
Comment 8
Figure 10: It is unclear how the number of extreme events has been calculated, particularly in Figures 10e–f. The reported values (~6000 events per year) appear unrealistically high and are difficult to interpret. Please clarify how extremes are defined and counted (e.g., grid-point exceedances versus spatially aggregated events) and verify whether these values are physically meaningful.
- We would do so in a revised version. The number of extremes is counted for each 10-year period and summed across all grid points. Thus, there are ~6000 extreme events over the entire coastline in 10 years.
Comment 9
Figure 11: The reported correlation values (approximately −0.2 to +0.2) appear weak, and it is unclear whether they are statistically significant. Given that the correlations are generally low, the robustness and interpretability of this analysis are questionable. If these relationships are weak or non-significant, this section could be substantially condensed.
- True. In lines 327-328, we explicitly state that the correlations are generally low. We acknowledge that the impact is weak and that other drivers may be more important.
Technical Corrections
Comment 1
Lines 6–7: Please ensure that all abbreviations are defined at first occurrence throughout the manuscript, including in the Abstract (e.g., CCLM, CORDEX). In addition, mathematical symbols should be used where appropriate (e.g., “×” instead of “x”).
- We would do so in a revised version.
Comment 2
Line 60: The sentence beginning with “Her dissertation…” would benefit from revision.Line 87: Please revise “In the previously cited study” to “In the previous study, …” and place the citation at the end of the sentence to improve flow.
- We would do so in a revised version.
Comment 3
Line 182: Please explicitly clarify that ERA5 is a reanalysis-based product and not a purely observational dataset.
- It’s an observational-based reanalysis dataset. We would clarify this in a revised version.
Comment 4
Line 298: Please verify whether this should be 40 °E rather than 40 °S.
- The reviewer is right. It should be 40°E. We would change that in a revised version.
Citation: https://doi.org/10.5194/egusphere-2026-806-AC2
-
AC2: 'Reply on RC2', Nele Tim, 15 Jun 2026
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 816 | 366 | 74 | 1,256 | 83 | 113 |
- HTML: 816
- PDF: 366
- XML: 74
- Total: 1,256
- BibTeX: 83
- EndNote: 113
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The work of Tim et al uses high resolution climate model simulations to investigate past and future change in extreme rainfall over southern Africa, while also exploring the influence of the Agulhas Current on the coastal regional precipitation. The authors find that the model replicate the observed extreme rainfall over the region and that extreme rainfall is projected to increase. The research presented in this paper aims to contribute to our understanding of extreme rainfall in the region and is of interest for the community. Overall, the paper is well written; however, some restructuring would improve the narrative flow. Additionally, the data used are not well suited for case study analysis, and the methodology requires further improvements and clarification. I also recommend strengthening the conclusions. Below are some suggestions and recommendations:
L37: I’m unclear on what the authors intend by the phrase 'models are burdened…”
L39: The intent of figure 1 is not clear in the introduction and how it fits within the general background.
L42-46: The sentences are not clear, and more context is needed, such as time periods, scenarios and seasons.
L83: Suggest replacing “negative trends’ by “downward trends’
L85: This manuscript (https://doi.org/10.1029/2020EA001466) could be interesting here.
L106-108: I recommend having a look at this paper https://doi.org/10.1002/wcc.70025
L146-149: I recommend adding a box of these regions with labels in one of the figures.
L150: Is this value representing the 99th percentile of wet days?
L51-153: Could you clarify or provide justification for why the order of operations varies depending on the size of the regions?
L163: For the GEV fit, is it assumed that the time series is stationary or a scale fit is included as covariate (e.g. section 4.3.2 https://ascmo.copernicus.org/articles/6/177/2020/)
L170: The authors highlight challenges related to observational data for model evaluation in lines 84-86. Are ERA5, TAMSAT, and CHIRPS the most suitable datasets for this purpose? Could alternatives such as ERA5-Land or Multi-Source Weather (MSWX) offer improvements? For reference, Tim et al. (2023; doi:10.5194/wcd-4-381-2023) utilized different datasets in their analysis.
L174: Would incorporating the seasonal rainfall cycle for the regions identified in Section 3.2 provide additional value here (and section 4.2)? Additionally, I would assume that section 4.1 would be already covered in Tim et al. 2023.
L196: “This study…” It is not clear to me which study is referenced here.
L205: Please clarify what you do with the resolution mismatch of the datasets. This should be clarified in the methods section.
L169: Should section 4 be under Results section?
L206: Figure4. Should the authors also include the regions identified in Section 3.2?
L209: “In our coastal domain…” Please provide boxes identifying the regions in the map.
L216-221: In here the authors should not compare return periods of two datasets with different lengths and periods! Furthermore, it is unclear whether this analysis and its findings contribute meaningful insights.
L222: Is this because of the short sample?
L229-230: I suggest adding this information in the methods section and the reasons. Which Niño index was used and from which dataset was it calculated, what are the regions with high correlation between El Nino and rainfall (see for e.g. 10.1088/1748-9326/ade60e)? As it stands, the analysis appears arbitrary and lacks context within the current narrative. Additionally, I suggest including a paragraph with a literature review on the relationship between El Niño and extreme precipitation to provide necessary context in the introduction.
L237: I recommend merging Sections 4 and 5 to create a more cohesive and accessible narrative. In its current form, the structure is somewhat disjointed and challenging for readers to follow. Additionally, I recommend including a detailed description of the analysis and methods in the methodology section to ensure the study is easily replicable.
L242-247: I am not sure how relevant it is to provide dates of extreme rainfall from climate models that have been run in a climatological setup as the model is not initialized to reproduce the real-world sequence of weather.
L247-248: “Thus, the heaviest rainfall occurs over the Mozambique Channel…”. Mozambique channel is over the ocean and the figure are not showing it!
L250: Would this information be better presented as timeseries for the regions of interest introduced in section 3.2? The current map is ‘too noisy’ and I’m not sure what is the main take away of this information.
L284: Why were this region chosen? It is unclear whether this region experiences the greatest impact from the Agulhas Current.
L293: I’m surprised by the fact that tropical cyclones contribute to extreme rainfall in this region! Are there any references supporting this?
L295: I suggest including these details in the methods. As mentioned before, the analysis appears arbitrary and lacks context within the current narrative.
L296-299: I suggest aligning the definition of tropical cyclones and cut-off-lows with the literature. Cut-off-low do also form north of 30°S as it is the case of Durban floods in April 2022 and others.
L330-389: I am not sure what is the validity of studying specific events with a climate model as they’re not designed to reproduce exact individual historical events exactly as they happened and they have internal variability.
L391-392: This is not the correct.
L392-397: Sentence is unclear.
L400-402: I’m not sure how is this information relevant here. Should this be in the introduction. Furthermore, “Tropical cyclones are projected to intensify” is not clear what aspect of tropical cyclones are projected to intensify and for which ocean basin. And the correct reference should be Seneviratne et al 2021 (https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_Chapter11.pdf)
L390: The conclusion needs to be revised considering the above recommendations and suggestions with clearer acknowledgment that the results are based on a single model and simulation, and with the associated uncertainties and limitations explicitly discussed and contextualized (see for e.g. L48-L51). Additionally, do these results improve or are they better/worse than the ones from CORDEX models? What can we learn from these higher resolution simulations for the region that we didn't know before?
And lastly, I suggest improving the colours in the maps. Here is a visual guide that may be helpful
https://www.ipcc.ch/site/assets/uploads/2022/09/IPCC_AR6_WGI_VisualStyleGuide_2022.pdf